Document nkY4pV5q9LEb1qQ2KkXR0wVG6

DRAFT - DO NOT CITE OR QUOTE EPA/600/R-10/038A www.epa.gov/iris EPA's Reanalysis of Key Issues Related to Dioxin Toxicity and Response to NAS Comments NOTICE THIS DOCUMENT IS AN EXTERNAL REVIEW DRAFT. It has not been formally released by the U.S. Environmental Protection Agency and should not at this stage be construed to represent Agency policy. It is being circulated for comment on its technical accuracy and policy implications. National Center for Environmental Assessment Office of Research and Development U.S. Environmental Protection Agency Cincinnati, OH DISCLAIMER This document is distributed solely for the purpose of pre-dissemination peer review under applicable information quality guidelines. It has not been formally disseminated by EPA. It does not represent and should not be construed to represent Agency determination or policy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. ABSTRACT This draft report details EPA's technical response to the key comments and recommendations included in the 2006 NAS report, "Health Risks from Dioxin and Related Compounds: Evaluation of the EPA Reassessment," focusing on the NAS comments regarding TCDD doseresponse assessment. After systematically evaluating the epidemiologic studies and rodent bioassays on TCDD, this draft report utilized a TCDD physiologically-based pharmacokinetic model to simulate TCDD blood concentrations, the dose metric used in the dose-response analyses. The draft report develops an oral reference dose (RfD) of 7 x 10-10 mg/kg-day based on two epidemiologic studies that associated TCDD exposures with decreased sperm concentration and sperm motility in men who were exposed during childhood (Mocarelli et al., 2008, 199595) and increased thyroid-stimulating hormone levels in newborn infants (Baccarelli et al., 2008, 197059). EPA also classifies TCDD as carcinogenic to humans, based on numerous lines of evidence, including primarily: multiple occupationally- and accidentally-exposed epidemiologic cohorts showing an association between TCDD exposure and certain cancers or increased mortality from all cancers and extensive evidence of carcinogenicity at multiple tumor sites in both sexes of multiple species of experimental animals. Based on a cancer mortality analysis of an occupational cohort (Cheng et al., 2006, 523122), EPA also develops an oral cancer slope factor of 1 x 106per (mg/kg-day) when the target risk range is 10-5 to 10-7. While this draft report provides limited sensitivity analyses of several steps in the cancer and noncancer dose-response assessment, it concludes that a comprehensive uncertainty analysis is infeasible at this time. Preferred Citation: U.S. Environmental Protection Agency (U.S. EPA). (2010) EPA's Reanalysis of Key Issues Related to Dioxin Toxicity and Response to NAS Comments. EPA/600/R-10/038A. NAS comments are published by the National Research Council of the National Academies and available from the National Technical Information Service, Springfield, VA, and online at http://www.epa.gov/ncea. This document is a draftfor review purposes only and does not constitute Agency policy. ii DRAFT--DO NOT CITE OR QUOTE CONTENTS LIST OF TABLES......................................................................................................................... ix LIST OF FIGURES.....................................................................................................................xiii LIST OF ABBREVIATIONS AND ACRONYMS................................................................... xvii PREFACE.................................................................................................................................... xxi AUTHORS, CONTRIBUTORS, AND REVIEWERS.............................................................. xxii EXECUTIVE SUMMARY....................................................................................................... xxvi 1. INTRODUCTION................................................................................................................. 1-1 1.1. SUMMARY OF KEY NAS (2006, 198441) COMMENTS ON DOSERESPONSE MODELING IN THE 2003 REASSESSMENT...................................... 1-2 1.2. EPA'S SCIENCE PLAN.............................................................................................. 1-4 1.3. OVERVIEW OF EPA'S RESPONSE TO NAS (2006, 198441) "HEALTH RISKS FROM DIOXIN AND RELATED COMPOUNDS: EVALUATION OF EPA's 2003 REASSESSMENT".................................................................................. 1-5 1.3.1. TCDD Literature Update.................................................................................. 1-6 1.3.2. EPA's 2009 Workshop on TCDD Dose Response........................................... 1-7 1.3.3. Overall Organization of EPA's Response to NAS Recommendations............. 1-9 2. TRANSPARENCY AND CLARITY IN THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS.................................................................................. 2-1 2.1. SUMMARY OF NAS COMMENTS ON TRANSPARENCY AND CLARITY IN THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS................................................................................................................... 2-1 2.2. EPA's RESPONSE TO NAS COMMENTS ON TRANSPARENCY AND CLARITY IN THE SELECTION OF KEY DATA SETS FOR DOSEREsPONSE ANALYSIS..............................................................................................2-2 2.3. STUDY INCLUSION CRITERIA FOR TCDD DOSE-RESPONSE ANALYSIS................................................................................................................... 2-4 2.3.1. Study Inclusion Criteria for TCDD Epidemiologic Studies............................. 2-6 2.3.2. Study Inclusion Criteria for TCDD In Vivo Mammalian Bioassays................2-8 2.4. EVALUATION OF KEY STUDIES FOR TCDD DOSE RESPONSE....................2-10 2.4.1. Evaluation of Epidemiological Cohorts for Dose-Response Assessment 2-10 2.4.1.1. Cancer..............................................................................................2-11 2.4.1.2. Noncancer....................................................................................... 2-87 2.4.2. Summary of Animal Bioassay Studies Included for TCDD DoseResponse Modeling....................................................................................... 2-134 2.4.2.1. Reproductive Studies.....................................................................2-135 2.4.2.2. Developmental Studies..................................................................2-149 2.4.2.3. Acute Studies.................................................................................2-168 2.4.2.4. Subchronic Studies........................................................................2-176 2.4.2.5. Chronic Studies (Noncancer Endpoints)......................................2-191 2.4.2.6. Chronic Studies (Cancer Endpoints)............................................ 2-204 This document is a draftfor review purposes only and does not constitute Agency policy. iii DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 2.4.3. Summary of Key Data Set Selection for TCDD Dose-Response Modeling....................................................................................................... 2-211 3. THE USE OF TOXICOKINETICS IN THE DOSE-RESPONSE MODELING FOR CANCER AND NONCANCER ENDPOINTS..................................................................... 3-1 3.1. SUMMARY OF NAS COMMENTS ON THE USE OF TOXICOKINETICS IN DOSE-RESPONSE MODELING APPROACHES FOR TCDD............................ 3-1 3.2. OVERVIEW OF EPA'S RESPONSE TO THE NAS COMMENTS ON THE USE OF TOXICOKINETICS IN DOSE-RESPONSE MODELING APPROACHES FOR TCDD........................................................................................ 3-3 3.3. PHARMACOKINETICS (PK) AND PK MODELING............................................... 3-4 3.3.1. PK Data and Models in TCDD Dose-Response Modeling: Overview and Scope................................................................................................................. 3-4 3.3.2. PK of TCDD in Animals and Humans............................................................. 3-6 3.3.2.1. Absorption and Bioavailability....................................................... 3-6 3.3.2.2. Distribution...................................................................................... 3-6 3.3.2.3. Metabolism and Protein Binding...................................................... 3-9 3.3.2.4. Elimination.................................................................................... 3-11 3.3.2.5. Interspecies Differences and Similarities....................................... 3-11 3.3.3. PK of TCDD in Humans: Interindividual Variability..................................... 3-12 3.3.3.1. Life Stage and Gender.................................................................... 3-13 3.3.3.2. Physiological States: Pregnancy and Lactation.............................. 3-16 3.3.3.3. Lifestyle and Habits........................................................................ 3-17 3.3.3.4. Genetic Traits and Polymorphism.................................................. 3-18 3.3.4. Dose Metrics and Pharmacokinetic Models for TCDD.................................. 3-18 3.3.4.1. Dose Metrics for Dose-Response Modeling.................................. 3-18 3.3.4.2. First-Order Kinetic Modeling......................................................... 3-22 3.3.4.3. Biologically-Based Kinetic Models............................................... 3-26 3.3.4.4. Applicability of PK Models to Derive Dose Metrics for DoseResponse Modeling of TCDD: Confidence and Limitations........3-42 3.3.4.5. Recommended Dose Metrics for Key Studies............................... 3-45 3.3.5. Uncertainty in Dose Estimates........................................................................ 3-47 3.3.5.1. Sources of Uncertainty in Dose Metric Predictions....................... 3-47 3.3.5.2. Qualitative Discussion of Uncertainty in Dose Metrics................. 3-49 3.3.6. Use of the Emond PBPK Models for Dose Extrapolation from Rodents to Humans....................................................................................................... 3-51 4. CHRONIC ORAL REFERENCE DOSE..............................................................................4-1 4.1. NAS COMMENTS AND EPA'S RESPONSE ON IDENTIFYING NONCANCER EFFECTS OBSERVED AT LOWEST DOSES................................. 4-1 4.2. NONCANCER DOSE-RESPONSE ASSESSMENT OF TCDD................................ 4-6 4.2.1. Determination of Toxicologically Relevant Endpoints................................... 4-6 4.2.2. Use of Toxicokinetic Modeling for TCDD Dose-Response Assessment........4-7 This document is a draftfor review purposes only and does not constitute Agency policy. iv DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 4.2.3. Noncancer Dose-Response Assessment of Epidemiological Data..................4-9 4.2.3.1. Baccarelli et al. (2008, 197059)....................................................... 4-9 4.2.3.2. Mocarelli et al. (2008, 199595)...................................................... 4-10 4.2.3.3. Alaluusua et al. (2004, 197142).....................................................4-11 4.2.3.4. Eskenazi et al. (2002, 197168)....................................................... 4-12 4.2.4. Noncancer Dose-Response Assessment of Animal Bioassay Data...............4-13 4.2.4.1. Use of Kinetic Modeling for Animal Bioassay D ata....................4-13 4.2.4.2. Benchmark Dose Modeling of the Animal Bioassay Data............4-14 4.2.4.3. POD Candidates from Animal Bioassays Based on HED and BMD Modeling Results.................................................................. 4-16 4.3. RfD DERIVATION.................................................................................................... 4-18 4.3.1. Toxicological Endpoints.................................................................................4-19 4.3.2. Exposure Protocols of Candidate PODs.........................................................4-20 4.3.3. Uncertainty Factors (UFs)............................................................................... 4-21 4.3.4. Choice of Human Studies for RfD Derivation...............................................4-22 4.3.4.1. Identification of POD from Baccarelli et al. (2008, 197059).........4-24 4.3.4.2. Identification of POD from Mocarelli et al. (2008, 199595).........4-25 4.3.4.3. Identification of POD from Alaluusua et al. (2004, 197142).........4-27 4.3.5. Derivation of the R fD .....................................................................................4-27 4.4. UNCERTAINTY IN THE RfD.................................................................................. 4-28 5. CANCER ASSESSMENT..................................................................................................... 5-1 5.1. QUALITATIVE WEIGHT-OF-EVIDENCE CARCINOGEN CLASSIFICATION FOR 2,3,7,8-TETRACHLORODIBENZO-p-DIOXIN (TCDD)........................................................................................................................ 5-1 5.1.1. Summary of National Academy of Sciences (NAS) Comments on the Qualitative Weight-of-Evidence Carcinogen Classification for 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD)................................................. 5-1 5.1.2. EPA's Response to the NAS Comments on the Qualitative Weight-ofEvidence Carcinogen Classification for TCDD................................................ 5-2 5.1.2.1. Summary Evaluation of Epidemiologic Evidence of TCDD and Cancer....................................................................................... 5-3 5.1.2.2. Summary of Evidence for TCDD Carcinogenicity in Experimental Animals.................................................................... 5-10 5.1.23. TCDD Mode of Action................................................................... 5-10 5.1.3. Summary of the Qualitative Weight of Evidence Classification for TCDD............................................................................................................. 5-20 5.2. QUANTITATIVE CANCER ASSESSMENT........................................................... 5-21 5.2.1. Summary of NAS Comments on Cancer Dose-Response Modeling.............. 5-21 5.2.1.1. Choice of Response Level and Characterization of the Statistical Confidence Around Low Dose Model Predictions.......5-21 5.2.1.2. Model Forms for Predicting Cancer Risks Below the Point of Departure (POD)............................................................................ 5-22 This document is a draftfor review purposes only and does not constitute Agency policy. v DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 5.2.2. Overview of EPA Response to NAS Comments on Cancer DoseResponse Modeling......................................................................................... 5-23 5.2.3. Updated Cancer Dose-Response Modeling for Derivation of Oral Slope Factor.............................................................................................................. 5-24 5.2.3.1. Dose-Response Modeling Based on Epidemiologic Cohort D ata................................................................................................ 5-24 5.2.3.2. Dose-Response Modeling Based on Animal Bioassay Data.......... 5-35 5.2.3.3. EPA's Response to the NAS Comments on Choice of Response Level and Characterization of the Statistical Confidence Around Low Dose Model Predictions........................ 5-50 5.2.3.4. EPA's Response to the NAS Comments on Model Forms for Predicting Cancer Risks Below the POD....................................... 5-51 5.3. DERIVATION OF THE TCDD ORAL SLOPE FACTOR AND CANCER RISK ESTIMATES.................................................................................................... 5-75 5.3.1. Uncertainty in Estimation of Oral Slope Factors from Human Studies.........5-77 5.3.1.1. Uncertainty in Exposure Estimation............................................... 5-78 5.3.1.2. Uncertainty in Shape of the Dose-Response Curve....................... 5-82 5.3.1.3. Uncertainty in Extrapolating Risks below Reference Population Exposure Levels.......................................................... 5-83 5.3.1.4. Uncertainty in Cancer Risk Estimates Arising from Background DLC Exposure........................................................... 5-84 5.3.1.5. Uncertainty in Cancer Risk Estimates Arising from Occupational DLC Coexposures.................................................... 5-85 5.3.2. Other Sources of Uncertainty in Risk Estimates from the Epidemiological Studies................................................................................. 5-86 5.3.2.1. Effect of Added Background TEQ on TCDD Dose-Response......5-88 5.3.3. Approaches to Combining Estimates from Different Epidemiologic Studies............................................................................................................. 5-90 5.3.3.1. The Crump et al. (2003, 197384) Meta-analysis............................ 5-90 5.3.3.2. EPA's Decision Not to Conduct a Meta-analysis.......................... 5-92 6. FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS FROM NAS EVALUATION OF THE 2003 REASSESSMENT.............................................................. 6-1 6.1. INTRODUCTION........................................................................................................ 6-1 6.1.1. Historical Context for Quantitative Uncertainty Analysis................................ 6-1 6.1.2. Definition of Terms........................................................................................... 6-3 6.1.3. Key Elements of a Quantitative Uncertainty Analysis..................................... 6-6 6.1.3.1. Quantitative Model............................................................................ 6-6 6.1.3.2. Marginal Distributions over Model Parameter................................. 6-6 6.1.3.3. Dependence between Parameter Uncertainties: Aleatoric and Epistemic (Uncertainty and Variability).......................................... 6-7 6.1.3.4. Model Uncertainty.............................................................................6-8 6.1.3.5. Sampling Method.............................................................................. 6-9 This document is a draftfor review purposes only and does not constitute Agency policy. vi DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 6.1.3.6. Method for Extracting andCommunicating Results....................... 6-9 6.2. EPA APPROACHES FOR ORAL CANCER AND NONCANCER ASSESSMENT........................................................................................................... 6-10 6.3. HIGHLIGHTS OF NAS REVIEW COMMENTS ON UNCERTAINTY QUANTIFICATION FOR THE 2003 REASSESSMENT........................................ 6-12 6.4. FEASIBILITY OF CONDUCTING A QUANTITATIVE UNCERTAINTY ANALYSIS FOR TCDD............................................................................................ 6-15 6.4.1. Feasibility of Conducting a Quantitative Uncertainty Analysis under the RfD Methodology........................................................................................... 6-15 6.4.1.1. Feasibility of Conducting a Quantitative Uncertainty Analysis for the Point of Departure............................................................... 6-16 6.4.1.2. Feasibility of Conducting a Quantitative Uncertainty Analysis with Uncertainty Factors................................................................ 6-19 6.4.1.3. Uncertainty Reduction Using Quantitative Data for Species Extrapolation.................................................................................. 6-21 6.4.1.4. Conclusion on Feasibility of Quantitative Uncertainty Analysis with the RfD Approach................................................... 6-22 6.4.2. Feasibility of Conducting a Quantitative Uncertainty Analysis for TCDD under the Dose-Response Methodology......................................................... 6-23 6.4.2.1. Feasibility of Quantitatively Characterizing the Uncertainties Encountered when Determining Appropriate Types of Studies (Epidemiological, Animal, Both, and Other)................................. 6-24 6.4.2.2. Uncertainty in TCDD Exposure/Dose in Epidemiological Studies............................................................................................ 6-25 6.4.2.3. Uncertainty in Toxicity Equivalence (TEQ) Exposures in Epidemiological Studies................................................................. 6-28 6.4.2.4. Uncertainty in Background Feed Exposures in Bioassays............. 6-29 6.4.2.5. Feasibility of Quantifying the Uncertainties Encountered When Choosing Specific Studies and Subsets of Data (e.g., Species and Gender)....................................................................... 6-31 6.4.2.6. Feasibility of Quantifying the Uncertainties Encountered when Choosing Specific Endpoints for Dose-Response Modeling........................................................................................ 6-31 6.4.2.7. Feasibility of Quantifying the Uncertainties Encountered when Choosing a Specific Dose Metric (Trade-Off between Confidence in Estimated Dose and Relevance of MOA)............... 6-32 6.4.2.8. Feasibility of Quantifying the Uncertainties Encountered When Choosing Model Type and Form......................................... 6-34 6.4.2.9. Threshold MOA for Cancer........................................................... 6-36 6.4.2.10. Feasibility of Quantifying the Uncertainties Encountered when Selecting the BMR................................................................ 6-37 6.5. CONCLUSIONS REGARDING THE FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS................................................................................... 6-38 This document is a draftfor review purposes only and does not constitute Agency policy. vii DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 6.5.1. Summary of NAS Suggestions and Responses............................................... 6-38 6.5.2. How Forward? Beyond RfDs and Cancer Slope Factors to Development of Predictive Human Dose-Response Functions............................................. 6-41 REFERENCES........................................................................................................................... R-1 APPENDIX A: DIOXIN WORKSHOP..................................................................................... A-1 APPENDIX B: EVALUATION OF CANCER AND NONCANCER EPIDEMIOLOGICAL STUDIES FOR INCLUSION IN TCDD DOSE-RESPONSE ASSESSMENT..................................................... B-1 APPENDIX C: KINETIC MODELING..................................................................................... C-1 APPENDIX D: EPIDEMIOLOGICAL KINETIC MODELING............................................... D-1 APPENDIX E: NONCANCER BENCHMARK DOSE MODELING...................................... E-1 APPENDIX F: CANCER BENCHMARK DOSE MODELING................................................F-1 APPENDIX G: ENDPOINTS EXCLUDED FROM REFERENCE DOSE DERIVATION BASED ON TOXICOLOGICAL RELEVANCE................... G-1 APPENDIX H: CANCER PRECURSOR BENCHMARK DOSE MODELING...................... H-1 APPENDIX I: EFFECT OF BACKGROUND EXPOSURE ON BENCHMARK-DOSE MODELING........................................................................................................I-1 This document is a draftfor review purposes only and does not constitute Agency policy. viii DRAFT--DO NOT CITE OR QUOTE LIST OF TABLES 2-1. Summary of epidemiological cancer studies (key characteristics).............................2-212 2-2. Epidemiological cancer study selection considerations and criteria............................2-215 2-3. Epidemiological noncancer study selection considerations and criteria......................2-219 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling.............2-223 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling......2-228 2-6. Animal bioassays selected for cancer dose-response modeling................................... 2-232 2- 7. Animal bioassay studies selected for noncancer dose-response modeling................... 2-234 3-1. Partition coefficients, tissue volumes, and volume of distribution for TCDD in humans........................................................................................................................... 3-56 3-2. Blood flows, permeability factors and resulting half lives ( t/) for perfusion losses for humans as represented by the TCDD PBPK model of Emond et al. (2005, 197317; 2006, 197316).................................................................................................. 3-56 3-3. Toxicokinetic conversion factors for calculating human equivalent doses from rodent bioassays............................................................................................................. 3-57 3-4. Equations used in the concentration and age-dependent model (CADM; Aylward et al., 2005, 197014)...................................................................................................... 3-58 3-5. Parameters of the Concentration and age-dependent model (CADM; Aylward et al., 2005, 197014).......................................................................................................... 3-59 3-6. Confidence in the CADM model simulations of TCDD dose metrics.......................... 3-60 3-7. Equations used in the TCDD PBPK model of Emond et al. (2006, 197316)................. 3-61 3-8. Parameters of the PBPK model for TCDD..................................................................... 3-63 3-9. Regression analysis results for the relationship between log10 serum TCDD at the midpoint of observations and the log10of the rate constant for decline of TCDD levels using Ranch Hand data........................................................................................ 3-66 3-10. Confidence in the PBPK model simulations of TCDD dose metrics............................. 3-66 3-11. Overall confidence associated with alternative dose metrics for cancer and noncancer dose-response modeling for TCDD using rat PBPK model......................... 3-67 This document is a draftfor review purposes only and does not constitute Agency policy. ix DRAFT--DO NOT CITE OR QUOTE LIST OF TABLES (continued) 3-12. Overall confidence associated with alternative dose metrics for cancer and noncancer dose-response modeling for TCDD using mouse PBPK model..................3-67 3-13. Contributors to the overall confidence in the selection and use of dose metrics in the dose-response modeling of TCDD based on rat and human PBPK models...........3-67 3-14. Contributors to the overall uncertainty in the selection and use of dose metrics in the dose-response modeling of TCDD based on mouse and human PBPK models.....3-68 3-15. Comparison of human equivalent doses from the Emond human PBPK model for the 45-year-old and 25-year-old gestational exposure scenarios..................................3-68 3- 16. Impact of toxicokinetic modeling on the extrapolation of administered dose to HED, comparing the Emond PBPK and first-order body burden models....................3-69 4- 1. POD candidates for epidemiologic studies of TCDD.....................................................4-33 4-2. Models run for each study/endpoint combination in the animal bioassay benchmark dose modeling............................................................................................. 4-33 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered dose, first-order body burden HED, and blood concentration.................................................................................................................. 4-34 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kg)............................................................... 4-38 4-5. Candidate points of departure for the TCDD RfD using blood-concentrationbased human equivalent doses....................................................................................... 4-49 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays possessing candidate points-of-departure for the TCDD RfD...........4-53 4- 7. Basis and derivation of the TCDD reference dose..........................................................4-57 5- 1. Cancer slope factors calculated from Becher et al.(1998, 197173), Steenland et al. (2001, 197433) and Ott and Zober (1996, 198408) from 2003 Reassessment Table 5-4.................................................................................................................................. 5-94 5-2. Cox regression coefficients and incremental cancer-mortality risk for NIOSH cohort data...................................................................................................................... 5-95 This document is a draftfor review purposes only and does not constitute Agency policy. x DRAFT--DO NOT CITE OR QUOTE LIST OF TABLES (continued) 5-3. Comparison of fat concentrations, risk specific dose estimates and associated oral slope factors based on upper 95th percentile estimate of regression coefficient of all fatal cancers reported by Cheng et al. (2006, 523122) for selected risk levels.......5-96 5-4. Comparison of fat concentrations, risk specific dose estimates and associated central tendency slope estimates based on best estimate of regression coefficient of all fatal cancers reported by Cheng et al. (2006, 523122) for selected risk levels.... 5-97 5-5. Kociba et al. (1978, 001818) male rat tumor incidence data and blood concentrations for dose-response modeling.................................................................. 5-97 5-6. Kociba et al. (1978, 001818) female rat tumor incidence data and blood concentrations for dose-response modeling................................................................... 5-98 5-7. NTP (1982, 594255) female rat tumor incidence data and blood concentrations for dose-response modeling........................................................................................... 5-98 5-8. NTP (1982, 594255) male rat tumor incidence data and blood concentrations for dose-response modeling................................................................................................ 5-99 5-9. NTP (1982, 594255) female mouse tumor incidence data and blood concentrations for dose-response modeling................................................................... 5-99 5-10. NTP (1982, 594255) male mouse tumor incidence data and blood concentrations for dose-response modeling......................................................................................... 5-100 5-11. NTP (2006, 197605) female rat tumor incidence data and blood concentrations for dose-response modeling........................................................................................ 5-100 5-12. Toth et al. (1979, 197109) male mouse tumor incidence data and blood concentrations for dose-response modeling................................................................. 5-101 5-13. Della Porta et al. (1987, 197405) male mouse tumor incidence data and blood concentrations for dose-response modeling................................................................. 5-101 5-14. Della Porta et al. (1987, 197405) female mouse tumor incidence data and blood concentrations for dose-response modeling................................................................. 5-101 5-15. Comparison of multi-stage modeling results across cancer bioassays using blood concentrations.............................................................................................................. 5-102 5-16. Individual tumor points of departure and slope factors using blood concentrations ... 5-104 5-17. Multiple tumor points of departure and slope factors using blood concentrations.....5-105 This document is a draftfor review purposes only and does not constitute Agency policy. xi DRAFT--DO NOT CITE OR QUOTE LIST OF TABLES (continued) 5-18. Comparison of cancer BMDs, BMDLs, and slope factors for combined or selected individual tumors for 1, 5, and 10% extra risk........................................................... 5-106 5-19. TCDD human-equivalent dose (HED) BMDs, BMDLs, and oral slope factors (OSF) for 1, 5, and 10% extra risk..............................................................................5-107 5-20. Illustrative RfDs based on tumorigenesis in experimental animals............................. 5-108 5-21. Illustrative RfDs based on hypothesized key events in TCDD's MOAs for liver and lung tumors............................................................................................................ 5-109 5-22. Comparison of prinicipal epidemiological studies....................................................... 5-110 5-23. Added background TEQ exposures to blood TCDD/TEQconcentrations in rats........5-112 5-24. Effect of added background TEQ exposure on BMDL01 for cholangiocarcinomas in rats...........................................................................................................................5-113 5- 25.NIOSH cohort septile data with added TEQ background.............................................5-113 6- 1. Key sources of uncertainty..............................................................................................6-44 6-2. PODs and amenability for uncertainty quantification....................................................6-45 This document is a draftfor review purposes only and does not constitute Agency policy. xii DRAFT--DO NOT CITE OR QUOTE LIST OF FIGURES 2-1. EPA's process to select and identify in vivo mammalian and epidemiologic studies for use in the dose-response analysis of TCDD..............................................2-248 2-2. EPA's process to evaluate available epidemiologic studies using study inclusion criteria for use in the dose-response analysis of TCDD.............................................2-249 2-3. EPA's process to evaluate available animal bioassay studies using study inclusion criteria for use in the dose-response analysis of TCDD.............................................2-250 3-1. Liver/fat concentration ratios in relation to TCDD dose at various times after oral administration of TCDD to mice...................................................................................3-70 3-2. First-order elimination rate fits to 36 sets of serial TCDD sampling data from Seveso patients as function of initial serum lipid TCDD..............................................3-71 3-3. Observed relationship of fecal 2,3,7,8-TCDD clearance and estimated percent body fat..........................................................................................................................3-72 3-4. Unweighted empirical relationship between percent body fat estimated from body mass index and TCDD elimination half-life--combined Ranch Hand and Seveso observation..................................................................................................................... 3-73 3-5. Relevance of candidate dose metrics for dose-response modeling, based on mode of action and target organ toxicity of TCDD................................................................ 3-74 3-6. Process of estimating a human-equivalent TCDD lifetime average daily oral exposure (dH) from an experimental animal average daily oral exposure (dA) based on the body-burden dose metric..................................................................................... 3-75 3-7. Human body burden time profiles for achieving a target body burden for different exposure duration scenarios........................................................................................... 3-76 3-8. Schematic of the CADM structure.................................................................................3-77 3-9. Comparison of observed and simulated fractions of the body burden contained in the liver and adipose tissues in rats................................................................................ 3-78 3-10. Conceptual representation of PBPK model for rat exposed to TCDD.......................... 3-79 3-11. Conceptual representation of PBPK model for rat developmental exposure to TCDD............................................................................................................................. 3-80 3-12. TCDD distribution in the liver tissue.............................................................................3-81 This document is a draftfor review purposes only and does not constitute Agency policy. xiii DRAFT--DO NOT CITE OR QUOTE LIST OF FIGURES (continued) 3-13. Growth rates for physiological changes occurring during gestation............................3-82 3-14. Comparisons of model predictions to experimental data using a fixed elimination rate model with hepatic sequestration (A) and an inducible elimination rate model with (B) and without (C) hepatic sequestration............................................................3-83 3-15. PBPK model simulation of hepatic TCDD concentration (ppb) during chronic exposure to TCDD at 50, 150, 500, 1,750 ng TCDD/BW using the inducible elimination rate model compared with the experimental data measured at the end of exposure..................................................................................................................... 3-84 3-16. Model predictions of TCDD blood concentration in 10 veterans (A-J) from Ranch Hand Cohort................................................................................................................... 3-85 3-17. Time course of TCDD in blood (pg/g lipid adjusted) for two highly exposed Austrian women (patients 1 and 2 )...............................................................................3-86 3-18. Observed vs. Emond et al. (2005, 197317) model simulated serum TCDD concentrations (pg/g lipid) over time (ln = natural log) in two Austrian women.........3-87 3-19. Comparison of the dose dependency of TCDD elimination in the Emond model vs. observations of nine Ranch Hand veterans and two highly exposed Austrian patients........................................................................................................................... 3-88 3-20. Sensitivity analysis was performed on the inducible elimination rate..........................3-89 3-21. Experimental data (symbols) and model simulations (solid lines) of (A) blood, (B) liver and (C) adipose tissue concentrations of TCDD after oral exposure to 150 ng/kg-day, 5 days/week for 17 weeks in mice..............................................................3-90 3-22 Comparison of PBPK model simulations with experimental data on liver concentrations in mice administered a single oral dose of 0.001-300 pg TCDD/kg .... 3-91 3-23. Comparison of model simulations (solid lines) with experimental data (symbols) on the effect of dose on blood (cb), liver (cli) and fat (cf) concentrations following repetitive exposure to 0.1-450 ng TCDD/kg, 5 days/week for 13 weeks in mice........3-92 3-24. Comparison of experimental data (symbols) and model predictions (solid lines) of (A) blood, (B) liver and (C) adipose tissue concentrations of TCDD after oral exposure to 1.5 ng/kg-day, 5 days/week for 17 weeks in m ice....................................3-93 This document is a draftfor review purposes only and does not constitute Agency policy. xiv DRAFT--DO NOT CITE OR QUOTE LIST OF FIGURES (continued) 3-25. Comparison of experimental data (symbols) and model predictions (solid lines) of (A) blood concentration, (B) liver concentration, (C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary elimination (% dose) of TCDD after oral exposure to 1.5 ng/kg-day, 5 days/week for 13 weeks in mice.............................3-94 3-26. Comparison of experimental data (symbols) and model predictions (solid lines) of (A) blood concentration, (B) liver concentration, (C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary elimination (% dose) of TCDD after oral exposure to 150 ng/kg-day, 5 days/week for 13 weeks in mice............................3-95 3-27. PBPK model simulations (solid lines) vs. experimental data (symbols) on the distribution of TCDD after a single acute oral exposure to A-B) 0.1, C-D) 1.0 and E-F) 10 pg of TCDD/kg of body weight in mice..................................................3-96 3-28. PBPK model simulation (solid lines) vs. experimental data (symbols) on the distribution of TCDD after a single dose of 24 pg/kgBW on GD 12 in mice..............3-97 3-29. Comparison of the near-steady-state body burden simulated with CADM and Emond models for a daily dose ranging from 1 to 10,000 ng/kg-day in rats and hum ans.......................................................................................................................... 3-98 3-30. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure scenarios, with target concentrations shown for each; profiles generated with Emond human PBPK model.................................................................3-99 3- 31. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure scenarios, showing continuous intake levels to fixed target concentration; profiles generated with Emond human PBPK model..........................3-100 4-1. EPA's process to select and identify candidate PODs from key epidemiologic studies for use in the noncancer risk assessment of TCDD..........................................4-58 4-2. EPA's process to select and identify candidate PODs from key animal bioassays for use in noncancer dose-response analysis of TCDD................................................ 4-59 4-3. Exposure-response array for ingestion exposures to TCDD.......................................... 4-60 4- 4. Candidate RfD array........................................................................................................4-61 5- 1. Mechanism of altered gene expression by A hR ...........................................................5-114 5-2. TCDD's hypothesized modes of action in site-specific carcinogenesis.......................5-115 This document is a draftfor review purposes only and does not constitute Agency policy. xv DRAFT--DO NOT CITE OR QUOTE LIST OF FIGURES (continued) 5-3. EPA's process to select and identify candidate OSFs from key animal bioassays for use in the cancer risk assessment of TCDD............................................................5-116 5-4. Dose-response model shape.......................................................................................... 5-117 5-5. Comparison of individual and population dose-response curves; a simple illustration.....................................................................................................................5-118 5-6. Multistage benchmark dose modeling of NTP (2006, 197605) cholangiosarcoma data............................................................................................................................... 5-119 5-7. Multistage benchmark dose modeling of NTP (2006, 197605) combined tumor data............................................................................................................................... 5-120 5-8. Estrogen receptor-mediated response-modeling plot from Kohn and Melnick (2002, 199104): low-dose region shown...................................................................... 5-121 5-9. Representative endpoints for each of the hypothesized key events following AhR activation for TCDD-induced liver tumors.................................................................. 5-122 5-10. Representative endpoints for two hypothesized key events following AhR activation for TCDD-induced lung tumors...................................................................5-123 5- 11.Candidate oral slope factor array...................................................................................5-124 6- 1. Back-casted vs. predicted TCDD serum levels for a worker subset............................... 6-46 6-2. Distribution of in vivo unweighted REP values in the 2004 database............................ 6-47 This document is a draftfor review purposes only and does not constitute Agency policy. xvi DRAFT--DO NOT CITE OR QUOTE LIST OF ABBREVIATIONS AND ACRONYMS 2,4,5-T 2,4-D AA ACOH AHH AhR AhR-/AIC ANL ANOVA APE ASAT AUC bHLH-PAS Bmax BMD BMDL BMDS BMI BMR BPS BROD b-TSH BW C CADM Cc CI CSAF CSLC Cx CYP Da:HED DEN df DLC DRE/XRE DRL DSA E2 EDx EGFR 2,4,5-trichlorophenoxyacetic acid 2,4-dichlorophenoxyacetic acid ascorbic acid acetanilide-4-hydroxylase aryl hydrocarbon hydroxylase aryl hydrocarbon receptor AhR-deficient Akaike Information Criterion Argonne National Laboratory analysis of variance airborne particulate extract aspartate aminotransferase area under the curve basic helix-loop-helix, Per-Arnt-Sim equilibrium maximum binding capacity benchmark dose benchmark dose lower confidence bound Benchmark dose software body mass index benchmark response balanopreputial separation benzyloxy resoufin-O-deethylase blood thyroid-stimulating hormone body weight cerebellum concentration- and age-dependent elimination model cerebral cortex confidence interval chemical-specific adjustment factor cumulative serum lipid concentration connexin cytochrome P450 ratio of administered dose to HED diethylnitrosamine degrees of freedom dioxin-like compound dioxin/xenobiotic response elements differential reinforcement of low rate delayed spatial alteration 17P-estradiol effective dose eliciting x percent response epidermal growth factor receptor This document is a draftfor review purposes only and does not constitute Agency policy. xvii DRAFT--DO NOT CITE OR QUOTE LIST OF ABBREVIATIONS AND ACRONYMS (continued) EPA ER EROD ERa EU FFA FR FSH FT4 GD GSH GSH-Px GST H HCH HED HQ HR Hsp90 IARC IGF IL ILSI i.p. IRIS KABS LASC LD50 LED LEDx LH LOAEL LOAELhed LOEL MCH MCMC MCV MOA MOE MROD MTD NAS NIOSH Environmental Protection Agency estrogen receptor 7-ethoxyresorufin-O-deethylase estrogen receptor alpha European Union free fatty acid fixed-ratio follicle stimulating hormone free thyroxine gestation day glutathione stimulating hormone glutathione stimulating hormone peroxidase glutathione-S-transferase hippocampus hexachlorocyclohexane human equivalent dose hazard quotient hazard ratio heat shock protein 90 International Agency for Research on Cancer insulin-like growth factor interleukin International Life Sciences Institute intraperitoneal Integrated Risk Information System oral absorption parameters lipid-adjusted serum concentration lethal dose eliciting x percent response lower confidence effective dose lower bound of the 95% confidence interval on the dose that yields an x% effect luteinizing hormone lowest-observed-adverse-effect level HED estimate based on LOAELs lowest-observed-adverse level mean corpuscular hemoglobin Markov Chain Monte Carlo mean corpuscular volume mode of action margin of exposure 7-methoxyresorufin-O-deethylase maximum tolerated dose National Academy of Sciences National Institute for Occupational Safety and Health This document is a draftfor review purposes only and does not constitute Agency policy. xviii DRAFT--DO NOT CITE OR QUOTE LIST OF ABBREVIATIONS AND ACRONYMS (continued) NOAEL NOEL NRC NTP OR OSF PA PAI2 PBMC PBPK PCB PCDD PCDF PEPCK PF PHAH PK PND POD pp PRA PRE PROD RAR REP RfC RfD RL RL RR RR RT-PCR RXR S SA SAhRM S-D SD SIR SMR SOD SRBC SSB no-observed-adverse-effect level no-observed-effect level National Research Council National Toxicology Program odds ratio oral slope factor permeability x area plasminogen activator inhibitor 2 peripheral blood mononuclear cells physiologically based pharmacokinetic polychlorinated biphenyl polychlorinated dibenzo-p-dioxin polychlorinated dibenzofuran phosphoenolpyruvate carboxykinase adipose tissue:blood partition coefficient polyhalogenated aromatic hydrocarbons pharmacokinetic postnatal day point of departure phosphotyrosyl protein probabilistic risk assessment body:blood partition coefficient 7-pentoxyresorufin-O-deethylase retinoic acid receptor relative potency reference concentration reference dose reversal learning risk level rate ratios relative risk reverse transcription polymerase chain reaction retinoid X receptor saline superoxide anion SRM for AhRs Sprague-Dawley standard deviation standardized incidence ratio standardized mortality ratio superoxide dismutase sheep red blood cell single-strand break This document is a draftfor review purposes only and does not constitute Agency policy. xix DRAFT--DO NOT CITE OR QUOTE LIST OF ABBREVIATIONS AND ACRONYMS (continued) SWHS T4 TBARS TCB TCDD TCP TEF TEQ TGFa TK TNF-a TOTTEQ TSH TT4 TWA US. NRC UDP UDPGT UED UF UFa UFd UFh UFl UFs UGT UGT1 Vd WHO ZS@Z Seveso Women's Health Study thyroxine thiobarbituric acid-reactive substances 3,3',4,4'-tetrachlorobiphenyl 2,3,7,8 -Tetrachlorodibenzo-p-dioxin 2,4,5-trichlorophenol toxicity equivalence factor toxicity equivalence transforming growth factor a toxicokinetic tumor necrosis factor alpha total toxicity equivalence thyroid stimulating hormone total thyroxine time-weighted average U.S. Nuclear Regulatory Commission uridine diphosphate UDP-glucoronosyl transferase upper confidence bound for the effective dose uncertainty factor interspecies extrapolation factor database factor human interindividual variability LOAEL-to-NOAEL UF subchronic-to-chronic UF UDP-glucuronosyltransferase uridine diphosphate glucuronosyltransferase I volume of distribution World Health Organization zero slope at zero dose This document is a draftfor review purposes only and does not constitute Agency policy. xx DRAFT--DO NOT CITE OR QUOTE PREFACE This report was developed by the U.S. Environmental Protection Agency's (EPA) Office of Research and Development (ORD), National Center for Environmental Assessment (NCEA). Sections of the report, including Section 6 and the updated literature search, were developed through a collaborative effort between NCEA and the Department of Energy's Argonne National Laboratory (ANL). In 2003, EPA, along with other federal agencies, asked the National Academy of Sciences (NAS) to review aspects of the science in EPA's draft dioxin reassessment entitled, "Exposure and Human Health Reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds," and, in 2004, EPA sent the 2003 draft dioxin reassessment to the NAS for their review. In 2006, the NAS released the report of their review entitled, "Health Risks from Dioxin and Related Compounds: Evaluation of the EPA Reassessment." The NAS identified three areas in EPA's 2003 draft reassessment that required substantial improvement to support a more scientifically robust risk characterization. These three areas were: (1) justification of approaches to dose-response modeling for cancer and noncancer endpoints; (2) transparency and clarity in selection of key data sets for analysis; and (3) transparency, thoroughness, and clarity in quantitative uncertainty analysis. The NAS provided EPA with recommendations to address their key concerns. This draft report details EPA's response to the key comments and recommendations included in the 2006 NAS report. In 2008, prior to developing this draft report, EPA, in collaboration with ANL, developed and published a literature database of peer-reviewed studies on TCDD toxicity, including in vivo mammalian dose-response studies and epidemiologic studies. EPA subsequently requested public comment on this database. EPA and ANL then convened a scientific workshop in 2009. The Workshop goals were to identify and address issues related to the dose-response assessment of TCDD and to ensure that EPA's response to the NAS focused on the key issues and reflected the most meaningful science. This draft report provides a technical response to the 2006 NAS report. It utilizes a TCDD physiologically-based pharmacokinetic model in its development of dose-response analyses of TCDD toxicological and epidemiological literature. This draft report presents new analyses of both the potential cancer and noncancer human health effects that may result from exposures to TCDD. The draft report develops an oral reference dose (RfD) for TCDD. It also presents a new cancer oral slope factor. Federal agencies and White House offices have been provided an opportunity for review and comment on this draft report prior to its public release. This draft dioxin report is being released for public comment and will also be provided to EPA's Science Advisory Board (SAB) for independent external peer review. The SAB will convene an expert panel composed of scientists knowledgeable about technical issues related to dioxins and risk assessment. The SAB is expected to hold their first public meeting on July 13-15, 2010. This document is a draftfor review purposes only and does not constitute Agency policy. xxi DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS PRIMARY AUTHORS National Center for Environmental Assessment, U.S. Environmental Protection Agency, Cincinnati, OH Belinda Hawkins Janet Hess-Wilson Glenn Rice (Project Co-Lead) Jeff Swartout (Project Co-Lead) Linda K. Teuschler CONTRIBUTING AUTHORS National Center for Environmental Assessment, U.S. Environmental Protection Agency, Cincinnati, OH Scott Wesselkamper Michael Wright Bette Zwayer National Health and Environmental Effects Research Laboratory, U.S. Environmental Protection Agency, Research Triangle Park, NC Hisham El-Masri Argonne National Laboratory, Argonne, IL Margaret MacDonell Emory University, Atlanta, GA Kyle Steenland Resources for the Future, Washington, DC Roger M. Cooke University of Montreal; BioSimulation Consulting, Newark, DE Claude Emond University of Montreal, Montreal, Canada Kannan Krishnan This document is a draftfor review purposes only and does not constitute Agency policy. xxii DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued) CONTRIBUTORS National Center for Environmental Assessment, U.S. Environmental Protection Agency, Washington, DC Karen Hogan Ravi Subramaniam Leonid Kopylev Paul White Argonne National Laboratory, Argonne, IL Maryka H. Bhattacharyya Andrew Davidson Mary E. Finster David P. Peterson Clark University, Worcester, MA Dale Hattis Colorado State University, Fort Collins, CO Raymond Yang Bruce Allen Consulting, Chapel Hill, NC Bruce C. Allen ICF International, Durham, NC Robyn Blain Rebecca Boyles Patty Chuang Cara Henning Baxter Jones Penelope Kellar Mark Lee Nikki Maples-Reynolds Amalia Marenberg Garrett Martin Margaret McVey Chandrika Moudgal Bill Mendez Ami Parekh Andrew Shapiro Audrey Turley National Toxicology Program, Research Triangle Park, NC Michael Devito Penn State University, University Park, PA Jack P. Vanden Heuvel Risk Sciences International, Ottawa, Ontario Jessica Dennis Dan Krewski Greg Paoli Salomon Sand Natalia Shilnikova Paul Villenueve University of California-Berkeley, Berkeley, CA Brenda Eskenazi This document is a draftfor review purposes only and does not constitute Agency policy. xxiii DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued) CONTRIBUTORS (continued) University of California-Irvine, Irvine, CA Scott Bartell REVIEWERS This document has been provided for review to EPA scientists and interagency reviewers from other federal agencies and White House offices. INTERNAL REVIEWERS National Center for Environmental Assessment, U. S. Environmental Protection Agency Glinda Cooper, Washington, DC Eva McLanahan, Research Triangle Park, NC Ila Cote, Research Triangle Park, NC Susan Rieth, Washington, DC Lynn Flowers, Washington, DC Reeder Sams, Research Triangle Park, NC Martin Gehlhaus, Washington, DC Paul Schlosser, Research Triangle Park, NC Kate Guyton, Washington, DC Jamie Strong, Washington, DC Samantha Jones, Washington, DC John Vandenberg, Research Triangle Park, NC Matthew Lorber, Washington, DC ACKNOWLEDGMENTS National Center for Environmental Assessment, U.S. Environmental Protection Agency Jeff Frithsen, Washington, DC Maureen Johnson, Washington, DC Annette Gatchett, Cincinnati, OH Peter Preuss, Washington, DC Andrew Gillespie, Cincinnati, OH Linda Tuxen, Washington, DC Marie Nichols-Johnson, Cincinnati, OH Colorado State University, Fort Collins, CO William H. Farland ECFlex, Inc., Fairborn, OH Dan Heing Heidi Glick Amy Prues Lana Wood This document is a draftfor review purposes only and does not constitute Agency policy. xxiv DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued) ACKNOWLEDGMENTS (continued) IntelliTech Systems, Inc., Fairborn, OH Cris Broyles Luella Kessler Debbie Kleiser Stacey Lewis Linda Tackett National Institute of Environmental Health Sciences, Research Triangle Park, NC Linda S. Birnbaum Christopher J. Portier National Toxicology Program, Research Triangle Park, NC Nigel Walker 2009 Dioxin Workshop Participants This document is a draftfor review purposes only and does not constitute Agency policy. xxv DRAFT--DO NOT CITE OR QUOTE 1 EXECUTIVE SUMMARY 2 3 4 OVERVIEW 5 The U.S. Environmental Protection Agency (EPA) is committed to the development of 6 risk assessment information of the highest scientific integrity for use in protecting human health 7 and the environment. Scientific peer review is an integral component of the process EPA uses to 8 generate high quality toxicity and exposure assessments of environmental contaminants. To this 9 end, EPA asked the National Academy of Sciences (NAS) to review its comprehensive human 10 health risk assessment external review draft entitled, Exposure and Human Health Reassessment 11 o f 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and Related Compounds (U.S. EPA, 2003, 12 537122; "2003 Reassessment"). This current document, EPA's Reanalysis o fKey Issues Related 13 to Dioxin Toxicity and Response to NAS Comments, directly and technically responds to key 14 comments and recommendations pertaining to TCDD dose-response assessment published by the 15 NAS in their review (NAS, 2006, 198441). This document only addresses issues pertaining to 16 TCDD dose-response assessment. 17 In May 2009, EPA Administrator Lisa P. Jackson announced the "Science Planfor 18 Activities Related to Dioxins in the Environment"("Science Plan") that addressed the need to 19 finish EPA's dioxin reassessment and provide a completed health assessment on this high profile 20 chemical to the American public as quickly as possible.1 The Science Plan states that EPA will 21 release a draft report that responds to the recommendations and comments included in the NAS 22 review of EPA's 2003 Reassessment, and that, in this draft report, EPA's National Center for 23 Environmental Assessment, Office of Research and Development, will provide a limited 24 response to key comments and recommendations in the NAS report (draft response). This draft 25 response is to focus on dose-response issues raised by the NAS and include analyses of relevant 26 new key studies. The draft response is to be provided for public review and comment and for 27 independent external peer review by EPA's Science Advisory Board. Following completion of 28 this report, EPA is to review the impacts of the response to comments report on its 2003 29 Reassessment. 1Available at http://www.epa.gov/dioxin/scienceplan. This document is a draftfor review purposes only and does not constitute Agency policy. xxvi 1 This draft document comprises EPA's report that responds both directly and technically 2 to the recommendations and comments on TCDD dose-response assessment included in the NAS 3 review of EPA's 2003 Reassessment. Because new data are analyzed in this report and toxicity 4 values are derived, this document will follow the IRIS process for review, clearance and 5 completion; however, it is not a traditional IRIS document. Information developed in this 6 document is intended to not only respond to the NAS review, but also to expand EPA's 7 knowledge of TCDD cancer and noncancer dose-response based on the most current literature, 8 existing methods, and adherence to EPA risk assessment guidance documents. 9 In addition to this document, three separate EPA activities address additional NAS 10 comments pertaining to toxicity equivalence factors (TEFs) and background exposure levels. 11 Information on the application of the dioxin TEFs is published elsewhere by EPA for both 12 ecological (U.S. EPA, 2008, 543774) and human health (U.S. EPA, 2009, 192196) risk 13 assessment. EPA does not directly address TEFs herein, but makes use of the concept of toxicity 14 equivalence (TEQ)2as applicable to the analysis of exposure dose in epidemiologic studies and 15 to discussions on the effect of background TEQ on TCDD dose response. Furthermore, 16 information on updated background levels of dioxin in the U.S. population has been recently 17 reported by EPA (Lorber et al., 2009, 543766), addressing the NAS recommendations pertaining 18 to the assessment of human exposures to TCDD and other dioxins. 19 The NAS identified three key recommendations requiring substantial improvement to 20 support a scientifically robust characterization of human responses to exposures to TCDD. 21 These three key areas are (1) improved transparency and clarity in the selection of key data sets 22 for dose-response analysis, (2) further justification of approaches to dose-response modeling for 23 cancer and noncancer endpoints, and (3) improved transparency, thoroughness, and clarity in 24 quantitative uncertainty analysis. The NAS also encouraged EPA to calculate a Reference Dose 25 (RfD), and provided numerous specific comments on various aspects of EPA's 2003 26 Reassessment. The three key recommendations specifically pertain to dose-response assessment 27 and uncertainty analysis. Therefore, EPA's response to the NAS in this document is focused on 2Toxicity equivalence (TEQ) is the product of the concentration of an individual dioxin like compound in an environmental mixture and the corresponding TCDD TEF for that compound. These products are summed to yield the TEQ of the mixture. This document is a draftfor review purposes only and does not constitute Agency policy. xxvii 1 these issues. EPA thoroughly considered the recommendations of the NAS and responds with 2 scientific and technical evaluation of TCDD dose-response data via: 3 4 an updated literature search that identified new TCDD dose-response studies (see 5 Section 2); 6 a kickoff workshop that included the participation of external experts in TCDD health 7 effects, toxicokinetics, dose-response assessment and quantitative uncertainty analysis; 8 these experts discussed potential approaches to TCDD dose-response assessment and 9 considerations for EPA's response to NAS; a Workshop Report was developed 10 (U.S. EPA, 2009, 543757, see Appendix A); 11 detailed TCDD-specific study inclusion criteria and processes for the selection of key 12 studies (see Section 2.3) and epidemiologic and animal bioassay data for TCDD 13 dose-response assessment (see Section 2.4.1, Appendix B, and Section 2.4.2, 14 respectively); 15 kinetic modeling to quantify appropriate dose metrics for use in TCDD dose-response 16 assessment (see Section 3 and Appendices C and D); 17 dose-response modeling for all appropriate noncancer and cancer data sets (see 18 Section 4.2/Appendix E and Section 5.2.3/Appendix F, respectively); 19 thorough and transparent evaluation of the selected TCDD data for use in the derivation 20 of an RfD and an oral slope factor (OSF) (see Sections 4.2 and 5.2.3, respectively); 21 the development of an RfD (see Section 4.3); 22 the development of a revised OSF (see Section 5.3) with an updated cancer weight of 23 evidence determination for TCDD based on EPA's 2005 Cancer Guidelines (U.S. EPA, 24 2005, 086237) (see Section 5.1.2); 25 consideration of nonlinear dose-response approaches for cancer, including illustrative 26 RfDs for cancer precursor events and tumors (see Section 5.2.3.4) ; and 27 discussion of the feasibility and utility of quantitative uncertainty analysis for TCDD 28 dose-response assessment (see Section 6). 29 30 Each of the activities listed above is briefly described in this Executive Summary, and is 31 described in detail in the related sections of this document. 32 33 PRELIMINARY ACTIVITIES UNDERTAKEN BY EPA TO ENSURE THAT THIS 34 TECHNICAL RESPONSE REFLECTS THE CURRENT STATE-OF-THE-SCIENCE 35 As part of the development of this document, EPA undertook two activities that included 36 public involvement: an updated literature search and a scientific expert workshop. The adverse This document is a draftfor review purposes only and does not constitute Agency policy. xxviii 1 health effects associated with TCDD exposures are documented extensively in epidemiologic 2 and toxicologic studies. As such, the database of relevant information pertaining to the 3 dose-response assessment of TCDD is vast and constantly expanding. Responding directly to the 4 NAS recommendation to use the most current and up-to-date scientific information related to 5 TCDD, EPA, in collaboration with Argonne National Laboratory (ANL), developed an updated 6 literature database of peer-reviewed studies on TCDD toxicity, including in vivo mammalian 7 dose-response studies and epidemiologic studies. An initial literature search for studies 8 published since the 2003 Reassessment was conducted to identify studies published between the 9 year 2000 and October 31, 2008. EPA published the initial literature search results in the Federal 10 Register in November 2008 and invited the public to review the list and submit additional 11 peer-reviewed relevant studies. Additional studies identified by the public and through 12 continued work on this response have been incorporated into the final set of studies for TCDD 13 dose-response assessment (updated through October 2009). EPA believes that the 14 implementation of this rigorous search strategy ensures that the most current and relevant studies 15 were considered for the technical response to NAS and TCDD dose-response assessment 16 included herein. 17 To assist in responding to the NAS, EPA, in collaboration with ANL, convened a 18 scientific expert workshop ("Dioxin Workshop") in February 2009 that was open to the public. 19 The primary goals of the Dioxin Workshop were to identify and address issues related to the 20 dose-response assessment of TCDD and to ensure that EPA's response to the NAS focused on 21 the key issues, while reflecting the most meaningful science. EPA and ANL assembled expert 22 scientists and asked them to identify and discuss the technical challenges involved in addressing 23 the NAS comments, discuss approaches for addressing these key recommendations, and to assist 24 in the identification of important published and peer-reviewed literature on TCDD. The 25 workshop was structured into seven scientific topic sessions as follows: (1) quantitative 26 dose-response modeling issues, (2) immunotoxicity, (3) neurotoxicity and nonreproductive 27 endocrine effects, (4) cardiovascular toxicity and hepatotoxicity, (5) cancer, (6) reproductive and 28 developmental toxicity, and (7) quantitative uncertainty analysis of dose-response. External 29 co-chairs (i.e., scientists who were not members of EPA or ANL) were asked to facilitate the 30 sessions and then prepare summaries of discussions occurring in each session. The session 31 summaries formed the basis of a final workshop report (U.S. EPA, 2009, 543757, Appendix A of This document is a draftfor review purposes only and does not constitute Agency policy. xxix 1 this document). Some of the key outcomes from the workshop include the following 2 recommendations: 3 4 to further develop study selection criteria for evaluating the suitability of developing 5 dose-response models based on animal bioassays and human epidemiologic studies; 6 to use kinetic modeling to identify relevant dose metrics and dose conversions between 7 test animal species and humans, and between human internal dose measures and human 8 intakes; 9 to consider newer human or animal (e.g., NTP, 2006, 197605) publications when 10 evaluating quantitative dose-response models for cancer; 11 to consider both linear and nonlinear modeling in the cancer dose-response analysis. 12 13 The discussions held during the Dioxin Workshop helped inform, guide, and focus EPA's 14 response to NAS. 15 16 EPA'S APPROACH TO CONSIDERING TRANSPARENCY AND CLARITY IN THE 17 SELECTION OF KEY STUDIES AND DATA SETS FOR DOSE-RESPONSE 18 MODELING 19 One of the key NAS recommendations to EPA was to utilize a clear and transparent 20 process for the selection of key studies and data sets for dose-response assessment. EPA agrees 21 with the NAS and believes that clear delineation of the study selection process and decisions 22 regarding key studies and data sets will facilitate communication of critical decisions made in the 23 TCDD dose-response assessment. EPA developed detailed processes and TCDD-specific 24 criteria for the selection of key dose-response studies. These criteria are based on common 25 practices and current guidance for point of departure (POD) identification and RfD and OSF 26 derivation and also consider issues specifically related to TCDD. Following the selection of key 27 studies, EPA employed additional processes to further select and identify cancer and noncancer 28 datasets from these key studies for use in dose-response analysis of TCDD. 29 For the study evaluation and key data set selection, EPA has undertaken different 30 approaches for the epidemiologic and in vivo animal bioassay studies. The significant 31 differences between animal and human health effects data and their use in EPA risk assessment 32 support development of separate criteria for study inclusion and different approaches to study 33 evaluation. For the vast majority of compounds on EPA's Integrated Risk Information System This document is a draftfor review purposes only and does not constitute Agency policy. xxx 1 (IRIS, U.S. EPA, 2009, 192196), cancer and noncancer toxicity values have been derived using 2 animal bioassay data; thus, some of the TCDD-specific study inclusion criteria for animal 3 bioassay data are based on EPA's common practices and guidance for POD selection and RfD 4 and OSF derivation. Far fewer IRIS toxicity values have been derived from human data, 5 although some examples do exist.3 The modeling and interpretation of such human data have 6 been conducted on a case-by-case basis because each cohort is uniquely defined and has its own 7 set of exposure conditions, significant confounders, and biases that may need to be considered in 8 dose-response modeling. 9 Figure ES-1 presents EPA's study evaluation process for the epidemiologic studies 10 considered for this TCDD dose-response assessment, including specific study inclusion criteria 11 (see Section 2.3.1). EPA applied TCDD-specific epidemiologic study inclusion criteria to all 12 epidemiologic studies published on TCDD and dioxin-like compounds (DLCs) that had been 13 identified in the TCDD literature database (see Section 2.4.1, Appendix B). The studies were 14 initially evaluated using five considerations (see Figure ES-1) that provide the most relevant 15 kinds of information needed to consider the feasibility of quantitative human health risk 16 analyses. Then EPA required that the studies meet three study inclusion criteria: 1) the study is 17 published in the peer-reviewed scientific literature and includes an appropriate discussion of 18 strengths and limitations; 2) the exposure is primarily to TCDD, rather than dioxin-like 19 compounds (DLCs), and is properly quantified so that dose-response relationships can be 20 assessed; and 3) the effective dose and oral exposure must be reasonably estimable. To meet the 21 third criterion, information is required on long-term exposures for cancer, and, for noncancer, 22 information is required regarding the appropriate time window of exposure that is relevant for a 23 specific, nonfatal health endpoint. Therefore, the study should include an appropriate latency 24 period between TCDD exposure and the onset of the effect. Only studies meeting these 25 three criteria were included in EPA's TCDD dose-response analyses (see Section 2.4.3). 26 Figure ES-2 presents EPA's study evaluation process for mammalian bioassays 27 considered for TCDD dose-response assessment, including the specific study inclusion criteria 28 (see Section 2.3.2). EPA applied TCDD-specific in vivo mammalian bioassay study inclusion 3Examples of toxicity values on IRIS from human data include benzene, beryllium and compounds, chromium IV, and 1,3-butadiene that have RfDs, Reference Concentrations, Inhalation Unit Risks and/or OSFs all based on occupational cohort data and the methyl mercury RfD that is based on high fish consuming cohorts (U.S. EPA, 2009, 192196). This document is a draftfor review purposes only and does not constitute Agency policy. xxxi 1 criteria to all of the bioassay studies of TCDD that had been identified in the TCDD literature 2 database (see Section 2.4.2). After ascertaining that a study had been published in the 3 peer-reviewed literature, EPA applied dose requirements to the lowest tested average daily doses 4 in each study, with specific requirements for cancer (<1 pg/kg-day) and noncancer 5 (<30 ng/kg-day) studies to ensure that only low-dose TCDD bioassays would be considered for 6 quantitative assessment. These dose requirements were used to eliminate those studies that 7 would not be selected for development of an RfD or an OSF because the lowest doses tested 8 were too high relative to other TCDD bioassays. EPA also required that the bioassays exposed 9 animals via the oral route to TCDD only and that the purity of TCDD was specified. Finally, the 10 studies were evaluated using four considerations (see Figure ES-2) regarded as providing the 11 most relevant information for development of quantitative human health risk analyses from 12 animal bioassay data. Only the bioassay studies meeting these criteria and considerations were 13 included in EPA's TCDD dose-response analyses (see Section 2.4.3). 14 Applying the study inclusion criteria for both epidemiologic and mammalian bioassay 15 datasets resulted in a list of key noncancer and cancer studies that were considered for 16 quantitative dose-response analyses of TCDD. Endpoints from these studies that were not 17 considered to be toxicologically relevant were eliminated from consideration (see Section 4.2.1, 18 Appendix G). The study/endpoint dataset combinations from the remaining studies were then 19 subjected to dose-response assessment, and PODs for use in developing RfDs or OSFs were 20 identified. PODs included no-observed-adverse-effect levels (NOAELs), lowest-observed21 adverse-effect levels (LOAELs) or lower bound benchmark dose levels (BMDLs). The most 22 sensitive PODs were selected as candidates for derivation of the RfD and OSF. 23 24 USE OF KINETIC MODELING TO ESTIMATE TCDD DOSES 25 NAS recommended that EPA utilize state-of-the-science approaches to finalize the 26 2003 Reassessment. Although NAS concurred with EPA's use of first-order body burden 27 models in the 2003 Reassessment, analyses of recent TCDD literature and comments by experts 28 at the Dioxin Workshop suggested that the understanding of TCDD kinetics had increased 29 significantly since the release of EPA's 2003 Reassessment. These advances led to the 30 development of several pharmacokinetic models for TCDD (Aylward et al., 2005, 197114; e.g., This document is a draftfor review purposes only and does not constitute Agency policy. xxxii 1 Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316) and 2 resulted in EPA's incorporation of TCDD kinetics in the dose-response assessment of TCDD. 3 The evaluation of internal dose in exposed humans and other species is facilitated by an 4 understanding of pharmacokinetics (i.e., absorption, distribution, metabolism, and excretion). 5 TCDD pharmacokinetics are influenced by three distinctive features: (1) TCDD is highly 6 lipophilic, (2) TCDD is slowly metabolized, and (3) TCDD induces binding proteins in the liver. 7 The overall impact of these factors results in preferential storage of TCDD in adipose tissue, a 8 long half-life of TCDD in blood due to slow metabolism, and sequestration in liver tissue when 9 binding induction becomes significant. As these kinetic features control target tissue levels of 10 dioxin, they become important in relating toxicity in animals to possible effects in humans. 11 Consideration of pharmacokinetic mechanisms is critical to the selection of the dose 12 metrics of relevance to dose-response modeling of TCDD. Earlier assessments for TCDD, 13 including the 2003 Reassessment, used estimates of body burden as the dose metric for 14 extrapolation between animals and humans. These body burden calculations used a simple 15 one-compartment kinetic model based on the assumption of a first-order decrease in the levels of 16 administered dose as a function of time. However, the assumption of a constant half-life value 17 for the clearance of TCDD from long-term or chronic exposure is not well-supported 18 biologically given the dose-dependant elimination observed in rodents and humans. The 19 dynamic disposition and redistribution of TCDD between blood, fat, and liver as a function of 20 time and dose is better described using biologically-based models. Additionally, these models 21 provide estimates for other dose metrics (e.g., serum, whole blood, or tissue levels) that are more 22 biologically relevant to response than body burden estimated based on an assumption of 23 first-order elimination over time. 24 EPA considered the following possible dose metrics for TCDD: administered dose, 25 first-order body burden, lipid-adjusted serum concentration (LASC), whole blood concentration, 26 tissue concentration, and functional-related metrics of relevance to the mode of action (MOA) 27 (e.g., receptor occupancy) (see Section 3.3.4.1). After careful evaluation of these dose metrics, 28 EPA chose to use TCDD concentration in whole blood as the dose metric for assessing TCDD 29 dose response in this document. Although LASC is generally considered to be the most relevant 30 metric, whole blood concentration was chosen because of the structure of the PBPK models, in 31 which the target tissue compartments are connected to the whole blood compartment rather than This document is a draftfor review purposes only and does not constitute Agency policy. xxxiii 1 to the serum compartment; LASC is related to whole blood by a scalar, so use of either is 2 equivalent in the model. Whole blood concentrations also reflect TCDD dose to target tissues 3 and, are biologically-relevant measures of internal dose. EPA used the time-weighted average 4 whole-blood concentration over the relevant exposure periods for all continuous dosing 5 protocols, dividing the area under the time-course concentration curve (AUC) by the exposure 6 duration.4 7 Several biologically-based kinetic models for TCDD exist in the literature. The more 8 recent pharmacokinetic models explicitly characterize the concentration-dependent elimination 9 of TCDD (Carrier et al., 1995, 197618; Carrier et al., 1995, 543780; Emond et al., 2004, 197315; 10 Emond et al., 2005, 197317; Emond et al., 2006, 197316; Aylward et al., 2005, 197114). The 11 biologically-based pharmacokinetic models describing the concentration-dependent elimination 12 (i.e., the pharmacokinetic models of Aylward et al. (2005, 197114) and Emond et al. (2005, 13 197317; 2006, 197316) are relevant for application to simulate the TCDD dose metrics in 14 humans and animals exposed via the oral route. The rationale for considering the application of 15 the Aylward et al. (2005, 197114) and Emond et al. (2004, 197315; 2005, 197317; 2006, 16 197316) models was largely based on the fact that both models reflect research results from 17 recent peer-reviewed publications, and both models are formulated with dose-dependent hepatic 18 elimination consistent with the physiological understanding of TCDD kinetics. Dose-response 19 modeling based on body burden of TCDD in adult animals and humans can be conducted with 20 either of the models, provided the duration of the experiment is at least one month, due to 21 limitations in the Aylward et al. (2005, 197114) model. The predicted slope and body burden 22 over a large dose range are quite comparable between the two models (generally within a factor 23 of two). 24 Results of simulations of serum lipid concentrations or liver concentrations vary for the 25 two models to a larger extent (up to a factor of 7), particularly for simulations of short duration. 26 These differences reflect two characteristics of the Emond et al. (2006, 197316) model: first, 27 quasi-steady-state is not assumed in the Emond et al. (2006, 197316) model; second, the serum 28 lipid composition used in the model is not the same as the adipose tissue lipids. The Aylward 4For the Seveso cohort, which had a high single exposure followed by low-level background exposures leading to a gradual decline in the internal TCDD concentrations, EPA estimated dose as the mean of the peak exposure and the average exposure over a defined critical exposure window (see Section 4.2.2). This document is a draftfor review purposes only and does not constitute Agency policy. xxxiv 1 et al. (2005, 197114) model does not account for differential solubility of TCDD in serum lipids 2 and adipose tissue lipids, nor does it account for the diffusion-limited uptake by adipose tissue. 3 Based on this evaluation, EPA determined that the Emond et al. (2006, 197316) model 4 performed better than the Aylward et al. (2005, 197114) model with respect to the ability to 5 simulate serum lipid and tissue concentrations during exposures that do not lead to the onset of 6 steady-state condition in the exposed organism. Additionally, of the two selected models, the 7 pharmacokinetic model developed by Emond et al. (2006, 197316) is more 8 physiologically-based, as compared to the Aylward et al. (2005, 197114) model, and models the 9 blood compartment directly in the rat, mouse, and human; there are also gestational and life-time 10 nongestational forms of the Emond et al. (2006, 197316) model. In this document, EPA chose 11 the Emond rodent physiologically-based pharmacokinetic (PBPK) model to estimate blood 12 TCDD concentrations based on administered doses (see Section 3.3.4, Appendix C). 13 To enhance the biological basis of the PBPK model of Emond et al. (2006, 197316), 14 three minor modifications,were made before its use in the computation of dose metrics for 15 TCDD: 1) recalculation of the volume of the "rest of the body compartment" after accounting for 16 volume of the liver and fat compartments; 2) calculation of the rate of TCDD excreted via urine 17 by multiplying the urinary clearance parameter by blood concentration in the equation instead of 18 by the concentration in the rest of the body compartment; and 3) recalibration for the human 19 gastric nonabsorption constant to yield observed oral bioavailability of TCDD (Poiger and 20 Schlatter, 1986) (see Section 3.3.4.4 for details). The modified PBPK model was evaluated 21 against all published data used in the original model. EPA assumed that the same blood TCDD 22 levels that led to effects in animals would also lead to effects in humans; therefore, the Emond 23 human PBPK model was used to estimate the lifetime average daily oral doses (consistent with 24 the chronic RfD and OSF) that would correspond to the blood TCDD concentrations estimated to 25 have occurred during the animal bioassays. EPA used the same Emond human PBPK model to 26 estimate the lifetime average daily doses that would correspond to the TCDD blood or tissue 27 concentrations reported in the epidemiological studies (Appendix D). These estimates are the 28 Human Equivalent Doses (HEDs) that are used to develop candidate RfDs and OSFs for TCDD. 29 Because TCDD elimination is inducible in the Emond model, ratios of daily averaged 30 intake to long-term blood concentrations are not linear. Because of the nonlinearity of blood 31 concentration and ingested dose in the Emond Human PBPK model, the cancer risk is only This document is a draftfor review purposes only and does not constitute Agency policy. xxxv 1 approximately linear with the TCDD blood concentration and low TCDD oral ingestion doses, 2 but is not linear with ingested TCDD at higher doses. Thus, to use these estimates in human 3 health risk assessment, risk-specific TCDD oral intake levels corresponding to the target risk 4 levels should be calculated (see Section 5.2.3.1.2.1). 5 6 DERIVATION OF AN RfD FOR TCDD 7 The NAS specifically recommended that EPA derive an RfD for TCDD. Through a 8 transparent study selection process, EPA identified key studies from both human epidemiologic 9 studies and animal bioassays. To select candidate PODs for its RfD methodology, EPA applied 10 additional processes to the key human epidemiologic studies and animal bioassays. Figure ES-3 11 (exposure-response array) shows the entire candidate PODs graphically in terms of 12 human-equivalent intake (ng/kg-day). The human study endpoints are shown at the far left of the 13 figure and, to the right, the rodent endpoints are arranged by the following study categories: less 14 than 1 year, greater than 1 year, reproductive, and developmental. 15 For each noncancer epidemiologic study that EPA selected as key, EPA evaluated the 16 dose-response information developed by the study authors to determine whether the study 17 provided noncancer effects and TCDD-relevant exposure data for a toxicologically-relevant 18 endpoint. If such data were available, EPA identified a NOAEL or LOAEL as a candidate POD. 19 Then, EPA used the Emond human PBPK model to estimate the continuous oral daily intake 20 (ng/kg-day) that would lead to the relevant blood TCDD concentrations associated with the 21 candidate POD. If all of this information was available, then the result was included as a 22 candidate POD. 23 Through this process, EPA identified health effects from the following 24 four epidemiologic studies to be considered as the basis for the RfD: Eskenazi et al. (2002, 25 197168)(reproductive--increased length of menstrual cycle), Alaluusua et al. (2004, 197142) 26 (developmental--tooth development), Mocarelli et al. (2008, 199595) (reproductive--decreased 27 sperm concentrations and motility), and Baccarelli et al. (2008, 197059) 28 (developmental--increased thyroid-stimulating hormone levels in neonates). All four studies are 29 from the Seveso cohort, whose members were exposed environmentally to high peak 30 concentrations of TCDD as a consequence of an industrial accident. This complicated the 31 estimation of average daily doses associated with these specific endpoints, however EPA was This document is a draftfor review purposes only and does not constitute Agency policy. xxxvi 1 able to calculate candidate PODs for derivation of an RfD from each of these human studies (see 2 Section 4.2.3). The Alaluusua et al. (2004, 197142) and Eskenazi et al. (2002, 197168) studies 3 had PODs well above the Mocarelli et al. (2008, 199595) and Baccarelli et al. (2008, 197059); 4 because the LOAEL in Eskenazi et al. (2002, 197168) is almost 2 orders of magnitude higher 5 than the LOAELs for Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 199595), it was 6 not considered further as a candidate POD for derivation of the RfD. 7 Figure ES-4 summarizes the strategy employed for identifying and selecting candidate 8 PODs from the key animal bioassays EPA identified for use in noncancer dose-response analysis 9 of TCDD (see Section 4.2.4). For each noncancer endpoint, EPA first evaluated the 10 toxicological relevance of each endpoint, rejecting those judged not to be relevant for RfD 11 derivation (Section 4.2.1, Appendix G). Next, initial PODs (NOAELs, LOAELs, and BMDLs) 12 based on the first-order body burden metric, and expressed as continuous human-equivalent oral 13 daily doses (HEDs), were determined for all relevant endpoints. 14 Because there were very few NOAELs, and BMDL modeling was largely unsuccessful 15 due to data limitations, the next stage of evaluation was carried out using LOAELs only. 16 Endpoints not observed at the LOAEL (i.e., reported at higher doses) with BMDLs greater than 17 the LOAEL were eliminated from further analysis, as they would not be considered as candidates 18 for the final POD on either a BMDL or NOAEL/LOAEL basis (i.e. the POD would be higher 19 than the PODs of other relevant endpoints). In addition, all endpoints with HEDs for LOAELs 20 (LOAELHEDs) beyond a 100-fold range of the lowest identified LOAELhed were eliminated 21 from further consideration, as they would not be potential POD candidates either (i.e. the POD 22 would be higher than the PODs of other relevant endpoints). For the remaining endpoints, EPA 23 then determined final potential PODs (NOAELs, LOAELs, and BMDLs) based on TCDD blood 24 concentrations obtained from the Emond rodent PBPK models. HEDs were then estimated for 25 each of these PODs using the Emond human PBPK model. From these HEDs, a PODhed was 26 selected for each study as the basis for the candidate RfD, to which appropriate uncertainty 27 factors were applied following EPA guidelines. The resulting candidate RfDs were then 28 considered in the final selection process for the RfD. Other endpoints occurring at slightly 29 higher doses representing additional effects associated with TCDD exposure (beyond the 30 100-fold LOAEL range) were evaluated, modeled, and included in the final candidate RfD array This document is a draftfor review purposes only and does not constitute Agency policy. xxxvii 1 to examine endpoints not evaluated by studies with lower PODs. In addition, BMD modeling 2 based on administered dose was performed on all endpoints for comparison purposes. 3 For BMD modeling, EPA has used a 10% BMR for dichotomous data for all endpoints; 4 no developmental studies were identified with designs that incorporate litter effects, for which a 5 5% BMR would be used (U.S. EPA, 2000, 052150). For continuous endpoints in this document, 6 EPA has used a BMR of 1 standard deviation from the control mean whenever a specific 7 toxicologically-relevant BMR could not be defined. Importantly, the 2003 Reassessment defined 8 the ED01 as 1% of the maximal response for a given endpoint, not as a 1% change from control. 9 Because RfD derivation is one goal of this document, the noncancer modeling effort undertaken 10 here differs substantially from the modeling in the 2003 Reassessment. Evaluation of BMD 11 modeling performance, goodness-of-fit, dose-response data, and resulting BMD and BMDL 12 estimates included statistical criteria as well as expert judgment of their statistical and 13 toxicological properties. EPA has reported and evaluated the BMD results using the standard 14 suite of goodness-of-fit measures from the benchmark dose modeling software (BMDS 2.1). 15 These include chi-square p-values, Akaike's Information Criterion (AIC), scaled residuals at 16 each dose level and plots of the fitted models. In some cases, when restricted parameters hit a 17 bound, EPA used likelihood ratio tests to evaluate whether the improvement in fit afforded by 18 estimating additional parameters could be justified. Goodness-of-fit measures are reported for 19 all key data sets in Appendix E. (See Section 4.2.4.2 for a more complete description of the 20 benchmark dose modeling criteria for model evaluation.) 21 For selection of the POD to serve as the basis of the RfD, EPA gave the epidemiologic 22 studies the highest consideration because human data are preferred in the derivation of an RfD, 23 given that the underlying epidemiologic and animal bioassay data are of comparable quality. 24 This preference for epidemiologic study data also is consistent with reccomendations of panelists 25 at the Dioxin Workshop (see U.S. EPA, 2009, 543757, Appendix A). Figure ES-5 arrays the 26 candidate RfDs from both the human and animal bioassays. The human studies included in 27 Figure ES-5 (Alaluusua et al., 2004, 197142; Baccarelli et al., 2008, 197059; Mocarelli et al., 28 2008, 199595) each evaluate a segment of the Seveso civilian population (i.e., not an 29 occupational cohort) exposed directly to TCDD released from an industrial accident. In this 30 document, EPA uses the Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 199595) This document is a draftfor review purposes only and does not constitute Agency policy. xxxviii 1 studies as co-critical studies in deriving the RfD (Section 4.3).5 In the Seveso cohort exposures 2 were primarily to TCDD, the chemical of concern, with apparently minimal DLC exposures 3 beyond those associated with background intake,6 making these studies highly appropriate for 4 use in RfD derivation for TCDD. In addition, health effects associated with TCDD exposures 5 were observed in humans, the species of concern whose health protection is represented by the 6 RfD, eliminating the uncertainty associated with interspecies extrapolation. The cohort members 7 who were evaluated included infants (exposed in utero) and adults who were exposed when they 8 were less than 10 years of age. The inclusion of these studies among the RfDs derived also may 9 characterize noncancer health effects associated with TCDD exposures in potentially vulnerable 10 populations, thus accounting for some part of the intraspecies uncertainty in the RfD. Finally, 11 the two virtually identical RfDs from different endpoints in the Baccarelli et al. (2008, 197059) 12 and Mocarelli et al. (2008, 199595) studies provide an additional level of confidence in the use 13 of these data for derivation of the RfD for TCDD. 14 Although the human data are preferred, Figure ES-5 presents a number of animal studies 15 with RfDs that are lower than the human RfDs. To a large extent, this is expected because a 16 10-fold interspecies uncertainty factor is generally used to extrapolate from test-animal species to 17 humans, intended to provide a conservative estimate of an RfD that would be derived directly 18 from human data. Two of the rat bioassays among this group of studies--Bell et al. (2007, 19 197041) and NTP (2006, 197605)--are of particular note. Both studies were recently conducted 20 and very well designed and conducted, using 30 or more animals per dose group; both also are 21 consistent with and, in part, have helped to define the current state of practice in the field. 22 Bell et al. (2007, 197041) evaluated several reproductive and developmental endpoints, initiating 23 TCDD exposures well before mating and continuing through gestation. NTP (2006, 197605) is 24 the most comprehensive evaluation of TCDD chronic toxicity in rodents to date, evaluating 25 dozens of endpoints at several time points in all major tissues. Thus, proximity of the RfDs 26 derived from these two high quality, recent studies, provide additional support for the use of the 27 human data for RfD derivation. 5The candidate RfD for Alaluusua et al. (2004, 197142) was approximately 2 orders of magnitude higher than the RfDs for Mocarelli et al. (2008, 199595) and Baccarelli et al. (2008, 197059), thus, it was not included as a co-critical study for the RfD. 6As an example, note the lack of statistically significant effects reported by Baccarelli et al. (2008, 197059; Figure 2 C and D) in regression models based on either maternal plasma levels of non-coplaner PCBs or total TEQ on neonatal TSH levels. This document is a draftfor review purposes only and does not constitute Agency policy. xxxix 1 There are several animal bioassay candidate RfDs at the lower end of the RfD range in 2 Figure ES-5 that are more than 10-fold below the human-based RfDs. Two of these studies 3 report effects that are analogous to the endpoints reported in the three human studies and support 4 the RfDs based on human data. Specifically, decreased sperm production in Latchoumydandane 5 and Mathur (2002, 197498) is consistent with the decreased sperm counts and other sperm 6 effects in Baccarelli et al. (2008, 197059), and missing molars in Keller et al. (2007, 198526; 7 2008, 198531; 2008, 198033) are similar to the dental defects seen in Alaluusua et al. (2004, 8 197142). Thus, because these endpoints have been associated with TCDD exposures in humans, 9 these animal studies would not be selected for RfD derivation in preference to human data 10 showing similar effects. 11 Another characteristic of the remaining studies in the lower end of the candidate RfD 12 distribution is that they are dominated by mouse studies (comprising 6 of the 8 lowest 13 rodent-based RfDs). EPA considers the candidate RfD estimates based on mouse data to be 14 much more uncertain than either the rat or human candidate RfD estimates. The EPA considers 15 the Emond mouse PBPK model to be the most uncertain of toxicokinetic models used to estimate 16 the PODs because of the lack of key mouse-specific data, particularly for the gestational 17 component (see Section 3.3.4.3.2.5). The LOAELHEDs identified in mouse bioassays are low 18 primarily because of the large toxicokinetic interspecies extrapolation factors used for mice, for 19 which there is more potential for error. The ratio of administered dose to HED (Da:HED) ranges 20 from 65 to 1,227 depending on the duration of exposure. The Da:HED for mice is, on average, 21 about four times larger than that used for rats. In addition, each one of the mouse studies has 22 other qualitative limitations and uncertainties that make them less desirable candidates as the 23 basis for the RfD than the human studies. 24 The most relevant human PODs are based on the Baccarelli et al. (2008, 197059) and 25 Mocarelli et al. (2008, 199595) studies, which exhibited similar LOAELs of 0.024 and 26 0.020 ng/kg-day, respectively. For Baccarelli et al. (2008, 197059), EPA defined a LOAEL as 27 the group mean of 39 ppt TCDD in neonatal plasma which corresponds to thyroid-stimulating 28 hormone (TSH) values above 5 pU/mL. The World Health Organization (WHO, 1994) 29 established the 5 pU/mL standard as an indicator of potential iodine deficiency and potential 30 thyroid problems in neonates. Increased TSH levels are indicative of decreased thyroid hormone 31 (T4 and/or T3) levels. For TCDD, the toxicological concern is not likely to be iodine uptake This document is a draftfor review purposes only and does not constitute Agency policy. xl DRAFT--DO NOT CITE OR QUOTE 1 inhibition, but rather increased metabolism and clearance of T4, as evidenced in a number of 2 animal studies (e.g., Seo et al., 1995, 197869). Clinically, a TSH level of >4 gU/mL in a 3 pregnant woman is followed up by an assessment of free T4, and treatment with L-thyroxine is 4 prescribed if T4 levels are low (Glinoer and Delange, 2000). This is to ensure a sufficient supply 5 of T4 for the fetus, which relies on maternal T4 exclusively during the 1st half of pregnancy 6 (Chan et al., 2005; Morreale de Escobar et al., 2000; Calvo et al., 2002). Adequate levels of 7 thyroid hormone also are essential in the newborn and young infant as this is a period of active 8 brain development (Glinoer and Delange, 2000; Zoeller and Rovet, 2004). Thyroid hormone 9 disruption during pregnancy and in the neonatal period can lead to neurological deficiencies. 10 Baccarelli et al. (2008, 197059) showed, in graphical form, how the TSH distribution in 11 each of three categorical exposure groups (reference, zone A, and zone B--representing 12 increasing TCDD exposure) shifted to higher TSH values with increasing exposure. The 13 individuals comprising the above 5 gU/mL group were from all three categorical exposure 14 groups, not just from the highest exposure group. Therefore, EPA was able to designate a 15 LOAEL independently of the nominal categorical exposure groups for TSH values above 16 5 gU/mL. Baccarelli et al. (2008, 197059) did not estimate the equivalent oral intake associated 17 with TCDD serum concentrations, rather they provided neonatal serum TCDD concentrations for 18 the groups above and below 5 gU/mL. EPA estimated the maternal intake at the LOAEL from a 19 maternal serum-TCDD/TSH regression model presented in Baccarelli et al. (2008, 197059) by 20 estimating the maternal TCDD lipid adjusted serum concentration (LASC) at which neonatal 21 TSH exceeded 5 gU/mL. EPA then used the Emond PBPK model to estimate the continuous 22 daily TCDD intake that would result in this TCDD LASC. The resulting predicted maternal 23 daily intake rate established the LOAEL (0.024 ng/kg-day). EPA did not defined a NOAEL 24 because it is not clear what maternal intake should be assigned to the group below 5 gU/mL. 25 For Mocarelli et al. (2008, 199595), EPA defined a LOAEL as the lowest exposed group 26 mean of 68 ppt (1st-quartile) corresponding to decreased sperm concentrations (20%) and 27 decreased motile sperm counts (11%) in men who were 1-9 years old at the time of the Seveso 28 accident (initial TCDD exposure event). Although a decrease in sperm concentration of 29 20% likely would not have clinical significance for an individual, EPA's concern is that such 30 decreases associated with TCDD exposures could lead to shifts in the distributions of these 31 measures in the general population. Such shifts could result in decreased fertility in men at the This document is a draftfor review purposes only and does not constitute Agency policy. xli DRAFT--DO NOT CITE OR QUOTE 1 low end of these population distributions. In the group exposed due to the Seveso accident, 2 individuals one standard deviation below the mean are just above the cut-off used by clinicians 3 (20 million/ml) to indicate follow-up for potential reproductive impact in affected individuals, 4 indicating that a number of individuals in the exposed group likely had sperm concentrations less 5 than 20 million/ml; EPA could not obtain the individual data to determine the exact number of 6 men in this category. EPA judged that the impact on sperm concentration and quality reported 7 by Mocarelli et al. (2008, 199595) is biologically significant given the potential for functional 8 impairment as a consequence of potential shifts in the distribution of these male fertility 9 measures in an exposed population. 10 For Mocarelli et al. (2008, 199595), TCDD LASC levels were measured within 11 approximately one year of the initial exposure event. Because effects were only observed in men 12 who were under 10 years of age at the time of exposure, EPA assumed a maximum 10-year 13 critical exposure window for elicitation of these effects. EPA has estimated a continuous daily 14 oral intake of 0.020 ng/kg-day associated with the designated LOAEL from the lowest exposure 15 group (68 ppt), (see Section 4.2.3.2). The reference group is not designated as a NOAEL 16 because there is no clear zero-exposure measurement for any of these endpoints, particularly 17 considering the contribution of background exposure to DLCs, which futher complicates the 18 interpretation of the reference group response as a true "control" response (see discussion in 19 Section 4.4). However, males less than 10 years old can be designated as a sensitive population 20 by comparison to older males who were not affected. 21 The two human studies, Baccarelli et al. (2008, 197059) and Mocarelli et al. (2008, 22 199595), have similar LOAELs of 0.024 and 0.020 ng/kg-day, respectively. Together, these 23 two studies constitute the best foundation for establishing a POD for the RfD, and are designated 24 as co-principal studies. Therefore, increased TSH in neonates (Baccarelli et al., 2008, 197059) 25 and male reproductive effects (decreased sperm count and motility) are designated as cocritical 26 effects. Although the exposure estimate used in determination of the LOAEL for Mocarelli et al. 27 (2008, 199595) is more uncertain than the Baccarelli et al. (2008, 197059) exposure estimate, the 28 slightly lower LOAEL of 0.020 ng/kg-day from Mocarelli et al. (2008, 199595) is designated as 29 the POD. 30 EPA used a composite UF of 30 for both studies. EPA applied a factor of 10 for UFLto 31 account for lack of a NOAEL. EPA also applied a factor of 3 (1005) for UFHto account for This document is a draftfor review purposes only and does not constitute Agency policy. xlii DRAFT--DO NOT CITE OR QUOTE 1 human interindividual variability because the effects were elicited in sensitive populations. A 2 further reduction to 1 was not made because the sample sizes in these two epidemiologic studies 3 were relatively small, which, combined with uncertainty in exposure estimation, may not fully 4 capture the range of interindividual variability. The resulting RfD for TCDD in standard units is 5 7 x 10-10 mg/kg-day. 6 7 WEIGHT-OF-EVIDENCE STATEMENT FOR CARCINOGENICITY 8 The NAS recommended that EPA update its cancer classification for TCDD and the 9 weight-of-evidence (WOE) statement to reflect the current state of the science and incorporate 10 the latest EPA Cancer Guidelines (U.S. EPA, 2005, 086237). Several notable new studies 11 addressing TCDD's carcinogenic potential have been published since the release of EPA's 12 2003 Reassessment, including several new studies of the Seveso epidemiologic cohort and an 13 NTP 2-year cancer bioassay in female rats (NTP, 2006, 197605). 14 Under the 2005 Guidelinesfor Carcinogen Risk Assessment (U.S. EPA, 2005, 086237) 15 TCDD is characterized as carcinogenic to humans, based on the available data as of 2009 (see 16 Section 5.1.2). When evaluating the carcinogenic potential of a compound, EPA employs a 17 WOE approach in which all available information is evaluated and considered. In the case of 18 TCDD, EPA based the classification on numerous lines of evidence, including: multiple 19 occupationally- and accidentally-exposed epidemiologic cohorts showing an association between 20 TCDD exposure and certain cancers or increased mortality from all cancers; extensive evidence 21 of carcinogenicity at multiple tumor sites in both sexes of multiple species of experimental 22 animals; consensus that the mode of TCDD's carcinogenic action in animals involves aryl 23 hydrocarbon receptor (AhR)-dependent key precursor events and proceeds through modification 24 of one or more of a number of cellular processes; the human AhR and rodent AhR are similar in 25 structure and function, and human and rodent tissue and organ cultures respond to TCDD in a 26 similar manner and at similar concentrations; and general scientific consensus that AhR 27 activation is anticipated to occur in humans and may progress to tumors. 28 Most evidence suggests that the majority of toxic effects of TCDD are mediated by 29 interaction with the AhR. EPA considers interaction with the AhR to be a necessary, but not 30 sufficient, event in TCDD carcinogenesis. Although AhR binding and activation by TCDD is 31 considered to be a key event in TCDD carcinogenesis, the sequence of key events following AhR This document is a draftfor review purposes only and does not constitute Agency policy. xliii DRAFT--DO NOT CITE OR QUOTE 1 activation that ultimately leads to the development of cancer is unknown (See Section 5.1.2.3). 2 Therefore, EPA has determined that TCDD's mode of action, as defined by the 2005 Cancer 3 Guidelines, is unknown. Since the mode of action for TCDD carcinogenesis is not known, EPA 4 has used a low dose linear extrapolation approach in the development of a cancer oral slope 5 factor. 6 7 DERIVATION OF CANDIDATE OSFs FROM EPIDEMIOLOGIC STUDIES AND 8 ANIMAL BIOASSAYS 9 In response to the NAS concerns that EPA evaluate data published since the 10 2003 Reassessment and better justify its approach to cancer dose-response modeling, EPA has 11 developed candidate OSFs using epidemiologic studies and animal bioassays for TCDD, 12 including both new evaluations of data from the 2003 Reassessment and also the assessment of 13 new studies. The BMR level that has been used for the POD in deriving the cancer OSF is 14 one percent extra risk, which is close to the observable response data for most data sets and, 15 therefore, best represents low dose cancer risks (see Section 5.2.3.2.6.11). EPA has chosen a 16 single BMR for consistency across studies. 17 There are several well-studied occupationally-exposed epidemiologic cohorts showing an 18 association between TCDD and increased all-cancer mortality, and several epidemiologic 19 cohorts exposed to TCDD as a consequence of industrial accidents showing an association 20 between TCDD and cancer or cancer mortality (see Section 5.2.3.1). The 2003 Reassessment 21 included cancer dose-response analyses based on the following three occupational cohorts: the 22 NIOSH cohort, an occupational cohort subject to chronic TCDD exposures (Steenland et al., 23 2001, 197433); the Hamburg cohort, an occupational cohort also subject to chronic TCDD 24 exposures (Becher et al., 1998, 197173); and the BASF cohort, an occupational cohort subject to 25 peak TCDD exposures through clean-up following an industrial accident (Ott and Zober, 1996, 26 198101). In this document, EPA determined that each of these studies met the epidemiologic 27 study inclusion criteria. Thus, after further evaluating the OSFs presented in the 2003 28 Reassessment for these three studies, EPA accepted those OSF estimates and retained them as 29 candidate OSFs in this document. These OSF estimates are arrayed in Figure ES-6, along with 30 the other OSFs calculated by EPA in this document. EPA also determined that three additional 31 studies met the epidemiologic study inclusion criteria: Cheng et al. (2006, 523122) and Collins This document is a draftfor review purposes only and does not constitute Agency policy. xliv DRAFT--DO NOT CITE OR QUOTE 1 et al. (2009, 197627) (NIOSH cohort) and Warner et al. (2002, 197489) (Seveso cohort). EPA 2 determined that the data presented in Collins et al. (2009, 197627) were not sufficient to derive 3 an OSF, and EPA was unable to derive a credible OSF from the data presented by Warner et al. 4 (2002, 197489) (see discussions in Section 5.2.3.1). 5 EPA did derive an OSF from Cheng et al. (2006, 523122), as detailed in Text Box ES-1. 6 In Table ES-1, EPA presents estimates of OSFs for specific TCDD intake rates based on target 7 risk levels of 1 x 10-2, through 1 x 10-7 based on Cheng et al. (2006, 523122). Note that there 8 are two nonlinear steps in the estimation of risk-specific doses from the Cheng et al. model. 9 First, fat-AUC (AUCRL) and the incremental cancer mortality risk (RD) do not have a linear 10 relationship (Equation 5-4); however, the relationship becomes virtually linear below an 11 incremental risk of 10-3 (see Table ES-1). Second, TCDD fat concentration is not linear with 12 oral intake in the Emond human PBPK model (see Section 3); this relationship also is close to 13 linear below the 10-5 risk level. The resulting predicted cancer-mortality risk is approximately 14 linear with daily oral intake at low doses. 15 EPA also identified candidate OSFs for TCDD from key animal bioassays (see 16 Section 5.2.3.2). Based on the inclusion criteria, EPA selected five key rodent cancer bioassays 17 suitable for quantitative dose-response assessment. These included Della Porta et al. (1987, 18 197405), Kociba et al. (1978, 001818), NTP (1982, 543764), and Toth et al. (1979, 197109) that 19 were evaluated in the 2003 Reassessment, and the new NTP (2006, 197605) rat chronic bioassay. 20 EPA conducted dose-response modeling for each tumor type separately (individual tumor 21 models) as well as for composite tumor incidence (multiple tumor models). The tumor types that 22 EPA analyzed are shown in Table ES-2. 23 For each in vivo animal cancer study that qualified for TCDD dose-response assessment, 24 EPA selected the species/sex/tumor dataset combinations characterized as having statistically 25 significant increases in tumor incidences, then used the Emond rodent PBPK model to estimate 26 blood concentrations corresponding to each study's average daily administered dose for use in 27 dose-response modeling. BMDL01s were then estimated for the blood concentration by 28 two different methodologies: (1) using the multistage cancer model for each species/sex/tumor 29 combination within each study, and (2) using a Bayesian Markov Chain Monte Carlo framework 30 that assumes independence of tumors, modeling all tumors together for each species/sex This document is a draftfor review purposes only and does not constitute Agency policy. xlv DRAFT--DO NOT CITE OR QUOTE Text Box ES-1. OSF Calculations Using Cheng et al. (2006, 523122) Information. To develop cancer risks for TCDD, EPA used the modeling results of the Cheng analysis, with conversion to oral intake using the Emond human PBPK model as follows. The slope (p) from the Cheng analysis is the slope of the linear relationship between the natural logarithm of the rate ratio (RR) and the cumulative fat TCDD concentration (fat-AUC). Conceptually, the slope (P) is similar to an OSF, except that it is expressed in terms of fat-AUC rather than intake. Also, the slope represents the incremental increase in cancer mortality (expressed as an RR) above the background TCDD exposure experienced by the NIOSH cohort rather than above zero. Using the upper 95% bound on p and assuming that the slope is the same below the NIOSH cohort background exposure level (approximately 5 ppt/yr TCDD fat concentration), EPA calculated risk-specific doses (as daily oral intakes) for TCDD for risk levels of concern to EPA. The risk-specific doses were estimated from the Emond human PBPK model for the lifetime-average TCDD fat concentrations corresponding to the fat-AUC predicted by the Cheng et al. model for each of the risk levels of concern. The steps in this computation are as follows: Background cancer mortality risk estimate (R0). EPA used an R0 of 0.112 as reported by Cheng et al. (2006, 523122) Total cancer mortality risk in the exposed group associated with a specified (extra) risk level (RL) of fatal cancer (TRr_). A TRr_ associated with any given extra risk level (e.g., 0.01, 1 x 10-6) can be calculated using the following relationship for extra risk: ER TRrl - R 1- R (Eq. ES-1) Incremental cancer mortality risk in the exposed population based on a given extra risk (RD). R D, is calculated as the difference between the total risk and background risk and expressed in terms of R _ and R 0 by combining Equations ES-2 and ES-1. r d = Tr rl R0 Rd= RL x (1 - R0) (Eq. ES -2) (Eq. ES -3) Cumulative TCDD concentration in the fat compartment for a given extra risk (AUCRL). AUCrl is then calculated by taking the natural logarithm of Equation 3 from Cheng et al. (2006, 523122), rearranging and substituting for RR1(RR = [RD+ R0]/R0): A U C rl = ln((RD + R^ R W (Eq. ES -4) where p * is the central-tendency regression slope or the 95% upper bound (p95) determined by summing the regression coefficient (P) and the product of 1.96 and the standard error of the regression coefficient, yielding an estimate of 6.0 x 10-6 per ppt-year lipid adjusted serum TCDD, as follows: P95 =P +1.96* SE (Eq. ES -5) Continuous daily TCDD intake associated with a given extra risk IDR 1. Because the fat concentrations generated by CADM are not linear with oral exposure at higher doses, a single oral slope factor to be used for all risk levels cannot be obtained; the response is approximately linear with fat concentrations and oral intake at lower doses. Instead, a risk-specific DRLmust be estimated by converting the respective AUCrl to the corresponding lifetime daily intake, using an appropriate human toxicokinetic model. EPA has chosen to use the Emond human PBPK model for this purpose because the CADM configuration does not facilitate this process and so that the dose conversions are consistent with those used in the derivation of the RfD. A DRL is obtained from the Emond model by finding the average lifetime daily intake corresponding to the AUCrl in the fat compartment. This document is a draftfor review purposes only and does not constitute Agency policy. xlvi DRAFT--DO NOT CITE OR QUOTE 1 combination within each study. The final selected models were subjected to goodness-of-fit tests 2 and visual inspection of fit to the raw data. Thus, for each sex/species combination within each 3 study, EPA generated a BMDL01 for each single tumor type and another BMDL01 for the 4 combined tumors. Using the Emond human PBPK model, BMDLHEDs were then calculated for 5 each of the BMDL01s, and using a linear extrapolation, OSFs were calculated by 6 OSF = 0.01/BMDLhed The highest OSF for a species/sex combination for either a single tumor 7 type or all combined tumors was selected as a candidate OSF. The OSF candidates from the key 8 animal bioassays are shown in Table ES-2. 9 10 DERIVATION OF TCDD ORAL SLOPE FACTOR AND RISK ESTIMATES 11 EPA was able to derive OSFs for tumor incidence data from five animal cancer 12 bioassays, as well as for cancer mortality data from four epidemiological cohort studies that were 13 selected for TCDD dose-response modeling using the study inclusion criteria (see Section 5.3). 14 These OSFs are arrayed in Figure ES-6. For the animal data, OSFs based on individual tumors 15 were developed for 28 study/sex/endpoint combinations, and the results ranged from 1.8 x 104to 16 5.8 x 106(per mg/kg-day). The OSFs based on combined tumors were developed for 17 seven study/sex combinations, and the results ranged from 3.2 x 105to 9.4 x 106 (per 18 mg/kg-day). EPA also developed OSFs based on four epidemiologic studies from three cohorts, 19 ranging from 3.75 x 105to 2.5 x 106(per mg/kg-day). 20 EPA has chosen to use the human data over the animal data as recommended by expert 21 panelists at EPA's 2009 Dioxin Workshop (U.S. EPA, 2009, 522927) and in the 2005 Cancer 22 Guidelines (U.S. EPA, 2005, 086237). OSFs derived from the human data are consistent with 23 the animal bioassay results; human OSFs fall within the same range as the animal bioassay 24 OSFs. 25 Among the human studies, the occupational TCDD exposures in the NIOSH and 26 Hamburg cohorts are assumed to be reasonably constant over the duration of occupational 27 exposure. In contrast, the TCDD exposure pattern for the Seveso and BASF accidents is acute, 28 high dose, followed by low-level background exposure. Such exposure patterns similar to those 29 experienced by the BASF and Seveso cohorts have been shown to yield higher estimates of risk 30 when compared to constant exposure scenarios with similar total exposure magnitudes (Kim 31 et al., 2003, 199146; Murdoch and Krewski, 1988, 548718; Murdoch et al., 1992, 548719). This document is a draftfor review purposes only and does not constitute Agency policy. xlvii DRAFT--DO NOT CITE OR QUOTE 1 Thus, EPA has judged that the NIOSH and Hamburg cohort response data are more relevant than 2 the BASF and Seveso data for assessing cancer risks from continuous ambient TCDD exposure 3 in the general population. 4 The NIOSH (Steenland et al., 2001, 197433; Cheng et al., 2006, 523122) and Hamburg 5 (Becher et al., 1998, 197173) cohort studies report cumulative TCDD levels in the serum for 6 cohort members. The most significant difference among the Cheng et al. (2006, 523122) 7 analysis and those of Steenland et al. (2001, 197433) and Becher et al. (1998, 197173) is the 8 method used to back-extrapolate exposure concentrations based on serum TCDD measurements. 9 Steenland et al. (2001, 197433) and Becher et al. (1998, 197173) back-extrapolated exposures 10 and body burdens using a first-order model with a constant half-life. In contrast, Cheng et al. 11 (2006, 523122) back-extrapolated body burdens using a kinetic modeling approach that 12 incorporated concentration- and age-dependent elimination kinetics. 13 Although all three of these are high-quality studies, the kinetic modeling used by Cheng 14 et al. (2006, 523122) is judged to better reflect TCDD pharmacokinetics, as currently 15 understood, than the first-order models used by Steenland et al. (2001, 197433) and Becher et al. 16 (1998, 197173). EPA believes that the representation of physiological processes provided by 17 Cheng et al (2006, 523122) is more realistic than the assumption of simple first-order kinetics 18 and this outweighs the attendant modeling uncertainties. Furthermore, the use of kinetic 19 modeling is consistent with recommendations both by the NAS and the Dioxin Workshop panel. 20 EPA, therefore, has decided to use the results of the Cheng et al. (2006, 523122) study for 21 derivation of the TCDD OSF based on total cancer mortality as calculated by EPA using data 22 and models from the Cheng et al. (2006, 523122) study, as described in Section 5.2.3.1.2. 23 Although the OSF is only strictly defined for exposures above the background exposure 24 experienced by the NIOSH cohort, which was assumed to be 0.5 pg/kg-day TCDD, or 25 5 pg/kg-day total TEQ, EPA assumes that the slope (risk vs. blood concentration) is the same 26 below those background exposure levels as it is above. Table ES-1 shows the oral slope factors 27 at specific target risk levels (OSFRLs) which range from 1.1 x 105to 1.3 x 106per (mg/kg-day). 28 EPA recommends the use of an OSF of 1 x 106per (mg/kg-day) when the target risk range is 10- 5 29 to 10-7. 30 This document is a draftfor review purposes only and does not constitute Agency policy. xlviii DRAFT--DO NOT CITE OR QUOTE 1 CONSIDERATION OF NONLINEAR DOSE-RESPONSE APPROACHES FOR 2 CANCER 3 The NAS focused much of its review on EPA's derivation of a cancer slope factor, 4 commenting extensively on the extrapolation of dose-response modeling below the POD. The 5 NAS questioned EPA's choice of a linear, nonthreshold model for extrapolating risk associated 6 with exposure levels below the POD, concluding that the current scientific evidence was 7 sufficient to justify the use of nonlinear methods when extrapolating below the POD for dioxin 8 carcinogenicity. 9 While, based on the 2005 Cancer Guidelines, EPA deemed linear extrapolation to be 10 most appropriate for TCDD, EPA carefully considered the NAS recommendation to provide risk 11 estimates using both linear and nonlinear methods. In this document, EPA has evaluated the 12 information available for identifying a threshold and for estimating the shape of the 13 dose-response curve below the POD (see Section 5.2.3.4). EPA presents a hypothetical sublinear 14 dose-response modeling example of rodent carcinogenicity. EPA also presents two illustrative 15 examples of RfD development (i.e., nonlinear method) for carcinogenic effects of TCDD, using 16 data derived from animal bioassays. EPA derives illustrative RfDs for cancer based on 17 combined tumor response and also on hypothesized key events in TCDD's MOA for female rat 18 liver and lung tumors. EPA identifies a number of limitations that prevent making strong 19 conclusions based on the nonlinear dose-response modeling exercises. 20 21 FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS 22 EPA also addresses the third key recommendation of the NAS, specifically, improving 23 transparency, thoroughness, and clarity in quantitative uncertainty analysis (see Section 6). In 24 summary, NAS suggested that EPA should 25 26 describe and define (quantitatively to the extent possible) the variability and 27 uncertainty for key assumptions used for each key endpoint-specific risk 28 assessment (choices of data set, POD, model, and dose metric), 29 incorporate probabilistic models to the extent possible to represent the range of 30 plausible values, 31 clearly state it when quantitation is not possible and explain what would be 32 required to achieve quantitation (NAS, 2006, 198441, p. 9). 33 This document is a draftfor review purposes only and does not constitute Agency policy. xlix DRAFT--DO NOT CITE OR QUOTE 1 Although the NAS summarized the shortfalls in the 2003 Reassessment categorically, the 2 elaborations within their report often contain the qualification "if possible" and do not take a 3 position with regard to the feasibility of many suggestions. With appreciation for the extent of 4 information available for dioxin, EPA's goal herein was to examine the feasibility of a 5 data-driven quantitative uncertainty analysis for TCDD dose-response assessment. 6 In examining feasibility of quantitative uncertainty analysis, EPA recognized that 7 different kinds of uncertainty require different statistical treatment. Cognitive uncertainty 8 concerns uncertainty that can be expressed as probabilities and may be operationalized using 9 either frequentist or Bayesian approaches. For example, classical statistical methods yield 10 distributions on model parameters which reflect sample fluctuations, assuming that the model is 11 true. This type of uncertainty can be taken into account in the BMDL estimation. Also, for 12 TCDD epidemiologic data, the dose reconstruction often involves assumptions that may be 13 amenable to data-driven uncertainty analysis if sufficient data can be retrieved; back14 extrapolated TCDD levels, biological half-life, body fat, and background levels are example 15 variables that could be included in such an analysis. In addition, a Monte Carlo analysis has 16 been examined to develop quantitative uncertainty distributions for the RfD (e.g., Swartout et al., 17 1998, 093460). Given a set of animal bioassay data, quantifying dose-response uncertainty may 18 be approached in different ways. The differences reflect different types of uncertainty that are 19 captured. A recent evaluation enumerates the following possible methodologies (Bussard et al., 20 2009, 543770): 21 22 Benchmark Dose Modeling (BMD): Choose the `best' model, and 23 assess uncertainty assuming this model is true. Supplemental results can compare 24 estimates obtained with different models, and sensitivity analyses can investigate 25 other modeling issues. 26 Probabilistic Inversion with Isotonic Regression (PI-IR): Define 27 model-independent `observational' uncertainty, and look for a model that captures 28 this uncertainty by assuming the selected model is true and providing for a 29 distribution over its parameters. 30 Non-Parametric Bayes (NPB): Choose a prior mean response (potency) 31 curve (potentially a "non-informative prior") and a precision parameter to express 32 prior uncertainty over all increasing dose-response relations, and update this prior 33 distribution with the bioassay data. This document is a draftfor review purposes only and does not constitute Agency policy. l DRAFT--DO NOT CITE OR QUOTE 1 Bayesian Model Averaging (BMA) (as considered here): Choose an 2 initial set of models, and then estimate the parameters of each model with 3 maximum likelihood. Use classical methods to estimate parameter uncertainty, 4 given the truth of the model. Determine a probability weight for each model 5 using the Bayes Information Criterion (BIC), and use these weights to average the 6 model results. 7 8 The first of the above methods involves standard classical statistical methods and captures 9 sampling uncertainty conditional on the truth of the model used. The other methods are "exotic" 10 in the sense that they attempt to capture uncertainty that is not conditional on the truth of a given 11 model. In this response document, EPA has not applied such methods, but recognizes that 12 quantitative uncertainty analysis is possible in these cases. 13 In contrast to cognitive uncertainty, Volitional uncertainty concerns uncertainty regarding 14 choices on the best course of action to take; volitional uncertainty cannot be analyzed by 15 sampling from a probability distribution and, thus, is not amenable to a complete quantitative 16 uncertainty analysis. Some of the choices made in TCDD dose-response assessment that are 17 volitional include: choice of occupational cohort data set or bioassay data set; choice of PODs 18 (e.g., ED01, ED05, and ED10); choice of species, strain, or sex within an animal bioassay; and 19 choice of dose metric (e.g., administered doses, blood concentrations, lipid-adjusted serum 20 concentrations). These volitional uncertainties cannot be quantified by sampling an input 21 distribution. 22 Although EPA has determined that a comprehensive quantitative uncertainty analysis is 23 not feasible because of the limitations discussed above, EPA believes the NAS was requesting 24 that dose-response modeling results be shown for specific choices of interest to TCDD 25 assessment. In response to the NAS concerns, this document provides some limited quantitative 26 comparisons. BMDs, BMDLs, and OSFs from the animal cancer bioassay benchmark dose 27 modeling assuming 1, 5, and 10% extra risk are compared in units of blood concentrations and 28 human equivalent doses (see Tables 5-18 and 5-19, respectively). In addition, central tendency 29 slope estimates and upper bound slope factor estimates based on Cheng et al. (2006, 523122) are 30 presented (see Tables 5-3 and 5-4). For the noncancer effects, key animal study PODs 31 (ng/kg-day) are shown based on different dose metrics: administered dose, first-order body 32 burden HED, and blood concentration (Tables 4-3 and 4-4). EPA has undertaken some limited 33 quantitative uncertainty analyses for the kinetic modeling, presenting a sensitivity analysis and This document is a draftfor review purposes only and does not constitute Agency policy. li DRAFT--DO NOT CITE OR QUOTE 1 uncertainty analysis in dose metrics derived for the risk assessment of TCDD and a detailed 2 discussion on the uncertainty in choice of PBPK model-driven dose metrics. (see Sections 3.3.3 3 and 3.3.5). TCDD kinetic doses from the Emond et al. (2005, 197317; 2006, 197316) PBPK 4 model that is primarily used in the technical analysis in this document are compared with those 5 predicted by the Aylward et al. (2005, 197114) model. 6 Uncertainty quantification is an emerging area in science. There are many examples of 7 highly vetted and peer-reviewed uncertainty analyses based on structured expert judgment. 8 Under this process, experts in effect synthesize a wide diversity of information in generating 9 their subjective probability distributions. Where considerable data exist for an environmental 10 pollutant, such as for the well-studied TCDD, it is natural to ask whether these extensive data can 11 be leveraged more directly in uncertainty quantification. This is an area where research could be 12 focused. Additional research topics relevant to dioxin that could further inform health 13 assessments include population variability of biokinetic constants and threshold mechanisms for 14 the mass action model. Further data and improved methodologies in these areas, combined with 15 developments illustrated elsewhere in this report, will help reduce or better quantify uncertainties 16 and strengthen EPA's understanding of potential health implications of environmental TCDD 17 exposures. This document is a draftfor review purposes only and does not constitute Agency policy. lii DRAFT--DO NOT CITE OR QUOTE Table ES-1. Comparison of fat concentrations, risk specific dose estimates and equivalent oral slope factors based on upper 95thpercentile estimate of regression coefficient of all fatal cancers reported by Cheng et al. (2006, 523122) for selected risk levels Risk level (RL) 1 x 10-2 5 x 10-3 1 x 10-3 5 x 10-4 1 x 10-4 5 x 10-5 1 x 10-5 5 x 10-6 1 x 10-6 5 x 10-7 1 x 10-7 AUCRL (ppt-yr) 1.262 x 104 6.432 x 103 1.307 x 103 6.546 x 102 1.311 x 102 6.558 x 101 1.312 x 101 6.559 x 100 1.312 x 100 6.559 x 10'1 1.312 x 10-1 FATrl (ng/kg) 1.803 x 102 9.189 x 101 1.867 x 101 9.352 x 100 1.873 x 100 9.368 x 10'1 1.874 x 10-1 9.370 x 10'2 1.874 x 10-2 9.370 x 10'3 1.874 x 10-3 Equivalent oral slope Risk specific doseb factors (OSFrl) per (Drl) (ng/kg-day) (mg/kg-day) 8.79 x 10-2 1.1 x 105 3.14 x 10-2 1.6 x 105 2.88 x 10-3 3.5 x 105 9.56 x 10-4 5.2 x 105 1.29 x 10-4 7.8 x 105 5.52 x 10-5 9.1 x 105 8.94 x 10-6 1.1 x 106 4.25 x 10-6 1.2 x 106 8.08 x 10-7 1.2 x 106 4.00 x 10-7 1.3 x 106 7.92 x 10-8 1.3 x 106 aBased on regression coefficient of Cheng et al. (2006, 523122, Table III), excluding observations in the upper 5% range of the exposures; where reported P = 3.3 x 10-6 ppt-years and standard error = 1.4 * 10-6. Upper 95th percentile estimate of regression coefficient (P95) calculated to be 6.04 x 10-6 = (3.3 x 10-6) + 1.96 x (1.4 x 10-6); background cancer mortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 523122). bTo calculate the extra cancer risk (ER) and OSF for any TCDD daily oral intake (D): 1. For D in ng/kg-d, look up the corresponding fat concentration (ng/kg = ppt) from the conversion chart (nongestational lifetime dose metrics) in Appendix C.4.1. 2. Calculate the AUC in ppt-yrs by multiplying the fat concentration by 70 years. 3. Calculate Extra Risk (ER) using the following equation: ER = [exp(AUC x 6.04E-6) x 0.112 - 0.112] - 0.888. 4. Calculate the OSF (mg/kg-d)-1 = 1E6 x (ER - D). Example for risk at the RfD: D = 7 x 10-4ng/kg-d; fat concentration = 6.93 ng/kg; AUC = 70 years x 6.93 ppt = 485 ppt-year; ER = exp(485 ppt-year x 6.04E-6 (ppt-yr)-1) x 0.112 - 0.112) - 0.888 = 3.7 x 10-4 OSF = 1E6 ng/mg x (3.7 x 10-4 - 7 x 10-4 ng/kg-d) = 5.3 x 105 (mg/kg-d)-1. This document is a draftfor review purposes only and does not constitute Agency policy. liii DRAFT--DO NOT CITE OR QUOTE Table ES-2. Tumor points of departure and oral slope factors using blood concentrations Study Sex/species: tumor sites BMDLp1HEpa OSF (ng/kg-day) (per mg/kg-day) NTP, (1982, 543764) Male mice: liver adenoma and carcinoma, lung 1.1E-03 9.4E+6 Toth et al., Male mice: liver tumors (1979, 197109) 1.9E-03 5.2E+6 NTP, (1982, 543764) Female mice: liver adenoma and carcinoma, thyroid adenoma, subcutaneous fibrosarcoma, all lymphomas 5.3E-03 1.9E+6 NTP, (1982, 543764) Female rats: liver neoplasitc nodules, liver adenoma and carcinoma, adrenal cortex adenoma or carcinoma, thyroid follicular cell adenoma 5.7E-03 1.8E+6 Kociba et al., Female rats: liver adenoma carcinoma, oral (1978, 001818) cavity, lung 7.3E-03 1.4E+6 NTP, (1982, 543764) Male rats: thyroid follicular cell adenoma, adrenal cortex adenoma 9.6E-03 1.0E+6 Della Porta et al., Male mice: Hepatocellular carcinoma (1987, 197405) 3.1E-02 3.2E+5 NTP, (2006, 197605) Female rats: liver cholangiocarcinoma, hepatocellular adenoma, oral mucosa squamous cell carcinoma, lung cystic keratinizing epithelioma, pancreas adenoma, carcinoma 2.3E-02 4.4E+5 Kociba et al., Male rats: adrenal cortex adenoma, tongue (1978, 001818) carcinoma, nasal/palate carcinoma 3.1E-02 3.2E+5 aBMDLHEDs are from the multiple tumor analyses, with the exception of Toth et al. (1979, 197109) and Della Porta et al. (1987, 197405) which are the result of modeling single tumor sites. This document is a draftfor review purposes only and does not constitute Agency policy. liv DRAFT--DO NOT CITE OR QUOTE Figure ES-1. EPA's process to evaluate available epidemiologic studies using study inclusion criteria for use in the dose-response analysis of TCDD. EPA applied its TCDD-specific epidemiologic study inclusion criteria to all studies published on TCDD and DLCs. The studies were initially evaluated using five considerations regarded as providing the most relevant kind of information needed for quantitative human health risk analyses. For each study that was published in the peer-reviewed literature, EPA then examined whether the exposures were primarily to TCDD and if the TCDD exposures could be quantified so that dose-response analyses could be conducted. Finally, EPA required that the effective dose and oral exposure be estimable: (1) for cancer, information is required on long-term exposures, (2) for noncancer, information is required regarding the appropriate time window of exposure that is relevant for a specific, nonfatal health endpoint, and (3) for all endpoints, the latency period between TCDD exposure and the onset of the effect is needed. Only studies meeting these criteria were included in EPA's TCDD dose-response analysis. This document is a draftfor review purposes only and does not constitute Agency policy. lv DRAFT--DO NOT CITE OR QUOTE Figure ES-2. EPA's process to evaluate available animal bioassay studies using study inclusion criteria for use in the dose-response analysis of TCDD. EPA evaluated all available in vivo mammalian bioassay studies on TCDD. Studies had to be published in the peer-reviewed literature. Next, to ensure working in the low-dose range for TCDD dose-response analysis, EPA applied dose requirements to the lowest tested average daily doses in each study, with specific requirements for cancer (<1 gg/kg-day), and noncancer (<30 ng/kg-day) studies. Third, EPA required that the animals were exposed via the oral route to only TCDD and that the purity of the TCDD was specified. Finally, the studies were evaluated using four considerations regarded as providing the most relevant kind of information needed for quantitative human health risk analyses from animal bioassay data. Only studies meeting all of these criteria and considerations were included in EPA's TCDD dose-response analysis. This document is a draftfor review purposes only and does not constitute Agency policy. lvi DRAFT--DO NOT CITE OR QUOTE Figure ES-3. Exposure-response array for ingestion exposures to TCDD. ( y ) S 6 6 I- ` | B } 0 O 0 S CM CM CM CM O OO o O OO o CM CD 00 X ( y ) e z o o s ` 'Ie i 9 iie a ( y ) l o o s ` |B 18 e r e s i l o Ie 10 n n ( y ) o o o s ` |B 18 U IL U V ( y ) 1 .0 0 3 ` |B 18 U 8 U IB H B |B 1 8 U 8 U !H 8 I|/\| |B } 0 0 0 H ( y ) 1 . 0 0 3 ` |B 1 8 |> |S A /\0 > |J B 1/\| (lA l) q ` B 8 0 0 3 `Z O O S " I e CD l _____ |e } 0 ! 1 ( y ) 6 Z 6 I- ` | B } 0 B j j n | / \ | ( y ) z o o s ` |B }0 m s ( y ) 3 0 0 3 `J n q }B |/\| p u e p o } B i ( y ) 8 Z 6 I- ` |B 1 8 e q p o ( y ) 9 0 0 3 `d lN (lA l) 3 8 6 1 - ` d l N (lA l) 6 Z 6 1 ` Ie i s q i o i ( d ) S 6 6 I- ` |B 1 8 | | B M 8 g ( y ) s o o s ` |B } 0 u o y o jQ |B 18 8 J 0 1 1 B J -- 11 -- X X IE g 00 00 M- 0 0 o OOO o o OoOo o CM CM CM CM CM ( y ) B S 6 6 1 ` |B } 0 U 0 | 0 6 j| g u b a ( y ) z o o s ` Ie 10 n p o ( y ) 1 .0 0 3 ` |B 18 O U B J J ( y ) 1. 8 6 1 . ` |B 1 8 I U 0 1 U B Q ( 0 ) 8 Z 6 1 ` Ie 1 0 s o a ( 0 ) 9 8 6 1 - ` |B }0 O |jd B O 0 Q ( y ) Z 6 6 I- ` | e } 0 n (lA l) 9 8 6 1 - ` |B } 0 0 1IP A A |B 1 8 Z O jM O |B jL U S |B 1 8 Bnsnn|B|v |B 1 8 |||8 J B 0 0 B g |B 1 8 1118 J B O O 0 1/\| CM x-- o X-- CM CO OOo O Oo 1 1+ + + i i o LU o LU O LU O LU O LU O LU O LU O x-- x - ' x-' (p-6>)/6u `9sop iu9|BAmb9-UBUjnH) 91By UO|lS9BU| This document is a draftfor re\'iew purposes only and does not constitute Agency policy. lvii DRAFT: DO NOT CITE OR QUOTE Figure ES-4. EPA's process to select and identify candidate PODs from key animal bioassays for use in noncancer dose-response analysis of TCDD. For each noncancer endpoint found in the studies that qualified for TCDD dose-response assessment using the study inclusion criteria, EPA first determined if the endpoint was toxicologically relevant. If so, EPA determined the NOAEL, LOAEL, and BMDL Human Equivalent Dose (HED) based on 1st-order body burdens for each endpoint. These potential PODs were examined for statistical relevance and included when the endpoint was observed at the LOAEL. If the BMDL was less than the LOAEL, and if the endpoint was less than the minimum LOAEL x 100, EPA then calculated NOAELs, LOAELs, or BMDLs based on blood concentrations from the Emond rodent PBPK model. Then, for all of the candidate PODs, HEDs were estimated using the Emond human PBPK model. Finally, the lowest group of the toxicologically relevant candidate PODs was selected for final use in derivation of an RfD. This document is a draftfor review purposes only and does not constitute Agency policy. lviii DRAFT--DO NOT CITE OR QUOTE Animal Bioassays see i-'|b10 HEM6S 9002 |e 10 uoyojo 9661 '|B}8 oss 0002 |Bie ejonsj 9Z61 |B}3 eqiDO B9661 '|B je ueieBjjg uba OO2 IB10 lieg OOS IBie ni|3 1-002 lBle 3UEJJ 6Z6L |B je Aejjni/'-j 1-002 'IB je o>iesl|o 8002 ' IBlenriH 0002 |e js mwy 9002d lN 1,002 ie je ueujenejH 9002 '|Bje usu!H0!iA] 1.861 |Eje juojueo 61 leissoA 2002 |B}0 oio|-| 1002 leis !>]SMO>|jeiAi 002 '|B je ms 9861 |Bje oudeoeQ 661 |Bje n 9861 |B}0 31!MM 2861 d lN 2002Jngiew>8 ipiei 6Z61 '|B je ifloi q'B8002 '002 ib ie jeii:>i 8002 |B18 ZD!MO|B!UJS 9002 ie 10 !1 fc-002 lejeensnnieiv 8002 |B10 !||jeo:::.e:g 8002 |B10 HI3JB30O1AJ LOO COD O pL pL pL V- V- OOO CoO o T-- CN V- pLJ pL pLU pL pLLl V- (Aep-Bh/Buj) sjnsodxg |ejq Figure ES-5. Candidate RfD array. Human I This document is a draftfor re\'iew purposes only and does not constitute Agency policy. lix DRAFT: DO NOT CITE OR QUOTE This document is a draftfor review purposes only and does not constitute Agency policy. lx DRAFT: DO NOT CITE OR QUOTE Cancer Slope Factors for 2,3,7,8-TCDD 1.0E+07 Human 1 1 A T I 1 Mouse 1 "*l A 1 1 Rat tOTsO 1.0E+06 HO 0Q); 0 N CcnOT 1.0E+05 o1 o 1 1 1 1 A 1 1 1 ! i i i 1 1 1 1 l 1 1 1 1 1 1 1 -------- 1-------- & & K& .cf <S>& <V . er <X?pN ^vt Fx%V ^ &.< ^ (A'V<<>% \0<u & N? / > K* ' & *V(Tv'' 0' * r'T&** rvX? * ^ <v or < f .o,-<u cU'<u <Vu% \0<u A <^u \0<^u K0<^u ^ KrJ r$F ^ 'Nx<? S*' f?^`S -<&' E <?f Figure ES-6. Candidate oral slope factor array. 1 1. INTRODUCTION 2 3 4 Dioxins and dioxin-like compounds (DLCs), including polychlorinated dibenzo-dioxins, 5 polychlorinated dibenzofurans, and polychlorinated biphenyls are structurally and 6 toxicologically related halogenated dicyclic aromatic hydrocarbons.7 Dioxins and DLCs are 7 released into the environment from several industrial sources such as chemical manufacturing, 8 combustion, and metal processing; from individual activities including the burning of household 9 waste; and from natural processes such as forest fires. Dioxins and DLCs are widely distributed 10 throughout the environment and typically occur as chemical mixtures. Additionally, they do not 11 readily degrade; therefore, levels persist in the environment, build up in the food chain, and 12 accumulate in the tissues of animals. Human exposure to these compounds occurs primarily 13 through the ingestion of contaminated foods (Lorber et al., 2009, 543766), although exposures to 14 other environmental media and by other routes and pathways do occur. 15 The health effects from exposures to dioxins and DLCs have been documented 16 extensively in epidemiologic and toxicologic studies. 2,3,7,8-Tetrachlorodibenzo-p-dioxin 17 (TCDD) is one of the most toxic members of this class of compounds and has a robust 18 toxicologic database. Characterization of TCDD toxicity is critical to the risk assessment of 19 mixtures of dioxins and DLCs because it has been selected repeatedly as the "index chemical" to 20 serve as the basis for standardization of the toxicity of components in a mixture of dioxins and 21 DLCs. The dose-response information for TCDD is used to evaluate risks from exposure to 22 mixtures of DLCs (Van et al., 1998, 198345; Van den Berg et al., 2006, 543769; also see the 23 World Health Organization's Web site for the dioxin toxicity equivalence factors [TEFs]),8 24 therefore, it is imperative to correctly assess the dose response of TCDD and understand the 25 uncertainties and limitations therein. 26 In 2003, the U.S. Environmental Protection Agency (EPA) produced an external review 27 draft of the multiyear comprehensive reassessment of dioxin exposure and human health effects 28 entitled, Exposure and Human Health Reassessment o f 2,3,7,8-Tetrachlorodibenzo-p-Dioxin 29 (TCDD) and Related Compounds (U.S. EPA, 2003, 537122). This draft report, herein called the 7For further information on the chemical structures of these compounds, see U.S. EPA (2003, 537122; 2008, 543774). 8Available at http://www.who.int/ipcs/assessment/tef_update/en/. This document is a draftfor review purposes only and does not constitute Agency policy. 1-1 DRAFT--DO NOT CITE OR QUOTE 1 "2003 Reassessment," consisted of (1) a scientific review of information relating to sources of 2 and exposures to TCDD, other dioxins, and DLCs in the environment; (2) detailed reviews of 3 scientific information on the health effects of TCDD, other dioxins, and DLCs; and (3) an 4 integrated risk characterization for TCDD and related compounds. 5 In 2004, EPA asked the National Research Council of the National Academy of Sciences 6 (NAS) to review the 2003 Reassessment. The NAS Statement of Task was as follows 7 The National Academies' National Research Council will convene an expert committee that will review EPA's 2003 draft reassessment of the risks of dioxins and dioxin-like compounds to assess whether EPA's risk estimates are scientifically robust and whether there is a clear delineation of all substantial uncertainties and variability. To the extent possible, the review will focus on EPA's modeling assumptions, including those associated with the dose-response curve and points of departure; dose ranges and associated likelihood estimates for identified human health outcomes; EPA's quantitative uncertainty analysis; EPA's selection of studies as a basis for its assessments; and gaps in scientific knowledge. The study will also address the following aspects of EPA's 2003 Reassessment: (1) the scientific evidence for classifying dioxin as a human carcinogen; and (2) the validity of the nonthreshold linear dose-response model and the cancer slope factor calculated by EPA through the use of this model. The committee will also provide scientific judgment regarding the usefulness of toxicity equivalence factors (TEFs) in the risk assessment of complex mixtures of dioxins and the uncertainties associated with the use of TEFs. The committee will also review the uncertainty associated with the 2003 Reassessment's approach regarding the analysis of food sampling and human dietary intake data, and, therefore, human exposures, taking into consideration the Institute of Medicine's report Dioxin and Dioxin-Like Compounds in the Food Supply: Strategies to Decrease Exposure. The committee will focus particularly on the risk characterization section of EPA's 2003 Reassessment report and will endeavor to make the uncertainties in such risk assessments more fully understood by decision makers. The committee will review the breadth of the uncertainty and variability associated with risk assessment decisions and numerical choices, including, for example, modeling assumptions, including those associated with the dose-response curve and points of departure. The committee will also review quantitative uncertainty analyses, as feasible and appropriate. The committee will identify gaps in scientific knowledge that are critical to understanding dioxin reassessment (NAS, 2006, 198441, p. 43, Box 1-1). 8 9 In 2006, the NAS published its review of EPA's 2003 Reassessment entitled Health Risksfrom 10 Dioxin and Related Compounds: Evaluation o f the EPA Reassessment (NAS, 2006, 198441). 11 12 1.1. SUMMARY OF KEY NAS (2006, 198441) COMMENTS ON DOSE-RESPONSE 13 MODELING IN THE 2003 REASSESSMENT 14 While recognizing the effort that EPA expended to prepare the 2003 Reassessment, the 15 NAS committee identified three key areas that they believe require substantial improvement to 16 support a scientifically robust risk assessment. These three key areas are This document is a draftfor review purposes only and does not constitute Agency policy. l -2 DRAFT--DO NOT CITE OR QUOTE 1 transparency and clarity in selection of key data sets for analysis; 2 justification of approaches to dose-response modeling for cancer and noncancer 3 endpoints; and 4 transparency, thoroughness, and clarity in quantitative uncertainty analysis. 5 6 In their Public Summary, the NAS made the following overall recommendations to aid 7 EPA in addressing their key concerns: 8 9 EPA should compare cancer risks by using nonlinear models consistent with a receptor 10 mediated mechanism of action and by using epidemiological data and the new National 11 Toxicology Program (NTP) animal bioassay data (NTP, 2006, 197605). The comparison 12 should include upper and lower bounds, as well as central estimates of risk. EPA should 13 clearly communicate this information as part of its risk characterization (NAS, 2006, 14 198441, p. 9). 15 EPA should identify the most important data sets to be used for quantitative risk 16 assessment for each of the four key end points (cancer, immunotoxicity, reproductive 17 effects, and developmental effects). EPA should specify inclusion criteria for the studies 18 (animal and human) used for derivation of the benchmark dose (BMD) for different 19 noncancer effects and potentially for the development of RfD (reference dose) values and 20 discuss the strengths and limitations of those key studies; describe and define 21 (quantitatively to the extent possible) the variability and uncertainty for key assumptions 22 used for each key end-point-specific risk assessment (choices of data set, POD [point of 23 departure], model, and dose metric); incorporate probabilistic models to the extent 24 possible to represent the range of plausible values; and assess goodness-of-fit of 25 dose-response models for data sets and provide both upper and lower bounds on central 26 estimates for all statistical estimates. When quantitation is not possible, EPA should 27 clearly state it and explain what would be required to achieve quantitation (NAS, 2006, 28 198441, p. 9). 29 When selecting a BMD as a POD, EPA should provide justification for selecting a 30 response level (e.g., at the 10%, 5%, or 1% level). In either case, the effects of this 31 choice on the final risk assessment values should be illustrated by comparing point 32 estimates and lower bounds derived from selected PODs (NAS, 2006, 198441, p. 9). 33 EPA should continue to use body burden as the preferred dose metric but should also 34 consider physiologically based pharmacokinetic modeling as a means to adjust for 35 differences in body fat composition and for other differences between rodents and 36 humans (NAS, 2006, 198441, p. 9). 37 Although EPA addressed many sources of variability and uncertainty qualitatively, the 38 committee noted that the 2003 Reassessment would be substantially improved if its risk 39 characterization included more quantitative approaches. Failure to characterize This document is a draftfor review purposes only and does not constitute Agency policy. 1-3 DRAFT--DO NOT CITE OR QUOTE 1 variability and uncertainty thoroughly can convey a false sense of precision in the 2 conclusions of the risk assessment (NAS, 2006, 198441, p. 5). 3 4 Importantly, the NAS encouraged EPA to calculate an RfD as the 2003 Reassessment 5 does not contain an RfD derivation. The committee suggested that 6 7 .. .estimating an RfD would provide useful guidance to risk managers to help 8 them (1) assess potential health risks in that portion of the population with intakes 9 above the RfD, (2) assess risks to population subgroups, such as those with 10 occupational exposures, and (3) estimate the contributions to risk from the major 11 food sources and other environmental sources of TCDD, other dioxins, and DLCs 12 for those individuals with high intakes (NAS, 2006, 198441, p. 6). 13 14 The NAS made many thoughtful and specific recommendations throughout their review; 15 additional NAS recommendations and comments pertaining to the dose-response assessment of 16 TCDD will be presented and addressed in various sections throughout this document. 17 18 1.2. EPA'S SCIENCE PLAN 19 In May 2009, EPA Administrator Lisa P. Jackson announced the "Science Planfor 20 Activities Related to Dioxins in the Environment"("Science Plan") that addressed the need to 21 finish EPA's dioxin reassessment and provide a completed health assessment on this high profile 22 chemical to the American public as quickly as possible.9 The Science Plan states that EPA will 23 release a draft report that responds to the recommendations and comments included in the NAS 24 review of EPA's 2003 Reassessment, and that, in this draft report, EPA's National Center for 25 Environmental Assessment, Office of Research and Development, will provide a limited 26 response to key comments and recommendations in the NAS report (draft response to comments 27 report). This draft response is to focus on dose-response issues raised by the NAS and include 28 analyses of relevant new key studies. The draft response is to be provided for public review and 29 comment and for independent external peer review by EPA's Science Advisory Board. 30 Following completion of this report, EPA is to review the impacts of the response to comments 31 report on its 2003 Reassessment. Available at http://www.epa.gov/dioxin/scienceplan. This document is a draftfor review purposes only and does not constitute Agency policy. 1-4 DRAFT--DO NOT CITE OR QUOTE 1 This draft document comprises EPA's report that responds both directly and technically 2 to the recommendations and comments on TCDD dose-response assessment included in the NAS 3 review of EPA's 2003 Reassessment. This document focuses on TCDD only. Because new data 4 are analyzed in this report and toxicity values are derived, this document will follow the IRIS 5 process for review, clearance and completion; however, it is not a traditional IRIS document. 6 Information developed in this document is intended to not only respond to the NAS review, but 7 also to expand EPA's knowledge of TCDD cancer and noncancer dose-response based on the 8 most current literature, existing methods, and adherence to EPA risk assessment guidance 9 documents. Following completion of this document, EPA will consider its contents as it reviews 10 the TCDD risk assessment information presented in the 2003 Reassessment and moves forward 11 towards completion of the dioxin reassessment. 12 13 1.3. OVERVIEW OF EPA'S RESPONSE TO NAS (2006, 198441) "HEALTH RISKS 14 FROM DIOXIN AND RELATED COMPOUNDS: EVALUATION OF EPA's 2003 15 REASSESSMENT" 16 In their key recommendations, the NAS commented that EPA should thoroughly justify 17 and communicate approaches to dose-response modeling, increase transparency in the selection 18 of key data sets, and improve the communication of uncertainty (particularly quantitative 19 uncertainty). They also encouraged EPA to calculate an RfD. These main areas of improvement 20 refer to issues specifically related to TCDD dose-response assessment (and uncertainty analysis); 21 therefore, as noted in the Science Plan, EPA's response to the NAS is particularly focused on 22 these issues. 23 EPA thoroughly considered the recommendations of the NAS and responds with 24 scientific and technical evaluation of TCDD dose-response data via: 25 26 an updated literature search that identified new TCDD dose-response studies (see 27 Section 2); 28 a kickoff workshop that included the participation of external experts in TCDD health 29 effects, toxicokinetics, dose-response assessment and quantitative uncertainty analysis; 30 these experts discussed potential approaches to TCDD dose-response assessment and 31 considerations for EPA's response to NAS (U.S. EPA, 2009, 543757, Appendix A); 32 detailed study inclusion criteria and processes for the selection of key studies (see 33 Section 2.3) and epidemiologic and animal bioassay data for TCDD dose-response 34 assessment (see Section 2.4.1/Appendix B and Section 2.4.2, respectively); This document is a draftfor review purposes only and does not constitute Agency policy. 1-5 DRAFT--DO NOT CITE OR QUOTE 1 kinetic modeling to quantify appropriate dose metrics for use in TCDD dose-response 2 assessment (see Section 3 and Appendices C and D); 3 dose-response modeling for all appropriate noncancer and cancer data sets (see 4 Section 4.2/Appendix E and Section 5.2.3/Appendix F, respectively); 5 thorough and transparent evaluation of the selected TCDD data for use in the derivation 6 of an RfD and an oral slope factor (OSF) (see Sections 4.2 and 5.2.3, respectively); 7 the development of an RfD (see Section 4.3); 8 the development of a revised OSF (see Section 5.3) with an updated cancer weight of 9 evidence determination for TCDD based on EPA's 2005 Cancer Guidelines (2005, 10 086237) (see Section 5.1.2); 11 consideration of nonlinear dose-response approaches for cancer, including illustrative 12 RfDs for cancer precursor events and tumors (see Section 5.2.3.4); and 13 discussion of the feasibility and utility of quantitative uncertainty analysis for TCDD 14 dose-response assessment (see Section 6). 15 16 Each of these activities is described in detail in subsequent sections of this document. 17 In addition to this document, it should be noted that three separate EPA activities address 18 other TCDD issues, specifically related to the application of dioxin TEFs and to TCDD and DLC 19 background exposure levels. Information on the application of the dioxin TEFs is published 20 elsewhere by EPA for both ecological (U.S. EPA, 2008, 543774) and human health risk 21 assessment (U.S. EPA, 2009, 192196). As a consequence, EPA does not directly address TEFs 22 herein, but makes use of the concept of toxicity equivalence10as applicable to the analysis of 23 exposure dose in epidemiologic studies. Furthermore, this document does not address the NAS 24 recommendations pertaining to the assessment of human exposures to TCDD and other dioxins. 25 Information on updated background levels of dioxin in the U.S. population has been recently 26 reported (Lorber et al., 2009, 543766). 27 28 1.3.1. TCDD Literature Update 29 EPA has developed a literature database of peer-reviewed studies on TCDD toxicity, 30 including in vivo mammalian dose-response studies and epidemiologic studies. An initial 31 literature search for studies published since the 2003 Reassessment was conducted by the U.S. 32 Department of Energy's Argonne National Laboratory (ANL) through an Interagency Agreement 10Toxicity equivalence (TEQ) is the product of the concentration of an individual DLC in an environmental mixture and the corresponding TCDD TEF for that compound. These products are summed to yield the TEQ of the mixture. This document is a draftfor review purposes only and does not constitute Agency policy. 1-6 DRAFT--DO NOT CITE OR QUOTE 1 with EPA. ANL used the online National Library of Medicine database (PubMed) and identified 2 studies published between the year 2000 and October 31, 2008. Supporting references published 3 since the release of the 2003 Reassessment were also identified. Supporting studies were 4 classified as studies pertaining to TCDD kinetics, TCDD mode-of-action, in vitro TCDD studies, 5 and TCDD risk assessment approaches. The literature search strategy explicitly excluded studies 6 addressing (1) analytical/detection data and cellular screening assays; (2) environmental fate, 7 transport and concentration data; (3) dioxin-like compounds and toxic equivalents; 8 (4) nonmammalian dose-response data; (5) human exposure analyses only, including body 9 burden data; and (6) combustor or incinerator or other facility-related assessments absent 10 primary dose-response data. EPA published the initial literature search results in the Federal 11 Register on November 24, 2008 (73 FR 70999; November 24, 2008) and invited the public to 12 review the list and submit additional peer-reviewed in vivo mammalian dose-response studies for 13 TCDD, including epidemiologic studies that were absent from the list (U.S. EPA, 2008, 519261). 14 Submissions were accepted by the EPA through an electronic docket, email and hand delivery, 15 and were evaluated for use in TCDD dose-response assessment. The literature search results and 16 subsequent submissions were used during a 2009 scientific workshop, which was open to the 17 public and featured a panel of experts on TCDD toxicity and dose-response modeling (discussed 18 below). Additional studies identified during the workshop and those collected by EPA scientists 19 during the development of this report through October 2009 have been incorporated into the final 20 set of studies for TCDD dose-response assessment. 21 22 1.3.2. EPA's 2009 Workshop on TCDD Dose Response 23 To assist EPA in responding to the NAS, EPA and ANL convened a scientific workshop 24 (the "Dioxin Workshop") on February 18-20, 2009, in Cincinnati, Ohio. The goals of the 25 Dioxin Workshop were to identify and address issues related to the dose-response assessment of 26 TCDD and to ensure that EPA's response to the NAS focused on the key issues and reflected the 27 most meaningful science. The Dioxin Workshop included seven scientific sessions: quantitative 28 dose-response modeling issues, immunotoxicity, neurotoxicity and nonreproductive endocrine 29 effects, cardiovascular toxicity and hepatotoxicity, cancer, reproductive and developmental 30 toxicity, and quantitative uncertainty analysis of dose-response. During each session, EPA asked 31 a panel of expert scientists to perform the following tasks: This document is a draftfor review purposes only and does not constitute Agency policy. 1-7 DRAFT--DO NOT CITE OR QUOTE 1 2 Identify and discuss the technical challenges involved in addressing the NAS comments 3 related to the dose-response issues within each specific session topic and the TCDD 4 quantitative dose-response assessment. 5 Discuss approaches for addressing the key NAS recommendations. 6 Identify important published, independently peer-reviewed literature--particularly 7 studies describing epidemiologic studies and in vivo mammalian bioassays expected to 8 be most useful for informing EPA's response. 9 10 The sessions were followed by open comment periods during which members of the 11 audience were invited to address the expert panels. The session's Panel Co-chairs were asked to 12 summarize and present the results of the panel discussions--including the open comment 13 periods. The summaries incorporated points of agreement as well as minority opinions. Final 14 session summaries were prepared by the session Panel Co-chairs with input from the panelists, 15 and they formed the basis of a final workshop report (U.S. EPA, 2009, 543757, Appendix A of 16 this report). Because the sessions were not designed to achieve consensus among the panelists, 17 the summaries do not necessarily represent consensus opinions; rather reflect the core of the 18 panel discussions. Some of the key discussion points from the workshop that influenced EPA's 19 development of this document are listed below (see Appendix A for detail): 20 21 In the development of study selection criteria, more relevant exposure-level (i.e., dose) 22 decision points using tissue concentrations could be defined. 23 A linear approach to body-burden estimation, which was utilized in the 2003 24 Reassessment (U.S. EPA, 2003, 537122), does not fully consider key toxicokinetic issues 25 related to TCDD--e.g., sequestration in the liver and fat, age-dependent elimination, and 26 changing elimination rates over time. Thus, kinetic/mechanistic modeling could be used 27 to quantify tissue-based metrics. In considering human data, lipid-adjusted serum levels 28 may be preferable over body burden, although the assumptions used in the back 29 calculation of the body burden in epidemiologic cohorts are of concern. In considering 30 rat bioassay data, lipid-adjusted body-burden estimates may be preferable. 31 New epidemiologic studies on noncancer endpoints have been published since the 32 2003 Reassessment that may need to be considered (e.g., thyroid dysfunction literature 33 from Wang et al. (2005, 198734) and Baccarelli et al. (2008, 197059)). 34 The 1% of maximal response (ED01) that was utilized in the 2003 Reassessment has not 35 typically been used in dose-response assessment. Some alternative ideas were as follows: 36 (1) the POD should depend on the specific endpoint; (2) for continuous measures, the 37 benchmark response (BMR) could be based on the difference from control and consider This document is a draftfor review purposes only and does not constitute Agency policy. 1-8 DRAFT--DO NOT CITE OR QUOTE 1 the adversity level; and (3) for incidence data, the BMR should be set to a fixed-risk 2 level. 3 The quantitative dose-response modeling for cancer could be based on human or animal 4 data. There are new publications in the literature for four epidemiological cohort studies 5 (Dutch cohort, NIOSH cohort, BASF accident cohort, and Hamburg cohort). The 6 increase in total cancers could be considered for modeling human cancer data. However, 7 non-Hodgkin's lymphoma and lung tumors are the main TCDD-related cancer types seen 8 from human exposure. In reviewing the rat data, the NTP (2006, 197605) data sets are 9 new and can be modeled. Although the liver and lungs are the main target organs, 10 modeling all cancers, as well as using tumor incidence in lieu of individual rats as a 11 measure, should be considered. 12 Both linear and nonlinear model functions should be considered in the cancer 13 dose-response analysis because there are data and rationales to support use of either 14 below the POD. 15 For quantitative uncertainty analysis, consider the impacts of choices among plausible 16 alternative data sets, dose metrics, models, and other more qualitative choices. Issues to 17 consider include how much difference these choices make and, also, how much relative 18 credence should be put toward each alternative as a means to gauge and describe the 19 landscape of imperfect knowledge with respect to possibilities for the true dose response. 20 This may be difficult to do quantitatively because the factors are not readily expressed as 21 statistical distributions. However, the rationale for accepting or questioning each 22 alternative in terms of the available supporting evidence, contrary evidence, and needed 23 assumptions, can be delineated. 24 25 1.3.3. Overall Organization of EPA's Response to NAS Recommendations 26 The remainder of this document is divided into five sections that address the 27 three primary areas of concern resulting from the NAS (2006, 198441) review. Section 2 28 describes EPA's approach to the recommendation for transparency and clarity during selection of 29 key data sets--including criteria for the selection of key dose-response studies, evaluations of the 30 important epidemiologic studies and animal bioassays, and a summary of the key studies used 31 for subsequent dose-response modeling. Sections 3, 4, and 5 present EPA's response to the NAS 32 recommendation to better justify the approaches used in dose-response modeling of TCDD. 33 Section 3 discusses the toxicokinetic modeling EPA conducted to support the dose-response 34 analyses. Section 4 presents EPA's approach to noncancer data set selection, dose-response 35 modeling, and derivation of an RfD for TCDD, and contains a qualitative discussion of the 36 uncertainties associated with the RfD. Section 5 presents an updated cancer weight-of-evidence 37 summary, EPA's approach to cancer data set selection, dose-response modeling, derivation of an This document is a draftfor review purposes only and does not constitute Agency policy. 1-9 DRAFT--DO NOT CITE OR QUOTE 1 OSF for TCDD, and a qualitative discussion of the uncertainties associated with the OSF, 2 including an evaluation of illustrative nonlinear approaches to cancer assessment of TCDD. 3 Finally, Section 6 discusses the feasibility of conducting a quantitative uncertainty analysis of 4 TCDD dose response. This document is a draftfor review purposes only and does not constitute Agency policy. 1-10 DRAFT--DO NOT CITE OR QUOTE 1 2. TRANSPARENCY AND CLARITY IN THE SELECTION OF KEY DATA SETS 2 FOR DOSE-RESPONSE ANALYSIS 3 4 5 This section addresses transparency and clarity in the study selection process and 6 identifies key data sets for 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) dose-response analysis. 7 Section 2.1 summarizes the National Academy of Sciences (NAS) committee's comments 8 specifically regarding this issue. Section 2.2 presents U.S. Environmental Protection Agency's 9 (EPA's) response to those comments and describes EPA's approach to ensuring transparency and 10 clarity in the selection of studies for subsequent dose-response analyses. Section 2.3 describes 11 the TCDD-specific study inclusion criteria and evaluation process EPA used in this document for 12 determining the eligibility of both epidemiologic and experimental animal studies for TCDD 13 dose-response analysis. Section 2.4.1 summarizes epidemiologic data and evaluates the 14 suitability of these data for TCDD dose-response analyses. Section 2.4.2 summarizes animal 15 bioassay data that have met the study inclusion criteria for TCDD dose-response assessment. 16 Finally, Section 2.4.3 identifies key TCDD epidemiologic and animal bioassay studies that were 17 determined using the study inclusion criteria. Study/endpoint combination data sets for 18 developing TCDD toxicity values for noncancer and cancer effects are further evaluated in 19 Sections 4 and 5 of this document, respectively. 20 21 2.1. SUMMARY OF NAS COMMENTS ON TRANSPARENCY AND CLARITY IN 22 THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS 23 The NAS committee proposed that EPA develop a clear and readily understandable 24 methodology for evaluating and including epidemiologic and animal bioassay data sets in 25 dose-response evaluations. The NAS committee recommended the development and application 26 of transparent initial criteria to judge whether or not specific epidemiologic or animal bioassay 27 studies be included in TCDD dose-response analysis. 28 Specific NAS comments on the topic of study evaluation and inclusion criteria include 29 30 EPA should specify inclusion criteria for the studies (animal and human) used for 31 derivation of the benchmark dose (BMD) for different noncancer effects and 32 potentially for the development of RfD values and discuss the strengths and 33 limitations of those key studies (NAS, 2006, 198441, p. 27). This document is a draftfor review purposes only and does not constitute Agency policy. 2-1 DRAFT--DO NOT CITE OR QUOTE 1 .. .in its [EPA's] evaluation of the epidemiological literature of carcinogenicity, it 2 did not outline eligibility requirements or otherwise provide the criteria used to 3 assess the methodological quality of other included studies (NAS, 2006, 198441, 4 p. 56). 5 With regard to EPA's review of the animal bioassay data, the committee 6 recommends that EPA establish clear criteria for the inclusion of different data 7 sets (NAS, 2006, 198441, p. 191). 8 . t h e committee expects that EPA could substantially improve its assessment 9 process if it more rigorously evaluated the quality of each study in the database 10 (NAS, 2006, 198441, p. 56). 11 EPA could also substantially improve the clarity and presentation of the risk 12 assessment process for T C D D .b y using a summary table or a simple summary 13 graphical representation of the key data sets and assum ptions. (NAS, 2006, 14 198441, p. 56). 15 16 2.2. EPA'S RESPONSE TO NAS COMMENTS ON TRANSPARENCY AND CLARITY 17 IN THE SELECTION OF KEY DATA SETS FOR DOSE-RESPONSE ANALYSIS 18 EPA agrees with the NAS committee regarding the need for a transparent and clear 19 process for selecting studies and key data sets for TCDD dose-response analyses. The 20 delineation of the study selection process and decisions regarding key data sets will facilitate 21 communication regarding critical decisions made in the TCDD dose-response assessment. In 22 keeping with the NAS committee's recommendation to use a transparent process and improve 23 clarity and presentation of the risk assessment process for TCDD, Figure 2-1 overviews the 24 approach that EPA has used in this document to develop a final list of key cancer and noncancer 25 studies for quantitative dose-response analysis of TCDD. The steps in Figure 2-1 are further 26 explained below. 27 28 Literature search for in vivo mammalian and epidemiologic TCDD studies 29 (2000-2008): EPA conducted a literature search to identify peer-reviewed, dose-response 30 studies for TCDD that have been published since the 2003 Reassessment. This search 31 included in vivo mammalian and epidemiological studies of TCDD from 2000 to 2008. 32 Additional details describing the conduct of this literature search are presented in 33 Section 1.3.1 of this document. 34 Federal Register Notice--Web publication of literature search for public comment: 35 In November 2008, EPA published a list of ~500 citations from results of this literature 36 search (U.S. EPA, 2008, 519261) and invited the public to review this preliminary list of 37 dose-response citations for use in TCDD dose-response assessment. EPA requested that 38 interested parties identify and submit peer-reviewed studies for TCDD that were absent This document is a draftfor review purposes only and does not constitute Agency policy. 2-2 DRAFT--DO NOT CITE OR QUOTE 1 from this list. Two parties identified additional references that were not included in the 2 2008 Federal Register notice and submitted additional references for EPA to consider. 3 These references were included in the final TCDD literature database considered by EPA 4 for TCDD dose-response analysis. 5 Initial study inclusion criteria development for TCDD in vivo mammalian 6 bioassays: EPA developed an initial set of draft criteria for evaluating the extensive 7 TCDD database of in vivo mammalian bioassays. These initial inclusion criteria had 8 three purposes. First, they provided a transparent and rigorous evaluation of the scientific 9 quality of each study in EPA's database, a deficiency in the 2003 Reassessment identified 10 by the NAS committee. Second, given the vast TCDD mammalian bioassay database, 11 they provided a transparent method for initially screening studies to be considered for 12 TCDD dose-response analyses. Third, they served as a starting point for discussions of 13 study inclusion criteria by expert panelists who were convened by EPA for its scientific 14 workshop on TCDD dose-response analysis (the Dioxin Workshop), described next (also 15 see the workshop report in Appendix A, U.S. EPA [2009b]). 16 Dioxin Workshop and expert refinement of TCDD in vivo mammalian bioassay 17 inclusion criteria: In February 2009, EPA convened "A Scientific Workshop to Inform 18 EPA's Response to NAS Comments on the Health Effects of Dioxin in EPA's 2003 19 Dioxin Reassessment." The goals of this 3-day public and scientific workshop were to 20 identify and address issues related to the dose-response assessment of TCDD. Sessions at 21 the workshop examined toxicities associated with TCDD, issues related to developing 22 dose-response estimates based on these data and associated uncertainties. At the 23 workshop, EPA presented the draft set of study inclusion criteria for evaluating the 24 extensive TCDD in vivo mammalian bioassay literature and asked workshop panelists to 25 discuss these criteria and make recommendations for their revision. Further details on 26 this workshop are presented in Section 1.3.2 of this document, and the complete report 27 from this workshop is available in Appendix A (U.S. EPA, 2009b), including detailed 28 summaries of the panels' comments on the inclusion criteria in relation to the various 29 toxic endpoints that were discussed. 30 Final development of inclusion criteria for TCDD in vivo mammalian studies: Based 31 on discussions at the Dioxin Workshop, the initial draft inclusion criteria for evaluating 32 the TCDD mammalian bioassay literature were revised and are presented in Section 2.3.2 33 (see Figure 2-3). An initial criterion is that studies for consideration must be publically 34 available and published in a peer-reviewed scientific journal. Because the methodology 35 EPA uses to develop reference doses (RfDs) and cancer oral slope factors (OSFs) relies 36 on identification of studies reporting potential adverse effects at low doses (relative to the 37 overall database), another important criterion shown in Section 2.3.2 identifies a 38 maximum value for the lowest TCDD dose tested in a bioassay. This maximum value 39 was used to eliminate those studies that could not be selected for development of an RfD 40 or an oral slope factor because tested doses were too high relative to other TCDD 41 bioassays. 42 Development of inclusion criteria for epidemiologic studies: Following the Dioxin 43 Workshop, EPA determined that an evaluation process was also needed for inclusion of 44 epidemiologic studies for TCDD dose-response assessment. These criteria were This document is a draftfor review purposes only and does not constitute Agency policy. 2-3 DRAFT--DO NOT CITE OR QUOTE 1 developed and are detailed in Section 2.3.1 (see Figure 2-2). Analogous to animal 2 bioassay data, epidemiologic studies for consideration must also be publically available 3 and published in a peer-reviewed scientific journal. In addition to assessing the 4 methodological considerations relative to epidemiologic cohorts and studies (e.g., 5 statistical power and precision of estimates, consideration of latency periods), key criteria 6 for use of a study in TCDD dose-response modeling were that the exposure be primarily 7 to TCDD and that the effective dose and oral exposure are reasonably estimable. 8 Final literature collection (October 2009): Additional literature was collected as it was 9 identified by EPA following the Dioxin Workshop through October 2009 to ensure the 10 consideration of all recently published data for this report. 11 Studies screened using inclusion criteria: The two sets of TCDD-specific study 12 inclusion criteria presented in Section 2.3 were used to evaluate all studies included in the 13 2003 Reassessment, studies identified in the 2000-2008 literature search, studies 14 identified through public comment and submission, and studies collected in 2009 as 15 identified by EPA during the development of this document. Section 2.4 presents results 16 of EPA's evaluation of epidemiologic and mammalian bioassay literature for both cancer 17 and noncancer endpoints. 18 Final list of key cancer and noncancer studies for quantitative dose-response 19 analysis of TCDD: Application of the study inclusion criteria concludes in Section 2.4 20 with development of a list of key noncancer and cancer studies that were considered for 21 quantitative dose-response analyses of TCDD in Sections 4 and 5, respectively. In those 22 sections, points of departure (PODs) are developed and evaluated for all biologically 23 relevant study/endpoint combinations from these final key study lists, and key data sets 24 and PODs for the development of TCDD toxicity values are identified. 25 26 2.3. STUDY INCLUSION CRITERIA FOR TCDD DOSE-RESPONSE ANALYSIS 27 One of the three major recommendations made by the NAS (2006, 198441) committee 28 was that EPA should provide greater clarity and transparency on the selection of studies that 29 were used in the quantitative dose-response modeling of TCDD in the 2003 Reassessment. In 30 this section, EPA describes TCDD-specific study inclusion criteria that have been developed to 31 evaluate epidemiologic studies and animal bioassays for TCDD dose-response assessment. 32 These criteria reflect EPA's goal of developing an RfD and a cancer OSF for TCDD through a 33 transparent study selection process; they are intended to be used by EPA for TCDD 34 dose-response assessment only. These criteria were applied to each of the ~500 studies listed in 35 Preliminary Literature Search Results and Requestfor Additional Studies on 36 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) Dose-Response Studies (U.S. EPA, 2008, 37 519261); studies identified and submitted by the public and by participants in the Dioxin This document is a draftfor review purposes only and does not constitute Agency policy. 2-4 DRAFT--DO NOT CITE OR QUOTE 1 Workshop (U.S. EPA, 2009, 522927); studies included in the 2003 Reassessment, and other 2 relevant published studies collected by EPA scientists through October 2009. 3 EPA has undertaken different approaches for epidemiologic versus in vivo animal 4 bioassay study evaluation and key data set selection. The significant differences between animal 5 and human health effects data and their use in EPA risk assessment support development of 6 separate criteria for study inclusion and different approaches to study evaluation. For the vast 7 majority of compounds on EPA's Integrated Risk Information System (IRIS), cancer and 8 noncancer toxicity values have been derived using animal bioassay data; therefore, approaches to 9 dose-response modeling and POD selection from in vivo mammalian bioassays have been 10 standardized and codified (U.S. EPA, 2000, 052150). The study criteria shown below and in 11 Figure 2-3 for animal bioassay data reflect EPA's preferences for TCDD-specific study 12 inclusion, some of which are based on common practices and guidance for POD selection and 13 RfD and OSF derivation. Far fewer IRIS toxicity values have been derived from human data, 14 although some examples do exist. For example, benzene, beryllium and compounds, chromium 15 IV, and 1,3-butadiene have RfDs, Reference Concentrations, Inhalation Unit Risks and/or OSFs 16 based on occupational cohort data and the methyl mercury RfD is based on high fish consuming 17 cohorts (U.S. EPA, 2009, 543757). The modeling and interpretation of such human data have 18 been conducted on a case-by-case basis because each cohort is uniquely defined and has its own 19 set of exposure conditions, significant confounders, and biases that may need to be considered in 20 dose-response modeling. For TCDD, not all data are from occupational cohorts, but include 21 cohorts exposed for relatively short time periods to high concentrations as a consequence of 22 industrial accidents, a scenario that has not commonly been used to establish EPA toxicity 23 values. 24 Because of these differences in data characteristics, divergent selection approaches are 25 used in this document to present and evaluate the epidemiologic studies (see Section 2.3.1) and 26 the in vivo animal bioassays (see Section 2.3.2). In Section 2.4.1, all of the available 27 epidemiologic studies on TCDD are summarized and evaluated for suitability for dose-response 28 modeling using the TCDD-specific study inclusion criteria below and shown in Figure 2-2; only 29 studies meeting the inclusion criteria are presented as key studies in Section 2.4.3 (see Tables 2-4 30 and 2-5 for the cancer and noncancer endpoints, respectively). In Section 2.4.2, because 31 summarizing and showing the evaluation of the thousands of available animal bioassays on This document is a draftfor review purposes only and does not constitute Agency policy. 2-5 DRAFT--DO NOT CITE OR QUOTE 1 TCDD was prohibitive, only studies first meeting the in vivo animal bioassays study inclusion 2 criteria below (and shown in Figure 2-3) are summarized. These studies are also presented as 3 key studies in Section 2.4.3 (see Tables 2-6 and 2-7 for cancer and noncancer endpoints, 4 respectively). 5 6 2.3.1. Study Inclusion Criteria for TCDD Epidemiologic Studies 7 This section identifies the process EPA used to select epidemiologic studies for defining 8 candidate PODs for TCDD dose-response modeling. These criteria are based on EPA's 9 approaches for deriving OSFs and RfDs. A discussion of the considerations used in selecting 10 epidemiologic data for quantitative dose-response modeling is valuable, particularly given EPA's 11 preference to use high-quality human studies over animal studies because such human studies are 12 regarded as providing the most relevant information needed for quantitative human health risk 13 analyses (U.S. EPA, 2005, 086237). As described by Hertz-Picciotto (1995, 065678), key 14 components needed for the use of an epidemiologic study as a basis for quantitative risk 15 assessment include issues regarding exposure assessment (a well-quantified exposure assessment 16 with exposures linked to individuals) and study quality ("strong biases," for example with 17 respect to inclusion criteria for membership in the cohort and follow-up procedures "ruled out or 18 unlikely" and "confounding controlled or likely to be limited"). The strength of the association, 19 either within the full study or within a high exposure subgroup, can also be considered in the 20 evaluation of suitability for dose-response modeling (Hertz-Picciotto, 1995, 065678). Stayner 21 et al. (1999, 198654), however, note that even weak associations could be useful in terms of 22 providing an estimate of a potential upper bound for a quantitative risk estimate. 23 EPA's method for applying the TCDD study inclusion criteria to epidemiologic data is 24 detailed below and in Figure 2-2. Based on the framework discussed above, EPA evaluated the 25 available epidemiologic cohorts and studies based on the five following considerations: 26 27 1. The methods used to ascertain health outcomes are clearly identified and unbiased, with 28 high sensitivity and specificity. 29 2. The risk estimates generated from the study are not susceptible to important biases 30 arising from an inability to control for potential confounding exposures or other sources 31 of bias arising from either study design or statistical analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-6 DRAFT--DO NOT CITE OR QUOTE 1 3. The study demonstrates an association between TCDD and an adverse health effect 2 (assuming minimal misclassification of exposure and absence of important biases) with 3 some suggestion of an exposure-response relationship. 4 4. The exposure assessment methodology is clearly described and can be expected to 5 provide adequate characterization of exposure, with assignment of individual-level 6 exposures within a study (e.g., based on biomarker data, or based on a 7 job-exposure-matrix approach). Limitations and uncertainties in the exposure assessment 8 are considered. 9 5. The size and follow-up period of a cohort study are large enough and long enough, 10 respectively, to yield sufficiently precise estimates for use in development of quantitative 11 risk estimates and to ensure adequate statistical power to limit the possibility of not 12 detecting an association that might be present (i.e., to avoid Type II Errors due to failing 13 to reject the null hypothesis when the null hypothesis is true). Similar considerations 14 regarding sample size and statistical precision and power apply to case-control studies. 15 16 Three specific study inclusion criteria were used to select studies for further evaluation 17 and potential TCDD quantitative dose-response assessment 18 19 1. The study is published in the peer-reviewed scientific literature and includes an 20 appropriate discussion of strengths and limitations. 21 2. The exposure is primarily to TCDD, rather than dioxin-like compounds (DLCs), and is 22 properly quantified so that dose-response relationships can be assessed. All 23 epidemiologic cohorts will have background exposures to DLCs through the food chain 24 and these exposures are not included in this criterion. 25 3. The effective dose and oral exposure must be reasonably estimable. The measures of 26 exposure must be consistent with the current biological understanding of dose. For 27 TCDD dose-response assessment, it is critical that reported dose is consistent with a dose 28 that is likely to be toxicologically relevant. The timing of the measurement of effects 29 (i.e., the response) also must be consistent with current biological understanding of the 30 effect and its progression. 31 For cancer endpoints, EPA assumes that cumulative TCDD dose estimates are 32 toxicologically relevant measures. Thus, cancer studies must provide information 33 about long-term TCDD exposure levels. Further, EPA reasons that measures of 34 cancer occurrence or death need to allow for examination of issues of latency 35 between the end of effective exposure and cancer detection or death. 36 For noncancer endpoints, exposure estimates and analysis must allow for examination 37 of issues of latency and other issues regarding the appropriate time window of 38 exposure relevant for specific endpoints. Also, to be consistent with the RfD 39 methodology, the response must be to a nonfatal endpoint. 40 This document is a draftfor review purposes only and does not constitute Agency policy. 2-7 DRAFT--DO NOT CITE OR QUOTE 1 Those studies that met these three inclusion criteria (see Sections 2.4.1, 2.4.3, and Appendix B) 2 were then subjected to further consideration for quantitative dose-response analyses. 3 4 2.3.2. Study Inclusion Criteria for TCDD In Vivo Mammalian Bioassays 5 This section identifies the criteria EPA applied to select nonhuman in vivo mammalian 6 studies for defining candidate PODs for use in TCDD dose-response modeling. These inclusion 7 criteria are based on EPA's approaches for deriving OSFs and RfDs from bioassay data 8 (U.S. EPA, 2005, 086237). EPA agrees with the NAS committee regarding the utility of an oral 9 RfD and the need for reevaluation of the OSF for TCDD, specifically in light of data that have 10 been published since the 2003 Reassessment was released. RfDs and OSFs are generally derived 11 using data sets that demonstrate the occurrence of adverse effects, or their precursors, in 12 low-dose range for that chemical. RfDs and OSFs are derived from a health protective 13 perspective for chronic exposures. Thus, when a group of studies is available on a chemical for 14 which a number of effects are observed at various doses across those studies, the studies using 15 the lowest exposures that show effects will typically drive the RfD and OSF derivations, all other 16 considerations being equal. Studies conducted at higher exposures relative to other available 17 studies are used as supporting evidence for the final RfD or OSF since they were conducted at 18 doses too high to impact the numeric derivations of toxicity values. EPA expresses RfDs and 19 OSFs in terms of average daily doses, usually as mg/kg-day and per mg/kg-day, respectively. 20 Thus, the study inclusion criteria for the animal bioassay data presented in this section include 21 requirements that average daily exposures in the studies are within a low dose range where, 22 relative to other studies, they could be considered for development of a toxicity value. These 23 low-dose requirements do not imply that TCDD studies conducted at higher doses are of poor 24 quality, simply that they are not quantitatively useful in the development of toxicity values 25 because other studies with lower exposures will drive the RfD and OSF derivations under current 26 EPA practice. Because EPA has identified ~2,000 studies on TCDD that may be considered for 27 this purpose, the development and application of these study inclusion criteria has been critical to 28 moving the risk assessment process forward. 29 EPA's method for applying study inclusion criteria for mammalian bioassays is detailed 30 below and in Figure 2-3. The first study inclusion criterion is that the study is published in the 31 peer-reviewed scientific literature. Then, two specific study inclusion criteria were used to select This document is a draftfor review purposes only and does not constitute Agency policy. 2-8 DRAFT--DO NOT CITE OR QUOTE 1 studies for further evaluation and potential TCDD quantitative dose-response analyses and 2 identification of candidate PODs: 3 4 1. The lowest dose level tested is <1 pg/kg-day for cancer studies and <30 ng/kg-day for 5 noncancer studies. 6 2. The study design consists of orally administered TCDD-only doses, and specifies the 7 purity and matrix used to administer the doses. 8 9 Then, EPA evaluated the remaining in vivo animal studies based on the following 10 four considerations. 11 12 1. The study tests mammalian species, identifying the strain, gender, and age of the tested 13 animals. 14 2. The study clearly documents testing protocol, including dosing frequency, duration, and 15 timing of dose administration relative to age of the animals. 16 3. The overall study design is consistent with standard toxicological principles and 17 practices. The control group or groups are appropriate, given the testing protocol, and are 18 well characterized. Clinical and pathological examinations conducted during the study 19 are endpoint-appropriate, particularly for negative findings. 20 4. The magnitude of animal responses is outside the range of normal variability exhibited by 21 control animals (e.g., greater than or less than one standard deviation). 22 23 Those studies that met the aforementioned considerations and inclusion criteria (see 24 Sections 2.4.2 and 2.4.3) were then subjected to dose-response analysis. 25 The criteria for dose requirements, although somewhat arbitrary, are intended to be 26 reasonable cutoffs that restrict the number of studies that would need to be modeled while 27 ensuring that all study/data set combinations that could be candidates for the cancer slope factor 28 or RfD were modeled. Thus, the dose range under consideration allows for liberal ranges of 29 no-observed-adverse-effect levels (NOAELs), lowest-observed-adverse-effect levels (LOAELs), 30 and benchmark dose lower confidence bound (BMDLs) for assessment of both cancer and 31 noncancer effects. 32 For cancer studies, the dose requirements were selected based on an initial evaluation of 33 available average daily doses administered in TCDD animal bioassays in which adverse effects 34 were observed. For example, in cancer studies, a sample of the relatively low ranges of tested This document is a draftfor review purposes only and does not constitute Agency policy. 2-9 DRAFT--DO NOT CITE OR QUOTE 1 average daily doses include 1-1,000 ng/kg-day (Toth et al., 1979), 1-100 ng/kg-day (Kociba 2 et al., 1978), 1.43-286 ng/kg-day (NTP, 1982, 543764) and 2.14-71.4 ng/kg-day (NTP, 2006, 3 197605) with statistically significant increases in tumor incidence via pair-wise or trend tests 4 found in all of these studies. The entire range of each these studies is <1 ^g/kg-day. The 5 linearized multistage model used by EPA to estimate OSFs is most appropriately applied to 6 studies from which PODs can be estimated as closely as possible to the experimental data. Thus, 7 given the dose ranges in these studies that are available for modeling, the restriction to 8 <1 p,g/kg-day for cancer was considered to be a reasonable cutoff. 9 For noncancer studies, dose ranges are more complex and vary according to study 10 endpoint. Examples of the lowest administered doses that might be considered as NOAELs or 11 LOAELs in POD determinations for noncancer endpoints include 1 ng/kg-day (Toth et al., 1979, 12 197109), 1.43 ng/kg-day (Cantoni et al., 1981, 197092), 1.07 ng/kg-day (Smialowicz et al., 2008, 13 198341) 1.43 ng/kg-day (NTP, 1982, 543764) and 2.14 ng/kg-day (NTP, 2006, 197605). Most 14 of the lowest tested doses in the TCDD studies have been designated as LOAELs (see 15 Section 4.1). Given the available database, it is likely that the same composite uncertainty factor 16 (e.g., of 300; 3 for UFA[interspecies], 10 for UFH[intraspecies], and 10 for UFL [LOAEL to 17 NOAEL]) would be applied to any animal noncancer LOAEL used to derive an RfD for TCDD. 18 This implies that any study that has a LOAEL of 30 ng/kg-day or more would result in a 19 candidate RfD that is more than an order of magnitude higher than the example doses of 20 1-2 ng/kg-day shown here. BMDLs that might be derived from such data also would not be 21 expected to be lower than these example doses of 1-2 ng/kg-day. Thus, a tested dose 22 <30 ng/kg-day is considered to be a reasonable cutoff where the lowest tested dose would never 23 be used as a POD to derive an RfD given that much lower tested doses (associated with adverse 24 effects) are available from other studies of acceptable quality. 25 26 2.4. EVALUATION OF KEY STUDIES FOR TCDD DOSE RESPONSE 27 2.4.1. Evaluation of Epidemiological Cohorts for Dose-Response Assessment 28 This section summarizes and evaluates studies for potential use in TCDD dose-response 29 assessment using the study evaluation considerations and inclusion criteria for epidemiologic 30 data (see Section 2.3.1). Those studies that meet the study inclusion criteria are are listed later in 31 this Section in Tables 2-4 and 2-5, for cancer and noncancer, respectively, and are considered in This document is a draftfor review purposes only and does not constitute Agency policy. 2-10 DRAFT--DO NOT CITE OR QUOTE 1 the dose-response modeling conducted later in this document (see Sections 4 and 5). The 2 following sections are organized by epidemiologic cohort. Following a brief summary of each 3 cohort, its associated studies are then summarized chronologically, assessed for methodological 4 considerations relative to epidemiologic cohorts and studies (e.g., statistical power and precision 5 of estimates, consideration of latency periods) and evaluated for suitability for TCDD dose6 response assessment. 7 8 2.4.1.1. C ancer 9 In the 2003 Reassessment, EPA selected three cohort studies from which to conduct a 10 quantitative dose-response analysis: the National Institute for Occupational Safety and Health 11 (NIOSH) cohort (Steenland et al., 2001, 197433), the BASF cohort (Ott and Zober, 1996, 12 198408), and the Hamburg cohort (Becher et al., 1998, 197173). Although these studies were 13 deemed suitable for quantitative dose-response analysis, the criteria EPA used to reach this 14 conclusion were unclear. In this section, the study selection criteria and methodological 15 considerations presented in Section 2.3 are systematically applied to evaluate a number of studies 16 to determine their suitability for inclusion in dose-response modeling. In addition to the 17 three cohorts used in previous TCDD quantitative risk assessment, considerations are applied to 18 other relevant TCDD epidemiological data sets that were identified through a literature review 19 for epidemiological studies of TCDD and cancer. Study summaries and suitability for 20 quantitative dose-response analysis evaluations are discussed below. 21 22 2.4.1.1.1. C an cer cohorts. 23 2.4.1.1.1.1. T he N I O S H cohort. 24 In 1978, the NIOSH undertook research that identified workers employed by U.S. 25 chemical companies that made products contaminated with TCDD between 1942 and 1982. 26 TCDD was generated in the production of 2,4,5-trichlorophenol and subsequent processes. This 27 chemical was used to make 2,4,5-trichlorophenoxyacetic acid (2,4,5-T), which was a major 28 component of the widely-used defoliant, Agent Orange. The NIOSH cohort is the largest cohort 29 of occupational workers studied to date and has been the subject of a series of investigations 30 spanning more than two decades. It is important to note that this cohort consists mostly of male 31 workers that were exposed to TCDD via daily occupational exposure, as compared to an acute This document is a draftfor review purposes only and does not constitute Agency policy. 2-11 DRAFT--DO NOT CITE OR QUOTE 1 accidental exposure scenario seen with other cohorts. The investigations have progressed from a 2 comparison of the mortality patterns of the cohort to the U.S. general population to 3 dose-response modeling using serum-derived estimates of TCDD that have been 4 back-extrapolated several decades. Analyses of cancer data from the NIOSH cohort that are 5 addressed in this section include Fingerhut et al. (1991, 197375), Steenland et al. (1999, 197437; 6 2001, 197433), Cheng et al. (2006, 523122), and Collins et al. (2009, 197627). 7 8 2.4.1.1.1.1.1. Fingerhut et al. (1991, 197375). 9 2.4.1.1.1.1.1.1. Study summary. 10 The investigation of Fingerhut and her colleagues published nearly two decades ago 11 attracted widespread attention (Fingerhut et al., 1991, 197375). This retrospective study 12 examined patterns of cancer mortality for 5,172 workers who comprised the NIOSH cohort, 13 which combined workers from the company-specific cohorts of Dow Chemical (Ott et al., 1987, 14 064994)(Cook, 1981) and the Monsanto Company (Zack and Gaffey, 1983, 548783; Zack and 15 Suskind, 1980, 065005). These workers were employed at 12 plants producing chemicals 16 contaminated with TCDD. Almost all workers in the cohort (97%) had production or 17 maintenance jobs with processes involving TCDD contamination. On average, workers were 18 employed for 2.7 years specifically in processes that involved TCDD contamination, and overall, 19 were employed for 12.6 years. The mortality follow-up began in 1940 and extended until the 20 end of 1987. Vital status was determined using records from the Social Security Administration, 21 the Internal Revenue Service, or the National Death Index. The ascertainment of vital status in 22 the cohort was nearly complete, with less than 1% of the cohort not followed up until death or 23 the end of the study period. 24 Comparisons of mortality were made relative to the U.S. male general population and 25 expressed using the standardized mortality ratio (SMR) metric and 95% confidence intervals 26 (CIs). Life-table methods were used to generate person-years of risk accrued by cohort members 27 at each plant. Person-years and corresponding deaths were tabulated across age, race, and year 28 of death strata, which permitted the SMRs to be examined for potential confounding from these 29 three characteristics. No unadjusted SMRs were presented in the paper. Cross-classification of 30 person-years and deaths was also done across several exposure-related groupings, including 31 duration of employment, years since first exposure, years since last exposure, and duration of This document is a draftfor review purposes only and does not constitute Agency policy. 2-12 DRAFT--DO NOT CITE OR QUOTE 1 exposure. Employment duration was categorized as <5, 5 - <10, 10- <15, 15- <20, 20- <25, 2 2 5 - <30, and >30 years. The variable "years since first exposure" (<10, 10- <20, and >20 years) 3 was used to evaluate associations in relation to different latency periods. The analysis was 4 jointly stratified by duration of employment and for varying latency intervals to evaluate whether 5 cohort members with higher cumulative TCDD levels had higher cancer mortality rates than 6 those cohort members with lower cumulative levels. 7 Overall, the cohort of workers had slightly elevated cancer mortality than the general 8 population (SMR = 1.15, 95% CI = 1.02-1.30). Comparisons to the general population, 9 however, yielded no statistically significant excess for any site-specific cancer. Cancer mortality 10 was examined for the subset of workers that worked for at least one year and had a latency 11 interval of at least 20 years (n = 1,520). The 1-year cut-point was selected based on analyses of 12 serum levels in a subset of 253 workers which revealed that every worker employed for at least 13 one year had a lipid-adjusted serum level that exceeded the mean (7 ppt). Relative to the 14 U.S. general population, statistically significant excesses in cancer mortality were observed for 15 all cancers (SMR = 1.46, 95% CI = 1.21-1.76), cancers of the respiratory system (SMR = 1.42, 16 95% CI = 1.03-1.92), and for soft tissue sarcoma (SMR = 9.22, 95% CI = 1.90-26.95) among 17 this subset of 1,520 male workers. The elevated SMR for soft tissue sarcoma, however, was 18 based on only three cases in this subset. 19 SMRs also were generated across joint categories of duration of exposure and period of 20 latency for deaths from all cancer sites (combined), and cancer of the trachea, bronchus, and 21 lung. Increased SMRs were observed in strata defined by longer exposure and latency, but no 22 statistically significant linear trends were found. 23 24 2.4.1.1.1.1.1.2. Study evaluation. 25 This cohort was the largest of four the International Agency for Research on Cancer 26 (IARC) considered in its 1997 classification of TCDD as a Group 1 human carcinogen (IARC, 27 1997, 537123). Duration of employment in processes that involved TCDD contamination was 28 used as a surrogate measure of cumulative exposure. In using this exposure metric, Fingerhut 29 et al. (1991, 197375) assumed that TCDD exposures were equivalent at all production plants. 30 Doses for individual cohort members were not reconstructed for these analyses, although they 31 were in subsequent analyses of this cohort. This document is a draftfor review purposes only and does not constitute Agency policy. 2-13 DRAFT--DO NOT CITE OR QUOTE 1 Workers in this cohort also were exposed to other chemicals, which could lead to bias 2 due to confounding if these exposures were associated with both TCDD exposure and the health 3 outcomes being examined. At one plant, workers were exposed to 4-aminobiphenyl. Previous 4 investigators also reported that workers at another plant were exposed to 2,4,5-T and 5 2,4-dichlorophenoxyacetic acid (2,4-D) (Bond et al., 1988, 197183; Bond et al., 1989, 064967; 6 Ott et al., 1987, 064994). Although this study did not examine the impact of confounding by 7 other occupational coexposures, subsequent analyses of this cohort showed that associations 8 between cumulative TCDD and all cancer mortality persisted after excluding workers exposed to 9 pentachlorophenols from the analyses (Steenland et al., 1999, 197437). Removal of workers 10 who died from bladder cancer also did not substantially change the dose-response association 11 between TCDD and cancer mortality from all other sites combined. This finding suggests that 12 exposures to 4-aminobiphenyl did not confound the association between cancer mortality and 13 TCDD exposure. Overall, there is little evidence of confounding by these co-exposures among 14 this cohort, however, exposure to other possible confounders, such as dioxin-like compounds, 15 was not examined. 16 The study collected no information on smoking behavior of the workers, and therefore, 17 the SMRs do not account for any differences in the prevalence of smoking that might have 18 existed between the workers and the general population. For several reasons, however, the 19 inability to take into account smoking is unlikely to have been an important source of bias. First, 20 mortality from other smoking-related causes of death such as nonmalignant respiratory disease 21 were not more common in the cohort than in the general population (SMR = 0.96, 22 95% CI = 0.54-1.58). Second, stratified analyses of workers with at least a 20-year latency 23 (assuming this subset shared similar smoking habits) revealed that excesses were apparent only 24 among those who were exposed for at least 1 year. Specifically, when compared to the general 25 population, the SMR among workers exposed for at least 1 year with a latency of 20 years was 26 1.46, (95% CI = 1.21-1.76) while those exposed for less than 1 year had an SMR of 1.02 27 (95% CI = 0.76-1.36). Third, for comparisons of cancer mortality between blue-collar workers 28 and the general population, smoking is unlikely to explain cancer excesses of greater than 29 10-20% (Siemiatycki et al., 1988, 198556). Finally, the investigators found no substantial 30 changes in the results for lung cancer when risks were adjusted for smoking histories obtained in 31 1987 from 223 workers employed at two plants. These data were used to adjust for the expected This document is a draftfor review purposes only and does not constitute Agency policy. 2-14 DRAFT--DO NOT CITE OR QUOTE 1 number of lung cancer deaths expected in the entire cohort (Fingerhut et al., 1991, 197375). 2 Following this adjustment, a small change was observed in the SMR for lung cancer in the 3 overall cohort from 1.11 (95% CI = 0.89-1.37) to 1.05 (95% CI = 0.85-1.30). Similarly, only a 4 slight change in the SMR for lung cancer in the higher exposure subcohort was noted from an 5 SMR of 1.39 (95% CI = 0.99-1.89 to 1.37 (95% CI = 0.98-1.87). 6 The use of death certificate information from the National Death Index is appropriate for 7 identifying cancer mortality outcomes. For site-specific cancers such as soft tissue sarcoma, 8 however, the coding of this underlying cause of death is more prone to misclassification (Percy 9 et al., 1981, 004891). Indeed, a review of tissues from four men concluded to have died from 10 soft-tissue sarcoma determined that two deaths had been misclassified (Fingerhut et al., 1991, 11 197375). A review of hospital data revealed that two other individuals had soft tissue sarcomas 12 that were not identified by death certificate information. The use of death certificate information 13 to derive SMRs for cancer as a whole is likely not subject to significant bias; the same might not 14 hold true, however, for some site-specific cancers such as soft tissue sarcoma. 15 Using the SMR metric to compare an occupational cohort with the general population is 16 subject to what is commonly referred to as the "healthy worker effect" (Choi, 1992, 594250; Li 17 and Sung, 1999, 198427). The healthy worker effect is a bias that arises because those healthy 18 enough to be employed have lower morbidity and mortality rates than the general population. 19 The healthy worker effect is likely to be larger for occupations that are more physically 20 demanding (Aittomaki et al., 2005, 197139; Checkoway et al., 1989, 027173), and the healthy 21 worker effect is considered to be of little or no consequence in the interpretation of cancer 22 mortality (McMichael, 1976, 073484; Monson, 1986, 001410). Few cancers are associated with 23 a prolonged period of poor health that would affect employability long before death. Also 24 recognized is that, as the employed population ages, the magnitude of the healthy worker effect 25 decreases as the absolute reduction in mortality becomes relatively smaller in older age groups 26 (McMichael, 1976, 073484). The mortality follow-up of occupational cohorts generally spans 27 several decades, which should minimize the associated healthy worker effect in such studies. 28 Bias could also be introduced in that workers who are healthier might be more likely to stay 29 employed and therefore accrue higher levels of exposure. In the NIOSH cohort, however, 30 mortality was ascertained for those who could have left the workforce or retired by linking 31 subjects to the National Death Index. Although internal cohort comparisons can minimize the This document is a draftfor review purposes only and does not constitute Agency policy. 2-15 DRAFT--DO NOT CITE OR QUOTE 1 potential for the healthy worker effect for the reasons presented above, for cancer outcomes, the 2 SMR statistic is a valuable tool for characterizing whether occupational cohort are more likely to 3 die of cancer than the general population. Moreover, stratified analyses across categories of 4 duration of exposure, or latency periods within a cohort can yield important insights about which 5 workers are at greatest risk. Perhaps most important, subsequent analyses of the NIOSH cohort 6 that presented risk estimates derived from external comparisons using the SMR were remarkably 7 consistent with rate ratios derived using an internal referent (Steenland et al., 1999, 197437). 8 9 2.4.1.1.1.1.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 10 This cohort meets most of the identified considerations for conducting a quantitative 11 dose-response analysis for mortality from all cancer sites combined. The NIOSH cohort is the 12 largest cohort of TCDD-exposed workers, exposure characterization at an individual level is 13 possible but not available in this particular study, and the follow-up period is long enough to 14 evaluate latent effects. Although there is no direct evidence of any important sources of bias, 15 confounding may be present due to a lack of consideration of dioxin-like compounds. For the 16 purpose of quantitative dose-response modeling, it is important to note that subsequent studies of 17 this cohort adopted methods that greatly improved the characterization of TCDD exposure in this 18 cohort and increased the follow-up interval (Cheng et al., 2006, 523122; Steenland et al., 2001, 19 197433). As such, for all practical purposes, due consideration for dose-response modeling 20 should focus on the more recently developed data sets. 21 For quantitative dose-response modeling for individual cancer sites, the data are much 22 more limited. A statistically significant positive association with TCDD was noted only for soft23 tissue sarcoma among those with more than 1 year of exposure and 20 years of latency 24 (SMR = 9.22, 95% CI = 1.90-26.95). However there were only three deaths from soft tissue 25 sarcoma among this exposed component of the cohort, and four deaths in total in the overall 26 cohort. Also, misclassification of outcome for soft-tissue sarcoma through death registries is 27 well recognized and supported with additional review of tissue from two of the men. 28 Specifically, tissues from the four men who died of soft-tissue sarcoma revealed that only two of 29 these cases were coded correctly. 30 Although subsequent analyses of the NIOSH cohort did not show evidence of 31 confounding by other occupational exposures, the design of this initial publication of the NIOSH This document is a draftfor review purposes only and does not constitute Agency policy. 2-16 DRAFT--DO NOT CITE OR QUOTE 1 cohort did not allow for examination of exposures to other possible confounders, such as dioxin 2 like compounds. Duration of exposure was used as a surrogate for cumulative TCDD exposure; 3 therefore, effective doses could not be estimated. Therefore, dose-response modeling was not 4 conducted for this study. 5 6 2.4.1.1.1.1.2. Steenland et al. (1999, 197437). 7 2.4.1.1.1.1.2.1. Study summary. 8 A subsequent analysis of the NIOSH cohort extended the follow-up interval of Fingerhut 9 et al. (1991, 197375) by 6 years (i.e., from 1940-1993) and improved characterization of TCDD 10 exposure (Steenland et al., 1999, 197437). A key distinction from the work of Fingerhut et al. 11 (1991, 197375) was the exclusion of several workers that had been included in the previous 12 mortality analyses. The authors excluded 40 workers who were either female, had never worked 13 in TCDD-exposed departments, or had missing date of birth information. An additional 14 238 workers were excluded as occupational data for characterizing duration of exposure were 15 lacking, preventing their use in a subcohort dose-response analysis. This subcohort was further 16 reduced by excluding workers from four plants (n = 591) because the information on the degree 17 of TCDD contamination in work histories was limited, preventing the characterization of TCDD 18 levels by job type. Thirty-eight additional workers were excluded from the eight remaining 19 plants because TCDD contamination could not be estimated. Finally, 727 workers were 20 excluded because they had been exposed to pentachlorophenol. In total, exposures were 21 assigned to 3,538 (69%) members of the overall cohort, a cohort substantially reduced from the 22 5,172 on which Fingerhut et al. (1991, 197375) reported. Steenland et al. (1999, 197437) also 23 evaluated the mortality experience of a subcohort of 608 workers with chloracne who had no 24 exposure to pentachlorophenol. 25 For each worker, a quantitative exposure score for each day of work was calculated based 26 on the concentration of TCDD (pg/g) present in process materials, the fraction of the day 27 worked, and a qualitative contact level based on estimates of the amount of TCDD exposure via 28 dermal absorption or inhalation. The authors derived a cumulative measure of TCDD exposure 29 by summing the exposure scores across the working lifetime history for each worker. The 30 authors validated this cumulative exposure metric indirectly by comparing values obtained for 31 workers with and without chloracne. Such a validation is appropriate, given that chloracne is This document is a draftfor review purposes only and does not constitute Agency policy. 2-17 DRAFT--DO NOT CITE OR QUOTE 1 considered a clinical sign of exposure to high doses of dioxin (e.g., Ott et al., 1993, 594322). 2 The median exposure score among those with chloracne was 11,546 compared with 77 among 3 those without (Steenland and Deddens, 2003, 198587). 4 Cancer mortality was compared using two approaches. As in Fingerhut et al. (1991, 5 197375), external comparisons were made to the U.S. general population using the SMR 6 statistic. The authors adjusted the SMR statistics for race, age, and calendar time. They also 7 applied life-table methods to characterize risks across the subcohort of 3,538 workers with 8 exposure data by categorizing the workers into seven cumulative exposure groups. The 9 cut-points for these categories were selected so that the number of deaths in each category was 10 nearly equal to optimize study power. Life-table analyses were extended further to consider a 11 15-year lag interval, which in a practical sense means that person-years at risk would not begin 12 to accrue until 15 years after the first exposure occurred. The person-years and deaths that 13 occurred in the first 15 years were included in the lowest exposure grouping. The Cox 14 proportional hazards model was used to characterize risk within the cohort. Cox regression was 15 used to provide an estimate of the hazard ratios and the 95% CIs for ischemic heart disease, all 16 cancers combined, lung cancer, smoking related cancers, and all other cancers. The authors also 17 performed Cox regression analyses using the seven categories of exposure, adjusting the 18 regression coefficients for year of birth and age. The regression models were run for both 19 unlagged and lagged (15 years) cumulative exposure scores. 20 Overall, when compared with the U.S. general population, a slight excess of cancer 21 mortality (from all sites) was noted in the 5,132 cohort study population (SMR = 1.13, 22 95% CI = 1.02-1.25). This result did not substantially differ from the earlier finding that 23 Fingerhut et al. (1991) published (SMR = 1.15, 95% CI = 1.03-1.30). Site-specific analyses 24 revealed statistically significant excesses relative to the U.S. general population for bladder 25 cancer (SMR = 1.99, 95% CI = 1.13-3.23) and for cancer of the larynx (SMR = 2.22, 26 95% CI = 1.06-4.08). In the chloracne subcohort (n = 608), SMRs of 1.25 27 (95% CI = 0.98-1.57) and 1.45 (95% CI = 0.98-2.07) were found for all cancer sites and for 28 lung cancer, repectively, relative to the general population. The authors also found statistically 29 significant excesses for connective and soft tissue sarcomas (SMR = 11.32, 30 95% CI = 2.33-33.10) and for lymphatic and hematopoietic malignancies (SMR = 3.01, 31 95% CI = 1.43-8.52). This document is a draftfor review purposes only and does not constitute Agency policy. 2-18 DRAFT--DO NOT CITE OR QUOTE 1 External comparisons made by grouping workers into septiles of cumulative TCDD 2 exposure and generating an SMR for each septile using the U.S. population as the referent group 3 suggested a dose-response relationship. For all cancer sites combined, workers in the highest 4 exposure score category had an SMR of 1.60 (95% CI = 1.15-1.82); increases also were 5 observed in the sixth (SMR = 1.34) and fifth (SMR = 1.15) septiles. The two-sided p-value 6 associated with the test for trend for cumulative TCDD exposure was statistically significant 7 (p = 0.02). A similar approach for lung cancer revealed virtually the same pattern. The 8 incorporation of a 15-year latency for the analyses of all cancer deaths, in general, produced 9 slightly higher SMRs across the septiles, although a slight attenuation of effect was noted in the 10 highest septile (SMRunlagged= 1.60 vs. SMRlagged= 1.54). For a 15-year lag, the lung cancer 11 SMRs were mixed compared to the unlagged results with some septile exposure categories 12 increasing and others decreasing relative to the lowest exposure group. 13 For the internal cohort comparisons using Cox regression analyses higher hazard ratios 14 were found among workers in the higher exposure categories than in the lowest septile. The 15 linear test for trend, however, was not statistically significant (p = 0.10). The associations across 16 the septiles for the unlagged exposure for the internal cohort comparisons were not as strong as 17 for the external cohort comparisons. The opposite was true, however, for cumulative exposures 18 lagged 15 years. 19 Relative to the lowest septile, stratified analyses revealed increased hazard ratios in the 20 upper septiles of the internal cohort comparisons for both smoking- and nonsmoking-related 21 forms of cancer. The test for linear trend was statistically significant for all other cancers (after 22 smoking-related cancers were excluded). These analyses suggest that the overall cancer findings 23 were not limited to an interaction between TCDD and smoking. Additional sensitivity analyses 24 by the authors indicated the findings for smoking-related cancers were largely unaffected by the 25 exclusion of bladder cancer cases. This observation suggests that the exposure to 26 4-aminobiphenyl, which occurred at one plant and might have contributed to an increased 27 number of bladder cancers, did not substantially bias the dose-response relationship between 28 TCDD and all cancers combined. 29 The investigators also evaluated the dose-response relationship with a Cox regression 30 model separately for each plant using internal cohort comparisons and found some heterogeneity. 31 This finding is not unexpected particularly given the relatively small number of cancer deaths at This document is a draftfor review purposes only and does not constitute Agency policy. 2-19 DRAFT--DO NOT CITE OR QUOTE 1 each plant, and given that exposures were quite low for one plant at which no positive 2 association was found. The variability among plants was taken into account by modeling plant 3 as a random effect measure in the Cox model, which produced little change in the slope 4 coefficient (P = 0.0422 vs. P = 0.0453). 5 6 2.4.1.1.1.1.2.2. Study evaluation. 7 This study represents a valuable extension of that by Fingerhut et al. (1991, 197375). 8 Internal comparisons were performed to help minimize potential biases associated with using an 9 external comparison group (e.g., healthy worker effect, and differences in other risk factors 10 between the cohort and the general population). That similar dose-response relationships were 11 found for internal and external comparison populations suggests that the bias due to the health 12 worker effect in the cohort might be minimal for cancer mortality. More importantly, the 13 construction of the cumulative exposure scores provides an improved opportunity to evaluate 14 dose-response relationships compared with the length of exposure and duration of employment 15 metrics that Fingerhut et al. (1991, 197375) used. 16 A potential limitation of the NIOSH study was the inability to account for cigarette 17 smoking. If cigarette smoking did contribute to the increased cancer mortality rates in this and 18 other cohorts, increased cancer mortality from exposure to TCDD would be expected only for 19 smoking-attributable cancers. This study demonstrates associations with TCDD for both 20 smoking- and nonsmoking-related cancers, including a stronger association for 21 nonsmoking-related cancers. Therefore, the data provide evidence that associations between 22 TCDD and cancer mortality are not likely due to cigarette smoking. 23 The findings regarding latency should be interpreted cautiously as the statistical power in 24 the study to compare differences across latency intervals was limited. Caution also should be 25 heeded, given that latency intervals can vary on an individual basis as they are often 26 dose-dependent (Guess and Hoel, 1977, 197464). The evaluation of whether TCDD acts as 27 either an initiating or promoting agent (or both) is severely constrained by the reliance on cancer 28 mortality data rather than incidence data. This constraint is due to the fact that survival time can 29 be quite lengthy and can vary substantially across individuals and by cancer subtype. For 30 example, the 5-year survival among U.S. males for all cancer sites combined ranged between 45 31 and 60% (Clegg et al., 2002, 594267). When only mortality data are available, evaluating the This document is a draftfor review purposes only and does not constitute Agency policy. 2-20 DRAFT--DO NOT CITE OR QUOTE 1 time between when individuals are first exposed and when they are diagnosed with cancer is 2 nearly impossible. 3 Starr (2003, 594271) suggested that Steenland et al. (1999, 197437) focused too heavily 4 on the exposures that incorporated a 15-year period of latency and that those who experienced 5 high exposures would inappropriately contribute person-years to the lowest exposure group 6 "irrespective of how great the workers' actual cumulative exposure scores may have been." 7 Most cancer deaths would, however, typically occur many years postemployment. Given that 8 the follow-up interval of the cohort was long and the average exposure duration was 2.7 years, at 9 the time of death, person-years for those with high cumulative exposures would be captured 10 appropriately. The median 5-year survival for all cancers is approximately 50% (Clegg et al., 11 2002, 594267), so applying a minimum latency of 5 years when using cancer mortality rather 12 than cancer incidence data is needed to assure that the exposure metric is capturing exposures 13 that occur before diagnoses. Increasing this latency period, for example to 10 or 15 years, would 14 eliminate consideration of exposures that occur in the period between tumor occurrence and 15 tumor detection (diagnosis), and allows for an appropriate focus on exposures that act either 16 early or late in the pathogenic process. If the association of TCDD with cancer is causal, effects 17 might become apparent only at high exposures and with adequate latency. As such, IARC has 18 concluded that a latency interval of 15 years could be too short (IARC, 1997, 537123). EPA 19 considers the Steenland et al. (1999, 197437) presentation to be balanced in that they provided 20 the range in lifetime excess risk estimated across the various models used. The authors' finding 21 that the models with a 15-year lag provided a statistically significant improvement in fit based on 22 the chi-square test statistic should not be readily dismissed. 23 24 2.4.1.1.1.1.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 25 This study meets most of the epidemiological considerations for conducting a 26 quantitative dose-response analysis for mortality from all cancer sites combined. This study 27 excludes a large number of workers who were exposed to pentachlorophenol, thus eliminating 28 the potential for bias from this exposure and used an improved methodology for assigning TCDD 29 exposures to the workers. However, given that exposures to other dioxin-like compounds were 30 not described, it is unclear if the exposures among this cohort were primarily to TCDD. 31 Therefore, dose-response modeling was not pursued for this study, but was for the subsequent This document is a draftfor review purposes only and does not constitute Agency policy. 2-21 DRAFT--DO NOT CITE OR QUOTE 1 NIOSH study by Steenland et al. (2001, 197433), which did examine exposure to dioxin-like 2 compounds. 3 4 2.4.1.1.1.1.3. Steenland et al. (2001, 197433). 5 2.4.1.1.1.1.3.1. Study summary. 6 In 2001, Steenland et al. published a risk analysis using the NIOSH cohort that for the 7 first time incorporated serum measures in the derivation of TCDD exposures for individual 8 workers. The authors applied the same exclusion criteria to the entire cohort of workers across 9 the 12 plants in the Steenland et al. (1999, 197437) study, which left 3,538 workers for which 10 risk estimates could be calculated. Cumulative TCDD serum levels were estimated on an 11 individual basis for all 3,538 workers by developing predictive models that used a subset of 12 170 workers for which both serum measures and TCDD exposures scores were available 13 (Steenland et al., 2001, 197433). Unlike previous analyses of the NIOSH cohort that considered 14 several different mortality outcomes, the analyses presented in Steenland et al. (2001, 197433) 15 focused exclusively on mortality from all cancers sites combined. The authors observed 16 256 cancer deaths in the cohort during the follow-up interval that extended from 1942 until the 17 end of 1993. All risks estimated in the Steenland et al. (2001, 197433) study were based on 18 internal cohort comparisons. 19 Characterization of TCDD exposure levels among the workers was based on serum 20 measures obtained in 1988 from 199 workers who were employed in one of the eight plants. The 21 researchers restricted the development of the model to include only those workers whose 22 measured serum levels were deemed to be greater than the upper range of background levels 23 (10 ppt), which resulted in 170 workers. 24 The authors developed a regression model that could estimate the level of TCDD at the 25 time of last exposure for the 170 workers. The model was developed based on the estimated 26 half-life of TCDD, the known work history of each worker, a pharmacokinetic model for the 27 storage and excretion of TCDD, and exposure scores for each job held by each worker over time. 28 The resulting equation follows 29 30 y lastexposure y 1988exp(kAf) (Eq. 2-1) 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-22 DRAFT--DO NOT CITE OR QUOTE 1 The first-order elimination rate constant (X) was based on a half-life of 8.7 years 2 previously reported for the Ranch Hands cohort (Michalek et al., 1996, 198893). The 3 background rate of TCDD exposure was assumed to be 6.1 parts per trillion (ppt), which was 4 based on the median level in a sample of 79 unexposed workers in the NIOSH cohort (Piacitelli 5 et al., 1992, 197275). This value was subtracted when TCDD values were back-extrapolated, 6 and then added again after the back-extrapolation was completed. A background level of 5 ppt 7 also was used in some of the analyses with minimal demonstrable effects on the results. 8 Sensitivity analyses also were incorporated to consider a 7.1-year half-life estimate that had been 9 developed for the earlier Ranch Hands study (Pirkle et al., 1989, 197861). 10 After back-extrapolating to obtain TCDD serums levels at the time of last exposure, the 11 investigators estimated cumulative (or "area under the curve") TCDD serum levels for every 12 cohort member. This estimation procedure was the same method Flesch-Janys et al. (1998, 13 197339) applied to the Hamburg cohort to derive a coefficient for relating serum levels to 14 exposure scores. The "area under the curve" approach integrates time-specific serum levels over 15 the employment histories of the individual workers. The slope coefficient was estimated using a 16 no-intercept linear regression model. This model is based on the assumption that a cumulative 17 score of zero is associated with no serum levels above background. 18 Cox regression was also used to model the continuous measures of TCDD. A variety of 19 exposure metrics were considered that took into account different lags, nonlinear relationships 20 (e.g., log-transform and cubic spline), as well as threshold and nonthreshold exposure metrics. 21 Categorical analyses were used to evaluate risks across TCDD exposure groups, while different 22 shapes of dose-response curves were evaluated through the use of lagged and unlagged 23 continuous TCDD measures. Categorical analyses of TCDD exposure were conducted using the 24 Cox regression model to derive estimates of relative risk (RR) as described by hazard ratios and 25 95% CIs. The reference group in this analysis was those workers in the lowest septile 26 cumulative exposure grouping (<335 ppt-years). The septiles were chosen based on cumulative 27 serum levels that considered no lag and also a 15-year lag. 28 The investigators also conducted dose-response analyses using the toxicity equivalence 29 (TEQ) approach. The TEQ is calculated as the sum of all exposures to dioxins and furans 30 weighted by the potency of each specific compound. In this study, TCDD was assumed to be 31 account for all dioxin exposures in the workplace. For background TEQ levels, the investigators This document is a draftfor review purposes only and does not constitute Agency policy. 2-23 DRAFT--DO NOT CITE OR QUOTE 1 used a value of 50 ppt in the dose-response modeling. This is based on the assumption that 2 TCDD accounted for 10% of the toxicity of all dioxins and furans (WHO, 1988, 594278), and is 3 equivalent to using a background level of 5 ppt/yr that was used in the derivation of cumulative 4 serum TCDD levels. A statistically significant dose-response pattern was observed for all cancer 5 mortality and TCDD exposure based on log of cumulative TEQs with a 15-year lag. A 6 comparison of the overall model chi-square values indicated that the fit of this model was not as 7 good as that for TCDD. 8 The hazard ratios among workers grouped by categories of cumulative TCDD exposure 9 (lagged 15 years) suggested a dose-response relationship. Steenland et al. (2001, 197433) found 10 statistically significant excesses in the higher exposure categories compared to the lowest septile. 11 The RR was 1.82, 95% CI = 1.18-2.82 for the sixth septile (7,568-20,455 ppt-years) and 1.62, 12 95% CI = 1.03-2.56) for the seventh septile (>20,455 ppt-years). Cox regression indicated that 13 log TCDD serum concentrations (lagged 15 years) was positively associated with cancer 14 mortality (P = 0.097, standard error (P) = 0.032, p < 0.003). A statistically significant 15 improvement in fit was observed when a 15-year lag interval was incorporated into the model 16 compared to a model with no such lag [Model x2with 4 degrees of freedom (df) = 7.5]. Results 17 were similar when using a half-life of 7.1 years rather than 8.7 years. The excess lifetime risk of 18 death from cancer at age 75 for TCDD intake (per 1.0-picogram per kilogram [pg/kg] of body 19 weight (BW) per day) was about 0.05-0.9% above a background lifetime risk of cancer death of 20 12.4%. The results from the best-fitting models provide lifetime risk estimates within the ranges 21 derived using data from the Hamburg cohort (Becher et al., 1998, 197173). 22 In both categorical and continuous analyses of TCDD based on a linear exposure metric, 23 the dose-response pattern tailed off at high exposures suggesting nonlinear effects. This 24 phenomenon could be due to saturation effects (Stayner et al., 2003, 054922) or, alternatively, 25 could have resulted from increased exposure misclassification of higher exposures (Steenland 26 et al., 2001, 197433). As the authors highlighted, some of the highest exposures might have 27 been poorly estimated as they occurred in workers exposed to short-term high exposures during 28 the clean-up of a spill. The choice of a linear model to develop data from a single time point can 29 also result in exposure misclassification in those individuals that have differences in the length of 30 exposure (Emond et al., 2005, 197317). Misclassification would be less likely at low 31 concentrations where dose-dependent elimination is minimal. This document is a draftfor review purposes only and does not constitute Agency policy. 2-24 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.1.3.2. Study evaluation. 2 An important consideration in the Steenland et al. (2001, 197433) study was the use of a 3 small subset of workers (n = 170) to infer exposures for the remainder of the cohort. This subset 4 comprised surviving members of the cohort (in 1988), and therefore, their age distribution would 5 have differed from the rest of the cohort. Furthermore, these workers were employed at a single 6 plant, at which the work histories were less detailed than at other plants; thus, the development of 7 the exposure scores differed between this plant and that of the others. Also, many of the workers 8 at this plant had the same job title and were employed during the same calendar period. The use 9 of serum data from this subset adds a level of uncertainty that is not readily characterized. 10 Despite this limitation, the use of these sera data to derive cumulative measures for all cohort 11 workers has merit given the strong correlation observed between the exposure scores, and TCDD 12 serum levels estimates at the time of last exposure (Spearman r = 0.90). 13 The authors performed an extensive series of sensitivity analyses and considered several 14 alternative exposure metrics to the simple linear model. The lifetime excess risk above 15 background was nearly twice as high for the log cumulative serum measures with a 15-year lag 16 when compared to the piecewise linear models with no lag. An important observation was that 17 the exposure metric based on cumulative serum (lagged 15 years) did not fit the data as well as 18 the cumulative exposure score used in earlier analyses (Steenland et al., 1999, 197437). A priori, 19 one would expect that a better fit would be obtained with serum-based measures because serum 20 levels are a better measure of relevant biological dose. As the authors noted, inaccuracies 21 introduced in estimating the external-based exposure scores could have contributed to a poorer 22 fit of the data. Alternatively, exposure misclassification error could be introduced if serum 23 samples based on the 170 workers were not representative of the entire cohort. Although the 24 serum-based measures did not fit the data as well as the exposures scores, the authors regarded 25 them as providing a reasonable fit based on an improvement in log likelihood of 3.99 (between 26 the log cumulative serum model and the log cumulative exposure score model). Moreover, the 27 serum-based measures enabled better characterization of risk in units (pg/kg-day) that can be 28 used in regulation exposures. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-25 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.1.3.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 This study meets all of the epidemiological considerations for conducting a quantitative 3 dose-response analysis for mortality from all cancer sites combined. As mentioned previously, 4 the NIOSH cohort is the largest assembled to date for which TCDD-related risks of cancer 5 mortality can be estimated. The use of serum-based measures provides an objective measure of 6 TCDD exposure. Repeated measures in other study populations have provided reasonable 7 estimates of the half-life of TCDD, which permitted back-extrapolation of exposures. 8 The authors have made extensive efforts to evaluate a wide variety of nonlinear and 9 linear models with varying lengths of latency and log transformations. The model chi-square test 10 statistics were fairly similar for the log cumulative serum (15-year lag) (Model %2(4df) = 11.3) 11 model and the piecewise linear model (no lag) (Model %2(5df) = 12.5). These models, however, 12 produced results with twofold differences in lifetime excess risks. These differences underscore 13 the importance of characterizing uncertainty in modeling approaches when conducting 14 dose-response analysis. 15 The Steenland et al. (2001, 197433) study characterizes risk in terms of pg/kg of body 16 weight per day. Given that tolerable daily intake dioxin levels are typically expressed in pg/kg 17 of body weight (WHO, 1988, 594278), the presentation of risks in terms of these units is an 18 important advance from the earlier analyses that used exposure scores (Steenland et al., 1999, 19 197437). Many of the Steenland et al. (2001, 197433) findings are consistent with earlier work 20 from this cohort, which is not surprising given that exposures scores were used to derive serum21 based levels for the cohort. The findings of excess lifetime risks obtained for the best- fitting 22 model are also consistent with those derived from the Hamburg cohort (Becher et al., 1998, 23 197173). This study meets the epidemiological considerations noted previously as there is no 24 evidence that the study is subject to bias from confounding due to cigarette smoking or other 25 occupational exposures. Given the considerable efforts to measure effective dose to TCDD 26 among the study participants, this study also meets the requisite dose-response modeling criteria 27 and will be used in quantitative dose-response analyses of cancer mortality. 28 This document is a draftfor review purposes only and does not constitute Agency policy. 2-26 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.1.4. Cheng et al. (2006, 523122). 2 2.4.1.1.1.1.4.1. Study summary. 3 Cheng et al. (2006, 523122) undertook a subsequent quantitative risk assessment of 4 3,538 workers in the NIOSH cohort using serum-derived estimates of TCDD. This 5 dose-response analysis was published after the 2003 Reassessment document was released. The 6 goal of this study was to examine the relationship between TCDD and cancer mortality (all sites 7 combined) using a new estimate of dose that estimated TCDD as a function of both exposure 8 intensity and age using a kinetic model. This physiologically based pharmacokinetic model has 9 been termed the "concentration- and age-dependent elimination model" (CADM) and was 10 developed by Aylward et al. (2005, 197014). This model describes the kinetics of TCDD 11 following oral exposure to humans by accounting for key processes affecting kinetics by 12 simulating the total concentration of TCDD based on empirical consideration of hepatic 13 processes (see Section 3.3). An important feature of this kinetic model is that it incorporates 14 concentration- and age-dependent elimination of TCDD from the body; consequently, the 15 effective half-life of TCDD elimination varies based on exposure history, body burden, and age 16 of the exposed individuals. The study was motivated by the reasoning that back-calculations of 17 TCDD using a first-order elimination model and a constant half-life of 7-9 years underestimated 18 exposures to TCDD among workers. This underestimate, in turn, would result in overestimates 19 of the carcinogenic potency of TCDD. 20 As with the earlier Steenland et al. (2001, 197433) analyses, the cohort follow-up period 21 was extended from 1942 until the end of 1993 and work histories were linked to a job exposure 22 matrix to obtain cumulative TCDD scores. Two cumulative serum lipid exposure metrics (in 23 ppt-years) were constructed using the data obtained from the sample of 170 workers. The first 24 replicated the metric used in a previous analysis of the cohort (Steenland et al., 2001, 197433) 25 and was based on a first-order elimination model with an 8.7-year half-life (Michalek et al., 26 1996, 198893). The second metric was based on CADM and had two first-order elimination 27 processes (Aylward et al., 2005, 197114). This metric assumes that the elimination of TCDD in 28 humans occurs at a faster rate when body concentrations are high and at slower rates in older 29 individuals (Aylward et al., 2005, 197114; Aylward et al., 2005, 197014). The model was 30 optimized using individuals for which serial measures of serum TCDD were available. These 31 measures were obtained from 39 adults with initial serum levels between 130 and 144,000 ppt This document is a draftfor review purposes only and does not constitute Agency policy. 2-27 DRAFT--DO NOT CITE OR QUOTE 1 (Aylward et al., 2005, 197014). This group included 36 individuals who had been exposed in the 2 Seveso accident and 3 exposed in Vienna, Austria. In practice, for serum levels greater than 3 1,000 ppt, the effective half-life would be less than 3 years, and for serum TCDD levels less than 4 50 ppt, the effective half-life would be more than 10 years (Aylward et al., 2005, 197014). 5 Results from the model indicate that men eliminate TCDD faster than women do as 6 demonstrated previously by Needham et al. (1994, 200030). These age- and 7 concentration-dependent processes were assumed to operate independently on TCDD in hepatic 8 and adipose tissues, and TCDD levels in liver and adipose tissue were assumed to be a nonlinear 9 function of body concentration. Cheng et al. (2006, 523122) calibrated CADM using a dose of 10 156 ng per unit of exposure score and assumed a background exposure rate of 0.01 ng/kg-month. 11 The average TCDD ppt-years derived from CADM with a 15-year lag was 4.5-5.2 times higher 12 than with the first-order elimination model. The two metrics, however, were highly correlated 13 based on a Pearson correlation coefficient of 0.98 (p < 0.001). Comparisons of fit between the 14 CADM and first-order elimination model were made using R2values and presented in Aylward 15 et al. (2005, 197014). 16 Cheng et al. (2006, 523122) compared the mortality experience of NIOSH workers to the 17 U.S. general population using the SMR statistic. SMR statistics also were generated separately 18 for each of the 8 plants and for all plants combined. Cox regression models were used to analyze 19 internal cohort dose-response. These models used age as the time variable, and penalized 20 smoothing spline functions of the CADM metric also were considered. The possible 21 confounding effects of other occupational exposures and other regional population differences 22 were assessed by repeating analyses after excluding one plant at a time. Lagged and unlagged 23 TCDD exposures were analyzed separately, and stratified analyses compared risk estimates for 24 smoking- and nonsmoking-related cancers. Cheng et al. (2006, 523122) adjusted the slope 25 estimates derived from the Cox model for potential confounding effects of race and year of birth. 26 Overall, a statistically significant excess in all cancer mortality in the cohort occurred 27 relative to the general population (SMR = 1.17, 95% CI = 1.03-1.32). The plant-specific SMRs 28 ranged from 0.62-1.87, with a statistically significant excess evident only for plant 10 29 (SMR = 1.87, 95% CI = 1.35-2.52). For lung cancer mortality, the overall SMR was not 30 statistically significant (SMR = 1.11, 95% CI = 0.89-1.37). A statistically significant excess for 31 lung cancer also was found for plant 10 (SMR = 2.35, 95% CI = 1.44-3.64). The SMRs between This document is a draftfor review purposes only and does not constitute Agency policy. 2-28 DRAFT--DO NOT CITE OR QUOTE 1 smoking- (SMR = 1.22, 95% CI = 1.01-1.45) and nonsmoking-related cancers (SMR = 1.12, 2 95% CI = 0.94-1.33) were comparable. 3 For the internal cohort analyses of serum-derived measures, the authors were able to 4 replicate the one-compartmental model used previously (Steenland et al., 2001, 197433). As had 5 been noted by Steenland et al. (2001, 197433), an inverse-dose-response pattern was seen for 6 individuals with high exposures (above 95th percentile); this type of pattern is often seen in 7 occupational studies (Stayner et al., 2003, 054922). Excluding these data produced a stronger 8 association between TCDD and all-cause mortality. In fact, only when the upper 2.5% or 5% of 9 observations was removed did a statistically significant positive association become evident with 10 the untransformed data. Similarly, when the model incorporated a lag of 15 years, a statistically 11 significant association was noted only for the untransformed TCDD ppt-years with the upper 5% 12 of observations removed. Stratified analyses revealed little difference between smoking- and 13 nonsmoking-related cancers, and the removal of one plant at a time from the analyses of TCDD 14 ppt-years changes did not substantially change the slope. 15 16 2.4.1.1.1.1.4.2. Study evaluation. 17 The authors reported that CADM provided an improved fit over the one-compartmental 18 model, but presented no evidence regarding any formal test of statistical significance. A 19 comparison of R2values presented in Aylward et al. (2005, 197014), however, does reveal that 20 the R2value increased from 0.27 (first-order compartmental model with an 8.7-year half-life) to 21 0.40 for CADM. TCDD exposures estimated using CADM were approximately fivefold higher 22 than the one-compartmental model estimates among cohort members with higher levels of 23 exposure. Differences in exposure estimates between the two metrics were less striking among 24 individuals with lower TCDD exposures. The net effect was that CADM produced a 6- to 25 10-fold decrease in estimated risks compared to estimates previously reported (Steenland et al., 26 2001, 197433). Nonetheless, the estimates produced by CADM span more than two orders of 27 magnitude under various assumptions. Further uncertainties arise from between-worker 28 variability of TCDD elimination rates, possible residual confounding, and the variability 29 associated with the use of data obtained from other cohorts. Nevertheless, the use of the CADM 30 model to estimate TCDD exposure is considered a significant advantage over the previous first 31 order body burden calculations. This document is a draftfor review purposes only and does not constitute Agency policy. 2-29 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.1.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 The value of including the NIOSH cohort data has already been established based on 3 investigations published by Steenland et al. (1999, 197437; 2001, 197433). The decision to 4 include data from the quantitative dose-response analysis that Cheng et al. (2006, 523122) 5 conducted relates to the added value that the CADM exposure estimates would provide. The 6 earlier modeling work of Aylward et al. (2005, 197014) provided some support for a modest 7 improvement of the fit of CADM over the first-order compartmental model, and they also 8 confirmed previous studies that found that TCDD elimination rates varied by age and sex. 9 Recent work by Kerger et al. (2006, 198651) also demonstrates that the half-life for TCDD is 10 shorter among Seveso children than the corresponding half-life for adults, and that body burdens 11 influence the elimination of TCDD in humans. That estimates of half-lives among men have 12 been remarkably consistent, with mean estimates ranging between 6.9 and 8.7 years 13 (Flesch-Janys et al., 1996, 197351; Michalek et al., 2002, 199579; Needham et al., 2005, 14 594295; Pirkle et al., 1989, 197861), however, is noteworthy. Based on the underlying strengths 15 of the NIOSH cohort data and efforts by Cheng et al. (2006, 523122) to improve estimates of 16 effective dose, these data support further dose-response modeling. 17 18 2.4.1.1.1.1.5. Collins et al. (2009, 197627). 19 2.4.1.1.1.1.5.1. Study summary. 20 In a recent study, Collins et al. (2009, 197627) investigated the relationship between 21 serum TCDD levels and mortality rates in a cohort of trichlorophenol workers exposed to 22 TCDD. These workers were part of the NIOSH cohort having accounted for approximately 45% 23 of the person-years in an earlier analysis (Bodner et al., 2003, 197135). The investigators 24 completed an extensive dioxin serum evaluation of workers employed by the Dow Chemical 25 plant in Midland, Michigan, that made 2,4,5-trichlorophenol (TCP) from 1942 to 1979 and 26 2,4,5-T from 1948 to 1982. Collins et al. (2004, 197267) developed historical TCDD exposure 27 estimates for all TCP and 2,4,5-T workers. This study represents the largest group of workers 28 from a single plant ever studied for the health effects of TCDD. Little information on how vital 29 status was ascertained, either in this paper or in the Bodner et al. (2003, 197135) report of 30 mortality in this cohort. Although the authors indicate that death certificates were obtained from This document is a draftfor review purposes only and does not constitute Agency policy. 2-30 DRAFT--DO NOT CITE OR QUOTE 1 the states in which the employees died, whether vital status was ascertained from company 2 records or through record linkage to the National Death Index is unclear. 3 The follow-up interval for these workers covered the period between 1942 and 2003. 4 Thus, the study included 10 more years of follow-up than earlier investigations of the entire 5 NIOSH cohort. Serum samples were obtained from 280 former workers collected during 6 2004-2005. A simple one-compartment first-order pharmacokinetic model and elimination rates 7 as estimated from the BASF cohort were used (Flesch-Janys et al., 1996, 197351). The "area 8 under the curve" approach was used to characterize workers' exposures over the course of their 9 working careers and provided a cumulative measure of exposure. Analyses were performed with 10 and without 165 of the 1,615 workers exposed to pentachlorophenol to evaluate the impact of 11 these exposures. 12 External comparisons of cancer mortality rates to the general U.S. population were made 13 using SMRs. Internal cohort comparisons of exposure-response relationships were made using 14 the Cox regression model. This model used age as the time variable, and was adjusted for year 15 of hire and birth year. Only those causes of death for which an excess was found based on the 16 external comparisons or for which previous studies had identified a positive association were 17 selected for dose-response analyses. 18 A total of 177 cancer deaths were observed in the cohort. For the external comparison 19 with the U.S. general population, overall, no statistically significant differences were observed in 20 all cancer mortality among all workers (SMR = 1.0, 95% CI = 0.8-1.1). Results obtained after 21 excluding workers exposed to pentachlorophenol were similar (SMR = 0.9, 95% CI = 0.8-1.1). 22 Excess mortality in the cohort were found for leukemia (SMR = 1.9, 95% CI = 1.0-3.2) and soft 23 tissue sarcoma (SMR = 4.1, 95% CI = 1.1-10.5). Although not statistically significant SMRs for 24 other lymphohemopoietic cancers included non-Hodgkin's lymphoma SMR = 1.3; 95%CI = 0.6, 25 2.5) and Hodgkin's disease (SMR = 2.2; 95% CI = 0.2, 6.4). 26 Internal cohort comparisons using the Cox regression model were performed for all 27 cancers combined, lung cancer, prostate cancer, leukemia, non-Hodgkin's lymphoma, and 28 soft-tissue sarcoma. Whether the internal comparisons excluded those workers exposed to 29 pentachlorophenol is not entirely clear from the text or accompanying table, but presumably they 30 do not. The RR was 1.002 (95% CI = 0.991-1.013) for all cancer mortality per 1 ppb-year 31 increase in cumulative TCDD exposure was not statistically significant. Except for soft tissue This document is a draftfor review purposes only and does not constitute Agency policy. 2-31 DRAFT--DO NOT CITE OR QUOTE 1 sarcomas, no statistically significant exposure-response trends were observed for any cancer site. 2 For soft tissue sarcoma, analyses were based on only four deaths. 3 4 2.4.1.1.1.1.5.2. Study evaluation. 5 A key limitation of this study is that SMRs were not derived for different periods of 6 latency for the external comparison group analysis. The original publication on the NIOSH 7 cohort found that SMRs increased when a 20-year latency period was incorporated (Fingerhut 8 et al., 1991, 197375), and similar patterns have been observed in other occupational cohorts 9 (Manz et al., 1991, 199061; Ott and Zober, 1996, 198101) and among Seveso residents 10 (Consonni et al., 2008, 524825). Additionally, dose-response analyses showed marked increases 11 in slopes with a 15-year latency period (Cheng et al., 2006, 523122; Steenland and Deddens, 12 2003, 198587). In this context, the absence of an elevated SMR for cancer mortality is 13 consistent with previous findings of the NIOSH cohort. While the cohort did have sufficient 14 follow-up, no evaluation of possible latent effects was presented and this is a major limitation of 15 this study. Further, the evaluation of the exposure metrics should be expanded from what was 16 presented in Collins et al. (2009, 197627) due to the previous analyses of the same workers 17 finding positive associations between cancer mortality and TCDD (Steenland et al., 2001, 18 197433). 19 Unfortunately, the Collins et al. (2009, 197627) study did not include a categorical 20 analysis of TCDD exposure and cancer mortality. This categorical analysis would have enabled 21 an evaluation of whether a nonlinear association exists between TCDD exposure and cancer risk. 22 The analyses of both Cheng et al. (2006, 523122) and Steenland et al. (2001, 197433) suggest an 23 attenuation of effects at higher doses, and several investigations have considered log-transformed 24 associations as a means to address nonlinearity. Also, the earlier plant-specific dose-response 25 analyses of Steenland et al. (2001, 197433) are not consistent with the findings for the Midland 26 plant that Collins et al. (2009, 197627) presented. These differences could be due to differences 27 in the construction of exposure metrics, additional follow-up, or lagging of exposures. 28 29 2.4.1.1.1.1.5.3. Suitability o f da ta f o r dose-response modeling. 30 The Collins et al. (2009, 197627) study uses serum levels to derive TCDD exposure 31 estimates and does not appear to be subject to important biases. The reliance on data from one This document is a draftfor review purposes only and does not constitute Agency policy. 2-32 DRAFT--DO NOT CITE OR QUOTE 1 plant offers some advantages over the multiplant analyses, as heterogeneity in exposure to other 2 occupational agents would be lower. The number of individuals who provided serum samples 3 (n = 280) is greater than the 170 individuals used to derive TCDD estimates for the NIOSH 4 cohort. The authors found a statistically signficant dose-response trend for soft tissue sarcoma 5 mortality and TCDD exposures. Therefore, this study is considered for quantitative 6 dose-response analysis. 7 8 2.4.1.1.1.2. T he B A S F cohort. 9 In 1953, dioxin contamination occurred as a result of an autoclave accident during the 10 production of trichlorophenol at the BASF plant in Ludwigshafen, Germany. A second dioxin 11 incident occurred in 1988 that was attributed to the blending of thermoplastic polyesters with 12 brominated flame retardants. Of the two events, the one on November 13, 1953, was associated 13 with more severe acute health effects, including chloracne that resulted in immediate 14 hospitalizations for seven workers. These adverse events were not linked to TCDD until 1957 15 when TCDD was identified as a byproduct of the production of trichlorophenol and was shown 16 to induce chloracne (Zober et al., 1994, 197572). Zober and colleagues (1998, 594300) noted 17 that with the 1988 accident, affected individuals did not exhibit clinical symptoms or chloracne, 18 but rather were identified through "analytical measures." In both instances, efforts were made to 19 limit the potential for exposure to employees. 20 21 2.4.1.1.1.2.1. Thiess and Frentzel-Beyme (1977, 594302) and Thiess et al. (1982, 064999). 22 2.4.1.1.1.2.1.1. Study summary. 23 A study of the mortality of workers employed at the BASF plant was first presented in 24 1977 (Thiess and Frentzel-Beyme, 1977, 594302) with subsequent updates in both 1982 (Thiess 25 et al., 1982, 064999), and in 1990 (Zober et al., 1990, 197604). In the first published paper 26 (Thiess et al., 1982, 064999), 74 employees involved in the 1953 accident were traced and their 27 death certificate information extracted. Of these, 66 suffered chloracne or severe dermatitis. 28 Observed deaths were compared to the expected number using three external reference groups: 29 the town of Ludwigshafen (n = 180,000), the district of Rhinehessia-Palatinate (n = 1.8 million), 30 and the Federal Republic of Germany (n = 60.5 million). Another comparison group was 31 assembled by selecting age-matched employees taken from other cohorts under study. This This document is a draftfor review purposes only and does not constitute Agency policy. 2-33 DRAFT--DO NOT CITE OR QUOTE 1 additional comparison was aimed at avoiding potential biases associated with healthy worker 2 effect when using an external referent. 3 During a follow-up interval of up to 26 years (1953-1979), 21 individuals died. Of 4 these, seven deaths were from cancer. The expected number of cancer deaths derived for the 5 three external comparison groups ranged between 4.1 and 4.2, producing an SMR of 1.7 6 (p-values ranged between 0.12 and 0.14). Excess mortality was found for stomach cancer based 7 on the external comparisons (p < 0.05); however, this was based on only three cases. No other 8 statistically significant excesses were found with the external comparisons made to the other 9 cohorts of workers. 10 11 2.4.1.1.1.2.1.2. Study evaluation. 12 In the Thiess et al. (1982, 064999) study, no TCDD exposures were derived for the 13 workers, thus no dose-reconstruction was performed. The findings from this study are limited by 14 the small size of the cohort. The 74 workers followed in this cohort represent the smallest 15 number of workers across the occupational cohorts (Becher et al., 1998, 197173; Fingerhut et al., 16 1991, 197375; Hooiveld et al., 1998, 197829; McBride, 2009, 198490; McBride et al., 2009, 17 197296; Michalek and Pavuk, 2008, 199573; Steenland et al., 2001, 197433) that have 18 investigated TCDD exposures and cancer mortality. Mechanisms of follow-up were excellent as 19 all individuals were traced, and death certificates were obtained from all deceased workers. 20 Although the study does compare the mortality experience to other occupational cohorts, 21 the paper provides insufficient information to adequately interpret the associated findings. For 22 example, a description of these occupations is lacking making it impossible to determine whether 23 these cohorts were exposed to other occupational carcinogens that might have confounded the 24 associations between TCDD exposure and cancer mortality. 25 26 2.4.1.1.1.2.1.3. Suitability o f datafor TCDD dose-response modeling. 27 Subsequent data assembled for the BASF cohort provide more detailed exposure 28 characterization and also include information for 243 male workers employed at the plant. As 29 such, this study did not meet the considerations for further dose-response analysis. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-34 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.2.2. Zober et al. (1990, 197604). 2 2.4.1.1.1.2.2.1. Study summary. 3 Zober et al. (1990, 197604) also examined the mortality patterns of 247 individuals 4 involved in the 1953 accident at the BASF plant. As detailed in their paper, the size of the 5 original cohort was expanded by efforts to locate all individuals who were exposed in the 6 accident or during the clean-up. Three approaches were followed in assembling the cohort. 7 Sixty-nine cohort members were identified from the company physician's list of employees 8 exposed as a result of the accident (Subcohort C1). Sixty-six of these workers were included in 9 the original study population of workers Thiess et al. (1982, 064999) examined. 10 Eighty-four other workers who were potentially exposed to TCDD due to their involvement in 11 demolitions or operations were added to the cohort. This group included 43 firemen, 18 plant 12 workers, 7 bricklayers, 5 whitewashers, 4 mechanics, 2 roofers, and 5 individuals in other 13 occupations (Subcohort C2). The cohort was further augmented through the Dioxin 14 Investigation Program, which sought to locate those who were involved in the 1953 accident and 15 were still alive in 1986. Current and former workers enrolled in the study were asked to identify 16 other current or former coworkers (including deceased or retired) who might have been exposed 17 from the accident. This third component of 94 workers (Subcohort C3) included 27 plant 18 workers, 16 plumbers, 10 scaffolders, 10 professionals, 7 mechanics, 6 transportation workers, 19 5 bricklayers, 5 laboratory assistant, 3 insulators, and 5 individuals in other occupations. A 20 medical examination was performed for those identified through the Dioxin Investigation 21 Program, and blood measures were obtained for 28 of these workers. 22 External comparisons of the workers' mortality experience to the general population of 23 the Federal Republic of West Germany were made using SMRs. Person-years were tabulated 24 across strata defined by calendar period, sex, and age group. Sixty-nine deaths including 25 twenty-three from cancer were detected among the workers during the 34-year follow-up period 26 (November 17, 1953 through December 31, 1987). Cause-specific death rates for these same 27 strata were available for the Federal Republic of West Germany. Stratified analyses were 28 conducted to examine variations in the SMRs according to years since first exposure (0-9, 29 10-19, and >20 years) for each of the three subcohorts, as well as 114 workers with chloracne. 30 Although it was consistent in magnitude with findings from the NIOSH cohort, a 31 statistically significant SMR for all cancer mortality was not observed (SMR = 1.17, This document is a draftfor review purposes only and does not constitute Agency policy. 2-35 DRAFT--DO NOT CITE OR QUOTE 1 90% CI = 0.80-1.66). The SMRs for each of the three subcohorts varied substantially. For 2 Subcohorts C1, C2, and C3, the SMRs were 1.30 (90% CI = 0.68-2.26), 1.71 3 (90% CI = 0.96-2.83), and 0.48 (90% CI = 0.13-1.23), respectively. The SMRs increased 4 dramatically when analyses were restricted to those with 20 or more years since first exposure in 5 Subcohort C1 (SMR = 1.67, 90% CI = 0.78-3.13) and Subcohort C2 (SMR = 2.38, 6 90% CI = 1.18-4.29). Meanwhile, in a subgroup analysis of those with chloracne, for the period 7 of 20 or more years after first exposure, a statistically significant excess in cancer mortality was 8 noted (SMR = 2.01; 90% CI = 1.22-3.15). 9 10 2.4.1.1.1.2.2.2. Study evaluation. 11 An important limitation of the study is the manner in which the cohort was constructed. 12 Subcohort C3 was constructed by identifying individuals who were alive in 1986. This resulted 13 in 97 active and retired employees who participated in the program, with 94 included in the 14 analysis. Although these individuals did identify other workers who might have also retired or 15 died, inevitably, some individuals who had died were not included in the cohort. This would 16 serve to underestimate the SMRs that were generated with external comparisons to the German 17 population. Indeed, cancer mortality rates in this subcohort were about half of what would have 18 been expected based on general population rates (SMR = 0.48, 90% CI = 0.13-1.23). 19 Additionally, more than half of Subcohort C2 were firemen (43 of 84), who would likely have 20 been exposed to other carcinogens as a consequence of their employment. Quantitative analyses 21 of epidemiologic data for firefighters have demonstrated increased cancer risk for several 22 different forms of cancer (Youakim, 2006, 197295). Therefore, potential confounding from 23 other occupational exposures of the firefighters could have contributed to the higher SMR in 24 Subcohort C2 cohort and is a concern. Data on cigarette smoking were not available either. No 25 excess for nonmalignant respiratory disease was found, however, suggesting this might not be an 26 important source of bias. 27 28 2.4.1.1.1.2.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 29 As with the Thiess et al. (1982, 064999) publication, worker exposure was not estimated. 30 Lack of exposure estimates precludes a quantitative dose-response analysis using these data. 31 Also, the study design is not well suited to characterization of risk using the SMR statistic. This document is a draftfor review purposes only and does not constitute Agency policy. 2-36 DRAFT--DO NOT CITE OR QUOTE 1 Mortality is also likely under-ascertained in the large component of the cohort that was 2 constructed through the identification of surviving members of the cohort. 3 4 2.4.1.1.1.2.3. Ott and Zober (1996, 198101). 5 2.4.1.1.1.2.3.1. Study summary. 6 Ott and Zober (1996, 198101) extended the analyses of the BASF cohort to include 7 estimates of individual-level measures of TCDD. The researchers also investigated associations 8 with cancer mortality and identified incident cancer cases. The cohort follow-up period of 9 39 years extended until December 31, 1992, adding 5 years to a previous study (Zober et al., 10 1990, 197604). Ott and Zober (1996, 198101) identified incident cases of cancer using 11 occupational medical records, death certificates, doctor's letters, necropsy reports, and 12 information from self-reported surveys sent to all surviving cohort members. Self-reported 13 cancer diagnoses were confirmed by contacting the attending physician. 14 This study characterized exposure by two methods: (1) determining chloracne status of 15 the cohort members and (2) estimating cumulative TCDD (pg/kg) levels. In 1989, serum 16 measures were sought for all surviving members of the 1953 accident, and serum TCDD levels 17 were quantified for 138 individuals. These serum levels were used to estimate cumulative 18 TCDD concentrations for all 254 members of the accident cohort. Ott et al. (1993, 594322) 19 published a description of the exposure estimation procedure, which was a regression model that 20 accounted for the circumstances and duration of individual exposure. The average internal 21 half-life of TCDD was estimated to be 5.8 years based on repeated serum sampling of 22 29 individuals. The regression model allowed for this half-life to vary according to the 23 percentage of body fat, and yielded half-lives of 5.1 and 8.9 years among those with 20% and 24 30% body fat, respectively. Previous analyses of this cohort had used a half-life of 7.0 years (Ott 25 et al., 1993, 594322). 26 TCDD half-life has been reported to increase with percentage of body fat in both 27 laboratory mammals (Geyer et al., 1990, 197700) and humans (Zober and Papke, 1993, 197602). 28 Ott and Zober (1996, 198101) contend that observed correlations with chloracne severity and 29 cumulative estimates of TCDD exposure indirectly validated this exposure metric. Specifically, 30 the mean TCDD concentration for those without chloracne was 38.4 ppt; for those with moderate 31 and severe forms of chloracne, the mean was 420.8 ppt and 1,008 ppt, respectively. This document is a draftfor review purposes only and does not constitute Agency policy. 2-37 DRAFT--DO NOT CITE OR QUOTE 1 Unlike for the NIOSH cohort, individual-level data were collected for other cancer risk 2 factors. These factors included body mass index at time of first exposure, history of 3 occupational exposure to P-naphthylamine and asbestos, and history of smoking. Smoking data 4 were available for 86% of the cohort. SMRs were based on the external referent population of 5 West Germany. For cancer incidence, Ott and Zober (1996, 198101) generated standardized 6 incidence ratios (SIRs) using incidence rates for the state of Saarland (1970-1991) as the 7 external referent. They calculated SMRs (and SIRs) for three categories of cumulative TCDD 8 levels: <0.1 pg/kg, 0.1-0.99 pg/kg and >1 pg/kg. The Cox regression model was used to 9 characterize risk within the cohort using a continuous measure of TCDD. These analyses 10 considered the potential confounding influence of age, smoking, and body mass index using a 11 stepwise regression modeling approach. The Cox modeling employed a stratified approach 12 using the date of first exposure to minimize possible confounding between calendar period and 13 exposure. The three first exposure groups were exposure within the first year of the accident, 14 exposure between 1 year after the accident and before 1960, and exposure after 1959. The Cox 15 regression estimates were presented in terms of conditional risk ratios (i.e., hazard ratios adjusted 16 for body mass index, smoking and age). 17 Although no statistically significant excesses relative to the general population were 18 detected for all cancer mortality, there was some suggestion of an exposure-response 19 relationship. In the 0.1-0.99 pg/kg and >1 pg/kg exposure groups, the all cancer SMRs were 1.2 20 (95% CI = 0.5-2.3) and 1.6 (95% CI = 0.9-2.6), respectively. Higher SMRs for cancer (all sites 21 combined) were also found with an increased interval since exposure first occurred. 22 Specifically, when observed versus expected counts of cancer were compared in the time interval 23 20 years after first exposure, the SMR in the highest exposure group (>1 pg/kg) was 1.97 24 (95% CI = 1.05-5.36). An excess in lung cancer also was noted with the same lag in this 25 exposure group (SMR = 3.06, 95% CI = 1.12-6.66). For cancer incidence, a statistically 26 significant increased SIR for lung cancer was observed in the highest exposure category 27 (SIR = 2.2, 95% CI = 1.0-4.3), but no other statistically significant associations were detected 28 for any other cancer site. No cases of soft-tissue sarcoma were found among the cohort members 29 in this analysis. 30 Based on internal cohort comparisons, Cox regression models also were used to generate 31 hazard ratios as measures of relative risk for TCDD exposures following adjustment for This document is a draftfor review purposes only and does not constitute Agency policy. 2-38 DRAFT--DO NOT CITE OR QUOTE 1 smoking, age and body mass index. A statistically significant association between TCDD dose 2 (per pg/kg) and cancer mortality was detected (RR = 1.22, 95% CI = 1.00-1.50), but not for 3 cancer incidence (RR = 1.11, 95% CI = 0.91-1.35). Statistically significant findings were 4 observed for stomach cancer mortality (RR = 1.46, 95% CI = 1.13-1.89) and incidence 5 (RR = 1.39, 95% CI = 1.07-1.69). 6 The Ott and Zober (1996, 198101) study also compared the relationship between TCDD 7 exposure categories and cancer mortality from all sites combined according to smoking status. 8 Associations were noted between increased exposure to TCDD and mortality from cancer among 9 smokers, but not among nonsmokers or former smokers. 10 11 2.4.1.1.1.2.3.2. Study evaluation. 12 The Ott and Zober (1996, 198101) study characterizes exposure to TCDD at an 13 individual level. Therefore, unlike in past studies involving this cohort, these data can provide 14 an opportunity for conducting quantitative dose-response modeling. As with the more recent 15 studies involving the NIOSH cohort, serum samples were obtained from surviving cohort 16 members and then used to back-extrapolate TCDD values for all cohort members. In the BASF 17 cohort, however, serum data were available for a much higher percentage of cohort members 18 (54%) than in the NIOSH cohort (5%). An additional study strength was the collection of 19 questionnaire data, which allowed for the potential confounding from cigarette smoking and 20 body mass index to be examined. 21 The Ott and Zober (1996, 198101) study also evaluates the relationship between TCDD 22 and cancer incidence. Most cohort studies of TCDD-exposed workers have relied solely on 23 mortality outcomes. The availability of incidence data better allows for period of latency to be 24 described, and moreover, to characterize risks associated with cancers that typically have long 25 survival periods. The authors provide few details on the expected completeness of ascertainment 26 for incident cancer cases, which makes determining any associated bias difficult. They do, 27 however, suggest that nonfatal cancers are more likely to have been missed in the earlier part of 28 the follow-up. The net result of differential case ascertainment over time makes evaluating 29 differences in risk estimates across different periods of latency impossible. 30 The small sample size of the cohort (n = 243 men) likely limited the statistical power to 31 detect small associations for some of the exposure measures. This also effectively limited the This document is a draftfor review purposes only and does not constitute Agency policy. 2-39 DRAFT--DO NOT CITE OR QUOTE 1 ability to analyze dose-response relationships quantitatively, particularly across strata such as 2 time since exposure. For site-specific analyses, the cancer site with the most cancer deaths was 3 the respiratory system (n = 11). Thus, quantitative dose-response analysis using these cohort 4 data would be limited to the evaluation of all cancer sites combined. 5 The most important limitation of this study is related to the construction of the 6 third component of the cohort. As mentioned earlier, this cohort was assembled by actively 7 seeking out surviving members of the cohort in the mid-1980s. The mortality experience of this 8 cohort is much lower than that of the general population over the entire follow-up, a result that is 9 expected given that the individuals were known to be alive as of 1986. The net result is likely an 10 underestimate of the SMR. 11 12 2.4.1.1.1.2.3.3. Suitability o f datafor TCDD dose-response modeling. 13 This study was included in the quantitative dose-response modeling for the 14 2003 Reassessment (U.S. EPA, 2003, 537122). The characterization of exposure data and 15 availability of other risk factor data at an individual level are appropriate for use in quantitative 16 dose-response analyses. 17 18 2.4.1.1.1.3. T h e H a m b u r g coh ort. 19 The Hamburg cohort has been the subject of several cancer risk assessments. As with the 20 NIOSH and BASF cohorts, analyses have progressed from basic comparisons of mortality 21 experience to general population rates to more sophisticated internal cohort analyses involving 22 the reconstruction of TCDD exposures using serum measures. This cohort consists of 23 approximately 1,600 workers who were employed in the production of herbicides at a plant in 24 Hamburg, Germany during 1950-1984 (Becher et al., 1998, 197173; Flesch-Janys et al., 1995, 25 197261). The herbicides produced included 2,4,5-T, P-hexachlorocyclohexane and lindane. The 26 production of TCP and 2,4,5-T was halted in 1954 following a chloracne outbreak. The plant 27 ceased operations in 1984. Approximately 20 different working areas were identified, which, in 28 turn, were grouped into five main areas based on putative TCDD exposure levels. One working 29 area was deemed to be extremely contaminated, having TCDD exposures at least 20-fold higher 30 than in other areas. In this section, the studies undertaken in this cohort that have examined 31 cancer mortality are summarized. This document is a draftfor review purposes only and does not constitute Agency policy. 2-40 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.3.1. Manz et al. (1991, 199061). 2 2.4.1.1.1.3.1.1. Study summary. 3 Manz et al. (1991, 199061) investigated patterns of mortality in the Hamburg cohort. 4 The study population consisted of 1,583 workers (1,184 men, 399 women) who were employed 5 for at least three months between 1952 and 1989. Casual workers were excluded as they lack 6 sufficient personal identifying information thereby not allowing for associations with mortality 7 outcomes to be examined. Vital status was determined using community-based registries of 8 inhabitants throughout West Germany. Cause of death until the end of 1989 was determined 9 from medical records for all cancer deaths and classified based on the ninth revision of the 10 International Classification of Diseases (WHO, 1978, 594329). Although Manz et al. (1991, 11 199061) present some data on cancer incidence for the cohort, the data are incomplete as 12 information was available on only 12 cases; 93 cancer deaths were observed in the cohort. 13 In this study, the authors used information on production processes to group workers into 14 categories of low, medium, or high exposure to TCDD. This information was based on TCDD 15 concentrations in precursor materials, products, waste, and soil from the plant grounds, measured 16 after the plant closed in 1984. The distribution of workers into the low, medium, and high 17 exposure groups was 186, 901, and 496, respectively. The authors examined the validity of the 18 three exposure categories using a separate group of 48 workers who provided adipose tissue 19 samples. The median exposure of the 37 volunteers in the high group was 137 and 60 ng/kg in 20 the remaining 11. Information about chloracne in the cohort was incomplete, and, therefore, was 21 not used as a marker of TCDD exposure. Other surrogate measures of exposure were considered 22 in this study, including duration of exposure and year of first employment. For the latter 23 measure, employment that began after 1954 was assumed to result in much lower exposures 24 given that production of 2,4,5-T and TCP stopped in 1954. 25 External comparisons of cancer mortality were made by calculating SMRs using the 26 general population of West Germany as a referent. Comparisons of mortality in the cohort also 27 were made to a separate cohort of 3,417 gas supply workers to avoid bias from a healthy worker 28 effect. Vital status and cause of death in the gas supply workers were determined using the same 29 methods as used in the Hamburg cohort. SMRs were calculated relative to both referent 30 populations (West Germany and gas supply workers) across low, medium, and high TCDD 31 exposure groups. The comparison of mortality to the gas supply workers, however, extended This document is a draftfor review purposes only and does not constitute Agency policy. 2-41 DRAFT--DO NOT CITE OR QUOTE 1 only until the end of 1985, whereas, comparisons to the general population extended until 1989. 2 Stratified analyses were undertaken to calculate SMRs for each of the three exposure groups for 3 categories of duration of employment (<20 versus >20 years) and date of entry into the cohort 4 (<1954 vs. >1954). 5 When compared to the general population, overall cancer mortality was elevated in male 6 cohort members (SMR = 1.24, 95% CI = 1.00-1.52) but not in females (SMR = 0.80, 7 95% CI = 0.60-1.05). A two-fold increase in female breast cancer mortality was noted although 8 it did not achieve statistical significance at the alpha level of 0.05 (SMR = 2.15, 9 95% CI = 0.98-4.09). The SMR among men was further increased when analyses were 10 restricted to workers who were employed for at least 20 years (SMR = 1.87, 11 95% CI = 1.11-2.95). Analyses restricted to those in the highest exposure group produced an 12 even higher SMR for those with at least 20 years of employment (SMR = 2.54, 13 95% CI = 1.10-5.00). Statistically significant excesses in risk were detected among those who 14 first worked before 1954, but not afterward. Furthermore, a dose-response trend was observed 15 across increasing exposure categories in the subset of workers employed before 1954. The 16 SMRs using the cohort of gas supply workers as the referent group for the low, medium, and 17 high groups in this subset were 1.41 (95% CI = 0.46-3.28), 1.61 (95% CI = 1.10-2.44), and 2.77 18 (95% CI = 1.59-4.53), respectively. This finding is consistent with what was known about 19 TCDD exposures levels at the plant, namely, that TCDD concentrations were much higher 20 between 1951 and 1954, with subsequent declining levels after 1954. 21 Generally speaking, patterns of excess mortality were similar when the cohort of gas 22 workers was used as a reference group. The overall SMR for men was 1.39 23 (95% CI = 1.10-1.75); and was 1.82 (95% CI = 0.97-3.11) when analyses were restricted to 24 workers with 20 or more years of employment. A dose-response trend also was observed across 25 exposure categories when analyses were restricted to those employed for at least 20 years. In 26 particular, with these analyses, no cancer deaths were observed among those in the lowest 27 exposure group, while the SMRs in the middle and high exposure groups were 1.36 28 (95% CI = 0.50-2.96) and 3.07 (95% CI = 1.24-6.33). 29 SMRs also were generated for several site-specific cancers relative to the West German 30 general population and the gas worker cohort. No statistically significant excesses were 31 observed using the general population reference. In contrast, statistically significant excesses This document is a draftfor review purposes only and does not constitute Agency policy. 2-42 DRAFT--DO NOT CITE OR QUOTE 1 were observed for lung cancer (SMR = 1.67, 95% CI = 1.09-2.44) and hematopoietic system 2 cancer (SMR = 2.65, 95% CI = 1.21-5.03) relative to the gas workers cohort. 3 4 2.4.1.1.1.3.1.2. Study evaluation. 5 The Manz et al. (1991, 199061) findings indicate an excess of all cancer mortality among 6 the workers with the highest exposures, particularly those who worked for at least 20 years and 7 were employed before 1954. The findings across categories of exposure within the subsets of 8 workers employed for at least 20 years and before 1954, particularly using the cohort of gas 9 supply workers, are consistent with a dose-response relationship. These elevated cancer 10 mortality rates found among those employed before 1954 were likely due to higher TCDD 11 exposures. Other carcinogenic coexposures, such as benzene, asbestos, and dimethyl sulfate, 12 could have occurred among this population. Given that no substantial changes in the production 13 processes at the Hamburg plant occurred after 1954, comparable levels of these coexposures 14 would be expected before and after 1954. Exposures to these other chemicals varied across 15 different departments/groups; therefore, confounding was unlikely since a strong association 16 between concentrations of these chemicals and TCDD exposures was not evident. No 17 information, however, was presented on potential exposure to other dioxin-like compounds 18 which may confound the associations that were detected. 19 Detailed information on workers' smoking behaviors was not collected. Limited 20 evidence indicated, however, that smoking prevalence between the Hamburg cohort and the gas 21 supply workers cohort was quite similar. A nonrepresentative sample of 361 workers in the 22 Hamburg cohort and the sample of 2,860 workers in the gas supply cohort indicated that the 23 self-reported smoking prevalence was 73% and 76%, respectively. This suggests that the 24 two cohorts are comprised predominantly of smokers. The similarity in overall smoking 25 prevalence indicates that comparisons of cancer mortality between the two groups are not unduly 26 influenced by an inability to adjust for smoking. 27 28 2.4.1.1.1.3.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 29 The data compiled for the Manz et al. (1991, 199061) study do satisfy many of the 30 considerations for conducting quantitative dose-response analysis; health outcomes appear to be 31 ascertained in an unbiased manner, and exposure was characterized on an individual-level basis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-43 DRAFT--DO NOT CITE OR QUOTE 1 However, as demonstrated in later studies, there was a large dioxin-like compound component 2 that was not quantified or assessed in this study. Dose-response associations between TCDD and 3 cancer mortality were detected, with stronger associations observed with increased periods of 4 latency and for those who first worked when TCDD was at higher levels. 5 The size of the cohort, although not as large as the NIOSH cohort, does offer sufficient 6 statistical power to evaluate TCDD-related risk for cancers from all cancer sites. The data are 7 limited, however, for characterizing cancer risks among women; only 20 cancer deaths occurred 8 in the 399 women included in the cohort. It is unlikely that the findings are biased by 9 confounding due to cigarette smoking since dose-response patterns were strengthened when 10 comparisons were made to the cohort of gas supply workers rather the general population 11 referent where smoking rates were likely lower. The inability to account for other occupational 12 exposure when TCDD exposures were much higher (pre-1955) could result in confounding if 13 these other exposures were related to TCDD and the health outcomes under consideration. This 14 data set would be suitable for quantitative dose-response modeling if the exposure 15 characterization of the cohort could be improved using biological measures of dose. 16 17 2.4.1.1.1.3.2. Flesch-Janys et al. (1995, 197261). 18 2.4.1.1.1.3.2.1. Study summary. 19 In 1995, Flesch-Janys et al. (1995, 197261) published an analysis of the male employees 20 from the Hamburg cohort that extended the follow-up to 40 years (1952-1992). Inclusion of 21 these three additional years of follow-up resulted in a sample size of 1,189 male workers. 22 The authors estimated a quantitative exposure variable for concentrations of TCDD in 23 blood at the end of exposure (i.e., when employment in a department ended) and above German 24 median background TCDD levels. The TCDD exposure assessment defined 14 production 25 departments according to TCDD levels in various products in the plant, in waste products, and in 26 various buildings. The time (in years) each worker spent in each department then was 27 calculated. Concentrations of TCDD were determined in 190 male workers using serum 28 (n = 142) and adipose tissue samples (n = 48). The authors used a first-order kinetic model to 29 calculate TCDD levels at the end of exposure for the 190 workers with available polychlorinated 30 dibenzo-p-dioxin (PCDD) and -furan (PCDF) at various time points. Half-lives were calculated 31 from an elimination study of 48 workers from this cohort, and the median TCDD background This document is a draftfor review purposes only and does not constitute Agency policy. 2-44 DRAFT--DO NOT CITE OR QUOTE 1 level was estimated at 3.4 ng/kg blood fat from the German population (Flesch-Janys et al., 2 1994, 197372; Papke et al., 1994, 198279). Using the one-compartment, first-order kinetic 3 model, the half-life of TCDD was estimated to be 6.9 years (Flesch-Janys, 1997, 197305). 4 Increased age and higher body fat percentage were associated with increased TCDD half-life, 5 while smoking was associated with a higher decay rate for most of the congeners examined 6 (Flesch-Janys et al., 1996, 197351). Cumulative TCDD exposures were estimated by summing 7 exposures over the time spent in all production departments and were expressed in terms of 8 ng/kg of blood fat. The authors also applied a metric of total toxicity equivalence (TOTTEQ) as 9 the weighted sum of all congeners where weights were TEQs that denoted the toxicity of each 10 congener relative to TCDD. 11 Similar to previous analyses on this cohort, comparisons were made using an external 12 referent group of workers from a gas supply company (Manz et al., 1991, 199061). In contrast to 13 previous analyses where SMR statistics were generated using this "external" reference, however, 14 Flesch-Janys et al. (1995, 197261) used Cox regression. The Cox regression models treated the 15 gas worker cohort as the referent group, and six exposure groups were defined by serum-derived 16 cumulative TCDD estimates. The groups were determined by using the first four quintiles with 17 the upper two exposure categories corresponding to the ninth and tenth deciles of the cumulative 18 TCDD. Internal cohort comparisons used those workers in the lowest quintile as the referent 19 group, as opposed to the cohort of gas workers. A similar approach was used to model TEQs. 20 No known TCDD exposures occurred in the gas workers, so they were assigned exposures based 21 on the median background levels in the general population. RRs were calculated based on 22 exposure above background levels; in other words, background levels were assumed to be 23 equivalent across all workers and also for those employed by the gas supply company. The RRs 24 derived using the Cox model were adjusted for total duration of employment, age, and year when 25 employment began. 26 The Cox regression with the cohort of gas workers as the referent exposure group yielded 27 a linear dose-response relationship between cumulative TCDD exposure and cancer mortality for 28 all sites combined (p < 0.01). The RRs for all-cancer mortality were 1.59, 1.29, 1.66, 1.60, 1.70, 29 and 3.30. For four of the six categories (excluding the referent group), the RRs were statistically 30 significant (p < 0.05); in the highest TCDD exposure category (344.7-3,890.2 ng/kg) the RR 31 was 3.30 (95% CI = 2.05-5.31). Similar findings were evident with TOTTEQ. A dose-response This document is a draftfor review purposes only and does not constitute Agency policy. 2-45 DRAFT--DO NOT CITE OR QUOTE 1 pattern for all cancer mortality (p < 0.01) based on the internal cohort comparisons was also 2 detected. 3 The authors performed an additional analysis to evaluate the potential confounding role 4 of dimethylsulfate. Although no direct measures of dimethylsulfate were available, the 5 investigators repeated analyses by excluding 149 workers who were employed in the department 6 where dimethylsulfate was present. A dose-response pattern persisted for TCDD (p < 0.01), and 7 those in the highest exposure group (344.7-3,890.2 ng/kg of blood fat) had a RR of 2.28 8 (95% CI = 1.14-4.59). 9 10 2.4.1.1.1.3.2.2. Study evaluation. 11 The Flesch-Janys et al. (1995, 197261) study used serum-based measures to determine 12 cumulative exposure to TCDD at the end of employment for all cohort members. They used the 13 standard one-compartment, first-order kinetic model and samples obtained from 190 male 14 workers. This quantitative measure of exposure permits an estimation of a dose-response 15 relationship. 16 Confounding for other occupational exposures is unlikely to have biased the results. A 17 dose-response relationship persisted after excluding workers exposed to dimethylsulfate. Other 18 potential exposures of interest included benzene and isomers of hexachlorocyclohexane. 19 Exposure to these agents, however, was highest in the hexachlorocyclohexane and lindane 20 department, where TCDD exposures were lower. Confounding was unlikely due to exposure to 21 these chemicals, since a strong association between concentrations of these chemicals and TCDD 22 exposures was not evident (due to considerable variability in concentrations across different 23 departments/groups). As outlined earlier, the study findings are unlikely to be biased for 24 cigarette smoking as cigarette smoking in the cohort was similar to that in the comparison 25 population. Moreover, more recent analyses of serum-based TCDD exposure measures found no 26 correlation with smoking status in this cohort (Flesch-Janys et al., 1995, 197261)--a necessary 27 condition for confounding. 28 The authors used an exposure metric that described cumulative TCDD exposure of 29 workers at the time they were last exposed. As a result, the authors were unable to characterize 30 risks associated with this metric for different periods of latency despite a sufficient follow-up This document is a draftfor review purposes only and does not constitute Agency policy. 2-46 DRAFT--DO NOT CITE OR QUOTE 1 period. Subsequent analyses constructed time-dependent measures of cumulative TCDD and 2 accounted for excretion of TCDD during follow-up. 3 In contrast to most risk assessments of TCDD exposure, this study modeled the 4 relationship between other dioxin-like compounds and the risk of cancer mortality using the 5 TOTTEQ metric. 6 7 2.4.1.1.1.3.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 8 The data used in this study satisfy most of the considerations developed for performing a 9 quantitative dose-response analysis. However, latency period was not examined in this study. 10 Dose-response analyses were, therefore, limited to a subsequent study of this cohort (Becher 11 et al., 1998, 197173), which did examine latency. 12 13 2.4.1.1.1.3.3. Flesch-Janys et al. (1998, 197339). 14 2.4.1.1.1.3.3.1. Study summary. 15 Flesch-Janys et al. (1998, 197339) undertook another analysis on this cohort that 16 incorporated additional sera data for 275 workers (39 females and 236 males). The follow-up 17 period was the same as that used in the 1995 analyses, with mortality follow-up extending until 18 December 31, 1992. Analyses were based on 1,189 males who were employed for at least 19 3 months from January 1, 1952 onward. The authors continued this dose-response analysis to 20 address limitations in their previous work. One limitation was that the previous method did not 21 account for the elimination of TCDD while exposures were being accrued during follow-up. A 22 second limitation was that the amount of time workers spent in different departments was not 23 considered. In the 1998 study, the "area under the curve" approach was used because it accounts 24 for variations in concentrations over time and reflects cumulative exposure to TCDD. The 25 authors used a first-order kinetic model to link blood levels and working histories to derive 26 department-specific dose rates for TCDD. The TCDD background level of 3.4 ng/kg blood fat 27 for the German population was used (Papke et al., 1994, 198279). The dose rates were applied 28 to estimate the concentration of TCDD at every point in time for all cohort members. A 29 cumulative measure expressed as ng/kg blood fat multiplied by years was calculated and used in 30 the SMR analysis. SMRs were calculated using general population mortality rates for the 31 German population between 1952 and 1992. No lag period was incorporated into the derivation This document is a draftfor review purposes only and does not constitute Agency policy. 2-47 DRAFT--DO NOT CITE OR QUOTE 1 of the SMRs. The SMRs were estimated for the entire cohort and for exposure groups based on 2 quartiles obtained from the area under the curve. Linear trend tests were also performed. The 3 overall SMR for cancer mortality in the cohort was 1.41 (95% CI = 1.17-1.68). This SMR value 4 was higher than the SMR of 1.21 reported for this same cohort with 3 fewer years of follow-up 5 (Manz et al., 1991, 199061). In terms of site-specific cancer mortality, excesses were found for 6 respiratory cancer (SMR = 1.71, 95% CI = 1.24-2.29) and rectal cancer (SMR = 2.30, 7 95% CI = 1.05-2.47). Increased risk for lymphatic and hematopoietic cancer (SMR = 2.16, 8 95% CI = 1.11-3.17) were also noted largely attributable (SMR = 3.73, 95% CI = 1.20-8.71) to 9 lymphosarcoma (i.e., non-Hodgkin's lymphoma). A dose-response relationship was observed 10 across quartiles of cumulative TCDD for all-cancer mortality (p < 0.01). The SMRs for these 11 quartiles were 1.24, 1.34, 1.34, and 1.73. Dose-response relationships were not observed for 12 lung cancer or hematopoietic cancers using this same metric. Dose-response relationships were 13 not observed with cumulative TEQ for any of the cancer sites examined (i.e., all cancers, lung 14 cancer, hematopoietic cancer). 15 16 2.4.1.1.1.3.3.2. Study evaluation. 17 The approach used in the Flesch-Janys et al. (1998, 197339) study offers a distinct 18 advantage over earlier analyses involving the same cohort. Three more years of follow-up were 19 available, and the characterization of exposure using the "area under the curve" better captures 20 changes in cumulative exposure using a person-years approach rather than cumulative TCDD at 21 the time of last exposure. As noted previously, other occupational exposures or cigarette 22 smoking are unlikely to have biased the study findings. A sufficient length of follow-up had 23 accrued, and dose-response associations were evident. Dioxin-like compounds were evaluated in 24 this study. For TCDD, the mean concentration was 101.3 ng/kg at the time of measurement. For 25 other higher chlorinated congeners, the corresponding mean (without TCDD) was 89.3 ng/kg. 26 27 2.4.1.1.1.3.3.3. Suitability o f datafor TCDD dose-response modeling. 28 The data used in this study satisfy most of the considerations developed for performing a 29 quantitative dose-response analysis. However, latency was not examined in this study. 30 Dose-response analyses were, therefore, limited to a subsequent study of this cohort (Becher This document is a draftfor review purposes only and does not constitute Agency policy. 2-48 DRAFT--DO NOT CITE OR QUOTE 1 et al., 1998, 197173) which did examine latency and supersedes the Flesch-Janys et al. (1998, 2 197339) study. 3 4 2.4.1.1.1.3.4. Becher et al. (1998, 197173). 5 2.4.1.1.1.3.4.1. Study summary. 6 The Becher et al. (1998, 197173) quantitative cancer risk assessment for the Hamburg 7 cohort was highlighted in the 2003 Reassessment as being appropriate for conducting 8 dose-response analysis. The integrated TCDD concentration over time, as estimated in the 9 Flesch-Janys et al. (1998, 197339) study, was used as the exposure variable. Estimates of the 10 half-life of TCDD based on the sample of 48 individuals with repeated measures were 11 incorporated into the model that back-calculated TCDD exposures to the end of the employment 12 (Flesch-Janys et al., 1996, 197351). This method took into account the age and body fat 13 percentage of the workers. In Becher et al. (1998, 197173), the analysis used the estimate of 14 cumulative dose (integrated dose or area under the curve) as a time-dependent variable. 15 Poisson and Cox regression models were used to characterize dose-response 16 relationships. Both models were applied to internal comparisons where a person-years offset 17 was used and to an external comparison where an offset of expected number of deaths was used. 18 The person-years offset was used to account for varying person-time accrued by workers across 19 exposure categories. The use of the expected number of deaths as an offset allows risks to be 20 described in relation to that expected in the general population. Within each classification cell of 21 deaths and person-years, a continuous value TCDD and TEQ levels based on the geometric mean 22 were entered into the Poisson model. For the Cox model, accumulated dose was estimated based 23 on area under the curve for TCDD, TEQ, TEQ without TCDD, and P-hexachlorocyclohexane. 24 These other coexposure metrics were adjusted for in the Cox regression analyses. Other 25 covariates considered included in the models were year of entry, year of birth, and age at entry 26 into the cohort. A background level of 3.4 ng/kg blood fat for the German population was used 27 (Papke et al., 1994, 198279). A variety of latencies was evaluated (0, 5, 10, 15, and 20 years), 28 and attributable risk and absolute risk were estimated. The unexposed cohort of gas workers was 29 used for most internal analyses. 30 Internal and external comparisons using the Poisson model found positive associations 31 with TCDD exposure and mortality from all cancers combined. The slope associated with the This document is a draftfor review purposes only and does not constitute Agency policy. 2-49 DRAFT--DO NOT CITE OR QUOTE 1 continuous measure of TCDD (pg/kg blood fat x years) for the internal comparison was 0.027 2 (p < 0.001), which decreased to 0.0156 (p = 0.07) after adjusting for age and calendar period. 3 The slope for the external comparison was 0.0163 (p = 0.055); this estimate was not adjusted for 4 other covariates. For TEQ, the slopes based on the internal comparisons were 0.0274 (p < 0.001) 5 in the univariate model and 0.0107 (p = 0.175) in the multivariate model after adjusting for age 6 and calendar period. The external estimate of slope for TEQ was 0.0109 (p = 0.164). Cox 7 regression of TCDD across six exposure categories, with a lag of 0 years, found a statistically 8 significant linear trend (p = 0.03) and those in the upper exposure group had a RR of 2.19 9 (95% CI = 0.76-6.29). These estimates were adjusted for year of entry, age at entry, and 10 duration of employment. A similar pattern was observed with the Cox regression analysis of 11 TEQ; the linear test for trend, however, was not statistically significant at the alpha level of 0.05 12 (p = 0.06). 13 Cox regression models that included both TCDD and TEQ (excluding TCDD) were 14 applied. In this model, the slope (P) for TCDD was 0.0089 (p = 0.058), while the coefficient for 15 TEQ (excluding TCDD) was -0.024 (p = 0.70). This suggests that confounding by other 16 dioxin-like compounds was unlikely and the increased risk of cancer was due to TCDD 17 exposure. For all TEQs combined, the slope was 0.0078 (p = 0.066). 18 The authors used multiple Cox models to evaluate the effect of latency. The slope 19 estimates for both TCDD and TEQ increased dramatically with increasing latency. The slope 20 estimates for TCDD increased from 0.0096 to 0.0160 (p < 0.05) when latency was increased 21 from 0 to 20 years. Similar changes in the TEQ slopes were noted (0.0093 to 0.0157). 22 Evaluations of dose-response curves found that the best-fitting curve was concave in shape, 23 thereby yielding higher risk at low exposure. Differences between the fit of the class of models 24 considered [i.e., RR(x,P) = exp (P log(kx = 1))], however, were small. 25 Attributable risks were generated only for TCDD, as the data suggested no effects with 26 other TEQs. The additional lifetime risk of cancer assuming a daily intake of 1 pg TCDD/kg 27 body weight/day was estimated to range between 0.001 and 0.01. 28 29 2.4.1.1.1.3.4.2. Study evaluation. 30 The Becher et al. (1998, 197173) study represent perhaps the most detailed analyses 31 performed on any cohort to date. The findings were robust, as similar patterns were found with This document is a draftfor review purposes only and does not constitute Agency policy. 2-50 DRAFT--DO NOT CITE OR QUOTE 1 and without using the gas supply worker cohort as the referent group. Exposures to other 2 potential confounding coexposures, such as dioxin-like compounds, were taken into account, and 3 workers with exposure to other carcinogens (e.g., lindane) were excluded. Furthermore, latency 4 was examined in this study, unlike earlier studies of this cohort. 5 6 2.4.1.1.1.3.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 7 This study was included in the quantitative dose-response modeling for the 8 2003 Reassessment (U.S. EPA, 2003, 537122). The data in the Becher et al. (1998, 197173) 9 study are suitable for conducting quantitative dose-response modeling. The exposure data 10 capture cumulative exposure to TCDD as well as exposures to other dioxin-like compounds. 11 The length of the follow-up is sufficient, and the study appears to not be subject to confounding 12 or other types of biases. Therefore, this study is utilized in quantitative dose-response analysis. 13 14 2.4.1.1.1.4. T he S eveso cohort. 15 Several studies have evaluated the morbidity and mortality effects of residents exposed to 16 TCDD following a July 10, 1976, accidental release through an exhaust pipe at a chemical plant 17 in the town of Meda near Seveso, Italy. The released fluid mixture contained 2,4,5-T, sodium 18 trichlorophenate, ethylene glycol, and sodium hydroxide. Vegetation in the area showed 19 immediate signs of damage, and in the days following the accident, residents developed nausea, 20 headaches, eye irritation, and dermal lesions, particularly children. 21 This accident transported TCDD up to 6 km from the plant. Soil samples taken near the 22 plant revealed average levels of TCDD that ranged from 15.5 pg/m2to 580.4 pg/m2in the most 23 contaminated area near the plant (referred to as Zone A) (Bertazzi et al., 2001, 197005). Zone A 24 covered 87 hectares and extended 2,200 m south from the plant. Another, more distant 25 contaminated zone (Zone B) covering 270 hectares also had contaminated soil levels, but the 26 TCDD concentration range was much lower (1.7-4.3 pg/m3). A reference zone (Zone R), which 27 surrounded the two contaminated areas, had lower TCDD soil levels (range: 0.9-1.4 pg/m ) and 28 included approximately 30,000 residents. Following the accident, most residents in Zone A left 29 the area. Although residents in Zone B remained, they were under strict regulations to avoid 30 consuming homegrown products. In total, 736, 4,737, and 31,800 individuals lived in Zones A, 31 B, and R, respectively. Within days of the accident, 3,300 animals (mostly poultry and rabbits) This document is a draftfor review purposes only and does not constitute Agency policy. 2-51 DRAFT--DO NOT CITE OR QUOTE 1 were found dead. Emergency slaughtering was undertaken to prevent TCDD from entering the 2 food chain, and within 2 years more than 80,000 animals had been slaughtered. Mechanisms 3 were put into place for long-term follow-up of these residents. Unlike the other studies based on 4 occupational cohorts, the follow-up of this population allows for risks to be characterized for 5 females. 6 The mortality studies from Seveso published to date have not incorporated serum TCDD 7 levels that were measured in individuals. Needham et al. (1997) describe the collection of serum 8 samples from a sample of the exposed population and control subjects in 1976. In 1988, human 9 exposure to TCDD was assessed by measuring small volumes of serum remaining from medical 10 examinations done in 1976. An examination of these data revealed some of the highest serum 11 TCDD levels ever reported, that the half-life of TCDD in this population was between 7 and 12 8 years, and that half-life varied between women and men. The half-life of TCDD in serum was 13 longer in women (~9 years) than in men (~7 years) (Needham et al., 1994, 200030). In this 14 report, the findings of studies that characterized cancer risks in relation to exposure to TCDD 15 from the 1976 accident are highlighted. These studies include comparisons of cancer mortality 16 rates to the general population based on zone of residence at the time of accident (Bertazzi et al., 17 2001, 197005; Consonni et al., 2008, 524825). More recent work done by Warner et al. (2002, 18 197489) investigated the relationship between serum-based measures of TCDD and breast cancer 19 among participants in the Seveso Women's Health Study (SWHS). 20 21 2.4.1.1.1.4.1. Bertazzi et al. (2001, 197005). 22 2.4.1.1.1.4.1.1. Study summary. 23 Several studies have reported on the mortality experience of Seveso residents. The more 24 recent publications having a longer follow-up of the cohort are evaluated here. In 2001, the 25 findings from a 20-year mortality study of Seveso residents was published (Bertazzi et al., 2001, 26 197005). The Bertazzi et al. (2001, 197005) study was an extension of the 10- and 15-year 27 follow-ups for mortality (Bertazzi et al., 1989, 197013; Bertazzi et al., 1997, 197097; Pesatori 28 et al., 1998, 523076) and the 10-year follow-up for cancer incidence (Bertazzi et al., 1993, 29 192445). 30 In this cohort, TCDD exposures were assigned to the population using a three-level 31 categorical variable representative of the individual's place of residence (Zones A, B, or R) at the This document is a draftfor review purposes only and does not constitute Agency policy. 2-52 DRAFT--DO NOT CITE OR QUOTE 1 time of the accident or when the person first became a resident of the zone, if that was after 2 1976. An external comparison to the province of Lombardy was made by generating rate ratios 3 (RR) using Poisson regression techniques. Person-years of follow-up were tabulated across 4 strata defined by age, zone of residence, duration of residence, gender, calendar time, and 5 number of years that had elapsed since the time of exposure. Mortality rates during the 6 preaccident period also were compared to evaluate potential changes in rates due to the accident 7 and to evaluate whether patterns were consistent before and after the accident. 8 No overall excess in mortality rates from all cancer sites combined was observed in 9 Zones A or B (combined) when compared to the reference population of Lombardy 10 (n = 9 million residents) (RR = 1.0, 95% CI = 0.9-1.2). Analyses of site-specific cancer 11 mortality revealed statistically significant excesses among residents in Zones A or B (combined) 12 for cancer of the rectum (RR = 1.8, 95% CI = 1.0-3.3) and lymphatic and hematopoietic 13 malignancies (RR = 1.7, 95% CI = 1.2-2.5). Lymphatic and hematopoietic malignancies were 14 elevated in women (RR = 1.8, 95% CI = 1.1--3.2) and in men (RR = 1.7, 95% CI = 1.0-2.8). 15 Analyses stratified by the number of years since first exposure (i.e., 1976) revealed 16 higher risk among men with an increased number of years elapsed. Similar to other studies, the 17 RR for all cancers (combined) was 1.3 (95% CI = 1.0-1.7) among men 15-20 years after first 18 exposure. No such increase after 15 years postexposure, however, was noted in women 19 (RR = 0.8, 95% CI = 0.6-1.2). 20 21 2.4.1.1.1.4.1.2. Study evaluation. 22 Ascertainment of mortality appears to be excellent. Vital status was established using 23 similar methods for both the exposed and reference populations. No individual data were 24 collected and, therefore, the possibility that confounding by individual characteristics such as 25 cigarette smoking cannot be entirely dismissed. Bertazzi et al. (2001, 197005) do note that the 26 sociodemographic characteristics of residents in the three zones were similar based on 27 independently conducted surveys, and no differences in chronic respiratory disease were found 28 across the different zones. If excess mortality was attributable to cigarette smoking, such 29 excesses would be expected to be evident during the entire study period. Latency analyses 30 revealed elevated risks 15-20 years postaccident. Finally, no excesses were observed for other 31 smoking-related cancers of the larynx, esophagus, pancreas, and bladder. The observed excesses This document is a draftfor review purposes only and does not constitute Agency policy. 2-53 DRAFT--DO NOT CITE OR QUOTE 1 in all cancer mortality do not appear to be attributed to differential smoking rates between the 2 two populations. 3 To examine potential for bias due to noncomparability in the two study populations, a 4 comparison of cancer mortality rates between the Seveso regions and the reference population of 5 Lombardy was conducted. Elevated rates for brain cancer mortality were noted in Seveso 6 relative to Lombardy, but the higher rates of leukemia mortality were found in Lombardy 7 relative to Seveso. That no excess was reported for all cancer sites combined lends credence to 8 the hypothesis that the exposure to TCDD from the accident increased rates of cancer after a 9 sufficient period of latency. 10 Stratified analyses were performed across several categorical variables including gender 11 and time since exposure. The numbers of cancer site-specific deaths are quite small in many of 12 the 5-year increments since first exposure. The study, therefore, has limited statistical power to 13 detect differences in mortality rates among the comparison groups for many cancer sites. 14 Bertazzi et al. (2001, 197005) assigned exposures based on zone of residence. Soil 15 sampling within each zone revealed considerable variability in TCDD soil levels within each 16 zone. Moreover, some individuals would have left the area shortly after the accident, and 17 determining the extent to which individuals in Zone B who were subject to the recommendations 18 near the time of the accident adhered to them is difficult. As a result, exposure misclassification 19 is possible, and the use of individual measures of TCDD level in serum is preferred over zone of 20 residence for determining exposure. As noted by the authors, the study is better suited to "hazard 21 identification" than to quantitative dose-response analysis. 22 23 2.4.1.1.1.4.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 24 Given the variability in soil TCDD levels within each zone and the lack of individual 25 level, no effective dose can be estimated for quantitative dose-response analyses. Uncertainty in 26 identifying the critical exposure window for the Seveso cohort is a key limitation. The 27 evaluation of this study indicates that this study is not suitable for quantitative dose-response 28 analysis. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-54 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.4.2. Warner et al. (2002, 197489). 2 2.4.1.1.1.4.2.1. Study summary. 3 To date, Warner et al. (2002, 197489) is the only published investigation of the 4 relationship between serum-based measures of TCDD and cancer in Seveso. Eligible 5 participants from the Seveso Women's Heath Study (SWHS; see Section 2.4.1.2.1.4 for details) 6 were women who, at the time of the accident in 1976, were 40 years of age or younger, had lived 7 in one of the most highly contaminated zones (A or B), and had adequate sera collected soon 8 after the explosion. Enrollment in SWHS was begun in March 1996 and lasted until July 1998. 9 Of the total 1,271 eligible women, 981 agreed to participate in the study. Cancer cases were 10 identified during interview and confirmed through review of medical records. Information on 11 other risk factors including reproductive history and cigarette smoking was obtained through 12 interview. 13 Serum volumes greater than 0.5 mL collected between 1976 and 1981 volume were 14 analyzed. Most sera were collected in 1976/77 (n = 899); samples were collected in 1978-1981 15 for 54 women, and in 1996/97 for 28 women. For most samples collected after 1977, serum 16 TCDD levels were back-extrapolated using a first-order kinetic model with a 9-year half-life 17 (Pirkle et al., 1989, 197861). For 96 women with undetectable values, a serum level that was 18 equal to one-half the detection level was used. 19 Analyses were based only on women who provided serum samples; no extrapolation of 20 values to a larger population was done. Risks were therefore generated using data collected at an 21 individual level. Serum TCDD was analyzed as both a continuous variable and a categorical 22 variable. The distribution of serum TCDD levels of the 15 cases of breast cancer was examined 23 in relation to the distribution of all women in the SWHS. The median exposure was slightly 24 higher among with the 15 cases of breast cancer (71.8 ppt) compared to those without (55.1 ppt), 25 and the exposure distribution among breast cancer cases appeared to be shifted to the right (i.e., 26 the exposures were higher but followed the same distribution); however, no formal test of 27 significance was conducted. 28 Warner et al. (2002, 197489) used Cox proportional hazards modeling techniques to 29 evaluate risk of breast cancer in relation to TCDD serum levels while controlling for a variety of 30 potential risk factors. In all, 21 women had been diagnosed with cancer, and of these, 15 cases 31 were cancer of the breast. The analysis revealed that for every 10-fold increase in TCDD This document is a draftfor review purposes only and does not constitute Agency policy. 2-55 DRAFT--DO NOT CITE OR QUOTE 1 log-serum levels (e.g., from 10 to 100 ppt) the risk of breast cancer increased by 2.1 2 (95% CI = 1.0-4.6). Risk estimates also were generated across four categories (<20, 20.1-44, 3 44.1-100, >100 ppt), with the lowest category used as the reference. The RRs estimated in the 4 third and fourth highest exposure categories were 4.5 (95% CI = 0.6-36.8) and 3.3 5 (95% CI = 0.4-28.0). Although statistical significance was not achieved for either category, 6 likely because of the small number of cases, the greater than threefold risk evident in both 7 categories is worth noting. Given that the reference category had only one incident case 8 underscores the limited inferences that can be drawn from these analyses. The authors adjusted 9 for numerous potential confounders, but observed no differences between the crude and adjusted 10 results; the authors, therefore, presented unadjusted risks. 11 12 2.4.1.1.1.4.2.2. Study evaluation. 13 The findings from the Warner et al. (2002, 197489) study differ from reports in earlier 14 studies in which mortality outcomes noted the absence of an SMR association. The design of 15 this study is much stronger than earlier ones, given the improved characterization of exposure, 16 the ability to compare incidence rates within the cohort, the ability to control for potential 17 confounding variables at an individual level, and the availability of incident outcomes. The use 18 of incident cases (versus mortality data) should also help minimize potential bias due to disease 19 survival. Another important advantage was the ability to measure TCDD near the time of the 20 accident, thereby reducing the potential for exposure measurement error. 21 A potentially important limitation of the Warner et al. (2002, 197489) study was that 22 information was collected only from those who were alive as of March 1996. Therefore, TCDD 23 and other relevant risk factor data could not be collected for those who had previously died of 24 breast cancer. Thirty-three women could not participate because they were either too ill or had 25 died. Of these, three died of breast cancer. Given that there were only 15 breast cancer cases, 26 the exclusion of these 3 cases could have dramatically impacted the findings in either direction. 27 Another limitation was that, at the time of the follow-up, most women were still 28 premenopausal and therefore, most of the cohort (average age = 40.8 years) had not yet attained 29 the age of greater risk of breast cancer (average age at diagnosis among the cases in this cohort 30 was 45.2 years). Although comparable data from Italy were not found, the median age of 31 diagnosis for breast cancer among U.S. women from 2003-2007 was 61 years (Altekruse et al., This document is a draftfor review purposes only and does not constitute Agency policy. 2-56 DRAFT--DO NOT CITE OR QUOTE 1 2010). An ongoing follow-up of the cohort should be completed by 2010, which should allow 2 for increased number of incident breast cancers to be identified. Given that the current analyses 3 were based only on 15 incident cases, this will substantially improve the statistical power of the 4 study. A secondary benefit is that the increased follow-up will allow for an investigation of 5 possible differential effects according to the age the women were at the time of exposure. 6 7 2.4.1.1.1.4.2.3. Suitability o f datafor TCDD dose-response modeling. 8 Several aspects of the Warner et al. (2002, 197489) study are weaknesses in the 9 consideration of this study for further dose-response modeling. Only 15 cases of breast cancer 10 were available, and no increases in risk were found with serum TCDD exposures between 20.1 11 and 44 ppt (n = 2) when compared to those with <20 ppt (n = 1). The average age at the time of 12 enrollment was 40.8 years while the average age at diagnosis among the cases was 45.2 years. 13 As most women had not yet reached the age when breast cancer cases are typically diagnosed, 14 additional follow-up of the cohort would improve the quantitative dose-response analysis and 15 strengthen this study. A key strength of this study, however, is that Warner et al. (2002, 197489) 16 includes an investigation of the relationship between individual serum-based measures of TCDD 17 and cancer in Seveso. Despite the weaknesses, this study meets the evaluation considerations 18 and criteria for inclusion and will be analyzed for quantitative dose-response modeling. 19 2.4.1.1.1.4.3. Pesatori et al. (2003, 197001). 20 2.4.1.1.1.4.3.1. Study summary. 21 Pesatori et al. (2003, 197001) published a review of the short- and long-term studies of 22 morbidity and mortality outcomes in the Seveso cohort in 2003. This paper presented cancer 23 incidence data from 1977 to 1991 for Seveso males and females residing in Zones A, B and R 24 relative to an external population (i.e., uncontaminated areas). Mortality data are also presented 25 for a 20-year follow-up (1976-1996) relative to the reference population. As in the original 26 Bertazzi et al. (2001, 197005) study, RRs were estimated using Poisson regression. No 27 associations were noted for zone of residence and all cancer mortality for either males or 28 females. Although no cases were reported in Zones A and B, soft tissues sarcoma was associated 29 with residence in males from Zone R (RR = 2.6, 95% CI = 1.1--6.3). Among males, residence in 30 Zones A and B was associated with lymphatic and hematopoietic cancer (RR = 1.9, This document is a draftfor review purposes only and does not constitute Agency policy. 2-57 DRAFT--DO NOT CITE OR QUOTE 1 95% CI = 1.1-3.1). This increased risk was due primarily to non-Hodgkin's lymphoma, which 2 accounted for 8 of the 15 incidence cases (RR = 2.6, 95% CI = 1.3-5.3). Among females, 3 increased incidence of multiple myeloma (RR = 4.9, 95% CI = 1.5-16.1), cancer of the vagina 4 (RR = 5.5, 95% CI = 1.3-23.8), and cancer of the biliary tract (RR = 3.0, 95% CI = 1.1-8.2) was 5 associated with residence in Zones A and B. 6 7 2.4.1.1.1.4.3.2. Study evaluation. 8 Study limitations of the Pesatori et al. (2003, 197001) study included exposure 9 misclassification from the use of an ecological measure of exposure (region of residency at time 10 of accident) and low statistical power for some health endpoints. For e.g., all of the RRs 11 presented above for specific cancer mortality among females in the Pesatori et al. (2003, 197001) 12 study were based on fewer than five incident cases. 13 14 2.4.1.1.1.4.3.3. Suitability o f da ta f o r TCDD dose-response modeling. 15 As with the studies of mortality among Seveso residents, the Pesatori et al. (2003, 16 197001) study does not capture TCDD exposure on an individual basis, and soil TCDD levels 17 considerably vary within each zone. Therefore, the quality of the exposure data is insufficient 18 for estimating the effective dose needed for quantitative dose-response analysis. 19 20 2.4.1.1.1.4.4. Baccarelli et al. (2006, 197036). 21 2.4.1.1.1.4.4.1. Study summary. 22 Given previous findings from Seveso, Baccarelli et al. (2006, 197036) examined t(14;18) 23 translocations in the DNA of circulating lymphocytes of healthy dioxin-exposed individuals. 24 These translocations are associated with the development of cancer, namely follicular 25 lymphomas. The study included 211 healthy subjects of the Seveso area, and 101 who had 26 developed chloracne. The investigators analyzed data from 72 high-TCDD plasma level 27 individuals (>10 ppt) and 72 low-TCDD plasma levels (<10 ppt). A three-level categorical 28 variable was used to evaluate dose-response. This variable was developed by dividing those 29 with exposures >10 ppt into two groups: 10- <50 ppt, and 50-475.0 ppt. Trained interviewers 30 administered a questionnaire that collected data on demographic characteristics, diet, and 31 residential and occupational history. This document is a draftfor review purposes only and does not constitute Agency policy. 2-58 DRAFT--DO NOT CITE OR QUOTE 1 The prevalence of t(14;18) was estimated as those individuals having a t(14;18) positive 2 blood sample divided by the t(14;18) frequency (number of copies per million lymphocytes). 3 Baccarelli et al. (2006, 197036) found that the frequency of t(14;18) was associated with plasma 4 TCDD levels, but no association between TCDD and the prevalence of t(14;18) was detected. 5 6 2.4.1.1.1.4.4.2. Study evaluation. 7 Whether the frequency of t(14;18) associated with plasma TCDD levels translates into an 8 increased risk of lymphoma is uncertain as prospective data of TCDD on those who developed 9 non-Hodgkin's lymphoma are lacking. Moreover, the t(14;18) translocation could be an 10 important event in the pre-B stage cell that contributes to tumorigenicity, however subsequent 11 exposure to carcinogenic agents might be necessary for t(14;18) cells to develop into a 12 malignancy (Hoglund et al., 2004, 199130). 13 14 2.4.1.1.1.4.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 15 Given that current TCDD plasma levels were measured for this study, it is unclear if the 16 effects of lymphocyte translocations may be due to initial high exposure or are a function of the 17 cumulative exposure for a longer exposure window. Additionally, whether the frequency of 18 t(14;18) associated with plasma TCDD levels translates into an increased risk of lymphoma is 19 unknown. Dose-response analysis for this outcome, therefore, was not conducted. 20 21 2.4.1.1.1.4.5. Consonni et al. (2008, 524825). 22 2.4.1.1.1.4.5.1. Study summary. 23 Consonni et al. (2008, 524825) analyzed cancer mortality in the Seveso cohort with the 24 addition of a 25-year follow up period. Similar analytic methods as Pesatori et al. (2003, 25 197001) were applied with 25 years of follow-up added to the analysis (Consonni et al., 2008, 26 524825). An important addition in this paper was the presentation of RRs for Zone R, which had 27 the lowest TCDD levels. Poisson regression models were used to calculate RRs of mortality 28 using Seregno as the reference population. Cancer deaths observed in Zones A and B were 42 29 and 244, respectively. 30 No statistically significant differences in all cancer mortality relative to the reference 31 population were noted in any of the zones (Zone A: RR = 1.03, 95% CI = 0.76-1.39; Zone B: This document is a draftfor review purposes only and does not constitute Agency policy. 2-59 DRAFT--DO NOT CITE OR QUOTE 1 RR = 0.92, 95% CI = 0.81-1.05; Zone R: RR = 0.97, 95% CI = 0.92-1.02). Statistically 2 significant excesses in mortality from non-Hodgkin's lymphoma (RR = 3.35, 3 95% CI = 1.07-10.46) and multiple myeloma (RR = 4.34, 95% CI = 1.07-17.52) were observed 4 in the area with the highest TCDD levels (Zone A). No other statistically significant increases in 5 cancer mortality relative to the reference population were apparent. The absence of elevated 6 breast cancer mortality among women in this study was noteworthy, as this finding differs from 7 the results of a study of Seveso women for which TCDD exposures were estimated using serum 8 samples (Warner et al., 2002, 197489). 9 10 2.4.1.1.1.4.5.2. Study evaluation. 11 Although no individual-level data on smoking were available, the potential for 12 confounding is likely minimal. Independent smoking surveys found that the smoking prevalence 13 rates in Desio, one of cities affected by the accident, were similar to those in districts just outside 14 the study area (Cesana et al., 1995, 594366). As mentioned earlier, one would expect elevated 15 RRs over the entire study period if smoking had biased the study results, and not just after 16 15-20 years since exposure to TCDD. 17 18 2.4.1.1.1.4.5.3. Suitability o f da ta f o r TCDD dose-response modeling. 19 The lack of individual-level exposure data precludes quantitative dose-response modeling 20 using these data. 21 22 2.4.1.1.1.5. C h a p a e v sk stu d y . 23 Industrial contamination of dioxin in the Chapaevsk region of Russia has been the focus 24 of research on the environmentally-induced cancer and other adverse health effects. The 25 Chapaevsk region is located in the Samara region of Russia and has a population of 83,000. The 26 region is home to a chemical plant that produced lindane and its derivatives between 1967 and 27 1987, which are believed to be responsible for local dioxin contamination. Soil sampling has 28 demonstrated a strong gradient of increased TCDD concentrations with decreased proximity to 29 the chemical plant (Revich et al., 2001, 199843). 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-60 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.5.1. Revich et al. (2001, 199843). 2 2.4.1.1.1.5.1.1. Study summary. 3 Revich et al. (2001, 199843) used a cross-sectional study to compare mortality rates of 4 Chapaevsk residents to two external populations of Russia and the region of Samara. Mortality 5 rates for all cancers combined among males in Chapaevsk were found to be 1.2 times higher 6 when compared to the Samara region as a whole and 1.3 times higher than Russia. Similar to 7 other studies, statistically significant excess was noted in men (SMR = 1.8, 95% CI = 1.6-1.9) 8 but not in women (SMR = 0.9, 95% CI = 0.8-1.1). Among men, the excess was highest for the 9 smoking-related cancers of the lung (SMR = 3.1, 95% CI = 2.6-3.5) and larynx (SMR = 2.3, 10 95% CI = 1.2-3.8) and urinary organs (SMR = 2.6, 95% CI = 1.7-3.6). Among females, there 11 was no increased SMR for all cancer sites combined, but excesses for breast cancer (SMR = 2.1, 12 95% CI = 1.6-2.7) and cancer of the cervix (SMR = 1.5, 95% CI = 1.0-3.1) were statistically 13 significant. 14 Revich et al. (2001, 199843) also compared age-standardized cancer incidence rates in 15 Chapaevsk to those in Samara. Although statistical tests examining these differences were not 16 reported, higher incidence rates were observed for all cancers combined, cancer of the lip, cancer 17 of the oral cavity, and lung and bladder cancer among males in Chapaevsk. Considerably lower 18 cancer incidence rates also were observed for prostate cancer, cancer of the esophagus, and 19 leukemia/lymphoma among males from Chapaevsk. Among females, incidence rates were 20 higher in 1998 for all cancers in Chapaevsk when compared to Russia and the Samara region, an 21 observation that appears somewhat counter to the presented SMR of 0.9 for all cancer mortality 22 from 1995-1998. Like mortality, rates of breast cancer incidence among women in Chapaevsk 23 were higher than in Russia, as were rates of cervical cancer. Leukemia/lymphoma rates were 24 higher among women in Chapaevsk than in those who lived in the reference populations of 25 Samara and Russia. This finding is contrary to the finding for males who had lower rates of 26 leukemia/lymphoma in Chapaevsk. 27 28 2.4.1.1.1.5.1.2. Study evaluation. 29 Although the Revich et al. (2001, 199843) findings suggest TCDD exposures in 30 Chapaevsk are quite high relative to other parts of the world (Akhmedkhanov, 2002, 197140), 31 evaluation of health outcomes to date have been based on ecological data only. This analysis did This document is a draftfor review purposes only and does not constitute Agency policy. 2-61 DRAFT--DO NOT CITE OR QUOTE 1 not adjust for the influence of other risk factors (e.g., smoking, reproductive characteristics) that 2 could contribute to increased cancer rates for lung cancer in men and breast cancer in women. 3 Given that both the SMRs and SIRs for cancer outcomes vary considerably between men and 4 women, this suggests the possibility that occupational exposures might be a contributing factor in 5 these adverse health outcomes. 6 Future research in Chapaevsk includes plans to conduct a breast cancer case-control 7 study. Women who were born from 1940 onward and who have been diagnosed with breast 8 cancer before the age of 55 were included in the study, although the plan to characterize TCDD 9 using serum is uncertain (Revich et al., 2005, 198777). 10 11 2.4.1.1.1.5.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 12 This study did not meet the considerations and criteria for inclusion in a quantitative 13 dose-response assessment. Given the lack of exposure data on an individual basis, no effective 14 dose can be estimated for this study population. As such, no dose-response modeling was 15 conducted. 16 17 2.4.1.1.1.6. T h e A i r F o rce H e a lth ( "R a n c h H a n d s " c o h o rt) stu d y . 18 Between 1962 and 1971, the U.S. military sprayed herbicides over Vietnam to destroy 19 crops that opposition forces depended upon, to clear vegetation from the perimeter of U.S. bases, 20 and to reduce the ability of opposition forces to hide. These herbicides were predominantly a 21 mixture of 2,4-D, 2,4,5-T, picloram, and cacodylic acid (Institute of Medicine, 2006, 594374). A 22 main chemical sprayed was Agent Orange, which was a 50% mixture of 2,4-D and 2,4,5-T. 23 TCDD was produced as a contaminant of 2,4,5-T and had levels ranging from 0.05 to 50 ppm 24 (Institute of Medicine, 1994, 594376). A series of studies have investigated cancer outcomes 25 among Vietnam veterans. A review of military records to characterize exposure to 26 Agent Orange led Stellman and Stellman (1986, 594380) to conclude that assignment of 27 herbicide levels should not be based solely on self-reports or a crude measure such as military 28 branch or area of service within Vietnam. Investigations have been performed on the Ranch 29 Hands cohort, which consisted of those who were involved in the aerial spraying of 30 Agent Orange between 1962 and 1971. More elaborate methods were used to characterize 31 exposures among these individuals, and these studies are summarized below. This document is a draftfor review purposes only and does not constitute Agency policy. 2-62 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.6.1. Akhtar et al. (2004, 197141). 2 2.4.1.1.1.6.1.1. Study summary. 3 Akhtar et al. (2004, 197141) investigated the incidence of cancer in the Ranch Hand 4 cohort, which was published after the release of the 2003 Reassessment document (U.S. EPA, 5 2003, 537122). The Ranch Hand Unit was responsible for aerial spraying of herbicides, 6 including Agent Orange, in Vietnam from 1962 to 1971. Cancer incidence in the Ranch Hand 7 cohort were compared to a cohort that included other Air Force personnel who served in 8 Southeast Asia during the same period but were not involved in the spraying of pesticides. 9 Health outcomes were identified during the postservice period that extended from the time each 10 veteran left Southeast Asia until December 31, 1999. In contrast to previous analyses of this 11 cohort, the Akhtar et al. (2004, 197141) study took into account concerns that both the 12 comparison and spraying cohorts had increased risks of cancer, and addressed the possibility that 13 workers with service in Vietnam or Southeast Asia might have increased cancer risk. The 14 authors addressed the latter concern by adjusting risk estimates for the time spent in Southeast 15 Asia and for the proportion of time spent in Vietnam. 16 The Ranch Hand cohort comprised 1,196 individuals, and the comparison cohort had 17 1,785 individuals. The comparison cohort was selected by matching date of birth, race, and 18 occupation (i.e., officer pilot, officer navigator, nonflying officer, enlisted flyer, or enlisted 19 ground personnel). TCDD levels were determined using serum levels collected from veterans 20 who completed a medical examination in 1987. For those who did not have a serum measure 21 taken in 1987, but provided one in subsequent years, TCDD levels were back-extrapolated to 22 1987 using a first-order kinetic model that assumed a half-life of 7.6 years. Those with 23 nonquantifiable levels were assigned a value of the limit of detection divided by the square root 24 of 2. A total of 1,009 and 1,429 individuals in the Ranch Hand and comparison cohorts, 25 respectively, provided serum measures that were used in the risk assessment. Veterans also were 26 categorized according to the time their tours ended. This date corresponded to changes in 27 herbicide use. These categories were before 1962 or after 1972 (no herbicides were used), 28 1962-1965 (before Agent Orange was used), 1966-1970 (when Agent Orange use was greatest), 29 and 1971-1972 (after Agent Orange was used). Information on incident cases of cancer in the 30 cohort was determined from physical examinations and medical records. Some malignancies 31 were discovered at death and coded from the underlying causes of death as detailed on the death This document is a draftfor review purposes only and does not constitute Agency policy. 2-63 DRAFT--DO NOT CITE OR QUOTE 1 certificate. A total of 134 and 163 incident cases of cancer were identified in the Ranch Hand 2 and comparison cohort, respectively. Akhtar et al. (2004, 197141) describe case ascertainment 3 verified by record review as being complete. 4 External comparisons were made based on the expected cancer experience derived from 5 U.S. national rates using SIRs and the corresponding 95% confidence interval. Person-years and 6 events were tabulated by 5-year calendar and age intervals. 7 When compared to the general population, no statistically significant excesses in all 8 cancer incidence were observed for either the Ranch Hand (SIR = 1.09, 95% CI = 0.91-1.28) or 9 the comparison cohort (SIR = 0.94, 95% CI = 0.81-1.10). Statistically significant differences 10 were found for three site-specific cancers in the Ranch Hands cohort relative to the general 11 population. Excesses were noted for malignant melanoma (SIR = 2.33, 95% CI = 1.40-3.65) 12 and prostate cancer (SIR = 1.46, 95% CI = 1.04-2.00). In contrast, a reduced SIR was found for 13 cancers of the digestive system (SIR = 0.61, 95% CI = 0.36-0.96). The excess in prostate cancer 14 was also noted in the comparison cohort (SIR = 1.62, 95% CI = 1.23-2.10) relative to the 15 general population. External comparisons were repeated by restricting the cohorts to the period 16 when Agent Orange was used (1966-1970). Again, no statistically significant excesses in all 17 cancer incidence were noted in the Ranch Hand (SIR = 1.14, 95% CI = 0.95-1.37) or 18 comparison cohort (SIR = 0.94, 95% CI = 0.80-1.11). Statistically significant excesses 19 continued to be observed for malignant melanoma (SIR = 2.57, 95% CI = 1.52-4.09) and 20 prostate cancer (SIR = 1.68, 95% CI = 1.19-2.33) in the Ranch Hand component of the cohort. 21 No other statistically significant differences were found among Ranch Hands personnel. 22 For internal cohort analyses, veterans were assigned to one of four exposure categories. 23 Those in the comparison cohort were assigned to the "comparison category." Ranch Hand 24 veterans that had TCDD serum levels <10 ppt were assigned to the "background" category. 25 Those with a TCDD levels >10 ppt had their TCDD level estimated at the end of their Vietnam 26 service with a first-order kinetic model that used a half-life of 7.6 years. These 27 back-extrapolated values that were less than 118.5 ppt were assigned to a "low" exposure group, 28 while those with values above 118.5 ppt were classified as "high" exposure. Akhtar et al. (2004, 29 197141) used Cox regression models to describe risks across the exposure groups using the 30 comparison category as the reference. Risks were adjusted for age at tour, military occupation, 31 smoking history, skin reaction to sun exposure, and eye color. Internal cohort analyses were This document is a draftfor review purposes only and does not constitute Agency policy. 2-64 DRAFT--DO NOT CITE OR QUOTE 1 restricted to those who spent no more than 2 years in Southeast Asia and Ranch Hand workers 2 who served exclusively in Vietnam, and the comparison cohort who served exclusively outside 3 of Vietnam. 4 Statistically significant excesses of cancer incidence (all sites combined) were observed 5 in the highest two exposure groups. A statistically significant trend test (p = 0.04) was detected 6 based on the RRs for the background-, low-, and high- exposure groups: 1.44 7 (95% CI = 0.82-2.53); 2.23 (95% CI = 1.24-4.00), and 2.02 (95% CI = 1.03-3.95). For 8 malignant melanoma, the RRs across the three increasing exposure categories were 2.99, 7.42, 9 and 7.51. The corresponding risk estimates for prostate cancer were 1.50, 2.17, and 6.04. 10 11 2.4.1.1.1.6.1.2. Study evaluation. 12 An important strength of this study is the manner in which TCDD exposure was 13 estimated. Serum data were available for most veterans, and therefore, generalizing exposure 14 from a small sample of cohort members is not a concern as was the case with the NIOSH and 15 Hamburg cohorts. Back-extrapolating to derive past exposures was based on a methodology that 16 has been applied in many of the cohorts, thereby facilitating risk comparisons. An additional 17 strength of the study is the examination of incidence as a measure of disease occurrence rather 18 than mortality. 19 In contrast to the previous analysis (Ketchum et al., 1999, 198120) the analysis by Akhtar 20 et al. (2004, 197141) was restricted to individuals who spent no more than 2 years in Southeast 21 Asia. Previous research had demonstrated that increased time spent in Southeast Asia was 22 associated with an increased risk of cancer. Confounding might have been introduced given that 23 the comparison cohort spent much more time in Southeast Asia than the Ranch Hands. To 24 illustrate, the median number of days spent in Southeast Asia was 790 for comparison cohort 25 members, and the median days for the Ranch Hand cohort in the background, low, and high 26 exposure groups were 426, 457, and 397, respectively. After restricting to those who spent at 27 most 2 years, statistically significant associations were observed for all cancer sites combined, 28 prostate cancer, and malignant melanoma using the internal cohort comparisons. 29 An important issue in the study is the high correlation between 2,4,5-T and 2,4-D, given 30 that both were used in equal concentrations in Agent Orange. As a result, distinguishing the 31 effects of each is impossible. This point is relevant, given that 2,4-D has been associated with This document is a draftfor review purposes only and does not constitute Agency policy. 2-65 DRAFT--DO NOT CITE OR QUOTE 1 prostate cancer in several studies. As a result, the dose-response association with prostate cancer 2 might be due to 2,4-D exposure and not TCDD. This issue also has implications for the 3 interpretation of the dose-response pattern for all cancer sites combined, given that incident 4 prostate cancers accounted for 4 of the 12 incident cases in the high-exposure group. 5 6 2.4.1.1.1.6.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 7 The ascertainment of incident cases and characterization of exposure to TCDD based on 8 serum measures are strong features of the cohort. Confounding by 2,4-D is a major concern. 9 Since delineating the independent effects of other Agent Orange contaminants is not possible, 10 quantitative dose-response analysis was not conducted on this study. 11 12 2.4.1.1.1.6.2. Michalek and Pavuk (2008, 199573). 13 2.4.1.1.1.6.2.1. Study summary. 14 Michalek and Pavuk (2008, 199573) recently published an updated analysis of the 15 incidence of cancer and diabetes in the cohort of Ranch Hand veterans. As with the Akhtar et al. 16 (2004, 197141) analysis, the study included a comparison cohort of other Air Force veterans who 17 served in Southeast Asia at the same time but were not involved with the spraying of herbicides. 18 This study extended previous analyses (Henriksen et al., 1997, 197645; Ketchum et al., 1999, 19 198120) by addressing the number of days of herbicide spraying, calendar period of service, and 20 the time spent in Southeast Asia. Veterans who attended at least one of five examinations were 21 eligible for inclusion. Incident cancer cases also were identified from medical records. 22 The methods used to determine TCDD exposures were as described above in the review 23 of the Akhtar et al. (2004, 197141) study. Blood measures also were taken in 1992, 1997, and 24 2002 for subjects with no quantifiable TCDD levels in 1987, those who refused in 1987, and 25 those new to the study. TCDD dose at the end of service in Vietnam was assigned to Ranch 26 Hands that had TCDD levels above background using a a first-order kinetic model and constant 27 half-life of 7.6 years. Each veteran was then assigned to one of four dose categories: comparison 28 veteran, background (i.e., Ranch Hands with 1987 levels of TCDD <10 ppt), low (Ranch Hands 29 with 1987 levels of TCDD 10.1-91 ppt), and high (Ranch Hands with 1987 levels of TCDD 30 >118.5 ppt). Serum TCDD estimates are available for 1,597 veterans in the comparison cohort, This document is a draftfor review purposes only and does not constitute Agency policy. 2-66 DRAFT--DO NOT CITE OR QUOTE 1 and 986 veterans in the Ranch Hand cohort. The comparison cohort was selected by matching 2 on date of birth, race, and occupation of the Ranch Hands. 3 Michalek and Pavuk (2008, 199573) used Cox regression to characterize risks of cancer 4 incidence across the three upper exposure categories using the comparison category as the 5 referent group. Risk estimates were adjusted for year of birth, race, smoking, body mass index at 6 the qualifying tour, military occupation, and skin reaction to sun exposure. Tests for trend for 7 increased risk of cancer were conducted by testing the continuous covariate log10TCDD. 8 Overall, no association between the TCDD exposure categories and RR of all-site cancer 9 was observed. Those in the highest exposure group had an RR of 0.9 (95% CI = 0.6-1.4). 10 Stratified analyses by calendar period of service showed more pronounced risk for those who 11 served before 1986 (when higher amounts of Agent Orange were used). A statistically 12 significant dose-response trend (p < 0.01) was observed for cancer risk and log10TCDD 13 exposure. The RRs for the background, low, and high groups used in these comparisons were 14 0.7 (95% CI = 0.4-1.3), 1.7 (95% CI = 1.0-2.9), and 1.5 (95% CI = 0.9-2.6). A statistically 15 significant increase, however, was noted when analyses were restricted to those who had sprayed 16 for at least 30 days before 1967 and spent time in Southeast Asia (RR = 2.2, 95% CI = 1.1-4.4). 17 18 2.4.1.1.1.6.2.2. Study evaluation. 19 Michalek and Pavuk (2008, 199573) used the same study population that Akhtar et al. 20 (2004, 197141), and so it has the same strengths and limitations as noted above. The follow-up, 21 however, extends an additional 5 years (until the end of 2004). The findings for the 22 dose-response analyses were not as compelling as the earlier Akhtar et al. (2004, 197141) 23 findings. 24 25 2.4.1.1.1.6.2.3. Suitability o f datafor TCDD dose-response modeling. 26 The key limitation precluding dose-response analysis for the Michalek and Pavuk (2008, 27 199573) study is the possible confounding from the inability to control for 2,4-D and other 28 agents used in Agent Orange. As such, quantitative dose-response analysis was not conducted 29 on this study. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-67 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.7. O th e r s tu d ie s o f p o te n tia l re le v a n c e to d o se -resp o n se m o d elin g . 2 2.4.1.1.1.7.1. Hooiveld et al. (1998, 197829)--Netherlands workers. 3 2.4.1.1.1.7.1.1. Study summary. 4 Hooiveld et al. (1998, 197829) re-analyzed the mortality experience of a cohort of 5 workers employed in two chemical plants in the Netherlands using 6 additional years of 6 follow-up from an earlier study (Bueno et al., 1993, 196993). The cohort consisted of those 7 employed between 1955 and June 30, 1985, and vital status was ascertained until 8 December 31, 1991 (i.e., 36 years of follow-up). These cohort members were involved in the 9 synthesis and formulation of phenoxy herbicides, of which the main product was 10 2,4,5-trichlorophenoxyacetic acid and monochloroacetic acid. This cohort, with a shorter 11 follow-up interval than the original study (t' Mannetje et al., 2005, 197593), was included in the 12 IARC international cohort. The cohort consisted of 1,167 workers, of which 906 were known to 13 be alive at the end of the follow-up. The average length of follow-up was 22.3 years, and only 14 10 individuals were lost to follow-up. 15 The authors used detailed occupational histories to assign exposures. Workers were 16 classified as exposed to phenoxy herbicides or chlorophenols and contaminants if they worked in 17 selected departments (i.e., synthesis, finishing, formulation, packing, maintenance/repair, 18 laboratory, chemical effluent waste, cleaning, shipping-transport, or plant supervision); were 19 exposed to the accident in 1963; or were exposed by proximity (i.e., if they entered an exposed 20 department at least once a week). The 1963 accident was the result of an uncontrolled reaction 21 in the autoclave in which 2,4,5-trichlorophenol was synthesized; an explosion resulted, with 22 subsequent release of PCDDs that included TCDD. Based on these methods of exposure 23 assignment, 562 workers were deemed to be exposed to phenoxy herbicides or chlorophenols, 24 and 567 were unexposed. Due to limited information, 27 workers were classified as having 25 unknown exposure. 26 TCDD exposures also were assigned using serum measured on a sample of workers who 27 were employed for at least 1 year and first started working before 1975. Dioxin-like compounds 28 including PCDDs were also measured in the serum samples but were not analyzed for this study. 29 Of the 144 subjects who were invited to provide samples, 94 agreed. TCDD levels were 30 back-extrapolated to the time of maximum exposure using a one-compartment, first-order kinetic 31 model that used a half-life estimate of 7.1 years. The mathematical model used was This document is a draftfor review purposes only and does not constitute Agency policy. 2-68 DRAFT--DO NOT CITE OR QUOTE 1 ln(TCDDmax) = ln(TCDD) + lag x ln(2)/7.1. The lag was defined as the number of years since 2 last exposure for those exposed by virtue of their normal job duties. For those exposed as a 3 result of the accident in 1963, the lag was defined as the number of years since the accident 4 occurred. 5 The authors made external comparisons of cohort mortality to the Netherlands population 6 using the SMR statistics. Poisson regression was used to perform internal cohort comparisons 7 using unexposed workers as the referent. RRs (measured using rate ratios) generated from the 8 Poisson model also were used to compare mortality based on low, medium, and high TCDD 9 serum-derived categories. The Poisson model included the following covariates as adjustment 10 factors: age, calendar period at end of follow-up, and time since first exposure. 11 When compared to the general population, workers had an excess mortality from cancer 12 (SMR = 1.5, 95% CI = 1.1-1.9), based on 51 cancer deaths. Generally, no excesses were 13 observed for site-specific cancers. The exception included eight deaths from cancers of the 14 urinary organs (SMR = 3.9, 95% CI = 1.7-7.6). Although not statistically significant, SMRs 15 comparable in magnitude to other studies were detected for non-Hodgkin's lymphoma 16 (SMR = 3.8, 95% CI = 0.8-11.0) and Hodgkin's disease (SMR = 3.2, 95% CI = 0.1-17.6). A 17 statistically significant excess of cancer mortality (n = 20 deaths among occupational workers) 18 also was also observed relative to the general population when analyses were restricted to those 19 exposed as a result of the 1963 accident (SMR = 1.7, 95% CI = 1.1-2.7). Three deaths from 20 prostate cancer were also noted among these workers (SMR = 5.2, 95% CI = 1.1-15.3), but no 21 excess was observed with any other cancer site. 22 Internal cohort comparison also demonstrated an increased risk of all cancer mortality 23 among those exposed to phenoxy herbicides, chlorophenols, and contaminants relative to those 24 unexposed (RR = 4.1, 95% CI = 1.8-9.0). A statistically significant increased risk was also 25 noted for respiratory cancer mortality (RR = 7.5, 95% CI = 1.0-56.1). Analyses across 26 categories of TCDD exposure revealed excesses in cancer mortality for all cancer sites 27 combined; however, no dose-response trend was apparent. 28 29 2.4.1.1.1.7.1.2. Study evaluation. 30 Several other studies that have characterized cohorts by TCDD levels have used the area 31 under the curve approach and thus have derived an exposure metric that is time dependent. This document is a draftfor review purposes only and does not constitute Agency policy. 2-69 DRAFT--DO NOT CITE OR QUOTE 1 Hooiveld et al. (1998, 197829) instead created an exposure metric to capture the maximum 2 exposure attained during the worker's employment. Characterizing risks using this metric 3 assumes that other TCDD exposures accrued during a workers' lifetime are not relevant 4 predictors of cancer risk. 5 6 2.4.1.1.1.7.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 7 One study limitation is that although dioxin-like compounds were measured in the serum 8 samples, Hooiveld et al. (1998, 197829) reported associations with mortality for TCDD only. 9 There is some utility to examining dose-response analyses using alternative exposure metrics as 10 those constructed in this cohort. However, the small number of identified cancer deaths, 11 limitations in terms of the exposure assignment (based on nonrepresentative sample, and 12 maximum exposure level) and concern over potential confounding by co-exposures preclude 13 using these data for a dose-response analysis. 14 15 2.4.1.1.1.7.2. t' Mannetje et al. (2005, 197593)--New Zealand herbicide sprayers. 16 2.4.1.1.1.7.2.1. Study summary. 17 t'Mannetje et al. (2005, 197593) described the mortality experience of a cohort of New 18 Zealand workers who were employed in a plant located in New Plymouth. The plant produced 19 phenoxy herbicides and pentachlorophenol between 1950 and the mid-1980s. This study 20 population also was included in the international cohort of producers and sprayers of herbicides 21 that was analyzed by IARC (Kogevinas et al., 1997, 198598; Saracci et al., 1991, 199190). In 22 this 2005 study, analyses were restricted to those who had worked at least 1 month; clerical, 23 kitchen, and field research staff were excluded. The authors followed up 1,025 herbicide 24 producers and 703 sprayers from 1969 and 1973, respectively, until the end of 2000. 25 The cohort consisted of two components: those involved with the production of 26 herbicides and those who were sprayers. For the herbicide producers, exposures were 27 determined by consulting occupational history records; no direct measures of exposure were 28 available. Each department of employment was assigned to one of 21 codes as in the IARC 29 international cohort (Saracci et al., 1991, 199190). Industrial hygienists and factory personnel 30 with knowledge of potential exposures in this workforce classified each job according to 31 potential to be exposed to TCDD, other chlorinated dioxins, and phenoxy herbicides. Exposure This document is a draftfor review purposes only and does not constitute Agency policy. 2-70 DRAFT--DO NOT CITE OR QUOTE 1 was defined as a dichotomous variable (i.e., exposed and unexposed). Among producers, 813 2 were classified as exposed, with the remaining 212 considered unexposed. 3 The "sprayer" component of the cohort includes those who were registered in the national 4 registry of applicators at any time from January 1973 until the end of 1984. For the sprayers, 5 detailed occupational information was lacking. Exposure was, therefore, based on an exposure 6 history questionnaire completed in a previous study of congenital malformations (Smith et al., 7 1982, 198586). This questionnaire, administered to 548 applicators in 1980 and 232 applicators 8 in 1982, achieved a high response rate (89%). Participants were asked to provide information 9 about 2,4,5-T-containing product use on an annual basis from 1969 up to the year the survey was 10 completed. As the use of 2,4,5-T ceased in the mid-1980s, data on occupational exposure to 11 TCDD among these workers are fairly complete. Virtually all sprayers (699 of 703) were 12 exposed to TCDD, higher chlorinated dioxins, and phenoxy herbicides. 13 Deaths among workers were identified through record linkage to death registrations in the 14 New Zealand Health Information Service. Electoral rolls, drivers' licenses, and social security 15 records also were consulted to confirm identified deaths. External comparisons of mortality 16 were made to the New Zealand population using the SMR statistic. The mortality follow-up for 17 the producers began on January 1, 1969 and extended until December 31, 2000. For the 18 sprayers, the follow-up period extended from January 1, 1973 until December 31, 2000. A total 19 of 43 cancer deaths occurred in the producer group and 35 cancer deaths occurred in the sprayer 20 group in the cohort. Where possible, stratified analyses by duration of employment and 21 department were conducted. The departments examined for producers included synthesis, 22 formulation and lab, maintenance and waste, packing and transport, other, and unexposed. 23 SMRs were generated using the New Zealand population as an external referent. A linear test 24 for trend was applied to evaluate dose-response trends according to categories of duration of 25 employment. Stratified analyses also were also done for sprayers who started working before 26 1973, as TCDD levels in 2,4,5-T produced at the New Zealand plant dropped dramatically after 27 1973. Although an SMR was presented for female producers, given that only one cancer death 28 was observed, this study can provide no insight on differential risks between the sexes. 29 Among TCDD-exposed producers, for all cancers combined, no statistically significant 30 excess mortality was found when compared to the general population (SMR = 1.24, 31 95% CI = 0.90-1.67). No dose-response trend in the SMRs for all cancers was observed with This document is a draftfor review purposes only and does not constitute Agency policy. 2-71 DRAFT--DO NOT CITE OR QUOTE 1 duration of employment (p = 0.44). No statistically significant elevated SMR was observed in 2 any of the duration of employment categories for any of the six specific departments examined. 3 A statistically significant positive linear trend, however, was noted among synthesis workers 4 (p = 0.04). There was some suggestion of reduced mortality in the upper exposure levels for 5 workers in the formulation and lab departments. For sprayers, the SMR for all cancer sites 6 combined was not elevated relative to the New Zealand general population (SMR = 0.82, 7 95% CI = 0.57-1.14), nor was a dose-response pattern observed with increasing duration of 8 employment (p = 0.86). Additionally, no statistically significant excess in cancer mortality for 9 all sites combined was evident in workers who were first employed either before 1973 10 (SMR = 0.75, 95% CI = 0.50-1.07) or from 1973 on (SMR = 1.81, 95% CI = 0.59-4.22). For 11 site-specific analyses of cancer mortality, an excess of multiple myeloma was observed among 12 production workers relative to the general population (SMR = 5.51, 95% CI = 1.14-16.1). This 13 SMR was based on three deaths. No statistically significant excess (or deficit) of mortality was 14 found for any other cancer site examined in either the sprayers or the producers. 15 16 2.4.1.1.1.7.2.2. Study evaluation. 17 The physical activity demands of spraying contribute to a healthy worker effect that 18 manifests itself in a lower SMR based on both external comparisons to the general population as 19 a referent, and the SMR generated for the producers in the cohort. The analyses conducted using 20 a simple dichotomy of exposure and duration of employment are limited, as nearly all of the 21 sprayers were unexposed. 22 The dose-response pattern with duration of employment coupled with the observation 23 that higher levels of exposure to TCDD occurred among workers in the synthesis department is 24 an important finding. These workers were also exposed to several other contaminants, however, 25 that include processing chemicals, technical products, intermediates, and byproducts (Kauppinen 26 et al., 1993, 594388). These included phenoxy herbicides and dioxin-like compounds such as 27 chlorinated dioxins. Since the dichotomous exposure measure was based on exposure to TCDD, 28 chlorinated dioxins and phenoxy herbicides, the associated dose-response analyses presented in 29 this study should be interpreted cautiously in light of the inability to either characterize or control 30 for these potential confounders. As such, these co-exposures might have contributed to the 31 dose-response pattern observed with increased duration of employment in the synthesis workers. This document is a draftfor review purposes only and does not constitute Agency policy. 2-72 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.7.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 Although the study authors completed a subsequent analysis of this cohort using 3 serum-derived TCDD (McBride, 2009, 198490), the lack of individual-level TCDD exposures 4 precludes dose-response modeling. 5 6 2.4.1.1.1.7.3. McBride et al. (2009, 198490)--New Zealand herbicide sprayers. 7 2.4.1.1.1.7.3.1. Study summary. 8 McBride et al. (2009, 198490) recently published the mortality experience of the New 9 Zealand cohort in relation to serum estimates of TCDD levels. This study included 10 1,599 workers who were employed between 1969 and November 1, 1989, which was the date 11 that 2,4,5-T was last used. As in their study published earlier in the same year (McBride et al., 12 2009, 197296), the follow-up period extended from the first day of employment until 13 December 31, 2004. Vital status was ascertained through record linkage to the New Zealand 14 Health Information Service Mortality Collection and the Registrar General's Index to Deaths for 15 deaths up to 1990. 16 All current and former workers who lived within 75 km of the plant were invited to 17 provide serum samples. A total of 346 of the eligible workers (68%) provided samples, which 18 represented 22% of the overall study population (346/1599). Based on the serum measures, 70% 19 (241/346) had been exposed to TCDD. This percentage is similar to the estimated 71% of 20 workers who were deemed to have been exposed based on a review of occupational records. The 21 mean serum TCDD value was 9.9 ppt. The highest exposures were observed for those employed 22 in the trichlorophenol operation (23.4 ppt). Values among unexposed workers averaged 4.9 ppt, 23 which is close to the background level of 3.9 ppt among individuals of similar age in the New 24 Zealand general population (Bates et al., 2004, 197113). Details on smoking histories of 25 individuals were also collected for the 346 individuals who provided serum, allowing for an 26 examination of the potential confounding role that smoking might have on derived risk estimates 27 for TCDD. 28 Cumulative exposure to TCDD as a time-dependent metric was estimated for each 29 worker. A detailed description of the methods used to derive TCDD exposure was described in 30 Aylward et al. (2009, 197187). The qualitative TCDD scores available for those with serum 31 measures were used to estimate the cumulative exposures based on a half-life of approximately This document is a draftfor review purposes only and does not constitute Agency policy. 2-73 DRAFT--DO NOT CITE OR QUOTE 1 7 years. A time-dependent estimate of TCDD exposure was derived and the area under the curve 2 was used to obtain cumulative workplace TCDD exposure above background levels. Model 3 performance appears modest as the model explained only 30% of the variance (adjusted R2) 4 when these TCDD exposure estimates were compared with actual serum levels (Aylward et al., 5 2009, 197187). 6 As with previous analyses of the cohort (McBride et al., 2009, 197296; t' Mannetje et al., 7 2005, 197593), external comparisons to the New Zealand general population were made using 8 the SMR statistic. The SMR statistic also was used to compare mortality across four exposure 9 groups relative to the general population, as defined by the serum TCDD estimates: 0-68.3, 10 68.4-475.0, 475.1-2085.7, and >2085.8 ppt-month. The proportional hazards model also was 11 used to conduct internal cohort comparisons across these same four exposure groups. In these 12 analyses, age was used as the time variable, and the covariates of date of hire, sex, and birth year 13 were included in the proportional hazards model. The cut-points for these four exposure 14 categories were chosen so that approximately equal numbers of deaths were included in each 15 category. 16 Consistent with earlier SMR analyses of the same cohort, no increased cancer mortality 17 was observed among "ever" exposed workers in this cohort when compared to the general 18 population (SMR = 1.1, 95% CI = 0.9-1.4). No statistically significant excess was noted for any 19 of the site-specific cancers, although there was some suggestion of increased risk of soft tissue 20 sarcoma (SMR = 3.4, 95% CI = 0.1-19.5), multiple myeloma (SMR = 2.2, 95% CI = 0.2-8.1), 21 non-Hodgkin's lymphoma (SMR = 1.6, 95% CI = 0.3-4.7), and cancer of the rectum 22 (SMR = 2.0, 95% CI = 0.7-4.4). No statistically significant increases in cancer mortality (all 23 sites combined) was found in any of the four exposure categories as measured by the SMR 24 statistic, nor was a dose-response trend noted with increasing exposure categories. No 25 dose-response trends (based on SMR analyses) were noted for five site-specific cancers 26 examined (i.e., digestive organs, bronchus, trachea and lung, soft tissue sarcomas, lymphatic and 27 hematopoietic tissue, and non-Hodgkin's lymphoma), although SMRs for three of the 28 four exposure categories exceeded 2.0 for non-Hodgkin's lymphoma. 29 In contrast to the external cohort comparisons, the RRs generated with the proportional 30 hazards model supported a dose-response trend, as rate ratios increased across increasing TCDD 31 exposure categories. The RRs and their 95% confidence intervals relative to the lowest of the This document is a draftfor review purposes only and does not constitute Agency policy. 2-74 DRAFT--DO NOT CITE OR QUOTE 1 four groups were 1.05 (95% CI = 0.48-2.26), 1.38 (95% CI = 0.64-2.97) and 1.58 2 (95% CI = 0.71-3.52). Neither the linear (p = 0.29) or quadratic (p = 0.82) test for trend, 3 however, was statistically significant. An increased risk of lung cancer mortality was observed 4 in the highest TCDD exposure category relative to the lowest (RR = 5.75, 5 95% CI = 0.76-42.24). The tests for trend for lung cancer, however, also were not statistically 6 significant. 7 A smoking survey was administered to a sample of surviving workers of this cohort, and 8 smoking prevalence was found to be slightly higher among those with higher cumulative 9 exposure (61%) compared to lower exposures (51-56%). These minor differences in smoking 10 prevalence unlikely was a strong enough confounder to explain the fivefold increase in risk of 11 lung cancer found in the highest exposure category. Although the smoking data assessment was 12 a strength of the study, it was limited to only sample of workers and was not available for those 13 who died of lung cancer. 14 15 2.4.1.1.1.7.3.2. Study evaluation. 16 Given high rates of emigration, loss to follow-up (22%) was a potential concern in this 17 study. If comparable emigration rates did occur among the general population then the SMRs 18 would be underestimated. It is unclear to what extent emigration occurred among the general 19 population and whether emigration in both the worker and general populations was dependent on 20 health status. If emigration rates were comparable among these two populations, the associated 21 bias from the under-ascertainment of mortality in the lost to follow-up group would likely 22 attenuate a positive association between TCDD and cancer mortality. Among the worker 23 population, there was not much evidence of differential loss to follow-up with respect to 24 exposure as average exposures were lower (3.2 ppt) among those loss to follow up compared to 25 those with complete follow-up (5.7 ppt). Previous studies among this population also found 26 slightly higher loss to follow-up rates among the unexposed (23%) compared to the exposed 27 (17%) workers (t' Mannetje et al., 2005, 197593). 28 McBride et al. (2009, 198490) did not present results using a continuous measure of 29 TCDD exposure (lagged or unlagged) as was done in most other occupational cohorts. 30 Additionally, the modeling did not consider the use of different periods of latency. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-75 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.1.7.3.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 There is no evidence that the authors considered exposure metrics that are consistent with 3 environmental cancer-causing agents such as exposure modeling that takes latency into account. 4 Given that past occupational cohort studies of TCDD-exposed workers have consistently 5 demonstrated stronger association with lag interval of 15 years, such an approach should be 6 applied to this cohort. This precludes this study from consideration for quantitative 7 dose-response modeling. 8 9 2.4.1.1.1.7.4. McBride et al. (2009, 197296)--New Zealand herbicide sprayers. 10 2.4.1.1.1.7.4.1. Study summary. 11 McBride et al. (2009, 197296) published an updated analysis of the mortality of the New 12 Zealand cohort. The follow-up period was from January 1, 1969 to December 31, 2004 13 extending the previous study by an additional 4 years. In contrast to the previous study where 14 the cohort comprised individuals employed for at least 1 month prior to 1982 (or 1984) 15 (t' Mannetje et al., 2005, 197593), the cohort in this study consisted of all those who worked at 16 least one day between January 1, 1969 and October 1, 2003. This resulted in a cohort of 17 1,754 workers, of which 247 died in the follow-up interval. Seventeen percent of the cohort 18 members were lost to follow-up, which could be a source of selection bias if loss to follow-up 19 was related to both the exposure metrics and the health outcome of interest. Previous data from 20 this cohort (t' Mannetje et al., 2005, 197593), however, showed fairly comparable loss to follow 21 up rates among the unexposed (23%) and the exposed populations (17%). 22 Comparisons to the New Zealand general population were made using the SMR statistic. 23 Stratified analyses were conducted by duration of employment (<3 months, >3 months), sex, 24 latency (<15 years, >15 years), and period of hire (<1976, >1976). The authors defined latency 25 as the period between the day last worked and the earliest of date of death, date of emigration or 26 loss to follow-up, or December 31, 2004. 27 The overall SMR for mortality from all cancer sites combined relative to the New 28 Zealand population was 1.01 (95% CI = 0.85-1.10). Although not statistically significant there 29 was suggestion of an increased risk of rectal cancer (SMR = 2.03; 95%CI = 0.88-4.01) among 30 the employees. SMRs for lymphatic and hematopoietic cancers (overall SMR = 1.21, 31 95% CI = 0.52-2.39) included 3.12 (95% CI = 0.08-17.37) for Hodgkin's disease, This document is a draftfor review purposes only and does not constitute Agency policy. 2-76 DRAFT--DO NOT CITE OR QUOTE 1 1.59 (95% CI = 0.43-4.07) for non-Hodgkin's lymphoma and 3.73, 95% CI = 1.20-8.71), and 2 1.66 (95% CI = 0.20-5.99) for multiple myeloma. No statistically significant excess of cancer 3 mortality was noted among workers employed for <3 months (SMR = 1.19, 4 95% CI = 0.65-2.00), or for >3 months (SMR = 0.98, 95% CI = 0.75-1.26). A statistically 5 significant excess of digestive cancers was found for those who worked fewer than 3 months 6 relative to the New Zealand population (SMR = 2.52, 95% CI = 1.15-4.78). No excesses were 7 observed for any site-specific cancers when analyses were restricted to those who worked for 3 8 or more months. No statistically significant elevated SMRs were found for all cancers 9 (combined) either for a latency period of fewer than 15 years (SMR = 1.14, 95% CI = 0.72-1.71) 10 or a latency period of >15 years (SMR = 0.96, 95% CI = 0.72-1.26). Similarly, no statistically 11 significant excess in cancer mortality was observed for all cancer sites combined, or any 12 site-specific cancer when analyses were stratified by date of hire (<1976, >1976) or by sex. The 13 SMR among women who were employed at the site was 0.68 (95% CI = 0.45-1.00). 14 15 2.4.1.1.1.7.4.2. Study evaluation. 16 High rates of emigration in New Zealand (9% among workers in the cohort) contributed 17 to a fairly high loss to follow-up (22% among workers) during the study period. The loss to 18 follow-up would reduce the overall mortality estimates among the workers, which could 19 underestimate the SMRs if loss to follow-up (and health status) was not comparable in the 20 general population. For example, it is unclear if workers and the general population who 21 emigrated were sicker than those remaining in the cohort. Previous data from the cohort workers 22 suggests that loss to follow-up rates were slightly higher among the low and unexposed 23 populations (McBride, 2009, 198490; t' Mannetje et al., 2005, 197593) worker population, so 24 presumably the highly exposed workers were not lost to follow-up more so than other workers. 25 26 2.4.1.1.1.7.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 27 This study extended the mortality follow-up and included stratified analyses to 28 investigate effect modification by period of latency, sex, and date of hire. A key limitation was 29 the lack of direct measures of exposure for study participants which precluded estimating 30 effective dose needed for dose-response modeling. This study did not meet the considerations 31 and criteria for inclusion in quantitative dose-response analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-77 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.1.2. K e y ch aracteristics o f epidem iologic ca n cer stu dies 2 See Table 2-1 at the end of the chapter for a comparison of the length of follow-up, 3 latency period used, the half-life for TCDD used, and the fraction of TEQs accounted for by 4 TCDD (when applicable) for each study. 5 6 2.4.1.1.3. F easibility o f TC D D can cer dose-respon se m od elin g --su m m ary discu ssion by 7 cohort. 8 2.4.1.1.3.1. U sing th e N I O S H c o h o r t in d o se -resp o n se m o d e lin 2. 9 It is important to evaluate the NIOSH cohort in cancer dose-response modeling of TCDD. 10 This cohort is the largest assembled to date, direct measures of TCDD based on sampling are 11 available, and the lengthy follow-up interval allows for latent effects to be taken into account. 12 Further, although this cohort consists mostly of male workers, these workers were occupationally 13 exposed to TCDD daily, as compared to the acute accidental exposures of other occupational 14 cohorts. Although the most recent analyses of a subset of the NIOSH cohort showed no 15 association between serum TCDD levels and cancer mortality, the study authors did not examine 16 latency effects (Collins et al., 2009, 197627). Incorporation of latency intervals is important in 17 light of the stronger dose-response relationships that consistently have been observed with a 18 15-20 year latency interval in previous investigations of the NIOSH and other cohorts 19 (Steenland et al., 2001, 197433). 20 Most published studies of the NIOSH cohort did not evaluate exposures to dioxin-like 21 compounds. An exception is the analysis by Steenland et al. (2001, 197433). Although 22 Steenland et al. (2001, 197433) did not incorporate individual-level data on dioxin-like 23 compounds, based on their previous work (Piacitelli et al., 1992, 197275) they assumed that TEQ 24 occupational exposures occurred as a result of TCDD alone in this population. TCDD exposures 25 provided a better fit to the data than the TEQ-based metric, and 15-year latencies improved the 26 fit for both metrics (relative to unlagged exposures). The lifetime risk estimates for an increase 27 in 10 TEQs (pg/kg of body weight/day/sex) ranged from 0.05-0.18%. The value added for this 28 measure is the incorporation of the contribution of other dioxin-like compounds to the 29 background rates. 30 Blue collar workers, such as those in the NIOSH cohort, typically have higher rates of 31 smoking than the general population (Bang and Kim, 2001, 197081; Lee et al., 2007, 594391). This document is a draftfor review purposes only and does not constitute Agency policy. 2-78 DRAFT--DO NOT CITE OR QUOTE 1 This potential source of confounding would be expected to produce a higher SMR for lung 2 cancer mortality, and could contribute to the excess noted in the cohort with longer lag intervals. 3 This bias, however, likely is not large as no statistically significant excess of nonmalignant 4 respiratory mortality was found in these workers. Any associated bias from smoking would be 5 expected to be smaller for comparisons conducted within the cohort, as fellow workers would be 6 expected to be more homogeneous with respect to their risk factor profile than with an external 7 general population referent group. Stratified analyses using both internal and external 8 comparison groups also did not identify important differences in associations with TCDD 9 exposure between smoking and nonsmoking cancers. Thus, fatal cancer risk estimates reported 10 for workers in the NIOSH cohort appear to provide a reasonable estimate of the carcinogenic 11 potency of TCDD. 12 Although the Steenland et al. (2001, 197433) study did not directly account for the 13 possible confounding effects of other occupational exposure, the authors did address this source 14 of potential bias. No known occupational exposures to carcinogens occurred, with the exception 15 of 4-aminobiphenyl, which occurred at one plant. Two deaths from mesothelioma also occurred 16 in the cohort, so some exposure to asbestos might also have occurred in the cohort (Fingerhut 17 et al., 1991, 197375). The statistical analyses suggested that the inability to control for other 18 occupational exposures would not have unduly affected risk estimates generated from internal 19 cohort comparisons. For instance, the removal of one plant at a time from the analysis did not 20 materially change dose-response estimates generated from the Cox model (Cheng et al., 2006, 21 523122). Moreover, adding a variable to represent plant in the Cox regression had little impact 22 on the risk estimates. Given that other occupational exposures varied by plant, a change in risk 23 estimates would be expected if such exposures were strong confounders. 24 The Cheng et al. (2006, 523122) analysis provides important information about the 25 impact of applying kinetic models to the data. The CADM TCDD kinetic model resulted in 26 dramatic decreases in the TCDD cancer mortality risk estimates when compared to the one-stage 27 compartmental model that had been applied. Although Cheng et al. (2006, 523122) suggested 28 that the CADM model provides a better fit to the data than the typically used simple 29 one-compartmental model, statistical comparisons of model fit were not reported. Therefore, 30 there is value in presenting the range in risk estimates across different models when 31 characterizing dose-response relationships. This document is a draftfor review purposes only and does not constitute Agency policy. 2-79 DRAFT--DO NOT CITE OR QUOTE 1 Finally, the half-life of TCDD is generally recognized to vary according to body fat 2 percentage, data that were not available for the NIOSH workers. The inability to account for 3 between-worker variability in body fat would introduce exposure measurement error. That body 4 fat percentage would not be expected to correlate with cumulative exposure to TCDD exposure, 5 however, would limit the potential for misdassification bias. The effect of any nondifferential 6 exposure measurement error likely would serve to attenuate the risk estimates of the study. 7 8 2.4.1.1.3.2. Using th e BA S F coh ort in d ose-resp on se m od eling. 9 The availability of blood lipid data for TCDD allows for characterization of cumulative 10 TCDD exposures in the BASF cohort. TCDD blood lipid data were collected for 90% of the 11 surviving members of the cohort (138 of 154) and these serum measures were used to generate 12 TCDD exposure estimates for all 254 cohort members. Therefore, the potential for 13 misclassification from extrapolating these exposures to the entire cohort may not be as likely as 14 for the NIOSH cohort where sera data were available for only a small fraction of workers. These 15 data were, however, collected long after the accident (36 years) and had to be back-extrapolated 16 to derive the initial exposures. 17 The data on this cohort included several risk factors such as cigarette smoking and body 18 mass index. One advantage is that cumulative TCDD levels by body mass index can be 19 estimates on an individual-level basis. As expected, the derived cumulative measures appear to 20 compare well with severity scores of chloracne. The finding that more pronounced risks are 21 found 15-20 years after first exposure are also consistent with findings from several other 22 cohorts (Bertazzi et al., 2001, 197005; Fingerhut et al., 1991, 197375; Manz et al., 1991, 23 199061). 24 One key limitation of the BASF cohort is its relatively small sample size (n = 243), which 25 limits the ability to evaluate dose-response relationships for site-specific cancers. Also, the 26 quality of the ascertainment of cancer incidence cannot be readily evaluated as the geographic 27 area of the cohort is not covered by a tumor registry. Ott and Zober (1996, 198101) state that 28 nonfatal cancers could have been more likely to be missed in early years, which could partially 29 contribute to the larger standardized incidence ratio found for cancer with longer latencies. 30 Commenting on risk differences derived from incident and decedent cancer outcomes is difficult. 31 Among those comprising the cohort, the ascertainment of incident outcomes was recognized to This document is a draftfor review purposes only and does not constitute Agency policy. 2-80 DRAFT--DO NOT CITE OR QUOTE 1 be less complete in early years. Although the ascertainment of mortality outcomes was generally 2 regarded to be good among the 243 workers, some workers who died or moved likely were 3 missed when the cohort was constructed. These deaths would have been more likely to have 4 occurred several years before the second component of the cohort was assembled. 5 The use of the SMR statistic for this study population is associated with important 6 sources of uncertainties. Deaths were surely missed, particularly for the third component of the 7 cohort that accounts for approximately 38% (94/247) of the entire cohort; this factor would serve 8 to underestimate the overall SMR. As mentioned before, this component of the cohort was 9 assembled through the recruitment of workers known to be alive in 1986. Despite this limitation, 10 the characterization of exposure data and availability of other risk factor data at an individual 11 level allow the development of quantitative dose-response analyses. 12 13 2.4.1.1.3.3. U sing th e H a m b u r g c o h o r t in d o se-resn o n se m o d elin g . 14 The Hamburg cohort lacked data on cigarette smoking, and, therefore, effect estimates 15 could not be adjusted for this covariate. Additional analyses that excluded lung cancers resulted 16 in an even stronger dose-response relationship between all cancer mortality and TCDD. Serum 17 levels of TCDD also were also not associated with smoking status in a subgroup of these workers 18 (Flesch-Janys et al., 1995, 197261) suggesting that smoking is not likely a confounder of the 19 association between all cancer mortality and TCDD. 20 An important limitation of the cohort is the reliance on blood and tissue measurements of 21 190 workers that likely represent a highly selective component of the cohort. This subset of 22 workers was identified at the end of the observation period, and therefore, excludes workers who 23 died or could not be traced. There are uncertainties in deriving department- and period-specific 24 estimates for a period that extends over three decades using this number of workers. 25 Additionally, the criteria applied to the reference population could have introduced some bias. 26 Workers were included only in the reference group if they had been employed for at least 27 10 years in a gas supply industry. The criteria were much different for the workers who were 28 exposed to TCDD (only 3 months of employment). As a result, the reference group likely would 29 be more susceptible to the healthy worker effect. Internal cohort comparisons, which should be 30 void of such bias, however, generally produced results similar to those based on the external 31 comparison population. Therefore, the Becher et al. (1998, 197173) study meets the criteria and This document is a draftfor review purposes only and does not constitute Agency policy. 2-81 DRAFT--DO NOT CITE OR QUOTE 1 additional epidemiological considerations which allowed for development of quantitative 2 dose-response analyses. 3 4 2.4.1.1.3.4. U sing th e S eveso c o h o r t in d o se -resp o n se m o d e lin 2. 5 Unlike many of the occupational cohorts that were examined, data from the Seveso 6 cohort are representative of a residential population whose primary exposure was from a single 7 TCDD release. A notable exception is the BASF cohort where workers were exposed primarily 8 through two accidents that occurred in the plant. The Seveso data, therefore, might permit 9 cancer dose-response investigations in women and children. 10 Uncertainty in identifying the critical exposure window for most of the outcomes related 11 to the Seveso cohort is a key limitation. An important feature of the Seveso cohort, however, is 12 that TCDD levels were much lower among those in the highest exposure zones in Seveso 13 (medians range from 56-136 ng/kg) (Eskenazi et al., 2004, 197160) than those in the 14 occupational cohorts who had TCDD exposures that were sometimes more than 1,000 ng/kg. 15 Given these dramatic differences in exposures, the standardized mortality ratios (after 16 incorporating a 15-20 year latency period) for all cancer sites combined are remarkably similar 17 between the Seveso and the occupational cohort analyses. Perhaps more importantly, the data 18 from Seveso might be more relevant for extrapolating to lower levels, given that exposures to 19 TCDD are two orders of magnitude higher than background levels (Smith and Lopipero, 2001, 20 198585). 21 The Warner et al. (2002, 197489) study found a positive association between serum 22 levels of TCDD and breast cancer. As noted previously, ascertainment of incident cases for all 23 cancers would allow for a dose-response relationship to be evaluated. Moreover, future breast 24 cancer analyses in this cohort should strengthen the quantitative dose response analyses of this 25 specific cancer site. The strengths of the Warner et al. (2002, 197489) study outlined earlier 26 suggest that this study should be considered for cancer dose-response modeling. 27 Earlier Seveso studies likely are unsuitable for conducting quantitative risk assessment. 28 These previous studies used an indirect measure of TCDD exposure, namely, zone of residence. 29 Soil concentrations of TCDD varied widely in these three zones (Zone A: 15.5-580.4 ppt; 30 Zone B: 1.7-4.3 ppt; and Zone R: 0.9-1.4 ppt), which could have resulted in considerable 31 exposure misclassification. The Warner et al. (2002, 197489) study greatly improved the This document is a draftfor review purposes only and does not constitute Agency policy. 2-82 DRAFT--DO NOT CITE OR QUOTE 1 characterization of TCDD exposure using serum measures, and also allowed for control of 2 salient risk factors that may have resulted in bias due to confounding. 3 At this time it is unclear whether any study has examined the relationship between cancer 4 and serum estimates of TCDD among Seveso males exposed from the 1976 accident. 5 6 2.4.1.1.3.5. U sing th e C h a p a e v sk r e la te d d a ta in d o se -resp o n se m o d elin g . 7 Currently, individual-level exposure data are lacking for residents of this area and there is 8 no established cohort for which cancer outcomes can be ascertained. These limitations, 9 therefore, preclude the inclusion of Chapaevsk data in a quantitative dose-response analysis. 10 11 2.4.1.1.3.6. U sing th e R a n c h H a n d s c o h o r t in d o se -resp o n se m o d elin g . 12 An important limitation of the Ranch Hands cohort for TCDD and cancer dose-response 13 modeling is an inability to isolate TCDD effects from the effects of other agents found in the 14 associated herbicides. Exposure to other dioxin-like compounds was not estimated in this study 15 and could confound the previously reported associations. As such, dose-response analyses on 16 this population were not conducted. 17 18 2.4.1.1.4. D iscu ssion o f g e n era l issu es rela ted to dose-respon se m odelin g 19 2 .4 1 .1 .4 .1 . A s c e r ta in m e n t o f exposu res. 20 Several series of epidemiological data have used serum measures to estimate TCDD 21 levels. Serum data offer a distinct advantage in that they provide an objective means to 22 characterize TCDD exposure at the individual level. The serum measures in the occupational 23 cohorts, however, are limited in two important ways. First, these samples are generally collected 24 from small subsets of the larger cohorts; therefore, using these measures to extrapolate to the 25 remainder of the cohort could introduce bias due to exposure misclassification. The 26 second limitation is related to estimating the half-life of TCDD. As noted previously, exposures 27 to TCDD were back-extrapolated several decades from serum samples collected among 28 surviving members of several cohorts. This approach was used in the NIOSH, Ranch Hands, 29 BASF, New Zealand, and Hamburg cohorts. The reported half-life of TCDD among these 30 populations was reported between 7.1 to 9.0 years and shown to vary with several individual 31 characteristics including age, body fat composition, and smoking. The derivation of half-lives This document is a draftfor review purposes only and does not constitute Agency policy. 2-83 DRAFT--DO NOT CITE OR QUOTE 1 from a sample of workers, and application of these estimates to retrospectively characterize 2 exposure can introduce uncertainty into the lifetime exposure estimates. It is important to note, 3 however, that sensitivity analyses results in several studies have been fairly consistent when 4 evaluating the impact of half-life of TCDD (Flesch-Janys et al., 1995, 197261; Steenland et al., 5 2001, 197433). 6 A unique advantage of the Seveso study is that serum measures were taken shortly after 7 the accident, and therefore characterization of TCDD exposure in this population does not 8 depend on assumptions needed to back-extrapolate exposures several decades. 9 10 2.4.1.1.4.2. L a te n c y in terva ls. 11 Many of the epidemiological studies indicate stronger associations between TCDD and 12 cancer outcomes once a latency period has been considered. Generally, risks are higher when a 13 lag period of 15-20 years is included. As noted previously, this observation is consistent with 14 many other environmental carcinogens such as radon, radiation, and cigarette smoking. That 15 recent exposures do not contribute to increased cancer risk provides some support that the 16 initiation and promotion phases might occur many years before death making recent exposures 17 irrelevant for these analyses. The ability to discriminate between models of varying latency, 18 however, was limited in many studies. The application of biologically based modeling could 19 provide additional important insights on which phase(s) of carcinogenesis TCDD exerts an 20 influence. Such modeling, however, would necessitate having data on an individual-level basis. 21 Ideally, this modeling would use cancer incident data rather than mortality outcomes, given that 22 for many cancers, the median survival time exceeds 5 years. 23 24 2.4.1.1.4.3. Use o f th e S M R m e tric. 25 The occupational cohorts and the studies in Seveso and Chapaevsk have made inferences 26 regarding the effects of TCDD on mortality using the SMR. When compared to the general 27 population, the healthy worker effect may result in a downward bias in the SMR. This often can 28 manifest as SMRs less than 1 for several causes of mortality. The effect of this bias is, however, 29 generally lower for cancer outcomes. Cancer outcomes, whether incidence or death, typically 30 occur later in life and do not generally affect an individual's ability to work at earlier ages. This document is a draftfor review purposes only and does not constitute Agency policy. 2-84 DRAFT--DO NOT CITE OR QUOTE 1 There are several approaches that can be taken to minimize potential biases introduced by 2 the healthy worker effect, which would account for workers being healthier than the general 3 population. Comparisons of mortality (or cancer incidence) can be made to other cohorts of 4 similar workers. If done properly, this can allow for some control of characteristics such as 5 sociodemographic characteristics and smoking as the two populations can be matched by these 6 factors. However, it may be the case that other working populations are exposed to other 7 harmful exposures, thereby making it difficult to estimate risk associated with a specific agent 8 (such as TCDD) in the cohort of interest. A second and preferred approach to control for the 9 healthy worker effect, should it prove feasible, is to conduct comparisons of health outcomes in 10 relation to exposure within the cohort. These comparisons are less likely to be influenced by 11 other potential confounding variables such as smoking, socioeconomic status, and other 12 occupational exposures that are generally more homogeneous within the cohort relative to 13 external populations. Moreover, the mechanisms used to identify health outcomes and follow 14 individuals over time are generally applied in the same manner to all cohort members. Taken 15 together, where different comparisons have been made to generate risk estimates, those that have 16 been conducted using internal cohort comparisons are preferable. 17 In addition to potential bias from the health worker effect, the comparison of SMRs 18 between studies is not always straightforward and is not recommended by some (Myers and 19 Thompson, 1998, 594395; Rothman, 1986, 046091). The SMR is the ratio of the observed 20 number of deaths to the expected number of deaths and is often referred to as the method of 21 indirect standardization. The expected number of deaths is estimated by multiplying the number 22 of person-years tabulated across individuals in the cohort, stratified by age, by rates from a 23 reference population that are available for the same strata. Therefore, each population cohort 24 will have an estimated number of cases derived using a different underlying age structure. As 25 outlined by Rothman (1986, 046091), the mortality rates might not be directly comparable to 26 each other, although the impact of such bias will be much less if the age-distribution of the 27 cohorts is similar. While it might be reasoned that the TCDD exposed workers would have 28 similar age distributions this is in fact not the case (Becher et al., 1998, 197173; Ott et al., 1993, 29 594322; Thiess et al., 1982, 064999). This may be due to exposure occurring both chronically, 30 as well as from acute exposures due to accidental releases that happened at various times at 31 different plants. This is evident with the Hamburg and the BASF cohorts, as most individuals This document is a draftfor review purposes only and does not constitute Agency policy. 2-85 DRAFT--DO NOT CITE OR QUOTE 1 comprising the BASF cohort were employed at the time of the accident (1953/1954), while most 2 of the Hamburg cohort (852/1048) was employed after 1954; the follow-up of these cohorts 3 ended at approximately the same time. 4 The method of direct standardization allows for a more meaningful comparison of 5 mortality rates to be made between cohorts. With this approach, weights (usually based on age 6 and sex) are drawn from a standard population and are, in turn, applied to disease rates for the 7 same strata observed in the cohort of interest. A comparison of weighted rates between different 8 cohorts would then be based on the same population standard. 9 Despite these limitations in comparing SMRs between studies, Armstrong (1995, 10 594397) argues that the comparisons are valid if the underlying stratum specific rates in each 11 exposure grouping are in constant proportion to external rates. Comparisons of the SMRs 12 between studies will be biased only if there is an interaction between age and TCDD (i.e., the RR 13 of disease due to exposure differs by age). For cancer outcomes, the finding that associations 14 become stronger after a period of latency is incorporated into the analyses suggests that this 15 assumption does not hold true. That is, risk estimates would be lower among young workers. 16 Similarly, for noncancer outcomes, some of the data from the Seveso cohort suggests differential 17 effects according to the age at exposure. 18 The use of the SMR might also be biased in that workers exposed to TCDD could be 19 subject to more intensive follow-up than the general population, and as a result, differential 20 coding biases with cause of death might occur. Moreover, some cohorts (e.g., the BASF cohort) 21 have been assembled, in part, by actively seeking out survivors exposed to accidental releases of 22 dioxins. As such, they would not include persons who have died or who were lost to follow-up. 23 This would result in underascertainment of deaths and SMRs developed from these data. The 24 use of an internal cohort comparison offers distinct advantages to overcome potential sources of 25 selection bias. Given these uncertainty about comparability across the different studies, 26 conducting a meta-analysis of cancer outcomes for TCDD using the SMR statistic is not 27 warranted for this analysis. 28 29 2.4.1.1.4.4. A l l ca n cers versu s site-specific. 30 An important consideration for quantitative dose-response modeling is the application of 31 models for all cancers combined, or for site-specific cancers. Consistency is often lacking for This document is a draftfor review purposes only and does not constitute Agency policy. 2-86 DRAFT--DO NOT CITE OR QUOTE 1 site-specific cancers, which might be due in large part to the relatively small number of cases 2 identified for site-specific cancers in the cohorts. Although the risk estimates produced for all 3 cancer sites have important limitations and uncertainties, the data are far more consistent in 4 terms of the magnitude of an association and latency intervals. The IARC evaluation has put 5 forth the possibility of a pleuripotential mode of action between TCDD and the occurrence of 6 cancer. Despite the criticism of this assertion by some (Cole et al., 2003, 197626), the general 7 consistency of an increased risk for all-cancer mortality across the occupational cohorts when 8 latency intervals have been incorporated, provides adequate justification for dose-response 9 quantification of all cancer sites combined. 10 11 2.4.1.1.4.5. S u m m a r y o f e p id e m io lo sic c a n cer s tu d y eva lu a tio n s f o r d o se-resp o n se 12 m o d elin g . 13 All epidemiologic cancer studies summarized above were evaluated for suitability of 14 quantitative dose-response assessment using the TCDD-specific considerations and study 15 inclusion criteria. The results of this evaluation are summarized in a matrix style array (see 16 Table 2-2) at the end of this section, and descriptively in Appendix B. Table 2-4 summarizes the 17 key epidemiologic cancer studies suitable for further TCDD dose-response analyses. 18 19 2.4.I.2. N on can cer 20 In this section, the available epidemiological data that could be used in a dose-response 21 analysis for noncancer endpoints are evaluated. Because many of the key studies also evaluated 22 cancer outcomes, the noncancer studies are presented in the same order as presented in 23 Section 2.4.1.1. Generally, the strengths and limitations of the cancer studies also apply to the 24 noncancer outcomes. In this section, key features of these studies that have direct relevance to 25 modeling of noncancer outcomes in particular are highlighted. To reduce redundancy, a detailed 26 overview of many of these cohorts and studies are not provided here. Instead, the reader should 27 refer to Section 2.4.1.1.1. 28 This document is a draftfor review purposes only and does not constitute Agency policy. 2-87 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1. N o n ca n cer cohorts. 2 2.4.1.2.1.1. T he N I O S H cohort. 3 2.4.1.2.1.1.1. Steenland et al. (1999, 197437). 4 2.4.1.2.1.1.1.1. Study summary. 5 The 1999 published report of NIOSH workers exposed to TCDD also conducted external 6 cohort comparisons to the U.S. general population using SMRs for mortality outcomes other than 7 cancer (Steenland et al., 1999, 197437). Analyses are based on 3,538 workers employed at 8 8 plants from 1942 to 1984. SMRs were based on a mortality follow-up that was extended until 9 the end of 1993. Cox regression analyses were used to compare mortality risk in relation to 10 TCDD exposure within the cohort. 11 12 2.4.1.2.1.1.1.2. Study evaluation. 13 Overall, no statistically significant differences in all-cause mortality (SMR = 1.03, 14 95% CI = 0.97-1.08) were observed. Mortality from ischemic heart disease (SMR = 1.09, 15 95% CI = 1.00-1.20) and accidents (SMR = 1.25, 95% CI = 1.03-1.50) was slightly elevated. 16 Based on the external comparison population, the dose-response relationship for ischemic heart 17 disease observed with the SMRs calculated across TCDD exposure septiles was not statistically 18 significant (p = 0.14). Overall, excess risk was not evident for diabetes, cerebrovascular disease, 19 or nonmalignant respiratory disease using the external population comparisons. Internal cohort 20 comparisons using the Cox regression model were performed using 0 and 15-year lag intervals. 21 A dose-response trend was observed for the derived ratios across the unlagged cumulative 22 TCDD exposure septiles for ischemic heart disease (p = 0.05) and diabetes (p = 0.02). For 23 ischemic heart disease mortality, those in the upper two septiles had rate ratios of 1.57 24 (95% CI = 0.96-2.56) and 1.75 (95% CI = 1.07-2.87), respectively, relative to those in the 25 lowest septile. In contrast, an inverse dose-response relationship was observed for diabetes 26 mortality. The inverse association found for diabetes is inconsistent with the positive association 27 reported in the Ranch Hands study (Michalek and Pavuk, 2008, 199573). However, previous 28 reports have questioned the use of death certificates as the means to ascertain outcome as 29 diabetes may be under-reported especially among descendents with diabetes who die from cancer 30 (McEwen and TRIAD, 2006, 594400). 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-88 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.1.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 The inverse association with diabetes precludes dose-response analysis for this outcome. 3 The dose-response relationship between TCDD exposure and ischemic heart disease mortality 4 was not statistically significant at the alpha level of 0.05 and was not observed in other cohorts. 5 Furthermore, fatal outcomes are not a suitable basis for development of an RfD. For these 6 reasons, dose-response analysis for this outcome is precluded. 7 8 2.4.1.2.1.1.2. Collins et al. (2009, 197627). 9 2.4.1.2.1.1.2.1. Study summary. 10 Collins et al. (2009, 197627) recently described the mortality experience of Dow 11 employees who worked in Midland, Michigan. This plant produced 2,4,5-trichlorophenol 12 between 1942 and 1979, and 2,4,5-T between 1948 and 1982. The cohort consisted of 13 1,615 workers exposed to TCDD from as early as 1942; the follow-up of the cohort extended 14 until 2003. 15 TCDD exposures were derived using serum samples obtained from 280 surviving 16 individuals. A simple one-compartment, first-order pharmacokinetic model was used to estimate 17 time-dependent TCDD measures. The area under the curve approach was then applied to 18 estimate cumulative TCDD exposure above background. A half-life of 7.2 years for TCDD 19 based on earlier work was incorporated into the exposure estimation (Flesch-Janys et al., 1996, 20 197351). 21 Collins et al. (2009, 197627) made an external comparison of the mortality rates of the 22 cohort to the U.S. general population using the SMR statistic. Noncancer causes of death 23 included all causes, diabetes, cerebrovascular disease, nonmalignant respiratory disease, cirrhosis 24 of the liver, and accidents. Overall, no statistically significant difference in all-cause mortality of 25 these workers was detected when compared to the general population (SMR = 0.9, 26 95% CI = 0.9-1.0). Except for cirrhosis of the liver (SMR = 0.4, 95% CI = 0.1-0.8), no 27 differences were found for any of the noncancer causes of death relative to the general 28 population. 29 Internal cohort analyses based on cumulative measures of TCDD were conducted for 30 mortality from diabetes, ischemic heart disease, and nonmalignant respiratory disease using the 31 Cox regression model. These models adjusted for possible confounders such as year of hire and This document is a draftfor review purposes only and does not constitute Agency policy. 2-89 DRAFT--DO NOT CITE OR QUOTE 1 birth year. No statistically significant association was found between continuous measure of 2 TCDD and these causes of death. 3 4 2.4.1.2.1.1.2.2. Study evaluation. 5 Given that the external comparisons may result in bias from the healthy worker effect, 6 results from the internal cohort comparisons using the Cox regression model are preferred. 7 These analyses were performed for diabetes, ischemic heart disease, and nonmalignant 8 respiratory disease. TCDD levels for these workers were estimated using a simple 9 one-compartment pharmacokinetic model (Aylward et al., 2007, 197175). The hazard ratios 10 generated from the Cox regression model were not statistically significant for any of the 11 three noncancer outcomes modeled. 12 13 2.4.1.2.1.1.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 14 No association of an increased risk for an adverse effect was observed with any of the 15 noncancer outcomes. In addition, since noncancer mortality was the endpoint being examined, 16 dose-response modeling based on this population was not conducted. 17 18 2.4.1.2.1.2. T he B A S F cohort. 19 2.4.1.2.1.2.1. Ott and Zober (1996, 198101). 20 2.4.1.2.1.2.1.1. Study summary. 21 In 1996, Ott and Zober published a report on the mortality experience of the cohort of 22 243 BASF male workers who were accidentally exposed to 2,3,7,8-TCDD in 1954 or in the clean 23 up that followed. The mortality follow-up of this cohort extended until the end of 1992. 24 External comparisons of mortality were made to the German population using the SMR statistic. 25 Internal cohort comparisons were also made by estimating cumulative TCDD for the cohort 26 using serum measures that were obtained from 138 workers. Ott et al. (1993, 594322) provided 27 a detailed account of the methodology to estimate TCDD. Briefly, a cumulative measure of 28 TCDD expressed in pg/kg was derived, by first estimating the half-life of TCDD using 29 individuals who had repeated serum measures; the half-life was estimated to be 5.8 years. 30 Individual-level data on body fat were used to account for the influence of body fat on decay 31 rates. Half-life estimates of TCDD varied (range: 5.1-8.9 years) and were dependent on body fat This document is a draftfor review purposes only and does not constitute Agency policy. 2-90 DRAFT--DO NOT CITE OR QUOTE 1 composition (20% and 30%, respectively). This approach differed from previous analysis of this 2 cohort that used a constant 7-year half-life (Ott et al., 1993, 594322). TCDD levels at the time of 3 serum sampling were then estimated as the product of TCDD concentration in blood lipid and 4 the total lipid weight for each worker. Nonlinear models then were applied to estimate the 5 contribution of duration of exposure to TCDD dose extrapolated to the time of exposure. 6 External comparisons to the German population using the SMR statistic also were 7 examined across dose categories. The noncancer causes of death examined by Ott and Zober 8 (1996, 198101) included all-cause mortality, diseases of the circulatory system, ischemic heart 9 disease, diseases of the digestive system, external causes, suicide, and residual causes of death. 10 Overall, no statistically significant differences in the SMR with the general population for 11 all-causes of death (SMR = 0.9, 95% CI = 0.7-1.1) were found. No statistically significant 12 differences were noted for any of the other causes of death examined. 13 Ott and Zober (1996, 198101) performed internal cohort comparisons using the Cox 14 regression model. These analyses found no dose-response patterns when cause-specific 15 mortality was examined across increasing cumulative TCDD exposure categories. Although an 16 inverse association for diseases of the respiratory system (SMR = 0.1, 95% CI = 0.0-0.8) was 17 detected, it was based only on 1 reported case. Many of these comparisons are limited by small 18 sample sizes as 92 deaths occurred in the cohort, and of these, 31 were from cancer. Also, the 19 third component of the cohort was identified primarily from former employees who were alive in 20 1986. As a result, the SMR based on the general population might be underestimated by the 21 exclusion of deceased workers. 22 23 2.4.1.2.1.2.1.2. Study evaluation. 24 As noted previously, caution should be exercised in the interpretation of SMR values of 25 noncancer outcomes as they could be influenced by the healthy worker effect. Although the 26 mechanism of identifying vital status appears to be excellent and unbiased, SMRs might be 27 underestimated for the cohort due to the manner in which they were constructed. Specifically, a 28 large component of the cohort was assembled by actively seeking out former workers who were 29 known to be alive in 1986. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-91 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.2.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 No dose-response patterns were observed between TCDD and the noncancer outcomes in 3 the Ott and Zober (1996, 198101) study. Therefore, dose-response modeling was not conducted. 4 5 2.4.1.2.1.3. T h e H a m b u r g coh ort. 6 2.4.1.2.1.3.1. Flesch-Janys et al. (1995, 197261). 7 2.4.1.2.1.3.1.1. Study summary. 8 Flesch-Janys et al. (1995, 197261) reported on the mortality experience of a cohort of 9 individuals employed by an herbicide-producing plant in Hamburg, Germany, covering the 10 period 1952 to 1992. As described in more detail in Section 2.4.1.1.1.3, the authors developed a 11 cumulative measure of TCDD using serum measures from 190 workers. This study also 12 examined the relationship between total TEQ and mortality. In the study population, the mean 13 TEQ without TCDD was 155 ng/kg, and for the mean TEQ including TCDD was 296.5 ng/kg. 14 Risks relative to the unexposed referent group of gas workers were estimated using Cox 15 regression across six exposed TCDD groups (i.e., the first four quintiles, and the ninth and 16 tenth deciles). A linear dose-response relationship was found with all causes of mortality and 17 cardiovascular mortality (p < 0.01). The RR for all cardiovascular deaths in the upper exposure 18 category was 1.96 (95% CI = 1.15-3.34), although there was no evidence of a linear 19 dose-response trend (p = 0.27). The dose-response relationship was most marked for ischemic 20 heart disease, with a RR of 2.48 (95% CI = 1.32-4.66) in the highest exposure group. A 21 dose-response relationship was also observed across TEQ groupings for all cause mortality, 22 cardiovascular disease mortality, and ischemic heart disease mortality. The authors did not 23 perform joint modeling of TEQ (without TCDD) and TCDD, so determining the extent that 24 dioxin-like compounds contributed to an increased risk of mortality is not possible. 25 26 2.4.1.2.1.3.1.2. Study evaluation. 27 The Flesch-Janys et al. (1995, 197261) study lacks information on other potential risk 28 factors for cardiovascular disease, which could result in confounding if those risk factors are also 29 related to TCDD exposure. Dose-response patterns were strong, however, and persisted across 30 numerous TCDD (and TEQ) exposure categories based on the use of an external reference group 31 (i.e., gas workers) or based on the internal comparison. The findings based on the internal This document is a draftfor review purposes only and does not constitute Agency policy. 2-92 DRAFT--DO NOT CITE OR QUOTE 1 comparison are noteworthy in that these groups should be more homogenous with respect to 2 confounding factors. As noted previously, the poor correlation between TCDD and smoking 3 among workers and similar smoking prevalence between the workers and the external gas 4 company workers suggest that smoking was not likely a confounder of the TCDD and 5 cardiovascular disease relationship. No other evaluation of noncancer mortality outcomes has 6 been undertaken in this cohort since 1995. 7 A strength of the Flesch-Janys et al. (1995, 197261) study was that it included the 8 collection of blood serum measures, which provided an objective measure of TCDD exposure. 9 Blood serum data, however, were obtained only for 16% of the cohort. The assumption of the 10 first-order kinetic elimination model is critical, given that measures were taken at the end of 11 follow-up. The model also assumed the half-life of TCDD was 6.9 years. If the kinetics are not 12 first order, or if the half-life estimate is inaccurate, estimates of TCDD levels during exposure 13 would be biased, particularly for workers having longer periods between exposure and PCDD 14 and PCDF assays. Sensitivity analyses completed by the authors suggest that such bias is not 15 likely to present because the results were unaffected when different model assumptions regarding 16 kinetic and half-lives were examined. The lack of an impact on RR estimates with varying 17 half-life estimates was similar to findings by Steenland et al. (2001, 197433). 18 19 2.4.1.2.1.3.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 20 Despite the aforementioned study strengths, the study focused on fatal outcomes such as 21 all cause mortality, cardiovascular disease mortality, and ischemic heart disease mortality. As 22 such, dose-response analysis was not conducted since these outcomes are not suitable for 23 development of an RfD. 24 25 2.4.1.2.1.4. T h e S ev e so W o m e n 's H e a lth S tu d y (S W H S ). 26 Eskenazi et al. (2000, 197162) presented an overview of the SWHS. The SWHS is the 27 first comprehensive epidemiologic study of the reproductive health of a female population 28 exposed to TCDD. The primary objective of the SWHS is to investigate the relationship of 29 TCDD and several reproductive endpoints, including endometriosis, menstrual cycle 30 characteristics, birth outcomes, infertility, and age at menopause. A second phase of follow-up This document is a draftfor review purposes only and does not constitute Agency policy. 2-93 DRAFT--DO NOT CITE OR QUOTE 1 that focuses on osteoporosis, thyroid hormone, breast cancer, diabetes, and metabolic syndrome 2 is expected to be completed in 2010. 3 Women were eligible for participation in the SWHS if they resided in Zones A and B (the 4 most contaminated areas) at the time of the explosion, were 40 years of age or younger at the 5 time of the explosion in 1976, and samples of their blood were collected and stored between 6 1976 and 1980. The enrollment of women in the SWHS began in March 1996 and continued 7 until July 1998. Of the 1,271 eligible women, 17 could not be found, 21 had died, and 12 were 8 too ill to participate. Of the 96% of the remaining women, 80% (n = 981) participated in the 9 study. Participation in the SWHS included a blood draw and an interview by a trained nurse who 10 was blind to subjects' TCDD level and zones of residence at the time of the accident. The 11 interview included detailed information on potential confounders including occupational, 12 medical, and reproductive, and pregnancy history. Also, women who were premenopausal were 13 asked to undergo a vaginal ultrasound and pelvic exam and to complete a daily diary on 14 menstruation. 15 Depending on the health outcome under study, TCDD exposures were characterized for 16 the women at different times. For example, TCDD exposure levels were estimated at the time of 17 the accident for some studies and at the time of conception for others. The SWHS study 18 population has been used to investigate associations between maternal TCDD levels and the 19 following health outcomes: menstrual cycle characteristics (Eskenazi et al., 2002, 197168); 20 endometriosis (Eskenazi et al., 2002, 197164); birth outcomes (Eskenazi et al., 2003, 197158); 21 age at menarche (Warner et al., 2004, 197490); age at menopause (Eskenazi et al., 2005, 22 197166); uterine leiomyomas (Eskenazi et al., 2007, 197170); and ovarian function (Warner 23 et al., 2007, 197486). An evaluation of the studies in chronological order is presented in this 24 section. 25 26 2.4.1.2.1.4.1. Eskenazi et al. (2002, 197168)--Menstrual cycle characteristics. 27 2.4.1.2.1.4.1.1. Study summary. 28 Eskenazi et al. (2002, 197168) evaluated serum TCDD exposures in relation to several 29 menstrual cycle characteristics in the SWHS. A total of 981 women who were 40 years of age or 30 younger at the time of the accident comprised the SWHS. The following exclusion criteria was 31 applied 44 years of age or older, women with surgical or natural menopause, those with Turner's This document is a draftfor review purposes only and does not constitute Agency policy. 2-94 DRAFT--DO NOT CITE OR QUOTE 1 syndrome, and those who in the past year had been pregnant, breastfed, or used an intrauterine 2 device or oral contraceptives. 3 A trained interviewer collected data on menstrual cycle characteristics using a 4 questionnaire. Women were asked to indicate how long their cycles were, whether the cycles 5 were regular (e.g., irregular cycle defined as length varied by more than 4 days), how many days 6 the menstrual flow lasted, and whether this flow was "scanty, moderate, or heavy." Information 7 was also collected on obstetric and gynecological conditions. TCDD exposures were derived 8 from serum samples collected in 1976-1985. The authors selected the earliest available serum 9 sample, and back-extrapolated to 1976 values using either the Filser model (Kreuzer et al., 1997, 10 198088) for women aged 16 years or younger in 1976 (n = 20) or the first-order kinetic model 11 (n = 6) (Pirkle et al., 1989, 197861). 12 Serum TCDD levels were transformed using the log10 scale, and the relationships 13 between these levels and length of menstrual cycle and days of menstrual flow were examined 14 using linear regression. The authors applied logistic regression to characterize the risk between 15 log10TCDD and heaviness of flow or regularity of cycle. In these analyses, moderate or heavy 16 flow and regular cycle were used as the reference categories. Stratified analysis was performed 17 by menarcheal status at the time of the accident. 18 Overall, the association with TCDD exposure (per 10-fold increase) and length of 19 menstrual cycle was not statistically significant for premenarcheal (P = 0.93, 95% CI = -0.01, 20 1.86) women or postmenarcheal women (P = -0.03, 95% CI = -0.61, 0.54). The corresponding 21 estimates found for days of menstrual flow were P = 0.18 (95% CI = -0.15, 0.51) and P = 0.16 22 (95% CI = -0.18, 0.50), respectively. Reduced flow was not associated with TCDD when 23 compared to moderate or heavy flow (odds ratio [OR] = 0.84, 95% CI = 0.44, 1.61); effect 24 modification by menarcheal status, however, was evident (p = 0.03). Specifically, women 25 exposed to TCDD who were premenarcheal had lower odds of reduced flow, while those 26 exposed to TCDD who were postmenarcheal did not. These findings counter the hypothesis that 27 TCDD exposure is related to ovarian dysfunction. Finally, statistically significant ORs were 28 found between serum TCDD levels (per 10-fold increase) and having an irregular cycle 29 (OR = 0.46, 95% CI = 0.23, 0.95). This inverse association was evident in both premenarcheal 30 women (OR = 0.50, 95% CI = 0.18, 1.38) and postmenarcheal women (OR = 0.41, 31 95% CI = 0.15, 1.16). This document is a draftfor review purposes only and does not constitute Agency policy. 2-95 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.1.2. Study evaluation. 2 Overall, the findings from the Eskenazi et al. (2002, 197168) study suggest that 3 exposures to TCDD can affect menstrual cycle characteristics among women who were exposed 4 before menarche. Exposures to TCDD were well characterized using serum samples available 5 on an individual-level basis, and the design allowed for the influence of other risk factor data to 6 be controlled for in regression analyses. Analysis of TCDD levels and the length of menstrual 7 cycle in premenarcheal women produced associations that were largely not statistically 8 significant at the alpha level of 0.05, but may have some biological significance. However, it is 9 unclear whether the endpoints that were measured constitute adverse health outcomes as they are 10 not definitive markers of ovarian dysfunction. Another source of uncertainty is measurement 11 error due to the subjective nature of menstrual flow reporting. Any resulting misclassification of 12 the outcome should be nondifferential, as the measurement error is unlikely to be dependent on 13 TCDD exposure. 14 15 2.4.1.2.1.4.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 16 The lack of a clear adverse health outcome related to TCDD exposure is a weakness of 17 this study. Although it is difficult to define the critical window of exposure for quantitative 18 exposure calculations, it can be estimated for the women that were premenarcheal at the time of 19 the accident as 13 years. Therefore, this study is suitable for further consideration for 20 quantitative dose-response modeling. 21 22 2.4.1.2.1.4.2. Eskenazi et al. (2002, 197164)--Endometriosis. 23 2.4.1.2.1.4.2.1. Study summary. 24 The SWHS provided the opportunity to investigate the association between serum TCDD 25 levels and endometriosis (Eskenazi et al., 2002, 197164). The rationale the authors provided for 26 undertaking this study was the experimental animal studies that suggested an association, the 27 high prevalence of endometriosis among infertile women where breast milk concentrations of 28 dioxin are high, and the unknown etiology of endometriosis. The study consisted of 601 women 29 who were younger than 30 years at the time of the Seveso accident. Stored sera that had been 30 collected between 1976 and 1980 were also available for these women. This document is a draftfor review purposes only and does not constitute Agency policy. 2-96 DRAFT--DO NOT CITE OR QUOTE 1 Given that laparoscopy could not be performed on women unless clinically indicated, no 2 "gold" standard was available for endometriosis diagnosis. Based on the results of a validation 3 study they conducted in a clinical population, the researchers classified women as having 4 endometriosis based on symptom report, gynecologic exam results, and vaginal ultrasound. 5 TCDD was measured in sera in 1976 for 93% of the women. Values for women whose 6 serum TCDD levels were collected after 1977 and had values exceeding 10 ppt were 7 back-extrapolated to 1976 using either the Filser model (<16 years of age) (Kreuzer et al., 1997, 8 198088) or a first-order kinetic model (>16 years) (Pirkle et al., 1989, 197861). These estimates 9 of TCDD were then modeled as both continuous (on a log scale) and categorical (<20, 20.1-100, 10 and >100 ppt) exposures. 11 Polytomous logistic regression was applied within the cohort used to generate RRs. In 12 relation to women in the lowest exposure category, the RR for endometriosis among women in 13 the middle and upper categories was 1.2 (90% CI = 0.3-4.5) and 2.1 (90% CI = 0.5-8.0), 14 respectively. The trend tests were not statistically significant for either the categorical (p = 0.25) 15 and continuous measures of TCDD (p = 0.84). 16 17 2.4.1.2.1.4.2.2. Study evaluation. 18 It is important to note that disease misclassification could have led to an underestimate of 19 the true risk of endometriosis if this misclassification was not differential with respect to TCDD 20 exposure. Also, younger women were likely to be under-represented as those who had never 21 been sexually active could not be examined due to cultural reasons. Other dioxin-like 22 compounds (PCDD, PCDFs, or polychlorinated biphenyls [PCBs]) were not considered because 23 of small serum volumes, but any potential TEQ exposures occurring in the population were 24 thought to be mostly attributable to TCDD in the exposed women. 25 26 2.4.1.2.1.4.2.3. Suitability o f datafor TCDD dose-response modeling. 27 Given that no statistically significant dose-response patterns were observed with either 28 log-transformed or across TCDD exposure categories, and that the elevated risks among those 29 with higher exposures had very wide confidence intervals (that included unity) quantitative 30 dose-response analyses were not recommended for this outcome. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-97 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.3. Eskenazi et al. (2003, 197158)--Adverse birth outcomes. 2 2.4.1.2.1.4.3.1. Study summary. 3 Eskenazi et al. (2003, 197158) examined the relationship between serum TCDD levels 4 and birth outcome measures. Analyses were based on 745 of the 981 women enrolled in the 5 SWHS who reported having been pregnant (n = 1,822). Most of these pregnancies 6 (888 pregnancies among 510 women) occurred after the accident. Analysis of spontaneous 7 abortions was restricted to 769 pregnancies among 476 women that did not end in abortion or in 8 ectopic or molar pregnancy. Congenital anomalies were evaluated for the 672 pregnancies that 9 did not end in spontaneous abortion. For the birth outcomes of fetal growth and gestational age, 10 analysis was performed using 608 singleton births from women without hypertensive pregnancy 11 disorders. 12 TCDD exposures were based on serum measures, most of which were taken shortly after 13 the accident. Serum was collected in 1976-1977 for 413 women, between 1978 and 1981 for 14 12 women, and in 1996 for 19 women. TCDD exposures based on serum samples collected from 15 1977 onward were back-extrapolated to 1976. 16 Statistical analyses were performed on pregnancies that ended between 1976 and the time 17 of interview. A continuous measure of log10TCDD (base 10 scale) was used to investigate 18 associations with adverse birth outcomes. Logistic regression was used to characterize the 19 relationship between TCDD exposure spontaneous abortions, small for gestational age, and 20 preterm birth (<37 weeks gestation). Linear regression was used to describe the relationship 21 between TCDD and birth weight (in grams) and gestational age (in weeks). 22 The risk estimates were adjusted for a series of characteristics that included sex of infant, 23 history of low birth weight child, maternal height, maternal body mass index, maternal 24 education, maternal smoking during pregnancy, and parity. No association was evident between 25 TCDD serum levels and spontaneous abortion for pregnancies between 1976 and 1998 26 (OR = 0.8, 95% CI = 0.6-1.2), or those between 1976 and 1984 (OR = 1.0, 95% CI = 0.6-1.6). 27 No statistically significant associations (ORs ranged from 1.2-1.8) were found between 28 log10TCDD levels and preterm delivery, small for gestational age. Although the mean change in 29 birth weight for pregnancies between 1976 and 1984 was fairly large (P = -92, 95% CI = -204 30 to 19), it also was not statistically significant at the alpha level of 0.05. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-98 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.3.2. Study evaluation. 2 This study was well-designed with well characterized exposures. Statistically significant 3 associations were not evident, although the birth-weight findings should be pursued with further 4 follow-up of the cohort. As the authors point out, those who were most vulnerable at the time of 5 the accident (the youngest) had not yet completed their childbearing years. While the study 6 lacked exposure data for the fathers, the authors indicated that only a small proportion were 7 believed to have high exposures to TCDD. The key limitation of the study was a reliance on 8 self-reported measures of pregnancy history, which may lead to some misclassification of the 9 birth outcomes. The observation that a large proportion of Seveso women had a voluntary 10 abortion because of fears of possible birth defects due to exposures from the accident suggest an 11 awareness bias is possible as a result of differential reporting of birth outcomes according to 12 exposure status. 13 14 2.4.1.2.1.4.3.3. Suitability o f da ta f o r TCDD dose-response modeling. 15 No statistically significant associations were found in the study; in addition, possible 16 awareness bias could have influenced the self-reported measures of birth outcomes. Therefore, 17 quantitative dose-response assessment was not considered for this study. 18 19 2.4.1.2.1.4.4. Warner et al. (2004, 197490)--Age at menarche. 20 2.4.1.2.1.4.4.1. Study summary. 21 Warner et al. (2004, 197490) examined the relationship between TCDD and age at 22 menarche in the SWHS cohort. As described earlier in this report, the SWHS comprised 23 981 participants. This study was restricted only to those who were premenarcheal at the time of 24 the accident (n = 282). The proportional hazards model was used to model TCDD exposures and 25 age at menarche. Age at menarche was determined by questionnaire administered by a trained 26 interviewer. Covariates examined as potential confounders included height, weight, body mass 27 index, athletic training at the time of interview, smoking, and alcohol consumption. 28 TCDD exposures were determined using serum samples collected from 257 of these 29 women between 1976 and 1977. For the remaining women, TCDD levels were quantified from 30 measures collected between 1978 and 1981 (n = 23) and in 1996 (n = 2). TCDD levels were 31 back-extrapolated to the time of the explosion in 1976. TCDD was modeled as both a This document is a draftfor review purposes only and does not constitute Agency policy. 2-99 DRAFT--DO NOT CITE OR QUOTE 1 continuous variable (log10TCDD) and a categorical variable based on quartile values (<55.9, 2 56-140.2, 140.3-300, >300 ppt). The lowest group was further subdivided into those with levels 3 <20, and >20 ppt; this cut-point represented background levels found in a sample of women 4 living in an unexposed area. 5 No association was found between the continuous measure of TCDD and age at 6 menarche (hazard ratio [HR] = 0.95, 95% CI = 0.83-1.09). Analyses restricted to those who 7 were younger than 8 in 1976 produced similar results (HR = 1.08, 95% CI = 0.89-1.30). 8 Additionally, no dose-response trend was observed with categorical measures of TCDD among 9 all women, as well as those under the age of 8. Although not statistically significant at the alpha 10 level of 0.05, TCDD exposures were later reported to be associated with age of menarche 11 (HR = 1.20, 95% CI = 0.98-1.60) when analyses were restricted to 84 women under the age of 5 12 at the time of the accident (Warner and Eskenazi, 2005). 13 14 2.4.1.2.1.4.4.2. Study evaluation. 15 An important strength of the Warner et al. (2004, 197490) study is the ability to 16 characterize TCDD exposures using serum samples that were collected shortly after the accident 17 occurred. The outcome of interest, age at menarche, was determined by asking women "At what 18 age did you get your first menstrual period?" Recent work suggests that self-reported measures 19 of age at menarche decades later have modest agreement with responses provided during 20 adolescence with recall varying by education and by history of an adverse birth outcome (Cooper 21 et al., 2005, 594401). In the Seveso study, bias would be introduced if recall varied according to 22 exposure levels. 23 24 2.4.1.2.1.4.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 25 Although the TCDD exposure characterization of study subjects was based on serum 26 data, and no major biases were introduced from the study design, the analyses produced largely 27 null associations. Therefore, quantitative dose-response assessment was not considered for this 28 study. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-100 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.5. Eskenazi et al. (2005, 197166)--Age at menopause. 2 2.4.1.2.1.4.5.1. Study summary. 3 Eskenazi et al. (2005, 197166) evaluated the relationship between age at onset of 4 menopause and serum levels of TCDD among women in the SWHS. Of the 981 women who 5 agreed to participate in SWHS, this analysis was restricted to those who had not reached natural 6 menopause before the time of the accident and who were at least 35 years of age at the time of 7 the interview. The recruitment and interview of women occurred approximately 20 to 22 years 8 after the accident (March 1996-July 1998). 9 The population was divided into quintiles of serum TCDD levels for the categorical 10 analysis. For most women (n = 564), TCDD levels were estimated from samples provided in 11 1976-1977. For the remaining women included in these analyses, TCDD levels were estimated 12 from samples collected between 1978 and 1982 (n = 28) and between 1996 and 1997 (n = 24). 13 As noted previously, exposure levels for women with post-1977 detectable levels of TCDD were 14 back-extrapolated to 1976 using either the first-order kinetic model (Pirkle et al., 1989, 197861) 15 (>16 years at time of accident) or the Filser model (<16 years at time of accident) (Kreuzer et al., 16 1997, 198088). Women were classified as premenopausal if they were still menstruating or if 17 they had amenorrhea as a result of pregnancy or lactation (at the time of interview) with an 18 indication of subsequent menstruation based on maintained diaries or further examination. 19 Subjects for which amenorrhea had persisted for at least 1 year with no apparent medical 20 explanation were classified into a natural menopause category. The category, surgical 21 menopause, pertained to women with a medically confirmed hysterectomy or an oophorectomy. 22 Finally, impending menopause was defined for subjects in which menstruation had been absent 23 for 2 months, but who provided evidence of subsequent menstruation, or had a secretory 24 endometrial lining, or indicated less predictable cycles in the previous 2-5 years. If participants' 25 menopausal status could not be determined, they were grouped into the "other" category. This 26 category included those for whom status could not be determined due to current use of oral 27 contraceptives, hormone replacement therapy, or previous cancer chemotherapy. 28 Statistical analysis was based on both a continuous measure of log-transformed TCDD 29 exposures and categories based on quintiles (<20.4 ppt; 20.4-34.2 ppt; 34.3-54.1 ppt; 30 54.2-118.0 ppt; >118.0 ppt). The Cox model was used to generate hazard ratios as estimates of 31 relative risks and their 95% confidence intervals examining natural menopause as the outcome. This document is a draftfor review purposes only and does not constitute Agency policy. 2-101 DRAFT--DO NOT CITE OR QUOTE 1 Several covariates previously identified as associated with menopausal status in the literature 2 were considered as potential confounders. These covariates included body mass index, physical 3 activity, premenopausal smoking, education, marital status, history of heart disease and other 4 medical conditions, and other reproductive characteristics. 5 The RRs were found to increase across the second through fourth quintiles (RRs = 1.1, 6 1.4, and 1.6, respectively) of serum TCDD categories in relation to those in the lowest category, 7 but not in the upper quintile (RR = 1.0, 95% CI = 0.6-1.8). A statistically significant test of 8 trend was detected across the first four quartiles (p = 0.04) but not across all five quintiles 9 (p = 0.44). A statistically significant association with onset of menopause was not detected 10 (RR = 1.02, 95% CI = 0.8-1.3) based on the logTCDD continuous measure. 11 12 2.4.1.2.1.4.5.2. Study evaluation. 13 The categorical exposure results from this study support a nonmonotonic 14 dose-related-association for earlier menopause with increased serum TCDD levels up to 15 approximately 100-ppt TCDD serum, but not above. Eskenazi et al. (2005, 197166) speculated 16 that the inverse "U" shape of the dose-response relationship is explained by the mimicking of 17 hormones at lower doses of a chemical, while at higher levels the toxic effect of a chemical does 18 not have the capacity to either inhibit or stimulate hormonal effects. 19 A study limitation is the potential for residual confounding due to adjustment based on 20 current smoking status and not at the time of onset of menopause. It is unclear to what extent 21 smoking status may differ between these two time periods and whether smoking is related to 22 TCDD exposures in this cohort. Exposures to other dioxin-like compounds were not considered 23 in this study because of small serum volumes, but any potential TEQ exposures occurring in the 24 exposed population were thought to be mostly attributable to TCDD in the exposed women. 25 26 2.4.1.2.1.4.5.3. Suitability o f datafor TCDD dose-response modeling. 27 To date, this study is the only one that has examined the relationship between TCDD 28 levels and onset of menopause. Although the findings suggest the possibility of a nonlinear 29 dose-response function, the log10TCDD exposure metric was not statistically significant, nor 30 were any category-specific hazard ratios statistically significant relative to the lowest category. 31 Therefore, a quantitative dose-response analysis was not undertaken. This document is a draftfor review purposes only and does not constitute Agency policy. 2-102 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.6. Warner et al. (2007, 197486)--Ovarian function. 2 2.4.1.2.1.4.6.1. Study summary. 3 Warner et al. (2007, 197486) investigated the association between serum TCDD levels 4 and ovarian function in subjects in the SWHS who were younger than 40 in 1976 and for whom 5 sera collected after the accident had been stored. These women were recruited from March 1996 6 until July 1998. Ovarian function analysis was limited to 363 women between 20 and 40 years 7 of age and who were not using oral contraceptives. Of these, 310 underwent transvaginal 8 ultrasound and were included in the functional ovarian cyst analysis. Ninety-six women were in 9 the preovulatory stage of their menstrual cycles and were included in the follicle analysis. For 10 the hormone analysis, 126 women who were in the last 2 weeks of their cycle were included. 11 The authors used logistic regression to examine the relationship between TCDD and the 12 prevalence of ovarian follicles greater than 10 mm. Linear regression models examined the 13 continuous outcome variables: number of ovarian follicles >10 mm and diameter of dominant 14 ovarian follicle. Covariates considered for inclusion in the model were age at ultrasound, age at 15 accident, age at menarche, marital status, parity, gravidity, lactation history, current body mass 16 index, age at last birth, and smoking history. For the serum hormone analyses, estradiol and 17 progesterone were measured in blood at the time of interview. Ovulation status was defined as a 18 dichotomous variable (yes/no) based on a serum progesterone cut-point value of 3 ng/mL. 19 The adjusted ORs across categories of TCDD exhibited no dose-response trend for the 20 presence of follicles in relation to TCDD in the follicular phase; also, no statistically significant 21 differences were noted in any of the upper exposure categories relative to those in the lowest. 22 The adjusted OR for the continuous measure of log10TCDD was 0.99 (95% CI = 0.4-2.2). A 23 similar nonstatistically significant finding was found for log10TCDD in relation to ovulation in 24 both the luteal (OR = 0.99, 95% CI = 0.5-1.9) and mid-luteal phases (OR = 1.03, 25 95% CI = 0.4-2.7). Analyses of progesterone and estradiol also were not related to serum 26 TCDD levels for either the luteal or mid-luteal phases (p = 0.51 and p = 0.47). 27 28 2.4.1.2.1.4.6.2. Study evaluation. 29 The investigators found no relationship between serum TCDD levels and serum 30 progesterone and estradiol levels among women who were in the luteal phase at the time of 31 blood draw. No association with number of ovarian follicles detected from ultrasound. This document is a draftfor review purposes only and does not constitute Agency policy. 2-103 DRAFT--DO NOT CITE OR QUOTE 1 Although no association was found, the authors suggested that the lack of significant results 2 could be because the women in SWHS were all exposed postnatally and the relevant and critical 3 time period for an effect might be in utero (animal studies support relevance of in utero 4 exposures). 5 6 2.4.1.2.1.4.6.3. Suitability o f da ta f o r TCDD dose-response modeling. 7 One limitation of the study was the lack of examination of confounding by dioxin-like 8 compounds. The absence of associations between TCDD and adverse health effects in this study 9 precludes conducting quantitative dose-response analyses. 10 11 2.4.1.2.1.4.7. Eskenazi et al. (2007, 197170)--Uterine leiomyoma. 12 2.4.1.2.1.4.7.1. Study summary. 13 Associations between TCDD exposures and uterine leiomyoma (i.e., fibroids) were also 14 examined among 956 women in the SWHS (Eskenazi et al., 2007, 197170). The sample 15 population was based on the on the original 981 SWHS participants excluding 25 women 16 diagnosed with fibroids before the date of the accident (July 10, 1976). Women who previously 17 had fibroids were identified both through the administered questionnaire and the review of 18 medical records. Transvaginal ultrasounds were performed for 634 women to determine if they 19 had fibroids at the time of follow-up. Similar to other SWHS studies, exposure to TCDD was 20 estimated using serum collected from women shortly after the time of the accident, between 21 1978 and 1981 and in 1996. TCDD levels were back-extrapolated to 1976 levels. 22 The study authors performed statistical analyses using two definitions of fibroids as 23 outcome measures. The first was fibroids detected before the study, and the second was fibroids 24 detected via ultrasound. A proportional odds method Dunson and Baird (2001, 197248) 25 developed was used to model the cumulative odds of onset of fibroids. This method combines 26 historical and current information of diagnoses of fibroids. Continuous and categorical measures 27 of TCDD were modeled. Regression models were adjusted for known or suspected risk factors 28 of fibroids including parity, family history of fibroids, age at menarche, body mass index, 29 smoking, alcohol use, and education. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-104 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.4.7.2. Study evaluation. 2 Categorical measures of TCDD suggested an inverse dose-response relationship with the 3 onset of fibroids. Relative to those with TCDD levels less than 20 ppt, those having TCDD 4 exposures between 20.1 and 75.0 ppt and greater than 75.0 ppt had RRs of 0.58 5 (95% CI = 0.41-0.81), and 0.62 (95% CI = 0.44-0.89), respectively. The continuous measure of 6 logi0TCDD produced a hazard ratio of 0.83 (95% CI = 0.65-1.07). 7 8 2.4.1.2.1.4.7.3. Suitability o f da ta f o r TCDD dose-response modeling. 9 The inverse association between TCDD and uterine fibroids supports the possibility of an 10 anti-estrogenic effect of TCDD. The observed direction of the reported associations precludes 11 quantitative dose-response modeling. 12 13 2.4.1.2.1.5. O th e r S eveso n o n ca n cer stu d ies. 14 2.4.1.2.1.5.1. Bertazzi et al. (1989, 197013); Consonni et al. (2008, 524825)--Mortality 15 outcomes. 16 2.4.1.2.1.5.1.1. Study summary. 17 Several studies have evaluated the mortality of Seveso residents exposed to TCDD 18 following the 1976 accident. The earlier section of this report described the designs of these 19 studies and discussed their findings as they relate to cancer mortality. In this section, some of 20 the findings for other causes of death are described. A key feature of these studies is that 21 patterns of mortality among Seveso residents were investigated according to their zone of 22 residence at the time of explosion relative to general population rates. 23 A 10-year mortality follow-up of residents of Seveso was published in 1989 (Bertazzi 24 et al., 1989, 197013). Poisson regression was used to derive RRs for those who had lived in 25 Zone A at the time of explosion using a referent group consisting of inhabitants who had lived in 26 the uncontaminated study area. Between 1976 and 1986, no statistically significant difference 27 was observed in all-cause mortality relative to the general population among those who lived in 28 the most highly exposed area (Zone A) at the time of the accident. This finding was evident in 29 both males (RR = 0.86, 95% CI = 0.5-1.4) and females (RR = 1.14, 95% CI = 0.6-2.1). A 30 statistically significant excess in circulatory disease mortality was found among males relative to 31 those in the referent population (RR = 1.75, 95% CI = 1.0-3.2); this increased risk was more This document is a draftfor review purposes only and does not constitute Agency policy. 2-105 DRAFT--DO NOT CITE OR QUOTE 1 pronounced when the follow-up period was restricted to the first 5 years after the accident 2 (1976-1981) (RR = 2.04, 95% CI = 1.04-4.2). Between 1982 and 1986, the RR decreased 3 substantially and was not statistically significant (RR = 1.19, 95% CI = 0.4-3.5). Among 4 females, a risk similar in magnitude was detected for circulatory disease mortality although it 5 was not statistically significant (RR = 1.89, 95% CI = 0.8-4.2). Contrary to the calendar 6 period-specific findings for males, the excess of circulatory mortality among females occurred 7 between 1982 and 1986 (RR = 2.91, 95% CI = 1.1--7.8) and not between 1976 and 1981 8 (RR = 1.12, 95% CI = 0.3--4.5). The number of deaths in this cohort with the 10 years of 9 follow-up was relatively small; in Zone A, 16 deaths were observed among males and 11 among 10 females. 11 The most recently published account of the mortality experience of Seveso residents 12 provides further information on follow-up of these residents until the end of 2001 (25 years after 13 the accident) (Consonni et al., 2008, 524825). Three exposure groups were considered: Zone A 14 (very high contamination), Zone B (high contamination), and Zone R (low contamination). The 15 reference population consisted of those residents who lived in unaffected surrounding areas, as 16 well as residents of five nearby towns. The authors used Poisson regression to compare 17 mortality rates for each zone relative to the reference population. 18 For all causes of death, no excess was found in Zone A, B, or R relative to the reference 19 population. Statistically significant excesses were noted for those who lived in Zone A relative 20 to the reference population for chronic rheumatic heart disease (RR = 5.74, 21 95% CI = 1.83--17.99) and chronic obstructive pulmonary disease (RR = 2.53, 22 95% CI = 1.20--5.32). These risks, however, were based on only 3 and 7 deaths, respectively. 23 For those in Zone A, no statistically significant excesses in mortality were noted for diabetes, 24 accidents, digestive diseases, ischemic heart disease, or stroke. Among Zone A residents, 25 stratified analysis by time since accident showed increased rates of circulatory disease 5--9 years 26 since the accident (RR = 1.84, 95% CI = 1.09--3.12). Increased mortality from diabetes relative 27 to the reference population was noted among females who lived in Zone B (RR = 1.78, 28 95% CI = 1.14--2.77). 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-106 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.5.1.2. Study evaluation. 2 The ascertainment of mortality in this cohort is nearly complete. Misclassification of 3 some health outcomes, such as diabetes, may occur due to use of death certificate data. 4 The characterization of exposure is based on zone of residence. Soil sampling indicated 5 considerable variability in TCDD soil levels, and therefore, the generation of risks based on zone 6 of residence likely does not accurately reflect individual exposure. Exposure misclassification 7 might also occur because residency in the areas does not necessarily reflect whether the 8 individual would have been present in the area at the time the accident occurred. Any exposure 9 misclassification would likely be nondifferential which would tend to bias the risk estimates 10 towards the null. 11 Although some excess of circulatory disease mortality was found, the finding was not 12 consistent between men and women. Moreover, excess circulatory disease mortality was more 13 pronounced among men within the first 5 years of exposure, while, for women, the excess was 14 more pronounced in years 5-10. Numerous other risk factors for circulatory disease were not 15 controlled for in these analyses and may be confounders if related to TCDD exposure. Taken 16 together, the possibility that TCDD increased circulatory disease mortality based on these data is 17 tenuous at best. 18 19 2.4.1.2.1.5.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 20 There is considerable uncertainty in these data due to the potential for outcome and 21 exposure misclassification. The lack of the individual-level TCDD levels and the examination of 22 fatal outcomes reported in this study are not a suitable basis for development of an RfD. For 23 these reasons, dose-response analysis for this outcome is not conducted. 24 25 2.4.1.2.1.5.2. Mocarelli et al. (1996, 197637; 2000, 197448)-- Sex ratio. 26 2.4.1.2.1.5.2.1. Study summary. 27 A letter to the editor was the first report of a possible change in the sex ratio from dioxin 28 among Seveso residents following the July 10, 1976 accident (Mocarelli et al., 1996, 197637). 29 The authors reported that 65% (n = 48) of the 74 total births that had occurred from April 1977 30 to December 1984 were females. This male to female ratio of 26:48 (35%) is significantly 31 different from the worldwide birth ratio of 106 males to 100 females (51%) (James, 1995, This document is a draftfor review purposes only and does not constitute Agency policy. 2-107 DRAFT--DO NOT CITE OR QUOTE 1 197722). Between 1985 and 1994, the Seveso male to female ratio leveled out at 60:64 (48%). 2 The authors suggested that the finding supported the hypothesis that dioxin might alter the sex 3 ratio through several possible mechanistic pathways. 4 Mocarelli et al. (2000, 197448) later reported on an investigation between serum-based 5 TCDD measures in parents and the sex ratio of offspring. In this study, serum samples were 6 collected from mothers and fathers who lived in the areas at the time of the explosion, were 7 between the ages of 3 and 45 at the time of the explosion, and produced offspring between 8 April 1, 1977 and December 31, 1996. The study population included 452 families and 9 674 offspring, and serum measures were available for 296 mothers and 239 fathers. An estimate 10 of TCDD at the time of conception was also examined in relation to male to female birth ratios. 11 TCDD exposure estimates between the years of 1976 and 1996 were estimated using Filser's 12 model (Kreuzer et al., 1997, 198088). 13 Mocarelli et al. (2000, 197448) used chi-square test statistics to compare observed sex 14 ratio to an expected value of 0.51 in this Seveso population. Concentrations of TCDD were 15 modeled as categorical variables in several ways. First, a dichotomous variable was used 16 whereby unexposed parents were defined as those who lived outside Zones A, B, and R or had a 17 serum TCDD concentration of less than 15 ppt; parents with exposures of 15 ppt or higher were 18 considered exposed. Second, a trichotomous exposure variable was created that consisted of 19 parents who (1) lived outside Zones A, B, and R or had serum concentrations of less than 15 ppt, 20 (2) had serum concentrations of 15-80 ppt, and (3) had serum concentrations that exceeded 21 80 ppt. These cut-points were chosen as they represented tertiles based on the distribution of 22 TCDD among parents. Analyses were conducted separately for paternal and maternal TCDD 23 levels. 24 The overall proportion of 0.49 male births (based on male to female ratio of 328:346) was 25 not significantly different from the expected proportion of 0.51 (p > 0.05). Statistically 26 significant differences were found, however, if both parents had TCDD levels >15 ppt (sex 27 ratio = 0.44) or just the father had serum TCDD levels >15 ppt (sex ratio = 0.44). No 28 statistically significant differences were found when the fathers had TCDD levels less than 29 15 ppt, irrespective of the maternal levels. A dose-response pattern in the sex ratio was found 30 across the paternal exposure categories. That is, the sex ratio decreased with increased paternal 31 TCDD levels (linear test for trend, p = 0.008). In the unexposed group, the sex ratio (male to This document is a draftfor review purposes only and does not constitute Agency policy. 2-108 DRAFT--DO NOT CITE OR QUOTE 1 female) was 0.56 (95% CI = 0.49-0.61), while in the highest exposure group 2 (281.0-26,400.0 ppt) the corresponding sex ratio was 0.38 (95% CI = 0.28-0.49). 3 Stratified analyses by age at paternal exposure revealed that the sex ratio was altered to a 4 greater degree among fathers who were younger than 19 at the time of the explosion. The male 5 to female ratio among the unexposed fathers was 0.56 (95% CI = 0.50-0.62), while it was 0.38 6 (95% CI = 0.30-0.47) for those younger than 19 when exposed and 0.47 (95% CI = 0.41-0.53) 7 for those exposed after 19. Regardless of the age at the time of exposure, however, fathers who 8 were exposed had a statistically significantly different birth ratio (they were more likely to father 9 girls) than those who were unexposed (p < 0.05). 10 Separate analysis of birth ratios based on paternal TCDD exposure estimated at the time 11 of conception did not show the same dose-response pattern but did show strong evidence of 12 consistently decreased male births relative to females. More specifically, the male to female 13 birth ratios among the four successive quartiles (first through fourth) were 0.41, 0.33, 0.33, 14 and 0.46. 15 16 2.4.1.2.1.5.2.2. Study evaluation. 17 Mocarelli et al. (2000, 197448) based the characterization of TCDD exposure on serum 18 samples, which is an objective method for characterizing dose. Unlike for the occupational 19 cohorts, serum measures for this study were taken close to the time of the accident, and 20 therefore, back-extrapolation of TCDD exposures is unnecessary. Exposure received before the 21 age of 19 at the time of the explosion were more strongly associated with a reduced male to 22 female ratio than those received after the age of 19. The cut off age of 19 seems to be somewhat 23 arbitrary, resulting in a highly uncertain critical exposure window. TCDD levels at the time of 24 conception did not demonstrate a dose-response relationship, but paternal exposures resulted in 25 consistently reduced male to female birth ratios (range: 0.33-0.46). 26 The study findings are unlikely to be influenced by age at conception as these values 27 were found, on average, to be similar across calendar years. This suggests that age at conception 28 was not an important confounder and that the birth ratio findings may be related to paternal 29 exposures. 30 The methods used to identify births appear to be appropriate. Even if some 31 under-ascertainment of births occurred, there is no reason to believe that ascertainment would be This document is a draftfor review purposes only and does not constitute Agency policy. 2-109 DRAFT--DO NOT CITE OR QUOTE 1 related to TCDD exposure and the sex of the baby. Therefore, no bias is suspected due to 2 incomplete birth ascertainment. 3 4 2.4.1.2.1.5.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 5 TCDD exposures were well-characterized, and internal cohort analyses demonstrate 6 association between paternal TCDD levels at the time of the accident and birth ratio. However, 7 the change in sex ratio was only statistically significant when exposure occurred before 19 years 8 of age. It is impossible to identify the relevant time interval over which TCDD dose should be 9 considered for dose-response analysis; specifically, it is difficult to discern whether the different 10 sex ratio is a consequence of the initial peak exposure before 19 years of age or a function of the 11 average cumulative exposure over this entire exposure window. Assuming the initial high 12 exposure is the correct exposure window, using the initial exposures in a dose-response model 13 would yield LOAELs that are too high to be relevant to factor into the RfD calculation. The 14 differences between the two dose estimates are quite large. Dose-response analysis for this 15 outcome, therefore, was not conducted. 16 17 2.4.1.2.1.5.3. Baccarelli et al. (2002, 197062; 2004, 197045)--Immunologic effects. 18 2.4.1.2.1.5.3.1. Study summary. 19 The relationship between TCDD and immunological effects was evaluated in a sample of 20 Seveso residents (Baccarelli et al., 2002, 197062; Baccarelli et al., 2004, 197045). Both studies 21 were based on findings from 62 individuals who were randomly selected from Zones A and B. 22 An additional 59 subjects were chosen from the surrounding noncontaminated areas. Residency 23 was based on where subjects lived at the time of the accident (July 10, 1976) (Landi, 1998, 24 594409). Frequency matching ensured that the two groups of subjects were similar with respect 25 to age, sex, and cigarette smoking status. 26 TCDD levels were determined by mass spectrometric analysis of plasma samples. 27 TCDD levels at the time of sampling were obtained, and estimates of levels at the time of the 28 accident also were estimated by assuming an 8.2-year half-life (Landi, 1998, 594409). The 29 plasma was also used to characterize levels of the immunoglobulins (Ig) IgG and IgM and the 30 complement components C3 and C4. One subject was excluded due to lack of an immunological This document is a draftfor review purposes only and does not constitute Agency policy. 2-110 DRAFT--DO NOT CITE OR QUOTE 1 evaluation. Analyses are, therefore, based on 58 subjects in the noncontaminated areas and 2 62 individuals from the contaminated areas. 3 Nonparametric tests were applied to test for differences between the two groups. 4 Multiple regression also was used to describe the relationship between the variables. Adjustment 5 was made for several potentially confounding variables that were collected via a questionnaire. 6 An inverse association was noted with increasing TCDD levels and plasma IgG levels; 7 this result remained statistically significant after adjusting for other potential confounding 8 variables in the regression models. Specifically, the slope coefficient and p-value for the 9 unadjusted model were -0.35 (p = 0.0002) and for the adjusted model the p-value was 0.0004. 10 The authors did not present the slope coefficient for the adjusted model in either paper but noted 11 minimal differences between the adjusted and unadjusted results. In the 2004 analysis, the 12 authors present IgG, IgM, IgA, C3, and C4 median and interquartile values across TCDD 13 exposure quintiles. Decreased levels of IgG were observed in the highest exposure groups. 14 Specifically, the median values across the five quintiles (for lowest to highest) were 1,526; 15 1,422; 1,363; 1,302; and 1,163. The Kruskal-Wallis test for differences across the TCDD 16 categories was statistically significant (p = 0.002), which is consistent with the findings for the 17 continuous measures of TCDD. This finding persisted after excluding those subjects with 18 inflammatory diseases and those who used antibiotics or nonsteroidal anti-inflammatory drugs. 19 For the other plasma measures, no dose-response relationship was apparent based on median 20 values for IgM, IgA, C3, or C4 across TCDD quintiles. The authors highlight the need for 21 additional research, particularly given the excess of lymphatic tumors noted in the area. 22 Exposure to other dioxin-like compounds for both the TCDD and nonexposed areas were 23 reported to be at background levels. 24 25 2.4.1.2.1.5.3.2. Study evaluation. 26 Both TCDD exposure and health outcome measures are well characterized. TCDD 27 exposures, in particular, are based on current serum measures and, therefore, are not dependent 28 on assumptions needed to back-extrapolate to earlier time periods of exposure. 29 A dose-response relationship between TCDD and IgG is well documented for the 30 unadjusted model, but no details are provided on the change in the slope coefficient when other 31 covariates were added to the model. This document is a draftfor review purposes only and does not constitute Agency policy. 2-111 DRAFT--DO NOT CITE OR QUOTE 1 Interpreting the inverse association between TCDD exposure and IgG in terms of clinical 2 significance is not possible. The IgG values reported are much higher than those subjects with 3 antibody immunodeficiency disorders. 4 5 2.4.1.2.1.5.3.3. Suitability o f da ta f o r TCDD dose-response modeling. 6 Although the data support an inverse dose-response association between IgG and TCDD, 7 because the relationship cannot be described in terms of clinical relevance with respect to a 8 specific adverse health outcome, these data were not suitable for quantitative dose-response 9 modeling. 10 11 2.4.1.2.1.5.4. Landi et al. (2003, 198362)--Gene expression. 12 2.4.1.2.1.5.4.1. Study sum m ary 13 The impact of TCDD on the aryl hydrocarbon receptor (AhR) was evaluated by Landi 14 et al. (2003, 198362) in a population-based study of Seveso residents. AhR, a mechanistically 15 based biomarker of dioxin response, must be present for manifestation of most of the toxic 16 effects of TCDD, including tumor promotion and immunological and reproductive system effects 17 (Safe, 1986; Puga et al., 2000). AhR activates the transcription of several metabolizing enzymes 18 in addition to certain genes (Whitlock, 1999). The primary objective of the study was to 19 determine whether plasma levels of TCDD and TEQ are associated with the AhR-dependent 20 pathway in lymphocytes among Seveso residents. The genes involved in the pathway that were 21 examined included: AhR, aryl hydrocarbon receptor nuclear translocator, CYPA1A1 and 22 CYP1B1 transcripts, and CYP1A1-associated 7-ethoxyresorufin O-deethylase (EROD). 23 Study recruitment occurred from December 1992 to March 1994. A total of 62 subjects 24 were randomly chosen from the highest exposed zones in Seveso (Zones A and B), while 59 25 were chosen from the noncontaminated area (non-ABR). Those chosen from the 26 noncontaminated zone were matched by age, sex, and smoking. Assignment of zones was based 27 on place of residence where subjects lived at the time of the accident in 1976. Subjects provided 28 data via questionnaire on a variety of sociodemographic and behavioral risk factors, including 29 cigarette smoking. Multivariate models were adjusted for a variety of confounders including; 30 adjustment for age, gender, date of assay, actin expression, postculture viability, experimental 31 group, and cell growth. This document is a draftfor review purposes only and does not constitute Agency policy. 2-112 DRAFT--DO NOT CITE OR QUOTE 1 TCDD levels were determined using high-resolution gas chromatography, and 21 other 2 dioxins, or dioxin-like compounds, were measured to examine TEQ. Eleven measurements 3 taken on the 121 subjects were deemed inadequate and excluded, but no further information was 4 provided on these exclusions. Nine subjects from Zone B and fourteen subjects from Zone ABR 5 had TCDD levels below that of detection, and were assigned a value equal to the lipid-adjusted 6 detection limit divided by the square root of 2. The toxic equivalent for the mixture of 7 dioxin-like compounds (i.e., TEQ) was calculated by summing the products of the concentration 8 of each congener by its specific toxic equivalency factor. 9 The subjects provided between 5 and 50 mL of whole blood, which was centrifuged to 10 separate mononuclear cells. The cells were frozen and later thawed. Cells were cultured, 11 removed from the culture medium, and resuspended in a stimulation medium, 14 mL of which 12 was used for RNA analysis. Reverse transcription-PCR was conducted and EROD was assayed. 13 Differences in gene expression and EROD activity observed for various cell culture conditions 14 were compared using paired t-tests. The unpaired Student's t-test was applied to test for 15 differences between groups, while a Bonferroni factor was used to account for multiple 16 comparisons. Data for continuous variables were log-transformed. 17 TCDD accounted for 26% of the TEQ among the study subjects, but varied by zone (35% 18 in zone A and 18% in zone non-ABR). After adjusting for potential confounding, AhR was 19 inversely related to plasma TCDD levels in uncultured cells (p < 0.03) and in mitogen-stimulated 20 cells (p < 0.05). EROD was lower in cells cultured from subjects with higher plasma TCDD and 21 TEQ levels, and the corresponding continuous measure of EROD was statistically significant 22 (p < 0.05). No statistically significant associations with TCDD or TEQ were found with ARNT 23 or CYP1B1 in uncultured cell medium, nor with CYP1A1 or CYP1B1 in mitogen-stimulated 24 cells. In general, females had lower AhR transcripts and higher levels of dioxin. 25 Collectively, the findings suggest that TCDD exposure might reduce AhR expression in 26 unstimulated cells. Therefore, TCDD could exert an influence on the AhR pathway regulation. 27 28 2.4.1.2.1.5.4.2. Study evaluation. 29 The study used biologically based measures of both TCDD exposures and biomarkers or 30 AhR. Subject recruitment was based on randomly sampling of the cohort study population; 31 some individuals with severe medical illnesses were excluded (Landi, 1998, 594409). Although This document is a draftfor review purposes only and does not constitute Agency policy. 2-113 DRAFT--DO NOT CITE OR QUOTE 1 few details are provided on the number of subjects excluded for these reasons, given the 2 objective nature of the biomarker outcomes that were evaluated, such exclusions are unlikely to 3 be an important source of bias. The exclusion rates were also reported to be low and comparable 4 across the zones (five subjects from the noncontaminated zone non-ABR and four subjects from 5 zone B). 6 A strength of the study was the examination of other dioxin-like compounds via the TEQ 7 analysis. A limitation of the study included the relatively small number of subjects which 8 resulted in the grouping of several covariates, including TCDD exposures, into a small number 9 of categories. As such, slope coefficients derived from modeling continuous measures were 10 emphasized in the data presentation. Another key limitation of the study is the uncertainty of 11 how effects on AhR translate into subsequent development of cancer and other chronic health 12 effects. 13 14 2.4.1.2.1.5.4.3. Suitability o f da ta f o r TCDD dose-response modeling. 15 It is unclear how associations between AhR biomarkers and TCDD levels translate into 16 an increased risk of cancer. Dose-response analysis for this outcome, therefore, was not 17 conducted. 18 19 2.4.1.2.1.5.5. Alaluusua et al. (2004, 197142)--Developmental dental effects. 20 2.4.1.2.1.5.5.1. Study summary. 21 Alaluusua et al. (2004, 197142) examined the relationship between TCDD and dental 22 defects, dental caries, and periodontal disease among Seveso residents who were children at the 23 time of the accident. Subjects were randomly selected from those individuals who had 24 previously provided serum samples in 1976, which was shortly after the accident. A total of 25 65 subjects who were less than 9.5 years of age at the time of the accident, and who lived in 26 Zones A, B, or R were invited to participate. Recruitment was initiated 25 years after the time of 27 the Seveso accident. An additional 130 subjects from the surrounding area (outside Zones A, B, 28 or R or "non-ABR zone") having the same age restriction were recruited. Subjects were 29 frequency matched for age, sex, and education. Questionnaires were administered to these 30 individuals to collect detailed information on dental and medical histories, education, and 31 smoking behaviors. Ten subjects who had completed at least high school were randomly This document is a draftfor review purposes only and does not constitute Agency policy. 2-114 DRAFT--DO NOT CITE OR QUOTE 1 excluded from the non-ABR zone to create groups with similar educational profiles. 2 Participation rates for the ABR and non-ABR zones were 74 and 58%, respectively. 3 One dentist who was blind to the patients' TCDD exposure levels assessed dental 4 aberrations. Dental caries was assessed using recommendations of the World Health 5 Organization. Periodontal status was described following a detailed evaluation of the surfaces of 6 the teeth. A radiographic examination was done to identify missing teeth, alveolar bone loss, 7 deformities in the roots, and jaw cysts. 8 Comparisons of the presence of dental enamel defects according to exposure status were 9 performed using logistic regression. Chi-square test statistics were applied to compare the 10 distributions in the prevalence of dental defects across several categorical covariates (i.e., 11 education, age, and serum TCDD level). For those who were younger than 5 at the time of the 12 accident, dental defects were more prevalent among patients in zone ABR (42%) than those in 13 the non-ABR zone (26%) (p = 0.14). Zone ABR is characterized by higher levels of soil TCDD 14 levels relative to non-ABR. Serum levels permitted an improved characterization of risk as they 15 were available at an individual level, rather than using a zone of residence. Defect prevalence 16 was highest among those in the upper serum TCDD category (700-26,000 ng/kg) with 60% of 17 subjects having dental defects. The continuous measure of serum TCDD was associated with 18 developmental dental defects (p = 0.007) and hypodontia (p = 0.05). 19 20 2.4.1.2.1.5.5.2. Study evaluation. 21 Although the subjects with serum measures were selected randomly, no direct measures 22 of TCDD were made in subjects from the unexposed area (i.e., non-ABR zones). That those who 23 resided in the non-ABR areas had lower TCDD exposures would be a reasonable assumption. 24 Alaluusua et al. (2004, 197142), however, provide few details about the sampling frame used to 25 identify these participants. Despite this fact, it is important to note that a dose-response pattern 26 was observed between TCDD exposure and presence of developmental defects in the ABR 27 population alone (p = 0.016). This finding is based on 27 subjects with developmental dental 28 defects. This positive association provides support for a quantitative dose-response modeling of 29 dental aberrations. The numbers of such subjects are small, however, with one, five, and 30 nine subjects having defects in the exposure groups of 31-226, 238-592, and 31 700-26,000 ng/kg TCDD, respectively. This document is a draftfor review purposes only and does not constitute Agency policy. 2-115 DRAFT--DO NOT CITE OR QUOTE 1 TCDD exposures were characterized using serum measures for those who resided in 2 zone ABR in 1976 (near the time of the accident). The authors could not account for additional 3 exposure to TCDD across subjects that might have occurred since the time of the accident, so 4 there is considerable uncertainty in delineating the critical exposure window for the reported 5 effects. In addition, the lack of exposure data for those in the non-ABR zone, however, makes 6 interpretation of the findings difficult. This difficulty is particularly evident, given that the 7 prevalence of dental defects was less among those in the low exposure category of zone ABR 8 (31-226 ng/kg TCDD) (10%) when compared to those in the non-ABR zone (26%). 9 10 2.4.1.2.1.5.5.3. Suitability o f da ta f o r TCDD dose-response modeling. 11 Most o f the considerations for conducting a dose-response analysis have been satisfied 12 with the study population, although, exposure assessment uncertainties are a limitation of this 13 study. For example, it is difficult to discern whether these health effects are a consequence of 14 the initial high exposure during childhood or a function of the cumulative exposure for this entire 15 exposure window beginning at the early age. If the latter is true, averaging exposure over the 16 critical window would add considerable uncertainty to effective dose estimates given the large 17 difference between initial TCDD body burden and body burden at the end of the critical 18 exposure window. Despite the uncertainty in defining the critical window of exposure, 19 dose-response analysis was conducted for this outcome. 20 21 2.4.1.2.1.5.6. Baccarelli et al. (2005, 197053)--Chloracne. 22 2.4.1.2.1.5.6.1. Study summary. 23 Baccarelli et al. (2005, 197053) published findings from a case-control study of 24 110 chloracne cases and 211 controls. The authors collected information on pigment 25 characteristics and an extensive list o f diseases. This study was performed to yield information 26 about the health status of chloracne cases, TCDD-chloracne exposure response, and factors that 27 could modify TCDD toxicity. TCDD was measured from plasma. Following adjustment for 28 confounding, TCDD was associated with chloracne (OR = 3.7, 95% CI = 1.5-8.8), and the risk 29 o f chloracne was considerably higher in subjects younger than 8 at the time of the accidents 30 (OR = 7.4, 95% CI = 1.8-30.3). Among individuals with lighter hair, the association between 31 TCDD and chloracne was stronger than among those with darker hair. This document is a draftfor review purposes only and does not constitute Agency policy. 2-116 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.5.6.2. Study evaluation. 2 Although a dose-response association was observed, chloracne is a rare health outcome 3 likely only to occur among those highly exposed. 4 5 2.4.1.2.1.5.6.3. Suitability o f da ta f o r TCDD dose-response modeling. 6 Given the very high TCDD levels needed to cause chloracne (e.g., Ott et al., 1993, 7 594322)), quantitative dose-response modeling to characterize risks for the general population 8 with much lower TCDD exposures would be of little value. Therefore, quantitative 9 dose-response assessment for the Baccarelli et al. (2005, 197053) study was not conducted. 10 11 2.4.1.2.1.5.7. Baccarelli et al. (2008, 197059)--Neonatal thyroid hormone levels. 12 2.4.1.2.1.5.7.1. Study summary. 13 Baccarelli et al. (2008, 197059) investigated the relationship between thyroid function 14 and TCDD among offspring of women of reproductive age who were exposed in the 15 1976 accident. This health endpoint is relevant because thyroid function is important for energy 16 metabolism and nutrients and for stimulating growth and development of tissues. Neonatal 17 thyroid function at birth is evaluated through blood thyroid-stimulating hormone (b-TSH). 18 The study population was drawn from 1,772 women who were identified as having lived 19 in the highly contaminated areas (Zones A or B) at the time of the accident or between 20 July 10, 1976 and December 31, 1947; were of fertile age (born after 1947); and were alive as of 21 January 1, 1994. A random sample of 1,772 unexposed women who lived in the reference area 22 was selected using frequency matching by year of birth to the exposed women, and residency in 23 the reference area at the time of the accident. The reference area represents the noncontaminated 24 areas that surround the three zones of decreasing exposure (Zones A, B and R). In total, 25 55,576 women had lived in the reference area. Population registry offices (n = 472) were 26 contacted to detect children born to these women. Records could be traced for virtually all 27 subjects (1761/1772 exposed; 1762/1772 unexposed). Children born outside the Lombardy area 28 were excluded as b-TSH could not be obtained for them. This accounted for 156 of the 29 1,170 children identified. The analyses were based on the remaining 56, 425, and 533 singletons 30 born between January 1, 1994, and June 30, 2005 in Zone A, B, and from the reference area, 31 respectively. This document is a draftfor review purposes only and does not constitute Agency policy. 2-117 DRAFT--DO NOT CITE OR QUOTE 1 Thyroid function is tested in all newborns by b-TSH measures in the region of Lombardy 2 where Seveso is located. These measures are obtained from blood samples taken 72 hours after 3 birth using a standardized protocol. The b-TSH levels were log transformed to approximate a 4 normal distribution. Linear regression analysis was used to conduct test for trends in mean 5 b-TSH levels across different covariates. Logistic regression was used to assess associations 6 between elevated b-TSH levels defined by the cutpoint of 5 pU/mL and residence in particular 7 zones of contamination. The 5 pU/mL cutpoint for TSH measurements in neonates was 8 recommended by WHO (1994) for use in neonatal population surveillance programs. Although 9 WHO established the standard for increased neonatal TSH in the context of iodine deficiency 10 disease, the toxicological implications are the same for TCDD exposure and include increased 11 metabolism and clearance of T4. Generalized estimating equations were used to adjust the 12 standard errors of the ORs for correlation between siblings. 13 The mean levels of b-TSH were positively associated with average soil TCDD 14 concentrations in the three areas (Zone A: 1.66 pU/mL; Zone B: 1.35 pU/mL; and Zone R: 15 0.98 pU/mL) (p < 0.001). Plasma TCDD levels also were shown to be much higher in a group of 16 51 newborns that had b-TSH levels >5 pU/mL. Compared to the reference population, adjusted 17 ORs were elevated for Zone B (OR = 1.90, 95% CI = 0.94-3.86) and Zone A (OR = 6.63, 18 95% CI = 2.36-18.6). These ORs were adjusted for gender, birth weight, birth order, maternal 19 age at delivery, hospital, and type of delivery. The adjusted ORs however differed only slightly 20 from those that were unadjusted (Zone B, OR = 1.79, 95% CI = 0.92-3.50; Zone A OR = 6.60, 21 95% CI = 2.45-17.8). Of the risk factors considered, both gender and birth weights were 22 associated with neonatal b-TSH. 23 The paper also included an analysis of children born to 109 women who were part of the 24 Seveso Chloracne Study (Baccarelli et al., 2005, 197053). A total of 51 children were born to 25 38 of these women, of these 12 lived in Zone A, 10 in Zone B, 20 in Zone R, and 9 from the 26 reference population. Several congeners including TCDD were measured in maternal plasma. 27 TCDD levels were extrapolated to the date of delivery using a first-order pharmacokinetic model 28 (Michalek et al., 1996, 198893). The elimination rate used was 9.8 years based on the mean 29 half-life estimate from a previous study of women in the Seveso region (Michalek et al., 2002, 30 199579). TEQs were calculated for a mixture of dioxin-like compounds by multiplying the 31 concentration of each congener by its toxicity equivalence factor. The maternal average TEQ This document is a draftfor review purposes only and does not constitute Agency policy. 2-118 DRAFT--DO NOT CITE OR QUOTE 1 was 44.8 ppt (range: 11.6-330.4) among 51 mothers. The measurement of noncoplanar PCBs 2 occurred only later in the study (1996) and, therefore, total mean TEQs (i.e., including the sum 3 of PCDDs, PCDFs, coplanar PCBs, and noncoplanar PCBs) are available only on a subset 4 (n = 37) of the population. Dioxin-like congeners were examined in this study as several studies 5 suggest associations between the sum of PCBs, or individual congeners having decreased 6 thyroxine (T4; Longnecker et al., 2000, 201463; Sandau et al., 2002, 594406), and increased 7 TSH (Alvarez-Pedrerol et al., 2008, 594407; Chevrier et al., 2007, 594408). The following 8 confounders were examined by the authors in the plasma dioxin models: maternal body mass 9 index, smoking habits, alcohol consumption, and neonatal age in hours at b-TSH measurement. 10 The authors used a linear model to examine the association between maternal TCDD 11 levels and b-TSH. The standardized regression coefficient obtained from this model was 0.47 12 (p < 0.001). For the evaluation of TEQs, a similar association was noted for PCDDs, PCDFs, 13 and coplanar PCBs (n = 51, P = 0.45, p = 0.005) but not with noncoplanar PCBs (n = 37, 14 P = 0.16, p = 0.45). Multivariate regression models that were adjusted for several covariates 15 (i.e., gender, birth weight, birth order, maternal age at delivery, hospital, and type of delivery) 16 found statistically significant associations with plasma TCDD, PCDDs, PCDFs, and coplanar 17 PCBs, but not with noncoplanar PCBs. The sum of all total TEQs from the measured 18 compounds was not statistically significant (n = 37, P = 0.31, p = 0.14). 19 20 2.4.1.2.1.5.7.2. Study evaluation. 21 The Baccarelli et al. (2008, 197059) study satisfies the epidemiological considerations 22 and criteria for determining whether dose-response modeling should be pursued. The outcome is 23 well defined, and a dose-response pattern was observed. The study also contained a substudy 24 that characterized TCDD and exposures to other dioxin-like congeners and used serum measures 25 for a sample of mothers. Results were consistent among the zone of residence analysis and the 26 substudy based on serum measures. 27 28 2.4.1.2.1.5.7.3. Suitability o f da ta f o r TCDD dose-response modeling. 29 Given the potential for exposure misclassification due to variability in TCDD soil levels 30 within each zone, modeling should rely on individual-level TCDD exposures derived from the 31 serum sampling substudy. The study data provide an opportunity for quantitative dose-response This document is a draftfor review purposes only and does not constitute Agency policy. 2-119 DRAFT--DO NOT CITE OR QUOTE 1 analyses as the critical exposure window of 9 months can be used for exposure assessment 2 purposes. 3 4 2.4.1.2.1.5.8. Mocarelli et al. (2008, 199595)-- Sperm effects. 5 2.4.1.2.1.5.8.1. Study summary. 6 Mocarelli et al. (2008, 199595) examined the relationship between TCDD and endocrine 7 disruption and semen quality in a cohort of Seveso men. A total of 397 subjects of the eligible 8 417 males (<26 years old in 1976) from Zone A and nearby contaminated areas were invited to 9 participate. Frozen serum samples were used to derive TCCD exposures. Also, 372 healthy 10 blood donors not living in the TCCD-contaminated area were invited to participate. The 11 researchers collected a health questionnaire and semen samples from participants. Analyses 12 were based on 257 individuals in the exposed group and 372 in the comparison group. 13 Semen samples were collected postmasturbatory at home. Ejaculate volume, sperm 14 motility, and sperm concentration were measured on these samples. Fasting blood samples also 15 were collected from the subjects for reproductive hormone analyses, including 17P-estradiol 16 (E2), follicle stimulating hormone (FSH), inhibin B, luteinizing hormone (LH), and testosterone. 17 The researchers estimated serum concentrations of TCDD from samples provided in 18 1976-1977, and also in 1997-1998 for individuals whose earlier samples had TCDD values that 19 exceeded 15 ppt. Serum concentrations for the comparison group were assumed to be less than 20 15 ppt in 1976 and 1977 and <6 ppt in 1998/2002 on the basis of serum results for residents in 21 uncontaminated areas. The exposed and comparison groups were divided into three groups 22 based on their age in 1976: 1-9, 10-17, and 18-26 years. Mocarelli et al. (2008, 199595) 23 applied a general linear model to the sperm and hormone data and included exposure status, age, 24 smoking status, body mass index, and occupational exposures as covariates. The study authors 25 thoroughly addressed the potential for confounding. 26 Men exposed between the ages of 1 and 9 had reduced semen quality 22 years later. 27 Reduced sperm quality included decreases in sperm count (p = 0.025), progressive sperm 28 motility (p = 0.001), and total number of motile sperm (p = 0.01) relative to the comparison 29 group. The opposite pattern was observed for several indices of semen quality among those aged 30 10-17 at the time of the accident; this included a statistically significant increase in sperm count 31 (p = 0.042). The clinical significance of this increase is unknown. For the hormone analyses, This document is a draftfor review purposes only and does not constitute Agency policy. 2-120 DRAFT--DO NOT CITE OR QUOTE 1 those in the exposed group had lower serum E2levels, and higher follicle stimulating hormone 2 concentrations. Neither testosterone levels nor inhibin B concentrations were associated with 3 TCDD exposure. 4 5 2.4.1.2.1.5.8.2. Study evaluation. 6 The findings of the Mocarelli et al. (2008, 199595) study support the hypothesis that 7 exposure to TCDD in infancy/prepuberty reduces sperm quality. The changes in serum E2 and 8 FSH concentrations are of unknown clinical significance, and cannot be considered adverse. 9 Although most semen analysis studies have low compliance rates in general population samples 10 (20-40%) (Jorgensen et al., 2001, 594402; Muller et al., 2004, 594403), the compliance rate in 11 this study was much higher (60%). Given that the compliance rates were similar between the 12 exposed and comparison groups and the strong differences detected across the two age groups, 13 selection bias appears unlikely in this study. 14 15 2.4.1.2.1.5.8.3. Suitability o f da ta f o r TCDD dose-response modeling. 16 Health outcomes are well defined in the Mocarelli et al. (2008, 199595) study, and 17 exposures are well characterized using serum data. Because the men exposed to elevated TCDD 18 levels between the ages of 1 and 9 had reduced semen quality 22 years later, it is difficult to 19 identify the relevant time interval over which TCDD dose should be considered. Specifically, it 20 is difficult to discern whether this effect is a consequence of the initial high exposure between 21 1 and 9 years of age or a function of the cumulative exposure for this entire exposure window 22 beginning at the early age. However, the differences between these two dose estimates (the 23 initial high exposure versus the cumulative exposure for the 9 year window) are minimal (i.e., 24 within an order of magnitude). Despite the uncertainty in estimating the critical window of 25 exposure, dose-response analysis for this outcome was conducted. 26 27 2.4.1.2.1.6. T h e C h a p a e v sk stu d y. 28 2.4.1.2.1.6.1. Revich et al. (2001, 199843)--Mortality and reproductive health. 29 2.4.1.2.1.6.1.1. Study summary. 30 Revich et al. (2001, 199843) describe a series of investigations that have evaluated 31 adverse health outcomes among residents of Chapaevsk where ecological measures of TCDD This document is a draftfor review purposes only and does not constitute Agency policy. 2-121 DRAFT--DO NOT CITE OR QUOTE 1 have been noted to be higher than expected. In the earlier cancer section of this report, the 2 cross-sectional comparisons of mortality that the authors carried out between Chapaevsk 3 residents and a general population reference were described. Although the general focus of this 4 paper is on cancer, the authors examined other adverse health outcomes. 5 For all-cause mortality, rates were found to be higher in Chapaevsk relative to the Samara 6 region and other nearby towns. The magnitude of this increase, however, was not quantified in 7 the review by Revich. Cardiovascular mortality accounted for nearly two-thirds of women's 8 deaths and almost half of those among men. The rates of cardiovascular mortality among 9 Chapaevsk men have been reported to be 1.14 times higher than those in Russia. 10 Revich et al. (2001, 199843) also reported on the occurrence of adverse reproductive 11 events. Although the authors indicated that official medical information was used to make 12 comparisons between regions, no details were provided about data quality, completeness, or 13 surveillance differences across areas. The presented rates for reproductive health outcomes 14 should be interpreted cautiously. A higher rate of spontaneous abortions (24.4 per 15 100 pregnancies finished by delivery) was found in Chapaevsk women relative to rates that 16 ranged between 10.6 and 15.2 found in five other areas. The frequency of preeclampsia also was 17 found to be higher in Chapaevsk women (44.1/100) relative to other towns, as was the proportion 18 of low birth-weight babies and preterm births. The percentage of newborns with low birth 19 weight was slightly larger in Chapaevsk (7.1%) when compared to other towns in Samara 20 (5.1-6.2%); observed differences, however, were not statistically significant. The authors also 21 reported on the sex ratio of newborns born between 1983 and 1997. These ratios (boys:girls) 22 were highly variable and ranged between 0.79 and 1.29. Given the annual variability of this ratio 23 on a year-to-year basis, it is unclear if this is largely due to natural fluctuations and to what 24 extent this may result from prior TCDD (or other contaminants) exposure TCDD and other 25 contaminants. 26 27 2.4.1.2.1.6.1.2. Study evaluation. 28 The review by Revich et al. (2001, 199843) highlights analyses that have been 29 undertaken using largely cross-sectional data. Although soil sampling measures appear to 30 demonstrate decreasing levels of TCDD in the soil with increasing distance from the plant, at this 31 time, no individual-level TCDD exposure data are available. Increased rates of mortality relative This document is a draftfor review purposes only and does not constitute Agency policy. 2-122 DRAFT--DO NOT CITE OR QUOTE 1 to the Samara region in Russia were observed among Chapaevsk men for all cancer sites 2 combined; this excess risk however, was not observed among women. Although the authors 3 provide compelling evidence of increased adverse events among residents of Chapaevsk, the 4 study lacks a discussion about the validity of comparing health data across regions, and suffers 5 from inherent limitations from ecological studies such as exposure misclassification. 6 7 2.4.1.2.1.6.1.3. Suitability o f datafor TCDD dose-response modeling. 8 As with the cancer outcomes presented in this study, the data for noncancer outcomes are 9 limited by the absence of TCDD levels on an individual-level basis and information on other 10 potential confounding variables that could have biased the comparisons. Additional studies are 11 being undertaken to evaluate the relationship between TCDD and the sexual and physical 12 development of boys. The cross-sectional nature of the data that were presented does not 13 provide the necessary level of detail needed to estimate effective dose given the lack of 14 individual-level exposure data. Therefore, a quantitative dose-response analysis was not 15 conducted. 16 17 2.4.1.2.1.7. T h e A i r F o rce H e a lth ( "R a n c h H a n d s " c o h o rt) stu d y . 18 2.4.1.2.1.7.1. Michalek and Pavuk (2008, 199573)--Diabetes. 19 2.4.1.2.1.7.1.1. Study summary. 20 Michalek and Pavuk (2008, 199573) examined both the incidence of cancer and the 21 prevalence of diabetes in the cohort of Ranch Hand workers exposed to TCDD. As noted 22 previously, these veterans were responsible for aerial spraying of Agent Orange in Vietnam 23 between 1962 and 1971. Exposure to TCDD was estimated using serum collected from 24 participants in 1987 and assayed for TCDD. Exposure to TCDD was estimated using a 25 first-order pharmacokinetic model with a half-life of 7.6 years and provided an estimate of 26 TCDD at the end of the tour of duty in Vietnam. Veterans were grouped into four categories: 27 comparison, background, low, and high. Diabetes was identified from diagnoses during the 28 post-Vietnam era from medical records. Overall, no differences were shown in the RR of 29 diabetes between the Ranch Hand unit and the reference group (RR = 1.21, p = 0.16). Stratified 30 analyses by days of spraying (<90 days, >90 days), however, revealed a significant increase in 31 risk of diabetes (RR = 1.32, p = 0.04) among those who sprayed for at least 90 days. A dose- This document is a draftfor review purposes only and does not constitute Agency policy. 2-123 DRAFT--DO NOT CITE OR QUOTE 1 response relationship was also evident when log10TCDD was modeled in the combined cohort. 2 Also, stratification by calendar period showed a dose-response relationship for those whose last 3 year of service was during or before 1969. 4 5 2.4.1.2.1.7.1.2. Study evaluation. 6 The Michalek and Pavuk (2008, 199573) study provides an opportunity to characterize 7 risks of diabetes as the study is not subject to some of the potential bias of case ascertainment 8 based on death certificates (D'Amico et al., 1999, 197389). The quality of the TCDD exposure 9 estimates is high, given that serum data were available at an individual-level basis for all Ranch 10 Hand and comparison veterans used in the cohort. Although disentangling the effects of 2,4-D 11 and TCDD is not possible because their concentrations in Agent Orange are equivalent, 2,4-D 12 has not been associated with diabetes. 13 14 2.4.1.2.1.7.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 15 The reported dose-response relationship between TCDD and diabetes is supported by 16 study strengths including the use of the individual-level level TCDD serum measures and the 17 identification of diabetes through medical records are important strengths of the Michalek and 18 Pavuk (2008, 199573) study. Nonetheless, the possible confounding from the inability to control 19 for 2,4-D and other agents used in Agent Orange precludes a quantitative dose-response analysis. 20 21 2.4.1.2.1.8. O th e r n o n ca n cer stu d ie s o f TC D D . 22 2.4.1.2.1.8.1. Ryan et al. (2002, 198508)-- Sex ratio. 23 2.4.1.2.1.8.1.1. Study summary. 24 Ryan et al. (2002, 198508) conducted an investigation on the sex ratio in offspring of 25 children of pesticide workers who were involved with the production of trichlorophenol and the 26 herbicide 2,4,5-T in Ufa, Bashkortostan, Russia. Ufa was the site of a state agrochemical plant 27 that has been in operation since the 1940s. Between 1961 and 1988, the plant employed more 28 than 600 workers, most in their early 20s. Females, however, accounted for about 15% of the 29 workforce that produced 2,4,5-T and 30% for 2,4,5-trichlorophenol. 30 Serum samples previously taken in 1992 among 60 men, women, and children from the 31 factory and city of Ufa showed TCDD exposures that were approximately 30 times higher than This document is a draftfor review purposes only and does not constitute Agency policy. 2-124 DRAFT--DO NOT CITE OR QUOTE 1 background levels (Ryan and Schecter, 2000, 594412). Blood data were subsequently measured 2 on a sample of 20 workers between 1997 and 2000, and on 23 2,4,5-trichlorophenol workers 3 between 1997 and 2001. In all, 84 individuals who provided blood samples formed the basis of 4 the analysis in this study. Of these, 55 were exposed to 2,4,5-T and 29 were exposed to 5 2,4,5-trichlorophenol. 6 Ryan et al. (2002, 198508) reviewed company records for these workers to determine the 7 number, sex, and date of birth of any children; birth data were available for 198 workers. 8 Awareness of the study led other workers who had not provided serum to provide information on 9 births that occurred 9 months after the time of first employment in the factory. 10 The authors calculated descriptive statistics for the 198 workers and compared them to 11 values for the city of Ufa between 1959 and 1996. Tests of statistical significance were made 12 using the z-test, and the chi-square test. The observed proportion of male births (0.40) among 13 the factory workers was much lower than that for the city of Ufa (0.51) (p < 0.001). Stratified 14 analyses revealed that this lower ratio was observed only among those paternally exposed to 15 TCDD. Specifically, the proportion of male births among exposed fathers was 0.38 and among 16 exposed mothers was 0.51. This pattern was observed in both the workers exposed to 2,4,5-T 17 (proportion of male births = 0.40) and 2,4,5-trichlorophenol (proportion of male births = 0.35). 18 19 2.4.1.2.1.8.1.2. Study evaluation. 20 The Ryan et al. (2002, 198508) findings are consistent with earlier work completed for 21 Seveso residents (Mocarelli et al., 2000, 197448). Although serum measures were available for 22 84 individuals, no dose-response of birth ratios was performed using exposure quantified at an 23 individual-level basis. This approach would have been preferred and consistent with that which 24 Mocarelli et al. (2000, 197448) used. All comparisons were made using an external comparison 25 group, namely the sex ratio observed in Ufa between 1959 and 1996. 26 Although serum measures were used to describe TCDD exposure for a sample of the 27 workers, individual-level dose estimates were not calculated for the study population. 28 Specifically, exposures were characterized many years after exposure, and no attempt was made 29 to back-extrapolate to the time of conception. The two groups of workers in the study also 30 reportedly had high exposure levels of 1,2,3,7,8-pentachlorodibenzo-p-dioxin. So, the group 31 level exposure classification (by plant) did not allow consideration of confounding due to other This document is a draftfor review purposes only and does not constitute Agency policy. 2-125 DRAFT--DO NOT CITE OR QUOTE 1 dioxin-like compounds. Another limitation of the study is that the study population is likely 2 nonrepresentative of all workers employed at the plant. Participants included only those willing 3 to provide serum samples and those who volunteered to participate in the study after learning 4 about it in a public forum. If participation was dependent on TCDD exposures and the 5 reproductive health of these subjects, then bias may have occurred. 6 7 2.4.1.2.1.8.1.3. Suitability o f da ta f o r TCDD dose-response modeling. 8 The findings are notable in their consistency with those found in Seveso residents by 9 Mocarelli et al. (2000, 197448). For the Ryan et al. (2002, 198508) study, serum data were 10 quantified at an individual-level basis. Risk estimates, however, were not derived in relation to 11 these exposures but instead in two separate subgroups (2,4,5-T and 2,4,5-trichlorophenol 12 workers). This important limitation precludes the use of these data for quantitative 13 dose-response modeling. 14 15 2.4.1.2.1.8.2. Kang et al. (2006, 199133)--Long-term health effects. 16 2.4.1.2.1.8.2.1. Study summary. 17 Kang et al. (2006, 199133) investigated the relationship between self-reported health 18 measures and serum-based measures of TCDD in a group of 1,499 Vietnam veterans and a 19 control group of 1,428 non-Vietnam veterans. The study subjects were identified from 20 (1) reports of Army Chemical Corps detachments in Vietnam between 1966 and 1971, 21 (2) personnel records of individuals involved in chemical operations who were on active duty 22 between 1971 and 1974, and (3) class rosters of personnel who were trained at Fort McClellan in 23 Alabama between 1965 and 1973. The comparison group was selected so that branch of service, 24 time period, and military occupation were similar to those of the subjects with the exception that 25 they did not serve in Vietnam. Although 2,872 Vietnam veterans and 2,732 non-Vietnam 26 veterans were identified as potential subjects, those who were deceased as of December 1998 27 and those who had previously participated in a pilot study were excluded. The study targeted 28 2,247 Vietnam and 2,242 non-Vietnam veterans. 29 Exposure to TCDD was characterized for subsets of the study population that provided 30 blood samples, specifically 795 of 1,085 (73%) Vietnam veterans and 102 of 157 (65%) 31 non-Vietnam veterans. Details on these individuals selected for participation in the serum dioxin This document is a draftfor review purposes only and does not constitute Agency policy. 2-126 DRAFT--DO NOT CITE OR QUOTE 1 study were not presented. The authors did state, however, that due to economic constraints, only 2 897' serum samples could be analyzed. Blood specimens were collected in 1999-2000 at 3 individuals' homes. TCDD concentrations were analyzed by laboratory staff blind to the group 4 status (i.e., Vietnam or non-Vietnam) of the study subjects. 5 Prevalent health outcomes were ascertained by self-reported information on selected 6 conditions diagnosed by a medical doctor. The following conditions were included: diabetes, 7 hepatitis (all types combined), heart disease, all cancer, nonmalignant chronic respiratory 8 diseases, and hypertension. Health-related quality of life was evaluated using the SF-36 survey 9 instrument (Ware et al., 1993, 004687). 10 Eligible veterans whose current residences (4,119 total) could be identified were 11 contacted for study participation. Survey participation rates were 72.9% for Vietnam veterans, 12 yielding data for 1,499 individuals, and 69.2% for non-Vietnam veterans, yielding data for 13 1,428 non-Vietnam veterans. The survey data showed that, relative to non-Vietnam veterans, 14 Vietnam veterans were more likely to be regular smokers and to be obese. They also were more 15 likely to be enlisted personnel, and a much higher proportion was 51 years of age or older 16 (83.4% vs. 58.4%). After adjusting for age, race, smoking status, rank, and body mass index, the 17 prevalence of self-reported health conditions was found to be statistically significantly higher in 18 the Vietnam group. The adjusted odds ratios (OR) were as follows: diabetes, OR = 1.16 19 (95% CI = 0.91, 1.49); hepatitis, OR = 1.85 (95% CI = 1.30, 2.64); heart condition, OR = 1.09 20 (95% CI = 0.87, 1.38); all cancer, OR = 1.46 (95% CI = 1.02, 2.10); nonmalignant respiratory 21 condition, OR = 1.41 (95% CI = 1.13, 1.76); and hypertension, OR = 1.06 (95% CI = 0.89, 1.27). 22 For those with Vietnam service, the mean serum TCDD concentrations were higher 23 among those who reported spraying herbicides (4.3 parts per thousand [ppt]) than those who did 24 not (2.7 ppt) (p < 0.001). The investigators did not back-extrapolate serum levels to the time 25 when individuals last sprayed. The adjusted ORs (adjusted for age, cigarette smoking, body 26 mass index, rank, and race) for most chronic health conditions examined revealed increased 27 prevalence among Vietnam sprayers relative to non-Vietnam sprayers. These ORs were: 28 diabetes, OR = 1.49 (95% CI = 1.10, 2.02); hepatitis, OR = 1.40 (95% CI = 0.92, 2.12); heart 29 condition, OR = 1.41 (95% CI = 1.06, 1.89); all cancer, OR = 1.36 (95% CI = 0.91, 2.04); 30 nonmalignant respiratory condition, OR = 1.57 (95% CI = 1.20, 2.07); and hypertension, 31 OR = 1.26 (95% CI = 1.00, 1.58). This document is a draftfor review purposes only and does not constitute Agency policy. 2-127 DRAFT--DO NOT CITE OR QUOTE 1 The investigators also examine the possibility of over-reporting of chronic health 2 conditions by comparing the prevalence of self-reported conditions among 357 Vietnam sprayers 3 who mean serum TCDD levels of 2.5 ppt compared to those who had levels less than 2.5 ppt. 4 Prevalence of diabetes, heart condition, and hypertension, was higher among those with mean 5 serum TCDD levels of 2.5 ppt, although no levels of statistical significance were reported. Data 6 for cancer were not presented. 7 8 2.4.1.2.1.8.2.2. Study evaluation. 9 Because data were collected from only half of the individuals in the study target 10 population, there is some potential for selection bias in this study. First, the study excluded those 11 who had died before 1999, excluding potentially important TCDD-related adverse health effects 12 that could result in death more than two decades after veterans had been actively spraying. 13 Second, survey participation rates were modest: 72.9% for Vietnam veterans and 69.2% for 14 non-Vietnam veterans. If those in poorer health were less inclined to participate, the prevalence 15 of the selected chronic health conditions would be understated. Selection bias due to study 16 participation could also be possible if, for example, those in poorer health also had high (or 17 lower) exposures than those not participating in the study. The lack of direct evidence of 18 differential participation and reports of comparable prevalence rates of hypertension and diabetes 19 to other general populations suggests that selection bias may be minimal. 20 Because the data collected are cross-sectional, they are ill-suited for evaluating the 21 relationship between the timing of exposure and the onset of disease. Whether any of the data 22 could help identify when the chronic health conditions were diagnosed is unclear. Given the 23 long period covered by the study, many of the self-reported health conditions likely were 24 diagnosed some time ago, perhaps closer to the time of potential TCDD exposure. Such detail is 25 needed to characterize health risks associated with specific TCDD levels, particularly given that 26 TCDD levels have been demonstrated to decrease from time of last exposure. 27 An important strength of the study is the availability of blood sera for a subset of the 28 study population, which allows for an objective determination of TCDD exposure. That serum 29 TCDD levels were available for only 897 subjects, however, limits the ability to examine the 30 relationship between measures of TCDD and prevalence of health outcomes without restricting 31 the sample size or extrapolating exposure levels to the whole study population. For example, This document is a draftfor review purposes only and does not constitute Agency policy. 2-128 DRAFT--DO NOT CITE OR QUOTE 1 among sprayers with available TCDD exposure data only 60 cases of diabetes and 69 cases of 2 heart disease were examined relative to exposure. Also, the small number of cancers precluded a 3 cancer site-specific analysis. Moreover, whether these TCDD levels are representative of the 4 larger eligible population is difficult to gauge, given that deceased veterans and those whose 5 current residences could not be determined were excluded. 6 The study relied on self-reported measures of disease prevalence. The ascertainment of 7 chronic health conditions using self-reported data can be fraught with difficulties. For example, 8 the sensitivity of self-reported data when compared to medical diagnosis has been shown to be 9 poor for conditions such as diabetes and hypertension (Okura et al., 2004). As Kang et al. (2006, 10 199133) state, prevalence studies are not be well suited to examine rare diseases with short 11 survival times such as cancer. In addition, self-reports of physician-diagnosed cancers by study 12 subjects often lacks the sensitivity needed in most epidemiological studies as they can be 13 influenced by a variety of factors including age and education (Navarro et al., 2006). 14 The potential for biases in the reporting of health outcomes between the sprayers and the 15 non-Vietnam veterans (i.e., differential by TCDD exposure status) also is plausible, given the 16 public attention that spraying of Agent Orange has received. Although the authors examined 17 whether over-reporting was related to outcome prevalence among herbicide sprayers (prior to 18 collection and determination of actual TCDD serum levels), the possibility exists that these 19 subjects reporting could be influenced by their perceived level of exposure from herbicide 20 spraying. The authors also examined the potential for misreported diabetes by conducting a 21 medical records review of 362 veterans. Seventy-nine percent of the self-reported diabetes cases 22 were confirmed with medical records. The documentation rate was also comparable between the 23 Vietnam veterans and the non-Vietnam veterans suggesting that differential reporting was not an 24 issue for this health outcome. 25 Because the Vietnam veterans group comprised professional sprayers, it is not 26 unreasonable to assume that they would have been exposed to other potentially harmful agents 27 either during their service in Vietnam, or from the end of their service to when they provided 28 data in 1999-2000. This study did not control for other, potentially relevant occupational 29 exposures. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-129 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.8.2.3. Suitability o f da ta f o r TCDD dose-response modeling. 2 Although the study demonstrates increased prevalence of several chronic health 3 conditions, these findings should be interpreted with caution due to potential for selection and 4 recall biases. The lack of demonstrated dose-response relationships with cancer or other 5 outcomes precluded the use of these data for characterizing the dose response from TCDD. 6 7 2.4.1.2.1.8.3. McBride et al. (2009, 198490; 2009, 197296)--Noncancer mortality. 8 2.4.1.2.1.8.3.1. Study summary. 9 The McBride et al. (2009, 198490) mortality study of New Zealand workers employed as 10 producer or sprayers with potential exposure to TCDD was described earlier in this report. 11 These individuals were employed at a plant that manufactured 2,4,-dichlorophenoxyacetic acid, 12 and later 2,4,5-T and 4-chloro-2-methyphenoxyacetic acid. In 1987, the plant closed and 2,4,5-T 13 production ceased in 1988. 14 The cohort consisted of 1,754 individuals who were employed for at least one day at the 15 New Plymouth site between January 1, 1969, and October 1, 2003. Vital status was determined 16 until the end of 2004. Comparisons of mortality were made to the New Zealand general 17 population using the SMR statistic. Exposure was characterized by duration of employment. 18 Person-years of follow-up were tabulated across strata defined by age, calendar period, duration 19 of employment, sex, latency, and period of hire. Analyses were stratified to compare risks by 20 duration of employment (<3 or >3 months), latency (<15 or >15 years), and period of hire 21 (<1976, >1976). 22 Overall, no statistically significant differences in all-cause mortality relative to the 23 general population were found among those who worked for at least 3 months (SMR = 0.92, 24 95% CI = 0.80-1.06) or for less than 3 months (SMR = 1.23, 95% CI = 0.91-1.62). No 25 statistically significant excesses were found for mortality from diabetes, cerebrovascular disease, 26 heart diseases, or accidents. The incorporation of a latency period of 15 years revealed no 27 statistically significant excesses for these same causes of death. Similarly, no excesses for any 28 cause of death were noted among those who were hired either before or after 1976. 29 In subsequent analyses of the same cohort that used estimated TCDD levels from serum 30 samples, McBride et al. (2009, 197296) found no excesses for all-cause mortality or mortality 31 from diabetes or heart disease. This document is a draftfor review purposes only and does not constitute Agency policy. 2-130 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.8.3.2. Study evaluation. 2 For the McBride et al. (2009, 198490) study, the size of the cohort is large enough to 3 characterize mortality risks relative to the general population for most common causes of deaths. 4 An important limitation of this study is the loss to follow-up of a substantial percentage of 5 workers (22%). This would have impacted statistical power by reducing the number of deaths 6 among the workers. If this incomplete ascertainment of mortality outcomes did not occur in a 7 similar fashion with the general population then the SMR may also be biased. 8 For noncancer causes of death, the use of the SMR statistic is more likely to be 9 influenced by the healthy-worker effect. Therefore, the findings obtained for these outcomes 10 should be interpreted with caution. Subsequent analyses published by the same authors 11 (McBride et al., 2009, 197296) provide improved characterization of TCDD exposure using 12 serum samples. 13 14 2.4.1.2.1.8.3.3. Suitability o f da ta f o r dose-response analysis. 15 Overall, no associations were evident between surrogate measures of TCDD (duration of 16 employment, year of hire) and noncancer mortality outcomes. Further, the use of mortality 17 endpoints is inconsistent with EPA RfD methodology. As such, these data do not support further 18 use in a quantitative dose-response analysis. 19 20 2.4.1.2.1.8.4. McBride et al. (2009, 197296)--Noncancer mortality. 21 2.4.1.2.1.8.4.1. Study summary. 22 McBride et al. (2009, 197296) further analyzed the cohort of New Zealand workers to 23 include estimates of TCDD exposure based on serum samples. Current and former employees 24 who were still alive and living within 75 km of the site were asked to provide serum samples. 25 Samples were collected from 346 workers representing 22% (346/1599) of the entire study 26 population. These serum measures were used to estimate cumulative TCDD levels for all 27 workers. The exposure assessment approach by Flesch-Janys et al. (1996, 197351) was used to 28 estimate time-dependent exposures based on area under the curve models. This was based on a 29 one-compartment first-order kinetic model with a half-life of 7.2 years. 30 Comparisons of mortality were made to the general population using the SMR statistic. 31 The Cox proportional hazards model was used to conduct an internal cohort analysis across This document is a draftfor review purposes only and does not constitute Agency policy. 2-131 DRAFT--DO NOT CITE OR QUOTE 1 four categories of cumulative TCDD levels for diabetes and ischemic heart disease mortality. 2 The RRs generated from these models were adjusted for sex, hire year, and birth year. No 3 diabetes deaths were observed among women, and therefore, analysis of this outcome was 4 limited to men. 5 Relative to the general population, no difference in the all-cause mortality experience was 6 observed in exposed cohort members (SMR = 1.0, 95% CI = 0.9-1.2). Similarly, no excess in 7 these workers was observed for heart disease (SMR = 1.1, 95% CI = 0.9-1.5); cerebrovascular 8 disease (SMR = 1.1, 95% CI = 0.6-1.9); diabetes (SMR = 0.7, 95% CI = 0.2-2.2); or 9 nonmalignant respiratory disease (SMR = 0.8, 95% CI = 0.4-1.4). For the internal cohort 10 analysis, the RR associated with cumulative categorical TCDD measure was 1.0 for both 11 diabetes and ischemic heart disease. 12 13 2.4.1.2.1.8.4.2. Study evaluation. 14 The McBride et al. (2009, 197296) study extends the earlier work the same authors 15 completed in two ways. First, serum measures were used to estimate cumulative TCDD with 16 methodology that has been applied to several other cohorts of workers exposed to TCDD. 17 Second, the authors used regression analyses that examined individual-level TCDD exposures in 18 relation to various outcomes as part of the internal cohort comparisons. For noncancer 19 outcomes, no dose-response associations with TCDD were observed with the internal 20 comparisons. Also, as found with earlier analyses of this same cohort, no excess noncancer 21 mortality relative to the New Zealand general population was observed. 22 Associations between TCDD and diabetes have been found previously in TCDD-exposed 23 populations, most notably in the Ranch Hands cohort (Michalek and Pavuk, 2008, 199573). In 24 this cohort, only five deaths from diabetes were identified, and of these, only three occurred 25 among those who were exposed to TCDD. The study, therefore, has limited statistical power to 26 characterize associations between TCDD and mortality from diabetes. Further, the identification 27 of diabetes deaths is subject to misclassification errors due to under-reporting (McEwen and 28 TRIAD, 2006, 594400). 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-132 DRAFT--DO NOT CITE OR QUOTE 1 2.4.1.2.1.8.4.3. Suitability o f datafor TCDD dose-response modeling. 2 McBride et al. (2009, 197296)) found no statistically significant associations in any of the 3 noncancer causes of death. Furthermore, the use of mortality endpoints is inconsistent with EPA 4 RfD methodology. Therefore, the data were not suitable for quantitative dose-response analysis 5 for these outcomes. 6 7 2.4.I.2.2. F easibility o f dose-respon se m odelin g f o r noncancer. 8 Relatively few study populations permit quantitative dose-response modeling to be 9 performed for noncancer outcomes. The serum collected among Seveso men and women 10 provide an opportunity to characterize risks for several health conditions in relation to TCDD 11 exposure. The collection of these serum samples, shortly after the accident does not require the 12 back-extrapolation of TCDD levels as in the occupational cohorts, which should reduce the 13 exposure assessment uncertainty and minimize the potential for exposure misclassification. 14 An added feature of the SWHS is the detailed collection of other risk factor data from 15 trained interviewers. These data allow for risk estimates to be adjusted for potential confounding 16 variables. For the evaluations of reproductive health outcomes, this adjustment is critical given 17 there are various documented risk factors for the different outcomes that were examined. For 18 some health outcomes, continued follow-up of the cohort is needed, given that several of the 19 Seveso studies suggest that those exposed at a very young age might be more susceptible to 20 subsequent adverse health effects. 21 The findings of positive associations and dose-response relationships with serum-based 22 measures of TCDD suggest several noncancer health outcomes could be associated with TCDD 23 exposure. These health outcomes include neonatal thyroid function, sex ratio, diabetes, and 24 semen quality. Although findings have suggested an association between TCDD and age at 25 menopause, they were not statistically significant and no dose-response trend was observed. 26 Weak or nonstatistically significant associations have been noted for endometriosis and 27 menstrual cycle characteristics and do not support quantitative dose-response analyses. 28 Associations between TCDD exposure and cardiovascular disease have been noted in 29 some, but not all, of the occupational cohorts, and also shortly after the accident among Seveso 30 residents. Findings from the cohort studies based on external comparisons using the SMR 31 statistic should be interpreted cautiously due to potential bias from the healthy worker effect. This document is a draftfor review purposes only and does not constitute Agency policy. 2-133 DRAFT--DO NOT CITE OR QUOTE 1 Because the magnitude of the healthy worker bias is recognized to be larger for cardiovascular 2 diseases than for cancer outcomes, risk estimates in some occupational cohorts might be 3 underestimated for cardiovascular outcomes. Information on cardiovascular risk factors 4 generally was not captured in these studies, and sensitivity analyses were generally designed to 5 examine risk estimates generated for cancer outcomes. 6 7 2.4.1.2.3. S u m m a ry o f epidem iologic n o n ca n cer stu d y evalu ation s f o r dose-respon se 8 modeling. 9 All epidemiologic noncancer studies summarized above were evaluated for suitability of 10 quantitative dose-response assessment using the TCDD-specific considerations and study 11 inclusion criteria. The results of this evaluation are summarized in a matrix style array (see 12 Table 2-3) at the end of the chapter, and descriptively in Appendix B. The key epidemiologic 13 noncancer studies suitable for further TCDD dose-response assessment are presented in 14 Table 2-5. 15 16 2.4.2. Summary of Animal Bioassay Studies Included for TCDD Dose-Response Modeling 17 This section summarizes studies that have already met the in vivo animal bioassay TCDD 18 study inclusion criteria (see Section 2.3.2). These studies are listed later in this section in 19 Tables 2-6 and 2-7, for cancer and noncancer, respectively, and are considered in the 20 dose-response modeling conducted later in this document (see Sections 4 and 5). The following 21 sections are organized by reproductive studies, developmental studies, and general toxicity 22 studies (subdivided by duration). They summarize the experimental protocol, the results, and the 23 NOAELs and LOAELs EPA has identified for each study. 24 To evaluate and discuss studies consistently, doses were converted to nanograms per 25 kilogram body weight per day (ng/kg-day) and were also adjusted for continuous exposure. 26 Some doses were adjusted based on daily dietary intake and body weight. For these studies, 27 EPA uses 10% of an animal's body weight as the daily feed rate. More commonly, doses were 28 adjusted from 5 days/week to a 7 days/week standard adjustment, in which case administered 29 doses were multiplied by 5 and divided by 7 to obtain continuous doses. To adjust for weekly 30 dosing, the weekly administered doses were multiplied by the administration frequency per week 31 (in days) and divided by 7 to give continuous doses. This document is a draftfor review purposes only and does not constitute Agency policy. 2-134 DRAFT--DO NOT CITE OR QUOTE 1 Other exposure protocols used a single loading dose followed by weekly maintenance 2 doses. To adjust these doses, the loading dose was added to the maintenance doses multiplied by 3 the administration frequency, and this sum was divided by the exposure duration to give a 4 continuous dosing rate. The doses administered in single dose studies were not averaged over 5 the observation period. 6 7 2.4.2.1. R eprodu ctive S tu dies 8 2.4.2.1.1. B ow m an et al. (1989, 543744; 1989, 543745) (a n d rela te d S ch a n tz a n d B ow m an 9 (1 9 8 9 ,198104) ; S ch a n tz e t al. (1986, 088206)). 10 Female rhesus monkeys (6 to 10 years old; 8 per treatment) were exposed to 0 or 5 ppt 11 (for 3.5 years), or 25 ppt (for 4 years) TCDD (purity not specified) (Bowman et al., 1989, 12 543744; Bowman et al., 1989, 543745; Schantz and Bowman, 1989, 198104; Schantz et al., 13 1986, 088206). Female monkeys were mated to unexposed males after 7 months (Cohort I) and 14 27 months (Cohort II) of exposure, then again 10 months postexposure (Cohort III). The average 15 daily doses to mothers were equivalent to 0, 0.15, and 0.67 ng/kg-day. The 0.67 ng/kg-day dose 16 group had reduced reproductive rates in both Cohorts I (p < 0.001) and II (p < 0.025; Bowman 17 et al., 1989, 543744). The mean number of days of offspring survival (p < 0.023) also decreased. 18 No effects on birth weight or growth, or physical evidence of toxicity (Bowman et al., 1989, 19 543745) were observed. Behavioral effects were observed in the offspring (Cohort I: 7, 6, and 20 0 offspring, respectively; Cohort II: 3, 5, and 0 offspring, respectively; Cohort III: 6, 7, and 3, 21 respectively). In the 0.67 ng/kg-day dose group, the number of offspring was insufficient to 22 form a group in either Cohorts I or II. Offspring in the 0.15 ng/kg-day dose group had alterations 23 in social behavior of the mother-infant pairs (mothers had increased care giving, which appeared 24 to be an effect of the infants and not due to the treatment of the mother) and peer group of the 25 offspring after weaning (Cohort I offspring were more dominant or aggressive and exhibited 26 more self-directed behavior; Bowman et al., 1989, 543745). The performance of learning tasks 27 was inversely related to the level of TCDD in the body fat. Schantz and Bowman (1989, 28 198104) examined effects using discrimination-reversal learning (RL) and delayed spatial 29 alteration (DSA). RL detected effects in the 0.15 ng/kg-day group as measured by retarded 30 learning of the shape reversal (p < 0.05), but DSA did not. Schantz et al. (1986, 088206) 31 combined the cohorts and looked at 5, 5, and 3 mother-infant pairs in the 0, 0.15, and This document is a draftfor review purposes only and does not constitute Agency policy. 2-135 DRAFT--DO NOT CITE OR QUOTE 1 0.67 ng/kg-day groups, respectively. They found that TCDD-exposed mother-infant pairs spent 2 more time in close, social contact compared to the controls (mutual ventral contact,p < 0.025; 3 nipple contact, p < 0.01) and infants had reduced locomotor activity (p < 0.05), but the 4 dose-effect was complex. Of note is that the control groups contained fewer males than did the 5 TCDD-exposed groups. 6 In a follow-up study, Rier et al. (2001, 199843) examined the DLC levels of sera 7 collected from some monkeys in this study. They reported that animals in this study had 8 elevated serum PCB77 and PCB126 levels and an increased serum TEQ. In fact, the fractional 9 contribution of serum TCDD levels to total serum TEQ was 30% in treated animals. In this 10 study, it is not possible to determine the contribution of TCDD alone to the developmental effect 11 due to the background contamination; thus, EPA has not developed a TCDD LOAEL from the 12 study. 13 14 2.4.2.I.2. F ran c e t al. (2 0 0 1 ,197353). 15 To study the effects of subchronic, low-dose exposure to TCDD on the regulation and 16 expression of the aryl hydrocarbon receptor (AhR), Franc et al. (2001, 197353) used rodent 17 models with varying sensitivities to TCDD. Female Sprague-Dawley rats, inbred Long-Evans 18 rats, and outbred Han/Wistar rats (8 per dose group) were dosed via oral gavage with 0, 140, 19 420, or 1,400 ng/kg TCDD (>99% purity) dissolved in corn oil once every 2 weeks for 22 weeks 20 (0, 10, 30, and 100 ng/kg-day average daily doses). Animals were sacrificed 10 days after the 21 final dosing. Body weights were recorded biweekly and just before sacrifice. After sacrifice, 22 liver and thymus weights were determined. Liver tissue samples were removed and either frozen 23 for RNA isolation followed by semiquantitative RT-PCR or homogenized and prepared for 24 subcellular fraction analysis. Radioligand binding and immunoblotting techniques were used to 25 measure AhR levels, and RT-PCR analysis was used to assess mRNA levels of AhR, aryl 26 hydrocarbon nuclear receptor (ARNT), and CYP1A1. 27 Long-Evans rats exhibited significant (p < 0.001) decreased weight gain over time as 28 compared to Sprague-Dawley and Han/Wistar rats as determined by repeated measures analysis 29 of variance (ANOVA). Because body weight gain varied indirectly with TCDD exposure, liver 30 and thymus tissue weights were normalized to body weight for data analysis. TCDD exposure 31 led to a significant (p < 0.05) increase in relative liver weights at all three TCDD doses and in all This document is a draftfor review purposes only and does not constitute Agency policy. 2-136 DRAFT--DO NOT CITE OR QUOTE 1 three rat strains, compared to the control groups. At the upper end of the TCDD dose range, 2 Sprague-Dawley rats dosed with 100 ng/kg-day showed the greatest increase in relative liver 3 weights (160% of the control values), while relative liver weights in Long-Evans and Han/Wistar 4 rats were similar to each other, and also were elevated above control values by 10-20%. At the 5 30 and 100 ng/kg-day doses, the relative thymus weights were significantly lower (p < 0.05) in 6 all rat strains compared to their corresponding controls, but the 10 ng/kg-day dose did not 7 produce a statistically significant effect in any strain. However, absolute thymus weight was 8 higher at all doses in Han/Wistar rats, which also had a higher control thymus weight. 9 Supporting observed differences in baseline TCDD sensitivity among the rat strains, liver 10 AhR levels in the control groups as measured by radioligand binding were similar for Sprague 11 Dawley and Han/Wistar rats, but were approximately two-fold higher for Long-Evans rats. A 12 significant (p < 0.05) two-fold, dose-dependent increase in radioligand binding of liver AhR was 13 observed at all TCDD doses relative to the control in Sprague-Dawley rats. At the 30 ng/kg-day 14 dose, the AhR level for Long-Evans rats was significantly (p < 0.05) increased to approximately 15 250% of the control level. 16 AhR protein levels measured in the liver cytosol by immunoblotting were highest in the 17 10 and 30 ng/kg-day TCDD dose groups for all three rat strains. Significant (p < 0.05) increases 18 in AhR levels were observed in the Sprague-Dawley rats that received 30 ng/kg-day, and in 19 Long-Evans rats that received either 10 or 30 ng/kg-day. A significant (p < 0.05) decrease in 20 AhR protein level was observed only at the 100 ng/kg-day dose in Han/Wistar rats. Liver AhR 21 protein was not detectable by immunoblotting in nuclear extracts for any strain or dose. The 22 study authors assert that AhR levels measured in cytosol correspond to measures in whole-tissue 23 lysates as demonstrated in their previous work. 24 Based on RT-PCR analysis, all three rat strains showed similar responses in liver AhR 25 mRNA following TCDD exposure. Liver AhR mRNA levels increased significantly (p < 0.05) 26 as compared to control levels in all rat strains at 10 and 30 ng/kg-day and in Long-Evans rats at 27 100 ng/kg-day. The study authors observed that statistically significant increases in AhR mRNA 28 levels in the liver were not always associated with statistically significant increases in AhR levels 29 for a given strain and dose, but that the opposite (increases in AhR levels associated with 30 increases in AhR mRNA levels) was always true. Changes in liver ARNT mRNA levels tended 31 to increase with increasing TCDD dose, and the increases were significant (p < 0.05) in the This document is a draftfor review purposes only and does not constitute Agency policy. 2-137 DRAFT--DO NOT CITE OR QUOTE 1 30 ng/kg-day dose groups of Long-Evans and Han/Wistar rats. At the 100 ng/kg-day TCDD 2 dose, all rat strains showed a decrease in ARNT mRNA in the liver relative to controls with 3 significant (p < 0.05) differences for the 100 ng/kg-day TCDD dose groups of Sprague-Dawley 4 and Han/Wistar rats. Liver CYP1A1 mRNA induction was not detectable in control animals. A 5 significant (p < 0.05) increase in liver CYP1A1 mRNA was observed in all rat strains 6 administered 10 or 30 ng/kg-day TCDD. Liver CYP1A1 mRNA levels also were significantly 7 (p < 0.05) elevated above controls in the 100 ng/kg-day groups although not to the same extent 8 as in the 30 ng/kg-day groups. For all rat strains, the largest up-regulation for AhR and ARNT 9 mRNA levels occurred in the 30 ng/kg-day TCDD dose groups. 10 The NOAEL for TCDD identified in this study is 10 ng/kg-day TCDD. At 10 ng/kg-day 11 TCDD, the change in relative liver weight, while significantly (p < 0.05) increased in 12 Sprague-Dawley rats, was determined (from Figure 5 in Franc et al., 2001, 197353) to be less 13 than 10% and judged by EPA not to be biologically relevant. Also, at 10 ng/kg-day TCDD, the 14 change in relative thymus weight, was not statistically significantly decreased in 15 Sprague-Dawley, Han-Wistar or Long-Evans rats. The study LOAEL is 30 ng/kg-day, based on 16 statistically and biologically significant increases in relative liver weight in Sprague-Dawley and 17 Long-Evans rats and statistically and biologically significant decreases in relative thymus weight 18 in Sprague-Dawley, Han-Wistar and Long-Evans rats. 19 20 2.4.2.I.3. H och stein e t al. (2 0 0 1 ,197544). 21 Adult female mink (12/treatment group) were administered dietary concentrations of 22 0.0006 (control), 0.016, 0.053, 0.180, or 1.40 ppb TCDD (purity >99.8%) for 132 days 23 (Hochstein et al., 2001, 197544). This dose is estimated to be equivalent to 0.03 (control), 0.8, 24 2.65, 9, and 70 ng/kg-day assuming a food consumption of 5% of body weight per day. Females 25 were mated with unexposed males beginning on treatment day 35. Females were allowed to 26 mate every fourth day during a 29-day mating period or until a confirmed mating. Mated 27 females were presented with a second male either the day after initial mating or 8 days later. In 28 the 70 ng/kg-day group, the treated animals were lethargic after 4 to 5 weeks, with several 29 having bloody (tarry) stools near the end of the trial. Two animals in the 70 ng/kg-day dose 30 group died prior to study termination. These animals had lost a large percentage of their body 31 weight (24-43%), and had pale yellow livers and intestinal hemorrhages. Histopathology from This document is a draftfor review purposes only and does not constitute Agency policy. 2-138 DRAFT--DO NOT CITE OR QUOTE 1 both mink indicated marked diffuse hepatocellular vacuolation. The mean body weight 2 decreased in all treatment groups including the control (losing an average of 3.29% of initial 3 body weight), compared to a dose-dependent loss of up to 26% in the 70 ng/kg-day group. 4 Mating and reproduction were considered subnormal in all groups. The number of females that 5 gave birth in the 0.03 (control), 0.8, 2.65, 9, and 70 ng/kg-day dose groups were 5/12, 0/12, 3/12, 6 8/12, and 0/11, respectively. The study authors speculated that the subnormal breeding and 7 reproductive performances in the control females likely were due to the indoor environment in 8 which the mink were housed. In the three groups that gave birth, there was a dose-dependent 9 decrease in kit body weight at birth, which was significant (p < 0.05) in the 9 mg/kg-day group 10 compared to the controls. The body weight in the kits was not significantly different at 3 or 11 6 weeks after birth. Three-week survival rates of 71, 47, and 11% were recorded for kits in the 12 0.03 (control), 2.65, and 9 ng/kg-day dose groups, respectively. Six-week kit survival rates were 13 62, 29, and 11% in the 0.03 (control), 2.65, and 9 ng/kg-day dose groups, respectively. 14 In the adult females, clinical signs of toxicity were noted in the 70 ng/kg-day group near 15 the end of the study and included alopecia and notably thickened, deformed, and elongated 16 toenails. There was a dose-dependent decrease in plasma total solids, total protein, and 17 osmolality that reached statistical significance (p < 0.05) in the two highest exposure groups. 18 Anion gap was significantly decreased (p < 0.05) and alanine aminotranferase was significantly 19 increased in the 70 ng/kg-day group compared to the controls. At terminal sacrifice, there was a 20 dose-related decrease in body weight. There was a dose-related increase in liver weight that 21 reached statistical significance (p < 0.05) in the 70 ng/kg-day dose group. The brains of 42% of 22 the animals in the 70 ng/kg-day dose group had localized accumulation of lymphatic cells within 23 the meninges with mild extension into the adjacent neuropil and mild gliosis. Of the 10 mink 24 surviving to study termination in the 70 ng/kg-day group, 3 had periportal hepatocellular 25 vacuolation. These same brain and liver lesions were not observed in the control mink. 26 As there were no litters produced in the low-dose group and pregnancy outcomes were 27 not dose related, the 0.8 ng/kg-day exposure level does not inform the choice of NOAEL or 28 LOAEL. Thus, the LOAEL for this study is 2.65 ng/kg-day (132-day maternal exposure 29 duration) based on reduced kit survival (47% of control at 6 weeks). A NOAEL cannot be 30 determined for this study. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-139 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.1.4. H u tt e t al. (2 0 0 8 ,198268). 2 Hutt et al. (2008, 198268) conducted a 3-month study investigating changes in 3 morphology and morphogenesis of pre-implantation embryos as a result of chronic exposure to 4 TCDD in female rats. The study authors administered 0 or 50 ng/kg TCDD (>99% purity) in 5 corn oil via oral gavage to groups of 3 pregnant Sprague-Dawley rats on gestation days 14 and 6 21 and on postnatal days 7 and 14. The resulting female pups were divided into groups of 3 and 7 administered 0 or 50 ng/kg TCDD (>99% purity) in corn oil (equivalent TCDD doses of 0 and 8 7.14 ng/kg-day) on postnatal day 21 and weekly thereafter until they reached 3 months of age. 9 Pups were then mated, fertilization was verified, and pre-implantation embryos were harvested 10 4.5 days later. Pre-implantation embryos were examined using immunofluorescence microscopy 11 to determine blastomere abnormalities. 12 No significant difference as compared to the control in pre-implantation embryotoxicity 13 was observed following exposure to TCDD. Morphologically normal pre-implantation embryos 14 were significantly (p < 0.05) reduced in 50 ng/kg TCDD exposed rats (15 of 41, 36.6%) 15 compared to the control group (31 of 39, 79.5%). Pre-implantation embryos of TCDD-exposed 16 rats included irregularities in mitotic spindles (13 of 18 were monopolar), chromosome patterns 17 in metaphase, blastomere size and shape, blastomere nuclei shape in interphase, f-actin, and 18 cytokinesis. The study authors concluded that the compaction stage of pre-implantation 19 embryogenesis is the most sensitive following exposure to TCDD. 20 A LOAEL for this study is 50 ng/kg (7.14 ng/kg-day adjusted dose) for a significantly 21 (p < 0.05) lower proportion of morphologically normal pre-implantation embryos during 22 compaction stage in female Sprague-Dawley pups weekly for 3 months. A NOAEL cannot be 23 determined for this study. 24 25 2.4.2.1.5. Iked a et al. (2 0 0 5 ,197834). 26 Ikeda et al. (2005, 197834) studied the effect of repeated TCDD exposure to F0 dams on 27 the male gonads of F1 generation and sex ratio in the F2 generation. Twelve female Holtzman 28 rats were treated with a single dose of 400 ng/kg TCDD (>98% purity) orally, via gavage, 29 followed by weekly treatment doses of 80 ng/kg TCDD (16.5 ng/kg-day adjusted for continuous 30 exposure of 10 weeks; specified 2 weeks premating, assumed 1 week for successful mating, 31 3 weeks of gestation, and specified 4 weeks to weaning) during mating, pregnancy, and This document is a draftfor review purposes only and does not constitute Agency policy. 2-140 DRAFT--DO NOT CITE OR QUOTE 1 lactational periods (total exposure duration approximately 10 weeks). Corn oil served as the 2 control in another group of 12 dams. Four dams were sacrificed on gestation day (GD) 20 to 3 evaluate the in utero toxicity of TCDD. Litter sizes from the remaining eight dams were 4 examined on postnatal day (PND) 2, and some of the F1 offspring were sacrificed to estimate 5 TCDD tissue concentrations. The remaining offspring were weaned on PND 28. Some of the F1 6 (number not specified) offspring were mated with untreated females on PND 98, following 7 which, litter size, sex ratio, weight, and anogenital distance of F2 pups were examined on 8 PND 2. Mated and unmated F1 males were sacrificed and the testes, epididymis, seminal 9 vesicle, and the ventral prostate were weighed; the cauda epididymis was weighed and examined 10 for sperm count. 11 All fetuses in the control and TCDD group as a result of in utero exposure in the 12 F0 generation survived. Litter size, sex ratio, and anogenital distance in the F1 generation on 13 PND 2 were not altered as a result of in utero TCDD exposure. Pup weight was significantly 14 (p < 0.05) lower in the TCDD-treated group than in controls. TCDD concentration in the 15 adipose tissue of the F0 dams on GD 20 was significantly (p < 0.05) higher than in the liver. 16 Adipose TCDD was significantly (p < 0.01) reduced at weaning, however, compared to 17 concentrations on GD 20. F1 pup liver TCDD concentration increased significantly (p < 0.01) 18 and was higher on PND 28 than PND2. The liver weight in F1 males increased by 14-fold at 19 PND 28 compared to PND 2, implying a transfer of approximately 850 pg of TCDD from the 20 dam to the F1 pup livers during lactation. TCDD also was detected in pup adipose tissue on 21 PND 28. Body weight of TCDD-exposed F1 males was significantly (p < 0.001) lower than 22 control males at weaning (PND 28). No significant differences in testis and cauda epididymis 23 weights were observed between the control and treated groups. Ventral prostate weight in the 24 F1 males exposed to TCDD, however, was approximately 60% lower than controls. No change 25 in weight of the body, brain, testes, cauda epididymis, or seminal vesicle was observed at 26 PND 120. Ventral prostate weight, however, was 16% lower than that of the control group 27 (p < 0.001). Sperm count in the cauda epididymis of the F1 males was not affected by TCDD 28 exposure. 29 Examination of F2 generation litters indicated no significant differences in litter size, pup 30 body weight, and anogenital distance between TCDD-treated or vehicle control groups. The 31 percentage of male F2 pups born to maternally and lactationally TCDD-exposed males was This document is a draftfor review purposes only and does not constitute Agency policy. 2-141 DRAFT--DO NOT CITE OR QUOTE 1 significantly (p < 0.05) lower (38%) than those sired by control group males (52%). Every 2 female mated with maternally TCDD-exposed F1 males delivered more female than male pups. 3 A LOAEL for TCDD of 16.5 ng/kg-day for an estimated 10 week exposure duration in 4 F0 rat dams is identified in this study for decreased development of the ventral prostate in the 5 F1 generation (60% lower than controls) and for significantly (p < 0.05) altered sex ratio 6 (decreased percentage of males) in the F2 generation. A NOAEL cannot be determined for this 7 study. 8 9 2.4.2.I.6. Ish ih ara e t al. (2 0 0 7 ,197677). 10 Ishihara et al. (2007, 197677) examined the effect of repeated TCDD exposure of 11 F0 males on the sex ratio of F1 offspring. Seven-week-old male ICR mice (n = 127) were 12 divided into three groups and treated via gastric intubation with an initial loading dose of either 2 13 or 2,000 ng TCDD/kg BW or an equivalent volume of sesame oil (vehicle) as control, followed 14 by a weekly maintenance doses of 0, 0.4, or 400 ng/kg until the animals were 12 weeks old. 15 One week after the last exposure, the animals were mated with untreated female mice. On the 16 day a vaginal plug was identified, F0 male mice were sacrificed and major organs including 17 testes, epididymis, and liver were removed and weighed. Organ tissues also were examined for 18 histopathological and immunohistochemical changes. Treatment levels, averaged over the 19 6 week period from start of treatment to mating (five maintenance doses), were 0, 0.095, and 20 950 ng/kg-day for the control, low dose and high dose groups, respectively. 21 All TCDD-treated males successfully impregnated untreated females and yielded viable 22 offspring. Mortality, pup weights, and mating and fertility indices were not affected by TCDD 23 exposure. There were no significant differences in body weights or in relative weights of testes, 24 epididymis, or livers in the TCDD-treated F0 males compared to the control group. The livers of 25 some animals (number not specified) in the high-dose group, however, were larger and heavier 26 than in the controls or the low-dose group. Hence, tissues from the high-dose animals were 27 selected for detailed immunohistochemical examination. 28 General histopathological findings in the TCDD-treated groups showed no changes in 29 cell morphology in germ, Sertoli, and Leydig cells of the testes. Arrangement of the germ cells 30 was normal and there was no difference in the epididymis spermatozoon number in either of the 31 TCDD-treated groups compared to controls. Livers of some of the animals in the high-dose This document is a draftfor review purposes only and does not constitute Agency policy. 2-142 DRAFT--DO NOT CITE OR QUOTE 1 group however, showed enlarged and vacuolated areas in the centrilobular area when compared 2 to the low-dose group and the control group. Immunohistochemical and quantitative 3 immunohistological findings showed a marked increase in staining intensity for cytochrome 4 P450 (CYP)1A1 in the cytoplasm of the hepatocytes in the centrilobular area of the high-dose 5 TCDD group compared to the cells in the low-dose and the control groups. In addition, 6 proportions of immunoreactive CYP1A1 areas in the liver sections of the high-dose group were 7 higher than in the low-dose and control groups. The proportions of immunoreactive CYP1A1 8 also varied across animals (n = 33) in the high-dose group. 9 In addition to the above findings, there was a dose-related decrease in the male/female 10 sex ratio. The proportion of male offspring of the high-dose group was significantly lower 11 (p < 0.05) than that observed in controls (46.2% versus 53.1%, respectively). Hepatic 12 immunoreactive CYP1A1 staining levels in individual F0 males were strongly correlated with 13 the sex ratio of their offspring. 14 A LOAEL for TCDD of 950 ng/kg-day for a 6 week exposure duration of F0 male mice 15 is identified for significantly (p < 0.05) decreased male/female sex ratio (i.e., higher proportion 16 of female offspring) in the F1 generation. The NOAEL is 0.095 ng/kg-day. 17 18 2.4.2.I.7. L atch o u m yca n d a n e a n d M a th u r (2 0 0 2 ,197498) (a n d related: L atch o u m yca n d a n e 19 e t al. (2 0 0 2 ,198365; 2 0 0 2 ,197839; 2003, 543746)). 20 Latchoumycandane and Mathur (2002, 197498) conducted a study to determine whether 21 treatment with vitamin E protected rat testes from TCDD-induced oxidative stress. Groups of 22 albino male Wistar rats (n = 6) were administered an oral dose of 0 (vehicle alone) 1, 10, or 23 100 ng TCDD/kg-day for 45 days, while another group of animals (n = 6) was co-administered 24 TCDD at the same doses, along with vitamin E at a therapeutic dose of 20 mg/kg-day for 25 45 days. At study termination, animals were fasted overnight, weighed, and sacrificed. Testis, 26 epididymis, seminal vesicles, and ventral prostate were removed, weighed, and preserved for 27 further examination. The left testis was used to determine daily sperm production, while the 28 right testis was used for biochemical studies. Superoxide dismutase, catalase, glutathione 29 reductase, and glutathione peroxidase activity were measured in the testes, along with production 30 of hydrogen peroxide and lipid peroxidation. This document is a draftfor review purposes only and does not constitute Agency policy. 2-143 DRAFT--DO NOT CITE OR QUOTE 1 Body weights of TCDD-treated rats did not differ significantly from the control group. 2 Testis, epididymis, seminal vesicle, and ventral prostate weights in the TCDD-treated groups, 3 however, decreased significantly (p < 0.05) when compared to controls. None of these changes 4 were observed in the TCDD-exposed groups receiving vitamin E. There was a dose-related 5 decrease in daily sperm production (p < 0.05) in all three TCDD-treated groups when compared 6 to the control group. In contrast, the TCDD treatment groups that also received vitamin E did 7 not show any significant changes in daily sperm production compared to the controls. The 8 TCDD-treated groups also showed significantly (p < 0.05) lower activities of the antioxidant 9 enzymes (superoxide dismutase, catalase, glutathione reductase, and glutathione peroxidase) than 10 the control group. Levels of hydrogen peroxide and lipid peroxidation increased significantly 11 (p < 0.05) in the testes of the rats treated with TCDD compared to the corresponding controls. 12 The TCDD-treated groups that had been co-administered vitamin E show no difference in 13 antioxidant enzyme activities or in reactive oxygen species production when compared with 14 controls. 15 A LOAEL for TCDD of 1.0 ng/kg-day for a 45-day exposure duration in rats is identified 16 in this study for significantly (p < 0.05) reduced sperm production and significantly (p < 0.05) 17 decreased reproductive organ weights. A NOAEL cannot be determined for this study. 18 19 2.4.2.I.8. M u rray et al. (1 9 7 9 ,197983). 20 Male (10-16 per treatment) and female (20-32 per treatment) Sprague-Dawley rats were 21 administered diets containing TCDD (purity >99%) to achieve daily concentrations of 1, 10, or 22 100 ng/kg-day through three generations. After 90 days of treatment, F0 rats were mated to 23 produce F1a offspring. Thirty-three days after weaning of the last F1a litter, the F0 rats were 24 mated again to produce F1b offspring. Some F0 rats were mated a third time for a cross-mating 25 study. The F1b and F2 rats were mated at about 130 days of age to produce the F2 and 26 F3 generations. No clinical signs of toxicity or changes in body weight and food consumption 27 were observed in F0 rats during the 90 days of treatment before mating. The 100 ng/kg-day 28 group was discontinued due to the lack of offspring. In the three surviving offspring (all males), 29 no changes in appearance, body weight, or food consumption occurred. A dose of 10 ng/kg-day 30 caused a consistent decreased body weight in both sexes of F1 and F2 rats, which was associated 31 with decreased food consumption. A significant (p < 0.05) decrease in fertility in F1 and F2 rats This document is a draftfor review purposes only and does not constitute Agency policy. 2-144 DRAFT--DO NOT CITE OR QUOTE 1 occurred, but not in F0 rats, administered 10 ng/kg-day. The number of live pups and gestational 2 survival index were significantly (p < 0.05) decreased in the 100 ng/kg-day F0 rats and in the 3 10 ng/kg-day F1 and F2 rats. The gestational survival index also was significantly (p < 0.05) 4 decreased in F2 rats administered 1 ng/kg-day. Postnatal survival was significantly (p < 0.05) 5 reduced only in F2 rats administered 10 ng/kg-day. Growth (as measured by body weight) was 6 affected at 10 ng/kg-day only in the third generation. In the 10 ng/kg-day group, a significant 7 (p < 0.05) decrease in relative thymus weight and increase in liver weight also occurred in F3rats 8 (weights were not measured in F2 rats). Additionally, mating 100 ng/kg-day TCDD-treated 9 females with untreated males increased the percent of implants resorbed as assessed by uterine 10 histopathology. 11 The reproductive LOAEL is 10 ng/kg-day, based on a significant (p < 0.05) decrease in 12 fertility (33-37% lower than controls); decrease in the number of live pups (18-27% lower than 13 controls); decrease in gestational survival (10-11% lower than controls); decrease in postnatal 14 survival (32% lower than controls); and decreased postnatal body weight (14-19% lower than 15 controls at weaning) in one or more generations. The reproductive NOAEL is 1 ng/kg-day. 16 17 2.4.2.I.9. R ie r e t al. (1 9 9 3 ,199987; 1 9 9 5 ,198566). 18 Reir et al. (1993, 199987; 1995, 198566) examined the impact of chronic TCDD 19 exposure on endometriosis in monkeys. Female rhesus monkeys (eight animals per treatment 20 group) were exposed to 0, 5, or 25 ppt TCDD (purity not specified) in feed for 4 years. 21 Previously, Bowman et al. (1989, 543745) determined that these dietary concentrations were 22 equivalent to 0, 0.15, and 0.67 ng/kg-day, respectively. Ten years after termination of TCDD 23 treatment, the presence of endometriosis was determined via laparoscopic surgical procedure, 24 and the severity of the disease was assessed. The study authors reported that three monkeys in 25 the 0.67 ng/kg-day exposure group died at 7, 9, and 10 years after termination of TCDD 26 treatment. Autopsy results attributed the deaths to widespread and severe peritoneal 27 endometriosis (all three monkeys) along with obstruction of the colon (one monkey) and 28 blockage of the jejunum (one monkey). Other deaths also occurred in the control group (1 death 29 from birthing complications and another from an unknown cause); in the 0.15 ng/kg-day dose 30 group (1 death due to natural causes with no endometriosis), and in the 0.67 ng/kg-day dose 31 group (1 death due to a breeding fight with no incidence of endometriosis). At study This document is a draftfor review purposes only and does not constitute Agency policy. 2-145 DRAFT--DO NOT CITE OR QUOTE 1 termination, 17 live animals plus the 3 that had previously died of endometriosis were evaluated 2 (total n = 20). 3 Incidence of endometriosis was significantly (p < 0.05) higher than in the control group 4 with 71 and 86 % incidence rates in the 0.15 and 0.67 ng/kg-day dose groups, respectively, 5 compared to 33% in the control group. Severity of endometriosis was also significantly 6 (p < 0.001) correlated with TCDD dose. Staging by rAFS indicated that untreated control 7 animals had either minimal or no incidence of endometriosis. In comparison, endometriosis was 8 absent in 2 of the 7 monkeys in the 0.15 ng/kg-day dose group, while only 1 of the 7 animals in 9 the high dose group was disease free. Moderate-to-severe disease was observed in 3 of the 10 7 animals in the 0.15 ng/kg-day dose group and 5 of the 7 animals in the 0.67 ng/kg-day dose 11 group. Moderate-to-severe disease was not observed in the control group. The authors also 12 compared the incidence and severity of endometriosis in TCDD-exposed animals with 13 304 normal, non-neutered females with no dioxin exposure and reported that the disease was not 14 present in monkeys that were less than 13 years of age, while the disease rate was 30% among 15 animals 13 years of age or older. The study authors report that these findings are in agreement 16 with human and rhesus studies demonstrating that the prevalence of detectable endometriosis can 17 increase with advanced age. 18 As noted previously, in a follow-up study, Rier et al. (2001, 198776) examined the DLC 19 levels of sera collected from some monkeys in this study. They reported that animals in this 20 study had elevated serum PCB77 and PCB126 levels and an increased serum TEQ; the fractional 21 contribution of serum TCDD levels to total serum TEQ was 30% in treated animals. They also 22 reported that the severity of the endometriosis corresponded to the serum PCB77 concentrations 23 rather than total TCDD. In this study, it is not possible to determine the contribution of TCDD 24 alone to the endometriosis due to the background contamination; thus, EPA has not developed a 25 TCDD LOAEL from the study. 26 27 2.4.2.1.10. S h i et al. (2 0 0 7 ,198147). 28 Pregnant Sprague-Dawley rat dams (3 per treatment group) were administered 0, 1, 5, 50, 29 or 200 ng/kg TCDD (purity >99%) in corn oil by gavage on GD 14 and GD 21 and on PND 7 30 and PND 14 for lactational exposure to pups (Shi et al., 2007, 198147). Ten female pups per 31 treatment were selected and administered TCDD weekly at the same dose levels through their This document is a draftfor review purposes only and does not constitute Agency policy. 2-146 DRAFT--DO NOT CITE OR QUOTE 1 reproductive lifespan (approximately 11 months). The corresponding equivalent daily TCDD 2 doses are 0, 0.14, 0.71, 7.14, and 28.6 ng/kg-day. Vaginal opening was slightly but significantly 3 (p < 0.05) delayed in 28.6 ng/kg-day females. Vaginal opening was also delayed, but not 4 significantly, in the 0.14 and 7.14 ng/kg-day groups. Reproductive senescence with normal 5 cyclicity was significantly (p < 0.05) accelerated beginning at 9 months in 7.14 and 6 28.6 ng/kg-day females. Serum estradiol concentrations were decreased at all time points across 7 the estrous cycle in a dose-dependent manner with a statistically significant decrease (p < 0.05) 8 in all but the lowest dose group. TCDD exposure, however, did not affect the number or size 9 distribution of ovarian follicles; responsiveness of the pituitary gland to gonadotropin-releasing 10 hormone, or serum profiles of FSH, LH, or progesterone. 11 A LOAEL for TCDD of 0.71 ng/kg-day for an 11-month exposure duration was 12 identified in this study based on significantly (p < 0.05) decreased estradiol levels in offspring. 13 The NOAEL for this study is 0.14 ng/kg-day. 14 15 2.4.2.1.11. Yang et al. (2 0 0 0 ,198590). 16 Yang et al. (2000, 198590) studied the impact of TCDD exposure on the incidence and 17 severity of endometriosis in female rhesus monkeys. Groups of 7- to 10-year old nulliparous 18 cynomolgus monkeys were treated with 0 (n = 5), 1, 5, or 25 (n = 6 per group) ng/kg BW TCDD 19 5 days per week via gelatin capsules for 12 months. Because the monkeys received one capsule 20 5 days per week, the doses adjusted for continuous exposure were 0, 0.71, 3.57, and 21 17.86 ng/kg-day. Prior to TCDD administration, all animals had endometriosis induced during 22 days 12-14 of the menstrual cycle by auto-transplantation of endometrial-strips in multiple 23 abdominal sites. All TCDD-treated and control groups were laparoscopically examined during 24 months 1, 3, and 6 to monitor the survival of endometrial implantations and to obtain peritoneal 25 fluid to determine the concentration and immunotype of endometrial growth regulator cytokines 26 interleukin-6 (IL-6) and interleukin-6 soluble receptor (IL-6sR). Because insufficient peritoneal 27 fluids were present in the treated and control monkeys, however, the study authors collected 28 blood samples at 6 and 12 months during laparoscopy for routine hematology and to assess the 29 circulating levels of IL-6 and IL-6sR. All animals were sacrificed at 12 months, and circulating 30 levels of gonadal steroids also were measured at the time of necropsy. This document is a draftfor review purposes only and does not constitute Agency policy. 2-147 DRAFT--DO NOT CITE OR QUOTE 1 No changes were observed among treatment levels in general toxicological endpoints 2 such as body weight changes, food consumption, hematological endpoints, general activity 3 levels, and caretaker interaction. In addition, TCDD did not impact circulating levels of gonadal 4 steroids measured during necropsy. Similarly, there were no differences in the number of 5 menstrual cycles, the length of the menstrual cycle, and bleeding intervals. Endometrial implants 6 were found in at least one site in all TCDD-treated and control monkeys during the 7 first laparoscopic examination. Follow-up laparoscopies revealed that there was a continuous 8 loss of endometrial implants over time in each dose group. At the 1-, 3-, and 6-month 9 examination, the number of endometrial losses was not significantly different among different 10 dose groups. At the 12-month examination, however, a significantly (p < 0.05) higher rate of 11 survival of endometrial implants was observed in the 3.57 and 17.86 ng/kg-day dose groups 12 compared to the control group. The highest rate of endometrial implant survival was observed in 13 the ovaries regardless of the dose group. In contrast, all lesions disappeared from the left broad 14 ligament, whereas two on the right broad ligament and one on the uterine fundus survived. 15 There was a dose-dependent divergence in the growth response of endometrial implants 16 following TCDD exposure. Both the maximum and minimum implant diameters in the 17 17.86 ng/kg-day dose group were significantly (p < 0.05) larger compared to controls. In 18 contrast, the maximum and minimum implant diameters in the 0.71 ng/kg-day dose group were 19 significantly (p < 0.05) smaller compared to controls. TCDD did not impact implant diameters 20 in the 3.57 ng/kg-day dose group when compared to controls. Histological examinations 21 revealed that endometrial glands and stromal cells were present in all surviving implants. 22 Sections examined in the 17.86 ng/kg-day of TCDD possessed cystic endometrial glands that 23 were more frequently observed in this dose group compared to other groups including controls. 24 In addition, circulating levels of IL-6 were significantly (p < 0.05) lower in monkeys exposed to 25 17.86 ng/kg-day TCDD both at 6 and 12 months compared to the control group. In contrast, 26 circulating levels of IL-6sR were significantly (p < 0.05) higher in animals treated with 3.57 and 27 17.86 ng/kg-day TCDD at 6 months, while the levels were higher only in the 17.86 ng/kg-day 28 TCDD group at 12 months. 29 A LOAEL for TCDD of 17.86 ng/kg-day for a 1 year exposure duration was identified in 30 this study for significantly (p < 0.05) increased endometriosis induced by endometrial implant 31 survival, significantly (p < 0.05) increased maximum and minimum implant diameters, and This document is a draftfor review purposes only and does not constitute Agency policy. 2-148 DRAFT--DO NOT CITE OR QUOTE 1 growth regulatory cytokine dysregulation (as assessed by significantly decreased IL-6 levels, 2 p < 0.05). A NOAEL of 3.57 ng/kg-day is identified in this study. 3 4 2.4.2.2. D evelo p m en ta l S tu dies 5 2.4.2.2.I. A m in e t al. (2 0 0 0 ,197169). 6 Amin et al. (2000, 197169) studied the impact of in-utero TCDD exposure on the 7 reproductive behavior in male pups. Groups of pregnant Harlan Sprague-Dawley rats (n = 108 8 divided into 4 cohorts; number of animals in the TCDD treatment group is ~3 per dose group) 9 were dosed via gavage with 0, 25, or 100 ng/kg-day TCDD (purity >98%) in corn oil on GDs 10 10-16. On the day of birth (PND 0), pups were examined for gross abnormalities and the 11 number of live pups, their weights, and sex were recorded from each litter. Litters consisting of 12 more than eight pups were reduced to eight, comprised of four males and four females when 13 possible. Litters consisting of fewer than five pups were excluded from the study to minimize 14 between-litter differences in growth rate, maternal behavior, and lactational exposure. After this 15 exclusion, approximately 10 to 11 litters per exposure group remained. All pups were weaned 16 on day 21 and one male and one female were retained to assess reproductive development, play 17 behavior, reproductive behavior, and saccharin preference behavior. Both male and female pups 18 were tested for saccharin preference between 189 and 234 days of age. A saccharin preference 19 test was conducted for 8 days. For the first 4 days, rats were provided bottles containing tap 20 water, and on days 5 and 6 the animals were provided a bottle containing water and a bottle 21 containing 0.25% saccharin solution. On days 7 and 8, the animals were provided water and a 22 bottle containing 0.50% of saccharin solution. A 0.50% saccharin solution was used because 23 previous studies have reported that male rats exhibited a greater reduction in preference for this 24 saccharin concentration compared to females, hence the sex difference in preference is more 25 marked at this saccharine dose. 26 None of the treated dams exhibited any signs of toxicity as a result of exposure to TCDD. 27 Gestational body weight, liver weight, litter size and percent live births were all comparable to 28 the corresponding control group. Birth rate and weaning weight of the pups also were not 29 affected by TCDD exposure. Sex-related water consumption, however, was significantly 30 (p < 0.001) affected during the first 4 days with female pups drinking more water per 100 g of 31 body weight compared to the respective male counterparts. Saccharin consumption was This document is a draftfor review purposes only and does not constitute Agency policy. 2-149 DRAFT--DO NOT CITE OR QUOTE 1 significantly (p < 0.001) affected, with females consuming greater amounts of saccharin solution 2 per 100 g body weight compared to the corresponding males. Additionally, both male and 3 female pups drank significantly (p < 0.001) more of the 0.25% saccharin solution compared to 4 the 0.50% saccharin solution. Females of all exposure groups consumed less of both the 0.25 5 and 0.50% saccharin solution compared to the same-sex control group. Comparisons of each 6 exposure group to the control group indicated that only the high TCDD exposure group 7 (100 ng/kg-day) different significantly (p < 0.05) compared to control in the consumption of 8 0.25% saccharin solution. In contrast, for the 0.50% saccharin solution, both the low and high 9 TCDD dose groups differed significantly (p < 0.05 andp < 0.01, respectively) compared to the 10 control group. The saccharin preference of TCDD-exposed male rats did not differ from that of 11 the male control group. The TCDD-exposed females' preference for saccharin solution, 12 however, was significantly reduced in both the 25 (p < 0.05) and the 100 ng/kg-day (p < 0.005) 13 dose group compared to that of the female controls. The study authors state that the reduction in 14 saccharin consumption and preference in females could be due to the anti-estrogenic action of 15 TCDD and that recent research reports suggest that TCDD can decrease the level of estrogen 16 receptor (ER) mRNA by blocking the ability of ER to transactivate from the estrogen response 17 element. 18 A LOAEL for TCDD of 25 ng/kg-day for 7 days of gestational exposure is identified for 19 significantly (p < 0.05) decreased preference in the consumption of 0.25% saccharin solution. A 20 NOAEL cannot be determined for this study. 21 22 2.4.2.2.2. B e ll et al. (2 0 0 7 ,197041). 23 Bell et al. (2007, 197041) examined the reproductive effects of TCDD in rats exposed 24 during development. Female CRL:WI (Han) rats were treated with TCDD (99% purity; 25 dissolved in acetone) in the diet at concentrations of 0 (acetone alone; n = 75), 28, 93, or 26 530 (n = 65/group) ng TCDD/kg diet, which provided average doses of 0, 2.4, 8, or 27 46 ng/kg-day, respectively. Rats were exposed to TCDD 12 weeks prior to mating, during 28 mating, and through pregnancy. Dams were switched to the control diet after parturition. Litters 29 from pregnant dams were reduced to a maximum size of eight on PND 4 and to five males (if 30 possible) on PND 21. These males were left untreated until sacrificed (25/group, one/litter) on 31 PND 70, while all remaining animals were sacrificed on PND 120. All sacrificed animals were This document is a draftfor review purposes only and does not constitute Agency policy. 2-150 DRAFT--DO NOT CITE OR QUOTE 1 necropsied and received a seminology examination. Prior to sacrifice, during weeks 12 and 13, 2 20 animals from each dose group were tested for learning ability and motor activity, and were 3 also administered a functional observation battery. During postnatal week 16, groups of 20 male 4 F1 rats from each treatment group were paired with untreated virgin females for 7 days, and 5 mated females were killed on GD 16 and examined for terminal body weights, pregnancy status, 6 number of corpora lutea, and number of intrauterine implantations. 7 The study authors found no evidence of direct maternal toxicity from exposure to TCDD. 8 In the high-dose groups, 8 of 27 dams suffered complete litter loss compared to 3 dams in the 9 control group, but the difference was not statistically significant. Pup survival at PND 4 was also 10 lower in the high-dose group, but the difference again was not statistically significant. 11 A dose-related decrease in mean pup body weight was observed on PND 1, and this trend 12 continued throughout the lactation period. High-dose male pups had lower body weights when 13 compared to controls at PND 21, with this trend continuing over the course of the study. 14 Balanopreputial separation (BPS) was significantly (p < 0.05) delayed compared to controls in 15 all three treatment groups by 1.8, 1.9, and 4.4 days in the low-, medium-, and high-dose groups, 16 respectively. The study authors reported that adjustment for lower body weights observed at 17 PND 21 and PND 42 did not affect the estimate of delay in BPS. No adverse effects from 18 maternal treatment were observed on learning or in functional observational battery performance. 19 Offspring in the high-dose group exhibited less activity when compared to controls (p < 0.05) 20 when they were subjected to a test of motor activity for 30 minutes. 21 The median precoital time was 2-3 days for all 20 F1 males that were mated during 22 postnatal week 16. The uterine and implantation data were similar in all dose groups and there 23 were no significant differences in the proportion of male offspring between groups. Epididymal 24 sperm counts and sperm motility did not differ significantly between dose groups in animals 25 sacrificed during postnatal week 10. The mean number of spermatids was significantly lower 26 (14%; p < 0.05) and the proportion of abnormal sperm was significantly (p < 0.05) higher in the 27 high-dose group when compared to controls on PND 70. These effects, however, were not seen 28 in animals sacrificed on PND 120. 29 Terminal body weights were significantly (p < 0.05) decreased in the high-dose group 30 (6.9 %) compared to controls on PND 120, while the depression in body weight in the 31 medium-dose group (5.5%) was not statistically significant. At PND 70, the relative and This document is a draftfor review purposes only and does not constitute Agency policy. 2-151 DRAFT--DO NOT CITE OR QUOTE 1 absolute testis weight of the high-dose group was less than the controls (12 and 18%, 2 respectively). Absolute spleen weight in the high-dose group was significantly higher (8%) on 3 PND 70, and increased significantly (p < 0.05) by 1-3% on PND 120 in all dose groups 4 compared to controls. Kidney weight in the low and medium-dose groups was significantly 5 (p < 0.05) greater than in controls (~2%) at PND 120. In addition to these organs, ventral 6 prostate (9.4%) and relative liver (~4.5%) weights were significantly (p < 0.05) higher than 7 controls on PND 120 in the medium- and low- and high-dose groups, respectively. On 8 PND 120, absolute brain weight was significantly (p < 0.05) less than the control in the 9 medium-dose group, while relative brain weight was significantly (p < 0.05) higher than the 10 control in the low- and high-dose group. Histological examination revealed no unusual findings. 11 A LOAEL for TCDD of 2.4 ng/kg-day following an estimated 17 week exposure duration 12 of dams was identified in this study for significantly (p < 0.05) delayed BPS. A NOAEL was not 13 identified in this study. 14 15 2.4.2.2.3. F ran czak et al. (2 0 0 6 ,197354). 16 Franczak et al. (2006, 197354) examined the impact of chronic TCDD exposure on the 17 onset of reproductive senescence in female rats. Pregnant Sprague-Dawley rats 18 (n = 2-3/dose group) were fed 50 or 200 ng/kg TCDD (>99% purity) or corn oil vehicle 19 (4 mL/kg) orally on GD 14 and 21 and PND 7 and 14 to provide in utero and lactational 20 exposure to TCDD. On PND 21, female pups (n = 7/dose group) were weaned and were 21 subsequently given weekly doses of 50 or 200 ng/kg-week TCDD by gavage (7.14 or 22 28.6 ng/kg-day adjusted for continuous exposure; administered doses divided by 7) or corn oil 23 vehicle. Exposure continued for up to 8 months, and animals were observed for changes in 24 estrus cycle at 4, 6, and 8 months. Rats were sacrificed at 8 months of age when the 25 TCDD-treated animals had entered the transition to reproductive senescence. Following 26 sacrifice, diestrus concentrations of serum LH, FSH, progesterone, and estradiol were measured, 27 and the ovaries were collected for examination. 28 Estrus cycles at 4 months exhibited normal cyclicity in both TCDD-exposed groups and 29 did not differ significantly from the control group. At 6 months, however, there was a tendency 30 (p < 0.1) toward loss of normal estrus cyclicity in animals treated with TCDD. At the 8 month 31 observation, estrus cyclicity was significantly (p < 0.05) different in both dioxin-exposed groups This document is a draftfor review purposes only and does not constitute Agency policy. 2-152 DRAFT--DO NOT CITE OR QUOTE 1 compared to controls (cumulative TCDD exposure is reported as 1.7 and 8 pg/kg for the 50 and 2 200 ng/kg dose groups, respectively). The study authors noted that although the low-dose 3 animals showed an increased prevalence of prolonged cycles, persistent estrus or diestrus was 4 observed in only 10% of the rats. Conversely, approximately 50% of the rats exhibited loss of 5 cyclicity in the high-dose group. There were no changes in the number and size distribution of 6 ovarian follicles or the number of corpora lutea at either dose. Progesterone levels at 8 months 7 tended to be higher (p < 0.08) in animals receiving either 7.14 or 28.6 ng/kg-day TCDD 8 compared to controls, while serum estradiol concentrations were significantly (p < 0.03) lower at 9 diestrus. Serum LH levels in TCDD-treated animals were comparable to those in the control 10 group, while FSH levels were elevated in rats receiving 7.14 ng/kg-day TCDD, but not in the 11 28.6 ng/kg-day dose group. 12 A LOAEL for TCDD of 7.14 ng/kg-day for an 8-month exposure duration was identified 13 for significantly (p < 0.03) decreased serum estradiol levels. A NOAEL cannot be determined 14 for this study. 15 16 2.4.2.2.4. H o jo e t al. (2 0 0 2 ,198785) (a n d related: Z a reb a et al. (2 0 0 2 ,197567)). 17 Hojo et al. (2002, 198785) studied the impact of prenatal exposure to TCDD on sexually 18 dimorphic behavior in rats. Thirty-six pregnant Sprague-Dawley rats were assigned according to 19 a randomized block design to groups receiving 0, 20, 60, or 180 ng/kg TCDD (98% purity) on 20 GD 8. Litters from pregnant dams were culled to 5 females and 5 males on PND 4 and allowed 21 to wean normally, at which time 5, 5, 6, and 5 litters from the 0, 20, 60, and 180 ng/kg TCDD 22 treatment groups, respectively, were maintained for examination of behavioral response. 23 Offspring were exposed to TCDD (from a single maternal exposure) for about 35 days through 24 gestation and lactation. After weaning at PND 21, offspring were fed ad libitum until PND 80, at 25 which time a fixed amount of food was supplied daily to maintain constant body weights. At 26 90 days old, the rats in these treatment groups were trained to press a lever to obtain food pellets 27 using two operant behavior procedures. Initially, each lever press was reinforced. The 28 fixed-ratio (FR) requirement was then increased every fourth session from the initial setting of 1 29 to values between 6 and 71. The responses for 30 days were studied under a multiple schedule 30 combining FR 11 and another schedule requiring a pause of at least 10 sec between responses 31 (differential reinforcement of low rate, or DRL 10-sec) This document is a draftfor review purposes only and does not constitute Agency policy. 2-153 DRAFT--DO NOT CITE OR QUOTE 1 Pup and dam body weights were not affected by TCDD exposure, and all pups were 2 successfully trained in the lever-press response within 3-4 days. Analyses of the FR procedure 3 data indicated that the male pups responded at a lower rate at all TCDD doses when compared to 4 the control group. In case of female pups, all TCDD-treated groups responded at a higher rate 5 than controls. None of these results were, by themselves, however, statistically significant. 6 Examination of the FR 11 and DRL 10-second data indicated that when considering the FR 7 component of this multiple procedure, males from all three treatment groups responded at lower 8 rates when compared to the controls. Conversely, all female pups responded at a higher rate than 9 controls. In addition, the treatment-by-sex interaction was significant (p = 0.036), with the 10 60 ng/kg female pups responding at a higher rate than the 60-ng/kg male pups. Examination of 11 the delayed response component in the multiple FR 11 and DRL 10-sec procedures indicated that 12 almost all TCDD treatment groups were affected. Like the FR component, male pups at all 13 TCDD dose groups responded at a lower rate compared to controls, while female pups at all dose 14 groups responded at a higher rate than controls. There was also a significant (p = 0.001) 15 sex-by-treatment interaction for the DRL 10-sec similar to the FR component. Following 16 behavioral testing, the animals were sacrificed and cortical depth measurements were taken in 17 selected right and left brain regions. Reduced cortical thickness and altered brain morphometry 18 were observed in both male and female offspring in the 180-ng/kg exposure group when 19 compared to controls (reported in a separate article; Zareba et al., 2002, 197567). 20 A nominal LOAEL for TCDD of 20 ng/kg for a single exposure on GD 8 is established 21 for this study based on abrogation of sexually dimorphic neurobehavioral responses. A NOAEL 22 cannot be derived for this study. 23 24 2.4.2.2.5. K attain en e t al. (2 0 0 1 ,198952). 25 Pregnant Line A, B, and C rats derived from Han/Wistar and Long-Evans rats 26 (4-8 pregnant dams/strain/treatment group) were administered a single gavage dose of 0, 30, 27 100, 300, or 1,000 ng/kg TCDD (purity >99%) in corn oil on GD 15 (Kattainen et al., 2001, 28 198952). On PND 1, the litters were culled to three males and three females. Offspring were 29 weaned on PND 28. Female pups were sacrificed on PND 35 and male pups were sacrificed on 30 PND 70. TCDD treatment did not affect body weight or cause clinical signs of toxicity in the 31 dams. In Line B offspring, body weights in the 1,000 ng/kg group were slightly decreased This document is a draftfor review purposes only and does not constitute Agency policy. 2-154 DRAFT--DO NOT CITE OR QUOTE 1 during PND 1-7, while Line C offspring had slightly decreased body weights throughout the 2 study period (data were not provided). The development of the third molar was affected the 3 most in Line C offspring. In 5 of 10 Line C females and 6 of 10 Line C males treated with 4 1,000 ng/kg TCDD, the lower third molar did not develop. In comparison, 1 of 19 Line A 5 females and 1 of 18 Line B females administered 1,000 ng/kg TCDD lacked the third molar at 6 sacrifice. Third molars were present in all the controls and all male Line A and B offspring 7 administered 1,000 ng/kg. Due to the lack of eruption of the third molar in the majority of 8 Line B and C control females (only 30% erupted), however, the effects of TCDD on third molar 9 eruption could only be evaluated in Line A female offspring (with 94% eruption). There was a 10 dose-dependent decrease in the eruption of the lower third molar in Line A female offspring with 11 a significant (p < 0.05) decrease observed in the 300 and 1,000 ng/kg dose groups. In the male 12 offspring, any third molar that developed erupted by PND 70. The mesiodistal length of the 13 existing lower third molar was reduced in a dose-dependent manner in both genders of all 14 three rat lines. In Line A and C females, the decrease was significant (p < 0.05) at all doses. The 15 size of the second molars was also significantly decreased with 1,000 ng/kg (p < 0.05) in all but 16 Line C males. 17 A developmental LOAEL for TCDD of 30 ng/kg for maternal exposure on GD 15 is 18 established for this study, based on impaired tooth development (significantly reduced 19 mesiodistal length of the lower third molar by approximately 12% to 38% [p < 0.05]). A 20 NOAEL could not be determined. 21 22 2.4.2.2.6. K e lle r e t al. (2 0 0 7 ,198526; 2 0 0 8 ,198531; 2 0 0 8 ,198033). 23 Keller et al. (2007, 198526; 2008, 198531; 2008, 198033) conducted three separate 24 experiments to assess the impact of TCDD on molar tooth development using different mouse 25 strains. In Experiment 1, Keller et al. (2007, 198526) used six inbred mouse strains (C57BL/6J, 26 BALB/cByJ, A/J, CBA/J, C3H/HeJ, and C57BL/10J) known to possess high affinity ligand 27 binding aryl hydrocarbon receptor alleles (b), two with b1 alleles (C57BL/6J and CBA/J), and 28 four with b2 alleles (BALB/cByJ, A/J, C3H/HeJ, and CBA/J). Females (number not specified) 29 from each strain were mated with males of the same strain. On GD 13, each pregnant female 30 was assigned to one of the four dose groups and treated with 0, 10, 100, or 1,000 ng TCDD/kg 31 BW via oral gavage. The control group received corn oil. GD 13 was chosen for dosing because This document is a draftfor review purposes only and does not constitute Agency policy. 2-155 DRAFT--DO NOT CITE OR QUOTE 1 the first morphological signs of tooth development occur on GD 11. The first visible signs of the 2 Ml (molar) occur on GDs 13-14 followed by final cuspal morphology, which is determined on 3 GD 15. The F1 offspring of females from each strain were weaned and separated by sex at PND 4 28 and were euthanized at PND 70. Each F1 mouse was examined for the presence or absence 5 of both maxillary (M3) and mandibular third molars (M3) on both the left and right sides. In 6 addition, all mice were scored as either normal or variant in M1morphology for both molar rows. 7 In Experiment 2 (Keller et al., 2008, 198531), dams from six inbred mouse strains 8 (C57BL/6J, BALB/cByJ, A/J, CBA/J, C3H/HeJ, and C57BL/10J) were orally dosed on GD 13 9 with 0, 10, 100, or 1,000 ng TCDD/kg BW in corn oil. GD 13 was used as the dosing day 10 because it coincided with the formation of Meckel's cartilage (a major signal center) in the 11 mouse mandible that is followed shortly by intramembranous bone formation on GD 15. The 12 A/J mouse strain was abandoned because the authors had difficulty rearing the offspring from 13 this strain. All offspring (n = 4 or 5 per treatment group) from the remaining strains were 14 euthanized at 70 days of age. Mandible size and shape from all selected offspring were 15 examined using geometric morphometric methods to assess the impact of TCDD exposure. 16 In Experiment 3 (Keller et al., 2008, 198033), dams from six inbred mouse strains 17 (C57BL/6J, BALB/cByJ, A/J, C3H/HeJ, CBA/J, and C57BL/10J) were treated with a single oral 18 dose of 0, 10, 100, or 1,000 ng TCDD/kg-BW in corn oil. GD 13 was chosen as the dosing day 19 because the first visible signs of the first molar (M1) occurs on GDs 13-14 and the final cuspal 20 morphology (the pattern of projections on the chewing surface of the tooth) is not determined 21 until after GD 15. Similar to Experiment 2, the A/J mouse strain was abandoned due to 22 difficulty in rearing offspring. All offspring (n = 107-110 in each of the five strains for all 23 treatment groups) were euthanized at 70 days of age and their molar size, shape, and asymmetry 24 traits were examined using geometric morphometric methods. 25 In Experiment 1, all four M3s were present in all dose groups in mice from C57BL/6J, 26 BALB/cByJ, and C57BL/10J strains. A similar response was observed in the A/J strain mice 27 with only 3 of 51 F1 mice exhibiting missing third molars. Approximately one-third of the mice 28 from the CBA/J and C3H/HeJ strains, however, were missing at least one M3 or M3molar. The 29 numbers of CBA/J mice missing one or both M3or M3molars were 0/29, 2/21, 6/29, and 30/30 30 in the 0, 10, 100, and 1,000 ng/kg groups, respectively. In the C3H/Hej animals, the numbers 31 missing one or both molars were 1/24, 3/28, 1/26, and 30/36, respectively. This document is a draftfor review purposes only and does not constitute Agency policy. 2-156 DRAFT--DO NOT CITE OR QUOTE 1 Maternal TCDD exposure was also found to affect the frequency of M1variants, but only 2 in the C57BL/10J strain, and the dose-response relationship was nonmonotonic. The proportions 3 of variants observed in the 0, 10, 100, and 1,000 ng/kg dose groups were 33, 68, 59, and 58%, 4 respectively. 5 A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 is identified for this study 6 for increased incidence (33%) of the M1variant in the C57BL/10J mouse strain. A NOAEL 7 cannot be determined in this study. 8 In Experiment 2 TCDD exposure of dams did not affect offspring survival or 10-week 9 body weight in any of the inbred mouse strains used. Analysis of variance (ANOVA) indicated 10 that although mandible size in both male and female offspring varied significantly (p < 0.0001) 11 among strains, it was not affected by TCDD exposure. In contrast, analysis of covariance 12 indicated that TCDD exposure significantly (p = 0.0033) decreased the mandible size in male 13 offspring in the C3H/HeJ strain at all treatment groups. The mean mandible size was similar 14 across all treatment groups in both sexes in all strains with male offspring exhibiting larger 15 mandibles compared to females. Males in the C3H/HeJ strain exhibited a significant (level not 16 reported) downward trend in mandible size throughout all treatment groups. Females in the 17 C3H strain also showed a similar trend in mandible size, but the trend was not significant. 18 ANOVA on mandible shape indicated that males had significantly (p < 0.0001) different 19 mandible shape in strain x treatment groups. In contrast, in female offspring, although the 20 mandible shape was significantly (p < 0.0001) different due to strains, treatment groups, and 21 litter, the strain x treatment interaction was not significant. Male offspring from the C3H/HeJ 22 and C57BL/6J mouse strains appear to be more sensitive to TCDD than BALB/cByJ or 23 CBA/J mice, with the C57BL/10J strain exhibiting intermediate sensitivity. In addition to these 24 analyses, Procrustes distance analysis also indicated that C3H/HeJ mice had the greatest 25 response to the highest dose of TCDD, followed by the C57BL/6J strain. Female offspring in the 26 C3H/HeJ and C57BL/6J strains also exhibited the largest change in Procrustes distance with 27 TCDD exposure. This trend, however, was not statistically significant (p = 0.29). 28 A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 was identified for this 29 study for significantly (p = 0.0033) decreased mandible shape and size in male C3H/HeJ mice. 30 A NOAEL cannot be determined in this study. This document is a draftfor review purposes only and does not constitute Agency policy. 2-157 DRAFT--DO NOT CITE OR QUOTE 1 In Experiment 3, effect of TCDD exposure on offspring survival or body weight was not 2 reported. Three-way ANOVA results showed significant (p < 0.0001) differences in molar size 3 among strains, sexes, and litters, but not between treatment groups. Molar size difference in 4 sex x strain interaction was significant (p = 0.03), whereas differences in sex x treatment and 5 sex x strain x treatment were not significant. Additionally, molar size in treatment x strain 6 interaction also was not statistically significant. Based on these results, the authors reported that 7 molar size varied significantly (p < 0.0001) among all five strains tested, with all strains 8 exhibiting similar trends in all four treatment groups. Strain differences in molar size were more 9 apparent in male offspring. A hormesis-like trend in molar size was observed in all strains 10 (except in BALBc/ByJ) and sexes with an increase at the 100 ng/kg dose and a decrease in the 11 1,000 ng/kg dose. In addition to lack of difference in molar size for all treatment groups in all 12 strains, fluctuating asymmetry in molar size also did not increase with increasing doses of 13 TCDD. 14 In contrast to these results on molar size, the Procrustes ANOVA indicated that molar 15 shape was significantly (p < 0.0001) affected by strain, sex, treatment, and litter size. Molar 16 shape in sex x strain and sex x strain x treatment interactions was also highly significant 17 (p < 0.0001). Based on these results, the authors concluded that differences between males and 18 females varied based on the strain, and that the effect of TCDD exposure on each strain also 19 differed for male and female offspring. Because molar shape in treatment x strain interaction 20 was significant (p < 0.0001), differences in molar shape between the three treatment groups and 21 the control group were analyzed for each strain using nonorthogonal contrasts. In male 22 offspring, contrasts between the control group and 1,000 ng/kg were statistically significant only 23 in the C3H/HeJ (p < 0.0001) and CBA/J (p < 0.03) strains. These results suggest that these 24 two strains are most susceptible to TCDD effect on molar shape, and similar results were 25 observed in female offspring of these two strains. The contrast in molar shape between the 26 control and the 100 ng/kg treatment group for the female C57BL/6J mice also was statistically 27 significant (p = 0.0096). On the whole, when considering Procrustes distance results for molar 28 shape, the C3H/HeJ male offspring had the largest response at the low and high doses, while the 29 female offspring had the largest response at low and mid doses. This observation in male 30 C3H/HeJ mice is consistent with that of TCDD-induced changes in mandible size from Keller 31 et al. (2008, 198531). This document is a draftfor review purposes only and does not constitute Agency policy. 2-158 DRAFT--DO NOT CITE OR QUOTE 1 A LOAEL for TCDD of 10 ng/kg maternal exposure on GD 13 is identified for this study 2 for significant (p < 0.0001) differences in molar shape in male C3H/HeJ mice. A NOAEL 3 cannot be determined in this study. 4 5 2.4.2.2.7. K u ch iiw a et al. (2 0 0 2 ,198355). 6 Kuchiiwa et al. (2002, 198355) studied the impact of in utero and lactational TCDD 7 exposure on serotonin-immunoreactive neurons in raphae nuclei on F1 male mouse offspring. 8 Twenty-one adult female ddY mice (seven per treatment group) were administered TCDD 9 (99.1% purity) by oral gavage once a week for 8 weeks at doses of 0, 4.9, or 490 ng/kg (0, 0.7, or 10 70 ng/kg-day average daily dose; administered doses divided by 7) or an equivalent volume of 11 olive oil vehicle (6.7 mL/kg) by gavage. Immediately following the final treatment, the mice 12 were housed with untreated male mice for mating. At approximately 20-21 days after mating, 13 3 female mice from each dose group, including the control group gave birth to 10-12 offspring. 14 One day after birth, each litter was culled to 10 offspring to accommodate similar lactational 15 TCDD exposure. On PND 28, the offspring were weaned, and three offspring from each TCDD 16 exposed group and the control group were selected for an immunocytochemical examination at 17 42 days of age. Following sacrifice of these offspring, the brain of each animal was removed 18 and every second serial section of the brain was processed for immunocytochemistry. In 19 addition to the serial sections of the brain, cells from 18 offspring (6 males per treatment group) 20 were used to assess the number of cells in the dorsal and median raphe nucleus, the 21 supralemniscal area, and the Nucleus raphe magnus. 22 Examination of external morphology, birth, and postnatal body weights indicated that 23 there were no differences between the male TCDD-exposed offspring and the control male 24 offspring. TCDD-exposed males, however, were aggressive toward other normal mice and were 25 also hypersensitive to soft touch. 26 Serotonin-immunoreactive neurons were found to be distributed throughout the entire 27 brainstem in 42-day-old males, and the general pattern in the TCDD-exposed animals was 28 consistent with those observed in control male offspring. Serotonergic neurons were identified 29 and counted in the caudal linear nucleus, the median and dorsal raphe nucleus, Nucleus raphe 30 pontis, interpeduncular nucleus, supralemniscal area, pedunculopontine segmental nuclei, deep 31 mensencephalic nucleus, Nucleus raphe magnus, pallidus, and obscurus, dorsal and medial to the This document is a draftfor review purposes only and does not constitute Agency policy. 2-159 DRAFT--DO NOT CITE OR QUOTE 1 facial nucleus and the ventrolateral medulla. Results from computerized cell counts (n = 6) 2 showed an average of 1,573.3 immunoreactive neurons in the raphe nuclei from the control 3 group versus 716.3 and 419.8 neurons in the low- and high-dose offspring, respectively. The 4 numbers of immunoreactive neurons in the individual raphe nuclei (dorsalis, medianus, magnus, 5 and B9) from the TCDD-exposed offspring were significantly (p < 0.01) lower than control 6 values, with the degree of reduction being dose-related. 7 In the absence of other relevant neurotoxicity endpoints, reduced serotonin is not an 8 adverse endpoint of toxicological significance in and of itself, thus, neither a NOAEL nor a 9 LOAEL can be established for this study. A lowest-observed-effect level (LOEL) of 10 0.7 ng/kg-day for an 8-week exposure duration is identified in this study for a significantly 11 (p < 0.01) lower number of serotonin-immunoreactive neurons in the raphe nuclei of male 12 offspring. A no-observed-effect level (NOEL) cannot be determined for this study. 13 14 2.4.2.2.8. L i et al. (2 0 0 6 ,199059). 15 Pregnant and pseudopregnant (obtained by mating normal estrous female mice with 16 vasectomized male mice) NIH mice (10 per treatment group) were exposed to 0, 2, 50, or 17 100 ng/kg-day of TCDD (purity 99%) during early gestation (GDs 1-8), preimplantation 18 (GDs 1-3), or peri-implantation to postimplantation (GDs 4-8) (Li et al., 2006). On GD 9, 19 animals were evaluated. The two highest TCDD doses (50 and 100 ng/kg-day) caused 20 significant (p < 0.05) early embryo loss independent of gestational exposure time. At 21 100 ng/kg-day, however, the embryo loss was greater when administered during GDs 1-8 or 22 GDs 1-3 compared to GDs 4-8 (p < 0.01). Uterine weight was significantly decreased in the 23 pseudopregnant mice when administered 50 or 100 ng/kg-day TCDD during GDs 1-8 24 (p < 0.001) or 1-3 (p < 0.01), but was only decreased at 100 ng/kg-day in pseudopregnant mice 25 when administered during GDs 4-8 (p < 0.01). Estradiol levels were increased at all TCDD 26 treatment levels (100% at the lowest dose), but statistical significance was not indicated. All 27 doses at all treatment times resulted in a significant reduction (p < 0.01) in serum progesterone 28 levels, with a 45% decrease at the lowest dose. Because the hormone effects were observed 29 following 4 days of treatment, the nominal doses were averaged over the entire test period of 30 8 days prior to measurement. The resulting average daily doses of TCDD were 0, 1, 25, and 31 50 ng/kg-day. This document is a draftfor review purposes only and does not constitute Agency policy. 2-160 DRAFT--DO NOT CITE OR QUOTE 1 A LOAEL of 2 ng/kg-day administered for 4 to 8 days is established in this study for a 2 significant (p < 0.01) decrease in progesterone (45% above control) and an approximate 2-fold 3 increase in estradiol levels (significance not indicated). A NOAEL cannot be determined. 4 5 2.4.2.2.9. M a rk o w sk i e t al. (2 0 0 1 ,197442). 6 Pregnant Holtzman rats (4-7 per treatment group) were administered a single gavage 7 dose of 0, 20, 60, or 180 ng/kg TCDD (purity not specified) in olive oil on GD 18 (Markowski 8 et al., 2001, 197442). One female rat from each liter (4-7 per treatment group) was assigned to 9 training on a wheel apparatus to respond on a lever for brief opportunities to run. Once animals 10 responded to an FR1 schedule of reinforcement, the requirement for lever pressing was increased 11 to FR2, FR5, FR10, FR20, and FR30 schedules. After each training session, the estrous cycle 12 stage was determined. Maternal body weight, length of gestation, number of pups per litter, and 13 sex distribution within litters were unaffected by treatment. For each of the FR schedules, there 14 was a significant dose-related (p = 0.0001) decrease in the number of earned run opportunities, 15 lever response rate, and total number of revolutions in the wheel in the adult female offspring. 16 There was no correlation between estrous cycle and responding for access to wheel running. 17 The developmental LOAEL for this study is a single dose of 20 ng/kg administered on 18 GD 18 for neurobehavioral effects. A NOAEL cannot be determined for this study. 19 20 2.4.2.2.10. M iettin en et al. (2 0 0 6 ,198266). 21 Miettinen et al. (2006, 198266) administered a single oral dose of 0, 30, 100, 300, or 22 1,000 ng/kg TCDD (purity >99%) in corn oil on GD 15 to pregnant Line C rats. The offspring 23 (24-32 per treatment group) were assigned to a sugar-rich cariogenic diet (via feed and drinking 24 water) and were orally inoculated three separate times with fresh cultures of Streptococcus 25 mutans. Three control groups varied with regard to TCDD exposure and administration of a 26 cariogenic diet. Two of the control groups received no TCDD, and the offspring were either 27 maintained on a normal diet without inoculation with S. mutans (C1; n = 48) or were given the 28 cariogenic diet with S. mutans inoculation (C2; n = 42). The final control group was maternally 29 exposed to 1,000 ng/kg TCDD with offspring fed a normal diet without S. mutans inoculation 30 (C3; n = 12). TCDD did not affect the maternal or offspring body weight. Survival of the 31 offspring was reduced in the 1,000 ng/kg dose group (50-58% survival compared to 83-95% in This document is a draftfor review purposes only and does not constitute Agency policy. 2-161 DRAFT--DO NOT CITE OR QUOTE 1 C1 and C2, respectively). All offspring administered 1,000 ng/kg were missing all lower 2 third molars. Two animals (8%) in the 100 ng/kg group were missing one of their lower 3 third molars. All doses, except the 100 ng/kg dose, caused a significant (p < 0.05) increase in the 4 number of caries lesions compared to group C2 (60, 79, 76, 83, and 91% in the C2, 30, 100, 300, 5 and 1,000 ng/kg groups, respectively). Group C3 (1,000 ng/kg TCDD exposure, normal diet) 6 animals also had increased caries lesions compared to C1 (8% versus 0%, respectively). There 7 were no changes in tooth mineral composition that could explain the increase in caries 8 susceptibility. 9 The developmental LOAEL from this study is a single dose of 30 ng/kg administered on 10 GD 15 based on the significant (p < 0.05) increase in dental caries in pups (30% above control). 11 A NOAEL cannot be determined from this study. 12 13 2.4.2.2.11. N oh ara et al. (2000, 200 0 2 7 ). 14 Pregnant Holtzman rats were administered 0, 12.5, 50, 200, or 800 ng/kg TCDD in corn 15 oil by gavage on GD 15 (Nohara et al., 2000, 200027). On PND 2, five males were randomly 16 selected from each litter and dose group. TCDD was detected in the thymus, spleen, and bone 17 marrow of the male pups on PND 21 and PND 49. TCDD was still detected in the thymus and 18 spleen on PND 120 but the levels decreased over time. The TCDD concentration was highest in 19 the thymus at all time points. There were no changes in the body, thymus, or spleen weights of 20 the male offspring on PND 5, PND 21, PND 49, or PND 120. On PND 5, there was a 200-fold 21 increase in CYP1A1 in the thymus of the high-dose male pups. CYP1A1 was only slightly 22 increased in the spleen. This induction decreased through PND 49. There was a slight (not 23 statistically significant) dose-dependent decrease in thymus cellularity in the male offspring at 24 PND 120. Spleen cellularity at PND 49 decreased in a dose-dependent manner (15-50% of the 25 control), with a statistically significant (p < 0.05) decrease observed in the high-dose group. A 26 slight but not significant reduction in spleen cellularity was noted in the high-dose group at 27 PND 21. The same effect was not observed at PND 120, nor was there any change in the percent 28 of B or T cells in the spleen. No changes in cytokine levels were observed in the 800-ng/kg 29 group. 30 Although a change in spleen cellularity on PND 49 (puberty) was observed, this effect 31 was transient and there were no coexisting changes in the percentage of splenic lymphocytes, This document is a draftfor review purposes only and does not constitute Agency policy. 2-162 DRAFT--DO NOT CITE OR QUOTE 1 spleen weight, and cytokine levels. Therefore, a developmental NOAEL of a single dose of 2 800 ng/kg administered on GD 15 is identified for this study. A LOAEL is not established. 3 4 2.4.2.2.12. O hsako et al. (2 0 0 1 ,198497). 5 Pregnant Holtzman rats (6 per treatment group) were administered 0, 12.5, 50, 200, or 6 800 ng/kg TCDD (purity >99.5%) in corn oil by gavage on GD 15 (Ohsako et al., 2001, 7 198497). On PND 2, five males were randomly selected from each litter. Two male offspring 8 from each litter were sacrificed on PND 49 and PND 120. Neither maternal nor male offspring 9 body weight was affected by TCDD treatment. TCDD was detected in both fat and testes at all 10 dose levels (including controls) with highest levels found in fat. There were no apparent 11 treatment-related effects on testicular weight, epididymal weight, daily sperm production, cauda 12 epididymal sperm reserves, luteinizing hormone, follicle stimulating hormone, or testosterone 13 levels. There was, however, a clear dose-dependent decrease in urogenital complex weight and 14 ventral prostate weight at both PND 49 and PND 120. For male offspring, statistically15 significant (p < 0.05) decreases were noted in urogenital complex weight at PND 120 in the 200 16 and 800 ng/kg groups, in ventral prostate weight at PND 49 in 800 ng/kg group, and at PND 120 17 in the 200 and 800 ng/kg groups. There was also a dose-dependent decrease in anogenital 18 distance (the length between the base of the genital tubercle and the anterior edge of the anus); 19 the decrease was not statistically significant at PND 49. At PND 120, however, male offspring 20 in all but the lowest dose group had significantly (p < 0.05) reduced anogenital distance 21 compared to the control animals. There was also a dose-dependent increase in 5aR-II mRNA 22 expression in the ventral prostate on PND 49 with significant increases (p < 0.05) in the 200 and 23 800 ng/kg animals. There was a significant (p < 0.01) decrease in the androgen receptor mRNA 24 in the ventral prostate on PND 49 at all doses tested. Similar effects were not observed on 25 PND 120 or in the caput epididymis on PND 49. 26 The developmental LOAEL for this study is a single dose of 50 ng/kg administered on 27 GD 15 for significantly (p < 0.01) reduced anogenital distance in male offspring (approximately 28 14%). The NOAEL for this study is 12.5 ng/kg. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-163 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.2.13. S ch an tz et al. (1 9 9 6 ,198781). 2 Schantz et al. (1996, 198781) studied the impact of in utero TCDD exposure on spatial 3 learning in male and female pups. Groups of pregnant Harlan Sprague-Dawley rats (n = 108, 4 divided into 4 cohorts; number of animals in each TCDD group approximately 4 per treatment 5 group) were dosed via gavage with 0, 25, or 100 ng/kg-day TCDD (purity >98%) in corn oil on 6 GDs 10-16. On the day of birth (post natal day [PND] 0), the pups were examined for gross 7 abnormalities and the number of live pups, weight, and sex were recorded for each litter. On 8 PND 2, litters were culled to eight animals and were balanced to include four males and 9 four females whenever possible. To minimize litter-size effects, litters with fewer than five pups 10 were excluded from the study. The exclusion of these litters resulted in 10-11 litters per 11 treatment group. Pups were weaned on PND 21 and one male and one female pup from each 12 litter were maintained for the learning tests. Pups were tested 5 days per week for spatial 13 learning and memory in a radial arm maze and a T-maze. A radial arm maze working memory 14 test and a T-maze DSA task were used a part of the testing process. 15 TCDD treatment did not affect dam gestational weight gain, dam liver weight, gestation 16 length, litter size, percentage of live births, birth weight, or postnatal growth of the pups 17 observed during the course of the study. Exposed pups, however, exhibited some signs of 18 toxicity in all exposure groups. Thymus weight was decreased and liver weight was increased in 19 the 100 ng/kg-day TCDD dose group. Also, liver microsomal 7-ethoxyresorufin-O-deethylase 20 (EROD) activity was markedly induced in pups from both the 25 and 100 ng/kg-day dose 21 groups. In the radial maze test, rats from all TCDD exposure groups displayed a significant 22 (p < 0.01) learning behavior as shown by progressively fewer errors from the first block of 23 sessions through the fourth session. The treatment by sex and treatment by session block 24 interactions were not significant. Comparisons between the average number of errors per session 25 block in the TCDD-exposed and control group indicated that both the 25 and the 100 ng/kg-day 26 dose groups made significantly (p < 0.05 andp < 0.001, respectively) fewer errors compared to 27 the control group. TCDD did not significantly affect adjacent arm selection behavior as 28 measured by C statistic; hence the reduction in errors observed did not appear to be accounted 29 for by an increased tendency to run into adjacent arms. Female pups had a significant (p < 0.05) 30 shorter radial arm maze latency, however, compared to the male pups. In the T-maze test, 31 TCDD did not significantly affect the percent of correct performance. All exposure groups This document is a draftfor review purposes only and does not constitute Agency policy. 2-164 DRAFT--DO NOT CITE OR QUOTE 1 performed best at the shortest delay, which showed a decline as the length of the intertrial delay 2 interval was increased. Additionally, all treated groups improved their performance over a 3 three-block session period. This finding indicated that animals in all groups could learn the task. 4 These observations were confirmed by a highly significant main effect of delay (p < 0.001) and 5 highly significant main effect of session blocks (p < 0.001). At the shortest 15-second delay, 6 average percent correct performance increased from 75 to 92%, while at the longest 40-second 7 delay, the average percent correct performance increased from 62 to 82%. A significant 8 (p < 0.05) main effect of exposure was evident in latency to respond in the T-maze. 9 Comparisons of the exposed group to control group, however, indicated that none of the 10 individual exposure groups differed significantly from the controls. Because no clear pattern 11 was observed in the various exposure groups, differences in latency to respond had no impact on 12 learning of the task. 13 Based on these results, the study authors state that the fact TCDD seems to have a 14 facilitatory effect on radial arm maze learning in rats should be interpreted with caution and 15 needs further evaluation using different and more varied learning tasks. No toxicologically 16 adverse endpoints were concurrently examined. Thus, a LOAEL and a NOAEL cannot be 17 determined for this study. 18 19 2.4.2.2.14. S eo e t al. (1 9 9 5 ,197869). 20 To study developmental effects of TCDD on thyroid hormone levels, time-mated female 21 Sprague-Dawley rat dams (n = 10-14/treatment group) were administered 25 or 100 ng/kg-day 22 of TCDD (>98% pure) in corn oil via gavage from GDs 10-16. Vehicle controls received 23 equivalent amounts of corn oil. The study also investigated PCB treatment outcomes. At birth, 24 pups were weighed and grossly examined for abnormalities. At 2 days of age, litters with fewer 25 than 5 pups were excluded from the analysis and the remaining litters were culled to 4 males and 26 4 females. Each treatment group contained 10 or 11 litters. Pups remained with the dams until 27 weaning. At weaning, 4-6 pups were retained for neurobehavioral tests (which were not 28 reported as part of this study). The remaining offspring were sacrificed, which provided 29 5-9 litters per treatment group. Data were collected from one male and one female where 30 possible. No signs of toxicity were evident in the dams; measurements on dams included 31 gestational weight gain, liver weight, litter size, and live births. Pup birth weight and weaning This document is a draftfor review purposes only and does not constitute Agency policy. 2-165 DRAFT--DO NOT CITE OR QUOTE 1 weight were unaffected by treatment. In pups sacrificed at weaning (21 days old), a significant 2 (p < 0.05) decrease occurred in thymus weight for the high-dose group, but not in thyroid, liver, 3 or brain weight. A significant (p < 0.05) decrease (20.4%) was observed in T4 in high-dose 4 females. Thyroid stimulating hormone and T3were unaffected by treatment. Uridine 5 diphosphate (UDP)-glucuronosyl transferase activity towards 4-nitrophenol significantly 6 (p < 0.05) increased in both treatment groups over control values, and the increase in the 7 high-dose group was significantly (p < 0.05) greater than in the low-dose group. Liver 8 microsomal EROD activity was significantly (p < 0.05) increased in both treatment groups, but 9 is considered to be an adaptive response and not adverse. 10 A LOAEL of 100 ng/kg-day for decreased thymus weights and decreased thyroxine is 11 identified for this study. A NOAEL of 25 ng/kg-day is established. 12 13 2.4.2.2.15. S im an ainen e t al. (2 0 0 4 ,198106). 14 Simanainen et al. (2004, 198106) studied the impact of in utero and lactational TCDD 15 exposure on the male reproductive system in three rat lines that are differentially sensitive to 16 TCDD. Groups of 5 to 8 pregnant Line A, B, and C C57BL/6N CYP1A2 dams were given a 17 single dose of 0, 30, 100, 300, or 1,000 ng/kg of TCDD (purity >99%) in corn oil on GD 15 via 18 oral gavage. Control animals were similarly dosed with a corn oil vehicle. One day after birth, 19 litters were randomly culled to include three males and three females to allow uniform postnatal 20 exposure. Offspring were weaned on PND 28. Dam and pup viabilities were monitored 21 throughout the study. Pup body weights were determined on PNDs 1, 4, 7, 14, and 28. 22 Anogenital distance and crown-rump length were measured on PNDs 1 and 4. On day 70, pups 23 were sacrificed and trunk blood was collected. Serum was collected for testosterone analysis. 24 The testes, cauda of the right epididymis, ventral prostrate, seminal vesicles, and thymus was 25 dissected and weighed. Absolute and relative organ weights were determined, and cauda 26 epididymis and testes were also preserved for sperm count analysis. 27 TCDD caused no mortality or overt signs of toxicity to the dams. Pup survival from 28 implantation to the day after birth also was not affected by TCDD exposure. Survival from the 29 day of implantation to the day after birth, however, was uncharacteristically lower in control 30 Line B rats (41%), resulting in a significant difference compared with the two lowest doses (30 31 and 100 ng/mg TCDD). The average survival percentage in the controls for Line A, B, and C This document is a draftfor review purposes only and does not constitute Agency policy. 2-166 DRAFT--DO NOT CITE OR QUOTE 1 rats was 85% (range 80-86%); 64% (41-86%); and 74% (63-85%); respectively. Percentage of 2 male pup survival in each line between PND 1 and PND 28 was 99% except for Line B males 3 exposed to 30 ng/kg TCDD and Line C males exposed to 30 or 100 ng/kg, where male survival 4 rate averaged 81% (range 81-83%). On PND 70, a significant (p < 0.05) reduction in body 5 weight was observed only in Line B and C rats at 1,000 ng/kg. In pups exposed to 1,000 ng/kg 6 TCDD, both absolute and relative weight of the ventral, anterior, and dorsolateral prostrate 7 decreased in all three lines at most postnatal time points measured. The change was most 8 consistent and significant (p < 0.05) in the ventral lobe. Animals exposed to 1,000 ng/kg TCDD 9 had an average decrease in absolute weight of the anterior prostrate of 37, 32, and 34% in 10 Lines A, B and C, respectively. Additionally, the average dorsolateral prostrate weight was also 11 decreased by 34, 28, and 39% in Lines A, B, and C, respectively. The effect on the ventral 12 prostrate was reversible with the only significant (p < 0.05) decrease in weight observed in 13 Line B rats at PND 70 in the 1,000 ng/kg TCDD dose group. The authors reported that TCDD 14 had no consistent effects on the weight of seminal vesicles. The absolute weights of the testis 15 and epididymis showed a significant (p < 0.05) increase on PNDs 28-49, but the relative testis, 16 epididymis, and cauda epididymis weights remained unchanged. In pups exposed to 17 1,000 ng/kg TCDD, severe malformation, including small caput and cauda and degeneration of 18 corpus epididymis, was observed. Malformations in the epididymis were observed in 6 of 19 44 Line C male rat offspring and 3 of 47 Line A male rat offspring. In Line A, B, and C rats at 20 PND 70in the 1,000 ng/kg TCDD dose group, daily sperm production was reduced by 9, 25, and 21 36% and cauda epididymal sperm reserves were reduced by 18, 42, and 49%, respectively. 22 Daily sperm reduction (17%) was significant (p < 0.05) in Line C rats at a TCDD dose of 23 300 ng/kg and in Line B and C rats at 1,000 ng/kg. A reduction in cauda epididymal sperm 24 reserves (25%) was significant (p < 0.05) in Line C rats at 300 and 1,000 ng/kg TCDD. 25 A LOAEL for TCDD of 300 ng/kg is identified for reduction in daily sperm production 26 and cauda epididymal sperm reserves in Line C rats. A NOAEL of 100 ng/kg is identified for 27 this study. 28 29 2.4.2.2.16. S u gita-K on ish i e t al. (2 0 0 3 ,198375). 30 Sugita-Konishi et al. (2003, 198375) examined the immunotoxic effects of lactational 31 exposure to TCDD in newborn mice. Eight pregnant female C57BL/6NCji mice were This document is a draftfor review purposes only and does not constitute Agency policy. 2-167 DRAFT--DO NOT CITE OR QUOTE 1 administered 0, 1.8, or 18 ng/L of TCDD via drinking water from parturition to weaning of the 2 offspring (for a total of 17 days). Based on an average water intake of 14-16 mL/day, the 3 average daily intake of TCDD for the dams was 1.14 and 11.3 ng/kg-day in the low- and 4 high-dose groups, respectively. In male offspring sacrificed at weaning (21 days after birth), 5 there was a statistically-significant (p < 0.05) decrease in relative spleen weight and a 6 statistically-significant (p < 0.005) increase in thymic CD4+ cells in the high-dose group. The 7 changes in relative spleen weight and thymic CD4+ cells were dose related, but effects in the 8 low-dose group did not achieve statistical significance. Changes in spleen weight and CD4+ cell 9 numbers were not observed in the female offspring. In a separate experiment, offspring infected 10 with Listeria monocytogenes following lactational TCDD exposure exhibited a statistically 11 significant increase in serum tumor necrosis factor alpha (TNF-a) 2 days after infection in both 12 sexes in the low- (p < 0.05) and high-dose (p < 0.005) groups. There was also a statistically 13 significant increase in serum interferon gamma in Listeria-infected high-dose females (p < 0.05). 14 The number of bacteria in the spleen was also significantly increased (p < 0.05) 2 days after 15 infection in the high-dose females compared to the controls, but not in males. Listeria levels in 16 the spleen returned to control levels by 4 days after infection in both sexes. 17 Based on these results, a LOAEL for TCDD of 11.3 ng/kg-day following a 17 day 18 exposure to dams was identified for significantly (p < 0.05) decreased spleen weight (in male 19 pups), a significant (p < 0.005) increase in thymic CD4+ cells (in male pups), and for increased 20 susceptibility to Listeria monocytogenes (in male and female pups). The NOAEL for this study 21 is 1.14 ng/kg-day. 22 23 2.4.2.3. A cu te S tu dies 24 2.4.2.3.I. B u rleson e t al. (1 9 9 6 ,196998). 25 Burleson et al. (1996, 196998) studied the impact of TCDD exposure on mice that were 26 challenged with the influenza virus 7 days after treatment with TCDD. Groups of 8-week-old 27 female B6C3F1 mice (n = 20, 2 replicate groups) were treated one time with 0, 1, 5, 10, 50, 100, 28 or 6,000 ng/kg TCDD (purity >99%, dissolved in corn oil) via oral gavage. In addition to the 29 treated groups, randomly selected animals were assigned as a sentinel group and screened for 30 numerous pathogens. Results of all tests performed on this sentinel group were negative. 31 Seven days after TCDD treatment, all animals were lightly anesthetized and infected intranasally This document is a draftfor review purposes only and does not constitute Agency policy. 2-168 DRAFT--DO NOT CITE OR QUOTE 1 with a highly lethal influenza A/Hong Kong/8/68 virus (H3N1; passage 14). The animals were 2 infected with sufficient H3N1 virus to achieve a 30% mortality rate in the control animals. 3 Animals were observed for mortality and morbidity for 21 days following viral infection. 4 Six mice from each treatment group were sacrificed on days 3, 9, and 12 postinfection, and body, 5 thymus, and wet lung weights were recorded. Influenza viral titers were examined by sacrificing 6 eight mice each at 2 hours and at 1, 4, 6, 7, 8, 9, 10, and 11 days post infection. 7 Exposure to TCDD resulted in significantly (p < 0.05) increased mortality in the 10, 50, 8 and 100 ng/kg dose groups. No statistically significant difference in the percentage alive was 9 observed between these dose groups. TCDD doses of 1 and 5 ng/kg did not alter mortality in 10 influenza infected animals. A time-related increase in the wet weights of the lungs in infected 11 mice as a result of increased edema also was reflected in an increase in the lung weight-to-body 12 weight ratio. The study authors stated that this ratio was not altered as a result of TCDD 13 exposure. TCDD-only exposures at 1, 10, or 100 ng/kg did not affect thymus weight. Similarly, 14 animals infected with the influenza virus following TCDD exposure also showed no loss in 15 thymic weight. Enhanced mortality in TCDD-treated animals was not correlated with an 16 increase in influenza virus titers. Additionally, animals treated with 1, 10, 100, or 1,000 ng/kg 17 did not affect pulmonary viral titer assays on days 6, 7, and 8 postinfection. The authors also 18 concluded that TCDD did not alter Hong Kong virus replication or clearance. 19 Although these results support immunotoxic effects induced by TCDD, the findings were 20 not reproduced by Nohara et al. (2002, 199021) using the identical study design, and the 21 translation of these findings to humans is dubious. Thus, no LOAEL/NOAEL was established. 22 A LOEL for TCDD of 10 ng/kg for a single exposure is identified for significantly (p < 0.05) 23 increased mortality in mice infected 7 days later with the influenza virus. The NOEL for this 24 study is 5 ng/kg. 25 26 2.4.2.3.2. C rofton et al. (2 0 0 5 ,197381). 27 Crofton et al. (2005, 197381) studied the impact of TCDD exposure in addition to the 28 impact of mixtures of thyroid disrupting chemicals and PCBs on serum total thyroxine (TT4) 29 concentration. Groups of female Long-Evans rats were dosed via oral gavage with 0, 0.1, 3, 10, 30 30, 100, 300, 1,000, 3,000, or 10,000 ng/kg-day TCDD (purity >99%) in corn oil (n = 14, 6, 12, 31 6, 6, 6, 6, 6, 6, and 4, respectively) for 4 consecutive days. On the day following the last dose, This document is a draftfor review purposes only and does not constitute Agency policy. 2-169 DRAFT--DO NOT CITE OR QUOTE 1 animals were sacrificed, trunk blood was collected, and serum obtained via centrifugation was 2 assayed for TT4 concentration using standard radioimmunoassay methods. 3 No visible signs of toxicity or changes in animal body weight as a result of TCDD 4 exposure were observed. Serum T4 levels showed a dose-dependent decrease, with the levels 5 dropping sharply beginning at 100 ng/kg-day dose. Percent serum T4 levels were 96.3, 98.6, 6 99.8, 93.3, 70.9, 62.5, 52.7, 54.7, and 49.1% in the 0.1, 3, 10, 30, 100, 300, 1,000, 3,000, and 7 10,000 ng /kg-day groups, respectively. 8 A LOAEL for TCDD of 100 ng/kg-day for 4 consecutive days of exposure is identified in 9 this study for a reduction in serum T4 levels (70.9% compared to 100% in controls). The 10 NOAEL for this study is 30 ng/kg-day. 11 12 2.4.2.3.3. K itch in a n d W oods (1 9 7 9 ,198750). 13 Female Sprague-Dawley rats (nine per control and four per treatment group) were 14 administered a single dose of 0, 0.6, 2, 4, 20, 60, 200, 600, 2,000, 5,000, or 20,000 ng/kg TCDD 15 (purity >99%) in corn oil. Animals were sacrificed 3 days after treatment and CYP level and 16 benzo(a)pyrene hydroxylase activity in the liver were measured. A significant (p < 0.05) 17 increase in cytochrome P450 levels occurred with doses of 600 ng/kg or greater and in 18 benzo(a)pyrene hydroxylase activity with doses of 2 ng/kg or greater. Cytochrome P450 was 19 significantly (p < 0.05) higher 1 month after a single exposure of 2,000 ng/kg (the only dose 20 measured), but not after 3 or 6 months. Aryl hydrocarbon hydralase (AHH; p < 0.05) and EROD 21 (p < 0.01) were both significantly increased through 3 months after treatment, and although 22 elevated at 6 months, the results were not significant. 23 CYP induction alone is not considered a significant toxicologically adverse effect given 24 that CYPs are induced as a means of hepatic processing of xenobiotic agents. Thus, no LOAEL 25 or NOAEL was established for this study because adverse endpoints (e.g., indicators of 26 hepatotoxicity) were not measured. The acute LOEL, however, is 2 ng/kg based on a significant 27 (p < 0.05) increase in benzo(a)pyrene hydroxylase activity (37% above control). The NOEL is 28 0.6 ng/kg. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-170 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.3.4. L i et al. (1 9 9 7 ,199060). 2 Female Sprague-Dawley rats (22 days old; 10 per treatment) were administered a single 3 oral dose of TCDD (>98% pure) in corn oil via gavage at doses of 3, 10, 30, 100, 300, 1,000, 4 3,000, 10,000, or 30,000 ng/kg. Vehicle controls received equivalent amounts of corn oil, while 5 nave controls were sham-treated only. In a preliminary time-course study, animals received a 6 single dose of 10,000 ng/kg and were sacrificed at 1, 2, 4, 8, 16, 24, 48, and 72 hours. The 7 time-course study showed two peaks in LH and FSH levels at 1 hour and 24 hours, with a 8 decrease to control values by 48 hours. Thus, in the dose-response study, animals were 9 sacrificed at 1 or 24 hours after treatment, blood was collected, and serum FSH and LH were 10 measured. The dose-response study demonstrated that the peak at 1 hour was related to the 11 vehicle as the peak also occurred in the vehicle controls, but did not occur in the nave controls. 12 At 24 hours, FSH was increased at 10 ng/kg and higher (>4-fold increase at 10 ng/kg). Doses of 13 10 to 1,000 ng/kg showed similar increases (not all reached statistical significance; p < 0.05). A 14 dose-dependent increase occurred for doses >3000 (p < 0.05) with a maximum increase of 15 20-fold over the vehicle control. At 24 hours, the LH response significantly (p < 0.05) increased 16 only for doses >300 ng/kg with a maximum increase of 15-fold above the vehicle control. The 17 study authors calculated an ED50 of 500 ng/kg for gonadotropin increase. The dose-dependent 18 release of LH was confirmed in in vitro studies, but did not occur with the same magnitude. The 19 increase did not occur in calcium-free medium and was unrelated to gonadotropin releasing 20 hormone. 21 Based on the increase in serum FSH, the LOAEL was 10 ng/kg and the NOAEL was 22 3 ng/kg. 23 24 2.4.2.3.5. L u cier et al. (1 9 8 6 ,198398) . 25 Adult female Sprague-Dawley rats (six per treatment) were administered a single gavage 26 dose of TCDD (purity not specified) in either corn oil or contaminated soil at doses of 15, 40, 27 100, 200, 500, 1,000, 2,000, 5,000 (corn oil), or 5,500 (contaminated soil) ng/kg. Animals were 28 sacrificed 6 days later and livers were removed for analysis. No clinical signs of acute toxicity 29 or changes in body weight were observed at any dose. AHH increased in a dose-dependent 30 manner with significant (p < 0.05) increases observed at 15 ng/kg or greater in corn oil or 31 40 ng/kg or greater in contaminated soil. Cytochrome P450 was significantly (p < 0.05) This document is a draftfor review purposes only and does not constitute Agency policy. 2-171 DRAFT--DO NOT CITE OR QUOTE 1 increased with doses of 1,000 ng/kg or greater in corn oil or 500 ng/kg or greater in contaminated 2 soil. A dose-dependent increase was observed for UDP glucoronyltransferase (significance of 3 individual doses not reported), with the results twice as high with corn oil than with 4 contaminated soil. The authors state that the results indicate bioavailability from soils is 50%. 5 Because the association between AHH activity and TCDD-mediated hepatotoxicity is 6 unknown and no adverse endpoints were measured, a LOAEL or NOAEL was not determined 7 for this study. The acute LOEL for this study is 15 ng/kg, based on the significant (p < 0.05) 8 increase (80% above control) in AHH. No NOEL is established. 9 10 2.4.2.3.6. N oh ara et al. (2 0 0 2 ,199021). 11 Male and female B6C3F1 (C57BL/6 x C3H), BALB/c, C57BL/6N, and DBA2 mice 12 (10-40 per treatment group) were administered a single dose of 0, 5, 20, 100, or 500 ng/kg 13 TCDD in corn oil via gavage. Seven days following TCDD treatment, mice were infected with a 14 mouse-adapted strain of influenza (A/PR/34/8; H1N1) at a plaque forming unit dose designed to 15 target approximately 30% mortality in each strain. TCDD did not affect the body weight or 16 survival in any of the infected mouse strains at any dose. 17 Therefore, no LOAEL is established in this study. The NOAEL is 500 ng/kg. 18 19 2.4.2.3.7. S im an ainen e t al. (2 0 0 3 ,198582). 20 Simanainen et al. (2003, 198582) studied the short-term effects of TCDD exposure to 21 determine the efficacy and potency relationships among three differentially susceptible rat lines. 22 The three rat lines used were A, B, and C, which were selectively bred from TCDD-resistant 23 Han/Wistar and TCDD-sensitive Long-Evans rats. The study authors reported that Line A rats 24 were most resistant to TCDD acute lethality followed by Line B and C. Groups of five or 25 six randomly selected rats (sex not specified) were treated with a single oral dose of TCDD 26 (purity >99%) in corn oil by oral gavage. The dose of TCDD was reported to range between 27 30 ng/kg and 3,000 pg/kg for Line A, 30 ng/kg and 1,000 pg/kg in Line B, and 30 ng/kg and 28 100 pg/kg for Line C. Control animals were similarly dosed with a corn oil vehicle. Rats were 29 sacrificed on day 8 postexposure, and trunk blood was collected and serum separated. Liver and 30 thymus were removed and weighed, and liver samples were collected and preserved. Liver This document is a draftfor review purposes only and does not constitute Agency policy. 2-172 DRAFT--DO NOT CITE OR QUOTE 1 EROD activity, serum aspartate aminotransferase (ASAT) activity, free fatty acid (FFA) 2 concentration, and total bilirubin concentration were determined. Teeth were also examined. 3 Relative thymus weights were reduced 25% at 300 ng/kg relative to controls in Line B 4 rats. Liver enzyme (CYP1A1) induction, as measured by EROD activity, was evident at all 5 exposure levels; CYP induction is considered to be an adaptive effect and not adverse in itself. 6 No other endpoints were affected below 1 pg/kg in any of the three rat lines. 7 A LOAEL for TCDD of 300 ng/kg is identified for decreased relative thymus weight in 8 Line B rats. A NOAEL of 100 ng/kg is identified for this study. 9 10 2.4.2.3.8. S im an ainen e t al. (2002, 201369). 11 To study the short-term effects of TCDD on hormone levels, adult female Long-Evans 12 (TCDD-sensitive) and Han/Wistar (TCDD-resistant) rats (n = 9-11/treatment) were administered 13 a single dose of TCDD (>99% pure) in corn oil via gavage at doses ranging from 30 ng/kg to 14 100 pg/kg. Vehicle controls received an equivalent amount of corn oil. The study also 15 examined other polychlorinated dibenzo-p-dioxins outcomes. Rats were sacrificed on day 8 16 postexposure, and trunk blood was collected and serum separated. Liver and thymus were 17 removed and weighed, and liver samples were collected and preserved. Liver EROD activity, 18 serum ASAT activity, FFA concentration, and total bilirubin concentration were determined. 19 Teeth were also examined. 20 Neither FFA or ASAT levels in Han/Wistar rats showed a dose-response relationship. In 21 Long-Evans rats, however, a significant (p < 0.05) dose-dependent increase in FFA occurred at 22 300 ng/kg TCDD. Serum ASAT sharply increased in Long-Evans rats between 3,000 and 23 10,000 ng/kg. Body weight change and relative thymus weights were significantly decreased 24 (p < 0.05) in Han/Wistar rats with doses >10,000 ng/kg and in Long-Evans rats with doses 25 >1,000 ng/kg. Liver EROD activity was significantly (p < 0.05) increased with all doses in both 26 strains. Serum T4 was significantly (p < 0.05) decreased in Long-Evans rats at concentrations 27 >300 ng/kg, but were not significantly affected in Han/Wistar rats. Serum bilirubin was 28 significantly (p < 0.05) increased with doses >10,000 ng/kg in Long-Evans rats and 29 >30,000 ng/kg in Hans/Wistar rats. Both strains of rat showed a dose-dependent increase in 30 mean severity of incisor tooth defects. The results indicate that TCDD was the most potent 31 congener tested in both rat strains. This document is a draftfor review purposes only and does not constitute Agency policy. 2-173 DRAFT--DO NOT CITE OR QUOTE 1 A LOAEL of 300 ng/kg for decreased T4 in the Long-Evans rat is identified for this 2 study. A NOAEL of 100 ng/kg is established. 3 4 2.4.2.3.9. S m ialow icz et al. (2 0 0 4 ,110937). 5 Smialowicz et al. (2004, 110937) examined the impact of TCDD exposure on 6 immunosuppression in mice. Groups of female (number not specified) C57BL/6N CYP1A2 7 (+/+) wild-type mice were administered a single dose of 0, 30, 100, 300, 1,000, 3,000, or 8 10.000 ng/kg TCDD (purity >99%) in corn oil via oral gavage. Control animals were similarly 9 dosed with a corn oil vehicle. To assess immune function, 7 days after TCDD administration, all 10 mice were immunized with sheep red blood cells (SRBCs) via injection into the lateral tail vein. 11 Five days after immunization, mice were sacrificed, blood was collected, and enzyme-linked 12 immunosorbant assays were performed. Additionally, spleen, thymus, and liver weights also 13 were measured. 14 Body and spleen weights of the wild-type mice were unaffected by the TCDD exposure. 15 A decrease in thymus weights of the mice appeared to be dose related. Only mice treated with 16 10.000 ng/kg TCDD, however, showed a statistically significant (p < 0.05) decrease in thymus 17 weights compared to corresponding controls. Liver weights also showed a dose-related increase 18 with only animals treated with 3,000 and 10,000 ng/kg TCDD showing statistical significance 19 (p < 0.05) compared to the control group. The antibody response to SRBCs indicated a 20 dose-related suppression in the wild-type mice, with animals treated with 1,000, 3,000, and 21 10.000 ng/kg TCDD showing statistically significant (p < 0.05) suppression compared to the 22 controls. 23 A LOAEL for TCDD of 1,000 ng/kg is identified in female C57BL/6N CYP1A2 (+/+) 24 wild-type mice for significant (p < 0.05) suppression of SRBCs. The NOAEL for this study is 25 300 ng/kg. 26 27 2.4.2.3.10. Vanden H e u v e l e t al. (1 9 9 4 ,197551). 28 Vanden Heuvel et al. (1994, 197551) examined the dose-response relationship between 29 TCDD exposure and induction of hepatic mRNA. Groups of 10-week-old female 30 Sprague-Dawley rats were administered TCDD (purity ~99%) in corn oil once at 0, 0.1, 0.05, 1, 31 10, 100, 1,000, or 10,000 ng/kg-BW. Four days after TCDD treatment, animals were sacrificed This document is a draftfor review purposes only and does not constitute Agency policy. 2-174 DRAFT--DO NOT CITE OR QUOTE 1 and livers were excised and preserved. Total hepatic RNA was extracted using guanidine 2 thiocyanate and DNA was removed using standard phenol-chloroform-isoamyl alcohol 3 partitioning procedures. Quantitative competitive RNA-PCR method was used to analyze 4 CYP1A1, UDP-glucuronosyltransferase I (UGT1), plasminogen activator inhibitor 2 (PAI2), 5 P-actin, and transforming growth factor a (TGFa). In addition to hepatic mRNA levels, 6 microsomal protein was assayed for EROD activity and livers were tested for TCDD 7 concentration. 8 CYP1A1 mRNA induction levels in the TCDD-treated groups were low in the low-dose 9 region and sharply increased to plateaus at higher doses. The lowest dose that showed a 10 statistically significant (p < 0.05) difference compared to controls was the 1 ng/kg dose, which 11 showed a three-fold increase in CYP1A1 mRNA levels. In contrast, a 130-fold increase 12 occurred at 100 ng/kg and a 4,000- and 7,000-fold increase occurred at 1,000 and 10,000 ng/kg, 13 respectively. A slight increase in the CYP1A1/p-actin levels was observed in the 0.1 ng/kg 14 group, but this increase was not significant. EROD activity exhibited a pattern similar to 15 CYP1A1 activity. EROD activity, however, was approximately 100-fold less sensitive 16 compared to mRNA levels in TCDD-treated groups. Statistical significance (p-value not 17 provided) in CYP1A1 level was observed at the 100 ng/kg dose compared to the 1 ng/kg dose. 18 The study authors reported that, despite this difference in CYP1A1 and EROD activity, the 19 correlation between CYP1A1 enzyme activity and mRNA levels was good. Dose-response 20 relationships for the induction of UGT1, PAI2, and TGFa mRNA differed from what had been 21 observed for CYP1A1 mRNA. UGT1 mRNA was induced, but at the much higher dose of 22 1,000 ng/kg. Additionally, the five-fold maximum induction of UGT1 mRNA was much less 23 than the 7,000-fold induction observed for CYP1A1 mRNA at the 10,000 ng /kg dose. The 24 authors state that this could be a result of the constitutive level of UGT1, which is much higher 25 than CYP1A1, which makes detecting induction of UGT1 in the low dose regions more difficult. 26 PAI2 and TGFa mRNA were not affected by TCDD in rat liver in the dose range tested. These 27 results indicate that dioxin-inducible genes have a quite dissimilar dose-response relationship. 28 Induction of CYP1A1 expression is not considered an adverse effect, as the role of 29 CYP1A1 in TCDD-mediated hepatotoxicity is unsettled. Therefore, in the absence of other 30 indicators of hepatotoxicity, a NOAEL/LOAEL cannot be determined for this study. A LOEL This document is a draftfor review purposes only and does not constitute Agency policy. 2-175 DRAFT--DO NOT CITE OR QUOTE 1 for TCDD of 1 ng/kg for a single exposure was identified for statistically significant (p < 0.05) 2 increase in CYP1A1 mRNA levels. The NOEL for this study is 0.1 ng/kg. 3 4 2.4.2.4. S u bch ron ic Stu dies 5 2.4.2.4.1. Chu e t al. (2001, 521829). 6 Adult female Sprague-Dawley rats (five per treatment group) were administered TCDD 7 (purity >99%) in corn oil by gavage at doses of 0, 2.5, 25, 250, or 1,000 ng/kg-day for 28 days 8 (Chu et al., 2001, 521829). The 1,000 ng/kg-day dose of TCDD caused a significant (p < 0.05) 9 decrease in body weight gain (36% lower than the control), increase in relative liver weight (40% 10 greater than the control), and decrease in relative thymus weight (50% lower than the control). 11 There was a significant (p < 0.05) increase in EROD activity, methoxy resoufin-O-deethylase 12 (MROD) activity, and UDP-glucuronosyl transferase (UDPGT) activity in the liver of female 13 rats receiving 250 or 1,000 ng/kg-day TCDD. In addition, significant (p < 0.05) increases in 14 serum cholesterol were observed in the 250 and 1,000 ng/kg-day dose groups, and liver ascorbic 15 acid (AA) also was significantly increased in the 1,000 ng/kg-day dose group. There was 16 ~1.5-fold increase in liver glutathione-^-transferase (GST), which was not statistically 17 significant. Other significant (p < 0.05) findings for the 1,000 ng/kg-day group included a 18 decrease in liver vitamin A (51% lower than the control), an increase in kidney vitamin A 19 (15.5-fold increase above the control), an increase in liver benzyloxy resoufin-O-deethylase 20 (BROD, 30-fold increase above control), a decrease in liver pentoxyresoufin-O-deethylase 21 (PROD, 37% lower than the control), increase in serum albumin (18% above the control), and a 22 decrease in mean corpuscular hemoglobin (MCH, 7% below the control) and mean corpuscular 23 volume (MCV, 7% below the control). 24 Based on the numerous significant (p < 0.05) liver-related biochemical changes and 25 significant (p < 0.05) increased relative liver weight, as well as significantly decreased body 26 weight and relative thymus weight, the LOAEL for 28 days of exposure in this study is 27 1,000 ng/kg-day and the NOAEL is 250 ng/kg-day. 28 29 2.4.2.4.2. Chu e t al., 2007. 30 Chu et al. (2007) examined the potential impact of TCDD on various organs and the 31 toxicological impacts as a result of interactions between TCDD and PCBs in rats. Groups of This document is a draftfor review purposes only and does not constitute Agency policy. 2-176 DRAFT--DO NOT CITE OR QUOTE 1 female Sprague-Dawley rats (n = 5 per treatment group) were treated daily for 28 days via 2 gavage with 0, 2.5, 25, 250, or 1,000 ng /kg-day TCDD (purity not specified) dissolved in corn 3 oil. Body weights were determined three times per week, and clinical observations were made 4 daily. At study termination, all animals were sacrificed and blood was analyzed for various 5 biochemical and hematological parameters. Liver, spleen, heart, thymus, brain, and kidneys 6 were removed and weighed. A small portion of the liver was homogenized and assayed for 7 BROD; EROD; MROD; and PROD. UDPGT, GST, and ascorbic acid levels also were 8 measured. Vitamin A levels in the liver, kidney, and lungs were analyzed as free retinol 9 (vitamin A), and histopathological analysis was conducted on various tissues. 10 Growth rate and thymic weights in rats treated with 1,000 ng/kg-day TCDD were 11 significantly (p < 0.05) inhibited compared to the control group. Enzyme analysis indicated that 12 measured levels of TCDD in the liver correlated with hepatic microsomal enzyme activity. The 13 authors reported that liver microsomal EROD and MROD activities were significantly (p < 0.05 14 for EROD activity, significance level for MROD not reported) increased in the 250 and 15 1.000 ng/kg-day TCDD dose groups compared to the control group. UDPGT levels were 16 significantly (significance level not reported) increased in the 250 and 1,000 ng/kg-day TCDD 17 dose groups compared to the controls. Serum albumin levels were significantly (p < 0.05) 18 increased in the 1,000 ng/kg-day TCDD dose group compared to the control group. Serum 19 cholesterol levels were significantly (level not reported) increased compared to the control group 20 at 250 ng/kg-day TCDD dose, while liver ascorbic acid concentrations were significantly (level 21 not reported) increased in the 1,000 ng/kg-day dose group. Hematological analysis indicated that 22 hemoglobin, packed cell volume, MCH, MCV, and platelet values were decreased in the 23 1.000 ng/kg-day TCDD dose group. Significant (p < 0.05) differences were observed only in 24 MCH and MCV levels compared to the control. Vitamin A levels in the liver and kidney were 25 significantly (p < 0.05) lower in the 1,000 ng/kg-day TCDD group compared to the control 26 group. Histopathological evaluation of various tissues indicated that liver, thyroid, and thymus 27 were the target organs. No TCDD-related affects were found in other tissues. A dose-dependent 28 alteration in the thymus consisted of reduced thymic cortex and increased medullar volume with 29 more animals exhibiting these changes at the 250 and 1,000 ng/kg-day dose level compared to 30 the control group. Alterations in thyroid included reduced follicles, reduced colloid density, and 31 increased epithelial height. A dose-dependent change in the thyroid was observed, with the This document is a draftfor review purposes only and does not constitute Agency policy. 2-177 DRAFT--DO NOT CITE OR QUOTE 1 highest impact evident in reduced follicles and reduced colloid density beginning at a dose of 2 25 ng/kg-day TCDD. Changes in liver were characterized by accentuated hepatic zones, 3 anisokaryosis of hepatocytes, increased cytoplasmic density, and vacuolation. These changes 4 were also dose dependent, with more animals exhibiting these histopathological changes with 5 increasing TCDD dose. Based on these results, the study authors concluded that exposure to 6 TCDD resulted in a wide range of adverse effects with the thyroid proving to be most sensitive. 7 A LOAEL for TCDD of 25 ng/kg for a 28-day exposure is identified for alterations in 8 thyroid, thymus, and liver histopathology. The NOAEL for this study is 2.5 ng/kg-day. 9 10 2.4.2.4.3. D eC aprio e t al. (1 9 8 6 ,197403). 11 Hartley guinea pigs (10 per sex per dose) were administered TCDD (purity not specified) 12 in the diet for 90 days at concentrations of 0, 2, 10, 76, or 430 ppt (equivalent to 0, 0.12, 0.61, 13 4.9, and 26 ng/kg-day in males and 0, 0.12, 0.68, 4.86, and 31 ng/kg-day in females calculated by 14 the study authors using food consumption and body weights). Other animals were administered 15 the high-dose diet (i.e., 430 ppt) for 11, 21, or 35 days and then administered the control diet 16 (i.e., no exposure) for the remainder of the 90 days for recovery analysis. Four high-dose males 17 died and two were sacrificed moribund by day 45; the remaining four animals were sacrificed on 18 day 46 for necropsy. Four high-dose females also died and two were sacrificed moribund by day 19 55 with the remaining females sacrificed on day 60 for necropsy. Animals in the 76- and 20 430-ppt groups had significantly (p < 0.05) reduced body weights. Organ weights were not 21 obtained in the 430-ppt group due to the early sacrifice, but in the 76-ppt group a significant 22 decrease in relative thymus weight (p < 0.05) was observed, and relative liver (p < 0.01) and 23 brain (p < 0.05) weights in males increased. Although a similar trend occurred in the females, 24 the results were not statistically significant. Males administered 76 ppt in the diet also had a 25 53% increase in triglycerides (p < 0.05). The same increase was observed in females, but was 26 not statistically significant. In the recovery groups, mortality during the recovery period after 11 27 or 21 days of treatment was 10% and after 35 days of treatment was 70%. Animals lost weight 28 during the treatment period. Although the body weight increased during the recovery period, the 29 body weight remained low compared to the control for the study duration. This document is a draftfor review purposes only and does not constitute Agency policy. 2-178 DRAFT--DO NOT CITE OR QUOTE 1 The LOAEL from this study is 4.9 ng/kg-day for 90 days of exposure, based on 2 decreased body weight (12-15 %; p < 0.05) and changes in organ weights (10-30%, significant 3 only in the males). The NOAEL is 0.61 ng/kg-day. 4 5 2.4.2.4.4. D evito et al. (1 9 9 4 ,197278). 6 Female B6C3F1 mice (5 per treatment) were administered 0, 1.5, 4.5, 15, 45, or 7 150 ng/kg TCDD (98% pure) in corn oil via gavage, 5 days a week for 13 weeks. This dose is 8 equivalent to 0, 1.07, 3.21, 10.7, 32.1, 107 ng/kg-day (adjusted for continuous exposure, 9 administered dose multiplied by 5 and divided by 7). Body weight was recorded weekly and 10 animals were sacrificed 3 days after the last treatment. Examinations were performed on the 11 lung, skin, uterus, and liver. No differences were observed in the liver or uterus weights or in the 12 estrogen receptor levels in these two tissues. A dose-dependent increase in EROD activity (an 13 indicator of CYP1A1 [CYP] induction) in the lung, skin, and liver was observed, with significant 14 (p < 0.05) increases even at the lowest dose. The TCDD doses used did not achieve maximal 15 EROD induction. A significant (p < 0.05) increase in liver acetanilide-4-hydroxylase (ACOH; 16 an indicator of CYP1A2 induction) also was observed with all doses. A maximum induction of 17 ACOH occurred with doses of 3.21 ng/kg-day and greater. A dose-dependent increase in 18 specific phosphotyrosyl protein (pp) levels also was observed. Levels of pp34 and pp38 were 19 significantly (p < 0.05) increased even at the lowest dose, while pp32 reached statistical 20 significance (p < 0.05) with doses of 4.5 ng/kg-day and above. 21 The role of CYPs and phosphorylated pp32, pp34, and pp38 in TCDD-mediated toxicity 22 is unknown, and changes in the activity or function of these proteins are not considered adverse 23 Therefore, no LOAEL or NOAEL is established. The 13-week LOEL is 1.07 ng/kg-day, based 24 on a significant (p < 0.05) increase in EROD, ACOH, pp34, and pp38 levels (all increased by at 25 least 2-fold). No NOEL is established for this study. 26 27 2.4.2.4.5. F attore e t al. (2 0 0 0 ,197446). 28 Fattore et al. (2000, 197446) examined TCDD-induced reduction of hepatic vitamin A 29 levels in a subchronic rat bioassay on Sprague-Dawley rats. Four experiments were conducted; 30 Experiments 1, 2, and 3 were conducted in both male and female rats, while Experiment 4 was 31 conducted only in female rats. The dosing regimens for each experiment were as follows This document is a draftfor review purposes only and does not constitute Agency policy. 2-179 DRAFT--DO NOT CITE OR QUOTE 1 E xperim en t 1 : Groups of six Iva:SIV 50 rats (male and female) were maintained on a diet 2 consisting of 0, 200, 2,000, or 20,000 ng TCDD/kg diet and 3-pg vitamin A/kg diet for 3 13 weeks. Assuming food consumption of 10% of body weight per day, the average daily 4 doses are 0, 20, 200, and 2,000 ng/kg-day TCDD. 5 E xperim en t 2 : Groups of six male and female rats were treated with 0 or 6 200 ng TCDD/kg-day and 3 pg vitamin A/kg diet for 13 weeks. 7 E xperim en t 3 : Groups of six male and female rats were fed 0, 200, or 8 1,000 ng TCDD/kg-day and 3 pg vitamin A/kg diet for 13 weeks. 9 E xperim en t 4 : Groups of female rats (number not specified; IVA;SIV 50 Sprague-Dawley 10 strain) were treated with TCDD for 26 and 39 weeks in addition to a 13-week dietary 11 treatment with 0 or 100 ng TCDD/kg-day and 3 pg vitamin A/kg diet for 13 weeks. 12 13 For a 13-week exposure duration employed in all four experiments, male and female rats 14 were treated at 0, 20, 100 (females only), 200, 1,000, or 2,000 ng/kg-day. In all 15 four experiments, liver from control and treated animals was analyzed at termination for free 16 retinol content to determine hepatic vitamin A levels. 17 18 R esults: 19 E xperim en t 1 : Liver and body weights in both treated males and females were significantly 20 affected at all but the lowest dose tested (20 ng/kg-day). Liver injury was severe, particularly 21 in female rats treated with 2,000 ng TCDD/kg-day. Dietary intake of vitamin A in male rats 22 was comparable to intake in controls, except in the 2,000 ng/kg-day group, which showed a 23 reduction of 16% in the dietary intake of vitamin A compared to controls. There was no 24 effect of TCDD on vitamin A intake in female rats. Hepatic vitamin A levels showed a 25 dose-dependent reduction with levels dropping sharply in the 200 and 2,000 ng/kg-day dose 26 groups, particularly in treated females. The reduction was significant at 200 ng/kg-day 27 (p < 0.05) and 2,000 ng/kg-day (p < 0.01) in males, and at 200 ng/kg-day (p < 0.5) and 28 2,000 ng/kg-day (p < 0.001) in females. The reductions ranged from 68-99% in males and 29 72-99% in females when compared to corresponding controls. 30 E xperim en t 2 : Changes in liver and body weights were not reported. Hepatic vitamin A 31 level in males and females were reduced by 70% and 99%, respectively, compared to 32 controls, in rats receiving 20 ng/kg-day (significance level in females: p < 0.01). 33 E xperim en t 3 : Similar to the results of Experiments 1 and 2, a dose-related trend of 34 significantly (p < 0.001) reduced hepatic vitamin A level was observed in both males and 35 females, with males exhibiting a particularly sharp drop at the 1,000 ng/kg-day dose 36 compared to controls. 37 E xperim en t 4 : Females treated with 100 ng/kg-day showed significant reductions in hepatic 38 vitamin A levels (p < 0.05-0.001) at all three treatment durations (13, 26, and 39 weeks). 39 This document is a draftfor review purposes only and does not constitute Agency policy. 2-180 DRAFT--DO NOT CITE OR QUOTE 1 A LOAEL for TCDD of 20 ng/kg-day for a 13-week subchronic exposure was identified 2 in this study for decreased hepatic vitamin A levels (27 and 24 % lower than the corresponding 3 control in female and male rats, respectively). This LOAEL is determined using data from 4 Experiment 1. A NOAEL was not identified in this study. 5 2.4.2.4.6. F ox et al. (1 9 9 3 ,197344). 6 Sprague-Dawley rats (6 per sex per dose) were gavaged with TCDD (purity not 7 specified) in corn oil using a dose-loading regime to achieve and maintain steady-state levels of 8 0.03, 30, or 150 ng/g in the liver. The regime consisted of an initial loading dose of 5, 2,500, or 9 12,000 ng/kg followed every 4 days with a maintenance dose of 0.9, 600, or 3,500 ng/kg. 10 Averaging the doses over the 14 days provides average daily doses of 0.55, 307, and 11 1,607 ng/kg-day (e.g., 5 ng/kg-day on day 1 and 0.9 ng/kg-day on days 5, 9, and 13 is 5 + 0.9 12 + 0.9 + 0.9/14 = 0.55 ng/kg-day). Body weight, liver weight, and liver gene expression were 13 measured at 7 and 14 days. A significant (p < 0.05) decrease in body weight occurred in 14 high-dose males (at 14 weeks only) and females (at 7 and 14 days). A significant (p < 0.05) 15 increase in absolute and relative liver weights was observed in mid- and high-dose males and 16 females at both 7 and 14 days. Although the liver of treated animals indicated moderate 17 vacuolization and swelling, there was no indication of necrosis. An increase in gene expression 18 (clone 1, CYP1A1, CYP1A2, and albumin) was observed in the mid- and high-dose groups. A 19 significant (p < 0.05) decrease in labeling index (indication of cell proliferation) occurred in both 20 females (all doses) and males (high-dose only) during week 1, but not during week 2. 21 The 14-day LOAEL is 307 ng/kg-day for significant (p < 0.05) increases in absolute and 22 relative liver weights (25--34%). The NOAEL is 0.55 ng/kg-day. 23 2.4.2.4.7. H assou n et al. (1 9 9 8 ,136626). 24 Female B6C3F1 mice (number not specified) received TCDD (>98% pure) in corn oil 25 5 days per week for 13 weeks via gavage at doses of 0, 0.45, 1.5, 15, or 150 ng/kg (equivalent to 26 0, 0.321, 1.07, 10.7, and 107 ng/kg-day adjusted for continuous exposure; administered dose 27 multiplied by 5 and divided by 7). Three days after the final dose, animals were sacrificed and 28 brains were removed for oxidative stress testing. Biomarkers for oxidative stress included 29 production of superoxide anion, lipid peroxidation, and DNA single-strand breaks. A significant This document is a draftfor review purposes only and does not constitute Agency policy. 2-181 DRAFT--DO NOT CITE OR QUOTE 1 (p < 0.05) increase was observed in superoxide anion production, lipid peroxidation as measured 2 by thiobarbituric acid-reactive substances (TBARS), and DNA single-strand breaks with all 3 doses tested. 4 No other indicators of brain pathology were assessed, and it is unfeasible to link the 5 markers of oxidative stress to a TCDD-induced toxicological outcome in the brain. Thus, no 6 LOAEL/NOAEL was established. The subchronic (13-week) LOEL is 0.32 ng/kg-day, based on 7 significant (p < 0.05) increases in superoxide anion production (80% above control); lipid 8 peroxide production (25% above the control); and DNA single-strand breaks (2-fold over the 9 control). No NOEL is established. 10 11 2.4.2.4.8. H assou n et al. (2 0 0 0 ,197431). 12 Hassoun et al. (2000, 197431) examined the effect of subchronic TCDD exposure on 13 oxidative stress in hepatic and brain tissues. Groups of 8-week-old female Harlan Sprague14 Dawley rats (6 rats/group) were administered TCDD (98% purity, dissolved in 1% acetone in 15 corn oil) via gavage at 0, 3, 10, 22, 46, or 100 ng/kg-day, 5 days/week for 13 weeks (0, 2.14, 16 7.14, 15.7, 32.9, or 71.4 ng/kg-day adjusted for continuous exposure; administered doses were 17 multiplied by 5 and divided by 7 days/week). Animals were sacrificed at the end of the study 18 period, and brain and liver tissues were collected and used to determine the production of 19 reactive oxygen species, lipid peroxidation, and DNA single-strand breaks (SSBs). 20 A dose-dependent effect was observed in both the liver and brain tissue as a result of 21 TCDD treatment. Based on the maximal induction of superoxide anion by various doses, more 22 production of superoxide anion was observed in the liver tissue when compared to the brain 23 tissue with an observed increase of 3.1- and 2.2-fold respectively, when compared to the control 24 group. A similar dose-dependent effect was observed in the induction of lipid peroxidation in 25 TCDD-treated animals with an approximately 1.8-fold increase in lipid peroxidation in both 26 tissues relative to the corresponding controls. A dose-dependent relationship was also observed 27 for DNA SSBs in both the hepatic and brain tissues at all TCDD-treated doses compared to 28 controls. Increases were statistically significant (p < 0.05) beginning at the lowest administered 29 dose. 30 Similar to the statement above, because no adverse endpoints were measured, no 31 LOAEL/NOAEL was established. However, a LOEL for TCDD of 2.14 ng/kg-day for a This document is a draftfor review purposes only and does not constitute Agency policy. 2-182 DRAFT--DO NOT CITE OR QUOTE 1 13-week exposure duration was identified in this study for significant increases (p < 0.05) in 2 superoxide anion, lipid peroxidation, and DNA SSBs in the liver and brain tissues. A NOEL 3 cannot be determined for this study. 4 5 2.4.2.4.9. H assou n et al. (2 0 0 3 ,198726). 6 Hassoun et al. (2003, 198726) examined the role of antioxidant enzymes in 7 TCDD-induced oxidative stress in various regions of the rat brain after subchronic exposure. 8 Groups of 8-week-old female Harlan Sprague-Dawley rats (12 rats/group) were administered 9 TCDD (98% purity, dissolved in 1% acetone in corn oil) via gavage at 0, 10, 22, or 46 ng/kg-day 10 (0, 7.14, 15.7, or 32.9 ng/kg-day adjusted for continuous exposure; administered doses were 11 multiplied by 5 and divided by 7) daily for 13 weeks. Animals were sacrificed at the end of the 12 study period and the brain was immediately removed and dissected to the following regions: 13 cerebral cortex (Cc), hippocampus (H), cerebellum (C), and brain stem including midbrain, pons, 14 and medulla. Four pooled samples from each region per dose (i.e., 3 animals/pooled sample) 15 were used in the study. Dissected regions were subsequently assayed for lipid peroxidation 16 (thiobarbituric acid reactive substances, or TBARS), superoxide dismutase, catalase, and 17 glutathione peroxidase. Because the cytochrome c reduction method was used to determine 18 superoxide anion (SA) production in brain tissues, superoxide dismutase (SOD) was added to 19 some of the brain tissue samples that had the highest SA production (tissue homogenates from 20 Cc and H from rats treated with 46 ng/kg-day TCDD). 21 A dose-dependent increase in the production of SA was observed in the Cc and H, but 22 significant changes in SA production were not observed in either the C or the mid-brain, pons, or 23 medulla brain stem cells. Similar to SA production, there was a dose-dependent increase in the 24 production of TBARS in the Cc and H regions of the brain, but no significant changes were 25 observed in either the C or the B sections of the brain. The study authors also measured the 26 activities of various enzymes as a result of TCDD treatment and reported a dose-dependent 27 increase in SOD activity in the C and B sections, while there was dose-dependent suppression in 28 SOD activity in Cc and H. In contrast, catalase activity was significantly (p < 0.05) increased in 29 H and Cc at the 10 ng/kg-day TCDD dose level compared to controls and the mid- and high-dose 30 animals. Catalase activity also was increased in a dose-dependent manner in the C section, but 31 no significant changes in the activity of this enzyme were observed in the B section at any of the This document is a draftfor review purposes only and does not constitute Agency policy. 2-183 DRAFT--DO NOT CITE OR QUOTE 1 three TCDD tested doses. The effects of subchronic exposure to different doses of TCDD on 2 glutathione stimulating hormone peroxidase (GSH-Px) showed a different response compared to 3 other enzymes. There was a dose-dependent increase in the activity of this enzyme in the C and 4 B regions of the brain, while a significant increase in the activity of GSH-Px occurred in Cc and 5 H only at the 10 ng/kg-day TCDD dose. In addition, the activity of this enzyme was suppressed 6 in a dose-dependent manner in the Cc and H at 22 and 46 ng/kg-day TCDD doses. Based on 7 these results, the study authors concluded that induction of oxidative stress by TCDD in the rat 8 brain occurs mainly in the Cc and H regions. 9 Similar to the statement above, because no adverse endpoints were measured, no 10 LOAEL/NOAEL was established. However, a LOEL for TCDD of 7.14 ng/kg-day for a 11 13-week exposure duration was identified for this study for increases in superoxide anion and 12 lipid peroxidation production, as well as increased activity in SOD, catalase, and GSH-Px. 13 14 2.4.2.4.10. K o cib a et al. (1 9 7 6 ,198594). 15 Adult Sprague-Dawley rats (12 per sex per treatment group) were administered TCDD 16 (purity not reported) in corn oil via gavage 5 days per week at doses of 0, 1, 10, 100, or 17 1,000 ng/kg-day (equivalent to 0, 0.71, 7.14, 71.4, or 714 ng/kg-day averaged over 7 days; 5/7 of 18 dose). Five animals per group were sacrificed at the end of treatment, and the remaining animals 19 were observed over 13 weeks post treatment (only initial results for the post-treatment period 20 were provided in the report). Body weights and food consumption were measured semiweekly. 21 Hematology and clinical chemistry were measured after 36-37 or 85-86 days of treatment and 22 59-60 days after termination of treatment. Forty-eight hour urine samples were collected from 23 select rats from 85-89 days of treatment and 52-56 days after cessation of treatment. Gross and 24 histopathological exams were conducted on the tissues. 25 Four high-dose females died during treatment. Two high-dose females and 26 two high-dose males died during the post-treatment period. Animals treated with 714 ng/kg-day 27 were less active during the treatment period, which became less evident during the 28 post-treatment period. Yellow discoloration of the external pinnae also was noted in this group, 29 both during treatment and during the post-treatment period. A significant (p < 0.05) reduction in 30 body weight and food consumption was observed in the 71.4 and 714 ng/kg-day groups. The 31 following significant (p < 0.05) hematology changes were observed in the high-dose This document is a draftfor review purposes only and does not constitute Agency policy. 2-184 DRAFT--DO NOT CITE OR QUOTE 1 (714 ng/kg-day) males at all measured time points: decreased packed cell volume, decreased red 2 blood cells, decreased hemoglobin, increased reticulocytes, and decreased thrombocytes. 3 Significant (p < 0.05) changes also occurred in the high-dose females, but the only consistent 4 observation was a decrease in thrombocytes and increased leukocytes. Significant changes in 5 clinical chemistry (p < 0.05) and urinalysis (p < 0.05) were more consistent between the sexes in 6 the high-dose group and included increases in total and direct serum bilirubin; increase in serum 7 alkaline phosphatase; decreased urinary creatinine; and increased urinary coproporphyrin, 8 uroporphyrin, and delta-amino-levulinic. The following significant (p < 0.05) changes were 9 observed in the 71.4 ng/kg-day group: decreased packed cell volume (4-9%) in males; decreased 10 red blood cells (2-10%) in males; decreased hemoglobin (2-13%) in males; increased urinary 11 coproporphyrin (2.2-fold increase during treatment) in females; increased urinary 12 delta-amino-levulinic (47% increase during treatment) in females; increased total and direct 13 serum bilirubin (48-61%) in females; and increased serum alkaline phosphatase (2-fold) in 14 females. The following significant (p < 0.05) changes in relative organ weights were observed 15 increased brain weight in 714 ng/kg-day males and females; increased liver weight in males 16 (71.4 and 714 ng/kg-day) and females (7.14, 71.4, and 714 ng/kg-day); increased spleen weight 17 in 714-ng/kg-day males and females; decreased thymus weight in 71.4 and 714 ng/kg males and 18 females; and increased testes weight in 714 ng/kg-day males. Microscopic changes were 19 observed in the thymus, and in other lymphoid tissues, and in the liver in rats treated with 20 71.4 ng/kg-day or greater. 21 The subchronic (13-week) LOAEL is 71.4 ng/kg-day, based on the numerous changes 22 noted in body weight, hematology, clinical chemistry, urinalysis, and histopathology. The 23 NOAEL is 7.14 ng/kg-day. 24 25 2.4.2.4.11. M a lly a n d C hipm an (2 0 0 2 ,198098). 26 Female F344 rats (3 per treatment group) were administered TCDD at concentrations of 27 0, 2.5, 25, or 250 ng/kg in corn oil via gavage for either 3 consecutive days or 2 days per week 28 for 28 days (Mally and Chipman, 2002, 198098). The average daily doses for the 28-day study 29 when adjusted for 7 days a week were 0, 0.71, 7.1, and 71 ng/kg-day (i.e., 2/7 of administered 30 dose). No clinical signs of toxicity were observed. Histological examination of the liver 31 revealed no abnormalities. All doses of TCDD reduced the number of connexin (Cx) 32 plaques This document is a draftfor review purposes only and does not constitute Agency policy. 2-185 DRAFT--DO NOT CITE OR QUOTE 1 and Cx32 plaque area in the liver, which was considered the target tissue. The reductions were 2 not statistically significant after the 3-day treatment, but were significant after the 28-day 3 treatment (p < 0.05). TCDD also caused a reduction in the Cx32 plaque number and area in the 4 thyroid after 28 days, but the results were not statistically significant. Although the reduction in 5 Cx32 plaque number and plaque area in the liver and thyroid occurred at all dose levels, there 6 was no relation to dose. TCDD did not induce hepatocyte proliferation. 7 In the absence of additional indicators of hepatotoxicity, changes in Cx32 plaques are not 8 clearly linked to TCDD-mediated hepatotoxicity, nor are they considered an adverse effect. 9 Additionally, no toxicologically-relevant endpoints were examined. Therefore, a NOAEL or 10 LOAEL cannot be determined. A 28-day LOEL at the lowest dose of 0.71 ng/kg-day for 11 significantly (p < 0.05) decreased Cx32 plaque area is evident (approximately 70% of the 12 controls). 13 14 2.4.2.4.12. S leza k et al. (2 0 0 0 ,199022). 15 Slezak et al. (2000, 199022) studied the impact of subchronic TCDD exposure on 16 oxidative stress in various organs of B6C3F1 female mice. Groups of 8- to 10-week-old female 17 B6C3F1 mice (number not specified) were administered TCDD (purity >98%, dissolved in corn 18 oil) via gavage at 0, 0.15, 0.45, 1.5, 15, or 150 ng/kg-day (0, 0.11, 0.32, 1.07, 10.7, or 19 107.14 ng/kg-day adjusted for continuous exposure) 5 days per week for 13 weeks. Three days 20 after the last treatment, the animals were sacrificed and organs were removed for the 21 measurement of oxidative stress indicators including SA, lipid peroxidation (TBARS), and 22 GSH-Px. Tissue TCDD concentrations also were measured. 23 The study authors reported that TCDD dose range resulted in overlapping tissue 24 concentrations for liver, lung, kidney and spleen. Liver had the highest TCDD concentration, 25 with each tissue demonstrating a dose-dependent increase in TCDD concentration. Compared to 26 controls, SA production was significantly (p < 0.05) lower at the 0.15 ng/kg-day TCDD dose, 27 while it was significantly (p < 0.05) higher at 15 and 150 ng/kg-day. A dose-dependent increase 28 in hepatic TBARS production was observed, although the rate of production was significant 29 (p < 0.05) only at the highest TCDD administered dose (150 ng/kg-day) compared to controls. 30 AA also followed the same pattern observed for SA and TBARS with AA production 31 significantly (p < 0.05) increased at the 15 and 150 ng/kg-day TCDD doses. Contrary to the SA, This document is a draftfor review purposes only and does not constitute Agency policy. 2-186 DRAFT--DO NOT CITE OR QUOTE 1 TBARS, and AA responses, GSH levels were decreased at 0.15 ng/kg-day, were increased at 2 0.45 and 150 ng/kg-day, and did not change at 1.5 or 15 ng/kg-day when compared to the control 3 group. Unlike the liver, there was no significant increase in SA production in the lung at any of 4 the TCDD tested doses; a dose dependent reduction, however, was observed at 0.45, 15, and 5 150 ng/kg-day compared to controls. GSH and AA production was decreased at 0.15 ng/kg-day, 6 while AA production was significantly (p < 0.05) increased at 15 and 150 ng/kg-day. Kidney 7 SA production showed a statistically significant (p < 0.05) increase only at the 15 and 8 150 ng/kg-day doses. GSH, like the liver and the lung, exhibited a decrease in production 9 following treatment at 0.15 ng/kg-day with this trend continuing at 0.45 and 1.5 ng/kg-day. AA 10 levels were significantly (p < 0.05) lower at all subchronic doses, except at 1.5 ng/kg-day dose. 11 SA levels in the spleen differed little from the control group at any of the TCDD doses. Total 12 GSH was higher only at the 150 ng/kg-day dose level, while the AA levels were significantly 13 (p < 0.05) decreased at 0.15, 1.5, and 150 ng/kg-day. 14 Similar to the statements regarding the Hassoun et al. studies above, because no adverse 15 endpoints were measured, no LOAEL/NOAEL was established. Therefore, a NOAEL or 16 LOAEL cannot be determined. However, a NOEL and LOEL of 1.07 and 10.7 ng/kg-day, 17 respectively, are identified in this study for increases in superoxide anion in the liver. 18 19 2.4.2.4.13. S m ialow icz et al. (2 0 0 8 ,198341). 20 Female B6C3F1 mice (8-15 per treatment group) were administered TCDD (purity 21 >98%) in corn oil by gavage at doses of 0, 1.5, 15, 150, or 450 ng/kg-day, 5 days a week for 22 13 weeks (1.07, 10.7, 107, or 321 ng/kg-day, adjusted for continuous exposure; i.e., 5/7 of the 23 dose) (Smialowicz et al., 2008, 198341). Mice were immunized 3 days after the final TCDD 24 exposure with an intravenous injection of an optimal concentration of 4 x 107 SRBCs and 25 sacrificed 4 days later. No TCDD-related effects on body weight were observed. There was a 26 dose-related decrease in relative spleen weight (9-19% lower than control values) with 27 statistically significant (p < 0.05) decreases at all but the lowest dose. Additionally, there was a 28 statistically significant (p < 0.05) increase in relative liver weight (5-21%) in all treatment 29 groups compared to controls. Statistically significant dose-dependent decreases were observed 30 in the antibody response to SRBCs (24-89% lower than control values), as measured by both the 31 number of plaque forming cells per 106cells and plaque forming cells per spleen. This document is a draftfor review purposes only and does not constitute Agency policy. 2-187 DRAFT--DO NOT CITE OR QUOTE 1 The 13-week LOAEL for this study is 1.07 ng/kg-day based on a significant (p < 0.05) 2 increase in relative liver weight (10%) and a significant (p < 0.05) decrease in antibody response 3 to SRBCs (24%). A NOAEL cannot be determined for this study. 4 5 2.4.2.4.14. Van B irgelen e t al. (1 9 9 5 ,197096: 1 9 9 5 ,19 8 0 5 2 ) 6 Van Birgelen et al. (1995, 197096; 1995, 198052) studied the impact of TCDD exposure 7 on various biochemical endpoints in rats. Groups of 7-week-old female Sprague-Dawley rats 8 (n = 8 per treatment group) were treated with 0, 200, 400, 700, 5,000, or 20,000 ng/kg TCDD 9 (purity >99%) in diet for 13 weeks. Daily TCDD intake based on food consumption, diet level, 10 and mean weight was estimated to be 0, 14, 26, 47, 320, or 1,024 ng/kg-day. Blood samples 11 were collected from treated animals and assayed for retinol (vitamin A), triiodothyronine, and 12 total (TT4) and free (FT4) thyroxine. At study termination, the animals were sacrificed and the 13 liver, thymus, spleen, and kidneys were removed and weighed. Parts of the liver were 14 homogenized and assayed to determine EROD; CYP1A1; CYP1A2; and UDPGT activity. Liver 15 samples also were analyzed for retinol content. 16 TCDD-treated animals showed a dose-related decrease in food consumption. Animals 17 treated with 1,024 ng/kg-day TCDD consumed 32% less food compared to controls. Similarly, a 18 dose-related decrease in body weight gain was observed in all animals treated with TCDD. 19 Animals treated with >47 ng/kg-day of TCDD showed a statistically significant (p < 0.05) 20 decrease in body weight gain. Relative liver weights were significantly (p < 0.05) increased in 21 the 320 and 1,024 ng/kg-day TCDD dose groups compared to the controls. Absolute and relative 22 thymus weights were significantly (p < 0.05) decreased at all TCDD dose groups compared to 23 the control group. Relative kidney and spleen weights were significantly (p < 0.05) higher in 24 animals dosed with >47 ng/kg-day of TCDD compared to the control group, with the greatest 25 increase occurring in animals treated with 1,024 ng/kg-day TCDD (121 and 173% higher than 26 controls for kidney and spleen, respectively). Cytochrome P450 enzymes, including EROD, 27 CYP1A2, CYP1A1, and UDPGT, exhibited statistically significant (p < 0.05) increases in 28 activity at all TCDD dose groups compared to the control group. TT4 and FT4 thyroid hormone 29 concentrations were statistically significantly (p < 0.05) decreased only at TCDD doses 30 >47 ng/kg-day. A dose-dependent increase was observed in the plasma retinol concentrations 31 with significant (p < 0.05) increases occurring at >47 ng/kg-day TCDD after a 13-week This document is a draftfor review purposes only and does not constitute Agency policy. 2-188 DRAFT--DO NOT CITE OR QUOTE 1 exposure. A dose-dependent reduction in liver retinoid levels also was observed after 13 weeks 2 of TCDD exposure with the levels dropping significantly (p < 0.05) at all TCDD-treated doses 3 compared to the control group. 4 A LOAEL for TCDD of 14 ng/kg for a 13-week exposure is identified for significantly 5 (p < 0.05) decreased absolute and relative thymus weights and significantly (p < 0.05) decreased 6 liver retinoid levels. A NOAEL cannot be determined for this study. 7 8 2.4.2.4.15. Vos e t al., (1 9 7 3 ,198367). 9 Vos et al. (1973, 198367) conducted a study to examine the immune response in 10 laboratory animals treated with TCDD. In one experiment, 10 female Hartley strain guinea pigs 11 were orally treated with 8 weekly doses of 0, 8, 40, 200, and 1,000 ng/kg TCDD in corn oil 12 (purity of TCDD not specified) (0, 1.14, 5.71, 28.6, and 143 ng/kg-day adjusted for continuous 13 exposure; administered dose divided by 7). At study termination, the animals were sacrificed, 14 and heart blood was used to determine total leukocyte and differential leukocyte counts. In 15 another experiment, the effect of TCDD on humoral immunity was determined by injecting 16 0.1 mL of tetanus toxoid into the right hind-foot pad on day 28 (1 left foot tetanus toxoid, 17 aluminum phosphate-adsorbed) and again on day 42 (1 left foot tetanus toxoid, unadsorbed). 18 Blood was collected (n = 10) on days 35 and 49, and the serum tetanus-antitoxin concentrations 19 were determined using a modified single radial immunodiffusion technique. 20 All guinea pigs receiving 1,000 ng/kg-day TCDD either died or were killed when 21 moribund between 24 and 32 days. These animals showed severe weight loss, lymphopenia, and 22 depletion of the lymphoid organs, especially the thymus. Microscopic observations revealed 23 severe atrophy of the thymic cortex with substantial destruction of lymphocytes, with the nuclear 24 debris being engulfed by macrophages. Large cystic Hassall bodies, filled with 25 polymorphonuclear leukocytes were observed in the medulla. All animals treated with 0, 8, 40, 26 or 200 ng/kg-day TCDD survived until study termination. Body weight gain was significantly 27 (p < 0.01) lower in the 200 ng/kg-day group. Absolute thymus weight was significantly reduced 28 in the 40 and 200 ng/kg-day treatment groups (p < 0.01 andp < 0.05, respectively). In contrast, 29 relative thymus weight was significantly (p < 0.01) reduced only in the 200 ng/kg-day dose 30 group. The absolute weight of the superficial cervical lymph nodes was significantly (p < 0.05) 31 decreased in the 200 ng/kg-day group, while the relative adrenal weight was significantly This document is a draftfor review purposes only and does not constitute Agency policy. 2-189 DRAFT--DO NOT CITE OR QUOTE 1 (p < 0.05) increased in the 200 ng/kg-day dose group. Total leukocyte count was significantly 2 (p < 0.05) decreased in the 40 ng/kg-day dose group and total lymphocyte count was 3 significantly decreased at 8, 40, and 200 ng/kg-day (p < 0.01, p < 0.05, andp < 0.05, 4 respectively). A significant (p-values not provided) monotonic dose-response relationship was 5 determined for body weight (decrease), relative thymus weight (decrease), relative adrenal 6 weight (increase), and total leukocyte and lymphocyte count (decrease). Microscopic 7 examination of the lymphoid organs and adrenals showed no effects, while slight cortical atrophy 8 of the thymus was observed at the 200 ng/kg-day dose. 9 Animals receiving the tetanus toxoid injection showed a small but significant increase in 10 serum tetanus antitoxin concentrations at the 8 and 40 ng/kg-day dose (p < 0.05 andp < 0.01, 11 respectively). Measurement at days 49 and 56 indicated that serum antitoxin levels had 12 decreased sharply and the significant (p < 0.05 on day 49 andp < 0.01 on day 56) effect was 13 seen only at the 200 ng/kg-day dose level. 14 A LOAEL for TCDD of 5.71 ng/kg-day for an 8-week exposure is identified in this study 15 for significantly (p < 0.01) reduced absolute thymus weight, significantly (p < 0.05) reduced 16 leukocyte and lymphocyte count, and significantly (p < 0.01) increased serum tetanus antitoxin 17 concentration. The NOAEL for this study is 1.14 ng/kg-day. 18 19 2.4.2.4.16. W hite et al. (1 9 8 6 ,197531). 20 White et al. (1986, 197531) studied the impact of TCDD exposure on serum complement 21 levels. Groups of female (C57BL/6 x C3H)F1(B6C3F1) mice were treated for 14 consecutive 22 days with TCDD in corn oil (purity of TCDD not specified) at doses of 0, 10, 50, 100, 500, 1,000 23 or 2,000 ng/kg-day via gastric intubation (n = 6-8). At study termination, blood was collected 24 from anesthetized animals and assayed for serum complement activity and complement 25 component C3 levels. 26 Serum complement activity between the 10 and 100 ng/kg-day doses was between 69 and 27 59% compared to the vehicle control group, with all treatment groups being significantly 28 (p < 0.05) low compared to the vehicle control. In contrast, C3 levels were comparable to the 29 vehicle control with levels ranging between 98 and 94% of the control group. The higher doses 30 of 500, 1,000, and 2,000 ng/kg-day, however, produced a marked decrease of the component This document is a draftfor review purposes only and does not constitute Agency policy. 2-190 DRAFT--DO NOT CITE OR QUOTE 1 hemolytic activity (45, 35, and 19% of the vehicle control) and of C3 levels (91, 81, and 74 % of 2 the vehicle control, respectively; significance level atp < 0.05). 3 A LOAEL for TCDD of 10 ng/kg-day for a 14-day exposure is identified in this study for 4 significantly (p < 0.05) lower serum complement activity. A NOAEL cannot be determined for 5 this study. 6 7 2.4.2.5. C hronic S tu d ies (N oncancer E n dpoints) 8 2.4.2.5.1. C antoni e t al. (1 9 8 1 ,197092). 9 CD-COBS rats (4 per treatment) were orally administered TCDD (purity not specified) 10 dissolved in acetone:corn oil (1:6) at doses of 0 (vehicle alone), 10, 100, or 1,000 ng/kg per week 11 (equivalent to 1.43, 14.3, and 143 ng/kg-day adjusted for continuous exposure, administered 12 dose by dividing the dose by 7) for 45 weeks. Urine was collected several times during 13 treatment and tested for porphyrin excretion. Twenty-four hours after the final dose, animals 14 were sacrificed and their livers, spleens, and kidneys were removed for analysis of total 15 porphyrins. All treatment groups had a significant (p < 0.05) increase in coproporphyrin 16 excretion beginning at 6, 3, or 2 months, respectively. Uroporphyrin excretion was significantly 17 (p < 0.05) increased in the 14.3 ng/kg-day group at 10 months and in the 143 ng/kg-day group 18 beginning at 6 months. The high-dose group also had a significant (p < 0.05) increase in 19 excretion of heptacarboxylic methyl ester beginning at 6 months. The high-dose group had a 20 marked porphyric state beginning at 8 months as indicated by a 70-fold increase above controls 21 in total urinary porphyrin excretion. This group also had a significant (p < 0.05) increase in total 22 porphyrins in the liver, kidneys, and spleen. 23 The 45-week LOAEL for this study is 1.43 ng/kg-day, based on a 2- to 3-fold increase in 24 urinary coproporphyrin excretion. No NOAEL was established for this study. 25 26 2.4.2.5.2. C routch et al. (2 0 0 5 ,197382). 27 Croutch et al. (2005, 197382) examined the impact of TCDD exposure on body weight 28 via insulin-like growth factor (IGF) signaling. Female Sprague-Dawley rats were randomly 29 assigned in groups of five to initial loading doses of TCDD (purity >98.5%, dissolved in corn 30 oil) at 0, 12.5, 50, 200, 800, or 3,200 ng/kg-day, followed by treatment with maintenance doses 31 equivalent to 10% of the initial loading dose every third day to maintain a pharmacokinetic This document is a draftfor review purposes only and does not constitute Agency policy. 2-191 DRAFT--DO NOT CITE OR QUOTE 1 steady state throughout the entire study (equivalent to: 14-day average = 0, 1.25, 5, 20, 80, or 2 320 ng/kg-day; 28-day average = 0, 0.85, 3.4, 13.6, 54.3, or 217 ng/kg-day; 63-day average = 0, 3 0.60, 2.4, 9.5, 38, or 152 ng/kg-day; and 128-day average dose = 0, 0.51, 2.0, 8.1, 32.5, or 4 130 ng/kg-day). Following 2, 4, 8, 16, 32, 64, or 128 days of initial dosing, the animals were 5 sacrificed, livers were removed and weighed, and trunk blood was collected to analyze glucose 6 content. Rat liver phosphoenolpyruvate carboxykinase (PEPCK) mRNA and protein levels also 7 were analyzed, and PEPCK activity was measured. 8 Body weights of TCDD-treated animals decreased after the second week of the 9 3.200 ng/kg-day TCDD loading dose, with significant differences beginning at week 9. There 10 was also a statistically significant (p < 0.05) difference in body weights at weeks 10, 11, 13, 18, 11 and 19 at the highest loading dose (3,200 ng/kg-day). PEPCK activity in the liver was also 12 decreased in a dose-dependent manner following TCDD administration at approximately 13 16 days. PEPCK inhibition was statistically significant (p < 0.05) on day 4 in rats treated with 14 either 800 or 3,200 ng/kg-day TCDD when compared to animals treated with a loading dose of 15 200 ng/kg-day. A similar statistically significant change was observed in animals treated with 16 3.200 ng/kg-day on day 16 when compared to the 200 ng/kg-day treatment group. In contrast, 17 differences in PEPCK activity at other doses or time points were not statistically significant. In 18 TCDD-treated animals, there was also a dose-dependent decrease in PEPCK mRNA expression 19 along with a decrease in PEPCK protein levels in the liver. In addition to body weight and 20 PEPCK activity changes, animals treated with 3,200 ng/kg-day TCDD showed a sharp decline in 21 circulating IGF-I levels on day 8 compared to the control group (corn oil) and TCDD-treated 22 animals at lower doses. In the highest dose animals, IGF-I levels continued to decline to 42% of 23 the control group by day 16 of the study. The IGF-I levels at the highest dose plateaued at an 24 average decrease of 66% through day 128 when compared to controls. Beginning at day 8, the 25 decrease in IGF-I was statistically significant at every time point through day 128 compared to 26 the control group, as well as groups treated with either 12.5 or 50 ng/kg-day TCDD. Similar 27 statistically significant decreases also were observed for the 800 ng/kg-day TCDD-treated groups 28 with an initial decrease of 37% on day 16 followed by a further decline to approximately 45% 29 thereafter compared to controls and the 12.5, 50, and 200 ng/kg-day dose groups. In contrast to 30 these results, circulating levels of insulin and glucose were unaffected by TCDD treatment, while This document is a draftfor review purposes only and does not constitute Agency policy. 2-192 DRAFT--DO NOT CITE OR QUOTE 1 the active or phosphorylated form of AMPK-a protein increased with dose as a result of TCDD 2 treatment. 3 A LOAEL for TCDD of 217 ng/kg-day for a 28-day exposure duration (because this 4 represented the most sensitive time for elicitation of effects) was identified in this study for 5 decreased body weight, significant (p < 0.05) inhibition of PEPCK activity, and reduced IGF-I 6 levels (42% lower than the control group). A NOAEL of 54.3 ng/kg-day was identified in this 7 study. 8 9 2.4.2.5.3. H assou n et al. (2002, 543725). 10 Hassoun et al. (2002, 543725) examined the potential of TCDD and other dioxin-like 11 chemicals to induce oxidative stress in a chronic rat bioassay. Groups of six Harlan 12 Sprague-Dawley female rats were treated with 0, 3, 10, 22, 46, or 100 ng/kg-day TCDD 13 (98% purity), 5 days a week via gavage for 30 weeks. The administered doses adjusted for 14 continuous exposure were 0, 2.14, 7.14, 15.7, 32.9, and 71.4 ng/kg-day, respectively 15 (administered doses were multiplied by 5 and divided by 7). At study termination, hepatic and 16 brain tissues from all treated rats were divided into two portions and examined for the production 17 of reactive oxygen species and SSBs in DNA. 18 When compared to controls, there was a dose-dependent increase in the production of 19 superoxide anion in TCDD-treated animals ranging from 21-998% and 66-257% in hepatic and 20 brain tissues, respectively. Hepatic tissues had statistically significant (p < 0.05) increases in 21 superoxide anion production at doses >7.14 ng/kg-day, while the brain tissue had a statistically 22 significant (p < 0.05) increase over controls at all doses. Similarly, increases in lipid 23 peroxidation were observed in hepatic and brain tissues with a 481% increase (p < 0.05) at 24 71.4 ng/kg-day in the hepatic tissue when compared to controls. The increase in lipid oxidation 25 in brain tissue ranged from 33-188% (p < 0.05) in the 2.14-71.4 ng/kg-day dose groups. DNA 26 SSBs were also observed in both hepatic and brain tissue in all treated groups. When compared 27 to the control group, there was a dose-dependent statistically significant (p < 0.05) increase in 28 DNA SSBs ranging from 58-322% and 29-137% in hepatic and brain tissues, respectively. 29 Nonmonotonic dose-response relationships were observed for superoxide production and lipid 30 peroxidation in liver tissues, with greater-than-linear increases in effect between the two highest 31 dose levels. This document is a draftfor review purposes only and does not constitute Agency policy. 2-193 DRAFT--DO NOT CITE OR QUOTE 1 As stated above, because no adverse endpoints were measured, no LOAEL/NOAEL was 2 established. However, a LOEL for TCDD of 2.14 ng/kg-day for a 30-week exposure duration is 3 identified in this study for significant (p < 0.05) increases in superoxide anion, lipid peroxidation 4 production, and DNA SSBs in the liver and brain tissues. A NOEL cannot be determined for this 5 study. 6 7 2.4.2.5.4. K o cib a e t al. (1978, 001818). 8 Sprague-Dawley rats (50 per sex per treatment group) were administered TCDD (purity 9 >99%) in the diet at doses of 0, 1, 10, or 100 ng/kg-day for 2 years. Body weights and food 10 consumption were routinely measured. Hematology, clinical chemistry, and urinalysis were 11 measured after 3, 12, or 23 months of treatment. Animals were routinely palpitated for tumors. 12 Gross and histopathological exams were conducted on the tissues of dead or dying animals or at 13 terminal sacrifice. Specific organs also were weighed. 14 The high-dose females had a statistically significant (p < 0.05) increase in mortality 15 compared to the controls during the second half of the study. Mortality changes in males were 16 variable and of questionable toxicological significance. A significant (p < 0.05) reduction in 17 body weight occurred in the 100 ng/kg-day males and females beginning at 6 months. Mid-dose 18 females also had reduced body weight, but to a lesser degree during the same time frame. There 19 were no consistent changes in food consumption. The following significant (p < 0.05) 20 hematology changes were observed in the high-dose animals: decreased packed cell volume in 21 males after 3 months and in females after 1 year, decreased red blood cells in females after 22 1 year and in males at terminal sacrifice, decreased hemoglobin in males after 3 months and in 23 females after 1 year, and decreased total white blood cell count in females after 1 year. Changes 24 in clinical chemistry (p < 0.05) occurred only in high-dose females and consisted of an increase 25 in serum alkaline phosphatase and gamma glutamyl transferase. Significant changes in 26 urinalysis occurred only in females and included increased urinary coproporphyrin in the mid27 and high-dose groups, increased urinary uroporphyrin in the mid- and high-dose groups, and 28 increased urinary delta-amino-levulinic acid in the high-dose group. Significant (p < 0.05) 29 changes in relative organ weights were observed, including increased liver weight in mid- and 30 high-dose females and decreased thymus weight in high-dose females. Mid- and high-dose rats 31 showed hepatocellular degeneration and inflammatory and necrotic changes in the liver. Thymic This document is a draftfor review purposes only and does not constitute Agency policy. 2-194 DRAFT--DO NOT CITE OR QUOTE 1 and splenic atrophy were noted in high-dose females. An increase in non-neoplastic lung lesions 2 was noted in mid-dose females and high-dose males and females. High-dose females had an 3 increase in uterine changes. High-dose males had a significant (p < 0.05) increase in the 4 incidence of stratified squamous cell carcinomas of the tongue. High-dose males and females 5 had a significant (p < 0.05) increase in the incidence of squamous cell carcinomas of the hard 6 palate/turbinates. 7 The chronic (2-year) LOAEL is 10 ng/kg-day, based on the numerous significant 8 (p < 0.05) changes noted in coproporphyrin excretion (67% increase above control) and an 9 increase in liver and lung lesions in female rats. The NOAEL is 1 ng/kg-day. 10 11 2.4.2.5.5. M a ro n p o t e t al. (1 9 9 3 ,198386). 12 An initiation-promotion study was performed in female Sprague-Dawley rats (8-10 rats 13 per group). Rats were initiated with saline or diethylnitrosamine (DEN), followed 2 weeks later 14 by promotion with biweekly administration of TCDD (purity not specified) in corn oil via 15 gavage for 30 weeks. The doses were stated to be equivalent to 3.5, 10.7, 35.7, or 16 125 ng/kg-day. Rats were sacrificed 7 days after the final treatment. A significant (p < 0.05) 17 decrease in body weight occurred in the 125 ng/kg-day group. A significant (p < 0.05) increase 18 in relative liver weight occurred in the 35.7 and 125 ng/kg-day groups. There was a significant 19 (p < 0.05) increase in the labeling index in the 125 ng/kg-day group, but only with DEN 20 initiation. In the TCDD-alone group, a 2-fold increase in labeling index occurred in the 21 125 ng/kg-day group that did not reach statistical significance. A significant (p < 0.05) trend for 22 increased alkaline phosphatase levels was observed in TCDD-treated animals, but despite a 23 50% increase in the highest dose group the increase was not statistically significant. Total 24 cholesterol and triglycerides were significantly (p < 0.05) higher in the 25 125 ng/kg-day TCDD-alone group. A significant (p < 0.05) increase in 5'-nucleotidase occurred 26 in the 35.7 and 125 ng/kg-day TCDD-alone groups. A dose-dependent increase in the incidence 27 and severity of liver toxicity as measured by microscopic lesions was observed. 28 The 30-week LOAEL is 35.7 ng/kg-day, based on a significant (p < 0.05) increase in 29 relative liver weight (12%, accompanied by increases in incidence and severity of liver lesions). 30 The 30-week NOAEL is 10.7 ng/kg-day. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-195 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.5.6. N a tio n a l T oxicology P ro g ra m (1982, 543764). 2 National Toxicology Program (NTP, 1982, 543764) conducted a carcinogenic bioassay of 3 TCDD on rats and mice. Fifty male and female Osborne-Mendel rats and male and female 4 B6C3F1 mice were treated twice per week with TCDD (purity not specified) in corn oil via oral 5 gavage at doses of 0, 5, 25, or 250 ng/kg for rats and male mice (1.4, 7.1, 71 ng/kg-day adjusted 6 for continuous exposure; administered doses multiplied by 2 and divided by 7) and 0, 20, 100, or 7 1,000 ng/kg for female mice (5.7, 28.6, or 286 ng/kg-day adjusted for continuous dosing; 8 administered doses multiplied by 2 and divided by 7) for 104 weeks. Seventy-five rats and mice 9 of each sex served as vehicle controls. One untreated control group of 25 rats and mice of each 10 sex was present in the TCDD treatment room and one untreated control group consisting of 11 25 rats and mice of each sex were present in the vehicle-control room. Animals surviving until 12 study termination were sacrificed at 105 or 108 weeks. A complete histopathological evaluation 13 was conducted on all animals. 14 Survival rates were not affected by TCDD exposure in rats or mice of either sex. Male 15 rats exhibited a dose-related depression in mean body weight after week 55, while the females 16 exhibited a dose-related body-weight depression after 45 weeks of TCDD exposure. However, 17 the magnitude of the body weight response is not indicated. Mean body weights in male and 18 female mice were comparable to the vehicle control group throughout the bioassay. Noncancer 19 histopathologic findings included increased incidences of liver lesions (termed toxic hepatitis) 20 from TCDD exposure, and were detected in the high-dose rats and high-dose mice of each sex. 21 A LOAEL for TCDD of 1.4 ng/kg-day for a 104-week exposure duration is identified for 22 increased incidences of liver lesions in mice of both sexes. A NOAEL cannot be determined for 23 this study. 24 25 2.4.2.5.7. N a tio n a l T oxicology P ro g ra m (2 0 0 6 ,197605). 26 Female Sprague-Dawley rats (81 control; 82 treatment group) were administered TCDD 27 (purity >98%) in corn oil:acetone (99:1) via gavage at doses of 0, 3, 10, 22, 46, or 28 100 ng/kg-day, 5 days per week for 105 weeks (0, 2.14, 7.14, 15.7, 32.9, or 71.4 ng/kg-day, 29 adjusted for continuous exposure) (NTP, 2006, 197605). In addition to this primary group, a 30 stop group of 50 animals was administered 100 ng/kg-day TCDD in corn oil:acetone (99:1) via 31 gavage for 30 weeks and then just the vehicle for the remainder of the study. Up to 10 rats per This document is a draftfor review purposes only and does not constitute Agency policy. 2-196 DRAFT--DO NOT CITE OR QUOTE 1 dose group were sacrificed and evaluated at 14, 31, or 53 (n = 8) weeks for biologically 2 noteworthy changes in the incidences of neoplasms or non-neoplastic lesions in the liver, lung, 3 oral mucosa, uterus, pancreas, thymus, adrenal cortex, heart, clitoral gland, ovary, kidney, 4 forestomach, bone marrow, mesentery gland, and pituitary gland. All interim sacrifice animals 5 also received a complete necropsy and microscopic examination, and the following organs were 6 weighed: the left kidney, liver, lung, left ovary, spleen, thymus (14 weeks only), and thyroid 7 gland. Out of 53 control animals and 53 or 54 animals per treatment group not used for interim 8 sacrifice analyses, at study termination the number of surviving animals had declined to 25 in the 9 control group and to 21, 23, 19, 22, and 21 in five treatment groups, respectively, due to 10 accidental deaths, moribund animals, or death due to natural causes. 11 Survival rate was not affected by TCDD treatment. Mean body weights in the high dose 12 primary study group and the 100 ng/kg stop group were less than the vehicle control group after 13 week 13 of the study. The mean body weights of animals in the 46 ng/kg-day group were less 14 than in the vehicle control at study termination (2 years), whereas animals in the 22 ng/kg-day 15 had lower mean body weights compared to controls during the last 10 weeks of study. In 16 addition to body weight changes, liver weights were also impacted as a result of TCDD 17 exposure. Absolute and relative liver weights were significantly (eitherp < 0.01 orp < 0.05) 18 higher in all dose groups compared to controls at the 14- and 31-week evaluation period, whereas 19 the relative liver weights were significantly (eitherp < 0.01 orp < 0.05) higher only at 20 >10 ng/kg-day at 53 weeks. 21 No clinical findings associated with TCDD treatment were observed. TCDD caused 22 changes in thyroid hormone levels at 14, 31, and 53 weeks. The following changes were 23 statistically significant (p < 0.05) compared to the vehicle control: decrease in TT4 at doses 24 >22 ng/kg-day at 14 and 31 weeks and at doses >46 ng/kg-day at 53 weeks; decrease in FT4 at 25 doses >22 ng/kg-day at 14 and 31 weeks; increase in total T3 at doses >46 ng/kg-day at 14 and 26 31 weeks and at doses >10 ng/kg-day at 53 weeks; and increase in TSH at doses >46 ng/kg-day 27 at 14 weeks. There was a statistically-significant (p < 0.05) increase in hepatocyte proliferation 28 at 14 weeks (22 ng/kg-day group only); 31 weeks (all doses); and 53 weeks (>46 ng/kg-day). 29 There were statistically significant (p < 0.01) dose-dependent increases in liver (includes EROD 30 [CYP1A1-associated] activity; 7-pentoxyresorufin-O-deethylase [PROD; CYP2B-associated] 31 activity; and acetanilide-4-hydroxylase [CYP1A2-associated] activity) and lung (EROD) This document is a draftfor review purposes only and does not constitute Agency policy. 2-197 DRAFT--DO NOT CITE OR QUOTE 1 cytochrome P450 enzyme activities in all treatment groups at all three evaluation periods 2 compared to the vehicle control group. The largest effect was an 82-fold induction of hepatic 3 EROD activity in the 46 ng/kg-day group at 31 weeks. 4 TCDD was detected at the greatest concentration in the liver, followed by fat tissue, with 5 tissue concentration increasing in both of these tissues in a dose-dependent manner. TCDD 6 tissue levels generally remained constant after the first measurement at week 14. Pathological 7 examination at week 14 revealed increased incidences of hepatocellular hypertrophy in animals 8 administered >10 ng/kg-day TCDD. Examinations at weeks 31 and 53 indicated that incidence 9 and or severity of hepatocellular hypertrophy was increased at all treatment doses although 10 incidences were statistically significant (p < 0.05) only at >10 ng/kg-day doses. The incidence of 11 non-neoplastic hepatic lesions (including inflammation, necrosis, multiple eosinophilic focus, 12 diffuse fatty change, pigmentation, toxic hepatopathy) in the liver increased at doses 13 >22 ng/kg-day beginning at 14 weeks. Severity of the lesions increased at 14 weeks at doses 14 >46 ng/kg-day and were also observed at lower dose levels during later evaluation periods (31 15 and 53 weeks). By terminal sacrifice, numerous non-neoplastic changes were noted in TCDD 16 treated rats, even at the lowest dose tested. 17 Noncancer cardiovascular and pulmonary effects were evident after 2 years of TCDD 18 exposure. Significantly increased incidences of minimal to mild cardiomyopathy were seen in 19 male and female rats at >10 ng/kg-day. In the lung, there was a significant (p < 0.01) 20 dose-dependent increase, when compared to the vehicle control, in the incidence of bronchiolar 21 metaplasia of the alveolar epithelium at all dose groups in the primary study. 22 A LOAEL for TCDD of 2.14 ng/kg-day adjusted dose for a 105-week exposure duration 23 is identified in this study for significantly (eitherp < 0.01 orp < 0.05) increased absolute and 24 relative liver weights, increased incidence of hepatocellular hypertrophy, and increased incidence 25 of alveolar to bronchiolar epithelial metaplasia. A NOAEL cannot be determined for this study. 26 27 2.4.2.5.8. R ie r e t al. (2 0 0 1 ,198776; 2001, 543773). 28 Female rhesus monkeys (8 per treatment group) were administered 0, 5, or 25 ppt TCDD 29 (purity not specified) in the diet for 4 years. Previously, Bowman et al. (1989, 543745) 30 determined that these dietary concentrations were equivalent to 0, 0.15, and 0.67 ng/kg-day, 31 respectively. Thirteen years after termination of TCDD treatment, serum concentrations of This document is a draftfor review purposes only and does not constitute Agency policy. 2-198 DRAFT--DO NOT CITE OR QUOTE 1 TCDD and dioxin-like polyhalogenated aromatic hydrocarbons (PHAH) were measured in 2 six control monkeys, six monkeys treated with 0.15 ng/kg-day, and three monkeys treated with 3 0.67 ng/kg-day (Rier et al., 2001, 198776). Even after 13 years without treatment, there was 4 significantly (p < 0.05) elevated serum levels of TCDD and other dioxin-like compounds in 5 treated monkeys. There was a significant increase in triglycerides and total lipids in the serum of 6 monkeys treated with either 0.15 or 0.67 ng/kg-day, but not in cholesterol or phospholipids. In 7 addition to these 15 animals, 8 other female monkeys (4 treated with 0.67 ng/kg-day TCDD that 8 died 7 to 11 years after treatment and 4 lead-treated animals with no history of PHAH exposure) 9 were evaluated for endometriosis. Elevated serum concentrations of TCDD were not correlated 10 with endometriosis. Increased serum levels of 3,3',4,4'-tetrachlorobiphenyl (TCB), however, 11 were associated with the presence and severity of endometriosis (p < 0.05). TCB was found in 12 none of the animals without endometriosis, including TCDD-treated animals, nor was it found in 13 control animals with endometriosis. Animals with elevated serum levels of TCB, 14 pentachlorobiphenyl, and total serum analyte TCDD equivalents (TEQ) had an increased 15 incidence of endometriosis, but severity was associated only with increased levels of TCB. EPA 16 did not develop a LOAEL for TCDD for this study, because of DLC contamination. 17 In a separate study that evaluated the same 15 monkeys 13 years after exposure, Rier 18 et al. (2001, 543773) examined effects on systemic immunity. Peripheral blood mononuclear 19 cells (PBMC) obtained from untreated monkeys secreted no detectable levels of TNF-a in 20 response to T-cell mitogen exposure. There was, however, a significant (p < 0.05) 21 dose-dependent increase in TNF-a production in PBMC from the TCDD-treated monkeys. 22 Although PBMC from treated monkeys with endometriosis produced more TNF-a than cells 23 from unexposed controls without the disease (median 128 pg/mL compared to not detected; 24 p < 0.01), PBMC from TCDD-treated animals without endometriosis also produced more TNF-a 25 than controls (median 425 pg/mL, p < 0.067). TNF-a production from the animals without 26 endometriosis, however, was much more variable and was not statistically significant compared 27 to controls. In addition, there was a dose-related but statistically insignificant decrease in PBMC 28 cytotoxicity against natural killer-sensitive RAJI cells in TCDD-treated animals compared to the 29 unexposed controls. The results were again related to TCDD exposure and not the presence of 30 endometriosis. TCDD alone was not associated with changes in PBMC surface antigen 31 expression, but increased serum levels of TCDD. 1,2,3,6,7,8-Hexachlorodibenzofuran and This document is a draftfor review purposes only and does not constitute Agency policy. 2-199 DRAFT--DO NOT CITE OR QUOTE 1 3,3',4,4',5-pentachlorobiphenyl were correlated with increased numbers of CD3+/CD25- and 2 CD3-/CD25+ leukocytes, as well as increased secretion of TNF-a in response to T-cell mitogen 3 exposure. Although TNF-a production is considered to be a general indicator of inflammation, 4 relative adversity of increased TNF-a secreted by PBMCs in and of itself cannot be substantiated 5 in the absence of concurrent physiological measurements of an inflammatory response. 6 Therefore, neither a LOAEL nor NOAEL can be determined for this study. 7 8 2.4.2.5.9. S ew a ll e t al. (1 9 9 3 ,197889). 9 Sewall et al. (1993, 197889) examined the impact of TCDD exposure on the hepatic 10 epidermal growth factor receptor (EGFR) as a critical effect in hepatocarcinogenicity. In 11 two separate experiments, groups of 6- to 8-week-old female Sprague-Dawley rats were 12 randomly assigned to the following groups: control group, receiving saline and corn oil; a 13 promoted group that received four different doses of TCDD along with saline; a DEN-only 14 initiated control group; and a DEN and TCDD initiated and promoted group that received 15 four different doses of TCDD. DEN was administered via intraperitoneal injection at a dose of 16 175 mg/kg [saline (S) vehicle] as the initiating agent to animals that were 70 days old. The 17 control animals received saline only. In the first experiment, each treatment group (S/TCDD and 18 DEN/TCDD) that included sham-operated or ovariectomized and intact animals were treated 19 with TCDD (purity >98%) at 125 ng/kg-day. In the second dose-response experiment, 20 DEN-initiated and saline control treatment groups (intact animals, 84 days old) were 21 administered TCDD (purity >98%) in corn oil via oral gavage once every 2 weeks for 30 weeks 22 at doses equivalent to 0, 3.5, 10.7, 35.7, or 125 ng/kg-day (n = 9). A week after the last 23 treatment, all animals were sacrificed and livers were harvested and fixed for 24 immunohistochemistry. Sections of the fixed liver were tested for EGFR binding, EGFR 25 autophosphorylation, immunolocalization of EGFR, and hepatic cell proliferation. 26 In the first experiment, intact animals treated with 125 ng/kg-day TCDD exhibited a 27 65% reduction in EGFR binding capacity. In contrast, the EGFR equilibrium maximum binding 28 capacity (Bmax) of the ovariectomized rats was not statistically different from the ovariectomized 29 control rats, and no changes in the Kdwere detected in any treatment group. In the 30 dose-response experiment with intact animals, a significant (p < 0.05) TCDD dose-dependent 31 decrease in the Bmax of EGFR was shown. A two-factor, five-level ANOVA indicated that the This document is a draftfor review purposes only and does not constitute Agency policy. 2-200 DRAFT--DO NOT CITE OR QUOTE 1 effect of TCDD exposure on EGFR Bmaxwas significant (p = 0.0001), whereas, the effect of 2 DEN treatment on EGFR Bmaxwas not significant. Comparative analysis using Fisher's 3 protected least significant difference indicated that the lowest TCDD dose resulting in a 4 statistically significant (p < 0.05) decrease in the EGFR Bmaxwas 10.7 ng/kg-day S/TCDD 5 group. At the highest TCDD dose of 125 ng/kg-day, the EGFR Bmaxwas reduced by 38% 6 compared to controls in both the DEN initiated and noninitiated groups. A two-factor, five-level 7 ANOVA showed no significant effect on EGFR Kdin either the DEN- or the TCDD-treated 8 groups. The EGFR autophosphorylation assay indicated that, with increasing TCDD dose, the 9 amount of EGFR autophosphorylation in DEN/TCDD-treated animals decreased. The study 10 authors state that this decrease is similar to the dose-response alterations observed for the EGFR 11 Bmax. Additionally, EGFR autophosphorylation in control and 125 ng/kg-day noninitiated 12 animals was similar to the corresponding dose levels for the DEN-treated animals, suggesting 13 that DEN treatment did not affect the EGFR or the EGFR response to TCDD under the 14 experimental conditions. The immunolocalization assay indicated that staining was more 15 apparent in the centrilobular and midzonal regions of the liver in the DEN initiated control 16 animals, whereas, the amount of hepatocyte plasma membrane staining in DEN/TCDD treated 17 animals substantially decreased. The cell proliferation assay showed a decrease in the cell 18 labeling index in the 3.5 ng/kg-day DEN/TCDD dose group that was statistically less (p < 0.05) 19 than the labeling index for the control group. In contrast, the labeling index for the 20 125 ng/kg-day DEN/TCDD treatment group was significantly (p < 0.05) higher compared to 21 controls. Except for the low-dose (3.5 ng/kg-day) group, a clear dose-response trend 22 (two mid-level doses were not statistically significant) was observed in the other three TCDD 23 treated groups. 24 The role of EGFR in TCDD-mediated hepatotoxicity is unknown, and as such, this 25 endpoint cannot be unequivocally linked to TCDD-induced hepatotoxicity nor labeled as 26 adverse. Thus, no LOAEL/NOAEL was established. A LOEL for TCDD of 3.5 ng/kg-day for a 27 30-week exposure duration was identified in this study for a significant (p = 0.0001 using 28 ANOVA) decrease in EGFR Bmaxlevels. A NOEL cannot be determined for this study. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 2-201 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.5.10. S ew a ll et al. (1 9 9 5 ,198145). 2 Sewall et al. (1995, 198145) studied the dose-response relationship for thyroid function 3 alterations in female rats as a result of TCDD exposure. Groups of female Sprague-Dawley rats 4 were initiated with DEN at 70 days of age at a dose of 175 mg/kg in a saline vehicle via an i.p. 5 injection. DEN was administered as a liver-initiating agent for a concurrent study to determine 6 TCDD promotion of hepatic preneoplastic foci. Saline-treated animals served as controls. At 7 84 days of age, both the DEN-initiated and the saline-noninitiated groups of animals were 8 administered TCDD (purity >98%) or corn oil vehicle via oral gavage once every 2 weeks for 9 30 weeks at dose levels equivalent to 0, 0.1, 0.35, 1.0, 3.5, 10.7, 35.7, or 125 ng/kg-day (n = 9 10 per group). One week after the last TCDD treatment, the animals were sacrificed and the thyroid 11 was removed and fixed for further analysis. Blood was drawn from the abdominal aortic vein, 12 and the serum was isolated and preserved for hormone analysis. Liver was also removed and 13 prepped for further analysis. Thyroid hormone analysis was performed to determine serum TSH, 14 T3, and T4 levels using radioimmunoassay kits. Histological examination was conducted on 15 eosin-stained sections of the thyroid tissue. RNA level in the hepatic tissue was determined 16 using a reverse transcription polymerase chain reaction (RT-PCR) technique. 17 TCDD treatment did not affect thyroid weight. A dose-dependent decrease in serum 18 T4 levels was observed in both noninitiated and DEN-initiated animals with T4 levels dropping 19 significantly (p < 0.05) at the 35 and 125 ng/kg-day TCDD doses in the noninitiated group. 20 Compared to the noninitiated control group, DEN alone did not significantly affect T4 levels. 21 Serum T3 level in the 125 ng/kg-day treatment group was slightly elevated but was not 22 significantly different from levels in the control group. TSH levels in DEN initiated rats were 23 increased at a dose of 3.5 ng/kg-day. In the noninitiated group, TSH level in the 125 ng 24 TCDD/kg-day group was 3.27 0.34 ng/mL (n = 9) compared to 1.3 0.18 ng/mL in the corn 25 oil control group (n = 7). This result, in conjunction with the T4 data, demonstrates that TCDD 26 had a similar effect on thyroid hormone levels in both the noninitiated and DEN initiated groups. 27 Histological sections examined for nodular lesions or neoplasms exhibited thyroid follicular 28 adenoma in one DEN/corn oil control animal. The DEN/TCDD-treated animals exhibited 29 diffuse follicular hyperplasia, with the size of colloidal follicles decreasing with TCDD 30 treatment. Other qualitative DEN/TCDD-related changes included increased frequency of 31 abnormally shaped follicles. The study authors reported that image analysis demonstrated a This document is a draftfor review purposes only and does not constitute Agency policy. 2-202 DRAFT--DO NOT CITE OR QUOTE 1 significant (p = 0.013) TCDD dose-related decrease in mean follicle size along with a significant 2 (p = 0.001) TCDD dose-related increase in parenchymal area. Additionally, like T4 and TSH 3 levels, DEN treatment alone or in combination with TCDD did not influence thyroid follicular or 4 C-cell morphology. 5 RT-PCR results for UGT1 and CYP1A1 mRNA levels indicated that the amount of 6 UGT1 mRNA at the 125 ng/kg-day dose was approximately 2.5-fold higher compared to the 7 concurrent controls. The study authors also stated that the maximal response for the UGT1 8 mRNA levels was reached at a dose between 1.0 and 3.5 ng TCDD/kg-day. In contrast, the 9 maximum induction of CYP1A1 mRNA was 260-fold higher at the 125 ng/kg-day compared to 10 the concurrent controls. 11 A LOAEL for TCDD of 35 ng/kg-day for a 30-week exposure duration was identified in 12 this study for a significant (p < 0.05) decrease in T4 levels. The NOAEL for this study is 13 10.7 ng/kg-day. 14 15 2.4.2.5.11. Toth et al. (1 9 7 9 ,197109). 16 Toth et al. (1979, 197109) examined the impact of TCDD exposure on the formation of 17 liver tumors in male mice. Ten-week-old, outbred Swiss/H/Riop male mice were administered 18 sunflower oil or TCDD (purity not specified; in sunflower oil) at 0, 7, 700 or 7,000 ng/kg (0, 1, 19 100, or 1,000 ng/kg-day adjusted for continuous dosing; administered dose divided by 7; n = 38, 20 44, 44, and 43, respectively) once per week via gastric tube for 1 year. Once exposure had 21 ceased, animals were followed for the rest of their lives. After spontaneous death or when mice 22 were moribund, autopsies were performed and all organs were examined histologically. 23 Average life span in the 1,000 ng/kg-day dose group decreased considerably (72%) when 24 compared to the control group. TCDD also caused dose-dependent, severe chronic and ulcerous 25 skin lesions (12, 30, and 58% in the 1, 100, and 1,000 ng/kg-day dose groups, respectively) that 26 was followed by generalized lethal amyloidosis (12, 23, and 40% in the 1, 100, and 27 1,000 ng/kg-day dose groups, respectively). 28 A LOAEL for TCDD of 1 ng/kg-day for 1-year exposure duration was identified in this 29 study for severe chronic and ulcerous skin lesions (12% higher than controls), and generalized 30 lethal amyloidosis (12% higher than controls). A NOAEL cannot be determined for this study. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 2-203 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.6. C hronic S tu d ies (C ancer E n dpoints) 2 2.4.2.6.I. D ella P orta et al. (1 9 8 7 ,197405). 3 Della Porta et al. (1987, 197405) studied the long-term carcinogenic effects of TCDD in 4 B6C3F1 (C57BL/6JDp x C3Hf/Dp) mice. Six-week-old male and female mice (initially about 5 15/sex/dose, and increased by approximately 30 to 40 per group within a few weeks) were 6 administered 0, 2,500, and 5,000 ng/kg TCDD (purity not provided) in corn oil by oral gavage 7 once per week for 52 weeks (0, 357, and 714 ng/kg-day adjusted for continuous exposure). At 8 ages 31 to 39 weeks, 41 male mice and 32 female mice in the 2,500 ng/kg dose group were 9 mistakenly administered a single dose of 25,000 ng/kg TCDD. TCDD treatment for the 10 2,500 ng/kg dose group was halted for 5 weeks (beginning the week after the 25,000 ng/kg dose 11 was administered in error) and resumed until exposure was terminated at 57 weeks. Mortality 12 was observed and body weights recorded at unspecified intervals until 110 weeks of age, when 13 all surviving animals were sacrificed and necropsied. Histopathological analysis was conducted 14 on the following organs and tissues: Harderian glands, pituitary, thyroid, adrenals, tongue, 15 esophagus, and trachea; lungs, liver, pancreas; spleen, kidneys, and bladder; testes, ovaries, and 16 uterus, mesenteric lymph nodes, small intestine, and all other organs with presumed pathological 17 changes. 18 Body weights of both male and female mice exposed to 2,500 and 5,000 ng/kg TCDD 19 were markedly lower than in the corresponding control groups (statistical significance not 20 reported). Relative to the controls, a significant (p < 0.001), dose-related decrease in survival 21 occurred in animals treated with either dose of TCDD. In the subset of animals treated 22 inadvertently with a single dose of 25,000 ng/kg TCDD, mortality in male mice increased shortly 23 after this treatment; females, however, did not show a mortality increase following the 24 inadvertent treatment. This mortality in male mice was associated with subcutaneous edema, 25 degenerative hepatocyte changes, and bile duct hyperplasia. The incidence of non-neoplastic 26 lesions (such as amyloidosis of the liver, spleen, adrenals, and pancreas), liver necrosis, and 27 nephrosclerosis, was increased in mice exposed to TCDD compared to controls (statistical 28 significance not reported). 29 The study authors used two statistical tests to analyze tumor incidence. Because of the 30 increased mortality in treated groups compared to controls, one test, which assumes all tumors 31 are fatal, overestimated the differences between the treated and control groups. The second test This document is a draftfor review purposes only and does not constitute Agency policy. 2-204 DRAFT--DO NOT CITE OR QUOTE 1 assumes that all tumors are incidental and resulted in an underestimation of TCDD effects. Both 2 tests were used to analyze the results for nonthymic lymphomas and hepatic adenomas and 3 carcinomas. Incidence of nonthymic lymphomas (6/45, 4/51, and 3/50 in the 0, 2,500, and 4 5.000 ng/kg dose groups, respectively in males and 17/49, 21/42, and 17/48 in the 0, 2,500, and 5 5.000 ng/kg dose groups, respectively in females) was significantly (p < 0.05 in males and 6 p < 0.01 in females) higher in TCDD-treated animals compared to the corresponding controls 7 using the fatal tumor test. However, the incidental tumor test showed that this higher incidence 8 was not significant. Similarly, a significantly (p < 0.001) higher incidence of hepatocellular 9 adenomas occurred in male mice using the fatal tumor test (10/43, 11/51, and 10/50 in the 0, 10 2.500, and 5,000 ng/kg dose groups, respectively), but the incidence was not significant when 11 assessed using the incidental tumor test. Hepatocellular carcinomas in males were significant 12 (p < 0.001) using either the fatal or incidental tumor tests (5/43, 15/51, and 33/50 in the 0, 2,500, 13 and 5,000 ng/kg dose groups, respectively). In female mice, hepatocellular adenomas were 14 significant using both the fatal (p < 0.01) and incidental (p < 0.001) tumor tests (2/49, 4/42, and 15 11/48 in the 0, 2,500, and 5,000 ng/kg dose groups, respectively). Similar results for female 16 mice were obtained for incidence of hepatocellular carcinomas (1/49, 12/42, and 9/48 in the 0, 17 2.500, and 5,000 ng/kg dose groups, respectively), which also were significant using both the 18 fatal (p < 0.01) and incidental (p < 0.05) tumor tests. TCDD-related incidences of other tumor 19 types in both sexes were uniformly low and comparable in the treatment and control groups. 20 These results indicate that TCDD is carcinogenic in male and female B6C3F1 mice, 21 causing hepatocellular adenomas and carcinomas in both sexes. 22 In addition to the long term bioassay results in mice described by Della Porta et al. (1987, 23 197405), carcinogenic effects of TCDD in a neonatal bioassay were reported in the same 24 publication. Briefly, groups of male and female B6C3F1 and B6CF1 (C57/BL6J x BALB/c) 25 mice were treated with 0, 1000, 30,000 or 60,000 ng/kg BW TCDD via intraperitoneal (i.p.) 26 injection beginning at postnatal day 10. Animals were treated once weekly for 5 weeks and then 27 observed until 78 weeks of age. However, because this study utilized i.p. injection as the route 28 of TCDD exposure, it does not qualify for further consideration based on the study selection 29 criterion that the study design consist of orally administered TCDD. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-205 DRAFT--DO NOT CITE OR QUOTE 1 2.4.2.6.2. K o cib a e t al. (1978, 001818). 2 As discussed above, Kociba et al. (1978, 001818) conducted a lifetime (2-year) feeding 3 study of male and female Sprague-Dawley rats using doses of 0, 1, 10, and 100 ng/kg-day. 4 There were 50 males and 50 females in each group. 5 With respect to the cancer endpoints examined, the most significant finding was an 6 increase in hepatocellular hyperplastic nodules and hepatocellular carcinomas in female rats. 7 The incidence of hepatocellular carcinomas was significantly elevated above the control 8 incidence at the 100 ng/kg-day dose, whereas increased incidence of hyperplastic nodules was 9 evident in the 10 ng/kg-day dose group. 10 There have been two reevaluations of slides of liver sections from the Kociba et al. study 11 (Goodman and Sauer, 1992, 197667; Sauer, 1990, 198829; Squire, 1990, 548781). The Squire 12 Review was requested by EPA as an independent review of the slides. The Sauer Review was 13 carried out using refined criteria for the diagnosis of proliferative hepatocellular lesions 14 (Maronpot et al., 1986, 013967; Maronpot et al., 1989, 548778). Liver tumor incidences for the 15 three evaluations are compared in Appendix F. Although there are some quantitative differences 16 between the evaluations, the lowest detectable effect for liver tumor incidence is consistently 17 observed at 10 ng/kg-day. 18 In the 10 ng/kg-day dose group, significant increases in the incidence of hyperplastic 19 nodules of the liver were observed in female rats (18/50 in the Kociba evaluation, 27/50 in the 20 Squire evaluation). Two females (2/50) had hepatocellular carcinomas. In the 1990 reevaluation 21 (Goodman and Sauer, 1992, 197667; Sauer, 1990, 198829), nine females (9/50) were identified 22 with hepatocellular adenomas and none with carcinomas; thus only one-third of the previously 23 observed "tumors" were identified when using the refined diagnostic criteria. As discussed 24 below, the tumor reclassification of Goodman and Sauer (1992, 197667) was used in the 25 dose-response modeling for the Kociba et al. (1978, 001818) data set. 26 In addition to nodules in the liver, increased incidence of stratified squamous cell 27 carcinoma of the tongue and nasal turbinates/hard palate, and keratinizing squamous cell 28 carcinoma of the lung were also observed in female rats in the 100 ng/kg-day dose group. 29 One possible cause for the induction of lung tumors in the Kociba feeding study may have been 30 the aspiration of dosed feed into the lungs. However the promotion of lung tumors has been 31 observed in mice treated systemically by intraperitoneal (i.p.) injections of TCDD (Beebe et al., This document is a draftfor review purposes only and does not constitute Agency policy. 2-206 DRAFT--DO NOT CITE OR QUOTE 1 1995, 548754). In addition the induction of hyperplastic and metaplastic lesions in rats has been 2 observed following chronic oral gavage treatment with TCDD (Tritscher et al., 2000, 197265). 3 More recently, chronic oral exposure to HCDD resulted in the induction of lung tumors in treated 4 female rats (Rozman, 2000, 548758). These data indicate that the induction of lung tumors in 5 the Kociba was most likely primarily the result of systemic chronic dietary exposure to TCDD 6 rather than due to a localized exposure to aspired dosed feed. 7 There was no detectable increase in liver tumor incidences in male rats in any of the dose 8 groups. The mechanism responsible for dioxin-mediated sex specificity for 9 hepatocarcinogenesis in rats is not clear, but may involve ovarian hormones (Lucier et al., 1991, 10 199007). 11 Although there was no increase in liver tumors in male rats in this study, in the 12 100 ng/kg-day group, there was an increased incidence of stratified squamous cell carcinoma of 13 the hard palate/nasal turbinate, stratified squamous cell carcinoma of the tongue, and adenoma of 14 the adrenal cortex. 15 Kociba et al. (1978, 001818) had reported that chemically related increases in 16 preneoplastic or neoplastic lesions were not found in the 1 ng/kg-day dose group. However, 17 Squire identified two male rats in the 1 ng/kg-day dose group with squamous cell carcinoma of 18 the nasal turbinates/hard palate, and one of these male rats had a squamous cell carcinoma of the 19 tongue. These are both rare tumors in Sprague-Dawley rats, and these sites are targets for 20 TCDD, implying that 1 ng/kg-day may not represent a NOEL. However, no dose-response 21 relationships were evident for tumors at these sites (Huff et al., 1991, 197981) 22 There is considerable controversy concerning the possibility that TCDD-induced liver 23 tumors are a consequence of cytotoxicity. Goodman and Sauer (1992, 197667) have extended 24 the reevaluation of the Kociba slides to include liver toxicity data and have reported a correlation 25 between the presence of overt hepatotoxicity and the development of hepatocellular neoplasms in 26 female rats. With the exception of two tumors in controls and one each in the low- and mid-dose 27 groups, all liver tumors occurred in livers showing clear signs of toxicity. However, male rat 28 livers exhibit cytotoxicity in response to high TCDD doses, yet they do not develop liver tumors. 29 Moreover, both intact and ovariectomized female rats exhibit liver toxicity in response to TCDD, 30 yet TCDD is a more potent promoter in intact but not ovariectomized rats (Lucier et al., 1991, 31 199007). Therefore, if cytotoxicity is playing a role in liver tumorigenesis, other factors must This document is a draftfor review purposes only and does not constitute Agency policy. 2-207 DRAFT--DO NOT CITE OR QUOTE 1 also be involved. Also, there is little information on the role of cytotoxicity in TCDD-mediated 2 cancer at other sites such as the lung and thyroid. 3 4 2.4.2.6.3. Toth et al. (1 9 7 9 ,197109). 5 In a study of 10-week-old outbred male Swiss/H/Riop mice, Toth et al. (1979, 197109) 6 administered oral gavage TCDD doses of 0, 7, 700, and 7,000 ng/kg-day in sunflower oil weekly 7 for 1 year (0, 1, 100, or 1,000 ng/kg-day adjusted for continuous dosing; see details above). All 8 mice (100/group) were followed for their entire lives. The study authors identified the effective 9 number of mice in each group to be the number of surviving animals when the 10 first tumor-bearing animal was identified. The average lifespan of the control, low, mid and high 11 dose groups was 588, 649, 633, and 424 days, respectively. 12 In the 100 ng/kg-day dose group, liver tumor incidence was twice that of the control 13 group and was statistically significant (p < 0.01%). A dose-related increase in liver tumor 14 incidence was observed (18, 29, 48, and 30% in the control and three TCDD-treated groups, 15 respectively) in all treated mice. Increases were not statistically significant, however, at 1 and 16 1,000 ng/kg-day. The study authors also stated that spontaneous and induced liver tumors were 17 not histologically different. Additionally, the ratio of benign hepatomas to hepatocellular 18 carcinomas in the control group was not affected by treatment and an increase was observed only 19 in the absolute number of liver tumors. Cirrhosis was not observed with the tumors. 20 21 2.4.2.6.4. N T P (1982, 543764) . 22 As discussed above, the NTP (1982, 543764) study was conducted using 23 Osborne-Mendel rats and B6C3F1 mice (NTP, 1982, 543764). Groups of 50 male rats, 24 50 female rats, and 50 male mice received TCDD as a suspension in corn oil:acteone (9:1) by 25 gavage twice each week at doses of 0, 5, 25, or 250 ng/kg-day (daily averaged doses of 0, 1.4, 26 7.1, or 71 ng/kg-day for rats and male mice and doses of 0, 5.7, 28.6, or 286 ng/kg-day for 27 female mice. 28 There were no statistically significant dose-related decreases in survival in any 29 sex-species group. TCDD-induced malignant liver tumors occurred in the high-dose female rats 30 and in male and female mice. These can be considered to result from TCDD exposure because 31 they are relatively uncommon lesions in control Osborne-Mendel rats (male, 1/208; female, This document is a draftfor review purposes only and does not constitute Agency policy. 2-208 DRAFT--DO NOT CITE OR QUOTE 1 3/208), are seen in female rats and mice of both sexes, and their increasing incidence with 2 increasing dose is statistically significant (Cochran-Armitage trend test, p = 0.004). Because 3 liver tumors were increased in both sexes of mice, this effect is not female-specific as was 4 observed in rats. Interestingly, liver tumor incidences were decreased in female rats in both the 5 NTP and Kociba low doses (not statistically significant compared with controls). For example, 6 the combined control incidence data were 11/161 (7%) compared with 4/99 (4%) in the low-dose 7 group. 8 The incidences of thyroid gland (follicular cell) tumors were increased in all three dose 9 groups in male rats. Because the responses in the two highest dose groups are highly significant, 10 the statistically significant elevation of incidence in the lowest dose group (Fisher exact 11 p-value = 0.042) is considered to be caused by exposure to TCDD, suggesting that thyroid tumor 12 incidence may be the most sensitive site for TCDD-mediated carcinogenesis. Because 13 71 ng/kg-day is above the maximum tolerated dose (MTD) (Huff et al., 1991, 197981), thyroid 14 tumors occur at doses more than 50 times lower than the MTD. 15 TCDD-induced neoplasms of the adrenal gland were observed in the 7.1 ng/kg-day/dose 16 group in male rats and in high-dose female rats. Fibrosarcomas of the subcutaneous tissue were 17 significantly elevated in high-dose female mice and female rats. One additional tumor type, 18 lymphoma, was seen in high-dose female mice. Lung tumors were elevated in high-dose female 19 mice; the increase was not statistically significant when compared with concurrent controls, but 20 the increase was dose related (Cochran-Armitage trend test, p = 0.004). 21 Huff (1992, 548757) concluded, based on the NTP bioassay results, that TCDD was a 22 complete carcinogen and induced neoplasms in rats and mice of both sexes. As was observed in 23 the Kociba study (1978, 001818), liver tumors were observed with greater frequency in treated 24 female rats, but in male rats the thyroid appears to be the most sensitive (increased tumor 25 incidence at doses as low as 1.4 ng/kg-day). 26 27 2.4.2.6.5. N T P (2 0 0 6 ,197605). 28 As discussed above, female Sprague-Dawley rats (53 control; 53 or 54 animals per 29 treatment group) were administered TCDD (purity >98%) in corn oil:acetone (99:1) via gavage 30 at doses of 0, 3, 10, 22, 46, or 100 ng/kg-day, 5 days per week for 105 weeks (0, 2.14, 7.14, 15.7, 31 32.9, or 71.4 ng/kg-day, adjusted for continuous exposure) (NTP, 2006, 197605). In addition to This document is a draftfor review purposes only and does not constitute Agency policy. 2-209 DRAFT--DO NOT CITE OR QUOTE 1 this primary group, a stop-dose group of 50 animals was administered 100 ng/kg-day TCDD in 2 corn oil:acetone (99:1) via gavage for 30 weeks and then just the vehicle for the remainder of the 3 study. At study termination, the number of surviving animals had declined to 25 in the control 4 group and to 21, 23, 19, 22, and 21 in five treatment groups, respectively, due to accidental 5 deaths, moribund animals, or death due to natural causes. 6 Incidence of hepatocellular adenomas was significantly (p < 0.001) increased in the 7 100 ng/kg-day dose group in the primary study and exceeded incidences seen in historical 8 vehicle control range at study termination. A dose-related increase in the incidence of 9 cholangiosarcoma was seen in the primary study group in animals receiving 22 ng/kg-day or 10 higher doses of TCDD. The high dose group of 100 ng/kg-day had the highest incidence of 11 cholangiosarcoma with a significant (p < 0.001) number of animals exhibiting multiple 12 cholangiosarcomas. Such an incidence was not seen in historical vehicle controls. In contrast, 13 only two cholangiosarcomas and hepatocellular adenomas were seen in the 100 ng/kg-day group 14 in the stop-exposure study. 15 In the lung, at 2 years, there was a significantly (p = 0.002) increased incidence of cystic 16 keratinizing epithelioma in the 100 ng/kg-day dose group of the primary study, while there were 17 no epitheliomas in the 100 ng/kg-day group of the stop-exposure study. There was also a 18 significant (p < 0.01) dose-dependent increase, when compared to the vehicle control, in the 19 incidence of bronchiolar metaplasia of the alveolar epithelium at all dose groups in the primary 20 study. Squamous metaplasia was also present in the 46 and 100 ng/kg-day dose groups in the 21 primary study, and was also observed in the 100 ng/kg-day dose group in the stop-exposure 22 study. 23 A positive trend in the incidence of gingival squamous cell carcinoma of the oral cavity 24 was seen at all doses (except 22 ng/kg-day), with the incidence significantly (p = 0.007) high in 25 the 100 ng/kg-day dose group. In addition, the occurrence of this lesion in the 46 and 26 100 ng/kg-day group of the primary study and 100 ng/kg-day group of the stop-exposure study 27 exceeded the historical control range. The incidence of gingival squamous hyperplasia was 28 significantly (eitherp < 0.01 orp < 0.05) increased in all dose groups of the primary study as 29 well as the 100 ng/kg-day group of the stop-exposure study. 30 In the uterus, at 2 years, there was a significantly (p = 0.032) higher rate of squamous cell 31 carcinoma in the 46 ng/kg-day group compared to vehicle controls. In addition there were This document is a draftfor review purposes only and does not constitute Agency policy. 2-210 DRAFT--DO NOT CITE OR QUOTE 1 two squamous cell carcinomas in the 100 ng/kg-day group of the stop-exposure study. No 2 squamous cell carcinomas have been reported in historical vehicle controls. 3 These results indicate that TCDD is carcinogenic to female Sprague-Dawley rats and 4 causes tumors at multiple sites. 5 6 2.4.3. Summary of Key Data Set Selection for TCDD Dose-Response Modeling 7 To meet the NAS' concerns regarding transparency and clarity in the identification of 8 TCDD studies for dose-response assessment, EPA has, in this section, developed and applied 9 two sets of criteria for animal bioassays and epidemiologic studies. EPA has collected and 10 evaluated these studies, including studies from the 2003 Reassessment and newer studies found 11 via literature searches and through public submissions. Tables 2-4 and 2-5 contain the final lists 12 of key cancer and noncancer studies, respectively, that have met EPA's inclusion criteria for 13 epidemiologic data. Tables 2-6 and 2-7 provide the final lists of key studies that have met EPA's 14 inclusion criteria for animal bioassay data for cancer and noncancer studies, respectively. 15 Collectively, these four tables contain the final set of key studies that EPA has used to develop 16 noncancer and cancer dose-response assessments for TCDD in Sections 4 and 5 of this 17 document, respectively. In Sections 4 and 5, additional evaluations are made to determine which 18 study/endpoint data sets are the most appropriate for development of the RfD and OSF for 19 TCDD, using statistical criteria, dose-response modeling results and decisions regarding 20 toxicological relevance of the endpoints. The approaches taken to select the final candidate 21 study/endpoint data sets are discussed in Sections 4 and 5 and are illustrated in Figures 4-1, 4-2 22 and and 5-3 of those sections. This document is a draftfor review purposes only and does not constitute Agency policy. 2-211 DRAFT--DO NOT CITE OR QUOTE 1 Table 2-1. Summary of epidemiological cancer studies (key characteristics) 2 Publication Length of follow-up Latency period Half-life for TCDD Fraction of TEQs accounted for by TCDD NIOSH cohort studies Fingerhut et al. 1942-1987 0, 20 years (1991,197375) N/A N/A Steenland et al. 1942-1993 0, 15 years (1999, 197437) N/A N/A Steenland et al. 1942-1993 0, 15 years (2001,197433) 8.7 years (Michalek et al., 1996, 198893) TCDD accounted for all occupational TEQ; 10% of background Cheng et al. 1942-1993 0, 10, 15 years 8.7 years (Michalek et al., N/A (2006, 523122) 1996, 198893), and CADM (Aylward et al., 2005, 197114) Collins et al. 1942-2003 None (2009,197627) 7.2 years (Flesch-Janys et al., N/A 1996,197351) BASF cohort studies Thiess et al. 1953-1980 None (1982, 064999) N/A N/A Zober et al. 1953-1987 Years since first N/A (1990, 197604) exposure: 0-9, 10-19, and 20+ N/A Ott and Sober 1953-1991 None (1996,198101) 5.8 years N/A Hamburg cohort studies Manz et al. 1952-1989 (1991,199061) None, used N/A duration of employment (<20, >20 years) N/A Flesch-Janys et 1952-1992 None al. (1995, 197261) 7.2 years Flesch-Janys et al. (1994, 197372) Mean TEQ without TCDD was 155 ng/kg; mean TEQ with TCDD was 296.5 ng/kg Flesch-Janys et 1952-1992 None al. (1998, 197339) 7.2 years Flesch-Janys et al. Mean concentration of (1996, 197351), also used TCDD was 101.3 ng/kg; for decay rates that were function TEQ (without TCDD) mean of age and fat composition exposure was 89.3 ng/kg Becher et al. 1952-1992 (1998,197173) 0, 5, 10, 15 and 20 years 7.2 years Flesch-Janys et al. (1996,197351) took into account age and fat composition Not described This document is a draftfor review purposes only and does not constitute Agency policy. 2-212 DRAFT--DO NOT CITE OR QUOTE 1 Table 2-1. Summary of epidemiological cancer studies (key characteristics) (continued) Publication Length of follow-up Latency period Half-life for TCDD Fraction of TEQs accounted for byTCDD Seveso cohort studies Bertazzi et al. (2001, 197005) 1976-1996 Periods N/A postexposure: 0, 0-4, 5-9, 10-14, 15-19 years N/A Warner et al. (2002, 197489) 1976-1998 None 8 years (Pirkle et al., N/A 1989,197861) Pesatori et al. (2003, 197001) 1976-1996 Period postexposure: 20 years N/A N/A Baccarelli et al. (2006, 1976-1998 Period 197036) postexposure: 22 years N/A N/A Consonni et al. (2008, 524825) 1976-2001 Periods N/A postexposure: 0, 0-4, 5-9, 10-14, 15-19, 20-24 years N/A Chapaevsk cohort studies Revich et al. (2001, 199843) Cross N/A sectional study (1995-1998) N/A N/A Ranch Hand cohort studies Akhtar et al. (2004, 197141) 1962-1999 None N/A N/A Michalek and Pavuk (2008,199573) 1962-2004 None, but 7.6 years stratified by period of service N/A This document is a draftfor review purposes only and does not constitute Agency policy. 2-213 DRAFT--DO NOT CITE OR QUOTE Table 2-1. Summary of epidemiological cancer studies (key characteristics) (continued) Publication Length of follow-up Latency period Half-life for TCDD Fraction of TEQs accounted for byTCDD New Zealand cohort studies t'Mannetje et al. (2005, 1969-2000 197593) (herbicide producers); 1973-2000 (herbicide sprayers) N/A N/A N/A McBride (2009, 198490) 1969-2004 None N/A N/A McBride et al. (2009, 1969-2004 None 197296) 7 years N/A Dutch cohort study Hoooiveld et al. (1998, 1955-1991 197829) Periods 7.1 years postexposure: 0-19 years, >19 years N/A This document is a draftfor review purposes only and does not constitute Agency policy. 2-214 DRAFT--DO NOT CITE OR QUOTE Table 2-2. Epidemiological cancer study selection considerations and criteria This document is a draftfor review purposes only and does not constitute Agency policy. 2-215 Cancer Exposure assessment methodology Effective dose Risk clear and & oral Methods estimates are adequately Study size exposure use to not characterizes and follow Published estimable & ascertain susceptible Association individual- up large in peer- Exposure consistent w/ health to biases between level enough to reviewed primarily current outcomes from TCDD and exposures. yield precise literature TCDD and biological were confounding adverse Limitations estimates of with quantified so understanding. unbiased, exposures or health effect, and risk and appropriate that dose- Latency and highly from study with uncertainties ensure discussion response appropriate Pass for sensitive design or exposure- in exposure adequate of relationship window(s) of dose- and statistical response assessment statistical strengths, can be exposure response specific. analysis. relationship. considered. power. limitations. assessed. examined. analyses? Considerations Criteria Y/N NIOSH Cohort Studies Fingerhut et al. (1991, 197375) all cancer sites, site-specific analyses VXX X V VX VN Steenland et al. (1999, 197437) all cancer sites combined, site-specific analyses V V V V V VV V Na Steenland et al. (2001, 197433) all cancer sites combined VV V V V VV VY Cheng et al. (2006, 523122) all cancer sites combined VV V V V VV VY Collins et al. (2009, 197627) all cancer sites combined, site-specific analyses V V V V V VV VY BASF Cohort Studies Thiess et al. (1982, 064999) all cancer sites combined, site-specific analyses V X X XX VX XN Zober et al. (1990, 197604) all cancer sites combined, site-specific analyses V V X XX VX XN DRAFT: DO NOT CITE OR QUOTE Table 2-2. Epidemiological cancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-216 Cancer Ott and Zober (1996, 198101) all cancer sites combined Hamburg Cohort Manz et al. (1991, 199061) all cancer sites combines, site-specific analyses Flesh-Janys et al. (2006, 197621) all cancer sites combined Flesh-Janys et al. (1998, 197339) all cancer sites combined, site-specific analyses Becher et al. (1998, 197173) all cancer sites combined Seveso Cohort Bertazzi et al. (2001, 197005) all cancer sites combined, site-specific analyses Pesatori et al. (2003, 197001) all cancer sites combined, site-specific analyses Methods use to ascertain health outcomes were unbiased, highly sensitive and specific. V V V V V V V Exposure assessment methodology Risk clear and estimates are adequately Study size not characterizes and follow susceptible Association individual- up large to biases between level enough to from TCDD and exposures. yield precise confounding adverse Limitations estimates of exposures or health effect, and risk and from study with uncertainties ensure design or exposure- in exposure adequate statistical response assessment statistical analysis. relationship. considered. power. Considerations VV V V VV V V VV V V VV V V VV V V VV XV VXX V Effective dose & oral exposure Published estimable & in peer- Exposure consistent w/ reviewed primarily current literature TCDD and biological with quantified so understanding. appropriate that dose- Latency and discussion response appropriate Pass for of relationship window(s) of dose- strengths, can be exposure response limitations. assessed. examined. analyses? Criteria Y/N VV VY VX VV VV VV VN XN V Nb VY VX VX XN XN DRAFT: DO NOT CITE OR QUOTE Table 2-2. Epidemiological cancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-217 Exposure assessment methodology Effective dose Risk clear and & oral Methods estimates are adequately Study size exposure use to not characterizes and follow Published estimable & ascertain susceptible Association individual- up large in peer- Exposure consistent w/ health to biases between level enough to reviewed primarily current outcomes from TCDD and exposures. yield precise literature TCDD and biological were confounding adverse Limitations estimates of with quantified so understanding. unbiased, exposures or health effect, and risk and appropriate that dose- Latency and highly from study with uncertainties ensure discussion response appropriate Pass for sensitive design or exposure- in exposure adequate of relationship window(s) of dose- and statistical response assessment statistical strengths, can be exposure response specific. analysis. relationship. considered. power. limitations. assessed. examined. analyses? Cancer Considerations Criteria Y/N Consonni et al. (2008, 524825) all cancer sites combined, site-specific analyses V V V X V VX XN Seveso Cohort-Women's Health Study Baccarelli et al. (2006, 197036) site specific analysis VVX V V VV V Nc Warner et al. (2002, 197489) breast cancer incidence VV V V V VV VY Chapaevsk Study Revich et al. (2001, 199843) all cancer sites combined, site-specific analyses X X X X V XX XN Ranch Hands Cohort Akhtar et al. (2004, 197141) all cancer sites combined, site-specific analyses V X V V V VX VN Michalek and Pavuk (2008, 199573) all cancer sites combined VXV V V VX V N DRAFT: DO NOT CITE OR QUOTE Table 2-2. Epidemiological cancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Cancer Others Hooiveld et al. (1998, 197829) all cancer sites combined, site-specific analyses t 'Mannetie et al. (2005, 197593) all cancer sites combined, site-specific analyses McBride et al. (2009, 197296) all cancer sites combined, site-specific analyses McBride et al. (2009, 198490) all cancer sites combined, site-specific analyses Methods use to ascertain health outcomes were unbiased, highly sensitive and specific. V V V V Exposure assessment methodology Risk clear and estimates are adequately Study size not characterizes and follow susceptible Association individual- up large to biases between level enough to from TCDD and exposures. yield precise confounding adverse Limitations estimates of exposures or health effect, and risk and from study with uncertainties ensure design or exposure- in exposure adequate statistical response assessment statistical analysis. relationship. considered. power. Considerations VV VX XV V V XX V X VXV X Effective dose & oral exposure Published estimable & in peer- Exposure consistent w/ reviewed primarily current literature TCDD and biological with quantified so understanding. appropriate that dose- Latency and discussion response appropriate Pass for of relationship window(s) of dose- strengths, can be exposure response limitations. assessed. examined. analyses? Criteria Y/N VV XX VX VV XN XN XN V Nd 2-218 DRAFT: DO NOT CITE OR QUOTE aThis study has been superseded and updated by Steenland et al. (2001, 197433). bBecher et al. (1998, 197173)) assessed this same cohort taking cancer latency into account, thereby superseding this study. cIt is unknown whether the frequency of t(14;18)translocations in lymphocytes relates specifically to an increased risk of non-Hodgkin's lymphoma. Given this lack of obvious adverse effect, dose-response analyses for this outcome were not conducted. dNo dose-response associations were noted. V= Consideration/criteria satisfied; X= Consideration/criteria not satisfied. Table 2-3. Epidemiological noncancer study selection considerations and criteria This document is a draftfor review purposes only and does not constitute Agency policy. 2-219 Noncancer NIOSH Cohort Steenland et al. (1999, 197437) mortality (noncancer) -ischemic heart disease Collins et al. (2009, 197627) mortality (noncancer) BASF Cohort Ott and Zober (1996, 198101) mortality (noncancer) Hamburg Cohort Flesch-Janys et al. (1995, 197261) mortality (noncancer) Seveso Cohort-Women's Health Study Eskenazi et al. (2002, 197168) menstrual cycle characteristics Eskenazi et al. (2002, 197164) endometriosis Eskenazi et al. (2003, 197158) birth outcomes Exposure assessment methodology Study size Risk clear and and Methods estimates adequately follow-up use to are not characterizes large ascertain susceptible Association health to biases between outcomes from TCDD and were confounding adverse unbiased, exposures health effect, individuallevel exposures. Limitations and enough to yield precise estimates of risk and highly or from with uncertainties in ensure sensitive study design exposure- exposure adequate and or statistical response assessment statistical specific. analysis. relationship. considered. power. Considerations Effective dose & oral exposure estimable & consistent w/ current Published in Exposure biological peer- primarily understanding. reviewed TCDD and Latency and literature quantified so appropriate with that dose- window(s) of appropriate response exposure Pass for discussion of relationships examined for a dose- strengths, can be Nonfatal response limitations. assessed. endpoint. analyses? Criteria Y/N VX VV V X VV VV VX VV XN XN VV X VV VV XN VV V VV VV XN VV XX XX V X X VV VV VX VV VV VV VY XN XN DRAFT: DO NOT CITE OR QUOTE Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-220 Noncancer Warner et al. (2004, 197490) age at menarche Eskenazi et al. (2005, 197166) age at menopause Warner et al. (2007, 197486) ovarian function Eskenazi et al. (2007, 197170) uterine leiomyoma Seveso Cohort-Other Studies Bertazzi et al. (2001, 197005) mortality (noncancer) Consonni et al. (2008, 524825) mortality (noncancer) Mocarelli et al. (2000, 197448) sex ratio Exposure assessment methodology Study size Risk clear and and Methods estimates adequately follow-up use to are not characterizes large ascertain susceptible Association individual- enough to health to biases between level yield outcomes from TCDD and exposures. precise were confounding adverse Limitations estimates unbiased, exposures health effect, and of risk and highly or from with uncertainties in ensure sensitive study design exposure- exposure adequate and or statistical response assessment statistical specific. analysis. relationship. considered. power. Considerations Effective dose & oral exposure estimable & consistent w/ current Published in Exposure biological peer- primarily understanding. reviewed TCDD and Latency and literature quantified so appropriate with that dose- window(s) of appropriate response exposure Pass for discussion of relationships examined for a dose- strengths, can be Nonfatal response limitations. assessed. endpoint. analyses? Criteria Y/N VV X VV VV XN VV X VV VV XN VV X VV VV XN VV V VV VV X Na VV VV VV X X V XV XV VV VX VX XV XN XN X Nb DRAFT: DO NOT CITE OR QUOTE Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-221 Baccarelli et al. (2002, 197062; 2004, 197045) immunological effects Landi et al. (2003, 198362) gene expression Alaluusua et al. (2004, 197142) oral hygiene Baccarelli et al. (2005, 197053) chloracne Baccarelli et al. (2008, 197059) neonatal thyroid function Mocarelli et al. (2008, 199595) semen quality Chapaevsk Study Revich et al. (2001, 199843) mortality (noncancer) and reproductive health Ranch Hands Cohort Michalek and Pavuk (2008, 199573) diabetes Exposure assessment methodology Study size Risk clear and and Methods estimates adequately follow-up use to are not characterizes large ascertain susceptible Association individual- enough to health to biases between level yield outcomes from TCDD and exposures. precise were confounding adverse Limitations estimates unbiased, exposures health effect, and of risk and highly or from with uncertainties in ensure sensitive study design exposure- exposure adequate and or statistical response assessment statistical specific. analysis. relationship. considered. power. VV VV V X VV VX VV V VV VV V VV VV V XV VV V VV VX X XV VX V VV Effective dose & oral exposure estimable & consistent w/ current Published in Exposure biological peer- primarily understanding. reviewed TCDD and Latency and literature quantified so appropriate with that dose- window(s) of appropriate response exposure Pass for discussion of relationships examined for a dose- strengths, can be Nonfatal response limitations. assessed. endpoint. analyses? VV VX XN XN VV VY VV V Nc VV VY VV VY VX XN VX VN DRAFT: DO NOT CITE OR QUOTE Table 2-3. Epidemiological noncancer study selection considerations and criteria (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Other Ryan et al. (2002, 198508) sex ratio Kane et al. (2006, 199133) long-term health consequences McBride et al. (2009, 198490) mortality (noncancer) McBride et al. (2009, 197296) mortality (noncancer) Exposure assessment methodology Study size Risk clear and and Methods estimates adequately follow-up use to are not characterizes large ascertain susceptible Association individual- enough to health to biases between level yield outcomes from TCDD and exposures. precise were confounding adverse Limitations estimates unbiased, exposures health effect, and of risk and highly or from with uncertainties in ensure sensitive study design exposure- exposure adequate and or statistical response assessment statistical specific. analysis. relationship. considered. power. Effective dose & oral exposure estimable & consistent w/ current Published in Exposure biological peer- primarily understanding. reviewed TCDD and Latency and literature quantified so appropriate with that dose- window(s) of appropriate response exposure Pass for discussion of relationships examined for a dose- strengths, can be Nonfatal response limitations. assessed. endpoint. analyses? XX XX XX XV X X X X XV VV VX VX VX VX VV VX XN XN XN XN 2-222 DRAFT: DO NOT CITE OR QUOTE "Categorical measures of TCDD suggest an inverse association between TCDD exposure and uterine fibroids. The observed direction of the reported associations precluded quantitative dose-response modeling. bThe somewhat arbitrary cut off age of 19 for statistically significant exposure associations results in a highly uncertain critical exposure window. It is difficult to determine whether effects are a consequence of the initial high exposure during childhood or a function of the cumulative exposure for this entire exposure window. The differences between these two dose estimates are quite large. cChloracne is recognized to occur following high TCDD exposure levels. This study provides limited relevance to TCDD RfD development, as exposure levels observed in the general population are much lower. V = Consideration/criteria satisfied. X = Consideration/criteria not satisfied. Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling This document is a draftfor review purposes only and does not constitute Agency policy. Health outcome Mortality from all cancers Mortality from all cancers Location, time period USA, 1942-1993 Cohort description NIOSH cohort including 3,538 occupationally exposed male workers at 8 plants in the United States; 256 cancer deaths Exposure assessment Cumulative serum lipid TCDD concentrations (CSLC) based on work histories, jobexposure matrix, and concentration and agedependent twocompartment model of elimination kinetics Exposure measures No exposure categories provided USA, 1942-1993 NIOSH cohort including 3,538 male workers, 256 cancer deaths CSLC based on work histories, job-exposure matrix, and a simple onecompartment first-order pharmacokineti c elimination model with 8.7year half-life CSLC (ppt-years) <335 335-520 520-1,212 1,212-2,896 2,896-7,568 7,568-20,455 >20,455 No. of cases/ deaths 256 cancer deaths 64 29 22 30 31 32 48 Effect Measure/ RR (95% CI) Risk factors Comments Reference The slope (P) was Available: age, Confounding by Cheng et al. 3.3 x 10-6 for lag year of birth, and smoking was (2006, of 15 years race considered indirectly 523122) excluding upper by analysis of 5% of TCDD Risks adjusted for: smoking-related and exposures. year of birth, age, smoking-unrelated The slopes ranged and race cancers. two orders of Other occupational magnitude exposures were depending on considered indirectly modeling by repeated analyses assumption removing one plant at a time. Based on indirect evaluation, there was no clear evidence of confounding. 1.00 1.26 (0.79-2.00) 1.02 (0.62-1.65) 1.43 (0.91-2.25) 1.46 (0.93-2.30) 1.82 (1.18-2,82) 1.62 (1.03-2,56) Available: date of birth and age Adjusted for: date of birth, and age was used as time scale in Cox model Included in U.S. EPA (2003, 537122) Steenland et al. (2001, 197433) 2-223 DRAFT: DO NOT CITE OR QUOTE Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. Health outcome Mortality from all cancers combined Location, time period Cohort description Exposure assessment Exposure measures Hamburg, Boehringer Cumulative Categorical Germany, cohort including TCDD serum exposures production approximately lipid (Cox model) period was 1,189 workers concentrations 0 - <1 1950-1984 employed in the based on area 1- <4 and production of under curve (in 4 - <8 mortality herbicides pg/kg years); 8 - <16 follow-up back- 16- <64 extended extrapolation to 64+ through date of last 1992 employment took into Continuous account age and exposure percent body TCDD (pg/kg fat; half-life years) value was 7.2 years No. of cases/ deaths 124 124 Effect Measure/ RR (95% CI) 1.0 1.12 (0.70-1.80) 1.42 (0.70-2.85) 1.77 (0.81-3.86) 1.63 (0.73-3.64) 2.19 (0.76-6.29) P = 0.0089, p = 0.0047 Risk factors Comments Reference Included in Becher et U.S. EPA (2003, al. (1998, 537122) 197173) Available: year of A large number of entry, age of models were fitted. entry, duration of These included employment, birth models for 5 cohort, P-HCH; different latency TEQ other than intervals (0, 5, 10, TCDD 15, and 20 years), as well as Available: year of multiplicative, entry, age of additive and power entry, duration of models, and employment, birth different offset cohort, P-HCH; variables (person TEQ other than years and expected TCDD deaths) 2-224 DRAFT: DO NOT CITE OR QUOTE Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Health outcome Location, time period Mortality Ludwig- and shafen, incidence Germany, for all 1954-1992 cancers combined, as well as for specific cancer sites Cohort description BASF cohort, 243 men exposed from accidental release that occurred in 1953 during production of trichlorophenol, or who were involved in clean-up activities Exposure assessment Exposure measures Cumulative Internal TCDD serum comparisons lipid based on concentrations continuous expressed in measure of pg/kg based on TCDD. TCDD half-life of 5.1-8.9 years, Cox regression model External comparisons exposure categories: <0.1, 0.1-0.99, 1.0-1.99 >2 pg/kg No. of cases/ Effect Measure/ deaths RR (95% CI) Risk factors Internal Available: age, cohort BMI, smoking analysis status and history 31 Date of 1st TCDD exposure of occupational exposure to cancer 1.22 (95% CI: deaths 1.00-1.50) amines and asbestos 47 incident 1.11 (95% CI: cancers 0.91-1.35) External cohort analyses Deaths 8 8 8 7 SMRs 0.8 (0.4-1.6) 1.2 (0.5-2.3) 1.4 (0.6-2.7) 2.0 (0.8-4.0) Comments Reference Included in Ott and U.S. EPA (2003, Zober 537122) (1996, 198101) Positive associations noted for digestive cancer, but not for respiratory cancer Associated between TCDD and increased SMRs found only among current smokers Last published account of this cohort 2-225 DRAFT: DO NOT CITE OR QUOTE DRAFT: DO NOT CITE OR QUOTE 2-226 This document is a draftfor review purposes only and does not constitute Agency policy. Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued) Health outcome Breast cancer incidence Location, time period Cohort description Italy 1976-1998 981 women from zones A and B with available archive serum samples, 15 breast cancer cases Exposure assessment TCDD serum lipid concentrations (ppt) collected between 1976 and 1981. For most samples collected after 1977, serum TCDD levels were backextrapolated using a first order kinetic model with a 9year half-life. Exposure measures No. of cases/ deaths Cases <20 ppt 1 20.1- 44 ppt 2 44.1- 100 ppt7 >100 ppt 5 Logi0TCDD also modeled as continuous variable 15 Effect Measure/ RR (95% CI) 1.0 1.0 (0.1-10.8) 4.5 (0.6-36.8) 3.3 (0.4-28.0) 2.1 (1.0-4.6) Risk factors Available: gravidity, parity, age at first pregnancy, age at last pregnancy, lactation, family history of breast cancer, age at menarche, current body mass index, oral contraceptive use, menarcheal status at explosion, menopause status at diagnosis, height, smoking, alcohol consumption. Comments Included in U.S. EPA (2003, 537122) Adjusted for age, which was used as time scale in Cox model; other covariates were evaluated but were not identified as confounders. Reference Warner et al. (2002, 197489) Table 2-4. Epidemiological studies selected for TCDD cancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Health outcome Mortality from all cancers and specific cancer types Location, time period Cohort description Midland, Subset of Michigan, NIOSH cohort USA. including 1,615 Follow-up occupationally period: exposed male 1942-2003. workers at 1 Serum plant in the collection United States; period: 177 cancer 2004-2005 deaths Exposure assessment Exposure measures Cumulative serum lipid Part per billion-year TCDD estimates of concentrations cumulative based on work TCDD histories, job exposure exposure matrix, and concentration and age- dependent two- compartment model of elimination kinetics. Serum samples were obtained from 280 former workers collected during 2004-2005. No. of cases/ Effect Measure/ deaths RR (95% CI) Risk factors Comments Reference 177 cancer deaths The slope of a Hazard ratios Confounding by Collins et proportional hazards adjused for age, smoking was not al. (2009, year of birth, and considered directly 197627) regression model hire year. due to a lack of data. for fatal soft Stratified analyses Relatively long tissue sarcoma used to examine follow-up period was 0.05872 potential impact (average = 36 (95% CI not of years). provided but for pentachlorophenol Potential outcome Chi-square exposure on misclassification for p = 0.0060) for mortality. soft tissue sarcoma every 1-part per due to potential billion-year inaccuracies on increase in death certificates. cumulative Data analyzed from exposure of one plant reduces TCDD. Slope heterogeneity estimates for all associated with fatal cancers, fatal multiplant analyses. lung, fatal More serum samples prostate, fatal (n = 280) analyzed leukemias and than used to derive fatal non-Hodgkin TCDD estimates for lymphomas were other NIOSH cohort not statistically analyses. significant 2-227 DRAFT: DO NOT CITE OR QUOTE Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling This document is a draftfor review purposes only and does not constitute Agency policy. Health outcome Location, time period b-TSH Italy, 1976; measured 72 children, hours after 1994-2005 birth from a heel pick (routine screening for all newborns in the region Cohort description Exposure assessment No. of Exposure cases/ measures deaths Population- Based on zone Population- based study: of residence, based study: 1,041 estimated mean singletons values from a (56 from previous study. Reference 533 zone A, 425 Maternal births from zone B plasma TCDD and 533 from levels estimated Zone B reference) at the date of born between Jan. 1, 1994- delivery using a first-order pharmacokineti Zone A June 30, c model and 2005. elimination rate Plasma estimated in dioxin study: Seveso women Plasma 51 children born to 38 women of fertile age who were part of the (half-life =9.8 years). dioxin study: Continuous maternal plasma Seveso TCDD 425 births 56 births Chloracne Study. Effect Measure/ RR (95% CI) Risk factors Population-based Available: gender, study birth weight, birth Mean b-TSH order, maternal age at delivery, Reference: hospital, type of 0.98 (95% CI: delivery. 0.90-1.08) Zone B: There was limited 1.66 (95% CI: evidence of 1.19-2.31) confounding, so Zone A: mean TSH results 1.35 (95% CI: presented here are 1.22-1.49) unadjusted. Association between neonatal b-TSH with plasma TCDD: adjusted p = 0.75 (p < 0.001) Comments An association with serum TCDD levels of mothers was found with b-TSH among the 51 births in the plasma dioxin study. Reference Baccarelli et al. (2008, 197059) 2-228 DRAFT: DO NOT CITE OR QUOTE Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-229 Health outcome Sperm conc. (million/mL) Progressive motility (%) Serum E2 (pmol/L) Location, Cohort time period description Exposure assessment Exposure measures Italy, 1976, 1998 135 exposed Serum TCDD TCDD (from zone (in ppt) from quartiles A) and 184 1976-1977 nonexposed samples (for men aged exposed men); 1-26 in 1976 background were values were included. assumed for These unexposed men subjects were based on serum selected from analysis of the cohort of residents in 257 exposed uncontaminated and 372 areas. unexposed people. No. of cases/ deaths Effect Measure/ RR (95% CI) Risk factors Comments Mean values were compared between the exposed and comparison groups for sperm concentration, volume, motility and count, FSH, E2, LH, and Inhibin B. Available: age, abstinence time, smoking status, education, alcohol use, maternal smoking during pregnancy, employment status, BMI, chronic exposure to solvents and other toxic substances. Results stratified by timing of exposure (1-9 yrs old vs. 10-17 yrs old in 1976). Reference Mocarelli et al. (2008, 199595) Adjusted for smoking status, organic solvents, age at time of tests, BMI, alcohol use, education, employment status and abstinence (days) for sperm data. Hormone data not adjusted for education level, employment status, and abstinence time. DRAFT: DO NOT CITE OR QUOTE Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Health outcome Dental defects Location, Cohort Exposure time period description assessment Seveso, Italy, 65 subjects Serum TCDD Dental <9.5 years (ng/kg) from exams old at time of 1976 samples administered Seveso for those who in 2001 explosion resided in Zone among those and residing ABR; no serum exposed to in zones levels for non TCDD in ABR; 130 ABR residents 1976 subjects (unexposed). recruited TCDD from the exposure non-ABR represent levels region as of 1976 (unexposed) (after accident) Exposure measures Non-ABR Zone 31-226 ng/kg 238-592 ng/kg 700-26000 ng/kg No. of cases/ deaths 10/39 1/10 5/11 9/15 <5 years of 25/75 age at time of accident Non-ABR Zone or 31-226 ng/kg serum TCDD 238-26,000 ng/kg serum TCDD Effect Measure/ RR (95% CI) Dental defect % 26% 10% 45% 60% Odds Ratios (among those <5 years of age at time of accident) 1.0 2.4 (1.3-4.5) Risk factors Comments Available: medical Dose-response history, age, sex, pattern observed education, smoking with dental defects in the ABR zone; however, the control population had a much higher prevalence of dental defects (26%) than those in the lowest exposure group (10%). Also assessed hypodontia and other dental and oral aberrations, but these were too rare to allow modeling by ABR zone. Reference Alaluusua et al. (2004, 197142) 2-230 DRAFT: DO NOT CITE OR QUOTE DRAFT: DO NOT CITE OR QUOTE 2-231 This document is a draftfor review purposes only and does not constitute Agency policy. Table 2-5. Epidemiological studies selected for TCDD noncancer dose-response modeling (continued) Health outcome Location, Cohort Exposure time period description assessment Menstrual Seveso, Italy, Women who Serum TCDD cycle follow-up were <40 (ng/kg) from characteristics: interview years from 1976 samples. menstrual cycle conducted in zones A or B TCDD length. 1996-1997 of in 1976, exposure level women A positive was back- exposed to association extrapolated to TCDD in the found among 1976 using the 1976 women who Filser or the accident were pre- first-order menarcheal kinetic models. at the time of accident (n = 134) Exposure measures Interquartile range was 64-322 ppt TCDD examined as continuous measure (per 10-fold increase in serum levels). No. of cases/ deaths Effect Measure/ RR (95% CI) Risk factors Lengthening of the menstrual cycle by 0.93 days (95% CI: 0.01, 1.86) Interview data: medical history, personal habits, work history, reproductive history, age, smoking, body mass index, alcohol and coffee consumption, exercise, illness, abdominal surgeries. Comments Reference Eskenazi et al. (2002, 197168) Table 2-6. Animal bioassays selected for cancer dose-response modeling This document is a draftfor review purposes only and does not constitute Agency policy. 2-232 Sex Average daily exposure dose levels Statistical significant tumors Species/strain route/duration n (ng/kg-day) Cancer types (pairwise with controls or trend tests) Reference Mouse/ B6C3F1 Male/Female Oral gavage once per week; 52 weeks Approximat 0, 351, and 714 ely 40 to 50 in each dose group including controls Females and males: hepatocellular adenomas and carcinomas Liver: adenomas and carcinomas in females and carcinomas in males (using incidental tumor statistical test) Della Porta et al. (1987, 197405) Rat/SpragueDawley Male/female Oral-lifetime feeding; 2 years 50 each (86 each in vehicle control group) 0, 1, 10, or 100 Females: liver, lung, oral cavity Males: adrenal, oral cavity, tongue Adrenal cortex: adenoma Liver: hepatocellular adenoma(s) or carcinoma(s); hyperplastic nodules Lung: keratinizing squamous cell carcinoma Oral cavity: stratified squamous cell carcinoma of hard palate or nasal turbinates Tongue: stratified squamous cell carcinoma Kociba et al. (1978, 001818); (Female liver tumors analysis updated in Goodman and Sauer, 1992, 197667) Mouse/ B6C3F1 Male/female Oral-gavage twice per week; 104 weeks 50 each (75 each in vehicle control group) 0, 1.4, 7.1, or 71 for males; 0, 5.7, 28.6, or 286 for females Females: hematopoietic system, liver, subcutaneous tissue, thyroid Males: liver, lung Hematopoietic system: lymphoma or leukemia Liver: hepatocellular adenoma or carcinoma Lung: alveolar/bronchiolar adenoma or carcinoma Subcutaneous tissue: fibrosarcoma Thyroid: follicular-cell adenoma NTP (1982, 543764) Rat/OsborneMendel Male/female Oral-gavage twice per week; 104 weeks 50 each (75 each in vehicle control group) 0, 1.4, 7.1, or 71 Females: adrenal, liver, subcutaneous tissue, thyroid Males: adrenal, liver, thyroid Adrenal: cortical adenoma, or carcinoma or NTP (1982, adenoma, NOS 543764) Liver: neoplastic nodule or hepatocellular carcinoma Subcutaneous tissue: fibrosarcoma Liver: neoplastic nodule or hepatocellular carcinoma Thyroid: follicular-cell adenoma or carcinoma DRAFT: DO NOT CITE OR QUOTE DRAFT: DO NOT CITE OR QUOTE 2-233 This document is a draftfor review purposes only and does not constitute Agency policy. Table 2-6. Animal bioassays selected for cancer dose-response modeling (continued) Species/strain Sex exposure route/duration n Average daily dose levels (ng/kg-day) Cancer types Statistical significant tumors (pairwise with controls or trend tests) Rat/Harlan SpragueDawley Female Oral-gavage 5 days per week; 2 years 53 or 54 0, 2.14, 7.14, 15.7, 32.9, or 71.4 Liver Lung Oral mucosa Pancreas Liver: hepatocellular adenoma Liver: cholangiocarcinoma Lung: cystic keratinizing epithelioma Oral mucosa: squamous cell carcinoma Pancreas: adenoma or carcinoma Mouse/ Outbred Swiss/H/Riop Male Gastric intubation once per week; 1 year 43 or 44 (vehicle control group = 38) 0, 1, 100, or 1,000 Liver Liver: tumors Reference NTP (2006, 197605) Toth et al. (1979, 197109) Table 2-7. Animal bioassay studies selected for noncancer dose-response modeling This document is a draftfor review purposes only and does not constitute Agency policy. 2-234 Species/ strain Exposure protocol Sex (exposure group) n Reproductive toxicity studies Monkey/ Rhesus Daily dietary exposure in female monkeys (3.5-4 years) F (F0, F1, F2, F3) 3 to 7 (F1) Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined 0, 0.15, or 0.67 0.15 0.67 Reproductive and developmental effects Rat/SpragueDawley, Long-Evans, Han/Wistar Biweekly oral Female gavage (22 weeks) 8 0, 10, 30 or 10 30 Body weight, 100 relative liver weight, relative thymus weight Mink Rat/SpragueDawley Daily dietary exposure (132 days) F 12 0.03 (control), None 0.8, 2.65, 9, or 70 Oral gavage Female (F0 3 (F0 and F1) 0 or 7.14 (GD 14 and and F1) 21, postpartum days 7 and 14), (Pups: once per week for 3 months) None 2.65 7.14 Reproductive effects Developmental effects LOAEL/NOAEL Endpoint(s) Reference Neurobehavioral effects (e.g., discriminationreversal learning affected) Bowman et al.(1989, 543744; 1989, 543745); Schantz and Bowman (1989, 198104); Schantz et al. (1986, 088206) Increased relative liver Franc et al. weight in Sprague- (2001, 197353) Dawley and Long-Evans Rats; Increased relative thymus weight in Sprague-Dawley, Han/Wistar and Long- Evans Rats Reduced kit survival Hochstein et al (2001, 197544) Lower proportion of morphologically normal pre-implantation embryos during compaction stage Hutt et al. (2008, 198268) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-235 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Reproductive toxicity studies (continued) Rat/Holtzman Com oil F (F0) gavage (initial F and M loading dose (F1 and followed by F2) weekly dose during mating, pregnancy, and lactation- about 10 weeks) 12 (F0) 0 or 16.5 Not specified (F1 and F2) None 16.5 (maternal exposure) Reproductive and developmental effects Mouse/ICR Sesame oil gavage (initial loading dose followed by weekly doses for 5 weeks) M (F0) 42 or 43 0, 0.095, or 950 0.1 100 Reproductive effects Rat/Wistar albino Olive oil M gavage (daily for 45 days) 6 0, 1, 10, or 100 None 1 Reproductive effects LOAEL/NOAEL Endpoint(s) Reference Decreased development Ikeda et al. of the ventral prostrate (2005, 197834) (F1), decreased sex ratio (percentage of males) (F2) Decreased male/female Ishihara et al. sex ratio (percentage of (2007, 197677) males) (F1) Reduced sperm production, decreased reproductive organ weights Latchoumycan dane and Mathur (2007, 197298) and related Latchoumycan dane et al. (2002, 198365; 2002, 197839; 2003, 543746) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-236 Species/ strain Exposure protocol Sex (exposure group) n Reproductive toxicity studies (continued) Rat/SpragueDawley Daily dietary F and M, exposure (F0) (3 generations) F and M, (F1 and F2) 10-32 (F0) 22 (F1) 28 (F2) Monkey/ Rhesus Daily dietary exposure (4 years) F Rat/SpragueDawley Maternal corn F (F0) oil gavage F (F1) (weekly on GD 14 and 21; PND 7 and 14) 8 3 (F0) 10 (F1) Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined 0, 1, 10, or 100 1 10 Reproductive and developmental effects 0, 0.15, or 0.67 None 0.15 0, 0.14, 0.71, 0.14 7.14, or 28.6 0.71 Reproductive effects Reproductive effects LOAEL/NOAEL Endpoint(s) Reference Decrease in fertility, Murray et al. decrease in the number (1979, 197983) of live pups, decrease in gestational survival; decrease in postnatal survival, decreased postnatal body weight in one or more generations Increased incidence of Rier et al. endometriosis (disease (1993,199987; ranged from moderate to 1995,198566) severe) Decrease serum estradiol levels (F1) Shi et al. (2007, 198147) Rhesus monkey/ Cynomolgus Offspring corn oil gavage (weekly for 11 months) Fed gelatin F capsules (5 days/week for 12 months) 6 (treatment) 0, 0.71, 3.57, 17.86 5 (controls) or 17.86 None Endometriosis effects Increased endometrial Yang et al. implant survival, (2000, 198590) increased maximum and minimum implant diameters, growth regulatory cytokine dysregulation DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-237 Species/ strain Exposure protocol Sex (exposure group) n Developmental toxicity studies Rat/Harlan SpragueDawley Com oil gavage (GD 10-16 F (F0) 80-88 (F1) Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined 0, 25, or 100 None 25 Developmental effects Rat/CRL:WI (Han) Rat/SpragueDawley Maternal daily F (F0) dietary M (F1) exposure for an estimated 20 weeks (12 weeks prior to mating through parturition) Maternal corn F (F0 and oil gavage F1) (GD 14 and 21; PND 7 and 14) 65 (F0 0, 2.4, 8, or 46 None treatments) 75 (F0 controls) at study initiation; following interim sacrifice ~30 animals were allowed to litter; F1 on PND 21 was ~7 2 or 3 (F0) 0, 7.14, or 28.6 None 7 (F1) 2.4 (maternal exposure) Reproductive and developmental effects 7.14 Developmental effects LOAEL/NOAEL Endpoint(s) Reference Decreased preference in the consumption of 0.25% saccharin solution (F1) Amin et al. (2000, 197169) Delayed BPS (F1) Bell et al. (2007, 197041) Decreased serum estradiol levels (F1) Franczak et al. (2006, 197354) Offspring corn oil gavage (weekly for 8 months) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-238 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Developmental toxicity studies Rat/Sprague- Maternal F (F0) Dawley single corn oil F and M gavage (GD 8) (F1) 12 (F0) 0, 20, 60, or 50 or 60 (F1) 180 None 20 (maternal exposure) Developmental effects Offspring exposed during gestation and lactation (35 days) Rat/ Han/Wistar and LongEvans Maternal single corn oil gavage (GD 15) F (F0) F and M (F1) Mouse/ C57BL/6J, BALB/cByJ, A/J, CBA/J, C3H/HeJ, and C57BL/10J Maternal single corn oil gavage (GD 13) F (F0) F and M (F1a, b, c) 4 to 8 (F0) 3F/3M per treatment group (F1) Dams not specified (F0); 23-36 (F1a); 4-5 (F1b); 107-110 (F1c) 0, 30, 100, 300, or 1,000 0, 10, 100, or 1,000 None None 30 (maternal exposure) Developmental effects 10 (maternal exposure) Developmental effects Mouse/ddY Maternal olive F (F0) oil gavage M (F1) (weekly for 8 weeks prior to mating) 7 (F0) 0, 0.7, or 70 3 (F1 immuno- cytochemical analysis) 6 (F1 cell number count) None 0.7 (LOEL) (maternal exposure) Neurotoxicity LOAEL/NOAEL Endpoint(s) Reference Abrogation of sexually dimorphic neuro behavioral responses (F1) Hojo et al. (2002, 198785) and related Zareba et al. (2002, 197567) Reduced mesiodistal Kattainen et length of the lower third al. (2001, molar (F1) 198952) Variation in M1 Keller et al. morphology in (2007, 198526; C57BL/10J males and 2008, 198531; females (F1a); 2008, 198033) decreased mandible shape and size in C3H/HeJ males (F1b); variation in molar shape in C3H/HeJ males (F1c) Decreased serotonin- Kuchiiwa et immunoreactive neurons al. (2002, in raphe nuclei of male 198355) offspring (F1) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-239 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) LOAEL/NOAEL strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Endpoint(s) Reference Developmental toxicity studies Mouse/NIH (pregnant and pseudo pregnant) Maternal sesame oil gavage daily for 8 days (GD 1-8) F 10 0, 2, 50, or 100 None 2 Developmental Decreased progesterone Li et al. (2006, effects and increased serum 199059) estradiol levels Rat/Holtzman Maternal F (F0 and single olive oil F1) gavage (GD 18) 4-7 (F0 and F1) 0, 20, 60, or 180 None 20 (maternal exposure) Behavioral effects Decreased training responses (F1) Markowski et al. (2001, 197442) Rat/Line C Maternal single corn oil gavage (GD 15) F (F0) F and M (F1) 24-32 (treatment) 12-48 (controls) 0, 30, 100, None 300, or 1,000 30 (maternal exposure) Developmental Increase in dental caries Miettinen et effects (F1) al. (2006, 198266) Rat/Holtzman Maternal single corn oil gavage (GD 15) F (F0) M (F1) Rat/Holtzman Maternal single corn oil gavage (GD 15) F (F0) M (F1) Not specified 0, 12.5, 50, (F0) 200, or 800 5 males and 3 females (F1) 6 (F0) 0, 12.5, 50, 5 males and 200, or 800 3 females (F1) 800 (maternal exposure) None Immunotoxicity Decreased spleen cellularity (F1) 12.5 (maternal exposure) 50 (maternal exposure) Developmental Decreased anogenital effects distance (F1) Nohara et al. (2000, 200027) Ohsako et al. (2001, 198497) Rat/Harlan SpragueDawley Maternal corn F(F0) oil gavage (GD 10-16 ~4 (F0); 0, 25, or 100 None 80-88 (F1) None Developmental Facilitatory effect on Schantz et al. effects radial arm maze learning (1996, 198781) (F1) Rat/Sprague- Maternal corn F and M Dawley oil gavage (F1) (GD 10-16) ~15 (F0); 5-9 (F1) 0, 25, or 100 25 100 Developmental Decreased thymus Seo et al. effects weight (1995,197869) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. 2-240 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) LOAEL/NOAEL strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Endpoint(s) Reference Developmental toxicity studies Rat/TCDD- Maternal com F (F0) 5-8 (F0) 0, 30, 100, 100 300 Reproductive Reduction in daily Simanainen et resistant oil gavage M (F1) 300, or 1,000 effects sperm production and al. (2004, Han/Wistar (GD 15) cauda epididymal sperm 198106) bred with reserves TCDD- sensitive Long-Evans Mouse/C57/6 NCji Maternal F (F0) drinking water F and M exposure (F1) (daily for 17-day lactational period) 8 (F0) 0, 1.14, or 11.3 1.14 Not specified (NOEL) (F1) (maternal exposure) 11.3 (LOEL) (maternal exposure) Immunotoxicity Increased susceptibility Sugitato Listeria (F1 males Konishi et al. and females); increase in (2003, 198375) thymic CD4+ cells (F1 males); decreased spleen weight (F1 males) Acute toxicity studies Mouse/B6C3F1 Com oil F gavage (single exposure) 20 0, 1, 5, 10, 50, 5 100, or 6,000 10 Immunotoxicity Increased mortality from Burleson et al. influenza infection (1996,196998) 7 days after a single TCDD exposure Rat/LongEvans Com oil gavage (4 consecutive days) F 14, 6, 12, 6, 6, 6, 6, 6, 6, and 4, respectively in control and treated groups 0, 0.1, 3, 10, 30, 100, 300, 1,000, 3,000, or 10,000 30 100 Thyroid effects Reduction in serum T4 Crofton et al. levels (2005, 197381) Rat/Sprague- Corn oil F Dawley gavage (single dose) 4 (treated); 9 (control) 0, 0.6, 2, 4, 20, 0.6 60, 200, 600, (NOEL) 2,000, 5,000, or 20,000 2 (LOEL) Enzyme induction Increased benzo(a)pyrene hydroxylase (BPH) Kitchin and Woods (1979, 198750) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-241 Species/ strain Exposure protocol Sex (exposure group) Acute toxicity studies (continued) Rat/SpragueDawley Com oil dose via oral gastric intubation (single dose) F 10 n Rat/SpragueDawley Corn oil gavage or TCDDcontaminated soil (single dose) F 6 Mouse/ B6C3F1 (BALB/c (C57BL/6N (and DBA2 Rat/TCDDresistant Han/Wistar bred; TCDDsensitive Long-Evans Corn oil M, F gavage (single dose) Corn oil M, F gavage (single dose) 10-40 9-11 Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined LOAEL/NOAEL Endpoint(s) Reference 0, 3, 10, 30, 100, 300, 1.000, 3,000, 10.000, or 30,000 3 0, 15, 40, 100, 200, 500, 1,000, 2,000, or 5,000 in corn oil None 0, 15, 44, 100, 220, 500, 1,100, 2,000, or 5,500 in contaminated soil 0, 5, 20, 100, 500 or 500 30-100,000 100 10 Hormonal Increased serum FSH Li et al. (1997, effects 199060) 15 (LOEL) Enzyme induction Induction of aryl hydrocarbon hydroxylase (at low dose in both treatment protocols) Lucier et al. (1986, 198398) None 300 Mortality and body weight changes No increased mortality Nohara et al. of virus-infected mice or (2002, 199021) treatment-related changes in body weight General Reduction in serum T4 toxicological levels endpoints, organ weights, dental defects Simanainen et al. (2002, 201369) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-242 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) LOAEL/NOAEL strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Endpoint(s) Reference Acute toxicity studies (continued) Rat/TCDDresistant Han/Wistar bred with TCDDsensitive Long-Evans Com oil M, F gavage (single dose) 5-6 Line A: 100 300 General Decreased thymus Simanainen et 30-3,000,000 toxicological weight al. (2003, Line B: endpoints, organ 198582) 30-1,000,000 weights, dental Line C: defects 30-100,000 Mouse/ Com oil C57BL/6N gavage (single CYP1A2 (+/+) dose) wild-type F Not specified 0, 30, 100, 300, 1000, 3000, or 10,000 300 1,000 Immunotoxicity Decreased antibody response to SRBCs Smialowicz et al. (2004, 110937) Rat/Sprague- Com oil F Dawley gavage (single dose) 5-15 0, 0.05, 0.1, 1, 0.1 1 Liver effects Increase in hepatic Vanden et al. 10, 100, 1,000, (NOEL) (LOEL) EROD activity and (1994,197551) or 10,000 CYP1A1 mRNA levels Subchronic toxicity studies Rat/Sprague- Com oil F Dawley gavage (daily for 28 days) 5 0, 2.5, 25, 250, 250 or 1,000 1,000 Body and organ weight changes Decreased body weight, increased relative liver weight and related biochemical changes, decreased relative thymus weight Chu et al. (2001, 521829) Rat/Sprague- Com oil F Dawley gavage (daily for 28 days) 5 0, 2.5, 25, 250, 2.5 or 1,000 25 Liver effects Alterations in thyroid, thymus, and liver histopathology Chu et al., 2007 Guinea pig/ Hartley Daily dietary exposure (90 days) M, F 10/sex 0, 0.12, 0.61, 4.9, or 26 (males); 0, 0.12, 0.68, 4.86, or 31 (females) 0.61 4.9 Body and organ weight changes Decreased body weight (male and females); increased relative liver weights (males); decreased relative thymus weight (males) DeCaprio et al. (1986, 197403) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-243 Species/ strain Exposure protocol Sex (exposure group) n Subchronic toxicity studies (continued) Mice/B6C3F1 Com oil gavage (5 days/week for 13 weeks) F 5 Rat/Iva:SIV 50-SpragueDawley Daily dietary exposure (13 weeks) M, F 6 Daily dietary exposure (13 weeks) M, F 6 Daily dietary exposure (13 weeks) M, F 6 Daily dietary exposure (13 weeks, 26, and 39 weeks) F 6 Rat/SpragueDawley Gavage loading/ maintenance doses (every 4 days for 14 days) M, F 6 Mouse/ B6C3F1 Corn oil gavage (5 days/week for 13 weeks) F Not specified Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined LOAEL/NOAEL Endpoint(s) Reference 0, 1.07, 3.21, None 10.7, 32.1, or 107 0, 20, 200, or None 2,000 0 or 200 1.07 (LOEL) 20 Body and organ weight changes; enzyme induction Increased EROD, ACOH and phosphotyrosyl proteins at all doses DeVito et al. (1994,197278) Liver effects Reduced hepatic vitamin A levels Fattore et al. (2000, 197446) 0, 200, or 1,000 0 or 100 0, 0.55, 307, or 0.57 1,607 327 Body and liver weight changes; hepatic cell proliferation Increased absolute and relative liver weight Fox et al. (1993,197344) 0, 0.32, 1.07, None 10.7, or 107 0.32 (LOEL) Brain effects Induction of biomarkers Hassoun et al. of oxidative stress at all (1998, 136626) doses DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-244 Sex Average daily Species/ Exposure (exposure dose levels NOAEL LOAEL Endpoint(s) LOAEL/NOAEL strain protocol group) n (ng/kg-day) (ng/kg-day) (ng/kg-day) examined Endpoint(s) Reference Subchronic toxicity studies (continued) Rat/Harlan SpragueDawley Com oil gavage (5 days/week for 13 weeks) F 6 0, 2.14, 7.14, None 15.7, 32.9, or 71.4 2.14 (LOEL) Liver and brain Induction of biomarkers Hassoun et al. effects of oxidative stress at all (2000, 197431) doses in liver and brain Rat/Harlan SpragueDawley Com oil gavage (5 days/week for 13 weeks) F 12 0, 7.14, 15.7, None or 32.9 7.14 (LOEL) Brain effects Induction of biomarkers Hassoun et al. of oxidative stress at all (2003, 198726) doses Rat/SpragueDawley Corn oil gavage (5 days/week for 13 weeks) M, F 12 0, 0.71, 7.14, 7.14 71.4 Liver effects, Reduced body weight Kociba et al. 71.4, or 714 body weight and food consumption, (1976, 198594) changes, and slight liver degeneration, hematologic and lymphoid depletion, clinical effects increased urinary porphyrins and delta aminolevulinic acid, increased serum alkaline phosphatase and bilirubin Rat/F344 Corn oil gavage (2 days/week for 28 days) F 3 0, 0.71, 7.14, None or 71.4 0.71 (LOEL) Clinical signs and histopathology Decreased Cx32 plaque Mally and number and area in the Chipman liver (2002, 198098) Mouse/ B6C3F1 Corn oil gavage (5 days/week for 13 weeks) F Not specified 0, 0.11, 0.32, 1.07 1.07, 10.7, or (NOEL) 107.14 10.7 (LOEL) Liver, lung, kidney, and spleen effects Increased hepatic superoxide anion Slezak et al. (2000, 199022) Mouse/ B6C3F1 Corn oil gavage (5 days/week for 13 weeks) F 8-15 0, 1.07, 10.7, None 1.07 Immunotoxicity Reduced antibody Smialowicz et 107, or 321 and organ response to SRBC, al. (2008, weight increased relative liver 198341) weight DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-245 Species/ strain Exposure protocol Sex (exposure group) Subchronic toxicity studies (continued) Rat/Sprague- TCDD in diet F Dawley (13 weeks) 8 n Guinea pig/ Hartley Corn oil gavage (weekly for 8 weeks) F 10 Mouse/ B6C3F1 Corn oil F gavage (daily for 14 days) Chronic toxicity studies Rat/CDCOBS Com oil gavage (weekly for 45 weeks) F Rat/SpragueDawley Loading/ F maintenance dose (every 3 days for different durations up to 128 days) Rat/SpragueDawley Corn oil gavage (5 days/week for 30 weeks) F 6-8 4 5 6 Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined LOAEL/NOAEL Endpoint(s) Reference 0, 14, 26, 47, None 320, or 1,024 14 0, 1.14, 5.71, 1.14 28.6, or 143 5.71 0, 10, 50, 100, None 500, 1,000, or 2,000 10 Multiple end points Decreased absolute and Van Birgelen relative thymus weights, (1995,197096; decreased liver retinoid 1995, 198052) levels Immunotoxicity Decreased total leukocytes and lymphocyte count, decreased absolute thymus and weight, increase in primary serum tetanus antitoxin Vos et al. (1973,198367) Immunotoxicity Reduction of serum complement activity White et al. (1986,197531) 0, 1.43, 14.3, None or 143 1.43 Hepatic porphyria Increased urinary porphyrin excretion Cantoni et al. (1981, 197092) 0, 0.85, 3.4, 13.6, 54.3, or 217 (28-day duration) 54.3 (28-day duration) 217 (28-day duration) Body weight Decreased body weight, changes and decreased PEPCK changes in activity, and reduced PEPCK activity IGF-I levels and IGF-I levels Croutch et al. (2005, 197382) 0, 2.14, 7.14, None 15.7, 32.9, or 71.4 2.14 (LOEL) Brain effects Induction of biomarkers Hassoun et al. of oxidative stress at all (2002, 543725) doses DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 2-246 Species/ strain Exposure protocol Sex (exposure group) Chronic toxicity studies (continued) Rat/SpragueDawley Daily dietary exposure (2 years) M, F 50 n Rat/Sprague- Biweekly Dawley gavage (30 weeks) F 9 Mouse/ B6C3F1; Rat/Osborne Mendel Corn oil M, F gavage (2 days/week for 104 weeks) 50 Rat/SpragueDawley Corn oil F gavage (5 days/week for 105 weeks) 53 Monkey/ Rhesus Daily dietary exposure (4 years) F 8 Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined LOAEL/NOAEL Endpoint(s) Reference 0, 1, 10, or 100 1 10 0, 3.5, 10.7, 35, or 125 10.7 35 0, 1.4, 7.1, or None 71 for rats and male mice; 0, 5.7, 28.6, or 286 for female mice 0, 2.14, 7.14, None 15.7, 32.9, or 71.4 1.4 2.14 0, 0.15, or None 0.67 0.15 Multiple endpoints measured Body and organ weight changes, clinical chemistry, hepatocellular proliferation Liver and body weight changes Increased urinary porphyrins, hepatocellular nodules, and focal alveolar hyperplasia Increased relative liver weight Increased incidences of liver lesions in mice (males and females) Kociba et al. (1978, 001818) Maronpot et al. (1993, 198386) NTP (1982, 543764) Liver and lung effects Increased absolute and relative liver weights, increased incidence of hepatocellular hypertrophy, increased incidence of alveolar to bronchiolar epithelial metaplasia General Elevated serum toxicological triglycerides and endpoints and total lipids reproductive effects NTP (2006, 197605) Rier et al. (2001, 198776; 2001, 543773) DRAFT: DO NOT CITE OR QUOTE Table 2-7. Animal bioassay studies considered for noncancer dose-response modeling (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Species/ strain Exposure protocol Sex (exposure group) Chronic toxicity studies (continued) Rat/Sprague- Biweekly Dawley gavage (30 weeks) F 9 n Rat/Sprague- Biweekly Dawley gavage (30 weeks) Mouse/Swiss/ H/Riop Sunflower oil gavage (weekly for 1year) F M 9 38-44 Average daily dose levels NOAEL LOAEL (ng/kg-day) (ng/kg-day) (ng/kg-day) Endpoint(s) examined LOAEL/NOAEL Endpoint(s) Reference 0, 3.5, 10.7, None 35, or 125 0, 0.1, 0.35, 1, 10.7 3.5, 10.7, 35, or 125 0, 1, 100, or None 1,000 3.5 (LOEL) 35 1 EGFR kinetics Decrease in EGFR and auto maximum binding phosphorylation, capacity hepatocellular proliferation Thyroid function Decreased serum T4 levels Sewall et al. (1993, 197889) Sewall et al. (1995,198145) Skin effects Dermal amyloidosis and Toth et al. skin lesions (1979, 197109) 2-247 ND = not determined. DRAFT: DO NOT CITE OR QUOTE 1 2 Figure 2-1. EPA's process to select and identify in vivo mammalian and 3 epidemiologic studies for use in the dose-response analysis of TCDD. EPA 4 first conducted a literature search to identify studies published since the 2003 5 Reassessment. Results were published and additional study submissions were accepted 6 from the public. Next EPA developed TCDD-specific study inclusion criteria for in vivo 7 mammalian studies and held a Dioxin Workshop where these criteria were discussed and 8 refined. Third, EPA developed two final sets of study inclusion criteria, one for in vivo 9 mammalian studies and another for epidemiologic studies. Finally, EPA applied these 10 two sets of criteria to all studies from the literature search, public submissions, 2003 11 Reassessment, and additional studies identified by EPA after the Dioxin Workshop 12 through October 2009. The studies that met these criteria formed a list of key studies for 13 EPA's consideration in TCDD dose-response assessment. This document is a draftfor review purposes only and does not constitute Agency policy. 2-248 DRAFT--DO NOT CITE OR QUOTE List of available epidemiologic studies on TCDD and DLCs 1 2 Figure 2-2. EPA's process to evaluate available epidemiologic studies using 3 study inclusion criteria for use in the dose-response analysis of TCDD. EPA 4 applied its TCDD-specific epidemiologic study inclusion criteria to all studies published 5 on TCDD and DLCs. The studies were initially evaluated using five considerations 6 regarded as providing the most relevant kind of information needed for quantitative 7 human health risk analyses. For each study that was published in the peer-reviewed 8 literature, EPA then examined whether the exposures were primarily to TCDD and if the 9 TCDD exposures could be quantified so that dose-response analyses could be conducted. 10 Finally, EPA required that the effective dose and oral exposure be estimable: (1) for 11 cancer, information is required on long-term exposures, (2) for noncancer, information is 12 required regarding the appropriate time window of exposure that is relevant for a specific, 13 nonfatal health endpoint, and (3) for all endpoints, the latency period between TCDD 14 exposure and the onset of the effect is needed. Only studies meeting these criteria were 15 included in EPA's TCDD dose-response analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-249 DRAFT--DO NOT CITE OR QUOTE List of available in vivo mammalian bioassay studies on TCDD 1 2 Figure 2-3. EPA's process to evaluate available animal bioassay studies using study 3 inclusion criteria for use in the dose-response analysis of TCDD. EPA evaluated all 4 available in vivo mammalian bioassay studies on TCDD. Studies had to be published in 5 the peer-reviewed literature. Next, to ensure working in the low-dose range for TCDD 6 dose-response analysis, EPA applied dose requirements to the lowest tested average daily 7 doses in each study, with specific requirements for cancer (<1 gg/kg-day) and noncancer 8 (<30 ng/kg-day) studies. Third, EPA required that the animals were exposed via the oral 9 route to only TCDD and that the purity of the TCDD was specified. Finally, the studies 10 were evaluated using four considerations regarded as providing the most relevant kind of 11 information needed for quantitative human health risk analyses from animal bioassay 12 data. Only studies meeting all of these criteria and considerations were included in 13 EPA's TCDD dose-response analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-250 DRAFT--DO NOT CITE OR QUOTE 1 3. THE USE OF TOXICOKINETICS IN THE DOSE-RESPONSE MODELING FOR 2 CANCER AND NONCANCER ENDPOINTS 3 4 5 A key recommendation from the National Academy of Sciences (NAS) for improving the 6 2003 Reassessment was that U.S. Environmental Protection Agency (EPA) should justify its 7 approaches to dose-response modeling for cancer and noncancer endpoints. Further, the NAS 8 suggested that EPA incorporate the most up-to-date and relevant state of the science for 9 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) dose-response assessment. 10 While EPA believes that at the time of its release, the 2003 Reassessment offered a 11 substantial improvement over the general state-of-the-science regarding dose-response modeling, 12 EPA agrees with the NAS that the justification of the approaches to dose-response modeling can 13 be improved and the methodologies updated to reflect the most current EPA practices and 14 science. In Section 3, EPA describes the use of toxicokinetic (TK)11 information in the 15 dose-response modeling of TCDD. Section 3.1 summarizes the NAS comments regarding the 16 use of TK in the dose-response approaches for TCDD. Section 3.2 overviews EPA's responses 17 to the NAS comments. Section 3.3 discusses TCDD kinetics, including TK models developed to 18 simulate disposition of this compound in rodents and humans (see Section 3.3.4), alternative 19 measures of dose that could be used in a TCDD dose-response analysis and uncertainties in the 20 TCDD dose estimates (see Section 3.3.5). Sections 4 and 5 of this document incorporate the TK 21 information into noncancer and cancer dose response modeling, respectively. 22 23 3.1. SUMMARY OF NAS COMMENTS ON THE USE OF TOXICOKINETICS IN 24 DOSE-RESPONSE MODELING APPROACHES FOR TCDD 25 The NAS commented on the appropriate use of TK models in dose-response modeling 26 for TCDD. Specifically, the committee requested that EPA consider using such models to 27 provide refined estimates of dose, for example, as the underlying science and predictive 28 capabilities of these models improved. 29 30 [Discussing Kinetic models],..the committee encourages further development and 31 use of these models as data become available to validate and further develop them 32 (NAS, 2006, 198441p. 59). "Toxicokinetics (TK) is that part of the pharmacokinetics (PK) where toxicity is resulted in the organism. This document is a draftfor review purposes only and does not constitute Agency policy. 3_1 DRAFT--DO NOT CITE OR QUOTE 1 Although the NAS basically agreed with EPA's use of body burden as a dose metric in 2 the 2003 Reassessment (e.g., see NAS, 2006, 198441, p. 7), the NAS was concerned about the 3 limitations of first order kinetic models, such as the one used in the 2003 Reassessment, to 4 estimate TCDD body burdens. 5 6 TCDD, other dioxins, and DLCs act as potent inducers of CYP, a property that 7 can affect both the hepatic sequestration of these compounds and their half-lives. 8 Hepatic sequestration of dioxin may influence the quantitative extrapolation of the 9 rodent liver tumor results because the body-burden distribution pattern in highly 10 dosed rats would differ from the corresponding distribution in humans subject to 11 background levels of exposure. EPA should consider the possible quantitative 12 influence of dose-dependent toxicokinetics on the interpretation of animal 13 toxicological data (NAS, 2006, 198441, p. 129). 14 15 The NAS also asked EPA to evaluate the impact of kinetic uncertainty and variability on 16 dose-response assessment. The NAS committee asked EPA to use TK models to examine both 17 interspecies and human interindividual differences in the disposition of TCDD, which would 18 better justify EPA dose-response modeling choices. 19 20 The Reassessment does not adequately consider the use of a PBPK model to 21 define species differences in tissue distribution in relation to total body burden for 22 either cancer or noncancer end points (NAS, 2006, 198441, p. 62). 23 24 EPA .. .should consider physiologically based pharmacokinetic modeling as a 25 means to adjust for differences in body fat composition and for other differences 26 between rodents and humans (NAS, 2006, 198441, p. 10). 27 28 The Reassessment does not provide details about the magnitudes of the various 29 uncertainties surrounding the decisions EPA makes in relation to dose metrics 30 (e.g., the impact of species differences in percentage of body fat on the 31 steady-state concentrations present in nonadipose tissues). The committee 32 recommends that EPA use simple PBPK models to define the magnitude of any 33 differences between humans and rodents in the relationship between total body 34 burden at steady-state concentrations (as calculated from the intake, half-life, 35 bioavailability) and tissue concentrations. The same model could be used to 36 explore human variability in kinetics in relation to elimination half-life. EPA 37 should modify the estimated human equivalent intakes when necessary (NAS, 38 2006, 198441, p. 73). 39 This document is a draftfor review purposes only and does not constitute Agency policy. 3-2 DRAFT--DO NOT CITE OR QUOTE 1 Finally, the NAS asked EPA to use TK considerations to better justify its choice of dose 2 metric. 3 4 EPA makes a number of assumptions about the appropriate dose metric and 5 mathematical functions to use in the Reassessment's dose-response analysis ... 6 but does not adequately comment on the extent to which each of these 7 assumptions could affect the resulting risk estimates...EPA did not quantitatively 8 describe how this particular selection affected its estimates of exposure and 9 therefore provided no overall quantitative perspective on the relative importance 10 of the selection (NAS, 2006, 198441, p. 51). 11 12 3.2. OVERVIEW OF EPA'S RESPONSE TO THE NAS COMMENTS ON THE USE OF 13 TOXICOKINETICS IN DOSE-RESPONSE MODELING APPROACHES FOR 14 TCDD 15 In response to the NAS recommendations regarding TCDD kinetics and choice of dose 16 metrics, this document presents an in depth evaluation of TCDD TK models, exploring their 17 differences and commonalities and their possible application for the derivation of dose metrics 18 relevant to TCDD. Initially, EPA discusses the application of first order kinetics to estimate 19 body burden as a dose metric for TCDD. This first order kinetic model is used to predict TCDD 20 body burden for all of the studies identified as Key Studies (see Section 2.4); this model uses a 21 constant half-life to simulate the elimination of TCDD from the body. However, given the 22 observed data indicating early influence of cytochrome P450 1A2 (CYP1A2) induction and 23 binding to TCDD in the liver and later redistribution of TCDD to fat tissue, the use of a constant 24 half-life for TCDD clearance following long term or chronic TCDD exposure is not biologically 25 supported. Therefore, using half-life estimates based on observed terminal steady state levels of 26 TCDD will not account for the possibility of an accelerated dose-dependent clearance of the 27 chemical during early stages following elevated TCDD exposures. The biological processes 28 leading to dose-dependent TCDD excretion are better described using physiologically based 29 pharmacokinetic (PBPK) models than by simple first order kinetic models. Additionally, as part 30 of its preparation for developing this document, EPA evaluated recent TCDD kinetic studies as 31 NAS advocated. Although the NAS agreed with continued use of body burden metric as the 32 dose metric of choice, EPA believes that the state-of-the-practice has advanced sufficiently to 33 justify the consideration of alternative dose metrics (other than administered dose) based on an 34 application of a physiologically-based TK model. This document is a draftfor review purposes only and does not constitute Agency policy. 3-3 DRAFT--DO NOT CITE OR QUOTE 1 EPA identified a number of advances in the overall scientific understanding of TCDD 2 disposition; many of these are documented in a summary discussion introducing the section on 3 TCDD kinetics (see Section 3.3). The increased understanding warranted an evaluation of 4 current kinetic modeling of TCDD to determine if the use of such models would improve the 5 dose-response assessment for TCDD. Justification of the final PBPK model choice is detailed in 6 Section 3.3. Through the choice of a published PBPK model to estimate dose metrics for dioxin, 7 EPA has addressed several of the NAS concerns. The PBPK model can be applied to estimate 8 dose metrics other than body burden that may be more directly related to response, e.g., tissue 9 levels, serum levels, blood concentrations, or dose metrics related to TCDD-protein receptor 10 binding. The selected PBPK model included explicit description of physiological and 11 biochemical parameters, therefore, it can also provide an excellent tool for investigating 12 differences in species uptake and disposition of TCDD. One of the criteria used to select a 13 PBPK model for TCDD kinetics was the availability of both human and animal models so that 14 differences in species uptake and disposition of TCDD can be investigated. Additionally, the 15 PBPK model includes quantitative information that is suitable for addressing the impact of 16 physiological (e.g., body weight [BW] or fat tissue volume), or biochemical (e.g., induction of 17 CYP1A2) variability on overall risk of TCDD between species, in response to another area of 18 concern in the NAS report. The sensitivity analysis and uncertainty in dose metrics derived for 19 the risk assessment of TCDD are also presented in Section 3.3. Detailed discussion on the 20 uncertainty in choice of PBPK model-driven dose metrics is also provided in Section 3.3. 21 22 3.3. PHARMACOKINETICS (PK) AND PK MODELING 23 3.3.1. PK Data and Models in TCDD Dose-Response Modeling: Overview and Scope 24 In general, the use of measures of internal dose in dose-response modeling is considered 25 to be superior to that of administered dose (or uptake) because the former is more closely related 26 to the response. The evaluation of internal dose, or dose metric, in exposed humans and other 27 animals is facilitated by an understanding of pharmacokinetics (i.e., absorption, distribution, 28 metabolism, and excretion). When measurements of internal dose (e.g., blood concentration, 29 tissue concentration) are not available in animals and humans, pharmacokinetic models can be 30 used to estimate them. The available data on the pharmacokinetics of TCDD in animals and This document is a draftfor review purposes only and does not constitute Agency policy. 3-4 DRAFT--DO NOT CITE OR QUOTE 1 humans have been reviewed (NAS, 2006, 198441; U.S. EPA, 2003, 537122; van Birgelen and 2 van, 2000, 523248). 3 It is evident based on these reviews and other analyses that three distinctive features of 4 TCDD play important roles in determining its pharmacokinetic behavior, as discussed below: 5 6 TCDD is very highly lipophilic and thus is more soluble in fat or other relatively 7 nonpolar organic media than in water. The n-octanol/water partition coefficient is a 8 commonly-used measure of lipophilicity equal to the equilibrium ratio of a substance's 9 concentration in n-octanol (a surrogate for biotic lipid) to the substance's concentration 10 in water (Leo et al., 1971, 019600). For TCDD, this coefficient is on the order of 11 10,000,000 or more (ATSDR, 1998, 197033). It follows that the solubility of TCDD in 12 the body's lipid fraction, i.e., the fatty portions of various tissues, including adipose, 13 organs, and blood, is extremely high. 14 TCDD is very slowly metabolized compared to many other organic compounds, with an 15 elimination half life in humans on the order of years following an initial period of 16 distribution in the body (Carrier et al., 1995, 197618; Michalek et al., 2002, 199579). 17 Most laboratory animals used for toxicologic testing tend to eliminate TCDD much more 18 quickly than people, although even in animals TCDD is eliminated much more slowly 19 than most other chemicals. 20 TCDD induces binding proteins in the liver that have the effect of sequestering some 21 of the TCDD. The ability of TCDD to alter gene expression and the demonstration that 22 the induction of CYP1A2 is responsible for hepatic TCDD sequestration suggest that 23 both pharmacokinetic and pharmacodynamic events must be incorporated for a 24 quantitative description of TCDD disposition (Santostefano et al., 1998, 200001). The 25 induction of these proteins implies that TCDD tends to be eliminated more rapidly in the 26 early years following short-term, high-level exposures than it is after those initial levels 27 have declined. Leung et al. (1988, 198815) and Andersen et al. (1993, 196991), in their 28 PBPK modeling, had taken into consideration the issue of liver protein binding. Recent 29 efforts of pharmacokinetic modeling have supported the concentration-dependent 30 elimination of TCDD in animals and humans (Aylward et al., 2005, 197014; Emond et 31 al., 2006, 197316). 32 33 Sections 3.3.2 and 3.3.3 present the salient features of TCDD pharmacokinetics in 34 animals and humans, with particular focus on mechanisms and data of relevance to interspecies 35 and intraspecies variability. Section 3.3.4 describes the various dose metrics for the 36 dose-response modeling of TCDD and the characteristics of pharmacokinetic models potentially 37 useful for estimating these metrics. Finally, Sections 3.3.5 and 3.3.6 summarize the results of 38 application of pharmacokinetic models to derive dose metrics as well as the uncertainty 39 associated with the predictions of dose metrics used in dose-response modeling. Dose metrics This document is a draftfor review purposes only and does not constitute Agency policy. 3-5 DRAFT--DO NOT CITE OR QUOTE 1 derived via PBPK modeling approaches are utilized in Sections 4 and 5 of this document for 2 noncancer and cancer TCDD dose-response modeling, respectively. 3 4 3.3.2. PK of TCDD in Animals and Humans 5 3.3.2.1. A b so rp tio n a n d B io availability 6 When administered via the oral route in the dissolved form, TCDD appears to be well 7 absorbed. Animal studies indicate that oral exposure to TCDD in the diet or in an oil vehicle 8 results in the absorption of >50% of the administered dose (Nolan et al., 1979, 543785; Olson et 9 al., 1980, 197976). Human data from Poiger and Schlatter (1986, 197336) indicate that >87% of 10 the oral dose (after ingestion of 105 ng [3H]-2,3,7,8-TCDD [1.14 ng/kg BW] in 6 mL corn oil) 11 was absorbed from the gastrointestinal tract. Lakshmanan et al. (1986, 548729), investigating 12 the oral absorption of TCDD, suggested that it is absorbed primarily by the lymphatic route and 13 transported predominantly by chylomicrons. 14 Oral absorption is generally less efficient when TCDD is more tightly bound in soil 15 matrices. Based on experiments in miniature swine, Wittsiepe et al. (2007, 548736) reported an 16 approximately 70% reduction in bioavailability when TCDD was administered in the form of 17 contaminated soil, relative to TCDD after extraction from the same soil matrix with solvents. 18 Working with soil from the prominent contamination site at Times Beach, Missouri, Shu et al. 19 (1988, 548739) reported an oral bioavailability of approximately 43% based on experiments in 20 rats. Percent dose absorbed by the dermal route is reported to be less than the oral route, whereas 21 absorption of TCDD by the transpulmonary route appears to be efficient (Banks and Birnbaum, 22 1991, 548742; see, for example; Banks et al., 1990, 548741; Diliberto et al., 1996, 143712; 23 Nessel et al., 1992, 548743; Roy et al., 2008, 548747; U.S. EPA, 2003, 537122). 24 25 3.3.2.2. D istribu tion 26 TCDD in systemic circulation equilibrates and partitions into the tissues where it is then 27 accumulated, bound, or eliminated. Whereas the bulk of the body tissues are expected to 28 equilibrate in a matter of hours, the adipose tissue will approach equilibrium concentrations with 29 blood much more slowly. Consistent with these assertions, a number of experimental and 30 modeling studies in rats and humans have shown that TCDD has a large volume of distribution 31 (Vd), i.e., the apparent volume in which it is distributed. The Vd corresponds to the volume of This document is a draftfor review purposes only and does not constitute Agency policy. 3-6 DRAFT--DO NOT CITE OR QUOTE 1 blood plus the product of internal tissue volumes and the corresponding tissue:blood partition 2 coefficients. This parameter is a key determinant of the elimination rate of TCDD in exposed 3 organisms. The tissue:blood partition coefficients of TCDD, in turn, are determined by the 4 relative solubility of TCDD in tissue and blood components (including neutral lipids, 5 phospholipids, and water). 6 Column 1 in Table 3-1 presents the tissue:blood partition coefficients for TCDD (Emond 7 et al., 2005, 197317; Wang et al., 1997, 104657). Column 3 of this table lists the physical 8 volume of each tissue, scaled to a person weighing 60 kg. The last column shows the 9 implications of the tissue volumes and tissue:blood partition coefficients for the effective 10 volumes of distribution for each tissue and for the body as a whole. It can be seen that, purely on 11 the basis of solubility space, the fat should be expected to contain about 94% of the TCDD in the 12 body, and that the body as a whole behaves as if it is about 1,200 liters in terms of 13 blood-equivalents (i.e., approximately 22-fold larger than its physical volume). 14 Maruyama et al. (2002, 198448) have published another set of tissue/blood partition 15 coefficients for TCDD and other dioxin congeners based in part on observations of tissue 16 concentrations measured in autopsy specimens from eight Japanese people without known 17 unusual exposures to TCDD. Their estimates of TCDD partition coefficients seem to be rather 18 large and variable, with a fat:blood value of 247 78 (standard deviation [SD]), a liver:blood 19 value of 9.8 5.7 and a muscle:blood value of 18 10.6. Depending on time of autopsy, tissue 20 samples may not be an accurate source of information on observed, in vivo partition coefficients 21 because weight loss is likely to occur pre and post mortem. In particular, a decline in fat stores 22 volume could lead to an increased concentration of dioxin in fat in autopsy specimens relative to 23 what would be observed in vivo. 24 The calculations shown in Table 3-1 do not include the additional amount that will be 25 bound to induced proteins in the liver. That induction and binding will tend to increase the 26 contribution of the liver on the effective volume of distribution (Birnbaum, 1986, 548749). 27 It is also of interest to point out some basic implications of the data in Table 3-1 for the 28 expected rates of perfusion-mediated transfer of TCDD between blood and each of the 29 organ/tissues. The rate of loss from a tissue (occurring primarily via blood flow) and the 30 corresponding half-life can be calculated using the following equations: 31 This document is a draftfor review purposes only and does not constitute Agency policy. 3-7 DRAFT--DO NOT CITE OR QUOTE 1 Rate constant for loss (hour'1) - Blood flow (liters / honr) (Eq. 3-1) Tissue volume (liters) x Tissue / Blood Partition Coefficent 2 t1/2 for tissue perfusion loss ln (2) Rate constant for loss 3 ln(2) x Tissue volume (liters) x Tissue/Blood Partition Coefficent Blood flow (liters/hour) (Eq. 3-2) 4 5 Because TCDD is highly lipophilic, its concentration in the aqueous portion of the blood 6 is very small, and TCDD tends to partition from blood components into cellular membranes and 7 tissues, probably in large part via diffusion. As a result, full equilibrium concentrations of 8 TCDD are not attained by the end of the transit time through organs from the arterial to venous 9 blood. For organs in which this occurs, diffusion coefficients or "permeability factors" have 10 been estimated to assess the fractional attainment of equilibrium concentration that occurs by the 11 time the blood leaving each organ reaches the venous circulation. Table 3-2 presents the 12 permeability factors and implications for perfusion half-lives for TCDD, per Emond et al. (2005, 13 197317; 2006, 197316). 14 Despite the high lipid bioconcentration potential of TCDD, the adipose tissue does not 15 always have the highest concentration (Abraham et al., 1988, 199510; Geyer et al., 1986, 16 064899; Poiger and Schlatter, 1986, 197336). Further, the ratios of tissue:tissue concentrations 17 of TCDD and related compounds (e.g., the liver:adipose ratio) may not remain constant during 18 nonsteady-state conditions. TCDD concentrations have been observed to decrease more rapidly 19 in the liver than in adipose tissue. For example, Abraham et al. (1988, 199510) found that the 20 liver:adipose tissue concentration ratio in female Wistar rats exposed to a subcutaneous TCDD 21 dose of 300 ng/kg decreased from 10.3 at 1 day postexposure to 0.5 at 91 days postexposure. It 22 should be noted that even at a ratio of 0.5, the amount of TCDD in the liver is greater than that 23 based on lipid content of the tissue alone, consistent with the presence of hepatic TCDD binding 24 proteins. The liver/adipose tissue ratio also was dose-dependent, such that the liver TCDD 25 burden increased from ~11% of the administered dose at low doses (i.e., 1-10 ng/kg) to ~37% of 26 the dose at an exposure level of 300 ng/kg. The increase in TCDD levels in liver, accompanied 27 by a decrease in concentration in the adipose tissue, is a particular behavior to be considered in This document is a draftfor review purposes only and does not constitute Agency policy. 3-8 DRAFT--DO NOT CITE OR QUOTE 1 high dose to low dose extrapolations. This behavior is essentially a result of dose-dependent 2 hepatic processes, as described below. 3 4 3.3.2.3. M eta b o lism a n d P rotein B in d in g 5 The metabolism of TCDD is slow, particularly in humans, and it is thought to be 6 mediated by the CYP1A2 enzyme that is inducible by TCDD (Olson et al., 1994, 198008; 7 Ramsey et al., 1982, 548750; Weber et al., 1997, 548753; Wendling et al., 1990, 548751). The 8 low rate of metabolism in combination with sequestration appear to account for the retention of 9 TCDD in liver, and these processes collectively contribute to the long half-life for elimination of 10 TCDD from the body. 11 Dynamic changes in TCDD binding in liver and partitioning to fat have been studied 12 extensively in rats and mice (Diliberto et al., 1995, 197309; 2001, 197238). Figure 3-1 shows 13 observations by Diliberto et al. (1995, 197309) of the ratio of liver concentrations to adipose 14 tissue concentrations for mice given doses spread over a 100-fold range and studied at four 15 different times following exposure. It can be seen that even for the lowest dose studied the 16 liver:fat concentration ratio is higher than would be expected based on the lipid contents of the 17 tissues (i.e., 0.06:1, corresponding to the ratio of human liver:blood and fat:blood partition 18 coefficients; see Table 3-1). Moreover, the relative concentration in the liver consistently rises 19 with dose, with the steepest rise observed during the first two weeks after dosing. If the 20 distribution of TCDD were governed solely by passive partitioning into fat, there should be no 21 such change in relative concentrations with dose. However, data presented in Figure 3-1 22 illustrate that at longer time points, the ratio of TCDD in the liver to TCDD in fat decreases, 23 indicating that a redistribution of the chemical occurs as time goes on for each applied dose. The 24 redistribution of TCDD tissue levels from liver to fat with increasing time suggests that binding 25 of the chemical in the liver (including via induction of CYP1A2) is an important kinetic 26 consideration at early exposure points with relatively high applied doses. 27 Experiments with CYP1A2 "knock-out" mice (i.e., congenic strains differing in only a 28 single gene that is "knocked out" in one of the strains) indicate that the inducible binding of 29 TCDD is attributable to CYP1A2 (Diliberto et al., 1997, 548755; 1999, 143713). As noted 30 previously, this enzyme is believed to make an important contribution to metabolism of TCDD. 31 Given the critical role of CYP1A2 induction in the kinetics of TCDD, dose-and time-dependent This document is a draftfor review purposes only and does not constitute Agency policy. 3-9 DRAFT--DO NOT CITE OR QUOTE 1 induction of this protein in rats has been examined and modeled (Emond et al., 2004, 197315; 2 Emond et al., 2006, 197316; Santostefano et al., 1998, 200001; Wang et al., 1997, 104657). 3 Accordingly, the amount of CYP1A2 in the liver can be computed as the time-integrated product 4 of inducible production and a simple first-order loss process (Wang et al., 1997, 104657) : 5 6 2AL= S(t)Ko - K 2 CA2 t (Eq. 3-3) at 7 8 where CYP2Ai is the concentration of the enzyme, K2is the rate constant for the first order loss, 9 CA2t is the concentration of CYP1A2 in the liver, K0 is the basal rate of production of CYP1A2 in 10 the liver, and S(t) is a multiplicative stimulation factor for CYP1A2 production in the form of a 11 Hill-type function: 12 13 S(t) = 1 + InA2(CAh-TCDD) (Eq. 3-4) (ICa2)h + (CAh_TCDD)h 14 15 where ICA2 corresponds to the concentration of the aryl hydrocarbon (Ah)-TCDD complex at 16 which half of the maximum fold stimulation of CYP2A production is reached, and h, the Hill 17 exponent, determines the curvature of the stimulation in relation to concentration of the 18 Ah-TCDD complex at relatively low doses. A value of 0.6 as the Hill exponent has been used by 19 Wang et al. (1997, 104657; 2000, 198738) and Emond et al. (2004, 197315; 2005, 197317; 2006, 20 197316), indicative of a negative cooperation, i.e., the curve is convex-upward (supralinear), 21 depicting a faster increase in the low-dose region compared to a straight line. Additional 22 parameters in this expression include InA2, the maximum fold increase in the CYP1A2 synthesis 23 rate over the basal rate that can occur at high levels of TCDD, and (CAh-TCDD), the concentration 24 of TCDD bound to the aryl hydrocarbon receptor (AhR). This concentration in turn depends on 25 the concentration of TCDD in the liver (CLif), the concentration of the AhR (AhLi) in liver, and 26 the dissociation constant for the Ah-TCDD receptor complex, KDAh: 27 28 CAh-TCDD AhL XCLif (Eq. 3-5) KDAh + CLif 29 This document is a draftfor review purposes only and does not constitute Agency policy. 3-10 DRAFT--DO NOT CITE OR QUOTE 1 3.3.2.4. E lim in ation 2 Elimination half-lives (i.e., the time taken for the concentration to be reduced to one-half 3 of its initial level) of TCDD range from 11 days in the hamster to 2,120 days in humans 4 (U.S. EPA, 2003, 537122). Hepatic metabolism and binding processes, fecal excretion, and 5 accumulation in adipose tissue collectively determine the dose-dependent elimination half-lives 6 in various species. Aylward et al. (2005, 197114) depicted the relationship between the 7 elimination rate versus initial level of lipid-corrected TCDD in serum for 36 people (see 8 Figure 3-2). Even though this analysis was done using the initial TCDD level, rather than the 9 geometric mean or midpoint level in the decline for each person, it indicated a 10 concentration-dependency of the half-life and elimination of TCDD in exposed individuals. 11 12 3.3.2.5. In terspecies D ifferen ces a n d S im ilarities 13 Among the pharmacokinetic determinants of TCDD, some are known to vary markedly 14 between species whereas others are not characterized sufficiently in this regard. Overall, the 15 qualitative determinants of the body burden and elimination half-lives appear to be similar across 16 species. Based on empirical observations for TCDD as well as with other PCDFs, Carrier et al. 17 (1995, 197618; 1995, 543780) argued that in rats, monkeys, and humans, the dose-dependent 18 changes in the fraction contained in liver and adipose tissue follow a similar pattern across 19 species. The authors suggested that the half-saturation body burden is around 100 ng/kg and the 20 plateau of liver dose (as fraction of body burden) appears to occur around 1,000 ng/kg. 21 Literature also indicates that AhR is conserved phylogenetically (Fujii-Kuriyama et al., 1995, 22 543727; Harper et al., 2002, 198124; Nebert et al., 1991, 543728) and is present in mammalian 23 species, including experimental animals and humans (Lorenzen and Okey, 1991, 198397; 24 Manchester et al., 1987, 198054; Okey et al., 1994, 548759; Roberts et al., 1985, 198706; 25 Roberts et al., 1986, 198780). These qualitative similarities in pharmacokinetic determinants and 26 outcome support the use of animal data to infer general patterns of the pharmacokinetic behavior 27 of TCDD in humans. However, quantitative differences in determinants, including 28 physiological, physicochemical, and biochemical, need to be taken into account. Even though 29 species-specific physiological parameters can be obtained from the literature, key data on 30 species-specific biochemical parameters (particularly binding constants, maximal capacity, 31 induction rates, and other parameters) are not available for humans at this time. However, these This document is a draftfor review purposes only and does not constitute Agency policy. 3-11 DRAFT--DO NOT CITE OR QUOTE 1 can be inferred by using a pharmacokinetic model fit to in vivo data on the rate of TCDD 2 elimination from specific compartments in humans (Aylward et al., 2005, 197014; Carrier et al., 3 1995, 197618; Carrier et al., 1995, 543780; Emond et al., 2004, 197315; Emond et al., 2005, 4 197317; Emond et al., 2006, 197316). 5 6 3.3.3. PK of TCDD in Humans: Interindividual Variability 7 TCDD pharmacokinetics and tissue doses vary across the human population as a function 8 of the interindividual variability of the key kinetic determinants. Because the NAS comments 9 focused on health effects associated with chronic, lifetime exposure, the key kinetic determinants 10 for such exposures include clearance, binding, and temporal changes in volume of distribution. 11 When considering the interindividual variability in pharmacokinetics and dose metrics of TCDD, 12 it is important to recognize that the elevated lipid-corrected serum concentrations in highly 13 exposed persons are associated with greater elimination rates, probably due to greater degrees of 14 induction of CYP1A2 in the liver and possibly other related metabolic enzymes (Abraham et al., 15 2002, 197034; Aylward et al., 2005, 197014; Emond et al., 2006, 197316; Grassman et al., 2000, 16 548762). 17 The interindividual variability in fat content is a critical parameter in pharmacokinetic 18 models given the characteristics of TCDD (see Section 3.3.2). Both metabolic elimination and 19 elimination via the GI tract depend on the fraction of TCDD in the body that is available outside 20 of adipose tissue. As body fat content rises, a smaller portion of the total body TCDD will be 21 contained in the relatively available fraction outside of the adipose tissue. Because elimination 22 of TCDD by both metabolism and fecal excretion depends on the small proportion of TCDD that 23 exists outside of fat tissue, people with larger proportions of body fat--including many older 24 people--will tend to require longer times to reduce TCDD levels by a given proportion than 25 leaner people (Emond et al., 2006, 197316; Rohde et al., 1999, 548764; Van der Molen et al., 26 1998, 548765; Van der Molen, et al., 1996, 548768). 27 The sections that follow highlight key aspects of interindividual variability in TCDD 28 pharmacokinetics, with an emphasis on the available data related to elimination half-lives and 29 volume of distribution. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 3-12 DRAFT--DO NOT CITE OR QUOTE 1 3.3.3.I. L ife S ta g e a n d G ender 2 The influence of the variability of fat content in human population on the distribution and 3 clearance of TCDD has been evaluated by several investigators. There are data showing an 4 inverse dependency of TCDD elimination rate on percent body fat. Figure 3-3 shows this 5 relationship in a study in which TCDD elimination via feces was measured in six people in 6 relation to their body fat content (Rohde et al., 1999, 548764). Observations of TCDD 7 elimination rates in a small number of men and women in the Seveso cohort (Aylward et al., 8 2005, 197114) provide a modest opportunity to compare TCDD elimination rates with actual 9 human data. Based on the partition coefficients reported by Emond et al. (2006, 197316), the 10 elimination rates for the men in the sampled group are expected to be greater than the elimination 11 rates in the women. Taking into consideration calculations similar to those shown in Table 3-2, 12 and fat proportions inferred from body mass indices using the equations of Lean et al. (1996, 13 548770), the Seveso men studied are expected to have an overall average of about 3.92% of their 14 TCDD body burden outside of fat, whereas the women are expected to have an average of only 15 2.36% outside of fat. On this basis, the TCDD elimination rates in the men are expected to be 16 3.92/2.36 = 1.66 times faster than the elimination rates in the women. By comparison, Michalek 17 et al. (2002, 199579) reported observed elimination rates in men and women that result in a 18 slightly lower ratio: 19 20 men:0.111 year 1- 0.010 (std.error) = 1.56 (Eq. 3-6) women:0.071 year 1 0.010 (std.error) 21 22 The central estimates for the elimination rates correspond to half lives of 6.5 and 9.6 years for 23 men and women, respectively. 24 A further point of comparison can be derived using the observed body mass index 25 (BMI)12 and TCDD elimination rate of each of the male Ranch Hand military veterans, whose 26 TCDD elimination rates were observed between 9 and 33 years after their time in Vietnam. The 27 average BMI over that time was 29.44 (based on 287 measurements for the 97 veterans, 28 tabulated in three periods by Michalek et al., 2002, 199579), and their average age was about 12The body mass index, or BMI, is calculated as the body weight in kilograms divided by the square of the height in meters. This document is a draftfor review purposes only and does not constitute Agency policy. 3-13 DRAFT--DO NOT CITE OR QUOTE 1 44.5 for the measurements. Based on these data, the corresponding average estimated percent 2 body fat is 29.7% using the Lean et al. (1996, 548770) formula for men. The observed average 3 TCDD elimination rate constant for these men for the period was 0.092 year-1 0.004 (standard 4 error), corresponding to a half life of 7.5 years. This half life is slightly longer than the central 5 estimate of the half life of 6.2 years (i.e., ln(2)/0.111) for the smaller group of Seveso males with 6 their slightly smaller estimated percent body fat. Figure 3-4 shows a simple plot of these data 7 and a fitted unweighted regression line characterizing the relationship between estimated fat 8 content and TCDD elimination rates. Variation in metabolic enzyme activities and other routes 9 of loss is also likely to be important, but there is little human quantitative information available 10 on these issues. 11 More recently, Kerger et al. (2006, 198651) estimated the slope of the relationship 12 between half-life and age to be 0.12 years (95% confidence interval, 0.10-0.14), which 13 corresponds to the rate of increase in TCDD half-life for each year of age. The authors 14 speculated that although age explained most of the variance in the individual half-life trends, it 15 was also correlated with TCDD concentration, BMI, and body fat mass. The regression model 16 developed by these authors discriminated between the high and low TCDD exposures or 17 concentrations. Thus, after accounting for the TCDD (concentration x age) term's effect on the 18 slope of age, the final model for TCDD concentration <700 ppt was 19 20 t1 / 2 = 0.35 + 0.12 x Age 21 For TCDD concentration >700 ppt, the final model was: (Eq. 3-7) 22 23 t1 / 2 = 0.35 + 0.088 x Age (Eq. 3-8) 24 25 where t1/2 is the half-life and Age is the age at time of subsequent sampling. Pharmacokinetic 26 information relevant to specific age groups is presented in the sections that follow. 27 28 3.3.3.1.1. P ren a ta l period. 29 Data to estimate TCDD elimination rates for fetuses are not available. Levels of TCDD 30 in fetal tissues for rats were experimentally estimated at different gestational periods and utilized 31 in a developmental model by Emond et al. (2004, 197315). There is information on body This document is a draftfor review purposes only and does not constitute Agency policy. 3-14 DRAFT--DO NOT CITE OR QUOTE 1 composition that is relevant to prediction of TCDD dose to fetus. These data, summarized as 2 part of the radiation dosimetry model of the International Commission on Radiological 3 Protection, are consistent with the idea that early fetuses are nearly all water and less than 4 1% lipid, and lipid levels rise toward parity with protein near the time of normal delivery. 5 Bell et al. (2007, 197050) reported that the disposition of TCDD into the fetus shows 6 dose dependency, with a greater proportion of the dose reaching the fetus at lower doses of 7 TCDD. Further, both CYP1A1 and CYP1A2 are highly inducible (~103-fold) in fetal liver, 8 whereas CYP1A2 shows much lower induction (10-fold) in maternal liver. It has been 9 speculated that this is due to the lower basal levels of CYP1A2 in fetal liver, as compared to 10 maternal liver (Bell et al., 2007, 197050). The greater relative disposition to the fetus at low 11 doses may be the result of higher bioavilalibty due to less hepatic sequestration and elimination 12 in the mother. 13 14 3.3.3.I.2. In fan cy a n d childhood. 15 Hattis et al. (2003, 548773) describe the general pattern of change of body fat content 16 with age in children. Central tendency values for percent body fat begin at about 12% at birth 17 and rise steeply to reach about 26% near the middle of the first year of life. Fat content then falls 18 to reach a minimum of approximately 15% at 5-8 years of age, followed by a sex-dependent 19 "adiposity rebound" that takes females to about 26% body fat while the males remain near 20 16-17% on average by age 20. The interindividual variability distributions about these central 21 values are complex, as some children experience the "adiposity rebound" earlier than others, and 22 this creates patterns that are not simply interpretable as unimodal normal distributions. Hattis et 23 al. (2003, 548773) did find it possible to fit distributions of body fat content inferred from 24 NHANES skin fold measures to mixtures of two normal distributions for children between age 5 25 and 18. 26 At least two groups of authors have published PBPK modeling results indicating 27 generally more rapid clearance of TCDD in children than in adults, a trend that is consistent with 28 the generally lower fat content of children (Kreuzer et al., 1997, 198088; Leung et al., 2006, 29 548779; Van der Molen et al., 2000, 548777). The rapid expansion of the adipose tissue 30 compartment can contribute, in part, to the reduced apparent half-life in children (Clewell et al., This document is a draftfor review purposes only and does not constitute Agency policy. 3-15 DRAFT--DO NOT CITE OR QUOTE 1 2004, 056269). This reduction may also be due to varying rates of metabolism and/or fecal lipid 2 excretion (Abraham et al., 1996, 548782; Kerger et al., 2007, 548784). 3 Furthermore, very young children have different modes and quantities of exposure 4 compared to adults. Lakind et al. (2000, 198094) characterize distributions of milk intake for 5 nursing infants to characterize distributions of TCDD exposure. This is also a corresponding 6 route of loss of TCDD stores for lactating women, as described in Section 3.3.3.2 below. 7 8 3.3.3.I.3. A d u lth o o d a n d o ld age. 9 The fraction of fat in relation to body weight in adulthood and old age can be computed 10 as a function of the BMI and age (e.g., Lean et al., 1996, 548770) : 11 % Body Fat (males) = 1.33 x BMI + 0.236 x Age - 20.2 (Eq. 3-9) 12 13 % Body Fat (females) = 1.21 x BMI + 0.262 x Age - 6.7 (Eq. 3-10) 14 15 The above equations are the result of analysis of data based on underwater weighing of 16 63 men and 84 women (age range 16.8-65.4). The salient observation with respect to TCDD for 17 these data is that age and BMI-dependent variability in fat content have implications for the 18 variability in TCDD elimination rates and internal dose among adults. 19 20 3.3.3.2. P h ysio lo g ica l S tates: P reg n a n cy a n d L actation 21 Data on body fat content in pregnant women at various stages of gestation (Pipe et al., 22 1979, 548786) have potential implications for TCDD elimination rates during pregnancy, even 23 though the relationship between these parameters has not been formally analyzed. 24 Lactation is viewed as an additional route of elimination for some chemicals such as 25 TCDD. According to a recent study, a breast-feeding woman expels through lactation an 26 estimated 8.76 kg fat per year [qf (kg/day), 0.8 kg milk/day with an average 3% lipid], and the 27 partition coefficient between blood lipid and milk fat (ABM) for TCDD is 0.92 (Milbrath et al., 28 2009, 198044; Wittsiepe et al., 2007, 548736). The estimated rate of elimination of TCDD due 29 to breast-feeding (kbfed) can then be computed as follows (Milbrath et al., 2009, 198044): 30 This document is a draftfor review purposes only and does not constitute Agency policy. 3-16 DRAFT--DO NOT CITE OR QUOTE 1 kbfed - qf XAtbfed (Eq. 3-11) Kbmx pbf- x BW 100 2 3 where 4 Atbfed (unitless) = the fraction of the year during which the woman was actively breast 5 feeding; 6 pbf = woman's percent body fat; and 7 BW = woman's body weight in kg. 8 9 Assuming no interaction between breast-feeding and other half-life determinants 10 Milbrath et al. (2009, 198044), the authors predicted a half-life of 4.3 years for TCDD in a 11 30-year-old, nonsmoking woman with 30% body fat if she did not breast-feed that year, and a 12 half-life of 1.8 years if she breast-fed for 6 months. 13 3.3.3.3. L ifestyle a n d H abits 14 One of the factors related to lifestyle and habits that could influence TCDD kinetics is 15 smoking. Smoking has been reported to enhance the elimination of dioxin and dioxin-like 16 compounds (Ferriby et al., 2007, 548789; Flesch-Janys et al., 1996, 197351). Milbrath et al. 17 (2009, 198044) accounted for interindividual variation in body composition as well as smoking 18 habits in an empirical model. The predicted half-life (years) for an individual i as a function of 19 age, smoking status, and percent body fat i was as follows 20 21 t1/2 (age, Sm0ke, PbfX - lfi(0age) + P(age) Xagei] XSFi X --- (Eq. 3-12) Pbfref( age,) 22 23 where 24 P(0age) = intercept constant derived from regressed data; 25 J3(age) = slope constant derived from regressed data; 26 agei = specific age i (years); 27 pbfi = individual percent body fat; This document is a draftfor review purposes only and does not constitute Agency policy. 3-17 DRAFT--DO NOT CITE OR QUOTE 1 pbfref(age.) = reference percent body fat; and 2 SFi = the unitless, multiplicative smoking factor. 3 4 3.3.3.4. G enetic Traits a n d P olym orph ism 5 One particular genetic locus that is potentially related to TCDD pharmacokinetics and 6 tissue dose is the gene for the AhR. Eight candidate AhR polymorphisms have been identified to 7 date (Connor and Aylward, 2006, 197632; Harper et al., 2002, 198124). Given the role of AhR 8 in regulating the induction of CYP1 isozymes (Baron et al., 1998, 548791; Connor and Aylward, 9 2006, 197632; Toide et al., 2003, 548792), the polymorphism might lead to interindividual 10 differences in metabolic clearance, the significance of which would depend upon the dose, fat 11 content, and exposure scenario. In this regard, it should be noted that the inducibility of aromatic 12 hydrocarbon hydroxylase in human tissues has been reported to be highly variable, up to 13 100-fold (Connor and Aylward, 2006, 197632; Smart and Daly, 2000, 548794; Wong et al., 14 1986, 548795). 15 Finally, the scientific literature contains values of K& (the dissociation constant of the 16 TCDD-AhR complex) ranging from about 1 to much higher values (corresponding to lower 17 binding affinity) (reviewed in Connor and Aylward, 2006, 197632). This provides suggestive 18 evidence for a heterogeneous human AhR, with functionally important polymorphisms (Micka et 19 al., 1997, 548797; Roberts et al., 1986, 198780), even though some of the range may be 20 attributed to experimental procedural differences and to other factors (Connor and Aylward, 21 2006, 197632; Harper et al., 2002, 198124; Lorenzen and Okey, 1991, 198397; Manchester et 22 al., 1987, 198054). 23 The various pharmacokinetic processes and determinants (see Sections 3.3.2 and 3.3.3), 24 individually or together, might influence the dose metrics of relevance to the dose-response 25 modeling of TCDD. 26 27 3.3.4. Dose Metrics and Pharmacokinetic Models for TCDD 28 3.3.4.1. D o se M etrics f o r D ose-R espon se M o d elin g 29 The dose metric related to a toxicologic endpoint can range from the maximal 30 concentration, the area under a time-course curve (AUC), or the time-averaged concentration of This document is a draftfor review purposes only and does not constitute Agency policy. 3-18 DRAFT--DO NOT CITE OR QUOTE 1 the toxic moiety in the body, blood, or target tissue, to an appropriate measure of the resulting 2 interactions in the target tissue (e.g., receptor occupancy or functional biomarkers related to 3 specific effects). A single dose metric, however, is unlikely to be sufficient for all endpoints and 4 exposure durations. Further, the ideal dose metric chosen on the basis of the mode of action 5 (MOA) may not be the dose metric for which model predictions can be obtained with a high 6 level of confidence. Consideration of these issues is critical to the selection of the dose metrics 7 of relevance to dose-response modeling of TCDD. 8 Figure 3-5 lists a range of alternative dose metrics for TCDD in terms of their relevance 9 based on considerations of pharmacokinetic mechanisms and MOA. The administered dose or 10 daily intake (ng/kg-day) is the least relevant dose metric for dose-response modeling o f TCDD. 11 This dose adjusts only for body weight differences between species. The administered dose, 12 when used with an uncertainty factor for kinetics (or kinetic adjustment factor, such as BW34) 13 and an uncertainty factor for dynamics, can also account for allometrically-predicted 14 pharmacokinetic (clearance) and pharmacodynamic differences between species in deriving the 15 human equivalent dose (HED). In effect, the use of kinetic and dynamic adjustment or 16 uncertainty factors facilitates the computation of HED. Such a calculation of HED is associated 17 with the steady-state blood concentration o f parent chemical in rats by accounting for species 18 differences in metabolic clearance. This is generally done by relating to body surface area or 19 metabolic rates, with no corresponding temporal changes in the volume o f distribution (see, for 20 example, Krishnan and Andersen, 1991, 548799). Such calculations of HED for TCDD may not 21 be appropriate given that (1) steady-state was not attained in all critical toxicological studies 22 chosen for the assessment, (2) the clearance is mainly due to enzyme(s) and processes whose 23 levels/rates do not necessarily vary across species or life stages as a function of body surface 24 differences, and (3) there is a likelihood of change in volume of distribution over time. 25 Furthermore, the use o f administered dose does not explicitly account for the dose-dependent 26 elimination o f TCDD from tissues as demonstrated in multiple studies (reviewed in 27 Sections 3.3.2 and 3.3.4). The use of administered dose in TCDD dose-response modeling is 28 unlikely to facilitate the characterization o f the true relationship between the response and the 29 relevant measures o f internal dose that are influenced by dose-dependent elimination and binding 30 processes. Additionally, the use o f administered dose to extrapolate across species or life stages This document is a draftfor review purposes only and does not constitute Agency policy. 3-19 DRAFT--DO NOT CITE OR QUOTE 1 would not effectively take into account the differences in fat content or the demonstrated dose2 dependent and species-dependent differences in elimination half-life of TCDD. 3 Dose metrics for TCDD may include absorbed dose, body burden, serum or whole blood 4 concentration, tissue concentration, and possibly functional-related metrics of relevance to the 5 MOA (e.g., receptor occupancy, change in protein levels). These measures can be calculated as 6 a current (terminal), average (over a defined period), or integral quantity. The applicability of 7 the integral measures, such as the AUC (i.e., the area under the curve of a plot of blood or 8 plasma concentration vs. time), traditionally used for analyzing chronic toxicity data, is 9 questionable in the case of TCDD. This is because of differences in lifespan and uncertainties 10 regarding the appropriateness of the duration to be specified for averaging the AUC in 11 experimental animals and humans for certain critical effects (NAS, 2006, 198441). 12 Among the alternative dose metrics, the absorbed dose accounts for differences in body 13 weight as well as species-specific differences in bioavailability. Thus, the absorbed dose is 14 equivalent to body burden. Body burden, or more appropriately the body concentration, 15 represents the amount of TCDD per kg body weight. TCDD body burdens, like other dose 16 measures, can be determined as the peak, the average over the period of the bioassays, or the 17 level at the end of the experiments. Thus, the terminal or average body burdens can be obtained 18 either using data or pharmacokinetic models and used in dose-response modeling. The body 19 burden is a measure of TCDD dose that reflects the net impact of bioavailability, uptake, 20 distribution, and elimination processes in the organism. It is essentially a function of the volume 21 of distribution and clearance processes, and as such it does take into account the temporal 22 changes in volume of distribution as well as the concentration-dependent clearance. These are 23 phenomena that are critical to the understanding of TCDD dose to the target. However, the body 24 burden may not accurately reflect the tissue dose (NAS, 2006, 198441), and as such does not 25 allow for analysis of species-specific differences in target organ sensitivity to TCDD. In 26 essence, the body burden represents only an "overall average" of TCDD concentration in the 27 body, without regard to the differential partitioning and accumulation in specific tissues, 28 including the target tissue(s). 29 Serum (or blood) concentration of TCDD is a dose metric that reflects both the body 30 burden and the dose to target tissues. Serum or blood concentration, at steady-state, would be 31 reflective of the impact of clearance processes, and expected to be directly proportional to the This document is a draftfor review purposes only and does not constitute Agency policy. 3-20 DRAFT--DO NOT CITE OR QUOTE 1 tissue concentrations of TCDD (NAS, 2006, 198441). This dose metric for lipophilic chemicals 2 such as TCDD is often expressed as a lipid-normalized value, to adjust for varying serum lipid 3 content (e.g.; DeKoning and Karmaus, 2000, 548801; Niskar et al., 2009, 548802) (Patterson et 4 al., 2009), particularly in human biomonitoring studies, thus of relevance to dose-response 5 modeling; however, the serum lipid-normalized concentrations of TCDD are not routinely 6 collected and reported in animal toxicologic studies. Serum lipid-adjusted of TCDD 7 concentration is calculated as the ratio of serum TCDD content over serum lipid content per unit 8 volume. Altenratively, TCDD serum lipid-normalized calculation can be estimated by using the 9 formula TL = (2.27 x TC) +TG + 62.3 mg/dL where the total lipid (TL) content of each sample 10 is estimated from its total cholesterol (TC) and triglyceride (TG) (Patterson et al., 2009). The 11 lipid-adjusted serum concentration, however, would be reflective of the lipid-adjusted 12 concentration of TCDD in other organs (reviewed in Aylward et al., 2008, 197068) depending 13 upon the extent of steady-state attained and the similarity of lipid composition across tissues in 14 each species. In essence, the serum lipid-normalized measure is representative of the amount of 15 TCDD per specified volume of total lipids, whereas the whole blood measure will be reflective 16 of the ensemble of free, lipid-bound and protein-bound TCDD in plasma and erythrocytes, which 17 may be species-specific. Even though these dose metrics are thought to be more closely and 18 directly related to the tissue concentrations associated with an effect, a less direct association 19 might occur at increasing doses when nonlinear processes dominate the kinetics and distribution 20 of TCDD into organs such as the liver. 21 Tissue concentration of TCDD, as free, bound, or total TCDD, is a more relevant 22 pharmacokinetic measure of dose, given that it provides a measure of exposure of the target cells 23 to the chemical. In this regard, the CYP1A2-bound fraction may be considered as a relevant 24 dose metric for certain toxic effects; however, the available data contain mixed results regarding 25 the mechanistic linkage of this dose metric to toxicity and carcinogenicity (reviewed in Budinsky 26 et al., 2006, 594248). In such cases, the use of alternative dose metrics (e.g., bound 27 concentration as well as the serum concentration) in dose-response modeling could be 28 considered. Other function-related biomarkers and dose metrics could facilitate the additional 29 consideration of pharmacodynamic aspects reflecting tissue- and species-specific sensitivity. 30 These metrics represent the most relevant measures of tissue exposure and sensitivity to TCDD. This document is a draftfor review purposes only and does not constitute Agency policy. 3-21 DRAFT--DO NOT CITE OR QUOTE 1 Empirical time-course data on the alternative dose metrics of TCDD associated with 2 epidemiologic and experimental (animal) studies are not available, requiring the use of 3 pharmacokinetic models to obtain estimates of these dose metrics. These models may be simple, 4 based on first order kinetics (see Section 3.3.4.2), or more complex based on physiochemical, 5 biochemical, and physiological parameters for simulating uptake, distribution (including 6 sequestration to proteins), and clearance of TCDD (see Section 3.3.4.3). Receptor occupancy 7 and functional biomarkers as dose metrics for TCDD require a clear understanding of mode of 8 action of TCDD and availability of relvant data. In the absence of such information, these 9 possible dose metrics can not be utilized at the present time. 10 11 3.3.4.2. F irst-O rder K in etic M odelin g 12 Figure 3-6 illustrates the process of estimating a human-equivalent TCDD oral exposure 13 from an experimental animal-administered dose, based on the assumption that body burden is the 14 effective dose metric for TK equivalence across species. The primary assumption is that the 15 time-weighted average (TWA) TCDD body burden over some critical time period is the 16 proximate toxicokinetically-effective dose eliciting a toxicologic effect.13 The process consists 17 of estimating the effective average body burden in the experimental animal over some time tA 18 (generally the experimental duration) using a TK model, then "back-calculating" a daily human 19 exposure level that would result in that average body burden over some time tH(the human 20 equivalent to tA). 21 The following closed-form equation is the general formula used to calculate a TCDD 22 terminal body burden in an experimental animal or human at time (t). 23 24 25 26 where BB(t) =BB(0) + d(1 ~ e )fa k 27 BB(t) = the body burden at time t (ng/kg); 28 BB(0) = the initial body burden (ng/kg); 29 d = the daily dose (ng/kg-day); 30 k = the whole-body elimination rate (days-1); (Eq. 3-13) 13The conversion depicted in Figure 3-6 does not account for toxicodynamic differences between species. This document is a draftfor review purposes only and does not constitute Agency policy. 3-22 DRAFT--DO NOT CITE OR QUOTE 1 t = the time at which the body burden is determined (days); and 2 fa = the fraction of oral dose absorbed (unitless). 3 , d q(1 - e ~ k --'A ) fa 4 For the experimental animal, B B ( t) is B B a (t) = B B a (0)e A A +-- ------------------- -- , and for kA 5 humans, this parameter is B B H (t ) = B B H (0)e_k H tH + d H (1--e-------- )f * H . kH 6 7 Setting B B H(t) = B B A(t) obtains the following expression: 8 9 B B h (0)e- k H 'H + d H (1 - e ~ k H lH >f a H = b b a (0)e- k --`A + d --(1 - e ^ kH kA 10 11 Rearranging yields the general solution for dH. 12 )f a -- (Eq. 3-14) 13 kH dH = dA fa A (1 - g- A > A ) + BBa (0)e k A 'A B B h (0)e - kH'H (Eq. 3-15) k A f a H (1- e ~ kH 'H ) 14 15 Assuming that initial body burdens are very small compared to BB(t) and that the fraction of 16 TCDD absorbed is the same for humans and experimental animals, and using the relationship k =M S 17 'i;2 , where ty2is the whole-body half-life, a simplified solution for dHis obtained. 18 19 dH d AhQA. (1 - e kA'A) (Eq. 3-16) ' 1/2h (1- e^kH`H) 20 21 The term 1-e~ktis the daily fraction eliminated. Therefore, dH can be seen to be the 22 average daily administered dose to the experimental animal times the ratio of the animal:human 23 half-life times the ratio of the animal:human daily fraction eliminated over the respective times, 24 tA and tH. For both species at (theoretical) steady state (t ^ <x>; daily fraction eliminated ^ 1), 25 the latter ratio approaches unity, reducing the animal:human conversion factor to the ratio of the This document is a draftfor review purposes only and does not constitute Agency policy. 3-23 DRAFT--DO NOT CITE OR QUOTE 1 half-lives. The latter approach was used in the 2003 Reassessment for conversion of animal 2 cancer slope factors to the human equivalent, where only lifetime exposures are relevant.14 3 However, for less-than-lifetime exposures eliciting noncancer effects, specific values for 4 tA and tHmust be considered. Furthermore, Eq. 3-16 computes dHon the basis of terminal body 5 burdens at times tA and tH. The more representative metric for toxicokinetic equivalence based 6 on average body burden over the respective time periods is given in Eq. 3-17. 7 L eH+dia[,-L id8 BB(t) = BB(0) 1 fe-kTdr +d f a 1 f (1- e-kT)dz =BB(0) (1- e ,+ d (Eq. 3-17) tJ k tJ kt k kt 9 10 On the basis of average body burden as given in Eq. 3-17, is transformed again assuming 11 minimal initial body burden (BB(0) ~ 0), as follows: 12 1 (l - e- k At A ) 13 t1/ 2A k A tA (Eq. 3-18) t1/ 2H 1 o33 ^- 1 0T 1 1 T --i ____ i 14 15 where tH0 is the initial human exposure time. 16 The value of tA is the duration of the experimental exposure period. For some gestational 17 exposures, if a critical exposure window is defined, tA will be the duration of the critical 18 exposure window. The value of tHis the human-equivalent duration corresponding to tA. 19 However, for tA less than lifetime (less than 2 years in rodents) and no defined susceptible life 20 stage, tHcannot begin at 0 (because typically animal experiments do not begin at age 0), but must 21 end at 25,550 days (70 years) to include the terminal (pseudo) steady-state level, at which the 22 BBH(t): dHratio is highest. Otherwise, starting tHat 0 would not be protective for less-than23 lifetime effects that could be manifest at any age in humans; the average is determined from the 24 terminal end of the human exposure period because the daily exposure achieving the target blood 25 concentration is smaller than for the same exposure period beginning at birth (i.e., dHwould be 14No conversions to human-equivalent exposures were attempted for other effects in the 2003 Reassessment. This document is a draftfor review purposes only and does not constitute Agency policy. 3-24 DRAFT--DO NOT CITE OR QUOTE 1 higher for earlier exposure periods) and is health protective for effects occurring after 2 shorter-term exposure.15 Figure 3-7 depicts the relationship of daily dose to TWA body burden 3 graphically for several exposure duration scenarios. For shorter durations occurring later in life, 4 the average body burden over the exposure period does not differ substantially from the 5 steady-state value. Even for half-lifetime exposures, the deviation of the average from steady 6 state is minimal. Only for lifetime exposures does the difference become more marked, but only 7 by about 15%. Note that in the 2003 Reassessment, a constant value of 3,000 was used for 8 BBH(t): dH, based on the relationship of continuous exposure to theoretical steady-state body 9 burden (t = lifetime, ty2= 2,593 days); this approach, while conservative, does not account for 10 exposure scenarios of different durations and does not strictly reflect the average body burden 11 dose metric. 12 The simulation in Figure 3-7 is based on a unit daily exposure to humans, such that the 13 target body burden represents BBH(tH):dHas a general scalar for calculating dHfrom any given 14 dA. Table 3-3 shows the resulting TK conversion factors for the rodent species and strains 15 comprising the bulk of the experimental animals in TCDD studies. Monkey and mink values are 16 not shown in this table because, for the former, only chronic exposures were evaluated and, for 17 the latter, no TCDD half-life information is available. Monkey (Rhesus) half-life estimates 18 range from about 200-500 days. A representative value of 365 days is used for this TCDD 19 assessment. The dA to dHconversion factor for the chronic monkey exposures (3.5-4 years) in 20 TCDD studies is 9.2-9.7 (BBA:dA = 279-263). 21 Application of first order kinetics for the risk assessment of TCDD can only be used to 22 estimate total body burdens or back-calculate administered dose from experimental data. Body 23 burden calculations using first order kinetics is based on the assumption of a first order decrease 24 in the levels of administered dose as function of time. In that sense, any loss of TCDD from the 25 body is described by using a rate constant that is not specific to any biological process. This 26 constant is usually estimated from estimates of half-life of TCDD. Assuming a constant half-life 27 value for the clearance for long-term or chronic TCDD exposure is not biologically supported 28 given the observed data indicating early influence of CYP1A2 induction and binding to TCDD 29 and later redistribution of TCDD to fat tissue. Abraham et al. (1988, 199510) found that the 30 liver:adipose tissue concentration ratio in female Wistar rats exposed to a subcutaneous TCDD 15See the following section (3.3.4.3) for a more detailed discussion o f this concept. This document is a draftfor review purposes only and does not constitute Agency policy. 3-25 DRAFT--DO NOT CITE OR QUOTE 1 dose of 300 ng/kg decreased from 10.3 at 1 day postexposure to 0.5 at 91 days postexposure. 2 Consequently, using half-life estimates based on observed steady-state levels of TCDD will not 3 account for the possibility of accelerated dose-dependent clearance of the chemical at the early 4 stages and thus would result in estimation of lower administered levels of the chemical. The 5 dynamic change in half-life due to dose-dependent elimination at the early stages of TCDD 6 exposure and its later redistribution to fat tissues for steady-state levels is better described using 7 biologically-based models, such as the PBPK models and concentration- and age-dependent 8 elimination (CADM) models (Aylward et al., 2005, 197014; Carrier et al., 1995, 197618; Carrier 9 et al., 1995, 543780; Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al., 10 2006, 197316). Additionally, these models provide estimates for other dose metrics (e.g., serum 11 or tissue levels) that are more biologically relevant to response than administered dose or total 12 body burden (see Section 3.3.4.3). 13 14 3.3.4.3. B io lo g ica lly-B a sed K in etic M o d els 15 The development and evolution of biologically-based kinetic models for TCDD have 16 been reviewed by EPA (2003, 537122) and Reddy et al. (2005, 594251). The initial PBPK 17 model of Leung et al. (1988, 198815) was developed with the consideration of TCDD binding to 18 CYP1A2 in the liver. The next level of PBPK models by Andersen et al. (1993, 196991) and 19 Wang et al. (1997, 104657) used diffusion-limited uptake and described protein induction by 20 interaction of DNA binding sites. The models of Kohn et al. (1993, 198601) and Andersen et al. 21 (1997, 197172) further incorporated extensive hepatic biochemistry and described zonal 22 induction of CYP by TCDD. TCDD PBPK models have evolved to include detailed descriptions 23 of gastrointestinal uptake, lipoprotein transport, and mobilization of fat, as well as biochemical 24 interactions of relevance to organ-level effects (Kohn et al., 1996, 022626; Roth et al., 1994, 25 198063). Subsequently, developed PBPK models either used constant hepatic clearance rate 26 (Maruyama et al., 2002, 198448; Wang et al., 1997, 104657; Wang et al., 2000, 198738) or 27 implemented varying elimination rates as an empirical function of body composition or dose 28 (Andersen et al., 1993, 196991; Andersen et al., 1997, 197172; Kohn et al., 1996, 022626; 29 Van der Molen et al., 1998, 548765; Van der Molen et al., 2000, 548777). The more recent 30 pharmacokinetic models explicitly characterize the concentration-dependent elimination of 31 TCDD (Aylward et al., 2005, 197014; Carrier et al., 1995, 197618; Carrier et al., 1995, 543780; This document is a draftfor review purposes only and does not constitute Agency policy. 3-26 DRAFT--DO NOT CITE OR QUOTE 1 Emond et al., 2004, 197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316). The 2 biologically-based pharmacokinetic models describing the concentration-dependent elimination 3 (i.e., the pharmacokinetic models of Aylward et al. (2005, 197014) and Emond et al. (2005, 4 197317; 2006, 197316) are relevant for application to simulate the TCDD dose metrics in 5 humans and animals exposed via the oral route. The rationale for considering the application of 6 Aylward et al. (2005, 197014) and Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) 7 models for estimating dose metrics for possible application to TCDD risk assessment is based on 8 the following considerations. 9 10 Both models represent research results from the more recent peer-reviewed publications. 11 Both models are relatively simple and less parameterized than earlier kinetic models for 12 TCDD. The Aylward et al. (2005, 197014) model is based on two-time scale TCDD 13 kinetics described by Carrier et al. (1995, 197618), and the Emond et al. (2004, 197315; 14 2005, 197317; 2006, 197316) PBPK models are reduced versions of earlier complex 15 PBPK models. Although simple, both the Aylward et al. (2005, 197014) and Emond et 16 al. (2004, 197315; 2005, 197317; 2006, 197316) models are still inclusive of important 17 kinetic determinants of TCDD disposition. 18 Both models are uniquely formulated with dose-dependent hepatic elimination consistent 19 with the physiological interpretations commonly accepted by the scientific community. 20 Both models and extrapolated human versions were tested against human data collected 21 in a variety of human exposure scenarios (Aylward et al., 2005, 197014; Emond et al., 22 2005, 197317). 23 Both models are capable of deriving one or more of the candidate dose-metrics that are of 24 interest to EPA's dose-response assessment of TCDD. 25 26 3.3.4.3.1. C A D M m odel. 27 3.3.4.3.1.1. M ode.l stru ctu re. 28 The pharmacokinetic model of Aylward et al. (2005, 197014), referred to as the CADM 29 model in this report, is based on an earlier model developed by Carrier et al. (1995, 197618; 30 1995, 543780) that describes the dose-dependent elimination and half-lives of polychlorinated 31 dibenzo-p-dioxins and furans. This model describes the TCDD levels in blood (body), liver, and 32 adipose tissue. Blood itself is not characterized physically as a separate compartment within the 33 model, and the distribution of TCDD to tissues other than adipose tissue and liver (usually less 34 than 4%) is not accounted for by the model. The original structure of the Carrier et al. (1995, This document is a draftfor review purposes only and does not constitute Agency policy. 3-27 DRAFT--DO NOT CITE OR QUOTE 1 197618; 1995, 543780) model was modified by Aylward et al. (2005, 197014) to include TCDD 2 elimination through partitioning from circulating lipids across the lumen of the large intestine 3 into the fecal content (see Figure 3-8). The most recent version of the Carrier model (Aylward et 4 al., 2005, 197014; 2008, 197068) includes fecal excretion of TCDD from two routes: 5 (1) elimination from circulating blood lipid through partitioning into the intestinal lumen; and 6 (2) elimination of unabsorbed TCDD from dietary intake. 7 A basic assumption of this model is that metabolic elimination of TCDD is a function of 8 its current concentration in the liver. The current concentration of TCDD in the liver increases 9 with increasing body burden in a nonlinear fashion as a result of the induction of (and binding of 10 TCDD to) specific proteins (i.e., CYP1A2). Consequently, the fraction of TCDD body burden 11 contained in the liver increases nonlinearly (with a corresponding decrease in the fraction 12 contained in adipose tissues) with increasing body burden of TCDD (Aylward et al., 2005, 13 197114; Carrier et al., 1995, 197618). 14 Of particular note is that the adipose tissue compartment of the model is considered to 15 represent the lipid contained throughout the body. It then assumes that the concentrations of 16 TCDD in lipids of plasma and various organs is essentially equivalent to that of adipose tissue, 17 and as such these concentrations are included in the adipose compartment of the model. Even 18 though this approximation is fairly reasonable given the available data, there is some concern 19 that the adipose compartment of this model also includes the lipid content of the liver to some 20 unknown extent. Removal of lipid volume from the liver would mathematically alter total 21 hepatic concentration and therefore would affect the estimated levels of the chemical available 22 for binding to proteins. 23 Distribution in the body is modeled to occur between hepatic and adipose/lipid 24 compartments, with the fraction of body burden in liver increasing according to a function that 25 parallels the induction of the binding protein CYP1A2. Elimination is modeled to occur through 26 hepatic metabolism (represented as a first-order process with rate constant K that decreases with 27 age) and through lipid-based partitioning of unmetabolized TCDD across the intestinal lumen 28 into the gut, which is also modeled as a first-order process. As the body burden increases, the 29 amount of TCDD in the liver increases nonlinearly, resulting in an increased overall elimination 30 rate. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 3-28 DRAFT--DO NOT CITE OR QUOTE 1 3.3.4.3.1.2. M a th e m a tic a l rep resen ta tio n . 2 The CADM model describes the distribution to tissues (including liver and adipose 3 tissue) based on exchange from blood at time intervals of one month. The model is based on 4 quasi-steady-state-approximation, and thus it is also based on the consideration that the 5 intertissue processes reach their equilibrium values "quasi-instantaneously." In this regard, 6 absorption and internal distribution reflective of kinetics at the cellular level (e.g., diffusion, 7 receptor binding, and enzyme induction) likely occur on a relatively fast time scale (a few hours 8 to a few days). However, the overall body concentration (i.e., body burden) varies slowly with 9 time such that it remains virtually unchanged during short time intervals. 10 The CADM model does not differentiate between binding to AhR and CYP1A2, and it 11 lacks explicit descriptions of CYP1A2 induction, a key determinant of TCDD kinetics. 12 However, the empirical equation in the CADM model is based on five parameters (i.e., fmin, fmax, 13 K, Wa, and Wl; see Tables 3-4 and 3-5) that allow the successful description of the behavior of 14 TCDD in liver and adipose tissue (i.e., TCDD half-lives in each compartment increase with 15 decreasing body burden). This observation implies that the model adequately accounts for the 16 ensemble of the processes. Essentially, the CADM model describes the rate of change in tissue 17 concentrations of TCDD as a function of total body burden such that the global elimination rate 18 decreases with decreasing body burden or administered dose. 19 20 3.3.4.3.1.3, P a r a m e te r estim ation . 21 The CADM model is characterized by its simplicity and fewer parameters compared to 22 physiologically-based models. Reflecting this simplicity, hepatic extraction is computed with a 23 unified empirical equation that accounts for all relevant processes (i.e., protein induction and 24 binding). 25 The key parameters (fmin, fmax, K, and ke) were all obtained by fitting to species-specific 26 pharmacokinetic data. The physiological parameters (such as tissue weights) used in the model 27 are within ranges documented in the literature. The fat content is described to vary as a function 28 of age, sex, and BMI. However, the BMI of the model is not allowed to change during an 29 individual simulation (which can range from 20 years to 70+ years) when in reality the 30 percentage of fat in humans changes over time. None of the TCDD-specific parameters were 31 estimated a priori or independent of the data set simulated by the model. This document is a draftfor review purposes only and does not constitute Agency policy. 3-29 DRAFT--DO NOT CITE OR QUOTE 1 3.3.4.3.I.4. M o d e l p e rfo rm a n c e a n d d e g re e o f eva lu a tio n . 2 The CADM model was not evaluated for its capabilities in predicting data sets not used 3 in its parameterization. In other words, one or more of the key input parameters (fhmin, fhmax, ke, 4 K) was or were obtained essentially by fitting to the species-specific pharmacokinetic data, such 5 that there was no "external" validation data set to which the model was applied. Despite the lack 6 of emphasis on the "external" validation aspect, the authors (Aylward et al., 2005, 197114); 7 (Carrier et al., 1995, 197618; Carrier et al., 1995, 543780) have demonstrated the ability of the 8 model to describe multiple data sets covering a range of doses and species. 9 The visual comparison of the simulated data to experimental values suggests that the 10 model could, to an approximate degree, correctly reproduce the whole set of data (e.g., 11 pharmacokinetic [PK] profile over a range of dose and time) and not just part of the PK curve, 12 essentially with the use of a single set of equations and parameters. 13 The pharmacokinetic data sets for TCDD that were used to calibrate/evaluate the CADM 14 model by Aylward et al. (2005, 197114; Carrier et al., 1995, 197618; Carrier et al., 1995, 15 543780) included the following: 16 17 Adipose tissue and liver concentrations of TCDD following a single oral dose of 1 pg/kg 18 in monkeys (McNulty et al., 1982, 543782); 19 Percent dose retained in liver for a total dose of 14 ng in hamsters (Van den Berg et al., 20 1986, 543781); 21 Elimination kinetics of TCDD in female Wistar rats following a single subcutaneous dose 22 of 300 ng/kg (data from Abraham et al., 1988, 199510); 23 Liver and adipose tissue concentrations (terminal measurements) in Sprague-Dawley rats 24 given 1, 10 or 100 ng TCDD/kg bw during 2 years (Kociba et al., 1978, 001818); and 25 Serum lipid concentrations of TCDD over a period of several years in 54 adults (29 men 26 and 25 women) from Seveso and in three Austrian patients (Aylward et al., 2005, 27 197114). 28 29 For illustration purposes, Figure 3-9 shows model simulations of rat data from Carrier et 30 al. (1995, 197618). Figure 3-2 (see Section 3.3.2.4) depicts the human data that were used by the 31 authors to support the concentration-dependent elimination concept; the model was 32 parameterized to fit approximately to these data (Aylward et al., 2005, 197114). This document is a draftfor review purposes only and does not constitute Agency policy. 3-30 DRAFT--DO NOT CITE OR QUOTE 1 The authors did not report any specialized analyses that quantitatively evaluated the 2 uncertainty, sensitivity, and/or variability of CADM model parameters and structure. 3 4 3.3.4.3.I.5. C o n fid en ce in C A D M m o d e l p r e d ic tio n s o f d o se m e tric s. 5 A qualitative level of confidence associated with the predictability and reliability of 6 absorbed dose and body burden for oral exposures in humans (as well as several animal species) 7 by this model can be ranked as high (see Table 3-6). This model, however, does not account for 8 the differential solubility of TCDD in serum lipids and adipose tissue lipids, nor does it account 9 for the diffusion-limited uptake by adipose tissue. Due to these limitations, the confidence 10 associated with the predictions of the serum lipid concentration of TCDD is considered medium, 11 particularly when it is not documented that steady-state is reached during the critical toxicologic 12 studies and human exposures. Furthermore, the CADM model does not facilitate the 13 computation of TCDD concentrations in specific internal organs (other than liver and adipose 14 tissue). The reliability of this model for simulating the liver concentration (free, bound, or total) 15 of TCDD at low doses is considered to be low. This low confidence level is a result of the 16 uncertainty associated with the key parameter /hmin. This parameter needs to be re-calibrated for 17 each study/species/population to effectively represent the free fraction of TCDD in liver and the 18 amount of TCDD contained in the hepatic lipids and bound to the liver proteins (whose levels 19 might be reflective of background exposures of various sources; see Carrier et al., 1995, 20 197618). The uncertainty related to the numerical value of this parameter in animals and 21 humans--particularly at very low exposures--raises concern regarding the use of this model to 22 predict TCDD concentration (free, bound, or total) in liver as the dose metric for dose-response 23 modeling. Although the use of the parameter /hmax permits the prediction of the dose to liver at 24 high doses, it does not specifically facilitate the simulation of the amount bound to the protein or 25 level of induction in liver. Because the CADM model is not capable of simulating enzyme 26 induction based on biologically-relevant parameters, its reliability for predicting the 27 concentration of TCDD bound specifically to the AhR is not known. Finally, due to the lack of 28 parameterization or verification with kinetic data in pregnant, lactating, or developing animals or 29 humans, the CADM model is unlikely to be reliable in the current form for use in predicting 30 potential dose metrics in these subpopulations or study groups that might form the basis of points 31 of departure (PODs) for the assessment. This document is a draftfor review purposes only and does not constitute Agency policy. 3-31 DRAFT--DO NOT CITE OR QUOTE 1 3.3.4.3.2. P B P K model. 2 3.3.4.3.2.I. M o d e l stru ctu re. 3 Emond et al. (2004, 197315; 2006, 197316) simplified the eight-compartment rat model 4 of Wang et al. (1997, 104657) to a four-compartmental model (liver, fat, rest of body and 5 placenta with fetal transfer) (Emond et al., 2004, 197315), and later to a three-compartment adult 6 model (liver, fat, rest of the body) (Emond et al., 2006, 197316) (see Figures 3-10 and 3-11). 7 Their rationale for simplification of the model was based on evaluating, critiquing, and 8 improving all earlier PBPK models by Wang et al. (1997, 104657). In general, the main reason 9 for the simplification was that extrapolation of a PBPK model to humans with these many (i.e., 10 eight compartments) compartments would be problematic due to the limited availability of 11 relevant human data for validation (Emond et al., 2004, 197315). One major difference from 12 earlier models, repeatedly emphasized by Emond et al. (2005, 197317; 2006, 197316), was their 13 description (included in their simplified PBPK models) of the dose-dependent, inducible 14 elimination of TCDD. The rationale for including TCDD binding and induction of CYP1A2 into 15 the model was earlier described by Santostefano et al. (1998, 200001). 16 The most recent version of the rat and human PBPK models developed by Emond et al. 17 (2006, 197316) describes the organism as a set of three compartments corresponding to real 18 physical locations--liver, fat, and rest of the body--interconnected by systemic circulation (see 19 Figure 3-10). The liver compartment includes descriptions of CYP1A2 induction, which is 20 critical for simulating TCDD sequestration in liver and dose-dependent elimination of TCDD. In 21 this model, the oral absorption of TCDD from the GI tract accounts for both the lymphatic (70%) 22 and portal (30%) systems. 23 The biological relationship between TCDD "sequestration" by liver protein and its 24 "elimination" by the liver is not entirely clear. TCDD is metabolized slowly by unidentified 25 enzymes. CYP1A2 is known to metabolize TCDD based on studies in CYP1A2 KO mice 26 (Diliberto et al., 1997, 548755; 1999, 143713), in which the metabolic profile is different 27 compared to wild-type mice. However, since several metabolites appear in the feces of CYP1A2 28 knock out mice, it is assumed that there are other enzymes involved in TCDD metabolism. 29 TCDD binds to the AhR and induces not only CYP1A2, but also CYP1A1, CYP1B1, and several 30 UGTs and transporters (Gasiewicz et al., 2008, 473406). Both hydroxylated and glucuronidated 31 hydroxyl metabolites are found in the feces of animals treated with TCDD (Hakk et al., 2009, This document is a draftfor review purposes only and does not constitute Agency policy. 3-32 DRAFT--DO NOT CITE OR QUOTE 1 594256). Because the exact enzymes involved with TCDD are unknown and yet the metabolism 2 is induced by TCDD, an assumption of increased the elimination rate of TCDD in proportion to 3 the induction of CYP1A2 is made. In the PBPK model, CYP1A2 is needed because TCDD 4 binds to rat, mouse, and human CYP1A2 (Diliberto et al., 1999, 143713; Staskal et al., 2005, 5 198276). Thus CYP1A2 induction is necessary to describe TCDD pharmacokinetics due to 6 TCDD binding. Hence, CYP1A2 can be used as a marker of Ah-receptor induction of "TCDD 7 metabolizing enzymes." Other models use AhR occupancy as a marker of induction of "TCDD 8 metabolizing enzymes" (Andersen et al., 1997, 197172; Kohn et al., 2001, 198767). 9 Figure 3-11 depicts the structure of the rat developmental-exposure PBPK model (Emond 10 et al., 2004, 197315). This model was developed to describe the relationship between maternal 11 TCDD exposure and fetal TCDD concentration during critical windows of susceptibility in the 12 rat. In formulating this PBPK model, Emond et al. (2004, 197315) reduced the original 13 8-compartment model for TCDD in adult rats by Wang et al. (1997, 104657) to a 4-compartment 14 (i.e., liver, fat, placenta, and rest of the body) model for maternal rat. Activation of the placental 15 compartment and a separate fetal compartment occurs during gestation (Emond et al., 2004, 16 197315). 17 18 3.3.4.3.2.2. M a th e m a tic a l rep resen ta tio n . 19 The key equations of the PBPK model of Emond et al. (2004, 197315) are reproduced in 20 Text Boxes 3-1 and 3-2, whereas those from Emond et al. (2005, 197317; 2006, 197316) are 21 listed in Table 3-7. The rate of change of TCDD in the various tissue compartments is modeled 22 on the basis of diffusion limitation considerations. Accordingly, mass balance equations are 23 used to compute the rate of change in the tissue (i.e., intracellular compartment) and tissue blood 24 (i.e., extracellular compartment). The membrane transfer of TCDD is computed using a 25 permeation coefficient-surface area cross product (PA) for each tissue. Metabolism and binding 26 of TCDD to the AhR and inducible hepatic protein (CYP1A2) are described in the liver. The 27 total mass in the liver was then apportioned between free dioxin (Clf) and bound forms of TCDD 28 (see Figure 3-12). The dose- and time-dependent induction of hepatic CYP1A2 in the liver is 29 described per Wang et al. (1997, 104657) and Santostefano et al. (1998, 200001). Accordingly, 30 the amount of CYP1A2 in the liver was computed as the time-integrated product of inducible 31 production and a simple first-order loss process (Wang et al., 1997, 104657): This document is a draftfor review purposes only and does not constitute Agency policy. 3 -3 3 DRAFT--DO NOT CITE OR QUOTE 1 2 -- f 2 =S (t)K0 - K 2CA2t (Eq. 3-19) dt 3 4 In this expression, CYP]A2 is the concentration of the enzyme (nmol/g), K2 is the rate constant for 5 the first order loss (hour-1), CA2t is the concentration of CYP1A2 in the liver (nmol/g), K0 is the 6 basal rate of production of CYP1A2 in the liver (nmol/g.hr), and S(t) (unitless) is a multiplicative 7 stimulation factor for CYP1A2 production in the form of a Hill-type function (see 8 Section 3.3.2.3): 9 10 S (t) = 1 + )I n A2 ( C Ah-TCDD (Eq. 3-20) (ICa 2 ) h + (Ca ,_ tcDD ) h 11 12 where, S(t) is the stimulation function, InA2 is the maximum fold of CYP1A2 synthesis rate over 13 the basal rate, CAh-TCDDis the concentration of AhR occupied by TCDD, and ICA2 is the 14 Michaelis-Menten constant of CYP1A2 induction (nM). The dose-dependent or variable 15 elimination of TCDD was described using the relationship: 16 CYP1A2induced - CYP1A2basal 17 KBILE LI x Kelv (Eq. 3-21) CYP1A2'basal 18 19 where CYP1A2inducedis the concentration of induced CYP1A2 (nmol/mL), CYP1A2basai is the 20 basal concentration of CYP1A2 (nmol/mL), and Kelv is the interspecies constant adjustment for 21 the elimination rate (hour-1). 22 There are various ways of formulating the dose-dependent elimination as a function of 23 the level of CYP1A2, and the above equation (used by the authors) can be viewed as one means 24 of describing this behavior quantitatively. The numerator in the equation above will always be 25 greater than zero when there is TCDD in the system (including TCDD derived from either 26 background exposures or defined external sources). Consequently, the rate of elimination will 27 correspond to a nonzero value for situations involving TCDD exposures. Furthermore, the 28 numerator in Eq. 3-21 should more appropriately be CYP1A2inducedrather than [CYP1A2induced- 29 CYP1A2Basal] to avoid the problem of lower levels of induction at low doses resulting in a lower This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 3 4 DRAFT--DO NOT CITE OR QUOTE 1 than basal rate of synthesis of CYP1A2. The above equation, however, does not describe 2 changes in elimination rate in direct proportionality with the CYP1A2 levels; also, the K elv value 3 by itself does not reflect a scalable basal metabolic rate. Rather, these two terms collectively 4 describe the outcome related to the TCDD elimination processes, based on fitting to observations 5 in rats (Santostefano et al., 1998, 200001). The impact of CYP1A2 induction and sequestration 6 on binding and elimination of TCDD is simulated using the Emond et al. (2004, 197315) model. 7 The gestational model consisted of a fetal compartment, and the transfer of TCDD 8 between the placental and fetal compartments was described as a diffusion-limited (rather than a 9 perfusion-limited) process (see Text Boxes 3-1 and 3-2).16 10 Text Box 3-1. Variation o f B ody W eight with Age: B W T. ( g ) ( 0.41 x Tim e B W initial x , 1402.5 + T im e ( 0.75 BW m other ') Cardiac O utput: Qc(mL / h) =Qcc x 60 1,000 A factor of 60 corresponds to the conversion of minutes to hours, and 1,000 is the conversion of body weight from g to kg. B lood Compartment: Cb(nmol / mL) = x + x + x + x x((Q f Cfb) (Qre Creb) (Qli Clib) (Qpla Cplab) + Lymph)) - (Cb Clru) Qc 11 16Diffusion limited, sometimes also known as "membrane limited," means a chemical's movement from one side of the membrane to the other is limited by the membrane. Thus, the membrane, in this case, is a limiting factor for uptake. Perfusion limited, also known as "flow limited" indicates that a chemical is so rapidly taken up (e.g., by the tissue from the blood) that the flow rate is the only limiting factor. This document is a draftfor review purposes only and does not constitute Agency policy. 3-35 DRAFT--DO NOT CITE OR QUOTE Text Box 3-2. Placenta Tissue Com partm ent (a) Tissue-blood subcompartment tdjAp ab (nmol / h) = Qpla(Ca - Cplab) + PApla(Cplab - Cplafree) Cplab = Aplab Wplab (b) Tissue cellular matrices dApa (nmol / h) =PApIa(CpIab - Cplafree) - dApla _ f " + dAfet _ pla dt dt dt Cpla(nmol / mL) = Apla Wpla Free TCDD Concentration in Placenta Cplafree(nmol / mL) =Clpla - (Cplafree x Ppla + Plabmax x Cplafree Kdpla + Cplafree D ioxin Transferfr o m Placenta to Fetuses dAPla _ fet (nm 0 / h) = ClPla fet x Cpla dt D ioxin Transferfr o m Fetuses to Placenta d^ ^ , ~ Fla (mol / h) = ClFi, x CfetV at Fetal D ioxin Concentration (Fetuses 5 = Per Litter) dA fet(,nmol, /.h,.) = dA-Fla fet - dAfet Fla dt dt dt Cfet(nmol / h) = Afet Wfet CfetV(nmol / mL) = Cfet Pfet 1 2 This document is a draftfor review purposes only and does not constitute Agency policy. 3-36 DRAFT--DO NOT CITE OR QUOTE 1 3.3.4.3.2.3. P a r a m e te r estim ation . 2 Table 3-8 lists the numerical values of the adult rat and human PBPK models of Emond 3 et al. (2005, 197317; 2006, 197316). The values for key input parameters of the rat gestational 4 model are summarized in Table 3-8 as well as Figure 3-13. 5 The parameters for the rat model were obtained primarily from Wang et al. (1997, 6 104657) except that the value of affinity constant for CYP1A2 was changed from 0.03 to 7 0.04 nmol/mL to get better fit to experimental data (Emond et al., 2004, 197315) and the variable 8 elimination parameter (Kelv) was obtained by optimization of model fit to kinetic data from 9 Santostefano et al. (1998, 200001) and (Emond et al., 2005, 197317; Emond et al., 2006, 10 197316; Wang et al., 1997, 104657). Wang et al. (1997, 104657) used measured tissue weights 11 whereas the tissue blood flows and tissue blood weights were obtained from International Life 12 Sciences Institute (ILSI, 1994, 046436). The partition coefficients (which were similar to those 13 of Leung et al., 1988, 198815; 1990, 192833), the permeability x area (PA) value for tissues, the 14 dissociation constant for binding to CYP1A2 (ICA2) and the Hill coefficient (h) were estimated 15 using a two-stage process of fitting to dose-response and time-course data on TCDD tissue 16 distribution (Wang et al., 1997, 104657). In the initial stage, the experimental data of arterial 17 blood concentrations were used as input to the individual compartment to estimate the 18 parameters; then, with the values obtained during stage one as initial estimates, those unknown 19 parameters were re-estimated by solving the entire model at once using an optimization route 20 (Wang et al., 1997, 104657). The receptor concentrations and dissociation constant of TCDD 21 bound to AhR were obtained by fitting the model to TCDD tissue concentration combining with 22 enzyme data reported by Santostefano et al. (1998, 200001) whereas the basal CYP1A2 in liver 23 was based on literature data (Wang et al., 1997, 104657). 24 The parameters for the human PBPK model were primarily based on the rat model 25 (Emond et al., 2005, 197317; Emond et al., 2006, 197316; Wang et al., 1997, 104657). 26 Specifically, the blood fraction in the tissues, the tissue:blood partition coefficients, tissue 27 permeability coefficient, the binding affinity of TCDD to AhR and CYP, and the maximum 28 binding capacity in the liver for AhR were all set equal to the values used in the rat model. The 29 species-specific Kelv was estimated by fitting to human data (Emond et al., 2005, 197317). 30 For the gestational rat model, the parameters describing the growth of the placental and 31 fetal compartments as well as temporal change in blood flow during gestation were incorporated This document is a draftfor review purposes only and does not constitute Agency policy. 3-37 DRAFT--DO NOT CITE OR QUOTE 1 based on existing data. Exponential equations for the growing compartments were used (see 2 Figure 3-13), except for adipose tissue for which a linear increment based on literature data was 3 specified. While physiological parameters for the pregnant rat were obtained from the literature, 4 all other input parameters were set equal to that of nonpregnant rat (obtained from Wang et al., 5 1997, 104657), see Tables 3-7 and 3-8. The current version of the rat gestational model contains 6 parameters for variable elimination from Emond et al. (2006, 197316; Table 3-8), and still 7 provides essentially the same predictions as the original publication (Emond et al., 2004, 8 197315). 9 10 3.3.4.3.2.4. M o d e l p e rfo rm a n c e a n d d e g re e o f eva lu a tio n . 11 The PBPK model of Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) had 12 parameters estimated by fitting to kinetic data, such that the resulting model consistently 13 reproduced the kinetic data. The same model structure with a single set of species-specific 14 parameters could reproduce the kinetics of TCDD following various doses and exposure 15 scenarios not only in the rat but also in humans. The simulations of the PBPK model of Emond 16 et al. (2006, 197316) have been compared with two sets of previously published rat data: blood 17 pharmacokinetics following a single dose of 10 pg/kg (the dose corresponding to the mean 18 effective dose for induction of CYP1A2) (Santostefano et al., 1998, 200001) (see Figure 3-14); 19 and hepatic TCDD concentrations during chronic exposure to 50, 100, 500, or 1,750 ng/kg 20 (Walker et al., 1999, 198615) (see Figure 3-15). It is relevant to note that the PBPK model of 21 Emond et al. (2004, 197315; 2006, 197316) is essentially a reduced version of the Wang et al. 22 (1997, 104657) model, and it therefore provides simulations of liver and fat concentrations of 23 TCDD that deviated by not more than 10-15% of those of Wang et al. (1997, 104657). The 24 nongestational model of Emond et al. (2004, 197315) simulated the kinetic data in liver, fat, 25 blood and rest of body of female Sprague-Dawley rats given a single dose of 10 pg TCDD/kg 26 (data from Santostefano et al., 1996, 594258) and in liver and fat of male Wistar rats treated with 27 a loading dose of 25 ng/kg followed by a weekly maintenance dose of 5 ng TCDD/kg by gavage 28 (data from Krowke et al., 1989, 198808). 29 The gestational rat PBPK model simulated the following PK data sets (Emond et al., 30 2004, 197315) : 31 This document is a draftfor review purposes only and does not constitute Agency policy. 3 -3 8 DRAFT--DO NOT CITE OR QUOTE 1 TCDD concentration in blood, fat, liver, placenta, and fetus of female Long-Evans rats 2 given 1, 10, or 30 ng/kg, 5 daysWeek, for 13 weeks prior to mating followed by daily 3 exposure through parturition (Hurst et al., 2000, 198806); 4 TCDD concentration in tissues (liver, fat), blood, placenta and fetus determined on 5 gestation day (GD) 16 and GD 21 following a single dose of 0.05, 0.8, or 1 pg/kg given 6 on GD 15 to pregnant Long Evans rat (Hurst et al., 2000, 199045); 7 Maternal and fetal tissue concentrations on GD 9, GD 16 and GD 21 after a single dose 8 of 1.15 pg TCDD/kg given to Long-Evans rats on GD 9 or GD 15 (Hurst et al., 1998, 9 134516); and 10 Fetal TCDD concentrations determined on GD 19 and GD 21 in rats exposed to 11 5.6 pg TCDD/kg on GD 18 (Li et al., 2006, 199059). 12 13 Furthermore, the scaled rat model was shown to be capable of simulating human data 14 from the Austrian and Seveso subjects (see Figures 3-16 and 3-17). In this regard, it is useful to 15 note that the computational version of the PBPK model of Emond et al. (2005, 197317; 2006, 16 197316) also contained the necessary equation to transform the model output of blood 17 concentration into serum lipid adjusted concentration of TCDD. 18 The human model of Emond et al. (2005, 197317; Emond model) has advantages for 19 improving the TCDD dosimetry used in existing human epidemiological studies because the 20 model predicts the redistribution of TCDD within the body (to stores in fat and liver) based on 21 physiological principles. However, because the dose-dependency of metabolic elimination in the 22 Emond model was not calibrated to human data, it is important to review the predictions of this 23 model using a database of human observations that is as extensive as possible and a spread of 24 internal TCDD concentrations that is as wide as possible. Thus, presented below is a 25 juxtaposition of modeled elimination rates from the Emond model with observations for 26 two highly exposed Austrian patients (severe intoxication of "unknown origin" (Geusau et al., 27 2001, 197444)) and nine of 10 Ranch Hand veterans17used for the original "validation" 28 comparisons presented in the Emond et al. (2005, 197317). 29 Figure 3-18 shows the time course of the declines in TCDD serum concentrations in 30 two highly-exposed Austrian subjects compared with the Emond model results. The comparison 31 in Figures 3-17 and 3-18 indicates that the Emond model adequately describes the rate of TCDD 17In preliminary comparisons, the simulation run for the 10thRanch Hand veteran appeared anomalous and was therefore excluded from this summary. This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 3 9 DRAFT--DO NOT CITE OR QUOTE 1 elimination for the more highly exposed Austrian patients, but predicts a somewhat faster rate of 2 decline than that observed for the less heavily exposed patient. 3 Figure 3-19 shows the results of combining the simulated and observed rates of loss for a 4 group of Austrian and Ranch Hand subjects evaluated by Emond et al. (2005, 197317), counting 5 only one data point per person. The X-axis in this figure is the TCDD serum concentration at the 6 midpoint of the observations for each subject. The error bars in the figure represent 1 standard 7 error. The results of this figure illustrate two points: (1) the Emond model simulation (open 8 squares) are generally very close to the actual data (solid circles) for the nine Ranch hands 9 (clustered toward lower left corner) and one of the the two Austrian patients (upper right corner); 10 and (2) both the Emond model simulation results and the actual data show a linear trend and 11 linear regression lines were plotted, respectively, as shown in Figure 3-19. 12 Table 3-9 presents the results of regression analyses of the observed rates of decline in 13 relation to the estimated TCDD serum levels at the midpoint of the observations for each subject 14 in the Ranch Hand study (see Figure 3-19). These results indicate that some appreciable dose 15 dependency of TCDD elimination is unequivocally supported. However, the central estimate of 16 the slope of the relationship between the log of the TCDD elimination rate and the log of the 17 TCDD level is only about 75% of that expected under the Emond et al. PBPK model 18 (i.e., 0.092 - 0.123 = 0.748). 19 Overall, the conclusion from the above analysis is that the Emond model is reasonable to 20 use, but the model might be improved by (1) include the two nondose-dependent pathways of 21 elimination documented in the Geusau papers (GI elimination via the feces and loss via the 22 sloughing of skin cells), and (2) reducing the extent of loss via the dose-dependent metabolism 23 pathway from the liver (Geusau et al., 2002, 594259; Harrad et al., 2003, 197324) so that overall 24 loss rates for the average elimination rates from the Ranch Hand veterans is maintained. 25 A sensitivity analysis of inputs used to estimate inducible elimination rate for a single 26 oral dose of 0.001 to 10 pg/kg in the rat indicated that the number of key parameters ranged from 27 seven at the low dose region to 12 at the high dose (see Figure 3-20)(Emond et al., 2006, 28 197316). The sensitive parameters identified included the oral absorption parameters (KABS), 29 volumes of liver and adipose tissue (WLIO, WFO), adipose tissue:blood partition coefficient 30 (PF), and the basal CYP1A2 level (CYP1A2 1A2). At high doses, the most sensitive parameters This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 4 0 DRAFT--DO NOT CITE OR QUOTE 1 also included those related to the maximal induction of CYP1A2 and AhR binding capacity (see 2 Figure 3-20) (Emond et al., 2006, 197316). 3 The gestational rat model described in Emond et al. (2004, 197315), upon 4 reparameterization, could simulate the kinetics of TCDD in mice. The initial changes to the rat 5 model parameters included: rest of the body:blood partition coefficient (PRE), basal 6 concentration (CYP1A2_1A2), delay in induction time (CYP1A2_1TAU) and adipose tissue 7 permeability coefficient (PAFF), in accordance with Wang et al. (2000, 198738) (see Table 3-8). 8 Subsequently, four parameters (adipose tissue:blood partition coefficient, CYP1A2 affinity 9 parameter, GI tract elimination transit constant (hour-1) and the interspecies metabolic parameter 10 Kelv (hour-1) were re-estimated based on visually fit of model simulations to the PK data from 11 Diliberto et al. (2001, 197238), following an oral dose 150 ng TCDD/kg/day, 5 days/week for 12 17 weeks (see Table 3-7). The resulting mouse model is capable of reproducing the kinetics of 13 TCDD in the adult (see Figures 3-21 through 3-27), as well as, to a very limited extent, the 14 kinetics during gestation (see Figure 3-28). 15 16 3.3.4.3.2.5. C o n fid en ce in P B P K m o d e l p r e d ic tio n s o f d o se m e tr ic s . 17 The PBPK model facilitates prediction of absorbed dose, body burden, and blood 18 concentration of TCDD for oral exposures in adult humans and rats (adult and developing) with 19 high confidence (see Table 3-10). The model output of blood concentration can be normalized to 20 lipid content representative of the study group (species, sex, age, lifestage, and diet). However, 21 the PBPK model of Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) does not simulate 22 plasma and erythrocyte TCDD concentrations separately, and it predicts tissue concentrations on 23 the basis of tissue:whole blood partition coefficients and not on the basis of serum 24 lipid-normalized values. 25 The reliability of this model for simulating the liver concentration of TCDD in rats is 26 considered to be high but it is considered to be medium for humans. Although empirical data on 27 bound or free concentrations were not used to evaluate model performance in humans, the 28 biological phenomena (consistent with available data) related to the hepatic sequestration, 29 enzyme induction, and dose-dependent elimination are described in the model. This is one of the 30 situations where PBPK models are uniquely useful; that is, they permit the prediction of system 31 behavior based on understanding of the mechanistic determinants, even though the required data This document is a draftfor review purposes only and does not constitute Agency policy. 3-41 DRAFT--DO NOT CITE OR QUOTE 1 cannot be directly obtained in the system (e.g., bound concentrations in the liver of exposed 2 humans). For these dose measures (i.e., bound concentration and total liver concentration), the 3 level of confidence can be further improved or diminished by the outcome of sensitivity analysis. 4 In this regard, the results of a focused sensitivity analysis indicate that the most sensitive 5 parameters of the human model are among the most uncertain (i.e., those parameters for which 6 estimates were not obtained in humans) with respect to prediction of liver TCDD concentration, 7 contrary to the animal model (see Section 3.3.6). 8 With respect to the mouse model, however, the level of confidence is low to medium, 9 given that it has not been verified extensively with blood, body burden, or tissue concentration 10 time-course or dose-response data. However, the mouse PBPK model, based on the rat model 11 that has been evaluated with several PK data sets, has been shown to reproduce well the limited 12 mouse liver kinetic data (see Figures 3-21 through 3-28; Boverhoff et al., 2005, 594260). The 13 same model structure has been used for simulating kinetics of TCDD in humans successfully. 14 Overall, the adult mouse model, given its biological basis combined with its ability to simulate 15 TCDD kinetics in multiple species, is considered to exhibit a medium level of confidence for 16 simulating dose metrics for use in high to low dose extrapolation and interspecies (mouse to 17 human) extrapolation. Even though similar considerations are applicable to gestational model in 18 mice, the confidence level is considered to be low since very limited comparison with empirical 19 data has been conducted (see Figure 3-28). Despite the uncertainty in these predictions, the 20 scaled rat gestational model, given its biological and mechanistic basis, might be of use in 21 predicting dose metrics in these groups that might form the basis of PODs in certain key studies. 22 23 3.3.4.4. A p p lica b ility o f P K M o d els to D erive D o se M etrics f o r D o se-R esp o n se M o d elin g o f 24 T C D D : C o n fid en ce a n d L im itation s 25 Both the CADM and PBPK models describe the kinetics of TCDD following oral 26 exposure to adult animals and humans by accounting for the key processes affecting kinetics, 27 including hepatic sequestration phenomena, induction, and nonlinearity in elimination, and 28 distribution in adipose tissue and liver. Both models can be used for estimating body burdens 29 and serum lipid adjusted concentrations of TCDD. However, there are several differences 30 between these two models. The PBPK model calculates the free and bound concentrations of 31 TCDD in the intracellular subcompartment of tissues. The total or receptor-bound This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 4 2 DRAFT--DO NOT CITE OR QUOTE 1 concentrations in liver are unambiguous and more easily interpretable with the PBPK model than 2 with the CADM model. In addition, the PBPK model computes bound and total concentrations 3 as a function of the free concentration in the intracellular compartment of the tissue. By contrast, 4 the CADM model simulates the total concentration based on empirical consideration of hepatic 5 processes. Consequently, the amount of TCDD bound to AhR or CYP1A2 cannot be simulated 6 with the CADM model. The CADM model computes only the total TCDD concentration in 7 liver, and describes TCDD elimination through partitioning from circulating lipids across the 8 lumen of the large intestine into the feces, while the PBPK model accounts for this process 9 empirically within its hepatic elimination constant. Elimination of TCDD via skin, a minor 10 process, is not described by either model. Thus, dose-response modeling based on body burden 11 of TCDD in adult animals and humans can be conducted with either of the models, provided the 12 duration of the experiment is at least one month, due to limitations in the CADM model. As 13 shown in Figure 3-29, the predicted slope and body burden over a large dose range are quite 14 comparable (generally within a factor of two). 15 Results of simulations of serum lipid concentrations or liver concentrations vary for the 16 two models to a larger extent (up to a factor of 7), particularly for simulations of short duration. 17 These differences reflect two characteristics of the PBPK model: first, quasi-steady-state is not 18 assumed in the PBPK model; second, the serum lipid composition used in the model is not the 19 same as the adipose tissue lipids. The CADM model does not account for differential solubility 20 of TCDD in serum lipids and adipose tissue lipids, nor does it account for the diffusion-limited 21 uptake by adipose tissue. Therefore, the PBPK model would appear to be superior to the CADM 22 model with respect to the ability to simulate serum lipid and tissue concentrations during 23 exposures that do not lead to the onset of steady-state condition in the exposed organism. 24 The CADM model is simple and based on fewer parameters than the PBPK model. 25 Because the CADM model is constructed by fitting to data, its performance is likely to be 26 reliable for the range of exposure doses, species, and life stages from which the parameter 27 estimates were obtained. On the other hand, the PBPK model structure and parameters are 28 biologically-based and can be adopted for each species and life stage. Accordingly, the PBPK 29 model has been adopted to simulate the kinetics of TCDD in the fetus and in pregnant rats, as 30 well as in adult humans and rats (Emond et al., 2004, 197315; Emond et al., 2005, 197317; 31 Emond et al., 2006, 197316). The time step for calculation and dosing in the CADM model This document is a draftfor review purposes only and does not constitute Agency policy. 3 .4 3 DRAFT--DO NOT CITE OR QUOTE 1 corresponds to 1 month. This requirement represents a constraint in terms of the use of this 2 model to simulate a variety of dosing protocols used in animal toxicity studies. This 3 requirement, however, is not a constraint with the PBPK models. So, simulating the body 4 burden and serum lipid concentrations for a longer duration of exposure, either model would 5 appear to be useful; but the PBPK model would be the tool of choice for simulating alternative 6 dose metrics of TCDD (e.g., blood concentration, total tissue concentration, bound 7 concentration) for various exposure scenarios (including single dose studies), routes and life 8 stages in the species of relevance, to TCDD dose-response assessment, particularly, mice, rats, 9 and humans. 10 Two minor modifications, to enhance the biological basis, were made to the PBPK model 11 of Emond et al. (2006, 197316), before its use in the computation of dose metrics for TCDD. 12 The first one involved the recalculation of the volume of the rest of the body as follows: 13 = - + x + + +14 WRE0 (0.91 (WLIB0 x WLI0 WFB0 WF0 WLI0 WF0)/(1 WREB0)) (3-22) 15 16 where 17 WRE0 = weight of cellular component of rest of body compartment (as fraction of 18 body weight); 19 WLI0 = weight of cellular component of liver compartment (as fraction of body 20 weight); 21 WF0 = weight of cellular component of fat compartment (as fraction of body 22 weight); 23 WREB0 = weight of the tissue blood component of the rest of body compartment (as 24 fraction of body weight); 25 WLIB0 = weight of the tissue blood component of the liver compartment (as fraction 26 of body weight); and 27 WFB0 = weight of the tissue blood component of the fat compartment (as fraction of 28 body weight). 29 30 In the original code, the weight of the rest of body compartment was calculated as the 31 difference between 91% of body weight and the sum total of the fractional volumes of blood, 32 liver tissue (intracellular component), and adipose tissue (intracellular component). The blood 33 compartment in the PBPK model is not explicitly characterized with a volume; as a result, the This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 4 4 DRAFT--DO NOT CITE OR QUOTE 1 total volume of the compartments is less than 91%. The recalculations shown above were used 2 to address this problem. Given the very low affinity of TCDD for blood and rest of the body, 3 reparameterizing the model resulted in less than a 1% change in output compared to the 4 published version of the PBPK model for chronic exposure scenarios (Emond et al., 2006, 5 197316). 6 The second minor modification related to the calculation of the rate of TCDD excreted 7 via urine. The original model code computed the rate of excretion by multiplying the urinary 8 clearance parameter with the concentration in the rest of the body compartment. Instead, the 9 code was modified to use the blood concentration in this equation. This resulted in the 10 re-estimation of the urinary clearance value in the rat and human models but it did not result in 11 any significant change in the fit and performance of the original model. 12 The revised parameter estimates of the rat, mouse, and human models are captured in 13 Table 3-8 with a footnote. 14 15 3.3.4.5. R eco m m en d ed D o se M etrics f o r K e y S tu dies 16 The selection of dose metrics for the dose-response modeling of key studies is largely the 17 result of (1) the relevance of a dose metric on the basis of current knowledge of TCDD's 18 mechanism of action for critical endpoints and (2) the feasibility and reliability of obtaining the 19 dose metric with available PK models. Secondarily, the goodness-of-fit of the dose-response 20 models (which reflects the relationship of the selected internal dose measures to the response) 21 can be used to inform selection of the most appropriate dose metric for use in deriving TCDD 22 toxicity values. 23 Body burden--even though this metric is based on mechanistic considerations--is a 24 somewhat distant measure of dose with respect to target tissue dose, and this metric represents 25 the "overall" average concentration of TCDD in the body. However, a benefit of body burden is 26 that this metric represents a dose measure for which the available PK models can provide highly 27 certain estimates. Thus, the overall confidence associated with the use of body burden in TCDD 28 assessment is categorized as medium. 29 The confidence in the ability of PK models to simulate blood concentration as a dose 30 metric is high, given that the models have been shown to consistently reproduce whole blood (or 31 serum lipid-normalized) TCDD concentration profiles in both humans and rats. Considering the This document is a draftfor review purposes only and does not constitute Agency policy. 3-45 DRAFT--DO NOT CITE OR QUOTE 1 facts that the PBPK models simulate whole blood rather than the serum lipid-normalized 2 concentrations of TCDD and that the study-specific values of serum lipid content are not known 3 with certainty, it is preferable to rely on TCDD blood concentrations as the dose metric. The 4 blood concentrations, if intended, can be normalized on the basis of appropriate total lipid levels. 5 However, based on mechanistic considerations, the confidence in their use would be somewhat 6 lower for hepatic effects. This conclusion reflects the concern regarding the inconsistent 7 relationship between the two variables with increasing dose levels and the fraction of 8 steady-state attained at the time of observation. For other systemic effects related to tissue 9 concentrations, the confidence in the use of TCDD serum or blood concentration is high, 10 particularly for chronic exposures, given the absence of data on organ-specific nonlinear 11 mechanisms. In general, the tissue concentration typically cannot be calculated as a reliable dose 12 metric with either the CADM or the Emond models. One exception is the use of the Emond 13 PBPK models to estimate levels in liver, a metric that is relevant based on MOA considerations. 14 However, it is noted that the hepatic TCDD level encompasses free and bound TCDD and it is a 15 highly complex entity for dose metric considerations. Finally, the AhR-bound concentration 16 may be evaluated for receptor-mediated effects. This dose metic can be obtained by PBPK 17 models, although uncertainties associated with lack of data for this dose metric renders it to be of 18 low confidence (see Table 3-10), The alternative dose metrics for dose-response modeling of 19 TCDD selected on the basis of MOA and PK modeling considerations are summarized in 20 Tables 3-11 and 3-12. 21 These measures of internal dose can be obtained as peak, average, integral (AUC), or 22 terminal values. For chronic exposures in rodents (ca. 2 years), the terminal and average values 23 would be fairly comparable under steady-state conditions. For less-than lifetime exposures, 24 however, the terminal and average values will differ, and therefore an overall average or 25 integrated value (AUC) would be more appropriate. Similarly, for developmental exposures, 26 these alternative dose metrics can be obtained with reference to the known or hypothesized 27 exposure window of susceptibility. 28 This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 4 6 DRAFT--DO NOT CITE OR QUOTE 1 3.3.5. Uncertainty in Dose Estimates 2 3.3.5.1. S o u rces o f U ncertainty in D o se M etric P rediction s 3 3.3.5.1.1. L im itation s o f available P K data. 4 3.3.5.1.1.1. A n im a l data. 5 The available animal data relate to blood, liver, and adipose tissue concentrations for 6 certain exposure doses and scenarios. Although these data are informative regarding the dose7 and time-dependency of TCDD kinetics for the range covered by the specific studies (see 8 Section 3.3.2), they do not provide the peak, average, terminal, or lipid-normalized values of 9 dose metrics associated with the key studies selected for this assessment. The limited available 10 animal PK data are useful, however, in the evaluation of the pharmacokinetic models (see 11 Section 3.3.4). 12 13 3.3.5.1.1.2. H u m a n data. 14 The human data on potential dose metrics are restricted to the serum lipid-adjusted 15 TCDD concentrations associated with mostly uncharacterized exposures (see Sections 3.3.2 and 16 3.3.3). While these data are useful in estimating half-lives in exposed human individuals, they 17 do not provide estimates of hepatic clearance or reflect target organ exposure. Some autopsy 18 data have been used to infer the partition coefficients; however, these data were collected 19 without quantification of the temporal nature of TCDD uptake (see Section 3.2). Despite the 20 limitations associated with the available human data, there has been some success in using these 21 data to infer the half-lives and elimination rates in humans using pharmacokinetic models 22 (Aylward et al., 2005, 197014; Carrier et al., 1995, 197618; Emond et al., 2006, 197316). 23 24 3.3.5.1.2. U ncertainties a sso cia ted w ith m o d el specification. 25 Uncertainty associated with model specification should be viewed as a function of the 26 specific application, such as interspecies extrapolation, intraspecies variability, or high dose to 27 low dose extrapolation. Because the use of pharmacokinetic models in this assessment is limited 28 to interspecies extrapolation and high dose to low dose extrapolation, it is essential to evaluate 29 the confidence in predicted dose metrics for these specific purposes. For interspecies 30 extrapolation, the PBPK and CADM models calculate differences in dose metric between an 31 average adult animal and an average adult human. Both models have a biologically and This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 4 7 DRAFT--DO NOT CITE OR QUOTE 1 mechanistically-relevant structure along with a set of parameters with reasonable biological 2 basis, and reproduce a variety of pharmacokinetic data on TCDD in both rodents and humans. 3 These models possess low uncertainty with respect to body burden, blood, and TCDD/serum 4 (lipid) concentration for the purpose of conducting rat to human extrapolation. However, for 5 other dose metrics, such as free, total, or bound hepatic concentrations, the uncertainty is higher 6 in the CADM model compared to the PBPK model due to model specification differences related 7 to the mechanisms of sequestration and induction in the liver (see Section 3.3.3). 8 For the purpose of high dose to low dose extrapolation in experimental animals, 9 confidence in both models is high with respect to a variety of dose metrics (see previous 10 discussion). The high confidence results from the use of the PBPK models to reproduce a 11 number o f data sets covering a wide range o f dose levels in rodents (rats, mice) including the 12 dose ranges o f most o f the key toxicological studies. Given that the TCDD levels during and at 13 the end o f exposures were not measured in most o f the key studies, use o f the PBPK models is 14 preferred because these models account for dose-dependent elimination, induction, and 15 sequestration. Despite the empirical nature of the specification of these key processes in PBPK 16 models, they essentially reproduce the dose-dependent behavior in rodents, supporting their use 17 in deriving dose metrics for dose-response modeling of TCDD. Overall, the confidence in the 18 use of the alternative dose metrics (identified in Table 3-10) is greater than the confidence in the 19 use o f administered dose for TCDD, for relating to the concentration within tissues to produce an 20 effect. The administered dose does not take into account interspecies differences in the volume 21 o f distribution and clearance or the complex nonlinear processes determining the internal dose. 22 The PBPK model of Emond et al. (2006, 197316) could benefit from further refinement 23 and validation, including a more explicit consideration o f nondose-dependent elimination 24 pathways. As indicated in Section 4, there is some uncertainty associated with the way the 25 elimination of TCDD is described in the existing human PBPK model. The current model 26 essentially treats all TCDD elimination as related to dose dependent metabolism in the liver. In 27 this regard, the classical and more recent PK data on TCDD may be useful in further improving 28 the confidence in their predictions. However, it is likely that there is nondose-dependent 29 elimination o f TCDD via feces and, to a lesser extent skin; juxtaposition o f available elimination 30 rate data with the PBPK model predictions suggests that the current PBPK model modestly 31 overestimates the dose dependency o f overall TCDD elimination. (The central estimate o f the This document is a draftfor review purposes only and does not constitute Agency policy. 3-48 DRAFT--DO NOT CITE OR QUOTE 1 slope of the relationship between the log of the TCDD elimination rate and the log of the TCDD 2 level is only about three-fourths of that expected using the unmodified PBPK model). Emond et 3 al. (2005, 197317) acknowledge that the model did not describe the elimination of TCDD from 4 the blood into the intestines, but it indirectly accounted for this phenomenon with the use of the 5 optimized elimination rate. 6 7 3.3.5.I.3. Im p a ct o f h um an in terin d ivid u a l variability. 8 The sources and extent of human variability suggested by the available data are presented 9 in Section 3.3.3, although there is some discussion of the impact of individual differences in 10 body fat content. The CADM model facilitates the simulation of body burden and serum lipid 11 concentrations on the basis of BMI and tissue weights of people, and the PBPK model simulates 12 alternative dose metrics in the fetus and in pregnant animals in addition to adult animals and 13 humans. However, neither of these models has been parameterized for simulation of population 14 kinetics and distribution of TCDD dose metrics. Therefore, at the present time, a quantitative 15 evaluation of the impact of human variability on the dose metrics of TCDD is not feasible, and 16 dose metric-based replacement of the default interindividual factor has not been attempted. 17 18 3.3.5.2. Q ualitative D iscu ssion o f U ncertainty in D o se M etrics 19 The usefulness of the CADM and PBPK models for conducting dose-response modeling 20 (rodent bioassays), interspecies (rodent to human) and intraspecies (high-dose to low-dose) 21 extrapolations is determined by their reliability in predicting the desired dose metrics. The 22 confidence in the model predictions of dose metrics is dictated by the extent to which the model 23 has been verified with empirical data relevant to the dose metric, supplemented by sensitivity 24 and uncertainty analyses. Analysis of sensitivity or uncertainty has not been conducted with the 25 CADM model. For the PBPK model, Emond et al. (2006, 197316) published the initial results 26 from sensitivity analyses of acute exposure modeling (see Section 3.3.3). One of the objectives 27 of a sensitivity analysis that is of highest relevance to this assessment is the identification of the 28 most critical model parameters with respect to the model output (i.e., dose metric). 29 If the model simulations have only been compared to entities that do not correspond to 30 the moiety representing the dose metric, or if the comparisons have only been done for some but 31 not all relevant dose levels, routes, and species, then the reliability in the predictions of dose This document is a draftfor review purposes only and does not constitute Agency policy. 3-49 DRAFT--DO NOT CITE OR QUOTE 1 metric can be an issue. The extent to which model results are uncertain will depend largely upon 2 the extent to which the dose metric is measurable (e.g., serum concentrations of TCDD) or 3 inferred (e.g., AhR-bound TCDD concentration). 4 With respect to TCDD body burden, whole-liver and blood concentration predictions in 5 the rat model, which are well-calibrated with measured data, uncertainty is relatively low. 6 Therefore the need for sensitivity and uncertainty analysis is less critical and confidence in these 7 dose metrics is high. For those dose metrics that are not directly measurable or are less easily 8 verified by available calibration methods, such as free-liver and AhR-bound concentrations, 9 sensitivity and uncertainty analyses are crucial for assessing the reliability of model predictions 10 and confidence is low. For the human model, calibration is largely dependent on blood (LASC) 11 TCDD meaurements, which are much less extensive than for the rat model. Because the blood 12 measurements are reported as LASC, uncertainty and variability in serum:blood and fat:serum 13 ratios also come into play when evaluating the adequacy of the whole-blood TCDD metric. 14 Furthermore, the human data are mostly representative o f much higher exposures than the 15 environmental exposures o f interest to the EPA. Because o f these additional uncertainties only 16 medium confidence can be held in the human model whole-blood TCDD concentration 17 predictions at higher exposures (observed effect range) and low-to-medium confidence at lower 18 exposures (background exposure range). 19 Sensitivity analysis for the Emond rat PBPK model predictions of liver TCDD 20 concentration indicated that hepatic CYP1A2 concentration is the most sensitive parameter 21 (Emond et al., 2006, 197316). For the Emond human PBPK model, the absorption parameters, 22 basal concentration of CYP1A2, and adipose tissue:blood partition coefficients were identified as 23 highly-sensitive parameters. 24 Confidence in the Emond rat and human PBPK models at high exposures is medium for 25 the purpose of rat-to-human extrapolation based on blood concentrations, given that the key 26 human model parameters are both sensitive and uncertain; confidence is low for lower 27 exposures. Conversely, confidence in the use of AhR-bound TCDD is low because of the large 28 uncertainty in the fraction of AhR-bound TCDD in the liver. 29 With regard to the predictability of body burden, the absorption and excretion parameters 30 were among the sensitive parameters in the rat. Several other parameters were also identified as 31 being sensitive in humans. Despite the sensitivity to these parameters and the uncertainty This document is a draftfor review purposes only and does not constitute Agency policy. 3-50 DRAFT--DO NOT CITE OR QUOTE 1 associated with individual parameter estimates, the overall confidence in the model predictions 2 of body burden appears to be high given the reproducibility of empirical data on tissue burdens 3 and blood concentrations of TCDD in various experiments by both models. Similar conclusions 4 can be drawn for blood concentration of TCDD predicted by the PBPK model, except that the 5 assigned value of blood (serum) lipid content will have additional impact on this dose metric to 6 the extent that the calibration data were in terms of LASC. Variability of total lipid levels and 7 variability of the contribution of phospholipids and neutral lipids to the total lipid pool across 8 species, lifestage and study groups is to be expected (Bernert et al., 2007, 594270; Poulin and 9 Theil, 2001, 594269). 10 Both conceptual (biological) relevance and prediction uncertainty are important in the 11 choice of dose metric for dose-response modeling and interspecies extrapolation. Conceptual 12 relevance has to do with how "close" the metric is to the observed effect, taking into account 13 both the target tissue and the MOA. In this context, a greater degree of confidence is held for 14 dose metrics that are more proximate to the event (i.e., specific effect). Prediction uncertainty 15 reflects the lack of confidence in the model predictions of dose metrics. Tables 3-13 and 3-14 16 provide a qualitative ranking of the importance and magnitude of each dose metric with respect 17 to these two sources of uncertainty. Conceptual relevance is low for the use of administered 18 dose in dose-response modeling because known (non-linear) physiological processes are ignored; 19 conversely, conceptual uncertainty is much lower for use of internal dose metrics more proximal 20 to the affected organs. 21 Table 3-13 presents a cross-walk of relevance, uncertainty and overall confidence 22 associated with the use of various dose metrics for dose-response modeling of TCDD. As shown 23 in Table 3-13, blood/serum levels have the highest overall confidence (medium) followed by 24 body burden (medium to low) for application in dose-response modeling. When using the mouse 25 PBPK model along with the human model (see Table 3-14), the contribution of the prediction 26 uncertainty to the overall uncertainty increases due to the limited comparison of the mouse 27 model simulations with empirical data. 28 29 3.3.6. Use of the Emond PBPK Models for Dose Extrapolation from Rodents to Humans 30 EPA has selected the Emond et al. (2004, 197315; 2005, 197317; 2006, 197316) PBPK 31 models, as modified by EPA for this assessment, for establishing toxicokinetically-equivalent This document is a draftfor review purposes only and does not constitute Agency policy. 3-51 DRAFT--DO NOT CITE OR QUOTE 1 exposures in rodents and humans.18 The 2003 Reassessment (U.S. EPA, 2003, 537122) 2 presented a strong argument for using the relevant tissue concentration as the effective dose 3 metric. However, no models exist for estimation of all relevant tissue concentrations. Therefore, 4 EPA has decided to use the concentration of TCDD in blood as a surrogate for tissue 5 concentrations, assuming that tissue concentrations are proportional to blood concentrations. 6 Furthermore, because the RfD and cancer slope factor are necessarily expressed in terms of 7 average daily exposure, the blood concentrations are expressed as averages over the relevant 8 period of exposure for each endpoint. Specifically, blood concentrations in the model 9 simulations are averaged from the administration of the first dose to the administration of the last 10 dose plus one dosing interval (time) unit in order to capture the peaks and valleys for each 11 administered dose. That is, for daily dosing, 24 hours of TCDD elimination following the last 12 dose is included in the average (the modeling time interval is one hour); for a weekly dosing 13 protocol, a full week is included. In addition, because of the accumulation of TCDD in fat and 14 the large differences in elimination kinetics between rodent species and humans, exposure 15 duration plays a much larger role in TK extrapolation across species than for rapidly-eliminated 16 compounds. Because of these factors, EPA is using discrete exposure scenarios that relate 17 human and rodent exposure durations. The use of discrete exposure scenarios was introduced 18 previously in Section 3.4.4.2 describing first-order kinetic modeling and is further described in 19 the following paragraphs. This section concludes with a quantitative evaluation of the impact of 20 exposure duration on the rodent-to-human TK extrapolation from both the human and rodent 21 "ends" of the process. 22 Figure 3-30 shows the TCDD blood concentration-time profile for continuous exposure 23 at 0.01 ng/kg-day, as predicted by the Emond human PBPK model, and the target TCDD 24 concentrations corresponding to the three discrete exposure scenarios used by EPA in this 25 document. The target concentrations are those that would be identified in the animal bioassay 26 studies that correspond to a particular POD (no-observed-adverse-effect level, lowest-observed27 adverse-effect level, or benchmark dose lower confidence bound) established for that bioassay. 28 That is, the target concentrations represent the toxicokinetically-equivalent internal exposure to 29 be translated into an equivalent human intake (or HED). 18The models will be referred to hereafter as the "Emond human PBPK model" and the "Emond rodent PBPK model," with variations when referring to individual species or components (e.g., gestational). This document is a draftfor review purposes only and does not constitute Agency policy. 3-52 DRAFT--DO NOT CITE OR QUOTE 1 For the lifetime exposure scenario, the HED is "matched" to the lifetime average TCDD 2 blood concentration from a lifetime animal bioassay result by determining the continuous daily 3 intake that would result in that average blood concentration for humans over 70 years. A table 4 for converting lifetime-average blood concentrations and other internal dose metrics to human 5 intake is presented in Appendix C.4. 6 For the gestational exposure scenario, the effective TCDD blood concentration (usually 7 the peak) determined for the particular POD in a particular developmental study is matched to 8 the average TCDD blood concentration over the gestational portion of the human gestational 9 exposure scenario. The HED is determined as the continuous daily intake, starting from birth 10 that would result in that average blood concentration over the 9-month gestational period for a 11 pregnancy beginning at 45 years of age. The choice of 45 years as the beginning age of 12 pregnancy is health protective of the population in that the daily exposure achieving the target 13 blood concentration is smaller than for earlier pregnancies. A table for converting average 14 gestational blood concentrations and other internal dose metrics to human intake for the 45-year15 old pregnancy scenario is presented in Appendix C.4. Also, a comparison of the 45-year old 16 pregnancy scenario to one beginning at age 25 is presented in Table 3-15. Using the 25 year-old 17 pregnancy scenario increases the HED by 30 to 60% for typical animal bioassay PODs (3 to 18 30 ng/kg). 19 For a less-than-lifetime exposure, the average TCDD blood concentration over the 20 exposure period in the animal bioassay associated with the POD is matched to the average over 21 the 5-year period that includes the peak concentration (58 years for an intake of 0.01 ng/kg-day). 22 The HED is determined as the continuous daily intake that would result in the target 23 concentration over peak 5-year period. The use of the peak is analogous to the approach in the 24 2003 Reassessment, where the terminal steady-state body burden played the same role. The 25 5-year average over the peak is taken to smooth out sharp peaks and more closely approximate a 26 plateau. The choice of peak is health protective because humans of any age must be protected 27 for short-term exposures, and the daily intake achieving a given TCDD blood concentration is 28 smallest when matched to the peak exposure as opposed to an average over shorter durations. 29 Thus, target concentrations for any exposure duration of less-than-lifetime must be averaged 30 backwards from the end of the lifetime scenario, rather than from the beginning. The only 31 exception would be if the short-term endpoints evaluated in the animal bioassay were associated This document is a draftfor review purposes only and does not constitute Agency policy. 3-53 DRAFT--DO NOT CITE OR QUOTE 1 with a specific life stage (such as for the gestational scenario). Note that this scenario lumps all 2 exposures from 1 day to over 1 year in rodents into the same less-than-lifetime category. 3 Conceptually, duration-specific scenarios could be constructed by defining equivalent rodent and 4 human exposure durations. However, for the most part, defining duration equivalents across 5 species is a somewhat arbitrary exercise, not generally based on physiologic or toxicologic 6 processes, but relying primarily on fraction-of-lifetime conversions. EPA defines "lifetime" 7 exposure as 2 years and 70 years for rodents and humans, respectively. So, a half-lifetime 8 equivalence of 1 year in rodents and 35 years in humans is defined easily. Also, considering a 9 subchronic exposure to be 10-15% of lifetime, leads to an equivalence of 90 days in rodents and 10 7-10 years in humans. However, in the practical sense with respect to the Emond human PBPK 11 model predictions, the difference in the dose-to-target-concentration ratios are not significantly 12 different from the peak 5-year average scenario, differing by less than 5%. A table for 13 converting less-than-lifetime average blood concentrations and other internal dose metrics to 14 human intake is presented in Appendix C.4. 15 The net effect of using three different scenarios for estimating the HED from rodent 16 exposures is that, for the same target concentration, the ratio of administered dose (to the rodent) 17 to HED will be larger for short-term exposures than for chronic exposures. Figure 3-31 is 18 similar to Figure 3-30, except that it shows the relationship of daily intake to a fixed target 19 TCDD blood concentration level. Figure 3-31 shows that, for human intakes of approximately 20 0.01 ng/kg-day, the difference in the defined scenarios is 40% or less, with a lifetime-scenario 21 daily intake of 0.014 ng/kg-day required to reach the same target concentration for a shorter-term 22 exposure of 0.01 ng/kg-day. The corresponding daily intake for the gestational scenario is 23 0.011 ng/kg-day. Because of the nonlinearities in the Emond human PBPK model, the 24 magnitude of the difference between the lifetime and less-than-lifetime exposure scenarios 25 increases at lower intake levels, but not to a substantial degree. 26 The differential effect of short- and long-term exposures is much more accentuated at the 27 rodent end of the exposure kinetic modeling. Analogous to the processes described in the 28 previous section for first-order body burden (see Section 3.4.2.2), the TCDD blood concentration 29 for single exposures is essentially the immediate absorbed fraction of the administered dose, 30 which will be somewhat lower than the administered dose, while for chronic exposure, the 31 TCDD blood concentration will reflect the long-term accumulation from daily exposure, which This document is a draftfor review purposes only and does not constitute Agency policy. 3-54 DRAFT--DO NOT CITE OR QUOTE 1 will be very much larger than the administered dose (expressed as a daily intake). Table 3-16 2 shows the overall impact of TK modeling on the extrapolation of administered dose to HED, 3 comparing the Emond PBPK and first-order body burden models. For comparison purposes, the 4 administered dose is fixed at 1 ng/kg-day for all model runs. Large animal-to-human TK 5 extrapolation factors (TKEF) are evident for short-term mouse studies, decreasing in magnitude 6 with increasing exposure duration. The only exception is the slightly lower extrapolation factor 7 for the mouse 1-day exposure, which is the result of the relatively short TCDD half-life (10 days) 8 in mice and the use of the peak TCDD blood concentration as representative of single exposures, 9 compared to the average TCDD blood concentration over the exposure period used for multiple 10 exposures. The TKEFs are lower for rats because of the slower elimination of TCDD in rats 11 compared to mice. Also, because of the nonlinear kinetics inherent in the Emond PBPK model, 12 the span of the HED (13-fold for mice) across these exposure durations is greater than the span 13 of the lipid-adjusted serum concentration (LASC; 4-fold for mice). Because of the dose14 dependence of TCDD elimination in the Emond model, the TKEFbecomes smaller with 15 decreasing intake. The result of this nonlinearity is that, although Table 3-16 shows much lower 16 TKEFs for the Emond PBPK model than for the first-order body burden metric, at much lower 17 HED levels the two models give much closer predictions. This document is a draftfor review purposes only and does not constitute Agency policy. 3 -5 5 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-1. Partition coefficients, tissue volumes, and volume of distribution 2 for TCDD in humans 3 Tissue Tissue/blood partition coefficient Tissue volume (liters, for a 60 kg person) Effective volume of distribution (Vd--liters of Percent blood equivalent) total Vd Blood 13 3 0.25 Fat 100 11.4 1.140 94.19 Liver 6 1.56 9 0.77 Rest of the body 1.5 38.64 58 4.79 Total 54.6* 1.210 100.00 4 5 *The total tissue volume presented here represents only 91% of body weight because some of the weight and 6 volume of the body is occupied by bone and other structures where TCDD uptake and accumulation do not occur to 7 a significant extent. 8 9 Source: Wang et al. (1997, 104657), Emond et al. (2005, 197317; 2006, 197316). 10 11 12 Table 3-2. Blood flows, permeability factors and resulting half lives (EA) for 13 perfusion losses for humans as represented by the TCDD PBPK model of 14 Emond et al. (2005, 197317; 2006, 197316) 15 Tissue Rate constant for Permeability (fraction of compartmental compartment blood flow) elimination (hour-1) U (hrs) Fat 0.12 0.0049 143 Liver 0.03 0.77 0.90 Rest of the body 0.35 3.84 0.18 This document is a draftfor review purposes only and does not constitute Agency policy. 3-56 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-3. Toxicokinetic conversion factors for calculating human equivalent 2 doses from rodent bioassays 3 Half-life (days)a Mouse 10 Rat (Wistar) Rat (other) 20 25 Guinea pig 40 Exposure duration (days) Conversion factor (CF)bBBA(#A):dAgiven in parentheses 1 3882 (0.77) 3815 (0.79) 3802 (0.79) 3783 (0.79) 7 1107 (2.71) 1020 (2.94) 1004 (2.99) 979 (3.07) 14 681 (4.41) 587 (5.11) 569 (5.27) 543 (5.53) 28 453 (6.62) 350 (8.56) 331 (9.06) 303 (9.90) 90 307 (9.76) 186 (16.1) 163 (18.4) 130 (23.0) 180 282 (10.6) 154 (19.5) 129(23.2) 93 (32.1) 365 270 (11.1) 141 (21.3) 115(26.0) 77 (38.9) 730 226 (11.3) 115 (22.2) 93 (27.4) 60 (42.5) O n O i -f^ aHalf-life for humans = 2,593 days (7.1 years). d = dA/CF; B B dtH )dH = 2,185 (1-180 days), 2,202 (365 days), 2,555 (730 days). This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 5 7 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-4. Equations used in the concentration and age-dependent model 2 (CADM; Aylward et al., 2005, 197014)a 3 Parameter Equation Hepatic Concentration (ng/kg) body (f max f. min ) * e bb<ody hepatic w, min K + C body Fat Concentration (ng/kg) c adipose ^Qbody *(1 ( Wa (f ( f max - Jfmin )' * c body) ) ( f min + K + Cbody )) Hepatic Elimination Excretion via gut of Unchanged TCDD (Exsorption) Exr _ hepatic = k e * Qbody * (1 - ( / min + (/ max / .) min * e body )) K + Cbody Exr _ gut = k a*Qa Change of TCDD due to bodyweight change (B W (t + dt) - B W (t)) ChangeTCDD _ B W = Qbodf BW (t) Amount in body as a function of time Qbocfy(t + dt) - Q' bbooddyy(t) = Exr _ hepatic + Exr _ gut + ChangeTCDD _ BW Adipose tissue growth w _ 1.2 * BM I + (0.23* Age) - 10.8* sex 100 Change of hepatic elimination constant with age k e = k e0 k eUope* Age 4 aFor abbreviations and parameter descriptions, see Table 3-5. This document is a draftfor review purposes only and does not constitute Agency policy. 3 -5 8 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-5. Parameters of the Concentration and Age-Dependent Model 2 (CADM; Aylward et al., 2005, 197014) 3 Parameter Value Units Comments/sources fhmina 0.01 unitless Minimumbody burden fraction in liver fhmaxa 0.7 unitless Maximumbody burden fraction in liver Ka 100 ng/kg Body burden at half-maximum of fraction liver ke Calculated per year ke= ke0- ke_slope*(age) with enforced minimum of ke min ke0 0.85 per year CADM-mean hepatic eliminationbase rate at age 0 ke slope k^e min 0.011 0.2 per year Change in keper year of age per year Minimum hepatic elimination rate wa(adipose weight fraction) Calculated unitless wa= [(1.2*BMI)+0.23*Age-10.8*sex]/100 wh (liver body weight fraction) 0.03 unitless Assumed constant ka (adipose clearance factor) 0.0025 per month Passive elimination rate from intestinal tract Monthly dose 0.15507069 ng per month Estimated absorption fraction 0.97 unitless From Moser and McLaghlan (2001, 198045) Body weight 70 kg Standard male weight Sex 1 unitless 1= male; 0 = female Time of administration 840 months Initial Cbody 0.2 ng/kg Estimated background young adults UMDES sampling Absorbed monthly dose 1 0.150418569 ng per month 4 5 aThe values of fhmin, fhmax, and K were obtained by best fit of the model simulations to the experimental data with 6 the method of least squares (Aylward et al., 2005, 197114; Carrier et al., 1995, 197618). This document is a draftfor review purposes only and does not constitute Agency policy. 3_5 9 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-6. Confidence in the CADMamodel simulations of TCDD dose 2 metrics 3 Dose metric Level of confidence Administered dose N/A Absorbed dose H Body burden H Serum lipid concentration M Total tissue (liver) concentration L Receptor occupancy (bound concentration) N/A O n O i -f^ aConcentration and age-dependent model (Aylward et al., 2005, 197014). H = high, M = medium, L = low, NA = not applicable. This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 6 0 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-7. Equations used in the TCDD PBPK model of Emond et al. (2006, 2 197316) 3 Aspect Equation Body weight growth with age BW,. (g) =BW T0 x f 0.41 x time --------------------- \ - ^ 1402.5 + time) Cardiac output Qc(mL / hr) = QCCAR x f 601 BW Y '75 ) A factor of 60 corresponds to the conversion of minutes to hours, and 1,000 is conversion of BW from grams to kilograms. Blood compartment \ ( Qf x Cfb ) + ( Qre x Creb ) + ( Qli x Clib ) + lymph] ( Cb x CLURI ) Cb(nmol / mL) = - Qc Qc Tissue compartment (fat, rest of the body) Tissue blood subcompartment dAtb (nmol / mL) =Qt(Ca Ctb) PAt^Ctb j Atb Ctb(nmol / mL) = Wtb Tissue cellular matrices (nmol / mL) =PAt ^Ctb - Ct(nmol / mL) = At Wt Liver tissue compartment Tissue blood subcompartment cdAlb (nmol / mL) =Qli(Ca - Clib) - PALI(Clib - Clifree) +inputoral dt Clib(nmol / mL) = WLIB Tissue cellular matrices d A (nmol / mL) =PALI(Clib - Clifree) - (KBILE LI x Clifree x WLI) dt Cli(nmol / mL) = Ali Wli Free TCDD concentration in liver Clifree(nmol / mL) = Cli - "Clifree x PLI + 1( LIBMAX x Clifree \ + ( CYP1A2 x Clifree Y ^ KDLI + Clifree J { KDLI1A2 + Clifree) \ Concentration bound to AhR in hepatic tissue LIBMAX x Clifree CtAhRbound ( nm / mL) = KDLI + Clifree All other induction processes and equations have been described and presented by Wang et al. (1997, 104657). This document is a draftfor review purposes only and does not constitute Agency policy. 3-61 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-7. Equations used in the TCDD PBPK model of Emond et al. (2006, 197316) (continued) Aspect Equation Gastrointestinal absorption and distribution of TCDD to the portal lymphatic circulation Amount of TCDD remaining in lumen cavity dLttm en (nm ol / h r) = [(K S T + K A B S ) x lum en J + intake Lumen in the amount of TCDD remaining in the GI tract (nmol); intake is the rate of intake of TCDD during a subchronic exposure (nmol/hr). Amount of TCDD eliminated in the feces dFeces (nmol / hr) = K S T x lum en dt Absorption rate of TCDD to the blood via the lymphatic circulation dL y m p h ( n m o l / h r ) = K A B S x lu m e n x 0.7 Absorption rate of TCDD by the liver via portal circulation dP ortal ( n m o l / h r ) = K A B S x lu m e n x 0.3 dt 2 3 Note: Key parameters and abbreviations are defined in Table 3-10. This document is a draftfor review purposes only and does not constitute Agency policy. 3 - 6 2 DRAFT--DO NOT CITE OR QUOTE Table 3-8. Parameters of the PBPK model for TCDD This document is a draftfor review purposes only and does not constitute Agency policy. 3-63 DRAFT: DO NOT CITE OR QUOTE Parameter Description Symbol Body weight (g) BW Cardiac output (mL/hour/kg) QCCAR Tissue (intracellular) volumes (fraction of BW) Liver WLI0 Fat WF0 Tissue blood volumes Liver (fraction of WLI0) WLIB0 Fat (fraction of WF0) WFB0 Rest of body (fraction of WRE0) WREB0 Placenta tissue fraction of tissue blood weight (unitless) WPLAB0 Tissue blood flow (fraction of cardiac output) Liver QLIF Fat QFF Placenta QPLAF Tissue permeability (fraction of tissue blood flow) Liver PALIF Fat PAFF Placenta diffusional permeability fraction (unitless) PAPLAF Rest of body PAREF Parameter values Human Human Mouse Mouse Rat Rat nongestationala gestationala nongestational gestational nongestational gestational Calculated Calculated 23-28b 23-28 125-250b 85-190b 15.36cd Calculated 275c 275c 311.4e 311.4e Calculated Calculated Calculated Calculated 0.0549f 0.069e 0.0549f Calculated 0.036e 0.069e 0.036e Calculated 0.266e 0.05e 0.03e N/A 0.266e 0.05e 0.03e 0.5g 0.266e 0.05e 0.03e N/A 0.266e 0.05 e 0.03 e 0.5e 0.266e 0.05e 0.03e N/A 0.266e 0.05e 0.03e 0.5e 0.26c 0.05c N/A 0.35e 0.12i N/A 0.03e 0.26c 0.05c Calculated 0.35e 0.12i 0.3g 0.03e 0.161f 0.07h N/A 0.161f 0.07h Calculated 0.183e 0.069e N/A 0.183e 0.069e Calculated 0.35e 0.12i N/A 0.03e 0.35e 0.12i 0.03g 0.03e 0.35e 0.091e N/A 0.0298e 0.35e 0.091e 0.3g 0.0298e Table 3-8. Parameters of the PBPK model for TCDD (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 3-64 DRAFT: DO NOT CITE OR QUOTE Parameter Description Partition coefficient Liver Fetus/blood partition coefficient (unitless) Placenta/blood partition coefficient (unitless) Fat Rest of body Metabolism constants Urinary clearance elimination (mL/hour) Clearance - transfer from mother to fetus (mL/hour) Liver (biliary elimination and metabolism; hour-1) Interspecies constant (hour-1) AhR Affinity constant in liver (nmol/mL) Binding capacity in liver (nmol/mL) Placenta binding capacity (nmol/mL) Affinity constant protein (AhR) in placenta (nmol/mL) Symbol PLI PFETUS PPLA PF PRE CLURI CLPLA_FET KBILE_LI Kelv KDLI LIBMAX PLABMAX KDPLA Parameter values Human Human Mouse Mouse Rat Rat nongestationala gestationala nongestational gestational nongestational gestational 6e 6e 6e 6 e 6e 6e N/A 4 J N/A 4J N/A 4J N/A 1.5j N/A 3g N/A 1.5J 100e 100e 400i 400i 100e 100e 1.5e 1.5e 3k 3k 1.5e 1.5e 4.17E-08l N/A Inducible 0.00111 4.17E-08l 16e Inducible 0.0011i 0.09i N/A Inducible 0.4i 0.09i 0.17i Inducible 0.4i 0.01J N/A Inducible 0.15e 0.01J 0.17i Inducible 0.15e 0.1e 0.35e N/A N/A 0.1e 0.35e 0.2J 0.1J 0.0001e 0.00035e N/A N/A 0.0001e 0.00035e 0.0002J 0.0001J 0.0001e 0.00035e N/A N/A 0.0001e 0.00035e 0.0002J 0.0001J Table 3-8. Parameters of the PBPK model for TCDD (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 3-65 DRAFT: DO NOT CITE OR QUOTE Parameter Description CYP1A2 induction parameters Dissociation constant CYP1A2 (nmol/mL) Degradation process CYP1A2 (nmol/mL) Dissociation constant during induction (nmol/mL) Basal concentration of CYP1A2 (nmol/mL) First-order rate of degradation (hour-1) Time delay before induction process (hour) Maximal induction of CYP1A2 (unitless) Other constants Oral absorption constant (hour-1) Gastric nonabsorption constant (hour-1) Symbol Parameter values Human Human Mouse Mouse Rat Rat nongestationala gestationala nongestational gestational nongestational gestational KDLI2 CYP1A2_1OUTZ CYP1A2_1EC50 CYP1A2_1A2 CYP1A2_1KOUT CYP1A2_1TAU CYP1A2_1EMAX 40> 1,600e 130e 1,600e 0.1e 0.25e 9,300` 40> 1,600e 130e 1,600e 0.1e 0.25e 9,300i 0.02i 1.6e 0.13e 1.5k 0.1e 1.5k 600e 0.02i 1.6e 0.13e 1.5k 0.1e 1.5k 600e 0.04J 1.6e 0.13e 1.6e 0.1e 0.25e 600e 0.04J 1.6e 0.13e 1.6e 0.1e 0.25e 600e KABS KST 0.061 0.01m 0.06i 0.01m 0.4s1 0.30i 0.4s1 0.30i 0.4Se 0.36e 0.4Se 0.36e aUnits for human nongestational parameters are L rather than mL and kg rather than g where applicable. bBody weight varies by study (Emond et al., 2004, 197315). cKrishnan and Andersen (2007). dUnits are L/kg/hr. eWang et al. (1997, 104657). fILSI (1994, 046436). gFixed. hLeung et al. (1990, 192833). `Optimized. JEmond et al. (2004, 197315). kWang et al. (2000, 198738). 'Lawrence and Gobas (1997, 199072). mCalculated to estimate 87% bioavailability of TCDD in humans (Poiger and Schlatter, 1986, 197336). 1 Table 3-9. Regression analysis results for the relationship between log10 2 serum TCDD at the midpoint of observations and the log10 of the rate 3 constant for decline of TCDD levels using Ranch Hand data 4 5 6 7 Table 3-10. Confidence in the PBPK model simulations of TCDD dose 8 metrics 9 Dose metric Human model Rat model Mouse model Administered dose N/A N/A N/A Absorbed dose H HM Body burden H HM Serum (blood)concentration H HM Total liver concentration M /L H M Receptor occupancy (bound concentration) 10 11 H = high, M = medium, L = low. L LL This document is a draftfor review purposes only and does not constitute Agency policy. 3-66 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-11. Overall confidence associated with alternative dose metrics for 2 cancer and noncancer dose-response modeling for TCDD using rat PBPK 3 model 4 End point Body burden Bound Blood or serum Liver concentration in concentration concentration liver L iv e r effects M H M /L Nonhepatic effects M H M /L 5 6 H = high, M = medium, L = low. 7 8 9 Table 3-12. Overall confidence associated with alternative dose metrics for 10 cancer and noncancer dose-response modeling for TCDD using mouse PBPK 11 model 12 End point Body burden Bound Blood or serum Liver concentration in concentration concentration liver L iv e r effects M ML Nonhepatic effects M M L 13 14 H = high, M = medium, L = low. 15 16 17 Table 3-13. Contributors to the overall confidence in the selection and use of 18 dose metrics in the dose-response modeling of TCDD based on rat and 19 human PBPK models 20 Dose metric Conceptual Relevance Prediction uncertainty Overall Confidence Administered dose L NA L Body burden M M M -L Blood concentration M L M L iv e r concentration L ML Receptor (A hR) H HL occupancy 21 22 H = high, M = medium, L = low, NA = not applicable, ? = if relevant to MOA of response. This document is a draftfor review purposes only and does not constitute Agency policy. 3-67 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-14. Contributors to the overall uncertainty in the selection and use 2 of dose metrics in the dose-response modeling of TCDD based on mouse and 3 human PBPK models 4 Dose metric Conceptual uncertainty Prediction uncertainty Administered dose H NA Absorbed dose HL Body burden MM Blood or serum concentration M M Tissue concentration L MH Receptor occupancy L(?) H 5 6 H = high, M = medium, L = low, NA = not applicable, ? = if relevant to MOA of response. 7 8 9 Table 3-15. Comparison of human equivalent doses from the Emond human 10 PBPK model for the 45-year-old and 25-year-old gestational exposure 11 scenarios 12 Animal bioassay POD (ng/kg-day) Species TCDD blood concentration HED HED 25-yr:45-yr 45 year-old 25 year-old ratio M ouse 8.800E-02 3 Rat 1.815E-01 6.79E-04 1.87E-03 1.03E-03 2.98E-03 1.5 1.6 M ouse 7.115E-01 30 Rat 1.367E+00 1.51E-02 4.22E-02 2.07E-02 5.41E-02 13 14 ^Determined from the Emond rodent PBPK models assuming a single exposure on GD13. 1.4 1.3 This document is a draftfor review purposes only and does not constitute Agency policy. 3-68 DRAFT--DO NOT CITE OR QUOTE 1 Table 3-16. Impact of toxicokinetic modeling on the extrapolation of 2 administered dose to HED, comparing the Emond PBPK and first-order 3 body burden models 4 Exposure duration (days) 1st-order BB HED (ng/kg-day) TKef Emond PBPK LASC HED (ng/kg) (ng/kg-day) TKef Mouse 1 2.57E-4 3,882 75.5 9.49E-4 1,054 14 1.47E-3 681 64.4 8.17E-4 1,224 90 3.25E-3 307 173 3.83E-3 261 365 3.70E-3 270 248 6.66E-3 150 730 4.43E-3 226 263 1.08E-2 93 Rat 1 2.63E-4 3,802 110 1.87E-3 535 14 1.76E-3 569 208 5.22E-3 192 90 6.13E-3 163 599 2.81E-2 36 365 8.68E-3 115 811 4.52E-2 22 730 5 1.07E-2 93 853 6.47E-2 15 This document is a draftfor review purposes only and does not constitute Agency policy. 3-69 DRAFT--DO NOT CITE OR QUOTE 1 2 4 D 7 Day Liver/Fat 14 Day Liver/Fat ta o 21 Day Liver/Fat I 35 Day Liver/Fat 3 ta V8I t--a 2 U 8O U c 1 a d 0 10 Dose ^g/Kg 3 4 Figure 3-1. Liver/fat concentration ratios in relation to TCDD dose at 5 various times after oral administration of TCDD to mice. 6 7 Source: D ilberto et al. (1995, 197309). This document is a draftfor review purposes only and does not constitute Agency policy. 3-70 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 3-2. First-order elimination rate fits to 36 sets of serial TCDD 3 sampling data from Seveso patients as function of initial serum lipid TCDD. 4 5 Source: A y lw a rd et al. (2005, 197014) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-71 DRAFT--DO NOT CITE OR QUOTE Fract Fecal Elim/Year 1 2 3 Figure 3-3. Observed relationship of fecal 2,3,7,8-TCDD clearance and 4 estimated percent body fat. 5 6 Source: R ohde et al. (1999, 548764) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-72 DRAFT--DO NOT CITE OR QUOTE TCDD Elimination Half-Life (Years) 1 % Body Fat 2 3 Figure 3-4. Unweighted empirical relationship between percent body fat 4 estimated from body mass index and TCDD elimination half-life--combined 5 Ranch Hand and Seveso observation. This document is a draftfor review purposes only and does not constitute Agency policy. 3-73 DRAFT--DO NOT CITE OR QUOTE Functional biomarkers Receptor occupancy Total tissue concentration Blood or serum concentration Absorbed Intake 1 2 Figure 3-5. Relevance of candidate dose metrics for dose-response modeling, 3 based on mode of action and target organ toxicity of TCDD. 4 This document is a draftfor review purposes only and does not constitute Agency policy. 3-74 DRAFT--DO NOT CITE OR QUOTE This document is a draftfor review purposes only and does not constitute Agency policy. 3-75 DRAFT: DO NOT CITE OR QUOTE Experimental Applied Dose \ d (1- e k )fa Body BurdenRat(t) = BB(0)e kt + k *- Body BurdenRat(t) Human d H Estimated 4 = t 1/2A dA t 1/2H (1 (1 - e kjA ) e - kHtH ) Exposure Body BurdenHuman ( t) Figure 3-6. Process of estimating a human-equivalent TCDD lifetime average daily oral exposure (dH) from an experimental animal average daily oral exposure (dA) based on the body-burden dose metric. The arrows represent mathematical conversions based on toxicokinetic modeling. BBA (TWA animal body burden) and BBH(TWA human body burden) are assumed to be toxicokinetically equivalent. See text for further explanation. 0 500 1000 1500 2000 2500 3000 1 2 3 Figure 3-7. Human body burden time profiles for achieving a target body 4 burden for different exposure duration scenarios. BB:d is BBH(tH):dHin 5 Figure 3-6. The curve depicted using the solid line illustrates the increase in the 6 human body burden over time for a hypothetical human administered a daily 7 TCDD dose where the time-weighted average human body burden estimate over 8 the lifetime is equal to the target body burden attained in a rodent bioassay. When 9 compared to shorter durations (dashed lines), a higher average daily TCDD dose 10 is required to yield a time-weighted average human body burden over a lifetime 11 that is equal to the target body burden attained in a rodent bioassay. The half 12 chronic exposure scenario (depicted using a dashed line) is equivalent to a 1-year 13 exposure in rodents. When compared to a chronic BBH, a lower value of dHis 14 needed to attain the target body burden in a rodent bioassay when the time15 weighted average is over the last 35 years of life; the dose to plateau ratio is also 16 smaller (i.e., dH,C< dH,SCto attain the target body burden in a rodent bioassay). 17 The shorter exposure scenario is equivalent to most other shorter rodent exposure 18 durations, from 1 day to subchronic, which are indistinguishable with respect to 19 the BB:d ratio (subchronic shown). This document is a draftfor review purposes only and does not constitute Agency policy. 3-76 DRAFT--DO NOT CITE OR QUOTE DISTRIBUTION ELIMINATION 1 2 3 Figure 3-8. Schematic of the CADM structure. 4 5 Source: A y lw a rd et al. (2005, 197014) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-77 DRAFT--DO NOT CITE OR QUOTE 1 Figure 3-9. Comparison of observed and simulated fractions of the body 2 burden contained in the liver and adipose tissues in rats. f h, fraction contained 3 in liver (observation) ();fh-sim, fraction contained in liver (simulation) (--);f at, 4 fraction contained in the adipose tissue (observation) (); f at-sim, fraction contained 5 in the adipose tissue (simulation) (--); and Cb,body concentration in ng TCDD/kg 6 body wt. 7 8 Source: Carrier et al. (1995, 197618); data from Abraham et al. (1988, 199510) 9 measured 7 days after dosing. This document is a draftfor review purposes only and does not constitute Agency policy. 3-78 DRAFT--DO NOT CITE OR QUOTE Blood systemic circulation 1 2 3 Figure 3-10. Conceptual representation of PBPK model for rat exposed to 4 TCDD. 5 6 Source: Em ond et al. (2006, 197316) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-79 DRAFT--DO NOT CITE OR QUOTE Arterial blood 1 2 Figure 3-11. Conceptual representation of PBPK model for rat 3 developmental exposure to TCDD. 4 5 Source: Em ond et al. (2004, 197315) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-80 DRAFT--DO NOT CITE OR QUOTE 1 2 3 Figure 3-12. TCDD distribution in the liver tissue. 4 5 Source: W ang et al. (1997, 104657) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-81 DRAFT--DO NOT CITE OR QUOTE 1 2 3 Figure 3-13. Growth rates for physiological changes occurring during 4 gestation. (a) Placental growth during gestation (calculated for n = 10 placenta). 5 Experimental data from Sikov (1970, 594274). (b) Blood flow rate in Placental 6 compartment during gestation. Experimental data from Buelke-Sam et al. (1982, 7 020478; 1982, 020477). (c) Fat fraction of body weight during gestation. 8 Experimental data came from Fisher et al. (1989, 065288), and (d) Fetal growth 9 during gestation. Experimental data obtained from Sikov (1970, 594274). 10 11 This document is a draftfor review purposes only and does not constitute Agency policy. 3-82 DRAFT--DO NOT CITE OR QUOTE This document is a draftfor review purposes only and does not constitute Agency policy. 3-83 Figure 3-14. Comparisons of model predictions to experimental data using a fixed elimination rate model with hepatic sequestration (A) and an inducible elimination rate model with (B) and without (C) hepatic sequestration. E X B L , experimental blood levels. M odel predictions w ere compared w ith the data o f Santostefano et al. (1998, 200001), w here fem ale rats w ere exposed to a single oral dose o f 10 pg o f T C D D /k g B W . E rro r bars are SD. Source: Edm o n d et al. (2006, 197316). D RA FT: DO NOT C IT E OR Q U O TE 1 2 00.00 10.00 1,750 ng TCDD/kg BW 500 ng TCDD/kg BW 150 ng TCDD/kg BW 50 ng TCDD/ka BW 0.10 = Time week 3 Figure 3-15. PBPK model simulation of hepatic TCDD concentration (ppb) 4 during chronic exposure to TCDD at 50, 150, 500, 1,750 ng TCDD/BW using 5 the inducible elimination rate model compared with the experimental data 6 measured at the end of exposure. 7 8 Source: Em o n d et al. (2006, 197316). This document is a draftfor review purposes only and does not constitute Agency policy. 3-84 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 3-16. Model predictions of TCDD blood concentration in 10 veterans 3 (A-J) from Ranch Hand Cohort. 4 Source: Em o n d et al. (2005, 197317). This document is a draftfor review purposes only and does not constitute Agency policy. 3-85 DRAFT--DO NOT CITE OR QUOTE 1 Figure 3-17. Time course of TCDD in blood (pg/g lipid adjusted) for two highly exposed Austrian women (patients 1 and 2). Sym bols represent measured concentrations, and lines represent model predictions. These data were used as part o f the model evaluation (G eusau et al., 2002, 594259) . Source: Em o n d et al. (2005, 197317) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-86 DRAFT--DO NOT CITE OR QUOTE Ln(pg/g TCDD) 1 2 Pgg P 3 4 Figure 3-18. Observed vs. Emond et al. (2005, 197317) model simulated 5 serum TCDD concentrations (pg/g lipid) over time (ln = natural log) in two 6 Austrian women. D ata from Geusau et al. (2002, 594259) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-87 DRAFT--DO NOT CITE OR QUOTE Ln(TCDD Serum Decline/Yr) Log(TCDD pg/g at Midpoint Obs) 1 2 Figure 3-19. Comparison of the dose dependency of TCDD elimination in the 3 Emond model vs. observations of nine Ranch Hand veterans and two highly 4 exposed Austrian patients. C ircles are observed data. This document is a draftfor review purposes only and does not constitute Agency policy. 3-88 DRAFT--DO NOT CITE OR QUOTE 1 WLIO a wfo Paraieler PF kst KABS CYP1A2J0UTZ CYP1A2JA2 ~ r ~ r --r ~ r T -8 -6 - A -2 0 2 Percent of change 10 Parameter WLIO WFO PF KST KABS KDLI LIE MAX KDLI2 CVP1A_10UTZ CYP1A2JEHAX C-TP1A2JEC50 CVP1A_1A2 B 1 _____ 1 -------- 1 -------------1 1 1-------------- 1 1 1 ------1 1------1 1 1 1 1 ~l 1 --------------------------------------------------------------------------1 -20 -15 -10 -5 0 5 10 15 20 Percent of change 3 4 Figure 3-20. Sensitivity analysis was performed on the inducible elimination 5 rate. The analysis w as performed at 0.001 pg/kg (A ) and at 10 pg/kg (B ). The 6 blue and white bars are results from -1 0 % and +10% changes, respectively. 7 8 Source: Em ond et al. (2006, 197316) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-89 DRAFT--DO NOT CITE OR QUOTE A 1 2 Figure 3-21. Experimental data (symbols) and model simulations (solid lines) 3 of (A) blood, (B) liver and (C) adipose tissue concentrations of TCDD after 4 oral exposure to 150 ng/kg-day, 5 days/week for 17 weeks in mice. Y -ax is 5 represents concentration in pg/g and X -a x is represents tim e in days. 6 7 Source: Experim ental data w ere obtained form D ilib erto et al. (2001, 197238) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-90 DRAFT--DO NOT CITE OR QUOTE Dose (ug/kg) 1 2 Figure 3-22Comparison of PBPK model simulations with experimental data 3 on liver concentrations in mice administered a single oral dose of 0.001-300 4 pg TCDD/kg. The sim ulations and experimental data w ere obtained 24 hour 5 post-exposure. 6 7 Source: D ata obtained from B o v erh o ff et al. (2005, 594260) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-91 DRAFT--DO NOT CITE OR QUOTE Dose (ng/kg) 1 2 Figure 3-23. Comparison of model simulations (solid lines) with 3 experimental data (symbols) on the effect of dose on blood (cb), liver (cli) and 4 fat (cf) concentrations following repetitive exposure to 0.1-450 ng TCDD/kg, 5 5 days/week for 13 weeks in mice. 6 7 Source: D ata obtained from D ilib erto et al. (2001, 197238) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-92 DRAFT--DO NOT CITE OR QUOTE A B 1 2 Figure 3-24. Comparison of experimental data (symbols) and model 3 predictions (solid lines) of (A) blood, (B) liver and (C) adipose tissue 4 concentrations of TCDD after oral exposure to 1.5 ng/kg-day, 5 days/week 5 for 17 weeks in mice. Y -a x is represents concentration in pg/g and X -a x is 6 represents time in days. 7 8 Source: Experim ental data w ere obtained form D ilib erto et al. (2001, 197238) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-93 DRAFT--DO NOT CITE OR QUOTE A B C 1 2 Figure 3-25. Comparison of experimental data (symbols) and model 3 predictions (solid lines) of (A) blood concentration, (B) liver concentration, 4 (C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary 5 elimination (% dose) of TCDD after oral exposure to 1.5 ng/kg-day, 6 5 days/week for 13 weeks in mice. Y -a x is represents concentration in pg/g and 7 X -a xis represents time in days. 8 Source: Experim ental data w ere obtained form D ilib erto et al. (2001, 197238) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-94 DRAFT--DO NOT CITE OR QUOTE A B C D 1 2 Figure 3-26. Comparison of experimental data (symbols) and model 3 predictions (solid lines) of (A) blood concentration, (B) liver concentration, 4 (C) adipose tissue concentration (D) feces excretion (% dose) and (E) urinary 5 elimination (% dose) of TCDD after oral exposure to 150 ng/kg-day, 6 5 days/week for 13 weeks in mice. Y -a x is represents concentration in pg/g and 7 X -axis represents time in days. 8 Source: Experim ental data w ere obtained form D ilib erto et al. (2001, 197238) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-95 DRAFT--DO NOT CITE OR QUOTE A B C D E F 1 2 Figure 3-27. PBPK model simulations (solid lines) vs. experimental data 3 (symbols) on the distribution of TCDD after a single acute oral exposure to 4 A-B) 0.1, C-D) 1.0 and E-F) 10 pg of TCDD/kg of body weight in mice. 5 L iv e r and adipose concentration for each dose w as measured after 72 hours. 6 Y -a x is represents the concentration in tissues (ng/g); insets A , C , and E represent 7 liver tissue, whereas B , D , and F correspond to adipose tissue. X -a x is represents 8 the time in hours. 9 Source: experimental data w ere obtained from Santostefano et al. (1996, 594258) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-96 DRAFT--DO NOT CITE OR QUOTE A C 1 2 Figure 3-28. PBPK model simulation (solid lines) vs. experimental data 3 (symbols) on the distribution of TCDD after a single dose of 24 pg/kgBW on 4 GD 12 in mice. Concentrations expressed as ng T C D D / g tissue. (A ) maternal 5 blood, (B ) maternal liver and (C ) maternal adipose tissue. Y -a x is represents the 6 tissue concentration whereas X -axis represents the time in hours. 7. 8 Source: Experim ental data w ere obtained from (Abbott et al., 1996, 155093) . This document is a draftfor review purposes only and does not constitute Agency policy. 3-97 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 3-29. Comparison of the near-steady-state body burden simulated 3 with CADM and Emond models for a daily dose ranging from 1 to 4 10,000 ng/kg-day in rats and humans. The rat model w as run for 13 w eeks and 5 the human model w as run from age 20 to 30. Th e tim e-averaged concentration 6 was used for each. This document is a draftfor review purposes only and does not constitute Agency policy. 3-98 DRAFT--DO NOT CITE OR QUOTE This document is a draftfor review purposes only and does not constitute Agency policy. 3-99 D RA FT: DO NOT C IT E OR Q U O TE Figure 3-30. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure scenarios, with target concentrations shown for each; profiles generated with Emond human PBPK model. This document is a draftfor review purposes only and does not constitute Agency policy. 3-100 D RA FT: DO NOT C IT E OR Q U O TE 0 20 40 60 Year Figure 3-31. TCDD serum concentration-time profile for lifetime, less-than-lifetime and gestational exposure scenarios, showing continuous intake levels to fixed target concentration; profiles generated with Emond human PBPK model. 1 4. CHRONIC ORAL REFERENCE DOSE 2 3 4 This section presents U.S. Environmental Protection Agency (EPA)'s response to the 5 National Academy of Sciences (NAS) recommendations that EPA more explicitly discuss the 6 modeling of noncancer endpoints and develop a reference dose (RfD) to address noncancer 7 effects associated with oral 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) exposures. Section 2 8 details the selection of the animal studies with the lowest TCDD doses associated with the 9 development of adverse noncancer effects and the selection of relevant epidemiologic studies of 10 adverse noncancer health effects. Section 3 discusses the kinetic modeling and estimation of 11 human equivalent daily oral doses that are used in TCDD RfD development in this section. This 12 section discusses the modeling of noncancer health effects data associated with TCDD exposure 13 and the derivation of an RfD. Specifically, Section 4.1 summarizes the NAS comments on 14 TCDD dose-response modeling and EPA's response, including justification of selected 15 noncancer effects and statistical characterization of modeling results. Section 4.2 presents the 16 TCDD dose-response modeling undertaken for identification of candidate points of departure 17 (PODs) for derivation of an RfD. In Section 4.3, EPA derives an RfD for TCDD. Finally, 18 Section 4.4 describes the qualitative uncertainties in the RfD. 19 20 4.1. NAS COMMENTS AND EPA'S RESPONSE ON IDENTIFYING NONCANCER 21 EFFECTS OBSERVED AT LOWEST DOSES 22 The NAS recommended that EPA identify the noncancer effects associated with low dose 23 TCDD exposures and discuss its strategy for identifying and selecting PODs for noncancer 24 endpoints, including biological significance of the effects. 25 26 With respect to noncancer end points, the committee notes that EPA does not use 27 a rigorous approach for evaluating evidence from studies... (NAS, 2006, 28 198441p. 47) 29 30 The Reassessment should describe clearly the following aspects: 31 1. The effects seen at the lowest body burdens that are the primary focus for any 32 risk assessment--the "critical effects." 33 2. The modeling strategy used for each noncancer effect, paying particular 34 attention to the critical effects, and the selection of a point of comparison based 35 on the biological significance of the effect; if the ED01 is retained, then the This document is a draftfor review purposes only and does not constitute Agency policy. 4-1 DRAFT--DO NOT CITE OR QUOTE 1 biological significance o f the response should be defined and the precision of 2 the estimate given... (N A S , 2006, 198441 p. 187). 3 4 In this document, E P A has developed a strategy for identifying the noncancer data sets 5 and P O D s that represent the most sensitive and biologically relevant endpoints for derivation of 6 an R fD for T C D D . E P A began this process by using the animal bioassays and human 7 epidem iologic studies that met its study inclusion criteria as sources of these data sets. 8 F o r all epidemiologic studies that were identified as suitable for further quantitative 9 dose-response analyses in Section 2.4.3, E P A has chosen to identify P O D s (i.e., estimates o f a 10 no-observed-adverse-effect level [ N O A E L ] or lowest-observed-adverse-effect level [ L O A E L ] ; 11 m odeling o f a benchm ark dose low er confidence bound [B M D L ] w as not possible given the data 12 presented in these studies). Fig u re 4-1 shows E P A ' s process to select and identify candidate 13 P O D s from these key epidem iologic studies. E P A first evaluated the dose-response inform ation 14 in the study to determine whether it provided an estimate o f T C D D dose and an observed 15 noncancer effect that w as relevant for R f D derivation. I f such data w ere available, then E P A 16 identified a N O A E L or L O A E L as a candidate P O D . F o r each o f these, E P A applied a human 17 kinetic model to estimate the continuous oral daily intake (ng/kg-day) associated w ith the P O D 18 that could be used in the derivation o f an R f D (see Section 4.2). I f all o f this inform ation w as 19 available, then the result w as included as a candidate P O D . 20 Figure 4-2 summarizes the strategy employed for identifying and selecting candidate 21 P O D s from the key anim al bioassays identified in Section 2.4.3 for use in noncancer 22 dose-response an alysis of T C D D . F o r each noncancer endpoint, E P A first evaluated the 23 toxicologic relevance o f each endpoint, rejecting those judged not to be relevant for R fD 24 derivation. Next, initial P O D s (N O A E L s, L O A E L s , and B M D L s ) based on the first-order body 25 burden metric (see Section 3.3.4.2) and expressed as human-equivalent doses (H E D s) were 26 determined for all relevant endpoints (summarized in Table 4-3). Because there were very few 27 N O A E L s and B M D L m odeling w as largely unsuccessful due to data lim itations, the next stage 28 o f evaluation w as carried out using L O A E L s only. W ithin each study, endpoints not observed at 29 the L O A E L (i.e., reported at higher doses) w ith B M D L s greater than the L O A E L w ere 30 eliminated from further analysis, as they w ould not be considered as candidates for the final P O D 31 on either a B M D L or N O A E L / L O A E L basis (i.e., the P O D w ould be higher than the P O D s o f This document is a draftfor review purposes only and does not constitute Agency policy. 4-2 DRAFT--DO NOT CITE OR QUOTE 1 other relevant endpoints). In addition, all endpoints with HED estimates based on LOAELs 2 (LOAELheds) beyond a 100-fold range of the lowest identified LOAELhed were eliminated 3 from further consideration, as they would not be potential POD candidates either (i.e., the POD 4 would be higher than the PODs of other relevant endpoints). For the remaining endpoints, EPA 5 then determined final potential PODs (NOAELs, LOAELs and BMDLs) based on TCDD blood 6 concentrations obtained from the Emond rodent physiologically based pharmacokinetic (PBPK) 7 models. HEDs were then estimated for each of these PODs using the Emond human PBPK 8 model. From these HEDs, a PODhed was selected19for each study as the basis for the candidate 9 RfD, to which appropriate uncertainty factors (UFs) were applied following EPA guidelines. 10 The resulting candidate RfDs were then considered in the final selection process for the RfD. 11 Other endpoints occurring at slightly higher doses representing additional effects associated with 12 TCDD exposure (beyond the 100-fold LOAEL range) were evaluated, modeled, and included in 13 the final candidate RfD array20to examine endpoints not evaluated by studies with lower PODs. 14 In addition, BMD modeling based on administered dose was performed on all endpoints for 15 comparison purposes. The final array of selected endpoints is shown in Table 4-4 (summary of 16 BMD analysis) and Table 4-5 (candidate RfDs). 17 The NAS recommended that EPA better justify the selection of response levels for 18 endpoints used to develop risk estimates. The NAS commented on EPA's decision to estimate 19 an ED01 (effective dose eliciting a 1% response) for noncancer bioassay/data set combinations as 20 a comparative tool across studies, suggesting that EPA identify and evaluate the levels of change 21 associated with adverse effects to define the benchmark response (BMR) level for continuous 22 noncancer endpoints. 23 24 The committee notes that the choice of the 1% response level as the POD 25 substantially affects ... the noncancer analyses.... The committee recommends 26 that the Reassessment use levels of change that represent clinical adverse effects 27 to define the BMR level for noncancer continuous end points as the basis for an 28 appropriate POD in the assessment of noncancer effects (NAS, 2006, 198441, 29 p. 72). 30 19I n t h e s t a n d a r d o r d e r o f c o n s i d e r a t i o n : B M D L , N O A E L , a n d L O A E L . 20H o w e v e r , s t u d i e s w i t h a l o w e s t d o s e t e s t e d g r e a t e r t h a n 3 0 n g / k g - d a y w e r e n o t i n c l u d e d i n t h e e x p a n d e d e v a lu a tio n . This document is a draftfor review purposes only and does not constitute Agency policy. 4-3 DRAFT--DO NOT CITE OR QUOTE 1 The committee concludes that EPA did not adequately justify the use of the 2 1% response level (the ED01) as the POD for analyzing epidemiological or animal 3 bioassay data for ... noncancer effects (NAS, 2006, 198441 p. 18). 4 5 In the 2003 Reassessment (U.S. EPA, 2003, 537122), EPA was not attempting to derive 6 an RfD when it conducted TCDD dose-response modeling. The 2003 Reassessment developed 7 ED01 estimates for noncancer effects in an attempt to compare disparate endpoints on a 8 consistent response scale. Importantly, the 2003 Reassessment defined the ED01 as 1% of the 9 maximal response for a given endpoint, not as a 1% change from control. Because RfD 10 derivation is one goal of this document, the noncancer modeling effort undertaken here differs 11 substantially from the modeling in the 2003 Reassessment. 12 The NAS committee was concerned with the statistical power to determine the shape of 13 the dose-response curve at doses far below observed dose-response information. EPA agrees 14 that the shape of the dose-response curve in the low-dose region cannot be determined 15 confidently when based on higher-dose information. An observed response above background 16 near (or below) the BMR level is needed for discrimination of the shape of the curve and for 17 accurate estimation of an EDx or BMDL. Although many of the ED01s presented in the 2003 18 Reassessment were near the lowest dose tested, responses at the lowest doses were often high 19 and much greater than a 1% response (i.e., 1% of the maximum response). The lack of an 20 observed response near the BMR level is often a problem in interpretation of BMD modeling 21 results. 22 In this document, EPA has used a 10% BMR for dichotomous data for all endpoints; 23 there were no developmental studies that accounted for litter effects, for which a 5% BMR would 24 be used (U.S. EPA, 2000, 052150). For continuous endpoints in this document, EPA has used a 25 BMR of 1 standard deviation from the control mean whenever a specific toxicologically-relevant 26 BMR could not be defined. For the vast majority of continuous endpoints, EPA could not 27 establish unambiguous levels of change representative of adversity, which EPA defines as "a 28 biochemical change, functional impairment, or pathologic lesion that affects the performance of 29 the whole organism, or reduces an organism's ability to respond to an additional environmental 30 challenge" (U.S. EPA, 2009, 192196). For body and organ weight change, EPA has previously 31 established a BMR of 10% change, which also is used in this document. This document is a draftfor review purposes only and does not constitute Agency policy. 4-4 DRAFT--DO NOT CITE OR QUOTE 1 The NAS commented on EPA's development of ED01estimates for numerous study/data 2 set combinations in the 2003 Reassessment, suggesting that EPA had not appropriately 3 characterized the statistical confidence around such model predictions in the low-response region 4 of the model. 5 6 It is critical that the model used for determining a POD fits the data well, 7 especially at the lower end of the observed responses. Whenever feasible, 8 mechanistic and statistical information should be used to estimate the shape of the 9 dose-response curve at lower doses. At a minimum, EPA should use rigorous 10 statistical methods to assess model fit and to control and reduce the uncertainty of 11 the POD caused by a poorly fitted model. The overall quality of the study design 12 is also a critical element in deciding which data sets to use for quantitative 13 modeling (NAS, 2006, 198441, p. 18). 14 15 EPA should ... assess goodness-of-fit of dose-response models for data sets and 16 provide both upper and lower bounds on central estimates for all statistical 17 estimates. When quantitation is not possible, EPA should clearly state it and 18 explain what would be required to achieve quantitation (NAS, 2006, 198441, 19 p. 10). 20 21 The NAS also commented that EPA report information describing the adequacy of 22 dose-response model fits, particularly in the low response region. For those cases where 23 biostatistical modeling was not possible, NAS recommended that EPA identify the reasons. 24 25 The Reassessment should also explicitly address the importance of statistical 26 assessment of model fit at the lower end and the difficulties in such assessments, 27 particularly when using summary data from the literature instead of the raw data, 28 although estimates of the impacts of different choices of models would provide 29 valuable information about the role of this uncertainty in driving the risk estimates 30 (NAS, 2006, 198441, p. 73). 31 32 To address this concern, in this document EPA has reported the standard suite of 33 goodness-of-fit measures from the benchmark dose modeling software (BMDS 2.1). These 34 include chi-square p-values, Akaike's Information Criterion (AIC), scaled residuals at each dose 35 level and plots of the fitted models. In some cases, when restricted parameters hit a bound, EPA 36 used likelihood ratio tests to evaluate whether the improvement in fit afforded by estimating 37 additional parameters could be justified. Goodness-of-fit measures are reported for all key data This document is a draftfor review purposes only and does not constitute Agency policy. 4-5 DRAFT--DO NOT CITE OR QUOTE 1 sets in Appendix E. (See Section 4.2.4.2 for a more complete description of the benchmark dose 2 modeling criteria for model evaluation.) 3 4 4.2. NONCANCER DOSE-RESPONSE ASSESSMENT OF TCDD 5 T h is section describes E P A ' s current effort to conduct an evaluation o f T C D D 6 dose-response for the noncancer endpoints from studies that met the study inclusion criteria. 7 D iscussions include benchmark dose modeling procedures, kinetic modeling, and P O D 8 candidates for derivation o f the R fD . Section 4.2.1 discusses the types o f endpoints that are 9 considered relevant by E P A 's Integrated R isk Information System and lists the study/endpoint 10 com binations that w ere not considered for the T C D D R f D derivation, w ith supporting text in 11 A ppendix G . Section 4.2.2 describes how E P A has used physiologically-based pharm acokinetic 12 (P B P K ) m odeling to estimate effective internal exposures as an alternative to using administered 13 doses or body burdens based on first-order kinetics. Section 4.2.3 details the dose-response 14 analysis o f the epidem iologic data, w ith supporting inform ation on kinetic m odeling in 15 A ppendix D . Section 4.2.4 details the dose-response an alysis for the anim al bioassay data; 16 A ppendix E provides the B M D S input tables (see Section E .1 ) and output for all m odeling, 17 including blood concentrations (see Section E .2 ) and administered dose (see Section E .3 ). 18 19 4.2.1. Determination of Toxicologically Relevant Endpoints 20 Th e N A S committee commented on the lo w dose model predictions and the need to 21 discuss the biological significance o f the noncancer health effects modeled in the 2003 22 Reassessment. In selecting P O D candidates from the animal bioassays for derivation o f the 23 candidate R fD s , E P A had to consider the toxicological relevance o f the identified endpoint(s) 24 from any given study. Some endpoints/effects may be sensitive, but lack general toxicological 25 significance due to not being clearly adverse (defined in E P A ' s Integrated R is k Information 26 System glossary as " a biochem ical change, functional impairment, or pathologic lesion that 27 affects the performance o f the w hole organism, or reduces an organism ' s ability to respond to an 28 additional environmental challenge" (U .S. E P A , 2009, 192196)), being an adaptive response or 29 not being clearly linked to downstream functional or pathological alterations. F o r example, C Y P 30 induction alone is not considered a significant toxicological effect given that C Y P s are induced 31 as part o f the hepatic m etabolism o f xenobiotic agents. A dditionally, the role o f C Y P induction This document is a draftfor review purposes only and does not constitute Agency policy. 4-6 DRAFT--DO NOT CITE OR QUOTE 1 in hepatotoxicity and carcinogenicity o f T C D D is unknown, thus, C Y P induction is not 2 considered a relevant P O D without obvious pathological significance. Another example is when 3 all oxidative stress markers are significantly affected, but no other indicators o f brain pathology 4 are assessed. In this case, it is im practicable to lin k the markers o f oxidative stress to a 5 toxicological outcome in the brain; thus, this endpoint is not considered a relevant P O D 6 candidate. It is standard E P A practice for R fD derivation to base a reference value on endpoints 7 that are adverse or are im m ediate precursors to an adverse effect. 8 Studies meeting the study selection criteria w ith endpoints that were not considered for 9 derivation o f a candidate R fD (because they were not considered to be toxicologically relevant 10 noncancer effects) are: K itch in and W oods (1979, 198750) , H assoun et al. (1998, 136626; 2000, 11 197431; 2002, 543725; 2003, 198726) , B u rleso n et al. (1996, 196998) , K u c h iiw a et al. (2002, 12 198355) , M a lly and Chipm an (2002, 198098), V anden H eu vel et al. (1994, 197551), D evito 13 et al. (1994, 197278) , L u c ie r et al. (1986, 198398) , Sugita-K onishi et al. (2003, 198375) , and 14 Sew all et al. (1993, 197889) . A ppendix G identifies the endpoints from these studies that w ere 15 not considered to be to xico lo g ically relevant for derivation o f an R f D (e.g., cytochrom e P450 16 induction, oxidative stress measures, gap junction disruption, m R N A induction, brain serotonin 17 le vels) and provides the rationales for the toxicological relevance decisions on the endpoints. 18 Note that for m any of these studies, other endpoints w ere exam ined that are toxicolog ically 19 relevant and w ere considered in the R f D derivation process. 20 21 4.2.2. Use of Toxicokinetic Modeling for TCDD Dose-Response Assessment 22 G iven that T C D D accumulates in fat w ith continuous exposure and is eliminated slow ly 23 from the body, but at very different rates across species, E P A has determined that the standard 24 U F approach or allometric scaling o f body weight for interspecies extrapolation is not 25 appropriate. Therefore, E P A has decided to use toxicokinetic modeling to estimate an effective 26 internal dose for equivalence across species. The toxicokinetic models chosen by E P A are the 27 rodent and human P B P K m odels described by Em o n d et al. (2004, 197315; 2006, 197316) and 28 modified by E P A for this assessment as described in Section 3.3.4 (hereafter referred to as the 29 "Em ond [rodent or human] P B P K m odel"). Both the rodent and human models have a 30 gestational component, w hich allow for more relevant exposure comparisons between general 31 adult exposures and the numerous gestational exposure studies. Ideally, a relevant tissue This document is a draftfor review purposes only and does not constitute Agency policy. 4-7 D R A F T -- D O N O T C IT E O R Q U O T E 1 concentration for each effect would be estimated. However, no models exist for estimation of all 2 relevant tissue concentrations. As virtually all TCDD is found in the adipose fraction of tissues, 3 or bound to specific proteins, a preferred approach to developing a dose metric would be to 4 account for the fat fraction of each tissue and protein binding; however, EPA has decided that the 5 modeling of such estimates is too uncertain and EPA has not found sufficient data to implement 6 this approach. Therefore, EPA has decided to use the concentration of TCDD in blood as a 7 surrogate for tissue concentrations, assuming that tissue concentrations are proportional to blood 8 concentrations. Furthermore, because the RfD is necessarily expressed in terms of average daily 9 exposure, the blood concentrations are expressed as averages over the relevant period of 10 exposure for each endpoint. For the animal bioassay studies, the relevant period of exposure is 11 the duration of dosing, starting at the age of the animals at the beginning of the study. For 12 humans, the relevant period of exposure is generally lifetime, which is defined as 70 years by 13 convention. However, EPA varied the averaging time for the equivalent human blood 14 concentrations to correspond to the test-animal exposure duration in the following manner. 15 16 For correspondence with animal chronic exposures,21 the human-equivalent 17 TCDD blood concentration is assumed to be the 70-year average. 18 For correspondence with animal gestational exposures, the human-equivalent 19 TCDD blood concentration is assumed to be the average over 45 years for a 20 female, beginning at birth, plus 9 months of gestational exposure. The choice of 21 45 years to beginning of pregnancy is health protective of the population in that 22 the TCDD daily oral intake achieving the target blood concentration is smaller 23 than for shorter averaging times.22 24 For correspondence with any other animal exposure duration, the 25 human-equivalent TCDD blood concentration is assumed to be the average over 26 the equivalent human exposure duration calculated backward from the peak 27 exposure plateau at or near the end of the 70-year scenario. The average is 28 determined from the terminal end of the human exposure period because the daily 29 oral intake achieving the target blood concentration is smaller than for the same 30 exposure period beginning at birth and is health protective for effects occurring 31 after shorter-term exposure. The determination of equivalent exposure durations 32 across species is problematic and somewhat arbitrary, so EPA uses the average 33 peak blood concentration as the human equivalent for all less-than-chronic animal 21Assumed to be >75% of nominal lifetime, or about 550 days in rodents. 22See Section 3.3.4.2 for a discussion of this issue, including a comparison of the 45-year old pregnancy scenario to one beginning at age 25 in Table 3-15. This document is a draftfor review purposes only and does not constitute Agency policy. 4-8 DRAFT--DO NOT CITE OR QUOTE 1 exposures (other than gestational).23 For the first-order kinetics model, the 2 average peak exposure is close to the theoretical steady-state asymptote (see 3 Section 3.3.4.2). However, for the Emond human PBPK model used by EPA in 4 this assessment, the timing of the peak exposure is dose-dependent and tends to 5 decline after 60 years in some cases. Therefore, the 5-year average TCDD blood 6 concentration that includes the peak ("5-year peak") is used as the relevant 7 dose-metric for the PBPK model applications. 8 9 4.2.3. Noncancer Dose-Response Assessment of Epidemiological Data 10 The following four epidemiologic studies describing noncancer endpoints were identified 11 in Section 2.4.3 as studies to be evaluated for development of PODs for derivation of candidate 12 RfDs: Baccarelli et al. (2008, 197059), Mocarelli et al. (2008, 199595), Alaluusua et al. (2004, 13 197142) and Eskenazi et al. (2002, 197168). Each of these studies described effects observed in 14 the Seveso cohort (see detailed study summaries in Section 2.4.1 and Table 2-5). Each study 15 modeled individual-level human exposure measures and provided information from which EPA 16 could determine an exposure window over which kinetic models could be used to quantify 17 TCDD exposures for dose-response assessment. EPA used kinetic information to estimate 18 group-mean daily TCDD intake rates for the exposure groups presented in these studies (see 19 Appendix D for details). EPA focused on identifying NOAELs and LOAELs for these studies; 20 EPA did not conduct Benchmark Dose modeling because the covariates identified by the study 21 authors could not be incorporated by modeling the grouped response data. EPA's development 22 of PODs for these studies is described in this section and shown in Table 4-1. 23 24 4.2.3.1. B acca relli e t al. (2 0 0 8 ,197059) 25 For Baccarelli et al. (2008, 197059), EPA was able to define a LOAEL as the group mean 26 of 39 ppt TCDD in neonatal plasma for thyroid stimulating hormone (TSH) values above 27 5 pU/mL. (See Section 2.4.1.2.1.5.7 for study details.) Baccarelli et al. (2008, 197059) did not 28 estimate the equivalent oral intake associated with TCDD serum concentrations and gave only 29 neonatal serum TCDD concentrations for the groups above and below the 5 pU/mL standard. 30 The study authors, however, developed a regression model relating the level of TSH in 3-day-old 23By comparison to a half-lifetime equivalent (1 year in rodents, 35 years in humans), the ratio of body burden (1st-order kinetic model) to oral intake does not differ significantly from the average-peak scenario; all shorter-term scenarios differ even less (see Section 3.3.4.2). These relationships, with respect to the 5-year peak, hold for the PBPK model results, as well (see Section 3). This document is a draftfor review purposes only and does not constitute Agency policy. 4-9 DRAFT--DO NOT CITE OR QUOTE 1 neonates to TCDD concentrations in maternal plasma at birth (given as lipid-adjusted serum 2 concentrations, LASC). The authors extrapolated maternal plasma concentrations from previous 3 measurements using a simple first-order pharmacokinetic model. Because there is limited 4 information regarding the relationship between maternal and neonatal serum TCDD levels, EPA 5 determined that there was too much uncertainty in estimating maternal intake from neonatal 6 TCDD serum concentrations, directly. Therefore, EPA determined the maternal intake at the 7 LOAEL from the maternal serum-TCDD/TSH regression model by finding the maternal TCDD 8 LASC at which neonatal TSH exceeded 5 pU/mL. EPA then used the Emond PBPK model 9 under the human gestational scenario (see Section 4.2.2) to estimate the continuous daily oral 10 TCDD intake that would result in a TCDD LASC corresponding to a neonatal TSH of 5 pU/mL 11 at the end of gestation; EPA established the resulting maternal intake (0.024 ng/kg-day) as the 12 LOAEL, shown in Table 4-1 as a candidate POD for derivation of candidate RfDs (PBPK 13 modeling details are shown in Appendix D). 14 15 4.2.3.2. M o ca relli e t al. (2 0 0 8 ,199595) 16 Mocarelli et al. (2008, 199595) reported decreased sperm concentrations (20%) and 17 decreased motile sperm counts (11%) in men who were 1-9 years old in 1976 at the time of the 18 accident (initial TCDD exposure event) (see Section 2.4.1.2.1.5.8 for study details). Men who 19 were 10-17 years old in 1976 were not adversely affected. Serum (LASC) TCDD levels were 20 measured within one year of the initial exposure. Serum TCDD levels and corresponding 21 responses were reported by quartile, with a reference group of less-exposed individuals assigned 22 a TCDD LASC value of 15 ppt (which was the mean of individuals outside the contaminated 23 area). The lowest exposed group mean was 68 ppt (1st quartile). Because effects were detected 24 only among boys under the age of 10, EPA assumes there is a maximum 10-year critical 25 exposure window for elicitation of these effects. However, for the exposure profile, with a high 26 initial pulse followed by an extended period of elimination with only background exposure, the 27 estimation of an average exposure resulting in the effect is problematic. Therefore, EPA 28 implemented a procedure for the estimation of the continuous daily TCDD intake associated with 29 the LOAEL in the Mocarelli et al. (2008, 199595) study using the following 5-step process: 30 This document is a draftfor review purposes only and does not constitute Agency policy. 4-10 DRAFT--DO NOT CITE OR QUOTE 1 1. U sin g the Em o n d human P B P K model, the initial (peak) blood T C D D concentrations 2 associated with the accident were back-calculated based on the time that had elapsed 3 between the explosion and the serum collection. A s serum measurements were taken 4 w ithin 1 year after the event, a lag time o f 0.5 years w as assumed. 5 2. The oral exposure associated w ith the peak blood T C D D concentration (peak exposure) 6 w as calculated using the Em ond P B P K model. 7 3. Starting w ith the peak exposure and accounting for background T C D D intake, the 8 average daily blood T C D D concentration experienced by a representative individual in 9 the susceptible population (boys under 10 years old) w as estimated using the Em on d 10 P B P K model. A ssu m in g a uniform distribution o f subject ages at the time o f the event, 11 the average age o f the exposed m ale children w ould be 5 years. Consequently, a critical 12 exposure w in do w for the cohort w as estimated to be, on average, 5 years (i.e., a boy aged 13 5 years w ould rem ain in this exposure w in dow for 5 more years until he w as 10 years o f 14 age). 15 4. U sin g the Em o n d P B P K model, the average daily T C D D intake rate needed to attain the 16 5-year average blood T C D D concentration in a boy 10 years old w as calculated. 17 5. Th e L O A E L P O D w as calculated as the average o f the peak exposure (0.032 ng/kg-day) 18 and the 5-year average exposure (0.0080 ng/kg-day), resulting in L O A E L o f 19 0.020 ng/kg-day, shown in Tab le 4-1 as a candidate P O D for derivation o f a candidate 20 R fD . However, neither o f the extremes w as used because (1) the peak exposure does not 21 account for the continuing internal exposure from T C D D given its slow elim ination, and 22 (2) the 5-year average does not reflect the influence o f the much higher peak exposure, 23 w h ich m ay be a significant factor in T C D D toxicity (K im et al., 2003, 199146) . 24 25 The P B P K modeling details are shown in Appendix D. 26 27 4.2.3.3. A la lu u su a e t al. (2 0 0 4 ,197142) 28 F o r A laluusua et al. (2004, 197142), the approach for estimation o f daily oral T C D D 29 intake is virtu ally identical to the approach used for the M o carelli et al. (2008, 199595) data. 30 (See Section 2.4.1.2.1.5.5 for study details.) A lalu u su a et al. (2004, 197142) reported dental 31 effects in m ale and fem ale adults w ho w ere less than 5 years o f age at the tim e o f the initial 32 exposure (1976). F o r the 75 boys and girls w ho were less than 5 years old at the time o f the 33 accident, 25 (3 3 % ) w ere subsequently diagnosed w ith some form o f dental enamel defect. F o r 34 the 38 individuals w ho w ere older than 5, only 2 (5 .3 % ) suffered dental enamel defects at a later 35 date. A w in do w o f susceptibility o f approxim ately 5 years is established. Serum measurements 36 for this cohort were taken within a year o f the accident. Serum T C D D levels and corresponding 37 responses were reported by tertile, with a reference group o f less-exposed individuals assigned a This document is a draftfor review purposes only and does not constitute Agency policy. 4-11 DRAFT--DO NOT CITE OR QUOTE 1 TCDD LASC value of 15 ppt (ng/kg); the tertile group means were 130, 383, and 1,830 ppt. 2 The incidence of dental effects for the reference group was 26% (10/39). The incidence of 3 dental effects in the 1st, 2nd and 3rd tertile exposure groups was 10% (1/10), 45% (5/11) and 4 60% (9/15), respectively. EPA judged that the NOAEL and LOAEL were 130 and 383 ppt 5 TCDD in serum. Following the same procedure used for the Mocarelli et al. (2008, 199595) 6 study (see Section 4.2.3.2), EPA estimated the continuous daily human oral TCDD intake 7 associated with each of the tertiles for both peak and average exposure across the critical 8 exposure window, assuming that the average age of the susceptible cohort at the time of the 9 accident was 2.5 years. Separate estimates for boys and girls were developed based on both the 10 peak intake and average intake across the critical exposure window (PBPK modeling details are 11 shown in Appendix D). The estimated averaged daily oral intakes for the tertiles, averaged for 12 boys and girls, are 0.20, 1.7, and 30 ng/kg-day for the peak exposure and 0.033, 0.15 and 13 1.5 ng/kg-day for the critical exposure window average. A study NOAEL at the second tertile of 14 0.12 ng/kg-day was identified as a candidate POD for derivation of a candidate RfD in Table 4-1. 15 16 4.2.3.4. E sken azi e t al. (2 0 0 2 ,197168) 17 The approach used to estimate daily TCDD intake in Eskenazi et al. (2002, 197168) 18 combines the approaches EPA used for Baccarelli et al. (2008, 197059), Mocarelli et al. (2008, 19 199595) and Alaluusua et al. (2004, 197142). Eskenazi et al. (2002, 197168) reported menstrual 20 effects in female adults who were premenarcheal in 1976 at the time of the initial exposure (see 21 Section 2.4.1.2.1.4.1 for study details). In Rigon et al. (2009), the median age at menarche was 22 shown to be 12.4 in Italian females with intergenerational decreases in age at menarche. Thus, 23 EPA established a window of susceptibility of approximately 13 years for this analysis. The 24 average age of the premenarcheal girls at the time of the initial exposure in 1976 was 6.8 years, 25 establishing an average critical-window exposure duration of 6.2 years for this cohort. Serum 26 samples were collected within a year of the accident from this cohort. However, serum TCDD 27 levels and corresponding responses were not reported by percentile and no internal reference 28 group was identified. As for Baccarelli et al. (2008, 197059), Eskenazi et al. (2002, 197168) 29 developed a regression model relating menstrual cycle length to plasma TCDD concentrations 30 (LASC) measured in 1976. The model estimated that menstrual cycle length was increased 31 0.93 days for each 10-fold increase in TCDD LASC, with a 95% confidence interval of -0.01 to This document is a draftfor review purposes only and does not constitute Agency policy. 4-12 DRAFT--DO NOT CITE OR QUOTE 1 1.86 days. E P A judged a 1-day increase in menstrual cycle length to be adverse; a normal 2 menstrual cycle length is 28 days. E P A then determined the 1976 T C D D serum level 3 corresponding to a 29-day menstrual cycle length in the exposed cohort from the regression 4 model developed by E sk e n a zi et al. (2002, 197168) . U sin g this serum level, the peak initial 5 exposure and average exposure over the 6.2 year w indow were calculated using the Em ond 6 human P B P K model, in the same manner as for M o carelli et al. (2008, 199595) and A lalu u su a 7 et al. (2004, 197142) . Th e resulting peak T C D D intake is 3.2 ng/kg-day. Th e average exposure 8 experienced by this cohort over the critical exposure w indow is estimated to be 0.12 ng/kg-day. 9 The average o f these two estimates is 1.64 ng/kg-day, w hich is designated as a L O A E L and 10 shown in Table 4-1. B e cau se the L O A E L is alm ost 2 orders o f magnitude higher than the 11 L O A E L s for B a cca re lli et al. (2008, 197059) and M o carelli et al. (2008, 199595), it w as not 12 considered further as a candidate P O D for derivation o f the R fD ( P B P K m odeling details are 13 shown in A pp end ix D ). 14 15 4.2.4. Noncancer Dose-Response Assessment of Animal Bioassay Data 16 E P A follow ed the strategy illustrated in Fig u re 4-2 to evaluate the anim al bioassay data 17 for T C D D dose-response. F o r the adm inistered average daily doses (ng/kg-day) in each animal 18 bioassay, E P A identified N O A E L s and/or L O A E L s based on the original data presented by the 19 study author. Section 2.4.2 identifies these values in the study sum m aries and in Table 2-7. 20 These became candidate P O D s for consideration in the derivation o f an R fD for T C D D . The 21 candidate R f D values associated w ith these candidate P O D s are presented in Table 4-5. 22 Additional P O D s were identified using B M D modeling. A ll P O D s were converted to H E D s 23 using the Em o n d P B P K models. Th e rem ainder o f this Section describes the steps in this process 24 and concludes w ith the P O D candidates from the animal bioassay data that were considered for 25 derivation o f the R fD . 26 27 4.2.4.1. U se o f K in etic M o d elin g f o r A n im a l B ioassay D ata 28 B lo o d concentrations corresponding to the adm inistered doses in each m ouse or rat 29 bioassay qualifying as a final R fD P O D candidate were estimated using the appropriate Em ond 30 rodent P B P K model. In each case, the simulation w as performed using the exposure and 31 observation durations, body weights, and average daily doses from the original studies. F o r all This document is a draftfor review purposes only and does not constitute Agency policy. 4-13 D R A FT-- DO NOT C IT E OR Q U O TE 1 multiple exposure protocols, the time-weighted average blood T C D D concentrations over the 2 exposure period were used as the relevant dose metric. F o r single (gestational and 3 nongestational) exposures, the initial peak blood T C D D concentrations were considered to be the 4 most relevant exposure metric. Gestational exposures were modeled using the species-specific 5 gestational component o f the Em ond rodent P B P K model. Bioassays employing exposure 6 protocols spanning gestational and postpartum life stages were modeled by sequential 7 application o f the gestational and nongestational models. 8 The Em ond P B P K models do not contain a lactation component, so exposure during 9 lactation w as not modeled explicitly. O n ly one bioassay (Sh i et al., 2007, 198147) considered as 10 a P O D candidate for R f D derivation included exposure during lactation. In Shi et al. (2007, 11 198147) pregnant anim als w ere exposed w eekly to T C D D throughout gestation and lactation. 12 Exp o su re w as continued in the offspring follow ing w eaning for 10 months. F o r assessm ent o f 13 maternal effects, the Em o n d gestational model w as used, terminating at parturition. F o r 14 assessm ent o f long-term exposure in the offspring, the Em o n d nongestational model w as used, 15 ignoring prior gestational and lactational exposure, w ith the assumption that the total exposure 16 during these periods w as sm all relative to exposure in the follow ing 10 months. Th e assumption 17 is conservative in that effects observed in the offspring w ould be attributed entirely to adult 18 exposure, w h ich is som ewhat less than the actual total exposure. 19 Th e model code, input files and P B P K m odeling results for each bioassay are reported in 20 Appendix C. These predicted T C D D blood concentrations were used for benchmark dose 21 m odeling o f bioassay response data and determination o f N O A E L s and L O A E L s . B M D 22 modeling w as performed, as described in Section 3.5.2.2.1, by substituting the modeled blood 23 concentrations for the adm inistered doses and calculating the corresponding B M D L . F o r each o f 24 these L O A E L , N O A E L , or B M D L blood-concentration equivalents, corresponding H E D s were 25 calculated using the Em ond human P B P K model for the appropriate gestational or nongestational 26 scenario as described previously (see Section 4.2.2). 27 28 4.2.4.2. B en ch m a rk D o se M o d elin g o f th e A n im a l B ioassay D ata 29 Benchm ark dose modeling w as performed using B M D S 2.1, B u ild 06/16/09 to estimate 30 B M D s and B M D L s for each study/endpoint combination. The input data tables for these 31 noncancer studies are shown in A ppendix E , Section E .1 , including both adm inistered doses This document is a draftfor review purposes only and does not constitute Agency policy. 4-14 D R A FT-- DO NOT C IT E OR Q U O TE 1 (ng/kg-day) and blood concentrations (ng/kg) and either incidence data for the dichotomous 2 endpoints or mean and standard deviations for the continuous endpoints. (See Section 4.2.4.1 3 and Sections 3.3.4 and 3.3.5 for a description of the development of TCDD blood concentrations 4 using kinetic modeling.) 5 Evaluation of BMD modeling performance, goodness-of-fit, dose-response data, and 6 resulting BMD and BMDL estimates included statistical criteria as well as expert judgment of 7 their statistical and toxicological properties. For the continuous endpoints, all available models 8 were run separately using both the assumption of constant variance and the assumption of 9 modeled variance. Saturated (0 degrees of freedom) model fits were rejected from consideration. 10 Parameters in models with power or slope parameters were constrained to prevent supralinear 11 fits, which EPA considers not to be biologically plausible and which often have undesirable 12 statistical properties (i.e., the BMDL diverges towards zero). However, if the constrained 13 parameters were estimated at their lower bounds, the unrestricted model was fit to the data, 14 primarily for elucidation of the degree of supralinearity present in the data. Depending on the 15 latter and the magnitude of the BMDL relative to the BMD, unrestricted model fits were 16 occasionally deemed acceptable. Table 4-2 shows each model and any restrictions imposed. 17 For the quantal/dichotomous endpoints, all primary BMDS dichotomous models were 18 run. The alternative dichotomous models were fit to several data sets, but the results were very 19 sensitive to the assumed independent background response and the fits were not accepted. The 20 confidence level was set to 95% and all initial parameter values were set to their defaults in 21 BMDS. For the continuous endpoints, one standard deviation was chosen as the default for the 22 BMR when a specific toxicologically-relevant BMR could not be defined. For the dichotomous 23 endpoints, a BMR of 10% extra risk was used for all endpoints. 24 24 The model output tables in Appendix E show all of the models that were run, both 25 restricted and unrestricted, goodness-of-fit statistics, BMD and BMDL estimates, and whether 26 bounds were hit for constrained parameters. After all models were run, the one giving the best 27 fit was selected using the selection criteria in the current BMDS draft guidance (U.S. EPA, 2000, 28 052150) where possible. Acceptable model fits were those with chi-square goodness-of-fit 29 p-values greater than 0.1. For continuous endpoints, a preference was held for models with an 30 asymptote term (plateau for high-dose response) because continuous measures do not continue to 24There were no developmental studies that accounted for litter effects, for which a 5% BMR would be used. This document is a draftfor review purposes only and does not constitute Agency policy. 4-15 DRAFT--DO NOT CITE OR QUOTE 1 rise (or fall) with dose forever; this phenomenon is particularly evident for TCDD. Unbounded 2 models, such as the power model, must account for the plateauing effect entirely in the shape 3 parameter, generally resulting in an abnormally supralinear fit. Also, for the continuous 4 endpoints, the p-value for the homogenous variance test (Test 2) was used to determine whether 5 constant variance (p > 0.1) or modeled (nonconstant) variance (p < 0.1) should be used. As 6 BMDS offers only one variance model, model fits for nonconstant variance models were not 7 necessarily rejected if the variance model did not fit well (Test 3 p-value < 0.05). Within the 8 group of models with acceptable fits, the selected model was generally the one with the lowest 9 BMDL, unless the AIC was much higher (ca. +2) than another model. However, particularly for 10 continuous models, the fit of the model to the control mean and standard deviation and in the 11 lower response range was assessed. Models with higher BMDLs or AICs but much better fit to 12 the lower response data were often chosen over the nominally best-fitting model. 13 For many data sets, no models satisfied the acceptance criteria and no clear BMD/BMDL 14 selection could be made. In theses cases, model fits were examined on an individual basis to 15 determine the reasons for the poor fits. On occasion, high doses were dropped and the models 16 were refit. Also, if a poor fit to the control mean was evident, the model was refit to the data 17 after fixing the control mean by specifying the relevant parameter in BMDS. However, these 18 techniques rarely resulted in better fits. If the fit was still not acceptable, the NOAEL/LOAEL 19 approach was applied to the study/data set combination. Most of the problems with BMD 20 modeling were a consequence of lack of response data near the BMR; many of the TCDD data 21 sets failed to show a response near the BMR, whether it was a 10% dichotomous relative change 22 or a continuous 1 standard deviation change. Responses at the lowest doses were generally much 23 higher than the BMR, resulting in a lack of anchoring at the critical response levels of interest 24 causing numerical problems in the estimation of BMDLs. 25 26 4.2.4.3. P O D C an didates f r o m A n im a l B io a ssa ys B a se d on H E D a n d B M D M o d elin g R esu lts 27 Table 4-3 summarizes the PODs that EPA estimated for each key animal study included 28 for TCDD noncancer dose-response modeling. After estimating the blood TCDD concentration 29 associated with a particular toxicity measure (NOAEL, LOAEL, or BMDL) obtained from a 30 rodent bioassay, EPA estimated a corresponding HED using the Emond human PBPK model 31 (described in Section 3). Table 4-3 summarizes the NOAEL, LOAEL, or BMDL (ng/kg) based This document is a draftfor review purposes only and does not constitute Agency policy. 4-16 DRAFT--DO NOT CITE OR QUOTE 1 on the administered animal doses for each key bioassay/data set combination. Table 4-3 also 2 summarizes the continuous daily HED corresponding to these administered doses as 1st order 3 body burdens and as blood concentrations. The doses in Table 4-3 are defined as follows, all in 4 units of ng/kg-day: 5 6 Administered Dose NOAEL: Average daily dose defining the NOAEL for the test species 7 in the animal bioassay 8 Administered Dose LOAEL: Average daily dose defining the LOAEL for the test species 9 in the animal bioassay 10 Administered Dose BMDL: BMDL for the test species based on modeling of the 11 administered doses from the animal bioassay 12 First-Order Body Burden HED NOAEL: Average daily dose defining the NOAEL for 13 humans derived from the animal bioassay using the first-order kinetics body-burden 14 model 15 First-Order Body Burden HED LOAEL: Average daily dose defining the LOAEL for 16 humans derived from the animal bioassay using the first-order kinetics body-burden 17 model 18 First-Order Body Burden HED BMDL: Human-equivalent BMDL from BMD modeling 19 of the animal bioassay data using first-order body burdens 20 Blood Concentration HED NOAEL: Average daily dose defining the NOAEL for 21 humans derived from the animal bioassay using the Emond human PBPK model 22 Blood Concentration HED LOAEL: Average daily dose defining the LOAEL for humans 23 derived from the animal bioassay using the Emond human PBPK model 24 Blood Concentration HED BMDL: Human-equivalent BMDL from BMD modeling of 25 the animal bioassay data using the Emond human PBPK model 26 27 An evaluation of key BMD analyses is presented in Table 4-4. Tables showing the best 28 model fit for each study/endpoint combination and the associated BMD/BMDL are shown in 29 Appendix E. As described above in Section 4.3.4.2, the BMD modeling was largely 30 unsuccessful, primarily because of a lack of response data near the BMR, poor modeled 31 representation of control values, or nonmonotonic responses yielding poor fits. The comments 32 column in Table 4-4 lists reasons for poor results. 33 This document is a draftfor review purposes only and does not constitute Agency policy. 4-17 DRAFT--DO NOT CITE OR QUOTE 1 4.3. RfD DERIVATION 2 Table 4-5 lists all the studies and endpoints considered for derivation of the RfD. These 3 studies were chosen from the entire list of candidate study/data set combinations (see 4 Section 2.4.3) based on the toxicologic relevance of the endpoints and covering a range of the 5 most conservative RfD candidates that includes three of the four human studies.25 Figure 4-3 6 (exposure-response array) shows all of the endpoints listed in Table 4-5 graphically in terms of 7 PODs in human-equivalent intake units (ng/kg-day). The human study endpoints are shown at 8 the far left of the figure and the rodent endpoints are arranged by category to the right. (Note the 9 two studies in guinea pigs were estimated using first-order body burden kinetics which are not 10 directly comparable to the PODs based on the mouse, rat and human studies that were generated 11 from the Emond PBPK model. There are no published models for TCDD disposition in guinea 12 pigs and EPA did not develop one for this assessment.) Figure 4-4 demonstrates the same 13 endpoints, arrayed by RfD value, showing the POD, applicable UFs and candidate RfD. 14 Table 4-5 illustrates the study, species, strain and sex, study protocol, and toxicologic 15 endpoints observed at the lowest TCDD doses. The table also identifies the human-equivalent 16 BMDLs (when applicable), NOAELs and LOAELs, as well as the composite UF that applies to 17 the specific endpoint, and finally, the corresponding candidate RfD.26 The NOAELS, LOAELs, 18 and BMDLs are presented as HEDs, based on the assumption that blood concentration is the 19 toxicokinetically-equivalent TCDD dose metric across species and serves as a surrogate for 20 tissue concentration.27 For rats and mice, these estimates relied on the two Emond PBPK 21 models--one for the relevant rodent species and one for the human--as described previously 22 (see Section 3.3.4.3). The two guinea pig studies that are included in Table 4-5 are given in 23 HED units based on the first-order body burden model described in Section 3.3.4.2; there is 24 currently no TCDD PBPK model for the guinea pig. The values listed for guinea pigs are not 25 directly comparable to those for rats and mice but are probably biased low, as first-order body 26 burden HED estimates for rats and mice are generally 2- to 5-fold lower than the corresponding 27 PBPK model estimates. The LOAELs for the human studies also rely on the Emond PBPK 28 model, as described in Sections 4.2.2 and 4.2.3. 25The RfD derived from the study of Eskenazi et al. (2002, 197168) was outside the RfD range presented in Table 4-5. 26Extra significant digits are retained for comparison prior to rounding to one significant digit for the final RfD. 27The procedures for estimating HEDs based on TCDD blood concentration are described in the preceding section. This document is a draftfor review purposes only and does not constitute Agency policy. 4-18 DRAFT--DO NOT CITE OR QUOTE 1 A s is evident from the Table 4-5, very few N O A E L s and even fewer B M D L s have been 2 established for low-dose T C D D studies. B M D modeling w as unsuccessful for all o f the 3 endpoints without a N O A E L , prim arily because o f the lack o f dose-response data near the B M R 4 (see discussion in Section 4.2). Therefore, the R fD assessment rests largely on evaluation of 5 L O A E L s to determine the P O D . 6 The rows in Table 4-5 are arranged in order o f increasing candidate R fD magnitude. 7 Endp oints projected to occur at higher exposure levels are still considered for qualitative support 8 o f the effects shown in Table 4-5. 9 10 4.3.1. Toxicological Endpoints 11 A s can be seen in Tab le 4-5, a w ide array o f toxicological endpoints has been observed 12 follow ing T C D D exposure, ranging from subtle developmental effects to overt chronic liv er 13 toxicity. Developm ental effects in rodents include dental defects, delayed puberty in m ales, and 14 several neurobehavioral effects. Reproductive effects reported in rodents include altered 15 hormone levels in fem ales and decreased sperm production in m ales. Im m unotoxicity endpoints 16 such as decreased response to S R B C challenge in m ice and decreased delayed-type 17 hypersensitivity response in guinea pigs are also observed. Lo n g e r durations o f T C D D exposure 18 in rodents elicit results such as organ and body w eight changes, renal toxicity, and liv er and lung 19 lesions. A dverse effects in human studies are also observed, w h ich include m ale reproductive 20 effects, increased T S H in neonates, and dental defects in children. Analogous results have been 21 observed in anim al bioassays for each o f these human endpoints. 22 A ll but two o f the study/endpoint combinations from animal bioassays listed in Table 4-5 23 are on T C D D -in d u c e d toxicity observed in m ice and rats; the other two study/endpoint 24 combinations are effects in guinea pigs. Although the effects o f T C D D have been investigated in 25 several other species (i.e., hamsters, monkeys, and m ink), those studies were not included for 26 final P O D consideration because the effect levels were greater than those in Table 4-5, or 27 because the effects could not be attributed solely to T C D D exposure (i.e., confounding by 28 dioxin-like compounds [D LC s]). 29 Three human studies were also included for final P O D consideration in the derivation of 30 an R fD and are presented in Table 4-5 as candidate R fD s. A ll three human study/endpoint 31 com binations are from studies on the Seveso cohort. Th e developmental effects observed in This document is a draftfor review purposes only and does not constitute Agency policy. 4-19 D R A FT-- DO NOT C IT E OR Q U O TE 1 these studies were associated with T C D D exposures either in utero or in early childhood between 2 1 and 10 years o f age. B a c ca re lli et al. (2008, 197059) reported increased levels o f T S H in 3 newborns exposed to T C D D in utero, indicating a possible dysregulation o f thyroid hormone 4 m etabolism . M o carelli et al. (2008, 199595) reported decreased sperm concentrations and 5 decreased m otile sperm counts in men w ho w ere 1 -9 years old in 1976 at the time o f the Seveso 6 accident (initial T C D D exposure event). A lalu u su a et al. (2004, 197142) reported dental effects 7 in adults w ho w ere less than 9.5 years o f age at the time o f the initial exposure (1976). 8 9 4.3.2. Exposure Protocols of Candidate PODs 10 Th e studies in Tab le 4-5 represent a w ide variety o f exposure protocols, involving 11 different methods o f administration and exposure patterns across virtu ally all exposure durations 12 and life stages. Both dietary and gavage adm inistration have been used in rodent studies, w ith 13 gavage being the predominant method. G avage dosing protocols vary quite w id ely and include 14 single gestational exposures, m ultiple daily exposures (for up to 2 w eeks, intermittent schedules 15 that include 5 days/week, once w eekly, or once every 2 w eeks), and loading/maintenance dose 16 protocols, in w h ich a relatively high dose is initially adm inistered follow ed by low er w eekly 17 doses. Th e intermittent dosing schedules require dose-averaging over time periods as long as 18 2 w eeks, w h ich introduces uncertainty in the effective exposures. In other words, the high unit 19 dose m ay be more o f a factor in eliciting the effect than the average T C D D tissue levels over 20 time. Although the loading/maintenance dose protocols are designed to maintain a constant 21 internal exposure, these protocols are som ewhat inconsistent w ith the constant daily T C D D 22 dietary exposures associated w ith human ingestion patterns. 23 Th e epidem iologic studies conducted in the Seveso cohort represent exposures over 24 different life stages including gestation, childhood, and young adulthood. The Seveso exposure 25 profile is essentially a high initial pulse T C D D exposure followed by a 2 0 -3 0 year period of 26 elimination. Effects are realized, or measured, 10-20 years follow ing the initial exposure; the 27 critical exposure w indow for susceptibility varies with effect and is often unknown. Therefore, 28 the effective exposure profiles for the Seveso cohort studies vary considerably. F o r the 29 M o carelli et al. (2008, 199595) and A lalu u su a et al. (2004, 197142) studies w here early 30 childhood exposures proximate to the initial event are associated w ith the outcomes, there is 31 some uncertainty as to the magnitude o f the effective doses. Although the effects are associated This document is a draftfor review purposes only and does not constitute Agency policy. 4-20 D R A FT-- DO NOT C IT E OR Q U O TE 1 with TCDD exposure in the first 10 years of life, it is not clear to what extent the initial peak 2 exposure is primarily responsible for the effects. It is also not clear if averaging exposure over 3 the critical window is appropriate given the large difference between initial TCDD body burden 4 and body burden at the end of the critical exposure window. The LOAELs for both Mocarelli 5 et al. (2008, 199595) and Alaluusua et al. (2004, 197142) are calculated as the average of the 6 peak exposure and average exposure across the critical exposure window (see Section 4.2 for 7 details). 8 For the gestational exposure study (Baccarelli et al., 2008, 197059), the critical exposure 9 window is strictly defined and relatively short (9 months) and occurs long after the initial 10 exposure (15-20 years). In addition, the maternal serum TCDD concentrations were measured 11 10-15 years after the initial exposure and are proximate to the actual pregnancies; consequently, 12 there is less uncertainty in the kinetic extrapolation between time of measurement and time of 13 birth (i.e., the critical exposure window). The narrow critical exposure window at a much later 14 time than the initial exposure (where the TCDD elimination curve is flattening) is assumed to 15 lead to a relatively steady-state exposure over the critical time period with much less uncertainty 16 in the magnitude of the effective dose. With the exception of Eskenazi et al. (2002, 197168) (see 17 Section 4.2), the effective doses for other effects reported for the Seveso cohort (see 18 Section 2.4.1.1.1.4) have not been quantified and are not represented in Table 4-5 because no 19 critical exposure windows can be identified or individual exposure estimates were not reported. 20 21 4.3.3. Uncertainty Factors (UFs) 22 The UF column in Table 4-5 shows the composite (total) UF that would be applied to the 23 POD for each endpoint. For the animal bioassays, a UF of 3 for the toxicodynamic component 24 of the interspecies extrapolation factor (UFA) was applied to all PODs. For both animal and 25 human studies, when a NOAEL was used as the POD, a factor of 10 was applied for human 26 interindividual variability (UFH). For all of the animal bioassay endpoints lacking a NOAEL, a 27 UF of 10 for the LOAEL-to-NOAEL UF (UFL) was included. For the human LOAELs, a UFLof 28 3 was applied because sensitive populations were identified. A subchronic-to-chronic UF (UFS) 29 of 1 and a database factor (UFD) of 1 are applied to all endpoints. A rationale for each UF is 30 provided for the derivation of the RfD below. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 4-21 DRAFT--DO NOT CITE OR QUOTE 1 4.3.4. Choice of Human Studies for RfD Derivation 2 For selection of the POD, the human studies are given the highest consideration, as 3 quality human data are always preferred by the EPA to animal data of comparable quality. The 4 human studies included in Table 4-5 (Alaluusua et al., 2004, 197142; Baccarelli et al., 2008, 5 197059; Mocarelli et al., 2008, 199595) each evaluate a segment of the Seveso civilian 6 population (i.e., not an occupational cohort) exposed directly to TCDD released from an 7 industrial accident. (The identification of PODs from these studies is detailed in 8 Sections 4.3.4.1, 4.3.4.2, and 4.3.4.3.) Thus, exposures were primarily to TCDD, the chemical of 9 concern, with apparently minimal DLC exposures beyond those associated with background 10 intake,28 making these studies highly appropriate for use in RfD derivation for TCDD. In 11 addition, health effects associated with TCDD exposures were observed in humans, the species 12 of concern whose health protection is represented by the RfD, eliminating the uncertainty 13 associated with interspecies extrapolation. The cohort members who were evaluated included 14 infants (exposed in utero) and adults who were exposed when they were less than 10 years of 15 age. These studies considered together associate TCDD exposures with health effects in 16 potentially vulnerable population subgroups. Their inclusion among the RfDs derived also may 17 characterize noncancer health effects associated with TCDD exposures in potentially vulnerable 18 populations, thus accounting for some part of the intraspecies uncertainty in the RfD. Finally, 19 the two virtually identical RfDs from different endpoints in different studies provide an 20 additional level of confidence in the use of these data for derivation the RfD for TCDD. 21 Although the human data are preferred, Table 4-5 presents a number of animal studies 22 with RfDs that are lower than the human RfDs. To a large extent, this is expected because a 23 10-fold interspecies uncertainty factor is generally used to extrapolate from test-animal species to 24 humans, intended to provide a conservative estimate of an RfD that would be derived directly 25 from human data. Two of the rat bioassays among this group of studies--Bell et al. (2007, 26 197041; RfD = 1.4E-9 mg/kg day based on delay in the onset of puberty) and NTP (2006, 27 197605; RfD = 4.6E-10 mg/kg day based on liver and lung lesions)--are of particular note. 28 Both studies were recently conducted. Both were very well designed and conducted, using 30 or 28As an example, note the lack of statistically significant effects reported by Baccarelli et al. (2008, 197059; Figure 2 C and D) in regression models based on either maternal plasma levels of noncoplaner PCBs or total TEQ on neonatal TSH levels. This document is a draftfor review purposes only and does not constitute Agency policy. 4-22 DRAFT--DO NOT CITE OR QUOTE 1 more animals per dose group (see Table 4-6 for a discussion of these studies' strengths and 2 weaknesses); both also are consistent with and, in part, have helped to define the current state of 3 practice in the field. Bell et al. (2007, 197041) evaluated several reproductive and 4 developmental endpoints, initiating TCDD exposures well before mating and continuing through 5 gestation. NTP (2006, 197605) is the most comprehensive evaluation of TCDD chronic toxicity 6 in rodents to date, evaluating dozens of endpoints at several time points in all major tissues. 7 Thus, proximity of the RfDs derived from these two high quality, recent studies provide 8 additional support for the use of the human data for RfD derivation. 9 There are several animal bioassay candidate RfDs at the lower end of the RfD range in 10 Table 4-5 that are more than 10-fold below the human-based RfDs. Two of these studies report 11 effects that are analogous to the endpoints reported in the three human studies and support the 12 RfDs based on human data. Specifically, decreased sperm production in Latchoumydandane and 13 Mathur (2002, 197498) is consistent with the decreased sperm counts and other sperm effects in 14 Baccarelli et al. (2008, 197059), and missing molars in Keller et al. (2007, 198526; 2008, 15 198531; 2008, 198033) are similar to the dental defects seen in Alaluusua et al. (2004, 197142). 16 Thus, because these endpoints have been associated with TCDD exposures in humans, these 17 animal studies would not be selected for RfD derivation in preference to human data showing the 18 same effects. 19 Another characteristic of the remaining studies in the lower end of the candidate RfD 20 distribution is that they are dominated by mouse studies (comprising 6 of the 8 lowest 21 rodent-based RfDs). EPA considers the candidate RfD estimates based on mouse data to be 22 much more uncertain than either the rat or human candidate RfD estimates. The EPA considers 23 the Emond mouse PBPK model to be the most uncertain of toxicokinetic models used to estimate 24 the PODs because of the lack of key mouse-specific data, particularly for the gestational 25 component (see Section 3.3.4.3.2.5). The LOAELHEDs identified in mouse bioassays are low 26 primarily because of the large toxicokinetic interspecies extrapolation factors used for mice, for 27 which there is more potential for error. The ratio of administered dose to HED (Da:HED) ranges 28 from 65 to 1,227 depending on the duration of exposure. The Da:HED for mice is, on average, 29 about four times larger than that used for rats. In addition, each one of the mouse studies has 30 other qualitative limitations and uncertainties (discussed above and in Table 4-6) that make them 31 less desirable candidates as the basis for the RfD than the human studies. This document is a draftfor review purposes only and does not constitute Agency policy. 4-23 DRAFT--DO NOT CITE OR QUOTE 1 4.3.4.I. Iden tification o f P O D f r o m B a cca relli e t al. (2 0 0 8 ,197059) 2 Baccarelli et al. (2008, 197059) reported increased levels of TSH in newborns exposed to 3 TCDD in utero, indicating a possible dysregulation of thyroid hormone metabolism. The study 4 authors related TCDD concentrations in neonatal blood to TSH levels, reporting group mean 5 TCDD concentrations associated with TSH levels above or below 5 p-Units TSH per mL of 6 serum (5 pU/mL). 7 The World Health Organization (WHO, 1994) established the 5 pU/mL standard as an 8 indicator of potential iodine deficiency and potential thyroid problems in neonates. Increased 9 TSH levels are indicative of decreased thyroid hormone (T4 and/or T3) levels. The 5 pU/mL 10 "cutoff' for TSH measurements in neonates was recommended by WHO (1994) for use in 11 population surveillance programs as an indicator of iodine deficiency disease (IDD). In 12 explaining this recommendation, WHO (1994) stated that: 13 14 "While further study of iodine replete populations is needed, a cutoff of 5pU/ml whole 15 blood... may be appropriate for epidemiological studies of IDD [iodine deficiency 16 disease.] Populations with a substantial number of newborns with TSH levels above the 17 cutoff could indicate a significant IDD problem." 18 19 For TCDD, the toxicological concern is not likely to be iodine uptake inhibition, but 20 rather increased metabolism and clearance of T4, as evidenced in a number of animal studies 21 (e.g., Seo et al., 1995, 197869). Clinically, a TSH level of >4 pU/mL in a pregnant woman is 22 followed up by an assessment of free T4, and treatment with L-thyroxine is prescribed if 23 T4 levels are low (Glinoer and Delange, 2000). This is to ensure a sufficient supply of T4 for the 24 fetus, which relies on maternal T4 exclusively during the 1st half of pregnancy (Chan et al., 2005; 25 (Calvo et al., 2002, 051690; Morreale et al., 2000, 019231). 26 Adequate levels of thyroid hormone also are essential in the newborn and young infant as 27 this is a period of active brain development (Glinoer and Delange, 2000; Zoeller and Rovet, 28 2004). Smaller reserves, higher demand, and shorter half-life of thyroid hormones in newborns 29 and young infants also could make this population more susceptible to the impact of insufficient 30 levels of T4 (Savin et al., 2003(Greer et al., 2002, 051202; Van Den et al., 1999, 016478). 31 Thyroid hormone disruption during pregnancy and in the neonatal period can lead to 32 neurological deficiencies. However, the exact relationship between TSH increases and adverse This document is a draftfor review purposes only and does not constitute Agency policy. 4-24 DRAFT--DO NOT CITE OR QUOTE 1 neurodevelopmental outcome is not w ell defined. A T S H level above 20 p U /L in a newborn 2 infant is cause for immediate intervention to prevent mental retardation, often caused by a 3 malformed or ectopic thyroid gland in the newborn (G linoer and Delange, 2000; Rovet, 2002; 4 W H O , 2007). Recent epidemiological data indicate concern for even low er level thyroid 5 hormone perturbations during pregnancy. F o r exam ple, H addow et al. (1999, 002176) reported 6 that women w ith subclinical hypothyroidism, w ith a mean T S H o f 13.2 p U /L had children with 7 IQ deficits o f up to 4 IQ points on the W ech sle r IQ scale. Neonatal T S H w ithin the first 8 72 hours o f birth (as w as evaluated by B a c ca re lli et al., 2008, 197059)is a sensitive indicator o f 9 both neonatal and maternal thyroid status (D e Lan g e et al., 1983). A n im a l m odels have recently 10 indicated that very modest perturbations in thyroid status for even a relatively short period o f 11 time can lead to altered brain development (e.g., A u so et al., 2004; La va d o -A u tric et al., 2003; 12 Sharlin et al., 2008, 2010; R o ylan d et al., 2008). 13 B a c ca re lli et al. (2008, 197059) discount iodine status in the population as a confounder, 14 as exposed and referent populations all lived in a relatively sm all geographical area. It is 15 u n lik ely that there w as iodine deficiency in one population and not in the other population based 16 on iodine levels in the soil. 17 B a c ca re lli et al. (2008, 197059) also showed, in graphical form, how the T S H distribution 18 in each o f three categorical exposure groups (reference, zone A , and zone B -- representing 19 increasing T C D D exposure) shifted to higher T S H values w ith increasing exposure. The 20 individuals com prising the above 5 p U /m L group were from all three categorical exposure 21 groups, not ju st from the highest exposure group. Therefore, E P A w as able to designate a 22 L O A E L independently o f the nominal categorical exposure groups; the L O A E L is designated as 23 the group mean o f 39 ppt T C D D in neonatal plasm a as a L O A E L for T S H values above 24 5 p U /m L. U sin g the Em o n d human P B P K m odel, the daily oral intake at the L O A E L is 25 estimated to be 0.024 ng/kg-day (see Section 4.2.3.1). A N O A E L is not defined because it is not 26 clear what maternal intake should be assigned to the group below 5 pU/m L. 27 28 4.3.4.2. Iden tification o f P O D f r o m M o ca relli e t al. (2 0 0 8 ,199595) 29 M o carelli et al. (2008, 199595) reported decreased sperm concentrations (2 0 % ) and 30 decreased m otile sperm counts (1 1 % ) in men w ho w ere 1 -9 years old in 1976 at the tim e o f the 31 Seveso accident (initial T C D D exposure event). Th e sperm concentrations and m otile sperm This document is a draftfor review purposes only and does not constitute Agency policy. 4-25 D R A FT-- DO NOT C IT E OR Q U O TE 1 counts in men who were 10-17 years old in 1976 were not affected. Serum (LASC) TCDD 2 levels were measured within one year of the initial exposure. Serum TCDD levels and 3 corresponding responses were reported by quartile, with a reference group of less-exposed 4 individuals assigned a TCDD LASC value of 15 ppt (which was the mean of the TCDD LASC 5 reported in individuals outside the contaminated area). The lowest exposed group mean was 6 68 ppt (1st-quartile). Mean sperm concentrations and motile sperm counts were reduced about 7 20% from the reference group. Further decrease in these values in the groups exposed to more 8 than 68 ppt was slight and reached a maximum of about 33%. 9 Although a decrease in sperm concentration of 20% likely would not have clinical 10 significance for an individual EPA's concern with the reported decreases in sperm concentration 11 and total number of motile sperm (relative to the comparison group) is that such decreases 12 associated with TCDD exposures could lead to shifts in the distributions of these measures in the 13 general population. Such shifts could result in decreased fertility in men at the low end of these 14 population distributions. While there is no clear cut-off indicating male fertility problems for 15 either of these measured effects. A sperm concentration of 20 million/ml is typically used as a 16 cut-off by clinicians to indicate follow-up for potential reproductive impact in affected 17 individuals. Low sperm counts are typically accompanied by poor sperm quality (morphology 18 and motility). For fertile men, between 50% and 60% of sperm are motile (Swan et al., 2003; 19 Slama et al., 2002; Wijchman et al., 2001). Any impacts on these reported levels could become 20 functionally significant. 21 For the 22-31 year-old men exposed to TCDD as a consequence of the Seveso accident, 22 the mean total sperm concentration was reported by Mocarelli et al. (2008, 199595) to be 23 53.6 million/ml, with a value of 21.8 million/ml at one standard deviation below the mean. In 24 the comparison group that consisted of men not exposed to TCDD by the Seveso explosion and 25 of the same age as the exposed men, the mean total sperm concentration was 72.5 million/ml 26 (31.7 million/ml at one standard deviation below the mean). In the group exposed due to the 27 Seveso accident, individuals one standard deviation below the mean are just above the cut-off 28 used by clinicians, indicating a that a number of individuals in the exposed group likely had 29 sperm concentrations less than 20 million/ml; EPA could not obtain the individual data to 30 determine the exact number of men in this category. EPA judged that the impact on sperm This document is a draftfor review purposes only and does not constitute Agency policy. 4-26 DRAFT--DO NOT CITE OR QUOTE 1 concentration and quality reported by M o carelli et al. (2008, 199595) is b io logically significant 2 given the potential for functional impairment. 3 E P A has designated the low est exposure group (68 ppt) as a L O A E L , w h ich translates to 4 a continuous daily oral intake of 0.020 ng/kg-day (see Section 4.2.3.2). The reference group is 5 not designated as a N O A E L because there is no clear zero-exposure measurement for any of 6 these endpoints, particularly considering the contribution o f background exposure to D L C s , 7 w hich further complicates the interpretation o f the reference group response as a true " control" 8 response (see discussion in Section 4.4). H ow ever, m ales less than 10 years old can be 9 designated as a sensitive population by comparison to older m ales w ho were not affected. 10 11 4.3.4.3. Iden tification o f P O D f r o m A la lu u su a e t al. (2 0 0 4 ,197142) 12 A lalu u su a et al. (2004, 197142) reported dental effects in m ale and fem ale adults w ho 13 w ere less than 9.5 years o f age, but not older, at the tim e o f the initial exposure (1976) in Seveso. 14 E P A used the same approach to estimate daily T C D D intake as w as used for the M o carelli et al. 15 (2008, 199595) data; a w in do w o f susceptibility o f about 5 years w as established. Serum 16 m easurements for this cohort w ere taken w ithin a year o f the accident. Serum T C D D levels and 17 corresponding responses w ere reported by tertile, w ith a reference group o f less-exposed 18 individuals assigned a T C D D L A S C value o f 15 ppt (ng/kg); the tertile group m eans w ere 130, 19 383, and 1,830 ppt. Both a N O A E L and L O A E L can be defined for this study. Th e N O A E L is 20 0.12 ng/kg-day, corresponding to the T C D D L A S C o f 130 ppt at the first tertile. Th e L O A E L is 21 0.93 ng/kg-day at the second tertile. The children in this cohort less than 5 years old can be 22 designated as a sensitive population by comparison to older individuals w ho were not affected 23 relative to the reference group. 24 25 4.3.5. Derivation of the RfD 26 Th e two human studies, B a c ca re lli et al. (2008, 197059) and M o carelli et al. (2008, 27 199595), have sim ilar L O A E L s of 0.024 and 0.020 ng/kg-day, respectively. Together, these 28 two studies constitute the best foundation for establishing a P O D for the R fD , and are designated 29 as coprincipal studies. Therefore, increased T S H in neonates in B a c ca re lli et al. (2008, 197059) 30 and m ale reproductive effects (decreased sperm count and m otility) in M o carelli et al. (2008, 31 199595) are designated as cocritical effects. A lthough the exposure estimate used in This document is a draftfor review purposes only and does not constitute Agency policy. 4-27 DRAFT--DO NOT CITE OR QUOTE 1 determination of the LOAEL for Mocarelli et al. (2008, 199595) is more uncertain than the 2 Baccarelli et al. (2008, 197059) exposure estimate, the slightly lower LOAEL of 3 0.020 ng/kg-day from Mocarelli et al. is designated as the POD. A composite UF of 30 is 4 applied to account for lack of a NOAEL (UFL= 10) and human interindividual variability 5 (UFh = 3); the resulting RfD in standard units is 7 x 10-10 mg/kg-day. Table 4-7 presents the 6 details of the RfD derivation. 7 8 4.4. UNCERTAINTY IN THE RfD 9 Exposure assessment is a key limitation of the epidemiologic studies (of the Seveso 10 cohort) used to derive the RfD. The Seveso cohort exposure profile consists of an initial high 11 dose followed by a drop in body burden to background levels over a period of about 20 years, at 12 which time the effects were observed. This exposure scenario is a mismatch with the constant 13 daily intake scenario addressed by the RfD methodology. The determination of an effective 14 average daily dose from the Seveso exposure scenario requires an understanding of the critical 15 time-window of susceptibility and the influence of the peak exposure on the occurrence of the 16 observed effects, particularly when the peak exposure is high relative to the average exposure 17 over the critical exposure window. For one of the principal studies (Mocaelli et al., 2008, 18 199595), a maximum susceptibility exposure window can be identified based on the age of the 19 population at risk. However, the influence of the peak exposure on the effects observed 20 years 20 later is unknown and the biological significance of averaging the exposure over several years, 21 with internal exposure measures spanning a 4.5-fold range, is unknown. EPA, in this 22 assessment, has averaged intermittent exposures for rodent bioassays over weekly dosing 23 intervals, but the peak and average body burdens varied by less than 50%. EPA has not 24 developed guidance for larger-interval averaging. Furthermore, because there is an assumption 25 of a threshold level of exposure below which the effects are not expected to occur, averaging 26 over large intervals could include below-threshold exposures. The process used by EPA to 27 estimate the LOAEL exposure for the Mocarelli study is a compromise between the extremes; as 28 such, there is some uncertainty in the estimate, perhaps in the range of 3- to 10-fold in either 29 direction. This uncertainty also holds for the LOAEL determined for the dental effects reported 30 in Alaluusua et al. (2004, 197142) and the increased menstrual cycle length reported in Eskenazi 31 et al. (2002, 197168 see Section 4.2.3.4); in both of those studies, the uncertainty is greater, as This document is a draftfor review purposes only and does not constitute Agency policy. 4-28 DRAFT--DO NOT CITE OR QUOTE 1 the difference between peak and average internal exposures is an order o f magnitude or more. 2 The L O A E L for increased T S H in neonates (B accarelli et al., 2008, 197059), however, is less 3 uncertain because the critical exposure w indow is much narrower (9 months) and the 4 developmental exposures occurred 10 to 15 years after the initial exposure, w hen internal T C D D 5 concentrations for the pregnant women likely were leveling off; that is, exposure over the critical 6 w indow w as more constant and estimation o f the relevant exposures w as less uncertain. 7 However, there is some uncertainty in the magnitude o f the exposures because they were 8 estimated from m easurements in sera taken several years prior to pregnancy. 9 Another source o f uncertainty using human epidem iologic data is the lack o f completely 10 unexposed populations. Th e available T C D D epidem iologic data w ere obtained by com paring 11 populations that experienced elevated T C D D exposures to populations that experienced low er 12 exposures, rather than to a population w ith no T C D D exposure. A n additional com plicating 13 factor is coexposure to D L C s , w h ich can behave in the same w ay as T C D D . Although the 14 accidental exposure to the Seveso w om en' s cohort w as prim arily to T C D D , background 15 exposure w as largely to D L C s . 29 E sk e n a zi et al. (2004, 197160) reported that T C D D com prised 16 only 2 0 % o f the total toxicity equivalence ( T E Q ) in the serum o f the reference group that w as 17 not exposed as a result o f the factory explosion, w h ich im plies that the effective background 18 T E Q exposure w as approxim ately 5-fold higher. 19 Th e higher background exposure could be significant at the low er T C D D exposure levels, 20 w ith the effect dim inishing as T C D D exposure increased. F o r dose-response modeling, the 21 effect o f a higher background dose (i.e., total T E Q ) , i f included, w ould be to shift the response 22 curve to the right (responses associated w ith higher exposures) but, prim arily, w ould reduce the 23 spread o f the exposures, w h ich w ould tend to alter the shape o f the dose response towards 24 sublinear. Both the right shift and the more sublinear shape w ould result in higher E D x 25 estimates, such as B M D s and B M D L s , from fitting dose-response models. However, for 26 determination o f a L O A E L , w hich is the case for all the human studies in Table 4-5, the impact 27 may be m inim al, as the L O A E L depends only on establishing that an effect o f sufficient 29M o c c a r e l l i ( 2 0 0 1 , 1 9 7 0 0 2 ) r e p o r t e d t h e r e l e a s e f r o m t h e S e v e s o p l a n t t o c o n t a i n a m i x t u r e o f T C D D , e t h y l e n e g ly c o l a n d so d iu m h y d ro x id e . A s th e se c h e m ic a ls a re n o t th o u g h t to p e rsist in th e e n v iro n m e n t o r in th e b o d y , c o e x p o su re to th e se a d d itio n a l c o n ta m in a n ts a lo n g w ith T C D D w o u ld n o t h a v e a s ig n ific a n t im p a c t o n lo n g e r-te rm T C D D d o se -re sp o n se . F o r a c u te e x p o su re , m a le re p ro d u c tiv e o r th y ro id h o rm o n e e ffe c ts a re n o t e v id e n t fo r e th y le n e g ly c o l (U .S . E P A , 2 0 0 9 , 1 9 2 1 9 6 ) . It is u n lik e ly th a t s o d iu m h y d ro x id e , b e in g p rim a rily a c a u stic a g e n t, w o u ld c a u se th e se e ffe c ts. This document is a draftfor review purposes only and does not constitute Agency policy. 4-29 D R A FT-- DO NOT C IT E OR Q U O TE 1 magnitude w as observed at some T C D D exposure level. In this case, the effect o f the increased 2 effective background exposure w ould be to inflate the " control" (zero -T E Q ) response, providing 3 the threshold for the response had been exceeded. The potential impact o f an inflated control 4 response w ould be to m ask a significant effect o f the added T C D D exposure, w hen the latter 5 effect is determined by comparison to the reference group response. To compensate for this, 6 E P A has been somewhat conservative in interpreting the magnitude o f responses defining 7 L O A E L s for the Seveso cohort studies. The actual magnitude o f the impact o f the D L C 8 background exposure is im possible to assess without knowing the true (T EQ -free) background 9 response. 10 A prim ary strength o f the T C D D database is that analogous effects have been observed in 11 anim al bioassays for m ost o f the human endpoints, increasing the overall confidence in the 12 relevance to hum ans o f the effects reported in rodents and the association o f T C D D exposure 13 w ith the effects reported in humans. Tab le 4-5 shows that lo w dose T C D D exposures are 14 associated w ith a w ide array o f toxicological endpoints in rodents including developmental 15 effects, reproductive effects, im m unotoxicity and chronic toxicity. E ffe cts reported in human 16 studies are sim ilar, including m ale reproductive effects, increased T S H in neonates and dental 17 defects in children; other human health effects such as fem ale reproductive effects and chloracne 18 have been observed at higher exposures (see Section 2.4.1). Other effects reported in rodent 19 studies such as liv e r toxicity and overt im m unological endpoints have not been reported in 20 human studies. H ow ever, w ith respect to im m unological effects, B a c ca re lli et al. (2002, 197062; 21 2004, 197045) evaluated im m unoglobin and com plem ent levels in the sera o f T C D D -e x p o se d 22 individuals from the Seveso cohort and found slightly reduced immunoglobulin in the highest 23 exposure groups but no effect on other im m unoglobulins or on C3 or C 4 com plem ent levels. 24 The latter finding indicates that at least one im m unological measure in humans is not a sensitive 25 endpoint, as it is for m ice, w ith large reductions in serum com plem ent at lo w exposure levels 26 (W hite et al., 1986, 197531) . 27 Although there is a substantial amount o f qualitative concordance o f effects between 28 rodents and humans, quantitative concordance is not evident in Table 4-5. The differential 29 sensitivity o f m ice and humans for the serum complement endpoint is one example. Other 30 examples o f differential sensitivity are developmental dental effects and thyroid hormonal 31 dysregulation. Developm ental dental defects are relatively sensitive effects in rodents, appearing This document is a draftfor review purposes only and does not constitute Agency policy. 4-30 D R A FT-- DO NOT C IT E OR Q U O TE 1 at exposure levels in mice (Keller et al., 2007, 198526; Keller et al., 2008, 198531; Keller et al., 2 2008, 198033) more than an order of magnitude lower than effect levels in humans (Alaluusua et 3 al., 2004, 197142). In contrast, thyroid hormone effects are seen in rats (Crofton et al., 2005, 4 197381) at 30-fold higher exposures than for humans (Baccarelli et al., 2008, 197059). Male 5 reproductive effects (sperm production) occur in rats (Latchoumycandane and Mathur, 2002, 6 197498) and humans (Mocaelli et al., 2008, 199595) at about the same dose. To what extent 7 these differential sensitivities depend on specifics of the comparison, such as species (mouse vs. 8 rat), life-stage (e.g., fetal vs. adult), endpoint measure (e.g., thyroxine [T4] vs. TSH) or 9 magnitude of the lowest dose tested, cannot be determined, so strong conclusions about 10 quantitative concordance cannot be made. 11 A number of qualitative strengths and limitations/uncertainties are associated with the top 12 animal bioassays listed in Table 4-5, as articulated in Table 4-6. Considering the issue of lowest 13 tested dose, the general lack of NOAELs and acceptable BMDLs is a primary weakness of the 14 rodent bioassay database. None of the 6 most sensitive rodent studies in Table 4-5, spanning a 15 30-fold range of LOAELs, had defined NOAELs or BMDLs. NOAELs or BMDLs were 16 established for only 4 of the next 10 rodent studies. In addition, many of these LOAELs are 17 characterized by relatively high responses with respect to the control population, so it is not 18 certain that a 10-fold lower dose (based on the application of UFL of 10) would be approximately 19 equivalent to a NOAEL. A major reason for the failure of BMD modeling was that the responses 20 were not "anchored" at the low end (i.e., first response levels were far from the BMR [see 21 Table 4-4]). Another major problem with the animal bioassay data was nonmonotone and flat 22 response profiles. The small dose-group sizes and large dose intervals probably contributed to 23 many of these response characteristics that prevented successful BMD modeling. Larger study 24 sizes with narrower dose intervals at lower doses are still needed to clarify rodent response to 25 TCDD. 26 Lower TCDD doses have been tested in rodents but almost entirely for investigation of 27 specialized biochemical endpoints30 that EPA does not consider to be adverse health effects (see 28 Appendix G). There is, however, a fundamental limit to the lowest dose of TCDD that can be 29 tested meaningfully, as TCDD is present in feed stock and accumulates in unexposed animals 30 prior to the start of any study. This issue is illustrated by the presence of TCDD in tissues of 30E n z y m e i n d u c t i o n , o x i d a t i v e s t r e s s i n d i c a t o r s , m R N A l e v e l s , e t c . This document is a draftfor review purposes only and does not constitute Agency policy. 4-31 DRAFT--DO NOT CITE OR QUOTE 1 unexposed control anim als, often at significant levels relative to the low est tested dose in low 2 dose studies (B e ll et al., 2007, 197041; O hsako et al., 2001, 198497) (V anden H eu vel et al., 3 1994, 594318, see T ext B o x 4-1). Some D L C s also have been m easured in anim al feeds and are 4 anticipated to accumulate in unexposed test animals further com plicating the interpretation of 5 low dose studies. 6 Text Box 4-1. Background levels of TCDD in Control Group Animals T C D D tis s u e le v e ls in c o n tro l a n im a ls a re ra re ly re p o rte d e ith e r e x p lic itly o r im p lic itly . V a n d e n H e u v e l e t al. (1 9 9 4 , 1 9 7 5 5 1 ), h o w e v e r, re p o rte d T C D D c o n c e n tra tio n s in liv e rs o f c o n tro l a n im a ls (1 0 -w e e k -o ld fe m a le S p ra g u e -D a w le y ra ts) o f 0 .4 3 p p t (n g /k g ) c o m p a re d to 0 .4 9 p p t in th e liv e rs o f a n im a ls g iv e n a s in g le o ra l T C D D d o s e o f 0 .1 n g /k g . A s s u m in g p ro p o rtio n a lity o f liv e r c o n c e n tra tio n to to ta l b o d y b u rd e n , th e b o d y b u r d e n o f u n tre a te d a n im a ls w a s 8 7 .8 % o f th a t o f tr e a te d a n im a ls . T h e e q u iv a le n t a d m in is te re d d o s e f o r u n tre a te d a n im a ls (do) c a n b e c a l c u l a t e d a s e q u a l t o 0 .8 7 8 x ( 0 .1 + d 0) , a s s u m i n g p r o p o r t i o n a l i t y o f b o d y b u r d e n t o a d m i n i s t e r e d d o s e a n d t h a t a l l a n im a ls s ta rte d w ith th e s a m e T C D D b o d y b u rd e n s . T h e c a lc u la tio n y ie ld s a v a lu e o f 0 .7 2 n g /k g f o r d 0, w h ic h re p re se n ts th e a c c u m u la te d T C D D fro m a ll so u rc e s in th e se a n im a ls p rio r to b e in g p u t o n a n d d u rin g te st. T h is v a lu e w o u ld ra is e th e n o m in a l 0 .1 n g /k g T C D D d o s e 8 -fo ld to 0 .8 2 n g /k g . T h e n e x t h ig h e r d o s e o f 1 n g /k g w o u ld b e n e a rly d o u b le d to 1 .7 2 n g /k g . T h e im p a c t o n h ig h e r d o s e s w o u ld b e n e g lig ib le , b e c a u s e th e ra tio o f tre a tm e n t d o s e to a p p a re n t b a c k g ro u n d e x p o s u re le v e ls in c re a s e s w ith h ig h e r tre a tm e n t le v e ls. B e ll e t al. (2 0 0 7 , 1 9 7 0 4 1 ) re p o rte d slig h tly h ig h e r le v e ls (0 .6 6 p p t) in th e liv e rs o f slig h tly o ld e r u n tre a te d p re g n a n t fe m a le S p ra g u e -D a w le y ra ts (m a te d a t 1 6 -1 8 w e e k s o f a g e a n d te ste d 17 d a y s la te r). O h sa k o e t al. (2 0 0 1 , 1 9 8 4 9 7 ) re p o rte d T C D D c o n c e n tra tio n s in th e f a t o f o ffs p rin g o f u n tre a te d p re g n a n t H o ltz m a n ra ts th a t w e re 4 6 % o f th e T C D D f a t c o n c e n tra tio n s in a n im a ls e x p o s e d in u te ro to 1 2 .5 n g /k g (s in g le e x p o su re o n G D 15). T h is le v e l o f T C D D w o u ld im p ly a v e ry la rg e b a c k g ro u n d e x p o su re , b u t q u a n tita tio n b a s e d o n sim p le k in e tic a ssu m p tio n s p ro b a b ly w o u ld n o t re fle c t th e m o re c o m p lic a te d in d ire c t e x p o su re sc e n a rio B e ll e t al. (2 0 0 7 , 1 9 7 0 4 1 ) a lso re p o rte d c o n c e n tra tio n s o f 0 .1 a n d 0 .6 p p t T C D D m e a s u re d in tw o s a m p le s o f fe e d sto c k . A s s u m in g th a t th e a v e ra g e o f 0 .3 5 p p t is re p re s e n ta tiv e o f th e e n tire s u p p ly o f fe e d s to c k a n d a fo o d c o n s u m p tio n fa c to r o f 1 0 % o f b o d y w e ig h t p e r d a y , th e a v e ra g e d a ily o ra l e x p o su re fro m fe e d to th e se a n im a ls w o u ld b e 0 .0 3 5 n g /k g . D is c rim in a tio n o f o u tc o m e s fro m lo n g e r-te rm re p e a te d e x p o s u re s m ig h t b e p ro b le m a tic a t e x p o su re le v e ls a ro u n d 0 .1 n g /k g -d a y . B a c k g ro u n d e x p o s u re w a s n o t m u c h o f a n is s u e f o r B e ll e t al. (2 0 0 7 , 1 9 7 0 4 1 ) , a s th e lo w e s t T C D D e x p o s u re le v e l w a s 2 .4 n g /k g -d a y (2 8 -d a y d ie ta ry e x p o su re ). N T P (2 0 0 6 , 5 4 3 7 4 9 ) re p o rte d T C D D c o n c e n tra tio n s in th e liv e r a n d fa t o f u n tre a te d fe m a le S -D ra ts a fte r 2 y e a rs o n te st th a t w e re 1 % a n d 2 .5 % o f th e le v e ls in th e liv e r a n d fa t o f th e lo w -d o se T C D D tre a tm e n t g ro u p (2 .1 4 n g /k g -d a y ; (N T P , 2 0 0 6 , 1 9 7 6 0 5 )), re sp e c tiv e ly . A ss u m in g p ro p o rtio n a lity o f fa t c o n c e n tra tio n a n d o ra l in ta k e , c o n tro l a n im a l e x p o s u re w o u ld h a v e b e e n a p p ro x im a te ly 0 .0 5 n g /k g -d a y , s im ila r to th e e stim a te fro m B e ll e t al. (2 0 0 7 , 1 9 7 0 4 1 ) . A s fo r th e la tte r stu d y , b a c k g ro u n d in ta k e fo r th e N T P (2 0 0 6 , 1 9 7 6 0 5 ) stu d y a n im a ls w o u ld n o t h a v e a la rg e e ffe c t o n th e d o s e -re s p o n s e a s s e s sm e n t g iv e n th e lo w e s t e x p o s u re le v e l o f 2 .1 4 n g /k g -d a y . In a ll o f th e se stu d ie s, e x c e p t th e 2 8 -d a y e x p o s u re in B e ll e t al. (2 0 0 7 , 1 9 7 0 4 1 ) , c o n tro l a n im a ls w e re g a v a g e d w ith c o rn o il v e h ic le . T C D D c o n c e n tra tio n s in c o rn o il w e re n o t re p o rte d in a n y o f th e stu d ie s. 7 This document is a draftfor review purposes only and does not constitute Agency policy. 4-32 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-1. POD candidates for epidemiologic studies of TCDD 2 Study POD (ng/kg-day) Critical effects Alaluusua et al. (2004, 197142) Baccarelli et al. (2008, 197059) 1.2E-01a(NOAEL) Dental effects in adults exposed to TCDD in childhood 2.4E-02b(LOAEL) Elevated TSH in neonates Eskenazi et al. (2002, 1.64E+00c(LOAEL Increased length of menstrual cycle in women exposed 197168) to TCDD in childhood Mocarelli et al. (2008, 2.0E-02d(LOAEL) Decreased sperm count and motility in men exposed to 199595) TCDD in childhood 3 4 aM e a n o f p e a k e x p o s u r e ( 0 . 1 5 n g / k g - d a y ) a n d a v e r a g e e x p o s u r e o v e r 1 0 - y e a r c r i t i c a l w i n d o w ( 0 . 0 0 9 3 n g / k g - d a y ) . 5 bM a t e r n a l e x p o s u r e c o r r e s p o n d i n g t o n e o n a t a l T S H c o n c e n t r a t i o n e x c e e d i n g 5 p U / m L . 6 cM e a n o f p e a k e x p o s u r e ( 3 . 2 n g / k g - d a y ) a n d a v e r a g e e x p o s u r e o v e r 1 0 - y e a r c r i t i c a l w i n d o w ( 0 . 1 2 n g / k g - d a y ) . 7 dM e a n o f p e a k e x p o s u r e ( 0 . 0 3 5 n g / k g - d a y ) a n d a v e r a g e e x p o s u r e o v e r 1 0 - y e a r c r i t i c a l w i n d o w ( 0 . 0 0 7 8 n g / k g - d a y ) . 8 9 Table 4-2. Models run for each study/endpoint combination in the animal 10 bioassay benchmark dose modeling 11 Model Restrictions imposed Continuous models Exponential M2-M5, not grouped Hill Linear Adverse direction specified according to the response data; power > 1 Adverse direction is automatic; n > 1 Adverse direction is automatic; degree of polynomial = 1 Polynomial Power Adverse direction is automatic; degree of polynomial unrestricted; restrict the sign of the power to nonnegative or nonpositive, depending on the direction of the responses Adverse direction is automatic; power >1 Dichotomous models Gamma Power >1 Logistic None Log-Logistic Slope >1 Log-Probit Multistage Probit Weibull None Beta >0, 2nddegree polynomial None Power >1 12 This document is a draftfor review purposes only and does not constitute Agency policy. 4-33 DRAFT--DO NOT CITE OR QUOTE Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered dose, first-order body burden HED, and blood concentration This document is a draftfor review purposes only and does not constitute Agency policy. 4-34 Study Endpoint A m in et al. (2000, Saccharin preference ratio, 197169) fem ale B ell et al. (2007, 197041) Balano-preputial separation in m ale pups Cantoni et al. (1981, U rinary coproporhyrins 197092) C hu et al. (2001, 521829) Tissue w eight changes C h u et al., 2007 Liv er lesions Crofton et al. (2005, Serum T4 197381) Croutch et al. (2005, D ecreased body w eight 197382) D eCaprio et al. (1986, 197403) D ecreased body w eight Fattore et al. (2000, D ecreased hepatic retinol 197446) F o x et al. (1993, 197344) Increased liver w eight Franc et al. (2001, O rgan w eight changes 197353) Franczak et al. (2006, Abnorm al estrous cycle 197354) H ojo et al. (2002, 198785) D R L response per m in H utt et al. (2008, 198268) Em byrotoxicity Administered dosea NOAEL LOAEL BMDLd -- 2.50E+01 5.10E+01 - 2.40E +00 2.87E +00 -- 1 .4 3 E + 0 0 1 .2 5 E --01 2.5 0 E +0 2 1.00E+03 -- 2.50E +00 2.50E+01 -- 3.00E+01 1.00E+02 3.01E+01 5.43E+01 2.17E +02 -- 6 .1 0 E --01 4 .9 0 E + 0 0 -- -- 2.00E+01 -- 5 .7 0 E --01 3 .2 7 E + 0 2 -- 1.00E+01 3.00E+01 1.59E+00 -- 7.14E +00 -- -- 2 .0 0 E + 0 1 2 .7 0 E --01 -- 7.14E +00 -- 1st-order body burden HEDb NOAEL LOAEL BMDLd -- 2 .4 9 E --02 5 .0 8 E --02 -- 1 .2 6 E --02 1 .5 0 E --02 -- 1 .2 4 E --02 1 .0 9 E --03 7 .5 5 E --01 3 .0 2 E + 0 0 -- 7 .5 5 E --03 7 .5 5 E --02 -- 1 .9 2 E --02 6 .4 0 E --02 1 .9 2 E --02 2 .2 2 E --01 8 .8 9 E --01 -- 4 .1 1 E --03 3 .3 0 E --02 -- -- 1 .2 3 E --01 -- 1 .4 2 E --03 8 .1 2 E --01 -- 6 .6 2 E --02 1 .9 9 E --01 1 .0 5 E --02 -- 5 .9 5 E --02 -- -- 5 .2 6 E --03 7 .1 1 E --05 -- 4.67E -02 -- Blood concentration HEDc NOAEL LOAEL BMDLd -- 1 .7 1 E --01 3 .2 0 E --01 -- 8 .8 3 E --02 4 .3 3 E --02 -- 6 .5 1 E --02 1 .6 0 E --03 ------ 3 .5 6 E --02 5 .7 6 E --01 -- 1 .7 2 E --01 7 .6 1 E --01 1 .4 0 E --01 ------ ------ -- 8 .0 1 E --01 -- ------ 4 .6 0 E --01 1 .4 5 E + 0 0 3 .3 7 E --02 -- 3 .2 5 E --01 -- -- 5 .5 0 E --02 7 .3 7 E --05 -- 2.57E-01 -- D RA FT: DO NOT C IT E OR Q U O TE Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered dose, 1st-order body burden HED and blood concentration HED (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-35 Study Ikeda et al. (2005, 197834) Ishihara et al. (2007, 197677) K attainen et al. (2001, 198952) K eller et al. (2007, 198526; 2008, 198531; 2008, 198033) K ocib a et al. (1976, 198594) K ocib a et al. (1978, 001818) Latchoum ycandane and M athur (2002, 197498) L i et al. (1997, 199060) L i et al. (2006, 199059) M arkow ski et al. (2001, 197442) M aronpot et al. (1993, 198386) Endpoint S e x ratio Se x ratio 3rd m o la r le n g th M issin g m andibular m olars Liver and hem atologic effects and body w eight changes Liv er and lung lesions, increased urinary porphyrins Sperm production Increased serum F S H H orm one levels (serum estradiol) FR 2 revolutions Increased relative liver w eight Administered dosea NOAEL LOAEL BMDLd -- 1.65E+01 -- 1 .00E -01 1.00E+02 -- -- 3.00E+01 2.14E +00 1.00E+01 1.88E+01 7.14E +00 7.14E+01 1 .0 0 E + 0 0 1 .0 0 E + 0 1 7 .3 0 E --01 1 .0 0 E + 0 0 1 .5 6 E --02 3.00E +00 1.00E+01 3.60E+03 -- 2.00E +00 1.08E+02 -- 2.00E+01 7.34E +00 1.07E+01 3.50E+01 -- 1st-order body burden HEDb NOAEL LOAEL BMDLd -- 1 .0 5 E --01 -- 3 .1 8 E --04 3 .1 8 E --01 -- -- 7 .8 9 E --03 5 .6 4 E --04 2 .5 8 E --03 4 .8 5 E --03 4 .5 3 E --02 4 .5 3 E --01 1 .0 7 E --02 1 .0 7 E --01 7 .8 4 E --03 3 .8 7 E --03 6 .0 3 E --05 7 .8 9 E --04 2 .6 3 E --03 9 .4 7 E --01 -- 9 .8 5 E --04 5 .3 3 E --02 -- 6 .2 5 E --03 2 .2 9 E --03 8 .9 7 E --02 2 .9 3 E --01 -- Blood concentration HEDc NOAEL LOAEL BMDLd -- 2.75E+00 -- ------ -- 8 .9 9 E --02 1 .7 1 E --03 9 .8 1 E --03 1 .7 0 E --02 2 .6 8 E --01 3 .1 0 E + 0 0 6 .4 6 E --02 6 .4 6 E --01 2 .0 0 E --02 1 .6 7 E --02 3 .8 3 E --05 2 .9 7 E --03 1 .7 2 E --02 2 .3 8 E + 0 1 -- 1 .5 7 E --03 3 .4 6 E --01 -- 5 .1 4 E --02 1 .1 8 E --02 ------ D RA FT: DO NOT C IT E OR Q U O TE Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered dose, 1st-order body burden HED and blood concentration HED (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-36 Study Endpoint M iettinen et al. (2006, 198266) M urray et al. (1979, 197983) N T P (1982, 200870) N T P (2006, 197605) N ohara et al. (2000, 200027) O hsako et al. (2001, 198497) Seo et al. (1995, 197869) Sew all et al. (1995, 198145) Shi et al. (2007, 198147) Sim anainen et al. (2002, 201369) Sim anainen et al. (2003, 198582) Sim anainen et al. (2004, 198948) Sm ialow icz et al. (2004, 198948) Sm ialow icz et al. (2008, 198341) Cariogenic lesions in pups Fertility index in f2 generation Liver lesions Liver and lung lesions D ecreased spleen cellularity A nogenital distance in pups Decreased thym us w eight Serum T4 Serum estradiol in fem ale pups Decreased serum T4 Decreased thym us w eight and change in E R O D activity Decreased daily sperm production D ecreased antibody response to S R B C s P F C per 10A6 cells Administered dosea NOAEL LOAEL BMDLd -- 3.00E+01 1.05E+01 1st-order body burden HEDb NOAEL LOAEL BMDLd -- 7 .8 9 E --03 2 .7 7 E --03 Blood concentration HEDc NOAEL LOAEL BMDLd -- 8 .9 3 E --02 9 .3 2 E --03 1 .0 0 E + 0 0 1 .0 0 E + 0 1 1 .6 3 E + 0 0 9 .4 3 E --03 9 .4 3 E --02 1 .5 4 E --02 2 .9 6 E --02 3 .8 8 E --01 4 .0 5 E --02 -- -- 8 .0 0 E + 0 2 1 .3 9 E + 0 0 2 .1 4 E + 0 0 -- 4.68E +00 5 .0 4 E --01 -- -- -- 2 .1 0 E --01 6 .4 7 E --03 2 .3 4 E --02 -- 2 .1 8 E --02 5 .5 0 E --03 -- -- -- 5 .3 4 E + 0 0 2 .2 1 E --02 1 .3 9 E --01 -- 5 .2 0 E --02 7 .3 8 E --03 -- 1 .2 5 E + 0 1 5 .0 0 E + 0 1 9 .7 5 E + 0 0 3 .2 9 E --03 1 .3 2 E --02 2 .5 7 E --03 2 .7 5 E --02 1 .7 8 E --01 1 .8 4 E --02 2.50E+01 1.00E+02 -- 2 .4 9 E --02 9 .9 6 E --02 -- 1 .6 7 E --01 9 .1 5 E --01 -- 1 .0 7 E + 0 1 3 .5 0 E + 0 1 5 .1 6 E + 0 0 8 .9 7 E --02 2 .9 3 E --01 4 .3 3 E --02 5 .1 5 E --01 1 .7 6 E + 0 0 1 .8 4 E --01 1 .4 3 E --01 7 .1 4 E --01 2 .2 4 E --01 1 .2 3 E --03 6 .1 3 E --03 1 .9 2 E --03 4 .7 1 E --03 2 .7 5 E --02 4 .9 5 E --03 1.00E+02 3.00E +02 -- 2 .6 3 E --02 7 .8 9 E --02 -- -- -- -- 1.00E+02 3.00E+02 2 .6 3 E --02 7 .8 9 E --02 1.00E+02 3.00E+02 -- 2 .6 3 E --02 7 .8 9 E --02 -- -- -- -- 3.00E +02 1.00E+03 -- 7 .7 3 E --02 2 .5 8 E --01 -- -- -- -- -- 1 .0 7 E + 0 0 4 .0 9 E --01 -- 5 .0 0 E --03 1 .9 1 E --03 -- 6 .3 8 E --03 2 .0 0 E --03 D RA FT: DO NOT C IT E OR Q U O TE Table 4-3. Summary of key animal study PODs (ng/kg-day) based on three different dose metrics: administered dose, 1st-order body burden HED and blood concentration HED (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Study Toth et al. (1979, 197109) V an B irgelen et al. (1995, 198052) V o s et al. (1973, 198367) W hite et al. (1986, 197531) Y a n g et al. (2000, 198590) Endpoint Skin lesions D ecreased liver retinyl palm itate Decreased delayed-type hypersensitivity response to tuberculin Decreased serum com plem ent Increased endom etrial im plant survival Administered dosea NOAEL LOAEL BMDLd -- 1.00E+00 2.15E +02 1st-order body burden HEDb NOAEL LOAEL BMDLd -- 3 .7 0 E --03 7 .9 4 E --01 Blood concentration HEDc NOAEL LOAEL BMDLd -- 1 .0 0 E --02 2 .1 8 E --01 - 1 .4 0 E + 0 1 9 .8 9 E + 0 2 -- 8 .6 3 E --02 6 .0 9 E + 0 0 -- 5 .2 5 E --01 5 .0 0 E + 0 0 1.14E+00 5.71E+00 6 .4 3 E --03 3 .2 2 E --02 -- 1 .0 0 E + 0 1 3 .5 9 E + 0 1 -- 2 .2 3 E --02 7 .9 8 E --02 -- 2 .8 3 E --02 4 .6 5 E --02 1 .7 9 E + 0 1 -- -- 6 .7 4 E --01 -- -- -- -- -- 4-37 D RA FT: DO NOT C IT E OR Q U O TE aA v e r a g e a d m in iste re d d a ily d o se o v e r th e e x p e rim e n ta l e x p o su re p e rio d . bH E D b a s e d o n 1st-o rd e r b o d y b u rd e n m o d e l d e s crib e d in S e c tio n 3 .2 .4 .4 . cH E D b a s e d o n E m o n d ro d e n t a n d h u m a n P B P K m o d e ls d e s crib e d in S e c tio n 3 .3 .6 . dB M R = 0 .1 fo r q u a n ta l e n d p o in ts a n d 1 stan d ard d e v ia tio n c o n tro l m e a n fo r co n tin u o u s e n d p o in ts, e x c e p t fo r b o d y a n d o r g a n w e ig h ts , w h e re B M R = 1 0% relative deviation from control m ean. - = value not established or not m odeled. Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kg)a This document is a draftfo r review purposes only and does not constitute Agency policy. 4-38 Study Amin et al. (2000, 197169) (rat) NOAEL/ LOAEL Endpoint Control First Max response responseb responsec Model fit detail BMD/ BMDL Saccharin 3.38E+00 consumed, female, (0.25%) (n = 10) 22% I (0.3 SD) 66% I Continuous linear, 9.15E+00 nonconstant variance 6.09E+00 (p = 0.55) Continuous power, 8.37E+00 nonconstant variance, 3.42E+00 unrestricted (p = NA) Saccharin consumed, female (0.50%) (n = 10) 49% I (0.7 SD) 80% I Continuous linear, 1.02E+01 nonconstant variance 6.57E+00 (p = 0.06) Continuous power, 6.57E+00 nonconstant variance, 1.15E+00 unrestricted (p = NA) Saccharin preference ratio, female (0.25%) (n = 10) 29% I (1.8 SD) 33% I Continuous linear, 1.16E+01 nonconstant variance 5.57E+00 (p = 0.002) Saccharin preference ratio, female (0.50%) (n = 10) 39% I (1.1 SD) 54% I Continuous linear, constant variance (p = 0.14) 8.14E+00 5.11E+00 Continuous power, constant variance, unrestricted (p = NA) 2.60E+00 1.06E-14 Comments BMDL > LOAEL; restricted power model, constrained parameter hit lower bound Saturated model; supralinear fit (power = 0.74) Restricted power model, constrained parameter hit lower bound Saturated model; supralinear fit (power = 0.40) BMDL > LOAEL; no response near BMR; near maximal response at LOAEL BMDL > LOAEL; near maximal response at LOAEL; restricted power model, constrained parameter hit lower bound Saturated model; supralinear fit (power = 0.28) D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. Study Bell et al. (2007, 197041) (rat) Cantoni et al. (1981, 197092) (rat) Crofton et al. (2005, 197381) (rat) NOAEL/ LOAEL Endpoint Control First Max response response response Model fit detail BMD/ BMDL Balano-preputial 1/30 5/30 15/30 Dichotomous log 2.25E+00 2.20E+00 separation in male logistic, restricted 1.39E+00 pups (p = 0.78) (n = 30 [dams]) Dichotomous log 2.00E+00 logistic, unrestricted 2.80E-01 (p = 0.50) Urinary uroporhyrins 1.85E+00 (n = 4) 2.4-fold T (5.7 SD) 87-fold T Continuous exponential (M2), nonconstant variance (p = 0.0003) 3.76E+00 2.76E+00 Urinary coproporhyrins (n = 4) 2.4-fold T 4.0-fold T Continuous (3.1 SD) exponential (M4), nonconstant variance (p = 0.49) 5.34E-01 1.80E-01 Continuous power, nonconstant variance, unrestricted (p = 0.61) 2.77E-02 2.03E-05 3.46E+00 Serum T4, 9.26E+00 (n = 4-14) 29% I (1.9 SD) 51% I Continuous exponential (M4), constant variance (p = 0.94) 5.19E+00 3.03E+00 Comments Adequate fit; constrained parameter bound hit; not litter based; selected Supralinear fit (slope = 0.93); selected No response near BMR; poor fits for all nonconstant variance models; constant variance poor representation of control SD; BMDL > LOAEL No response near BMR Supralinear fit (n = 0.30); poor model choice for plateau effect No response near BMR 4-39 D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-40 Study Franc et al. (2 0 0 1 , 197353) (rat) NOAEL/ LOAEL 6.58E +00 1 .4 5 E + 0 1 Endpoint S-D Rats, Relative Liver W eight L -E Rats, Relative Liver W eight S-D Rats, Relative Thym us W eight L -E Rats, Relative Thym us W eight H /W Rats, Relative Thym us W eight Control response First Max response response Model fit detail BMD/ BMDL 8.1% t 55% t (0.58 SD ) Continuous pow er, constant variance (p = 0.84) 9 .4 7 E + 0 0 4.59E +00 6.3% t 22% t (0.63 SD ) Continuous H ill, 7.72E +00 nonconstant variance, 1.22E+00 restricted (p = 0 .8 3 ) Continuous H ill, 7.22E +00 nonconstant variance, 1.15E+00 unrestricted (p = N / A ) 9.0% I 77% I (0.11 SD ) Continuous exponential (M 4), nonconstant variance (p = 0 .7 2 ) 1 .8 8 E + 0 0 9 .2 2 E -0 1 Continuous polynom ial, nonconstant variance (p = 0 .4 0 ) 4.78E +00 3 .8 9 E + 0 0 7.7% I 66% I (0.15 SD ) Continuous exponential (M 4), constant variance (p = 0 .2 3 ) 2.08E +00 5 .9 3 E -0 1 3.7% I 51% I (0.10 SD ) Continuous exponential (M 2), constant variance (p = 0 .7 0 ) 5 .0 9 E + 0 0 3 .1 3 E + 0 0 Comments Acceptable fit Constrained param eter hit low er bound; otherw ise acceptable fit; selected Supralinear fit (pow er = 0.55) Poor fit for responses in controls and low est exposure group Acceptable fit Poor fit for responses in controls and low est exposure group; doseresponse relationship not significant D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. 4-41 Study Hojo et al. (2002, 198785) (rat) Kattainen et al. (2001, 198952) (rat) Keller et al. (2007, 198526; 2008, 198531; 2008, 198033) (mouse) NOAEL/ LOAEL Endpoint DRL reinforce per 1.62E+00 min (n = 12) Control response DRL response per min (n = 12) 3rdmolar length in 2.23E+00 pups (n = 4-8) 3rdmolar eruption in pups (n = 4-8) 1/16 Missing molars 5.37E-01 (n = 23-36) 0/29 First response 55% T (1.0 SD) 105% 1 (2.4 SD) 15% 1 (4.2 SD) 3/17 2/23 Max response Model fit detail BMD/ BMDL 80% T Continuous exponential (M4), constant variance (p = 0.054) 1.32E+00 2.37E-03 105% 1 Continuous exponential (M4), constant variance (p = 0.48) 3.81E-01 1.55E-02 27% 1 Continuous Hill, 3.13E-01 nonconstant variance, 1.68E-01 restricted (p = 0.02) Continuous Hill, nonconstant variance, unrestricted (p < 0.001) 1.21E-02 13/19 Dichotomous log logistic, restricted (p = 0.98) 2.40E+00 1.33E+00 Dichotomous log 1.93E+00 logistic, unrestricted 1.84E-01 (p = 0.95) 30/30 Dichotomous 1 multistage (p = 0.26) 1.09E+00 7.62E-01 Comments Poor fit; near maximal response at lowest dose, BMD/BMDL ratio 100 No response data near BMR; maximal response at lowest dose, BMD/BMDL ratio 20 No response data near BMR; Constrained parameter lower bound hit BMDL could not be calculated Constrained parameter lower bound hit Supralinear fit (slope = 0.91) Poor fit at first response level; not most sensitive endpoint; other endpoints not amenable to BMD modeling D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-42 Study Kociba et al. (1978, 001818) (rat) Latchoumycandane and Mathur (2002, 197498) (rat) NOAEL/ LOAEL Endpoint 1.55E+00 Uroporphyrin per 7.15E+00 creatinine, females (n = 5) Urinary coproporphyrins, females (n = 5) Liver lesions (n = 50) Lung lesions (n = 50) Daily sperm 7.85E-01 production (n = 6) Li et al. (1997, 199060) (rat) 2.66E-01 FSH in female rats 7.99E-01 (n = 10) Control response First response 15% (0.48 SD) 67% (5.1 SD) Max response Model fit detail BMD/ BMDL 89% Continuous linear, constant variance (p = 0.79) 1.31E+01 9.29E+00 78% Continuous exponential (M4), nonconstant variance (p = 0.01) 1.57E+00 7.18E-01 Comments BMDL > LOAEL; otherwise adequate fit Poor fit; no response near BMR No data presented No data presented 29% I (1.0 SD) 3.6-fold (2.0 SD) 41% I Continuous Hill, constant variance, restricted (p = 0.96) 1.17E-01 1.32E-02 Continuous Hill, constant variance, unrestricted (p = N/A) 9.96E-02 1.23E-09 19-fold Continuous power, 2.00E+02 nonconstant variance, 1.36E+02 restricted (p < 0.01) Continuous power, 1.96E-01 nonconstant variance, 2.48E-02 unrestricted (p = 0.003) Near maximal response at LOAEL; constrained parameter bound hit; standard deviations given in paper interpreted as standard errors Slightly supralinear fit (n = 0.92) Power hit lower bound supralinear fit (power = 0.31) D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. 4-43 Study Li et al. (2006, 199059) (mouse) Markowski et al. (2001, 197442) (rat) NOAEL/ LOAEL Endpoint Serum estradiol 1.59E-01 (n = 10) Serum progesterone (n = 10) FR5 run 1.56E+00 opportunities (n = 4-7) FR2 revolutions (n = 4-7) FR10 run opportunities (n = 4-7) Control First Max response response response Model fit detail 2.0-fold 2.4-fold Continuous linear, (0.8 SD) constant variance (p = 0.16) BMD/ BMDL 1.61E+01 5.38E+00 33% I 61% I (2.0 SD) Continuous Hill, 9.46E-04 nonconstant variance 8.01E-11 (p = 0.39) 10% I (0.21 SD) 9% I (0.15 SD) 15% I (0.24 SD) 51% I 43% I 57% I Continuous Hill, constant variance (p = 0.94) Continuous power, constant variance, unrestricted (p = 0.13) Continuous Hill, constant variance (p = 0.65) Continuous power, constant variance, unrestricted (p = 0.16) Continuous exponential (M2) , constant variance (p = 0.30) 1.72E+00 9.08E-01 2.67E+00 1.03E-14 1.84E+00 5.99E-01 5.74E+00 1.03E-14 8.57E+00 2.89E+00 Comments BMDL > LOAEL; high control CV (1.25); near maximal response at low dose; nonmonotonic response; other model fits are step-function-like No response data near BMR; large CVs (>1) for treatment groups; poor fit for variance model; Hill coefficient at lower bound (step function) Constrained parameter upper bound hit Saturated model; supralinear fit (power = 0.39); BMD/BMDL ratio 100 Constrained parameter bound hit (upper bound) Supralinear fit (power = 0.32) BMDL > LOAEL D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-44 Study Miettinen et al. (2006, 198266) (rat) Murray et al. (1979, 197983) (rat) NTP (1982, 200870) (mouse) NTP (2006, 197605) (rat) NOAEL/ LOAEL Endpoint Cariogenic lesions 2.22E+00 in pups (n = 4-8) Control response 25/42 1.12E+00 Fertility in f2 gen. 5.88E+00 (no litters) (n = 20) Toxic hepatitis; 7.67E-01 males (n = 50) Hepatocyte 2.56E+00 hypertrophy (n = 53-54) Alveolar metaplasia (n = 52-54) 4/32 1/73 0/53 2/53 Oval cell hyperplasia 0/53 (n = 53-54) First response 23/29 0/20 5/49 19/54 19/54 4/54 Max response Model fit detail 29/32 Dichotomous log logistic, restricted (p = 0.60) Dichotomous log logistic, unrestricted (p = 0.73) 9/20 Dichotomous multistage (p = 0.08) 44/50 Dichotomous multistage (p = 0.04) 52/53 Dichotomous multistage (p = 0.02) 46/52 53/53 Dichotomous log logistic (p = 0.72) Dichotomous probit (p = 0.23) BMD/ BMDL 1.43E+00 5.17E-01 4.94E-02 2.73E+00 1.37E+00 2.78E+00 1.34E+00 9.27E-01 7.91E-01 6.50E-01 3.75E-01 5.67E+00 4.79E+00 Dichotomous Weibull 5.72E+00 (p = 0.08) 4.09E+00 Comments Constrained parameter lower bound hit; near maximal response at LOAEL; high control response Supralinear fit (slope = 0.47); BMDL could not be calculated Poor fit; nonmonotonic response; no response data near BMR No acceptable model fits; lowest BMDL shown Poor fits for all models No response near BMR Relatively poor fit for control and low dose groups; negative response intercept (same for logistic); BMDL > LOAEL Marginal fit; BMDL > LOAEL D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-45 Study NTP (2006, 197605) (rat) (continued) NOAEL/ LOAEL Endpoint Control response Gingival hyperplasia 2.56E+00 (n = 53-54) (continued) 1/53 Eosinophilic focus, multiple (n = 53-54) Liver fatty change, diffuse (n = 53-54) Liver necrosis (n = 53-54) 3/53 0/53 1/53 Liver pigmentation (n = 53-54) 4/53 Toxic hepatopathy (n = 53-54) 0/53 Ohsako et al. 1.04E+00 Ano-genital distance (2001, 198497) 3.47E+00 in male pups (rat) (n = 5) First response 7/54 8/54 2/54 4/54 9/54 2/54 12% I (1.0 SD) Max response Model fit detail 16/53 Dichotomous log logistic, restricted (p = 0.06) Dichotomous log logistic, unrestricted (p = 0.66) 42/53 Dichotomous probit (p = 0.46) BMD/ BMDL 5.85E+00 3.73E+00 7.05E-01 1.26E-05 5.58E+00 4.86E+00 Comments Poor fit; constrained parameter bound hit; BMDL > LOAEL Supralinear fit (slope = 0.37) Relatively poor fit to control response; BMDL > LOAEL 48/53 Dichotomous Weibull 3.92E+00 BMDL > LOAEL; otherwise (p = 0.72) 2.86E+00 adequate fit 17/53 53/53 53/53 17% I Dichotomous log probit, unrestricted (p = 0.80) Dichotomous log probit (p = 0.96) Dichotomous multistage (p = 0.69) Continuous Hill, constant variance, restricted (p = 0.15) Continuous Hill, constant variance, unrestricted (p = 0.056) 7.50E+00 Adequate fit; slightly supralinear; 3.50E+00 BMDL > LOAEL 2.46E+00 Adequate fit 1.89E+00 3.98E+00 BMDL > LOAEL; otherwise 3.06E+00 adequate fit 2.88E+00 8.03E-01 Constrained parameter lower bound hit; near maximal response at LOAEL 3.49E+00 Supralinear fit (n = 0.59) 3.05E-01 D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. 4-46 Study Sewall et al. (1995, 198145) (rat) NOAEL/ LOAEL Endpoint 7.11E+00 Serum T4 1.66E+01 (n = 9) Shi et al. (2007, 3.42E-01 Serum estradiol in 198147) 1.07E+00 female pups (rat) (n = 10) Smialowicz et al. (2008, 198341) (mouse) PFC per spleen 4.38E-01 (n = 15) PFC per 10A6 cells (n = 8-15) Toth et al. Skin lesions (1979, 197109) 5.73E-01 (n = 38-44) (mouse) Control First Max response response response Model fit detail BMD/ BMDL 9.1% I (0.6 SD) 40% I Continuous Hill, constant variance, restricted (p = 0.90) 1.03E+01 3.60E+00 Continuous Hill, constant variance, unrestricted (p = 0.86) 9.71E+00 1.97E+00 38% I (0.4 SD) 62% I Continuous exponential (M4), nonconstant variance (p = 0.69) 8.07E-01 3.54E-01 24% I (0.5 SD) 24% I (0.5 SD) 89% I 9.3-fold I Continuous power, unrestricted, nonconstant variance (p = 0.27) Continuous power unrestricted, constant variance (p = 0.48) 1.19E+01 3.76E+00 1.90E+00 2.16E-01 0/38 5/44 25/43 Dichotomous log logistic, restricted (p = 0.08) 6.41E+00 4.02E+00 Dichotomous log-logistic, unrestricted (p = 0.74) 5.97E-01 6.77E-02 Comments Constrained parameter hit lower bound; otherwise acceptable fit; selected Supralinear fit (power = 0.57) Adequate fit; selected BMDL > LOAEL; fit at control and low dose inconsistent with data; constrained parameters in other models hit lower bounds Constant variance test failed; observed control variance underestimated by 35%; poor fits for all nonconstant variance models Constrained parameter lower bound hit Supralinear fit (slope = 0.48) D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. Study Toth et al. (1979, 197109) (mouse) (continued) NOAEL/ LOAEL Endpoint Dermal amyloidosis 5.73E-01 (n = 38-44) (cont.) Van Birgelen et al. (1995, 198052) (rat) 7.20E+00 Hepatitic retinol (n = 8) Hepatitic retinyl palmitate (n = 8) Control First Max response response response Model fit detail BMD/ BMDL 0/38 5/44 17/43 Dichotomous log logistic, restricted (p = 0.05) 1.50E+01 8.75E+00 Dichotomous log 4.84E-01 logistic, unrestricted 5.31E-03 (p = 0.90) 44% I 96% I (0.74 SD) Continuous exponential (M4), nonconstant variance (p < 0.01) 2.49E+01 3.36E+00 Continuous power, 3.80E-01 nonconstant variance, 1.39E-02 unrestricted (p = 0.01) 80% I 99% I (1.4 SD) Continuous exponential (M4), nonconstant variance (p < 0.01) 1.42E+02 3.65E+01 Continuous power, 5.26E-02 nonconstant variance, 5.89E-05 unrestricted (p = 0.24) Comments Poor fit; constrained parameter lower bound hit; BMDL > LOAEL Supralinear fit (slope = 0.33) Poor fit Poor fit; supralinear fit (power = 0.14) Poor fit; no response near BMR Supralinear fit (power = 0.06) 4-47 D RA FT: DO NOT C IT E OR Q U O TE Table 4-4. TCDD BMDL analysis (NOAEL, LOAEL, BMD, and BMDL values given as animal whole blood concentrations in ng/kga) (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. Study White et al. (1986, 197531) (mouse) NOAEL/ LOAEL Endpoint 1.09E+00 Total hemolytic complement activity (CH50) (n = 8) Control response First response 41% I (2.6 SD) Max response Model fit detail BMD/ BMDL 81% I Continuous Hill, nonconstant variance, restricted (p = 0.002) 8.63E+00 1.50E+00 Continuous Hill, 1.48E-01 nonconstant variance, 4.35E-03 unrestricted (p = 0.07) Comments Poor fit; no response near BMR; constrained parameter bound hit; BMDL > LOAEL Supralinear fit (n = 0.25) aAnimal whole blood concentrations were used to determine the HEDs in Table 4-5. bMagnitude of response at first dose where response differs from control value (in the adverse direction); continuous response magnitudes given as relative to control plus change relative to control standard deviation; quantal response given as number affected/total number. cMagnitude of response maximally differing from control value (in the adverse direction). S-D = Sprague-Dawley. SD = standard deviation. 4-48 D RA FT: DO NOT C IT E OR Q U O TE Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human equivalent doses This document is a draftfor review purposes only and does not constitute Agency policy. 4-49 Study Species, strain (sex, if not both) Protocol L i et al. (2006, M ouse, N IH (F) G avage G D 1-3; 199059) n = 10 Sm ialow icz et M ouse, B 6C3F1 90-day gavage; al. (2008, (F) n =8-15 198341) K eller et al. M ouse, C B A /J (2007, 198526; and C3H /H eJ 2008, 198531; 2008, 198033)b G avage G D 13; n =23-36 (pups) Toth et al. M ouse, Sw iss/ 1-year gavage; (1979, 197109) H /Riop (M ) n = 38-44 Latchoum y- candane and M athur (2002, 197498) R at, W istar (M ) 45-day oral pipetting; n =6 N T P (1982, 200870) M ouse, B6C3F1 2-year gavage; (M ) n =50 W hite et al. M ouse, B 6C 3F1 14-day gavage; (1986, 197531) (F) n =6-8 L i et al. (1997, R at, S-D Single gavage; 199060) (F , 22 day-old) n = 10 D eCaprio et al. G uin ea pig, (1986, 197403) Hartley 90-day dietary; n = 10 Shi et al. (2007, R at, S-D (F) 198147) 11-m onth gavage; n = 10 M arkow ski et R at, H oltzm an G avage G D 18; al. (2001, n =4-7 197442) Endpoint Horm one levels in pregnant dam s (decreased progesterone, increased estradiol) Decreased S R B C response NOAELh e d (N) or BMDLh e d (B) (ng/kg-day) - loaelh e d (ng/kg-day) 1.6E -03 6 .4 E -0 3 M issin g m olars, m andibular shape changes in pups 9 .8 E -03 D erm al am yloidosis, skin lesions Decreased sperm production - 1 .0 E -02 1 .7 E -02 L iv er lesions Decreased serum com plem ent Increased serum F S H Decreased body w eight, organ w eight changes (liver, kidney, thym us, brain) D ecreased serum estradiol N eurobehavioral effects in pups (running, lever press, w heel spinning) - - 3 .0 E -0 3 (N) 4 .1 E -0 3 d (N) 4 .7 E -0 3 (N) 5 .0 E -0 3 (B) 2 .2 E -0 2 2 .8 E -0 2 1 .7 E -02 3 .3 E -0 2 d 2 .8 E -0 2 5 .1 E -0 2 oo oo oo CO CO CO RfD UFa (mg/kg-day) 300 5 .2 E -1 2 300 2 .1 E -1 1 300 3 .3 E -1 1 300 3 .3 E -1 1 300 5 .6 E -1 1 300 7 .4 E -1 1 300 9 .4 E -1 1 9 .9 E -1 1 1 .4 E -10 1 .6 E -10 300 1 .7 E -10 D RA FT: DO NOT C IT E OR Q U O TE Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human equivalent doses (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-50 Study Species, strain (sex, if not both) Protocol Endpoint NOAELh e d (N) or BMDLh e d (B) LOAELh e d (ng/kg-day) (ng/kg-day) Hojo et al. Rat, S-D (2002, 198785) Gavage GD 8; n = 12 Food-reinforced operant behavior in pups -- 5.5E--02 Vos et al. Guinea pig, (1973, 198367) Hartley (F) 8-week gavage; n = 10 Decreased delayed-type hypersensitivity response to tuberculin 6.4E-03d(N) 3.2E--02d Cantoni et al. Rat, CD-COBS 45-week gavage; (1981, 197092) (F) n =4 Increased urinary porhyrins -- 6.5E--02 Miettinen et al. Rat, Line C (2006, 198266) Gavage GD 15; n = 3-10 Cariogenic lesions in pups -- 8.9E--02 Kattainen et al. Rat, Line C (2001, 198952) Gavage GD 15; n = 4-8 Inhibited molar development in pups -- 9.0E--02 NTP (2006, 197605) Rat, S-D (F) 2-year gavage; n = 53 Liver and lung lesions -- 1.4E--01 Amin et al. Rat, S-D (2000, 197169) Gavage GD 10-16; Reduced saccharin consumption and n = 10 preference -- 1.7E--01 Mocarelli et al. Human (M) (2008, 199595) Childhood Decreased sperm concentration and sperm exposure; n = 157 motility, as adults -- 2.0E--02e Baccarelli et al. (2008, 197059) Human infants Gestational exposure; n = 51 Increased TSH in newborn infants 2.4E--02g Hutt et al. Rat, S-D (F) (2008, 198268) 13-week dietary; n =3 Embryotoxicity - 2.6E+00 Ohsako et al. Rat, Holtzman Gavage GD 15; (2001, 198497) n =5 Decreased ano-genital distance in male pups 2.8E--02 (N) 1.8E--01 Murray et al. Rat, S-D (1979, 197983) 3-generation dietary Reduced fertility and neonatal survival (f 0 and f 1) 3.0E--02 (N) 3.9E--01 oo oo oo CO CO CO RfD UFa (mg/kg-day) 300 1.8E--10 2.1E--10 300 2.2E--10 300 3.0E--10 300 3.0E--10 300 4.6E--10 300 5.7E--10 30f 6.7E--10 30f 8.2E--10 300 8.6E--10 9.2E--10 9.9E--10 D RA FT: DO NOT C IT E OR Q U O TE Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human equivalent doses (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-51 Study Species, strain (sex, if not both) Protocol Endpoint Franc et al. Rat, Long-Evans 22-week gavage; (2001, 197353) (F) n =8 Increased Relative Liver Weight; decreased relative thymus weight Chu et al., 2007 Rat, S-D (F) 28-day gavage, n =5 Liver lesions Bell et al. Rat, CRL:WI 17-week dietary; (2007, 197041) (Han) (M) n = 30 Delay in onset of puberty Van Birgelen et Rat, S-D (F) al. (1995, 198052) 13-week dietary; n =8 Decreased liver retinyl palmitate Kociba et al. Rat, S-D (F) (1978, 001818) 2-year dietary; n = 50 Liver and lung lesions, increased urinary porhyrins Fattore et al., Rat, S-D (2000, 197446) 13-week dietary; n=6 Decreased hepatic retinol Seo et al. (1995, Rat, S-D 197869) Gavage GD 10-16; Decreased serum T4 and thymus weight n = 10 Crofton et al. Rat, Long-Evans 4-day gavage; (2005, 197381) (F) n = 4-14 Decreased serum T4 Sewall et al. Rat, S-D (F) (1995, 198145) 30-week gavage; n =9 Decreased serum T4 Alaluusua et al. Human (2004, 197142) Childhood exposure; Dental defects n = 48 NOAELh e d (N) or BMDLh e d (B) loaelh e d (ng/kg-day) (ng/kg-day) 4.6E-01 (N) 3.4E-02 (B) 1.45E+00 3.6E-02 (N) 5.8E-01 4.3E-02 (B) 8.8E-02 5.3E-01 6.5E-02 (N) - 1.7E-01 (N) 1.7E-01 (N) 5.2E-01 (N) 1.8E-01 (B) 1.2E-01h (N) 6.5E-01 8.0E-01 9.1E-01 7.6E-01 1.8E+00 9.3E-011 oo oo oo oo CO CO CO CO oo oo oo CO CO CO RfD UFa (mg/kg-day) 1.1E-09 1.2E-09 1.4E-09 300 1.8E-09 2.2E-09 300 2.7E-09 5.6E-09 5.7E-09 6.1E-09 3J 3.9E-08 aExcept where indicated, UFA= 3 (for dynamics), UFH= 10, UFL= 10. bResults from 3 separate studies with identical designs combined. cUFL= 1 (NOAEL or BMDL). dHED determined from 1st-order body burden model; no PBPK model available for guinea pigs. eMean of peak exposure (0.0319 ng/kg-day) and average exposure over 10-year critical window (0.00802 ng/kg-day). fUFH= 3, UFl = 10. gMaternal exposure corresponding to neonatal TSH concentration exceeding 5 pU/mL. D RA FT: DO NOT C IT E OR Q U O TE Table 4-5. Candidate points of departure for the TCDD RfD using blood-concentration-based human equivalent doses (continued) hMean of peak exposure (0.200 ng/kg-day) and average exposure over 10-year critical window (0.0335 ng/kg-day). 1 Mean of peak exposure (1.71 ng/kg-day) and average exposure over 10-year critical window (0.153 ng/kg-day). jUFh = 3. S-D = Sprague-Dawley. This document is a draftfor review purposes only and does not constitute Agency policy. 4-52 D RA FT: DO NOT C IT E OR Q U O TE Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays possessing candidate points-of-departure for the TCDD RfD This document is a draftfor review purposes only and does not constitute Agency policy. 4-53 Study Strengths Limitations Remarks Bell et al. (2007, Large sample size of both rat dams and 197041) offspring/dose employed Several developmental effects tested Batch-to-batch variation of up to 30% in TCDD concentration in the diet Longer-term dosing of dams does not accurately define gestational period when fetus is especially sensitive to TCDD-induced toxicity Study is a significant addition to a substantial database on the developmental toxicity of TCDD in laboratory animals Cantoni et al. (1981, 197092) Experiments were designed to test qualitative and quantitative composition and the course of urinary excretion in TCDD-induced porphyria Small sample size of rats/dose employed (n = 4) Concurrent histological changes with tissue porphyrin levels were not examined TCDD used for dosing was of unknown purity Early study on porphyrogenic effects of TCDD DeCaprio et al. (1986, 197403) Subchronic oral dosing duration up to 90 days. Male and female guinea pigs tested Relatively small sample size of guinea pigs/dose employed (n = 10) No histopathological analyses performed TCDD used for dosing was of unknown purity Limited subchronic study; PBPK model not available for estimation of HED Franc et al. (2001, 197353) Three different rat strains with varying sensitivities to TCDD were utilized (SpragueDawley, Long Evans, Han/Wistar) Longer-term oral dosing up to 22 weeks Relatively small sample size of rats/dose employed (n = 8) Only female rats were tested Concurrent liver histopathological changes with liver weight changes were not examined Gavage exposure was only biweekly Limited subchronic study Hojo et al. (2002, Low TCDD dose levels used allowed for subtle 198785) behavioral deficits to be identified in rat offspring Preliminary training sessions in operant chamber apparatuses were extensive Neurobehavioral effects are exposure-related and cannot be attributed to presence of learning or discrimination deficits Relatively small sample size of rat dams/dose employed (n = 12) Small sample size of rat offspring/dose evaluated One of a few neurobehavioral toxicity studies; somewhat limited study size (n = 5-6) Neurobehavioral effects induced by TCDD at earlier or later gestational dosing dates are unknown because of single gavage administration on GD 8 Although BMD analysis was conducted, the model parameters were not constrained according to EPA guidance, so the results cannot be used D RA FT: DO NOT C IT E OR Q U O TE 1 Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays possessing candidate points-of-departure for the TCDD RfD (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-54 Study Strengths Limitations Remarks Keller et al. (2007, 198526; 2008, 198531; 2008, 198033) Six different inbred mouse strains were utilized Unknown sample size of mouse dams/dose/strain Endpoint similar to effects Large sample size of mouse offspring/dose/strain evaluated Low TCDD dose levels used compared to typical mouse studies allowed for identification employed All inbred strains possessed sensitive b allele at the Ahr locus (i.e., a potentially resistant subpopulation was not evaluated for comparison purposes) observed at higher exposure levels in humans; HED highly uncertain using mouse PBPK model of subtle sensitivity differences in presence of Morphological dental and mandibular changes absence of third molars, variant molar induced by TCDD at earlier or later gestational morphology, and mandible structure in offspring dosing dates are unknown because of single gavage administration on GD 13 Difficulties breeding A/J mice led to abandonment of that strain in the analysis (Keller et al., 2008a, b) Latchoumycandane and Mathur (2002, 197498) Compared to epididymal sperm counts, the testicular spermatid head count provides better quantitation of acute changes in sperm production and can indicate pathology Small sample size of rats/dose employed (n = 6) Oral pipette administration of TCDD may be a less efficient dosing method than gavage Endpoint has human relevance, similar to critical effects in principal human study for RfD Li et al. (2006, 199059) Female reproductive effects (i.e., early embryo loss and changes in serum progesterone and estradiol) were tested at multiple exposure times--early gestation, preimplantation, and peri- to postimplantation Small sample size of dams/dose (n = 10) Large dose-spacing interval (25-fold at lowest 2 doses) Endpoint has human relevance but HED highly uncertain using mouse PBPK model Markowski et al. (2001, 197442) Low TCDD dose levels used allowed for subtle behavioral deficits to be identified in rat offspring Several training sessions on wheel apparatuses were extensive Neurobehavioral effects are exposure-related and cannot be attributed to motor or sensory deficits Unknown sample size of rat dams/dose employed. Small sample size of rat offspring/dose evaluated (n = 4-7) One of a few neurobehavioral toxicity studies; somewhat limited study size TCDD used for dosing was of unknown purity and origin Only 2 treatment levels Neurobehavioral effects induced by TCDD at earlier or later gestational dosing dates are unknown because of single gavage administration on GD 18 D RA FT: DO NOT C IT E OR Q U O TE Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays possessing candidate points-of-departure for the TCDD RfD (continued) This document is a draftfor review purposes only and does not constitute Agency policy. 4-55 Study Strengths Limitations Remarks NTP (1982, 200870) Large sample size of mice and rats/dose employed Comprehensive 2-year bioassay that assessed body weights, clinical signs, and pathological changes in multiple tissues and organs Elevated background levels of hepatocellular tumors Comprehensive chronic in untreated male mice toxicity evaluations of TCDD Gavage exposure was only 2 days/week in rodents; HED highly Only 2 treatment levels uncertain using mouse PBPK model NTP (2006, 197605) Chronic exposure duration with several interim Single species, strain and sex sacrifices Lowest dose tested too high for establishing Large number of dose groups with close spacing NOAEL Large number of animals per dose group Comprehensive suite of endpoints evaluated Comprehensive biochemical, clinical and histopathological tests and measures Detailed reporting of results, with individual animal data presented as well as group summaries Study is the most comprehensive chronic TCDD toxicity evaluation in rats to date Shi et al. (2007, 198147) Study design evaluated TCDD effects on aging female reproductive system (i.e., exposure began in utero and spanned across reproductive lifespan) Several female reproductive endpoints were evaluated, including cyclicity, endocrinology, serum hormone levels, and follicular reserves Relatively small sample size of rats/dose employed (n = 10) Endpoint similar to effects observed at higher exposure levels in humans Smialowicz et al. Sheep red blood cell (SRBC) plaque forming (2008, 198341) cell assay is highly sensitive and reproducible across laboratories when examining TCDD Small sample size of animals/dose (n = 8) Only female mice were tested Thymus and spleen weights were only other immune response-related endpoints tested Limited immunotoxicity study Toth et al. (1979, Large sample size of mice/dose employed 197109) Chronic exposure duration Reporting of findings is terse and lacks sufficient detail (e.g., materials and methods, thorough description of pathological findings, etc.) Limited number of endpoints examined Only male mice were tested Limited chronic study; HED highly uncertain using mouse PBPK model D RA FT: DO NOT C IT E OR Q U O TE Table 4-6. Qualitative analysis of the strengths and limitations/uncertainties associated with animal bioassays possessing candidate points-of-departure for the TCDD RfD (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Study Strengths Limitations Remarks Vos et al. (1973, 198367) White et al. (1986, 197531) Three different animal species tested (guinea Small sample size of animals/dose employed in each Endpoints relevant to humans pigs, mice, and rats) experiment (n = 5-10) but study size limited; PBPK Effects of TCDD tested on both cell-mediated Only female guinea pigs and rats were tested, and model not available for and humoral immunity only male mice were tested estimation of HED Only one experimental assay was utilized to assess cell-mediated and humoral immunity in each animal species; humoral immunity was only investigated in guinea pigs TCDD used for dosing was of unknown purity Tcreoopmtarpelslheeetnemtacotoliyvmteipcfleucmonmcetnipotlnesmaelqeanusetsna(cCyeHo5f0t)heis sCSInimHgdan5ilv0lifisidcaaumacanptliltvelcyistoiyzdmee(poopfllneerlamtytesed/Cdn3owtseifistaehcmmotpoueltroasysaeufmdfree(ancdyt=)inb6ge-8) Endpoint similar to effects observed at higher exposure levels in humans; HED highly uncertain using mouse PBPK TCDD used for dosing was of unknown purity model 4-56 D RA FT: DO NOT C IT E OR Q U O TE 1 Table 4-7. Basis and derivation of the TCDD reference dose 2 Princi pal study detail Study POD (ng/kg-day) Critical effects M o carelli et al. (2008, 0.020 ( L O A E L ) 199595) Decreased sperm count (20% ) and motility (11% ) in men exposed to T C D D during childhood B a c ca re lli et al. (2008, 0.024 ( L O A E L ) 197059) Elevated T S H (> 5 p U /m L) in neonates RfD derivation P O D 0.020 ng/kg-day (2 .0 E -8 mg/kg-day) U F 30 (UFl = 10, UFh = 3) R f D 7 x 10-10 (7 E -1 0 ) mg/kg-day (2 .0 E - 8 - 30) Uncertainty factors LO AEL-to-N O A EL (UFl) 10 N o N O A E L established; cannot quantify low er exposure group in B a c ca re lli et al. (2008, 197059); magnitude o f effects at L O A E L sufficient to require a 10-fold factor. Human interindividual variability (UFh) 3 A factor o f 3 (1005) is used because the effects w ere elicited in sensitive populations. A further reduction to 1 w as not made because the sample sizes were relatively small, which, combined with uncertainty in exposure estimation, may not fully capture the range of interindividual variability. Interspecies extrapolation (UFa) Subchronic-to-chronic (UFs) 1 Hum an study. 1 Chronic effect levels are not w ell defined for humans; however, animal bioassays indicate that developmental effects are the most sensitive, occurring at doses low er than other effects noted in chronic studies. Considering that exposure in the principal studies encompasses the critical w indow of susceptibility associated with development, an U F to account for exposure duration is not warranted. Database sufficiency (UFD) 3 1 The database for T C D D contains an extensive range of human and animal studies that examine a com prehensive set o f endpoints. There is no evidence to suggest that additional data would result in a lower reference dose. This document is a draftfor review purposes only and does not constitute Agency policy. 4-57 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 4-1. EPA's process to select and identify candidate PODs from key 3 epidemiologic studies for use in the noncancer risk assessment of TCDD. For 4 each noncancer study that qualified for T C D D dose-response assessment using 5 the study inclusion criteria, E P A first evaluated the dose-response information 6 developed by the study authors for whether the study provided noncancer effects 7 and T C D D dose data for a toxicologically relevant endpoint. I f such data were 8 available, then E P A identified a N O A E L or L O A E L as a candidate P O D . Then, 9 E P A used a human kinetic model to estimate the continuous oral daily intake 10 (ng/kg-day) for the candidate P O D that could be used in the derivation of an R fD 11 based on the study data. I f all of this inform ation w as available, then the result 12 w as included as a candidate P O D . This document is a draftfor review purposes only and does not constitute Agency policy. 4-58 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 4-2. EPA's process to select and identify candidate PODs from key 3 animal bioassays for use in noncancer dose-response analysis of TCDD. Fo r 4 each noncancer endpoint reported in the studies that qualified for T C D D 5 dose-response assessment using the study inclusion criteria, E P A evaluated the 6 endpoint and eliminated it if it w as not toxicologically relevant for R fD deriation. 7 Th en , relevant endpoints not observed at the L O A E L (i.e ., reported at higher 8 doses) with B M D L s greater than the L O A E L were eliminated from further 9 analysis. Endpoints with L O A E L S greater than the m inim um L O A E L times 100 10 also w ere elim inated from further analysis. U sin g kinetic m odeling, E P A 11 developed human equivalent doses for each rem aining N O A E L / L O A E L / B M D L 12 associated w ith selected endpoints and included these as candidate P O D s. This document is a draftfor review purposes only and does not constitute Agency policy. 4-59 DRAFT--DO NOT CITE OR QUOTE Figure 4-3. Exposure-response array for ingestion exposures to TCDD. This document is a draftfor review purposes only and does not constitute Agency policy. 4-60 D RA FT: DO NOT C IT E OR Q U O TE Animal Bioassays ss ssssss 4 s ^ m \S>SaJ zb XSNySNNNN sss S S!SSS S < t z I 1 2', t>V t0!3 a I s O --" !-- _J_ 03. 4-- O --i x ,=:I-O - LL q ~ LL += O X Zffl' Z *IX X 9661 '|B je IIBM9S 9002 le uoyojo 9661- leis oes 0002 is J0 ejoned 8161 Ie 10 eqpo>| 09661- ie j U9|96jia uba 0 0 2 Ie ie usa 002 Ie I0 nqo 1.002 IE0 OUBJd _ _ 66 l IBJ0 AejjniAl sssss RSKa n 1002 |B 1 0>jl3SL|0 8002 "Ie Ie HnH 0002 P ie Ujiuv 9002 d lN KSSSSH I 5 aI |p assssSs 1.002 |e 10 ueuienex 9002 '10 10 uemneiiA] [,961, '10 10 luojues SSSSSSSK S S S_S_S J iK sssia ^ i sssss 61, '|G Ie SOA 2002 '|010 H [002 '|B10 !>|SMO>1JB[AJ 002 IBIS 'MS 986[ '|0 10 oudesea 66[ Ile !1 ESSSSSS 3 RSSSSiSS S iB S ' 9861 IBI 01IMAA 2861 d lN 2002-iniflBW >8 ip jB l 66L '10 10 Mioi q'B8002'/002'|BieJe||e>1 IS S S S S l 8002 '|G JZDiMOiejius 9002'10 10 !1 t?002 IBie ensnnieiv ~l 8002 '|0 10 ieiseeeg SS* 8002 '10 10 !l|eJ033O[Aj LOO CoO oh- CoO COD O CN LpU LpLJ LpLJ OLLJ pLLJ OLLJ pLLl LoL1 (Ep-6>1/6l4j) ejnsodxg |ejq Figure 4-4. Candidate RfD array. Human I This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-61 DRAFT: DO NOT CITE OR QUOTE [This page intentionally left blank.] 1 5. CANCER ASSESSMENT 2 3 4 5.1. QUALITATIVE WEIGHT-OF-EVIDENCE CARCINOGEN CLASSIFICATION 5 FOR 2,3,7,8-TETRACHLORODIBENZO-^-DIOXIN (TCDD) 6 5.1.1. Summary of National Academy of Sciences (NAS) Comments on the Qualitative 7 Weight-of-Evidence Carcinogen Classification for 2,3,7,8-Tetrachlorodibenzo8 ^-Dioxin (TCDD) 9 In its charge, the National Academy of Sciences (NAS) was requested to comment 10 specifically on U.S. Environmental Protection Agency (EPA)'s conclusion that 11 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) is best characterized as "carcinogenic to humans." 12 While indicating that distinction between the categories of "carcinogenic to humans" and "likely 13 to be carcinogenic to humans" is " .. .based more on semantics than on s c ie n c e ." (NAS, 2006, 14 198441, p. 141) and recommending that EPA " .s p e n d its energies and resources on more 15 carefully delineating the assumptions used in quantitative risk estimates for T C D D ." (NAS, 16 2006, 198441, p. 141) rather than on the qualitative cancer descriptor for TCDD, the NAS 17 provided the following comments: 18 19 . t h e classification of dioxin as "carcinogenic to humans" versus "likely to be 20 carcinogenic to humans" depends greatly on the definition and interpretation of 21 the specific criteria used for classification, with the explicit recognition that the 22 true weight of evidence lies on a continuum with no bright line that easily 23 distinguishes between these two categories. The committee agreed that, although 24 the weight of epidemiological evidence that dioxin is a human carcinogen is not 25 strong, the human data available from occupational cohorts are consistent with a 26 modest positive association between relatively high body burdens of dioxin and 27 increased mortality from all cancers. Positive animal studies and mechanistic data 28 provide additional support for classification of dioxin as a human carcinogen. 29 However, the committee was split on whether the weight of evidence met all the 30 necessary criteria described in the cancer guidelines for classification of dioxin as 31 "carcinogenic to humans." EPA should summarize its rationale for concluding 32 that dioxin satisfies the criteria set out in the most recent cancer guidelines for 33 designation as either "carcinogenic to humans" or "likely to be carcinogenic to 34 humans (NAS, 2006, 198441, p. 140). 35 36 If EPA continues to designate dioxin as "carcinogenic to humans," it should 37 explain whether this conclusion reflects a finding that there is a strong association 38 between dioxin exposure and human cancer or between dioxin exposure and a key 39 precursor event of dioxin's mode of action (presumably AhR binding). If EPA's 40 finding reflects the latter association, EPA should explain why that end point This document is a draftfor review purposes only and does not constitute Agency policy. 5-1 DRAFT--DO NOT CITE OR QUOTE 1 (e.g., AhR binding) represents a "key precursor event (NAS, 2006, 198441, p. 2 141). 3 4 5.1.2. EPA's Response to the NAS Comments on the Qualitative Weight-of-Evidence 5 Carcinogen Classification for TCDD 6 A cancer descriptor is used to express the conclusion of the weight of evidence regarding 7 the carcinogenic hazard potential of a compound. EPA agrees with the NAS committee that 8 cancer descriptors represent points along a continuum of evidence. Relatedly, EPA 9 acknowledges that there are gradations and borderline situations that cannot be communicated 10 through a descriptor and are best clarified by a full weight of evidence narrative. 11 The 2003 Reassessment contains a detailed discussion of TCDD carcinogenicity in both 12 humans (Part II, Chapter 7a; 8) and animals (Part II, Chapter 6; 8) as well as an overall summary 13 of TCDD carcinogenicity (Part III, Chapter 2.2.1). Since the release of the 2003 Reassessment, 14 the database pertaining to TCDD carcinogenicity has been strengthened and expanded by 15 numerous publications (U.S. EPA, 2008, 519261), including a new chronic bioassay in female 16 rats (NTP, 2006, 543749) and several new follow-up epidemiological investigations (see 17 Section 2.4.1 and references therein). Many of these studies have been published subsequent to 18 the NAS review. These new data are summarized and evaluated in Section 2.4 of this document. 19 As noted by the NAS, the 2003 Reassessment was released prior to EPA's publication of 20 the U.S. EPA Guidelinesfor Carcinogen Risk Assessment ("2005 Cancer Guidelines"; U.S. EPA, 21 2005, 086237). Using EPA's guidance at the time of its release (U.S. EPA, 1996, 198087), the 22 2003 Reassessment determined that the available evidence was sufficient to classify TCDD as a 23 "human carcinogen." The 1996 guidance suggested "human carcinogen" to be an appropriate 24 descriptor of carcinogenic potential when there is an absence of conclusive epidemiologic 25 evidence to clearly establish a cause-and-effect relationship between human exposure and 26 cancer, but there are compelling carcinogenicity data in animals and mechanistic information in 27 animals and humans demonstrating similar modes of carcinogenic action. 28 The 2005 Cancer Guidelines (U.S. EPA, 2005, 086237) are intended to promote greater 29 use of the increasing scientific understanding of the mechanisms that underlie the carcinogenic 30 process. The 2005 Cancer Guidelines expand upon earlier guidance applied in the 2003 31 Reassessment and encourage the use of chemical- and site-specific data versus default options, 32 the consideration of mode of action information and understanding of biological changes, fuller This document is a draftfor review purposes only and does not constitute Agency policy. 5-2 DRAFT--DO NOT CITE OR QUOTE 1 characterization of carcinogenic potential, and consideration of differences in susceptibility. The 2 2005 Cancer Guidelines also emphasize the importance of weighing all of the available evidence 3 in reaching conclusions about the human carcinogenic potential of an agent. As noted above, 4 additional information on TCDD carcinogenicity has been published since the release of the 5 2003 Reassessment. This information has expanded the TCDD database and provided additional 6 support for conclusions made in the 2003 Reassessment regarding the carcinogenic potential of 7 TCDD. 8 Under the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237), TCDD is characterized as 9 carcinogenic to humans, based on the available data as of 2009. The 2005 Cancer Guidelines 10 indicate that this descriptor is appropriate when there is convincing epidemiologic evidence of a 11 causal association between human exposure and cancer or when all of the following conditions 12 are met (a) there is strong evidence of an association between human exposure and either cancer 13 or the key precursor events of the agent's mode of action, but not enough for a causal 14 association, and (b) there is extensive evidence of carcinogenicity in animals, and (c) the mode(s) 15 of carcinogenic action and associated key precursor events have been identified in animals, and 16 (d) there is strong evidence that the key precursor events that precede the cancer response in 17 animals are anticipated to occur in humans and progress to tumors, based on available biological 18 information. 19 As noted above, the NAS commented that EPA should .explain whether this 20 conclusion reflects a finding that there is a strong association between dioxin exposure and 21 human cancer or between dioxin exposure and a key precursor event of dioxin's mode of action 22 (presumably AhR binding)" (NAS, 2006, 198441). When evaluating the carcinogenic potential 23 of a compound, EPA employs a weight of evidence approach in which all available information 24 is evaluated and considered in reaching a conclusion. The following sections provide a summary 25 of EPA's weight of evidence evaluation for TCDD. 26 27 5.1.2.1. S u m m a ry E valu ation o f E p idem iologic E vid en ce o f TC D D a n d C an cer 28 The available occupational epidemiologic studies provide convincing evidence of an 29 association between TCDD exposure and all cancer mortality. Among the strongest of these are 30 the studies of over 5,000 U.S. chemical manufacturing workers (the National Institute for 31 Occupational Safety and Health [NIOSH] cohort) (Aylward et al., 1997, 594365; Cheng et al., This document is a draftfor review purposes only and does not constitute Agency policy. 5-3 DRAFT--DO NOT CITE OR QUOTE 1 2006, 523122; C o llin s et al., 2009, 197627; Fingerhut et al., 1991, 197301; Steenland et al., 2 1999, 197437; Steenland et al., 2001, 198589) ; a study o f nearly 2,500 Germ an w orkers involved 3 in the production o f phenoxy herbicides and chlorophenols (the H am burg cohort) (B echer et al., 4 1996, 197121; B e ch er et al., 1998, 197173; F lesch -Jan ys et al., 1995, 197261; F lesch -Jan ys et 5 al., 1998, 197339; M an z et al., 1991, 199061; N agel et al., 1994, 594369) ; a study o f more than 6 2,000 Dutch workers in two plants involved in the synthesis and formulation o f phenoxy 7 herbicides and chlorophenols (the D utch cohort) (B ueno et al., 1993, 196993; H o oiveld et al., 8 1998, 197829) ; a smaller study o f roughly 250 workers involved in a chem ical accident cleanup 9 (the B A S F cohort) ed in a chem ical accident cleanup (the B A S F cohort) (Ott and Zober, 1996, 10 198101; T h iess et al., 1982, 0 64999; Zob er et al., 1990, 197604) ; and an international study o f 11 more than 18,000 w orkers exposed to phenoxy herbicides and chlorophenols (K o g evin as et al., 12 1997, 198598; Saracci et al., 1991, 199190) including new er studies o f sm aller subsets o f these 13 w orkers (M cB rid e, 2009, 198490; M c B rid e et al., 2009, 197296; t' M annetje et al., 2005, 14 197593) . Th e findings from these studies have been thoroughly described either in the 2003 15 R eassessm ent or in Section 2.4.1 o f this document. 16 A s noted in Section 2.4, there are considerable challenges inherent in addressing potential 17 sources o f confounding from sm oking and co-exposure to other carcinogens, (w hich could 18 produce inflated or spurious associations), the healthy w orker effect, (w hich could result in 19 attenuated effects through com parison w ith a referent background w ith an inappropriately high 20 background risk), and quantifying exposure to the populations included in many o f these 21 retrospective studies. Th e m ore recent studies o f these cohorts have made significant advances 22 in reducing the potential for bias from the healthy w orker effect through use o f internal cohort 23 analyses and/or controlling for potential confounders through statistical adjustment, restriction, 24 and use o f internal comparisons. Although some exposure assessment uncertainties remain, 25 some o f these studies have also collected individual-level T C D D exposure estimates that allow 26 quantification o f effective dose necessary for dose-response modeling. Overall, the occupational 27 data provide consistent support for an association between exposure to T C D D and increased 28 cancer mortality. 29 Additional epidemiologic evidence supporting an association between T C D D exposure 30 and cancer com es from studies investigating the m orbidity and mortality o f residents exposed to 31 T C D D follow ing an accidental release from a chem ical plant near Seveso, Italy (the Seveso This document is a draftfor review purposes only and does not constitute Agency policy. 5-4 D R A F T -- D O N O T C I T E O R Q U O T E 1 cohort) (Bertazzi et al., 1989, 197013; Bertazzi et al., 1993, 192445; Bertazzi et al., 1997, 2 197097; Bertazzi et al., 2001, 197005; Consonni et al., 2008, 524825; Pesatori et al., 1998, 3 523076; Pesatori et al., 2003, 197001; Warner et al., 2002, 197489). Pesatori et al. (2003, 4 197001) and Consonni et al. (2008, 524825) were not available at the time the 2003 5 Reassessment was released. Among individuals with relatively high exposure at Seveso 6 (Zones A and B combined), all-cancer mortality in the 20-year post-accident period and all 7 cancer incidence in the 15-year post-accident period failed to exhibit significant departures from 8 the expected 197001). However, an increased risk of all-cancer mortality was noted among men 9 15-20 years after first exposure; not only is the association similar in magnitude to other studies 10 (relative risk [RR] = 1.3; 95% confidence interval [CI] = 1.0-1.7) but also emphasizes the 11 importance of consideration of latency (Bertazzi et al., 2001, 197005). Furthermore, associations 12 between TCDD and some specific cancer sites were detected in this cohort, including increased 13 incidence (based on 15 years of follow-up) and mortality (based on 20 years follow-up) from 14 lymphatic and hematopoietic neoplasms in both males and females from Zones A and B 15 (Consonni et al., 2008, 524825). This excess was primarily due to non-Hodgkin's lymphoma. 16 Additionally, there was an increase in lung and rectal cancer mortality in men (Bertazzi et al., 17 2001, 197005) and limited evidence of increased liver cancer incidence in women based on the 18 15-year follow-up study (Bertazzi et al., 1993, 192445). In a separate analysis of 981 women in 19 Zone A, breast cancer incidence (n = 15) was associated (a 2-fold increase for a 10-fold increase 20 in serum TCDD) with TCDD measurements first collected in 1976 and 1977 (Warner et al., 21 2002, 197489). The authors also reported a 2-3-fold increase in all cancer incidence (n = 21) for 22 the two upper quartiles of TCDD exposure. 23 Overall, the newer studies of the Seveso cohort have reported significant increases in 24 cancer incidence and elevations in cancer mortality that were not evident in earlier studies of this 25 cohort. While these studies demonstrate an association between TCDD exposure and different 26 types of cancer, one of the main limitations is the small number of cancer cases to assess 27 site-specific associations with TCDD exposure. Ongoing studies in that cohort should help 28 further elucidate potential risk for specific cancer types (and other endpoints) associated with 29 TCDD exposures among this population. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 5-5 DRAFT--DO NOT CITE OR QUOTE 1 5.1.2.1.1. E viden ce f o r causality. 2 Th e evidence for causality for cancer from the human studies is briefly sum m arized in the 3 paragraphs that follow and is based on recommendations from the 2005 C an cer G uidelines. It 4 should be noted that there are methodological limitations o f the epidem iologic studies that may 5 temper some o f the conclusions regarding causality. These limitations include limited statistical 6 power, exposure assessment uncertainty, and lack of control of confounders (e.g., dioxin-like 7 compounds and sm oking) in some studies. There also is additional uncertainty in the evidence 8 for causality due to the lack o f organ specificity in T C D D associated cancers, as the most 9 consistent results occurred for all-cancer mortality; however, this would be consistent with a 10 hypothesized carcinogenic mode o f action o f T C D D as a promoter. D espite these uncertainties, 11 m any o f the more recent studies have greatly im proved exposure assessm ents compared to 12 earlier studies o f the same cohorts and have addressed the potential for confounding and other 13 types o f biases. 14 Temporality-- exposure must precede the effect for causal inference. G iv en the long 15 induction period for m any types o f cancers, exposure should precede the effect w ith a sufficient 16 latency (i.e., ty p ically 1 5 -2 0 years for environmental carcinogens). In all the occupational 17 studies review ed (w ith the exception o f (M cB rid e, 2009, 198490)), T C D D exposure has 18 preceded the effect w ith sufficient latency to be considered causally associated. In the studies of 19 the Seveso cohort, the follow -up exposure period has now reached 20 years, a latency sufficient 20 to address some carcinogenic endpoints. Since most o f the studies are based on occupational 21 exposures or accidental releases into the environment, tem porality is more readily established 22 due to the obvious determination o f the specific exposure w indow s prior to disease onset. 23 Strength of Association--refers to the magnitude o f measures o f association such as the 24 ratio o f incidence or mortality (e.g., standardized mortality ratio [SM R s], standardized incidence 25 ratios, R R s, or odds ratios) in addition to statistical significance considerations. E ffect estimates 26 that are large in magnitude are less likely to be due to chance, bias, or confounding. Reports of 27 modest risk, however, do not preclude a causal association and may reflect an agent of lower 28 potency, low er levels o f exposure or attenuation due to nondifferential exposure 29 m isclassification. The four occupational cohorts w ith the highest exposures (N IO S H , Hamburg, 30 Dutch, and B A S F ) consistently showed statistically significant, although moderate, elevations in 31 cancer m ortality. W hen the data w ere combined, the S M R for all four subcohorts w as 1.4 This document is a draftfor review purposes only and does not constitute Agency policy. 5-6 D R A F T -- D O N O T C I T E O R Q U O T E 1 [95% CI = 1.2--1.6] (IARC, 1997, 537123). Based on findings from the International Agency for 2 Research on Cancer (IARC) Working Group, increases in all cancer (combined) mortality of the 3 magnitude reported for TCDD have rarely been found in occupational cohort studies (IARC, 4 1997, 537123). Although these estimates are higher than the all-cancer mortality results among 5 Seveso men (RR = 1.1; 95% CI = 1.0--1.3), they are comparable to the risk estimated in this 6 population (RR = 1.3; 95% CI = 1.0-1.7) 15--20 years after first exposure.31 These consistent 7 results comparable in magnitude from the occupational cohorts and Seveso population are not 8 likely due to chance. 9 The occupational cohort studies also show an increased risk for lung cancer in the 10 previously mentioned four subcohorts. The relative risk for lung cancer in the combined highly 11 exposed subcohorts was estimated to be 1.4 (95% CI = 1.1--1.7) (IARC, 1997). This is 12 consistent with the lung cancer mortality findings for the highest exposed group of men in 13 Seveso (RR = 1.3; 95% CI = 1.0--1.7). Additionally, there was an increase in rectal cancer 14 mortality in the Seveso cohort (RR = 2.4; 95% CI = 1.2--4.6) (Bertazzi et al., 2001, 197005) with 15 a corresponding increase in incidence. Consistent relative risks of more than two were also 16 detected for rectal cancer in the Hamburg and New Zealand cohorts, but increased risks were not 17 found in the other cohorts. Although there was limited evidence of increased incidence or 18 morality from hepatobiliary cancers across the cohorts, liver cancer incidence was elevated in the 19 15-year post accident period among women in the Seveso cohort (RR = 2.4; 95% CI = 1.1--5.1, 20 (Warner et al., 2002, 197489)). An association in this population was also detected for between 21 breast cancer incidence (RR = 2.1; 95% CI = 1.0-4.6) and serum TCDD levels (per a 10-fold 22 increase in serum TCDD). Although findings were based on small numbers, three- and four-fold 23 increased risks of soft tissue sarcoma were detected among the NIOSH (Collins et al., 2009, 24 197627) and New Zealand cohorts (McBride, 2009, 198490). No other cases of this very rare 25 cancer were detected in the exposed populations from the other cohorts. 31I n a d d i t i o n t o c o n s i d e r a t i o n o f s t a t i s t i c a l s i g n i f i c a n c e t o a d d r e s s t h e p o s s i b i l i t y o f r a n d o m v a r i a b i l i t y ( i . e . , c h a n c e ) , m a n y o th e r fa c to rs a re im p o rta n t to c o n s id e r w h e n a sse ssin g c a u sa lity u s in g a w e ig h t o f e v id e n c e d e te rm in a tio n . A s n o te d in th e E P A 's C a n c e r G u id e lin e s , a n u m b e r o f fa c to rs b e s id e s s ta tis tic a l s ig n ific a n c e a re re le v a n t fo r a s s e s s in g e v id e n c e o f a d v e rse h e a lth e ffe c ts b a s e d o n h u m a n d ata. T h e se in c lu d e stre n g th o f a sso c ia tio n , te m p o ra lity , b io lo g ic a l g ra d ie n t (i.e ., d o s e -re s p o n s e c o n c o rd a n c e ), b io lo g ic a l p la u sib ility , etc .). In a n a ly z in g th e b o d y o f in fo rm a tio n in th e lite ra tu re , th e c o n s is te n c y o f th e m a g n itu d e o f re p o rte d ris k e s tim a te s (a c ro s s d iffe re n t s tu d ie s ) is c o n sid e re d w h e n a d d re ssin g c a u sa lity ; ra th e r th a n re ly in g so le ly o n sta tistic a l sig n ific a n c e . This document is a draftfor review purposes only and does not constitute Agency policy. 5-7 DRAFT--DO NOT CITE OR QUOTE 1 Elevated risk o f lymphohemopoietic cancer mortality w as noted among the Seveso cohort 2 ( R R = 1.7; 9 5% C I = 1.2, 2.5) (Consonni et al., 2008, 524825) . Increased S M R s for 3 lymphohemopoietic cancer comparable in magnitude (range: 1.6-2.2) were also detected among 4 the Ham burg and N e w Zealand occupational cohorts, but lim ited evidence (range: 1.0 to 1.2) o f 5 increased m ortality w as found in the B A S F , N IO S H and R an ch H ands em ployees (A khtar et al., 6 2004, 197141; Ott and Zober, 1996, 198101; Steenland et al., 1999, 197437) . M ost o f the 7 lymphohemopoietic cancer mortality risk w as reportedly due to non-Hodgkin' s lymphoma in 8 most o f the cohorts. Relative risks for non-Hodgkin's lymphoma among T C D D exposed 9 populations from the N IO S H , Hamburg, N ew Zealand, Dutch, and Seveso cohorts ranged from 10 1.2 to 3.8. Although statistical power w as lim ited in most o f these studies, relative risks 11 exceeded 3.0 for non-H odgkin' s lym phom a in three o f these cohorts (Consonni et al., 2008, 12 524825; F lesch -Jan ys et al., 1998, 197339; H o oiveld et al., 1998, 197829) . 13 Consistency-- the observation o f the same site-specific effect across several independent 14 study populations strengthens an inference o f causality. D espite differences across occupational 15 cohorts, most studies have consistently reported increases in all-cancer mortality w ith T C D D 16 exposure. Several o f these studies have also reported increases in lung cancer related to T C D D 17 exposure. A s noted above, there is also suggestive evidence o f an increased risk in all-cancer 18 and lung cancer m ortality among the Seveso cohort consistent in magnitude to the occupational 19 cohorts. Elevated risk o f lym phohem opoietic cancer mortality consistent in magnitude 20 (range: 1.6-2.2) w as also detected among the Seveso, Hamburg and N ew Zealand cohorts. A n 21 increased risk for non-H odgkin' s lym phom a w as found in tw o o f the occupational cohorts as 22 w ell as in the Seveso cohort, although the relative risks largely did not achieve statistical 23 significance. Am ong those studies detecting an association, consistent two-fold relative risks 24 w ere found for rectal cancer (B ertazzi et al., 2001, 197005; F lesch -Jan ys et al., 1998, 197339; 25 M cBride, 2009, 198490) and relative risks in excess o f three were detected for soft tissue 26 sarcom a (C o llin s et al., 2009, 197627; M cB rid e , 2009, 198490) . 27 Biological Gradient-- refers to the presence o f a dose-response and/or duration-response 28 between a health outcome and exposure o f interest. Several o f the occupational cohort studies 29 (Flesch -Jan ys et al., 1998, 197339; M an z et al., 1991, 199061; M ich a le k and Pavuk, 2008, 30 199573; Ott and Zober, 1996, 198101; Steenland et al., 1999, 197437) found evidence o f a 31 dose-response relationship for all cancers and various T C D D exposure m easures. Th e S M R This document is a draftfor review purposes only and does not constitute Agency policy. 5-8 D R A F T -- D O N O T C I T E O R Q U O T E 1 analyses based on internal comparisons within the occupational cohorts show a biological 2 gradient by comparing highly T C D D exposed w orkers to low or unexposed w orkers. A 3 biological gradient w as also demonstrated in the Seveso cohort by comparing highly exposed 4 individuals (Z o nes A and B ) to individuals in low er exposure zones (Z ones C and R ) . W arner et 5 al. (2002, 197489) also reported evidence o f a dose-response trend for breast cancer and 6 increasing T C D D exposures. 7 Biological Plausibility-- refers to the observed effect having some biological lin k to the 8 exposure. M ost evidence suggests that toxic effects o f T C D D are mediated by interaction with 9 the aryl hydrocarbon receptor (A h R ). A h R is a highly conserved protein among mammals, 10 including hum ans (F u jii-K u riy a m a et al., 1995, 543727; H arper et al., 2002, 198124; Nebert et 11 al., 1991, 543728) . Several hypothesized modes o f action have been presented for T C D D 12 induced tumors in rodents, all in vo lving A h R activation. The available evidence does not 13 preclude the relevance o f these hypothesized modes o f action to humans. 14 Specificity-- as originally intended, refers to increased inference o f causation i f a single 15 site effect, as opposed to m ultiple effects, is observed and associated w ith exposure. B ased on 16 current biological understanding, this is now considered one o f the w eaker guidelines for 17 causality. A s stated in the 2005 C an cer G uidelines, given the current understanding that m any 18 agents cause cancer at m ultiple sites, and cancers have m ultiple causes, the absence o f specificity 19 does not detract from evidence for a causal effect. G iv e n that the most consistent findings 20 associating T C D D and cancer are for all-cause cancer mortality, epidemiological evidence 21 suggests that T C D D lacks sp ecificity for particular tumor sites. A key event in T C D D ' s mode o f 22 action is binding to and activating A h R ; however, downstream events leading to tumor formation 23 are uncertain and m ay lik e ly be tissue specific. G iv e n that the A h R is highly conserved among 24 species and is expressed in various human tissues, the lack o f tumor site specificity does not 25 preclude a determination o f causality. 26 In summary, E P A finds the available epidem iological information provides strong 27 evidence o f an association between T C D D exposure and human cancer that cannot be reasonably 28 attributed to chance or confounding and other types o f bias, and w ith a demonstration of 29 temporality, strength o f association, consistency, biological plausibility, and a biological 30 gradient. Additional evidence from animal studies and from m echanistic studies (described 31 below ) provides additional support for the classification o f T C D D as carcinogenic to humans. This document is a draftfor review purposes only and does not constitute Agency policy. 5-9 D R A F T -- D O N O T C I T E O R Q U O T E 1 5.1.2.2. S u m m a ry o f E vid en ce f o r TC D D C arcin ogen icity in E x p erim en ta l A n im a ls 2 A n extensive database on the carcinogenicity o f T C D D in experimental animals is 3 described in detail in Part II, Chapter 6 o f the 2003 Reassessm ent. There is substantial evidence 4 that T C D D is carcinogenic in experimental anim als based on long-term bioassays conducted in 5 both sexes o f rats and m ice (K o c ib a et al., 1978, 0 01818; N T P , 1982, 594255; N T P , 2006, 6 543749) and in m ale hamsters (R ao et al., 1988, 199032) . A dditionally, N ational To xico logy 7 Program (N TP , 2006, 543749) has completed a new chronic bioassay in female Sprague D aw ley 8 rats. These studies are summarized in Section 2.4.2 o f this document. A ll studies have produced 9 positive results, w ith T C D D increasing the incidence o f tumors at sites distant from the site o f 10 treatment and at doses w ell b elow the m axim um tolerated dose. In both sexes o f rodents, w hen 11 administered by different routes and at lo w doses, T C D D caused tumors at m ultiple sites; tumors 12 w ere observed in liver, lung, lym phatic system, soft tissue, nasal turbinates, hard palate, thyroid, 13 adrenal, pancreas, and tongue. The most consistent and best characterized carcinogenic 14 responses to T C D D are in the rodent liver, lung, and thyroid (discussed below in 15 Section 5.1.2.3). 16 17 5.1.2.3. TC D D M o d e o f A ction 18 Th e 2005 C an ce r G u idelin es defines the term " mode o f action" as " a sequence o f key 19 events and processes, starting w ith interaction o f an agent w ith a cell, proceeding through 20 operational and anatomical changes, and resulting in cancer formation." A "key event" is an 21 em p irically observable precursor step that is itse lf a necessary element o f the mode o f action or is 22 a biologically based marker for such an element. Mode o f action is contrasted with "mechanism 23 o f action," w h ich im plies a more detailed understanding and description o f events, often at the 24 m olecular level. In the case o f T C D D , the terms `m echanism o f action' and `mode o f action' are 25 often used interchangeably in the scientific literature in reference to T C D D ' s interaction w ith the 26 A h R . A thorough discussion o f T C D D ' s interaction w ith the A h R can be found in the 2003 27 Reassessm ent (Part II, Chapter 2; Part III, Chapter 3), and is summarized below (see 28 Section 5.1.2.3.1). 29 M ost evidence suggests that the majority o f toxic effects o f T C D D are mediated by 30 interaction w ith the A h R . E P A considers interaction w ith the A h R to be a necessary, but not 31 sufficient, event in T C D D carcinogenesis. Th e sequence o f key events follow ing binding o f This document is a draftfor review purposes only and does not constitute Agency policy. 5-10 DRAFT--DO NOT CITE OR QUOTE 1 T C D D to the A h R and that ultimately leads to the development o f cancer is unknown. 2 Therefore, in the strictest sense, T C D D 's interaction with the A h R does not constitute a "mode 3 o f action" as defined by the 2005 Cancer Guidelines because information about the progression 4 o f necessary events is lacking. H ow ever, A h R binding and activation by T C D D is considered to 5 be a key event in T C D D carcinogenesis. 6 7 5.1.2.3.1. The a ry l hydrocarbon receptor (AhR). 8 W hile substantial evidence suggests that most toxic effects o f T C D D are mediated by 9 interaction with the A h R , less is known about the com plex responses that result in tumor 10 formation. Nonetheless, a picture is emerging w herein T C D D is considered a 11 "receptor-mediated carcinogen" in laboratory anim als (see Fig u re 5-1), acting in a manner 12 sim ilar to peroxisom e proliferators, phorbol esters, or estrogen (W oods et al., 2007, 543735) . 13 T C D D activates the A h R , a m em ber o f the b asic helix-loop-helix, Per-A rnt-Sim 14 (b H L H - P A S ) fam ily o f transcription factors. A h R is present in most cell types and in the 15 inactivated state is cytosolic and exists in a com plex w ith chaperone proteins, such as heat shock 16 protein 90 (H sp90). B in d in g o f T C D D to A h R leads to nuclear translocation and 17 heterodim erization w ith its partner protein Arnt, another b H L H - P A S fam ily member. The 18 A h R :A rn t heterodimer binds to specific cognate D N A sequence elements know n as 19 dioxin/xenobiotic response elements ( D R E / X R E ) present in the regulatory region o f specific 20 genes. Binding o f the A h R :A rn t heterodimer to these elements, and subsequent recruitment of 21 tissue specific transcriptional coactivator com plexes, leads to increased transcription o f specific 22 genes, known as "target genes." There is a battery o f genes affected in this manner and targets 23 include certain xenobiotic-m etabolizing enzym es, such as cytochrom e P450 (C Y P )1 A 1 , 24 C Y P 1 A 2 , C Y P 2 B 1 , and UDP-glucuronosyltransferase (U G T )1 A 6 (reviewed in Schwartz and 25 Appel, 2005, 543737) . In addition, genes affected by the T C D D /A h R -co m p le x code for both 26 inhibitory and stimulatory growth factors; their gene products affect cellular growth, 27 differentiation and hom eostasis and have been shown to contribute to carcinogenicity as w ell as 28 other forms o f toxicity (review ed in Popp et al., 2006, 197074) . 29 Detailed m olecular biology research has been performed to identify the extent o f the 30 genes regulated by A h R (W oods et al., 2007, 543735) ; how ever a com plex and still ill-defined 31 profile rem ains. Th e basic physiology o f A h R signaling is still poorly understood, despite being This document is a draftfor review purposes only and does not constitute Agency policy. 5-11 D R A FT-- DO NOT C IT E OR Q U O TE 1 highly conserved among vertebrate species (reviewed in Hahn, 2002, 099302). In fact, it is now 2 known that the AhR recognizes a large number of chemical structures, including nonaromatic 3 and nonhalogenated compounds (Denison and Nagy, 2003, 197226), which supports the 4 biological role of the AhR as a receptor that helps regulate the expression of genes necessary for 5 biotransformation of environmental chemicals (i.e., CYP1A1). However, the endogenous 6 physiological role of AhR is complicated, as evidenced by the numerous studies examining AhR 7 null (ArH -/-) mice, which demonstrate alterations in the liver, immue system, ovary, heart and 8 other organs (reviewed in Hahn, 2009, 477460). The endogenous function of AhR remains 9 unknown. 10 Given that the AhR is expressed in most tissues (Dolwick et al., 1993, 543762) with 11 tissue-specificity in terms of level of expression and the profile of target genes, there is 12 substantial complexity and difficulty associating TCDD-mediated transcription of specific target 13 genes and tissue-specific toxic responses, including cancer. It is important to note that the extent 14 of the response of individual TCDD target genes does not correlate with site-specific 15 tumorigenicity. For example, while TCDD is ineffective as a tumor promoter in ovariectomized 16 rats and does not stimulate liver cell proliferation in these animals, it is still capable of inducing 17 CYP1A2 in roughly the same magnitude as in the intact female rats (Lucier, 1991, 198691). 18 Similarly, CYP1A1 induction by TCDD is very similar in male and female rats even though 19 males are almost completely resistant to TCDD carcinogenicity (Wyde et al., 2002, 197009). 20 Some of AhR's effects on gene expression may be the result of interaction with other 21 transcription factors (such as the retinoblastoma protein(Ge and Elferink, 1998, 197702), NF-k B 22 (Tian et al., 1999, 198378) or with the tyrosine kinase c-Src (Blankenship and Matsumura, 1997, 23 543751) rather than via direct interaction with DNA. By far the most extensive studies involving 24 cross-talk between AhR and another transcription factor are those involving the estrogen receptor 25 alpha (ERa). The anti-estrogenic properties of TCDD have been well-documented, beginning 26 with the observations that TCDD repressed estradiol function in rat uterus and liver. The 27 AhR-ERa cross-talk can be manifested at several levels including direct protein interaction, 28 association of the receptors with the other's response element and altered metabolism of estradiol 29 by AhR ligand (Takemoto et al., 2004, 543753). The interactions between AhR/Arnt- and 30 estrogen receptor-dependent signaling pathways, which mediate anti-estrogenic effects of 31 dioxins and dioxin-like polychlorinated biphenyls (PCBs; Bock, 1994, 543755), is probably This document is a draftfor review purposes only and does not constitute Agency policy. 5-12 DRAFT--DO NOT CITE OR QUOTE 1 causal for the well-documented gender-specificity o f the carcinogenic effects o f these agents 2 (e.g., hepatocarcinogenicity o f T C C D in female as opposed to male rats) (Lucier, 1991, 198691) . 3 In addition, cross-talk between A hR /A rnt and other nuclear receptors, their coactivators, and 4 corepressors, has been described. In fact, cross-talk has been reported for A h R and numerous 5 signaling pathways involved in a broad range of physiological processes. The m olecular 6 mechanism s by w hich the A h R interferes with these signaling networks are multifaceted and 7 occur at m ultiple levels o f regulation (m any beyond transcriptional control) 8 (H aarm ann-Stem m ann et al., 2009, 197874) . It rem ains unknow n how any o f these m olecular 9 pathways in vo lving A h R signaling are linked to T C D D -m ed ia ted carcinogenesis. 10 Pertinent to human risk assessm ent, there are w ide inter- and intraspecies differences in 11 the toxicological responses to T C D D (E m a et al., 1994, 197313; Poland and G lover, 1990, 12 543759; Poland et al., 1994, 198439) some o f w h ich can be explained by polym orphism s in 13 A h R . F o r instance, there is a 10-fold difference in susceptibility to T C D D -in d u c e d toxicity 14 between the T C D D -se n sitiv e C 5 7 B L / 6 and the T C D D -resistan t D B A /2 strains o f m ice (Poland 15 and G lover, 1980, 543761) that can be explained by polym orphic variations in the ligand-binding 16 domain and in the C-term inal region o f the A h R m olecule o f each strain (D o lw ic k et al., 1993, 17 543762) . D epending on the system exam ined, the estimated affinity o f binding o f T C D D (and 18 related com pounds) to the human A h R is about 10-fold low er than that observed to the A h R 19 from " responsive" rodent species and is com parable to that observed to the A h R from 20 "nonresponsive" mouse strains (Ram adoss and Perdew, 2004, 198824) . Th is reduced affinity is 21 due, in part, to a single amino acid substitution w ithin the ligand binding domain o f the human 22 and "nonresponsive" mouse A h R s (Ram adoss and Perdew, 2004, 198824) . Although the affinity 23 o f binding o f T C D D and related compounds to the human A h R is reduced compared w ith rodent 24 A h R s, the qualitative and quantitative rank-order potency o f these chem icals is similar. The 25 considerable tissue and species variab ility in response to T C D D cannot be ascribed solely to 26 polym orphism s o f the A h R gene (G eyer et al., 1997, 543768; Pohjanvirta and Tuom isto, 1994, 27 543767), further complicating this key event in TCD D-m ediated carcinogenesis. 28 29 5.1.2.3.1.1. O th e r A h R con sideration s. 30 In addition to the potent agonist T C D D , there are many other exogenous ligands for the 31 A h R , including certain p o ly cy clic aromatic hydrocarbons, polychlorinated dibenzofurans, and This document is a draftfor review purposes only and does not constitute Agency policy. 5-13 D R A FT-- DO NOT C IT E OR Q U O TE 1 P C B s (Bock, 1994, 543755) . Several natural and endogenous compounds are also regulators of 2 A h R (C hiaro et al., 2008, 543771) . The classes o f endogenous compounds that have been shown 3 to induce C Y P 1 and/or activate A h R include: (a) tryptophan metabolites, other indole-containing 4 m olecules, and phenylethylam ines (G ielen and Nebert, 1971, 543775) ; (b) tetrapyrroles such as 5 bilirubin and biliverdin; (c) sterols such as 7-ketocholesterol and the horse steroid equilenin; 6 (d) fatty acid metabolites, including at least six different prostaglandins (Seidel et al., 2001, 7 543776) and lipoxin A 4; and (e) the ubiquitous second messenger cA M P (reviewed in M cM illan 8 and B rad field (2007, 543777) and B arou ki et al. (2007, 543778)). Several o f these endogenous 9 and exogeous compounds, including bilirubin, biliverdin, and P-naphthoflavone, that also bind to 10 the A h R are not carcinogenic in rodent m odels, therefore, some other key precursor event(s) 11 need to be identified. Further, the existence o f m ultiple ligands w ith varying affinity and 12 responses suggests that " selective receptor m odulators" (or S R M s) o f the A h R exist. S R M s are 13 ligands for a receptor that, upon binding, elicit a conform ational change in the receptor that 14 results in differential recruitm ent o f coregulatory m olecules to the target gene promoter region, 15 thereby im parting a different biological activity relative to the prototypical ligand. T h is 16 phenomenon has been m ost studied for nuclear receptors such as the E R a w ith the classic 17 exam ple being tam oxifen, w h ich has estrogen-like activity in the uterus but anti-estrogen-like 18 effects in the breast. Thus, the relative abilities o f compounds to stimulate gene expression or 19 other effects vary in promoter- and cell type-specific manners. It is now apparent that S R M s 20 exist for the A h R as w ell (S A h R M s , Fretland et al., 2004, 197357) . F o r exam ple, 21 6-m ethyl-1,3,8-trichlorodibenzofuran (6 - M C D F ), a S A h R M w hose structure is sim ilar to that o f 22 T C D D , can induce C Y P 1 A 1 gene expression in liver but does not lead to the toxic responses 23 associated w ith T C D D (F ritz et al., 2009, 594372). The existence o f S A h R M s further 24 complicates the role o f T C D D binding to A h R as a key event in TCD D -m ediated 25 carcinogenicity, and suggests that additional inform ation is necessary to elucidate the 26 carcinogenic mode o f action o f T C D D . 27 T C D D may have dose-dependent modes o f action. It has been demonstrated that 28 A h R -d eficien t (A hR -/-) m ice show no signs o f toxicity at doses o f T C D D approximating the 29 lethal dose eliciting 50% response ( L D 50) dose (200 pg/kg) in A h R +/+ m ice (Fernandez- 30 Salguero et al., 1996, 197650) . H ow ever, a single high exposure o f 2,000 pg/kg to 31 A h R -d eficien t m ice produced several m inor lesions including scattered necrosis and v ascu litis in This document is a draftfor review purposes only and does not constitute Agency policy. 5-14 D R A FT-- DO NOT C IT E OR Q U O TE 1 the liver and lungs. These data suggest that a pathway leading to toxicity exists, albeit at very 2 high doses, that is independent o f the A h R . H ow ever, these data also indicate that, at least in 3 m ice, the major in vivo effects o f T C D D are mediated through the A h R . The finding of 4 carcinogenicity in hamsters (R ao et al., 1988, 199032) is o f special interest since hamsters have 5 been found to be relatively resistant to the lethal effects o f T C D D (H en ck et al., 1981, 543779; 6 O lso n et al., 1980, 197976) . T o date, there have been no chronic bioassay studies o f T C D D 7 carcinogenicity in A hR-deficient transgenic animals. 8 There are additional insights into the com plexity o f T C D D 's m echanism o f action 9 involving A h R . Some biochem ical responses to T C D D treatment in isolated cells have been 10 reported in ce lls lacking Arnt, in cells expressing a mutated A rn t protein and in ce lls w ith highly 11 reduced levels o f A h R (K o llu ri et al., 1999, 548721; Puga et al., 1992, 543784), im p lying either 12 a non nuclear role o f the A h R in mediating these events or an AhR-independent process. 13 A dditionally, recent studies have linked A h R activation in the absence o f exogenous 14 ligand to a multitude o f biological effects, ranging from control o f m am m ary tum origenesis to 15 regulation o f autoimmunity (H ahn et al., 2009, 548725) . F in a lly , constitutively activated A h R in 16 rodents has been shown to induce stomach tumors (A ndersson et al., 2002, 197101) . T h is 17 indicates that A h R activation alone (i.e., in the absence o f ligand) is sufficient to induce tumors. 18 19 5.I.2.3.2. TC D D as a tu m or prom oter. 20 The role of T C D D as a tumor promoter is discussed in the 2003 Reassessm ent (Part II, 21 Chapter 6). Th e follow ing is a b rie f sum m ary o f the inform ation regarding T C D D as a tumor 22 promoter. 23 N um erous studies have exam ined the tumor promoting potential o f T C D D . U sin g the 24 traditional two-stage initiation-promotion study design in the liver, studies have demonstrated 25 that T C D D is a dose- and duration-dependent liver tumor promoter (Dragan and Schrenk, 2000, 26 197243; M aronpot et al., 1993, 198386; Pitot et al., 1980, 197885; Teeguarden et al., 1999, 27 198274; W a lk e r et al., 2000, 198733) (W alk er et al., 1998). T C D D has also tested positive for 28 tumor promoting ability in the two-stage models o f mouse skin tumorigenesis (Dragan and 29 Schrenk, 2000, 197243; I A R C , 1997, 537123), and in the lung (A nderson et al., 1991, 2 01 76 1 ; 30 B eebe et al., 1995, 548754) . O verall, the data demonstrate that T C D D is a tumor promoter and 31 potentially harbors only w eak initiating activity. This document is a draftfor review purposes only and does not constitute Agency policy. 5-15 DRAFT--DO NOT CITE OR QUOTE 1 T C D D is typically designated as a nongenotoxic and nonmutagenic carcinogen because it 2 does not damage D N A directly through the formation of D N A adducts, is negative in most 3 short-term assays for genotoxicity, and is a potent tumor promoter and a w eak initiator or 4 noninitiator in multistage m odels for chem ical carcinogenesis (C la rk et al., 1991, 594378; 5 Flodstrom and Ahlborg, 1991, 548728; Graham et al., 1988, 594375; Lu cie r, 1991, 198691; Pitot 6 et al., 1980, 197885; Poland et al., 1982, 199756) . H ow ever, m echanism s have been proposed 7 that support the possibility that T C D D might be indirectly genotoxic, either through the 8 induction o f oxidative stress or by altering the D N A-dam aging potential o f exogenous and 9 endogenous compounds, such as estrogens. In addition, there have been numerous reports 10 demonstrating T C D D -in d u c e d m odifications o f growth factor signaling pathways and cytokines 11 in experimental anim als and cell culture systems. Som e o f the altered signaling pathways 12 include those for epidermal growth factor, transform ing growth factor alpha, glucocorticoids, 13 estrogen, tumor necrosis factor-alpha, interleukin 1-beta, plasm inogen inactivating factor-2, and 14 gastrin. M an y o f these pathways are involved in cell homeostasis, proliferation, and 15 differentiation and provide plausible m echanism s responsible for the carcinogenic actions o f 16 T C D D . Unfortunately, inform ation on the etiology o f the different tumor types is lackin g to 17 equivocally lin k tumor promotion or indirect genotoxic action o f T C D D to a sp ecific m echanism 18 or mode o f T C D D carcinogenesis. 19 20 5.I.2.3.3. H yp o th esized m odes o f action o f T C D D in rodents. 21 T C D D has been shown to consistently induce m ultiple tumors in both sexes in several 22 rodent species. These tumors are observed in various tissues, including (but not limited to): 23 liver, lung, thyroid, lym phatic system , soft tissue, nasal turbinates, hard palate, adrenal, pancreas, 24 and tongue. W hile the mode o f action o f T C D D in producing cancer has not been elucidated for 25 any tumor type, the best characterized carcinogenic actions o f T C D D are in rodent liver, lung, 26 and thyroid. The hypothesized mode o f action for each o f these three tumor types is briefly 27 discussed below and is described in Figure 5-2. The hypothesized sequence o f events following 28 T C D D interaction with the A h R is m arkedly different for each o f these three tumor types. N o 29 detailed hypothesized mode o f action information exists for any o f the other reported tumor 30 types. Further, no single definitive mode o f action o f TCD D -m ediated carcinogenicity has been 31 identified. This document is a draftfor review purposes only and does not constitute Agency policy. 5-16 DRAFT--DO NOT CITE OR QUOTE 1 5.1.2.3.3.1. L iv e r tu m ors. 2 The mode o f action o f T C D D in producing liver cancer in rodents has not been 3 elucidated. One hypothesized mode o f carcinogenic action o f T C D D in the liver is mediated 4 through hepatotoxicity. G enerically speaking, T C D D activation o f the A h R leads to a variety of 5 changes in gene expression, w hich then lead to hepatotoxicity, follow ed by compensatory 6 regenerative cellular proliferation and subsequent tumor development (see Figure 5-2). The 7 details o f the m echanism o f TC D D -in d u ced hepatotoxicity have not been fully determined but 8 both C Y P induction and oxidative stress have been postulated to be involved (M aronpot et al., 9 1993, 198386; V ilu k se la et al., 2000, 198968) . Th e enhanced cell proliferation arising from 10 either altered gene expression or hepatotoxicity, or both, m ay lead to the promotion o f 11 hepatocellular tumors (W h ysn er and W illia m s, 1996, 197556) . Th e sensitivity o f fem ale rat liv er 12 to T C D D , w h ich apparently does not extend to the m ouse, depends on ovarian horm ones (Lu cier, 13 1991, 198691; W yd e et al., 2001, 198575) . T h is sensitivity has been ascribed to induction o f 14 estradiol m etabolizing enzym es (G raham et al., 1988, 594375) and is hypothesized to lead either 15 to generation o f reactive metabolites o f endogenous estrogen or to active oxygen species o f 16 estrogens. O xidative D N A damage has been im plicated in liv e r tumor promotion (U m em ura et 17 al., 1999, 198001) . 18 A dose-response relationship exists for T C D D -m ed ia ted hepatotoxicity, and this parallels 19 the dose-response relationship for tumor formation (or formation o f foci o f cellu lar alteration as a 20 surrogate o f tumor formation). H ow ever, the dose-response relationship for other 21 T C C D -in d u c e d responses such as enhanced gene expression is different from the dose-response 22 for tumor formation in terms o f both efficacy and potency (see Popp et al. (2006, 197074) for 23 review ). It is important to note that differences in potency between events (i.e., gene expression 24 versus cell proliferation) does not necessary im ply alternative m echanism s o f action. 25 26 5.I.2.3.3.2. L u n g tu m ors. 27 The mode o f action o f T C D D in producing lung cancer in rodents (predominantly 28 keratinizing squamous cell carcinoma, (Larsen, 2006, 548744)) has not been elucidated. One 29 hypothesized mechanism o f the carcinogenic action o f T C D D in the lung involves disruption of 30 retinoid homeostasis in the liver (see Figure 5-2). Retinoic acids and their corresponding nuclear 31 receptors, the retinoic acid receptors (R A R s ) and the retinoid X receptors (R X R s ) , w ork together This document is a draftfor review purposes only and does not constitute Agency policy. 5-17 D R A FT-- DO NOT C IT E OR Q U O TE 1 to regulate cell growth, differentiation, and apoptosis. It is hypothesized that T C D D , through 2 activation of the A h R , can affect parts of the complex retinoid system and/or other signaling 3 systems regulated by, and/or cross-talking with, the retinoid system (reviewed in (N ilsson and 4 H akansson, 2002, 548746)). Th ese effects are then hypothesized to lead to lung tumor 5 development; however, the mechanism s underlying this hypothesis are not well-defined. 6 Pulm onary squamous proliferative lesions have been reported follow ing oral exposure to T C D D 7 in rats (T ritsch er et al., 2000, 197265) . In general, squamous m etaplasia w ith some 8 inflammation is associated with significant forms of injury via inhalation of toxic compounds but 9 is also seen w ith vitam in A deficiency (T ritsch er et al., 2000, 197265) and gives some credence 10 to this hypothesis. 11 Another hypothesized m echanism for the carcinogenic action o f T C D D in the lung is 12 through induction o f m etabolic enzym es. Through activation o f A h R and subsequent induction 13 o f m etabolizing enzym es (such as C Y P 1 A 1 ), T C D D may enhance bioactivation o f other 14 carcinogens in lung (T ritsch er et al., 2000, 197265) . There have been few studies to support this 15 hypothesis; how ever, in a long-term continuous-application study o f carcinogenesis using 16 airborne particulate extract (A P E ) , squamous cell carcinom a occurred in 8 o f 17 AhR+ /+ m ice 17 (4 7 % ) w hile no tumors w ere found in A hR -/- m ice (M atsum oto et al., 2007, 548748) . In 18 addition C Y P 1 A 1 w as induced in AhR+ /+ m ice but not in A hR -/- m ice in this study. These 19 results suggest that A h R plays a significant role in A P E -in d u ce d carcinogenesis in A hR+ /+ m ice 20 and C Y P 1 A 1 activation o f carcinogenic p o lycyclic aromatic hydrocarbons (the primary 21 carcinogenic component o f A P E ) is also o f importance. 22 23 5.I.2.3.3.3. T h y r o id tu m ors. 24 The mode o f action o f T C D D in producing thyroid cancer in rodents has not been 25 elucidated. It is hypothesized that T C D D increases the incidence o f thyroid tumors through an 26 extrathyroidal mechanism (see Figure 5-2). The prevailing hypothesis for the induction of 27 thyroid tumors by T C D D involves the disruption o f thyroid hormone homeostasis via induction 28 o f Phase I I enzym es U G T s in the liv e r (review ed in B ro u w e r et al., 1998, 201801) by an 29 AhR-dependent transcriptional m echanism (B o c k et al., 1998, 548752; Nebert et al., 1990, 30 548756) . Th is induction o f hepatic U G T results in increased conjugation and elimination of 31 thyroxine (T 4 ), leading to reduced serum T 4 concentrations. T 4 synthesis is controlled by the This document is a draftfor review purposes only and does not constitute Agency policy. 5-18 DRAFT--DO NOT CITE OR QUOTE 1 thyroid stimulating hormone (T S H ) w hich is under negative and positive regulation from the 2 hypothalamus, pituitary, and thyroid via thyrotrophin-releasing hormone, T S H , T4, and 3 triiodothyronine. Consequently, the reduced serum T 4 concentrations lead to a decrease in the 4 negative feedback inhibition on the pituitary gland. T h is w ould then lead to a rise in secreted 5 T S H and stimulation o f the thyroid. The persistent induction o f U G T by T C D D and the 6 subsequent prolonged stimulation o f the thyroid could result in thyroid follicular cell hyperplasia 7 and hypertrophy o f the thyroid, thereby increasing the risk o f progression to neoplasia. Increases 8 in blood T S H levels are consistent with prolonged stimulation o f the thyroid and may represent 9 an early stage in the induction o f thyroid tumors identified in animal bioassays. Statistically 10 significant increases in neonatal blood T S H levels have been recently been reported in children 11 born to T C D D -e x p o se d mothers in the Seveso cohort (B a cca relli et al., 2008, 197059, discussed 12 in Section 2 .4.1.1.1.4.4). Support for this hypothesis com es from several studies showing that 13 T C D D decreases serum total thyroxine and free thyroxine concentrations in rats follow ing both 14 single dose and repeated dose exposures (B astom sky, 1977, 548760; B ro u w e r et al., 1998, 15 201 80 1 ; Pohjanvirta et al., 1989, 548766; Potter et al., 1983, 548769; Potter et al., 1986, 548771; 16 Sew all et al., 1995, 198145; V a n B irg elen et al., 1995, 198052) . Further support com es from 17 studies o f transgenic anim als in w h ich T C D D exposure resulted in a m arked reduction o f total 18 thyroxin and free T 4 levels in the serum o f AhR+ /- m ice but not A hR -/- m ice (N ishim ura et al., 19 2005, 197860) . A dditionally, gene expression o f U G T 1 A 6 , C Y P 1 A 1 , and C Y P 1 A 2 in the liv er 20 w as m arkedly induced by T C D D in AhR+/- but not AhR-/- m ice (N ishim ura et al., 2005, 21 197860) . 22 23 5.I.2.3.4. S u m m a ry o f TC D D m ode o f action in rodents. 24 O verall, there are inadequate data to support the conclusion that any o f the particular 25 mode o f action hypotheses described above is operant in TC D D -induced carcinogenesis. 26 How ever, the wealth o f scientific evidence available indicates that most, if not all, o f the 27 biological and toxic effects o f T C D D are mediated by the A h R . Although the receptor may be 28 necessary for the occurrence o f these events, it is not sufficient because other proteins and 29 conditions are known to affect the activity o f the receptor and its ability to alter gene expression 30 or to induce other effects. Certain studies could be interpreted to indicate AhR-independent 31 m echanism s, although these studies have not clearly ruled out involvem ent o f the A h R . The This document is a draftfor review purposes only and does not constitute Agency policy. 5-19 DRAFT--DO NOT CITE OR QUOTE 1 only consistent, but limited, evidence for TC D D -induced effects that do not involve the A h R 2 comes from studies using A hR-deficient transgenic animals. Here however, only m inor effects 3 occurred follow ing treatment w ith extrem ely high doses o f T C D D . Thus, a toxic response to 4 T C D D has A h R interaction as a key event, but there are various species-, cell-, development-, 5 gender-, and disease-dependent differences in the cellular m ilieu that can affect the nature and 6 extent of the response observed. 7 The findings that many AhR-m odulated effects are regulated with distinct specificity 8 supports the understanding that the m olecular and cellular pathways leading to any particular 9 toxic event are extremely complex. Precise dissection o f these events represents a considerable 10 challenge, especially in that a toxic response m ay depend on tim ely m odulation o f several genes 11 rather than o f ju st one particular gene, and possibly m odulation o f these genes in several rather 12 than ju st one cell type or tissue. 13 W h ile a defined m echanism at the m olecular level or a defined mode o f action for 14 T C D D -in d u c e d carcinogenicity is lacking, E P A concludes the follow ing 15 16 interaction with the AhR is a necessary early event in TCDD carcinogenicity in 17 experimental animals. 18 through interaction with the AhR, TCDD modifies one or more of a number of cellular 19 processes, such as induction of enzymes, changes in growth factor and/or hormone 20 regulation, and/or alterations in cellular proliferation and differentiation. 21 AhR activation is anticipated to occur in humans and may progress to tumors. AhR is 22 present in human cells and tissues, studies using human cells are consistent with the 23 hypothesis that the AhR mediates TCDD toxicity and no data exist to suggest that the 24 biological effects of AhR activation by TCDD are precluded in humans. 25 non-AhR mediated carcinogenic effects of TCDD are possible. 26 27 5.1.3. Summary of the Qualitative Weight of Evidence Classification for TCDD 28 Under the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237), TCDD is characterized as 29 carcinogenic to humans, based on the available data as of 2009. This conclusion is based on 30 Multiple occupational epidemiologic studies showing strong evidence of an association 31 between TCDD exposure and increased mortality from all cancers. 32 Epidemiological studies showing an association between TCDD exposure and certain 33 cancers in individuals accidentally exposed to TCDD in Seveso, Italy. This document is a draftfor review purposes only and does not constitute Agency policy. 5-20 DRAFT--DO NOT CITE OR QUOTE 1 Extensive evidence of carcinogenicity at multiple tumor sites in both sexes of multiple 2 species of experimental animals. 3 General scientific consensus that the mode of TCDD's carcinogenic action in animals 4 involves AhR-dependent key precursor events and proceeds through modification of one 5 or more of a number of cellular processes, such as induction of enzymes, changes in 6 growth factor and/or hormone regulation, and/or alterations in cellular proliferation and 7 differentiation. 8 The human AhR and rodent AhR are similar in structure and function and human and 9 rodent tissue and organ cultures respond to TCDD in a similar manner and at similar 10 concentrations. 11 General scientific consensus that AhR activation is anticipated to occur in humans and 12 may progress to cancers. 13 14 5.2. QUANTITATIVE CANCER ASSESSMENT 15 5.2.1. Summary of NAS Comments on Cancer Dose-Response Modeling 16 5.2.1.1. C h oice o f R esp o n se L e v e l a n d C h aracterization o f th e S ta tistica l C on fiden ce A ro u n d 17 L o w D ose M o d el P rediction s 18 The NAS commented on the low dose model predictions in the 2003 Reassessment, 19 including EPA's development of ED01 (effective dose eliciting x percent response) estimates for 20 numerous study/endpoint combinations. The committee also suggested that EPA had not 21 appropriately characterized the statistical confidence around such model predictions in the low- 22 response region of the model. 23 The committee concludes that EPA did not adequately justify the use of the 1% 24 response level (the ED01) as the POD for analyzing epidemiological or animal 25 bioassay data for both cancer and noncancer effects. The committee recommends 26 that EPA more explicitly address the importance of the selection of the POD and 27 its impact on risk estimates by calculating risk estimates using alternative 28 assumptions (e.g., the ED05) (NAS, 2006, 198441, p. 18) 29 30 It is critical that the model used for determining a POD fits the data well, 31 especially at the lower end of the observed responses. Whenever feasible, 32 mechanistic and statistical information should be used to estimate the shape of the 33 dose-response curve at lower doses. At a minimum, EPA should use rigorous 34 statistical methods to assess model fit, and to control and reduce the uncertainty of 35 the POD caused by a poorly fitted model. The overall quality of the study design 36 is also a critical element in deciding which data sets to use for quantitative 37 modeling (NAS, 2006, 198441, p. 18). 38 39 EPA should ... assess goodness-of-fit of dose-response models for data sets and 40 provide both upper and lower bounds on central estimates for all statistical This document is a draftfor review purposes only and does not constitute Agency policy. 5-21 DRAFT--DO NOT CITE OR QUOTE 1 estimates. W hen quantitation is not possible, E P A should clearly state it and 2 explain what w ould be required to achieve quantitation (N A S , 2006, 198441, p. 3 10). 4 5 The N A S also suggested that E P A report information describing the adequacy o f dose- 6 response model fits, particularly in the low-response region. Fo r those cases where biostatistical 7 modeling w as not possible, the N A S recommended that E P A identify the reasons. 8 9 The Reassessm ent should also explicitly address the importance o f statistical 10 assessm ent o f model fit at the low er end and the difficulties in such assessments, 11 particularly w hen using sum m ary data from the literature instead o f the raw data, 12 although estimates o f the im pacts o f different choices o f m odels w ould provide 13 valuable inform ation about the role o f this uncertainty in driving the risk estimates 14 (N A S , 2006, 198441, p. 73). 15 16 5.2.I.2. M o d e l F orm s f o r P red ictin g C an cer R isk s B elo w th e P o in t o f D ep a rtu re (POD) 17 Th e N A S focused m uch o f its review on E P A ' s derivation o f a cancer slope factor. 18 S p ecifically, the N A S commented extensively on the selection o f the appropriate point o f 19 departure (P O D ) and the extrapolation o f dose response m odeling b elow the P O D . 20 The N A S questioned E P A ' s choice of a linear, nonthreshold model for extrapolating risk 21 associated w ith exposure levels below the P O D , concluding that the current scientific evidence 22 w as sufficient to ju stify the use o f nonlinear methods when extrapolating below the P O D for 23 T C D D carcinogenicity. Th e committee further recom m ended that E P A include a nonlinear 24 model for low dose cancer risk estimates as a comparison to the results from the linear model. 25 26 The committee concludes that E P A 's decision to rely solely on a default linear 27 model lacked adequate scientific support. The report recommends that E P A 28 provide risk estimates using both nonlinear and linear methods to extrapolate 29 below P O D s (N A S , 2006, 198441, p. 5). 30 A fter review ing E P A ' s 2003 Reassessm ent and additional scientific data 31 published since com pletion o f the Reassessm ent, the committee unanim ously 32 agreed that the current weight o f scientific evidence on the carcinogenicity of 33 dioxin is adequate to ju stify the use o f nonlinear methods consistent w ith a 34 receptor-mediated response to extrapolate below the P O D . The committee points 35 out that data from N T P released after E P A generated the 2003 Reassessm ent 36 provide the most extensive information collected to date about T C D D 37 carcinogenicity in test animals, and the committee found the N T P results to be This document is a draftfor review purposes only and does not constitute Agency policy. 5-22 DRAFT--DO NOT CITE OR QUOTE 1 compelling. The committee concludes that EPA should reevaluate how it models 2 the dose-response relationships for TCDD... (NAS, 2006, 198441, p. 16). 3 4 Because EPA's assumption of linearity at doses below the 1% excess risk level 5 for carcinogenic effects of TCDD, other dioxins, and DLCs is central to the 6 ultimate determination of regulatory values, it is important to critically address the 7 available scientific evidence on the most plausible shape of the dose-response 8 relationship at doses below the POD (LED01). On the basis of a review of the 9 literature, including the detailed review prepared by EPA and presented in Part II 10 of EPA's Dioxin Risk Assessment and new literature available since the last EPA 11 review, the committee concludes that, although it is not possible to scientifically 12 prove the absence of linearity at low doses, the scientific evidence, based largely 13 on mode of action, is adequate to favor the use of a nonlinear model that would 14 include a threshold response over the use of the default linear assumption (NAS, 15 2006, 198441, p. 122). 16 17 On the whole, the committee concluded that the empirical evidence supports a 18 nonlinear dose-response below the ED01, while acknowledging that the possibility 19 of a linear response cannot be completely ruled out. The Reassessment 20 emphasizes the lack of such nonlinear models, hence its adoption of the approach 21 of linear extrapolation below the POD level. Although this approach remains 22 consistent with the cancer guidelines (U.S. EPA, 2005, 086237; see also 23 Appendix B), EPA should acknowledge the qualitative evidence of nonlinear dose 24 response in a more balanced way, continue to fill in the quantitative data gaps, 25 and look for opportunities to incorporate mechanistic information as it becomes 26 available. The committee recommends adopting both linear and nonlinear 27 methods of risk characterization to account for the uncertainty of dose-response 28 relationship shape below ED01 (NAS, 2006, 198441, p. 72). 29 5.2.2. Overview of EPA Response to NAS Comments on Cancer Dose-Response Modeling 30 EPA agrees with the NAS that the approaches to cancer dose-response modeling for 31 TCDD should be clearly communicated and justified. Furthermore, due to the abundance of new 32 information on TCDD carcinogenicity published since the 2003 Reassessment, EPA has 33 reevaluated the cancer dose-response modeling for TCDD presented in the 2003 Reassessment. 34 As detailed below in Section 5.2.3, EPA has conducted an updated cancer dose-response 35 assessment for TCDD that incorporates key NAS recommendations discussed in this document, 36 reflects the current state-of-the science in cancer dose-response modeling and integrates new 37 TCDD carcinogenic information. Detailed responses to the NAS comments summarized above 38 are found in Section 5.2.3.3. This document is a draftfor review purposes only and does not constitute Agency policy. 5-23 DRAFT--DO NOT CITE OR QUOTE 1 The 2003 Reassessm ent presents an extensive dose-response assessment o f T C D D and 2 provides a comprehensive summary o f dose-response relationships. The analyses and 3 discussions synthesized a considerable breadth o f data and model types, highlighting the 4 strengths and weaknesses o f the then-available scientific information. M odeling included both 5 administered dose and steady state body burden dose m etrics, taking into account variation in 6 half-lives o f T C D D across species. These body burden calculations used a simple one7 compartment kinetic model based on the assumption o f a first-order decrease in the levels of 8 administered dose as a function o f time. A n excess risk o f 1% w as chosen to model the cancer 9 data, but comparative results were also shown for 5% and 10% excess risk (see Table 8-2 o f the 10 2003 Reassessm ent). D o se response w as also explored thoroughly for a number o f in vitro and 11 biochem ical endpoints in addition to the in v iv o data analyses, and ranges o f these values were 12 presented (see Fig u res 8-1, 8-2 and 8-3 o f the 2003 Reassessm ent). Thus, the 2003 13 R eassessm ent provides an initial evaluation o f the carcinogenic database for T C D D and serves as 14 the foundation for the analyses presented below. 15 16 5.2.3. Updated Cancer Dose-Response Modeling for Derivation of Oral Slope Factor 17 Th e follow ing sections describe the dose-response analysis o f the cancer data from 18 epidem iologic cohort studies (see Section 2.4.1 and Tab le 2-4) and rodent bioassays (see 19 Section 2.4.2 and Tab le 2-6), concluding w ith the derivation o f oral slope factors for T C D D 20 based on epidem iologic data (see Section 5.2.3.1) and rodent bioassay data (see Section 5.2.3.2). 21 22 5.2.3.1. D o se-R esp o n se M o d elin g B a se d on E pidem iologic C oh ort D ata 23 Th e 2003 R eassessm ent included dose-response analyses and the development o f oral 24 slope factors from the following three occupational cohorts: the N IO S H cohort, the Hamburg 25 cohort, and the B A S F cohort. In this document, E P A determined that specific studies from each 26 o f these cohorts (B ech er et al., 1998, 197173; Ott and Zober, 1996, 198408; Steenland et al., 27 2001, 198589) met the epidem iologic study inclusion criteria (see Section 2.3.1 and 28 Section 2.4.1). In Section 5.2.3.1.1, the oral slope factors derived from these studies in the 2003 29 Reassessm ent are reviewed. Another study that met the current epidem iologic study inclusion 30 criteria (W arner et al., 2002, 197489) w as also briefly discussed in the 2003 Reassessm ent, but 31 an oral slope factor w as not derived from that study. In Section 5.2.3.1.2.2, E P A discusses its This document is a draftfor review purposes only and does not constitute Agency policy. 5-24 DRAFT--DO NOT CITE OR QUOTE 1 unsuccessful attempt to use the categorical results published by (W arner et al., 2002, 197489) to 2 develop an oral cancer risk estimate. 3 Since the publication o f the 2003 Reassessm ent, additional cancer epidem iologic studies 4 based on these cohorts have been published in the peer-reviewed literature. O f these, C o llin s et 5 al. (2009, 197627) and Cheng et al. (2006, 523122) met the epidem iologic study inclusion 6 criteria (see Section 2.3.1 and Section 2.4.1). In Section 5.2.3.1.2, E P A evaluates the suitability 7 o f deriving an oral slope factor from the Cheng et al. (2006, 523122) study and derives oral slope 8 factor estimates. Although the C o llin s et al. (2009, 197627) study met the study inclusion 9 criteria, E P A could not derive an oral slope factor from that study. In Section 5.2.3.1.2.3, E P A 10 discusses w hy an oral cancer risk estimate w as not developed using the positive results for the 11 soft-tissue sarcom a mortality published by C o llin s et al. (2009, 197627) . 12 13 5.2.3.1.1. E valu ation o f E pidem iologic S tu d ies U sed in th e 2003 R ea ssessm en tf o r O S F 14 D erivation . 15 In the 2003 Reassessm ent, E P A reported dose-response m odeling results for three 16 epidem iologic studies o f human occupational cohorts: the N IO S H cohort w ith data published by 17 Steenland et al. (2001, 198589) ; the Ham burg cohort w ith data published by B e ch er et al. (1998, 18 197173) ; and the B A S F cohort w ith data published by Ott and Zob er (1996, 198408) . E a c h o f 19 these studies is sum m arized in Section 2.4.1 o f this document and in the 2003 Reassessm ent 20 (Part II, Chapter 8; Part III, Chapter 5). Furthermore, E P A has evaluated the suitability o f these 21 studies for use in T C D D dose-response m odeling and concluded that each o f these studies meet 22 the inclusion criteria for epidemiology studies presented in Section 2.3.1. 23 E a c h o f these studies reports all cancer mortality as an outcome. Steenland et al. (2001, 24 198589) and B e ch er et al. (1998, 197173) analyzed cohorts o f prim arily m ale w orkers w ho 25 experienced occupational exposures to T C D D over long periods o f time, w hile Ott and Zober 26 (1996, 198408) studied a cohort o f prim arily m ale w orkers w ho w ere exposed to high T C D D 27 concentrations at a single point in time due to an industrial accident. 28 The authors o f all three of these studies measured, and then back-extrapolated, T C D D 29 levels in a subset o f w orkers to estimate exposures during employment and then the authors used 30 this information to estimate exposures in the remainder o f the cohort. These measured T C D D 31 sam ples generally w ere collected decades after the last know n occupational exposure. In each This document is a draftfor review purposes only and does not constitute Agency policy. 5-25 DRAFT--DO NOT CITE OR QUOTE 1 study, the authors relied on T C D D measures in the cohort to back-calculate serum lipid or body 2 fat levels o f T C D D using a simple one-compartment kinetic model based on the assumption o f a 3 first-order decrease in the levels o f exposure dose as a function o f time. The assumed half-life of 4 T C D D used in the models varied from study to study. None o f the studies sampled T C D D levels 5 from the entire cohort; for example, Ott and Zober collected samples from 138/243 workers 6 (57% o f the cohort), w hich w as the highest percentage o f w orkers sampled among the three 7 studies. Steenland et al. (2001, 198589) and B e ch e r et al. (1998, 197173) used the m easured and 8 back-extrapolated T C D D concentrations to estimate the exposures that were associated with 9 various occupations w ithin the cohort, and subsequently used this information to develop 10 exposure m atrices (i.e., the T C D D load per unit time for an occupation) that then could be used 11 to estimate the cum ulative dioxin dose for each cohort member. Ott and Zob er (1996, 198408) 12 used regression procedures w ith data on time spent at various occupational tasks to estimate 13 T C D D levels for all mem bers o f the cohort. Fo llo w in g the estimation o f w orker exposures in 14 each o f these three studies, the studies' authors divided these cohorts into exposure subgroups 15 based on the estimated T C D D levels. 16 In the 2003 Reassessm ent, E P A identified a P O D based on a 1% response in cancer 17 m ortality ( E D 01) for the Steenland et al. (2001, 198589) , and the Ott and Zob er (1996, 198408) 18 studies. E P A extrapolated from this P O D to low er doses using a straight line drawn from the 19 P O D to the origin-- zero increm ental dose, zero increm ental response-- to give a probability o f 20 extra risk. Because there w as insufficient evidence to support an assumption o f nonlinearity, 21 E P A chose to develop these m odels using a linear model. 22 23 5.2.3.1.1.1. S te e n la n d e t al. (2 0 0 1 ,198 5 8 9 ). 24 Steenland et al. (2001, 198589) developed dose-response m odels based on T C D D 25 exposures and all cancer mortalities from eight plants in the N IO S H cohort (see Section 26 2.4.1.1.1.1.3 for study details). Serum lipid levels o f T C D D in 1988 were measured in 27 193 w orkers at one o f these plants. Steenland and coauthors relied on a first-order kinetic model 28 (assum ing a constant 8.7 year half-life) to back-extrapolate to serum T C D D levels at the time o f 29 the last occupational exposure. The study authors assigned exposure estimates to each o f the 30 3,538 w orkers in the cohort based on a job-exposure matrix. T h is m atrix w as based on (1) an 31 estimated level o f contact w ith T C D D , (2) the degree o f T C D D contamination o f the products This document is a draftfor review purposes only and does not constitute Agency policy. 5-26 DRAFT--DO NOT CITE OR QUOTE 1 the workers produced, and (3) the fraction o f a workday during w hich the worker likely 2 contacted the TCD D-contam inated products. They then estimated each w orker's serum T C D D 3 levels as an area under the concentration curve (A U C ) for lipid-adjusted serum levels over time. 4 The mortality analysis w as conducted on 256 cancer decedents. 5 Several different dose-response models were fit to these data to provide estimates o f fatal 6 cancer risk. The best-fitting model w as a C o x regression exposure-response model using the 7 lo g (A U C ) o f T C D D lip id concentration (ppt-year) lagged by 15 years as the exposure metric. 8 Steenland and colleagues also developed a piecewise linear regression model with no lag, in 9 w hich two separate linear slopes were estimated. Th is analysis assumed a background exposure 10 o f 0.5 pg/kg-day. Th e lipid concentrations w ere converted to body burdens by dividing by 4. 11 Th e central tendency estimate and low er bound E D 01s from the piecew ise linear model and their 12 associated cancer slope factors for the most sensitive endpoint (m ale cancer m ortality) are 13 presented in Table 5-1. 14 15 5.2.3.1.1.2. B e c h e r e t al. (1 9 9 8 .1 9 7 1 7 3 ). 16 B ased on the H am burg cohort, B e ch er et al. (1998, 197173) reported a dose-response 17 analysis for all fatal cancers com bined (see Section 2 .4 .1.1.1 .3.4 for study details). Th e m ortality 18 analysis w as conducted in 1992 on 124 cancer decedents. Th e exposure variable in the study 19 w as the integrated blood levels for T C D D concentration over time ( A U C , ng/kg-years), as 20 estimated by F lesch -Jan ys et al. (1998, 197339) ; these w ere converted to body burdens by 21 dividing by 4. Estim ates o f the half-life o f T C D D , based on the sample o f 48 individuals w ith 22 repeated measures, w ere incorporated into a model that back-extrapolated T C D D exposures to 23 the end o f the em ploym ent after accounting for the w o rk ers' ages and body fat percentages. 24 These estimated exposure measures were then applied to the entire cohort, w hich consisted o f all 25 1,189 regular m ale em ployees w ho w ere em ployed for at least 3 months between 1952 and 1984 26 at the Boehringer chem ical plant in Ham burg, Germ any. 27 B e ch er et al. (1998, 197173) used a C o x regression approach for the dose-response 28 modeling and developed three models: a m ultiplicative model, an additive model, and a power This document is a draftfor review purposes only and does not constitute Agency policy. 5-27 DRAFT--DO NOT CITE OR QUOTE 1 model.32 The response variable in each model was the SMR for total cancer mortality. The 2 models were calculated with lag times of up to 20 years. The multiplicative model provided the 3 best fit; however, the study authors judged the fits for all three models to be acceptable. The 4 model results were used to calculate unit risk estimates derived as the risk of cancer death 5 through age 70 given a daily dose of 1 pg/kg body weight of TCDD minus the risk given no 6 exposure to TCDD. These calculations were based on background German cancer mortality 7 rates. The model results were used to calculate cancer risk estimates. The lower bound 8 estimates on the dose were not available for models published by Becher et al. due to the absence 9 of statistical parameter measures. The central tendency estimate ED01s from the three statistical 10 models and their associated cancer slope factors are presented in Table 5-1. 11 12 5.2.3.1.1.3. O tt a n d Z o b e r (1 9 9 6 ,1 9 8 4 0 8 ). 13 In the 2003 Reassessment, EPA also developed a dose-response analysis based on a study 14 reported by Ott and Zober (1996, 198408) for cancer incidence and mortality experienced by 15 243 men, who were exposed to TCDD in 1953 during an accident at the BASF plant in Germany 16 (see Section 2.4.1.2.1.2.1 for study details). The cohort was followed through 1992. TCDD 17 blood lipid levels were available for 138 of these men 30 years after the accident. These levels 18 were back-extrapolated and used to estimate the AUC for TCDD. Body burdens (ng/kg) were 19 estimated by dividing AUC by 4, and steady-state body burdens were estimated assuming a 20 constant half-life of approximately 7.1 years.33 Ott and Zober (1996, 198408) used Cox 21 proportional hazard approaches to estimate both cancer incidence and cancer mortality risk per 32The " m ultip licative m odel" set relative risk (R R ) equal to exp(ftd), where the dose d is the A U C . The " additive m odel" set R R = 1+ftd, and the " power m odel" set R R = exp(ft log (kd+1)). The values ft and k are estimated p aram eters. 33Based on the in itia l body burden (B 0) E P A estimated the body burden at tim e t using the fo llo w in g form ula: B (t) = B0e ket , where k eis an elim in ation constant equal to ln (2 )/(h a lf-life in years). This im plies that the A U C at tim e T after in itia l exposure is AUC --B0 (1 - e-keT ) . T in this case was 39 years (tim e fro m the accident in 1953 to e the follo w -u p in 1992). D ivid ing by a life tim e o f 71 years (mean age in 1954, 33 years, plus 38 years fro m 1954 to the follo w u p in 1992) yields the life tim e mean body burden as: B mean= B (1 - 71ke e keT) . In the 2003 Reassessment, E P A converted the steady-state body burden to u n its o f equivalent in itia l dose by d ivid ing by the constant 1 71A (1- e keT) . W ith the given values fo r h a lf-life and T , that constant is 0.1411 and 1/(the constant) is 7.09. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-28 DRAFT--DO NOT CITE OR QUOTE 1 unit TCDD dose.34 Ott and Zober reported conditional risk ratios for cancer mortality that were 2 slightly larger than the conditional risk ratios for cancer incidence, which is counter-intuitive. 3 The risk of cancer mortality would be expected to be greater than the risk of cancer incidence. 4 The conditional risk ratio (and 95%CI) for all cancer mortality (1.22; 1.00-1.50) exceeded the 5 conditional risk ratio for all incident cancer cases (1.11; 0.91-1.35). Similarly, the conditional 6 risk ratios for digestive cancer mortality (1.46; 1.13-1.89) and respiratory cancer mortality (1.09; 7 0.70-1.68) were also both larger than the conditional risk ratios for all digestive cancers (1.39; 8 1.07-1.69) and all respiratory cancers (1.02; 0.65-1.59). As expected, in this cohort, incident 9 cases exceeded cancer mortality for total cancers (47 vs. 31), digestive cancers (12 vs. 11) and 10 respiratory cancers (13 vs. 11). Ott and Zober also reported that conditional risks for mortality 11 for all cancer and lung cancer associated with cigarette smoking were also higher than the 12 respective incidence risks. In their Cox regression analysis, Becher et al. (1998, 197173) also 13 report that the regression coefficient for total cancer mortality (0.0096) was slightly larger than 14 the regression coefficient for total cancer incidence (0.0089). The finding of Ott and Zober and 15 Becher et al. that the risk of cancer mortality is greater than cancer incidence is possibly due to a 16 systematic difference in the reference population for incidence vs. the reference population for 17 mortality. The central tendency estimate and lower bound ED01s from the modeling and their 18 associated cancer slope factors are presented in Table 5-1. 19 20 5.2.3.I.2. E valu ation o f O th er E pid em io lo g ic S tu d ies C on sideredf o r O S F D erivation. 21 Three additional epidemiologic studies that met the study inclusion criteria (see 22 Section 2.3) for use in dose response modeling as set forth in this document are evaluated in this 23 section for the estimation of cancer risk estimates. These studies were either published after the 24 Reassessment (Cheng et al. (2006, 523122) and Collins et al., (1996, 197637)), or not used to 25 derive an OSF in the Reassessment (Warner et al., 2002, 197489). Each study is summarized in 26 Section 2.4.1. 27 34The model from Ott and Zober has risk proportional to e^'//hse with fi = ln(1.22). The corresponding slope for the mean (steady-state) body burden is 7.0851 * log(1.22) * 0.001 (the 0.001 converts nanograms to micrograms). This document is a draftfor review purposes only and does not constitute Agency policy. 5-29 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.1.2.1. C h e n s e t al. (2006, 5 2 3 1 2 2 ). 2 As discussed in Section 2.4.1.1.1.1.4, Cheng et al. (2006, 523122)) analyzed the 3 relationship between TCDD dose and all cancer mortality for the same subset of NIOSH workers 4 as analyzed previously by Steenland et al. (2001, 198589). In contrast to Steenland et al., Cheng 5 et al. (2006, 523122) used the "concentration- and age-dependent elimination model" 6 (concentration- and age-dependent elimination [CADM], discussed in Section 3.3; see also 7 Aylward et al. (2005, 197114)), rather than a constant 8.7-year half-life, and calculated serum8 derived TCDD estimates for use in dose-response analysis. An important feature of CADM is 9 that it incorporates concentration- and age-dependent elimination of TCDD from the body, 10 meaning that the effective half-life of TCDD elimination varies based on exposure history, body 11 burden, and age of the exposed individuals. As discussed in Section 3.3, the use of the CADM 12 model to simulate TCDD kinetics in the NIOSH cohort results in time-integrated body burden 13 estimates four to five times greater than those obtained with the simple first-order model, with 14 smaller differences between the two methods at lower exposures. 15 Following the estimation of dose using the CADM-derived AUC values, Cheng and 16 colleagues (Cheng et al., (2006, 523122); the "Cheng analysis") derived dose-response estimates 17 for the NIOSH cohort using linear Cox regression for both lagged and un-lagged exposure on 18 various subsets of the data (high-exposures trimmed). The results for the lagged-exposure 19 analysis are summarized in Table 5-2. For comparison, the Cox regression coefficient from the 20 analysis conducted by Steenland et al. (2001, 198589), which relied on a first-order elimination 21 rate model assuming a constant 8.7-year half-life, is also shown in the table. As in Steenland et 22 al. (2001, 198589),35 Cheng et al. (2006, 523122) found a much stronger relationship between 23 cancer mortality and exposure metrics lagged 15 years compared to the relationships for 24 unlagged exposures. Cheng et al. (2006, 523122) also noted that the dose-response relationship 25 plateaued above the 95th percentile of exposure. For exposures lagged 15 years, the regression 26 coefficient (P) of the linear slope derived by Cheng et al. (2006, 523122) was 3.3 x 10-6 per 27 ppt-year lipid-adjusted serum TCDD, with a standard error of 1.4 x 10-6 (Table III of Cheng et 28 al. (2006, 523122)). The upper 5% of the exposure range (individuals >252,950 ppt-year lipid 29 adjusted serum TCDD) was excluded in estimating this slope. Because this exclusion reduces 35 L a g g e d e x p o su re s m o d e le d o n ly fo r lo g -tr a n s fo r m e d se ru m c o n c e n tra tio n s, n o t fo r u n tra n sfo rm e d se ru m concentrations in the piece-w ise linear m odel. This document is a draftfor review purposes only and does not constitute Agency policy. 5-30 DRAFT--DO NOT CITE OR QUOTE 1 the upper portion of the response where the slope is shallow36, this likely better represents the 2 slope in the region of the curve where the fatal cancer risk is increasing with dose, which is the 3 equivalent of dropping the highest dose in an animal bioassay or using a piece-wise linear model 4 as in Steenland et al. (2001, 198589). 5 To develop cancer risks for TCDD, EPA used the modeling results of the Cheng analysis, 6 with conversion to oral intake using the Emond human PBPK model as follows. The slope (ft) 7 from the Cheng analysis is the slope of the linear relationship between the natural logarithm of 8 the rate ratio (RR) and the cumulative fat TCDD concentration (fat-AUC). Conceptually, the 9 slope (P) is similar to an OSF, except that it is expressed in terms of fat-AUC rather than intake. 10 Also, the slope represents the incremental increase in cancer mortality (expressed as an RR) 11 above the background TCDD exposure experienced by the NIOSH cohort rather than above zero. 12 Using the upper 95% bound on ft and assuming that the slope is the same below the NIOSH 13 cohort background exposure level (approximately 5 ppt/yr TCDD fat concentration), EPA 14 calculated risk-specific doses (as daily oral intakes) for TCDD for risk levels of concern to EPA. 15 The risk-specific doses were estimated from the Emond human PBPK model for the lifetime- 16 average TCDD fat concentrations corresponding to the fat-AUC predicted by the Cheng et al. 17 model for each of the risk levels of concern. The steps in this computation are as follows: 18 19 Background cancer mortality risk estimate (R0). EPA used an R0 of 0.112 as reported by 20 Cheng et al. (2006, 523122)37 21 Total cancer mortality risk in the exposed group associated with a specified (extra) risk 22 level (RL) of fatal cancer (TRRl). A TRRl, associated with any given extra risk level (e.g., 23 0.01, 1 x 10"6) can be calculated using the following relationship for extra risk: 24 25 ER TRrl - R0 (Eq. 5-1) 1- R0 26 36 Steenland et al. (2001, 198589) ; Steenland and Deddens (2003, 198587) found a slig htly negative slope fo r the higher exposures, stating that the phenomenon could be a result o f exposure m isclassification, depletion o f susceptible individuals or saturation o f receptor-mediated processes. 37 In Table IV , Cheng et al. (2006, 523122) report tw o estimates o f background fa ta l cancer risk, R 0, fo r males aged 75 years: 0.112 and 0.124. A R 0 estimate o f 12.4% was used by Steenland et al. (2001, 198589), and 11.2%, as estimated fo r a ll males in the 1999-2001 Surveillance Epidem iology and End R esult data set. E P A chose to use the more recent estimate o f 11.2% fo r the purpose o f predicting risk in the current U .S. population. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-31 DRAFT--DO NOT CITE OR QUOTE 1 Incremental cancer mortality risk in the exposed population based on a given extra risk 2 (RD). R d, is calculated as the difference between the total risk and background risk and 3 expressed in terms of R L and R 0 by combining Equations 5-2 and 5-1. 4 5 Rd = TRrl - Ro (Eq. 5-2) 6 7 Rd = RL x (1 - Ro) (Eq. 5-3) 8 9 Cumulative TCDD concentration in the fat compartment for a given extra risk (AUCRL). 10 AUCrl is then calculated by taking the natural logarithm of Equation 3 from Cheng et al. 11 (2006, 523122), rearranging and substituting for RR38 (RR = [RD + R0]/R0): 12 13 AUCrl = ln((Rb + Ro)/Ro)/p* (Eq. 5-4) 14 15 where fi* is the central-tendency regression slope or the 95% upper bound (figs) 16 determined by summing the regression coefficient (P) and the product of 1.96 and the 17 standard error of the regression coefficient, yielding an estimate of 6.0 x 10-6 per 18 ppt-year lipid adjusted serum TCDD, as follows: 19 20 P95 = fi +1.96* SE (Eq. 5-5) 21 22 Continuous daily TCDD intake associated with a given extra risk [DRL]. Because the fat 23 concentrations generated by CADM are not linear with oral exposure at higher doses, a 24 single oral slope factor to be used for all risk levels cannot be obtained; the response is 25 approximately linear with fat concentrations and oral intake at lower doses. Instead, a 26 risk-specific DRLmust be estimated by converting the respective AUCRLto the 27 corresponding lifetime daily intake, using an appropriate human toxicokinetic model. 28 EPA has chosen to use the Emond human PBPK model for this purpose because the 29 CADM configuration does not facilitate this process and so that the dose conversions are 30 consistent with those used in the derivation of the RfD. A DRLis obtained from the 31 Emond model by finding the average lifetime daily intake corresponding to the AUCRLin 32 the fat compartment.39 33 34 Note that there are two nonlinear steps in the estimation of risk-specific doses from the 35 Cheng et al. model. First, fat-AUC (AUCRL) and the incremental cancer mortality risk (RD) do 38 A s d e fin e d b y C h e n g et a l. (2 0 0 6 , 5 2 3 1 2 2 , p . 1063). 39 A lth o u g h th e N I O S H co h o rt e x p o su re s are rep o rted as L A S C , th e y are tre ated as fa t co n c e n tra tio n s in th e C h e n g analysis because fat in all tissues are m odeled as one com partm ent (hence equal) in C A D M . The translation to oral intake in the Em ond m odel is from the fat com partm ent, rather than the serum com partm ent, even though the serum and fat com partm ents are not equivalent, because the regression slope (P) in the C heng analysis is in term s o f the (equivalent) fat com partm ent. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-32 DRAFT--DO NOT CITE OR QUOTE 1 not have a linear relationship (see Eq. 5-5); however, the relationship becomes virtually linear 2 below an incremental risk of 10-3 (see Table 5-3). Second, TCDD fat concentration is not linear 3 with oral intake in the Emond human PBPK model (see Section 3); this relationship also is close 4 to linear below the 10-5 risk level. The resulting predicted cancer-mortality risk is approximately 5 linear with daily oral intake at low doses. Table 5-3 shows the AUCrl and DRLbased on the 6 95% upper-bound regression slope (fi95) from the Cheng analysis for a number of risk levels of 7 interest to the EPA. For comparative purposes, EPA has also shown the equivalent oral slope 8 factor (RL ^ Drl) for those same risk levels. Table 5-4 also shows analogous results based on 9 the best estimate of regression coefficient (fi = 3.3 x 10-6) for total fatal cancers from the Cheng 10 analysis. 11 12 5.2.3.I.2.2. Warner et al. (2002.197489). 13 Warner et al.(2002, 197489) is a study of 981 females exposed to elevated TCDD levels 14 following the Seveso accident of 1976 (see Section 2.4.1.1.1.4.2 for study details). The TCDD 15 exposure pattern involving a single period of elevated TCDD exposures followed by an extended 16 period of lower ambient level TCDD exposures and elimination is similar to that of the BASF 17 cohort (Ott and Zober, 1996, 198408). TCDD levels, measured or estimated in blood lipids 18 shortly after the time of the accident, were available for all women. These women were divided 19 into four exposure groups of <20, 20-44, 44.1-100, and >100 ppt. In this cohort, 21 total 20 cancers have been observed; 15 of these were breast cancer cases and 3 were thyroid cancer 21 cases. Cox proportional hazards modeling showed that the hazard ratio for breast cancer 22 associated with a 10-fold increase in serum TCDD levels (log10 (TCDD)) was significantly 23 increased to 2.1 (95% CI = 1.0-4.6). Rate ratios (95% CI) for cancer incidence in these 4 groups 24 were 1.0, 1.0 (0.2-5.5), 2.2 (0.5-10.8) and 2.5 (0.5-11.8). Using a Cox proportional hazards 25 model and assuming continuous exposure, the rate ratio was 1.7 (0.9-3.4) for each 10-fold 26 increase in serum TCDD; that is, a log10transformation of the exposure estimates in their 27 analysis was presented. They reported a test for trend ofp = 0.09. 28 EPA attempted to estimate an ED01from the modeled results of Warner et al. (2002, 29 197489) from the statistically significant hazard ratio for breast cancer. However, EPA had to 30 estimate the slope of the tangent to the log-linear relationship. Because the exponentiated slope 31 of a log-dose linear relationship is not constant but varies with dose, and because the lowest This document is a draftfo r review purposes only and does not constitute Agency policy. 5-33 DRAFT--DO NOT CITE OR QUOTE 1 exposure measure w as w ell-above the 1% response region o f interest, E P A could not confidently 2 estimate the tangent to the log-dose linear relationship. Th e transformation o f the lo g i0 dose 3 units to linear units of T C D D yielded an im plausibly low ED01 and correspondingly high cancer 4 risk that w as inconsistent w ith a visual inspection o f the untransformed plot. E P A w as not 5 confident in these values for health risk assessment because of uncertainties in the transformation 6 in the lo w response region o f the original model. Thus, E P A did not derive an E D 01 or oral slope 7 factor for this study. 8 9 5.2.3.I.2.3. C o llin s e t al. (2 0 0 9 .1 9 76 2 7). 10 C o llin s et al. (2009, 197627) investigated the relationship between serum T C D D levels 11 and mortality rates in the N IO S H cohort (see Section 2.4.1.1.1.1.5 for study details). The 12 investigators completed an extensive dioxin serum evaluation of w orkers em ployed by the D o w 13 C h em ical plant in M idland, M ichig an, that made 2,4,5-trichlorophenol ( T C P ) from 1942 to 1979 14 and 2 ,4 ,5 -T from 1948 to 1982. C o llin s et al. (2009, 197627) developed historical T C D D 15 exposure estimates for all 1,615 w orkers using serum sam ples from 280 form er w orkers that 16 w ere collected during 2 0 0 4 -2 0 0 5 . Investigators calculated a cum ulative m easure o f exposure 17 using a sim ple one-compartment first-order pharm acokinetic model and elim ination rates as 18 estimated from the B A S F cohort (Flesch -Jan ys et al., 1996, 197351) . Th e follow -up interval for 19 these w orkers covered the period between 1942 and 2003. Thus, the study included 10 more 20 years o f follow-up than earlier investigations o f the entire N IO S H cohort. A key limitation of 21 this study is that the derivation o f the S M R s and slope parameters did not include a lag period, 22 u nlike other analyses o f the N IO S H cohort (e.g., Cheng et al., 2006, 523122; Steenland et al., 23 2001, 198589) . 24 Although results were largely negative, statistically significant mortality in the cohort 25 w as found for soft-tissue sarcoma (S M R = 4.1, 95% C I = 1.1-10.5), based on only four deaths. 26 A regression coefficient o f 0.05872 (standard error not reported), and a hazard ratio o f 1.060 27 (9 5 % C I = 1.017 to 1.106) w ere reported by C o llin s et al. (2009, 197627) for soft-tissue sarcoma. 28 Although it met the dose-response study criteria, E P A could not calculate an upper bound on the 29 regression coefficient because the standard error w as not given. In addition, E P A w as unable to 30 estimate an extra-risk value because the reference population response w as not specified. Thus, 31 E P A did not derive an E D 01 or oral slope factor for this study. This document is a draftfor review purposes only and does not constitute Agency policy. 5-34 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.2. D o se-R esp o n se M o d elin g B a se d on A n im a l B io a ssa y D ata 2 Figure 5-3 provides a summary of the process EPA has utilized to select and identify 3 candidate TCDD OSFs from key animal bioassays that were identified in Section 2.4.3 of this 4 document. For each in vivo animal cancer study that qualified for TCDD dose-response 5 assessment using the study inclusion criteria specified in Section 2.3.2, EPA first selected the 6 species/sex/tumor data set combinations that had been characterized as having statistically 7 significant increases in tumor incidence by either a pair-wise test between the treated group and 8 the controls or by a trend test showing increases in tumors with increases in dose. Next, EPA 9 used the Emond animal kinetic model discussed in Section 3 to estimate blood concentrations 10 corresponding to each study's average daily administered doses for use in dose response 11 modeling. Benchmark dose lower confidence bounds (BMDL01s) were then estimated for the 12 blood concentrations by (1) using the multistage cancer model for each species/sex/tumor 13 combination within each study and (2) using a Bayesian Markov Chain Monte Carlo framework 14 that assumes independence of tumors, modeling all tumors together for each species/sex 15 combination within each study. The final selected models were subjected to goodness-of-fit tests 16 and visual inspections of fit to the raw data. Thus, for each sex/species combination within each 17 study, this process generated a BMDL01 for each single tumor type and another BMDL01 for the 18 combined tumors. Finally, using the Emond human kinetic model discussed in Section 3, human 19 equivalent doses (BMDLHEDs) were then estimated for each of the BMDL01s and, using a linear 20 extrapolation, OSFs were calculated by OSF = 0.01/BMDLhed. The highest OSF for a 21 species/sex combination for either a single tumor type or all combined tumors was selected as a 22 candidate OSF for TCDD cancer assessment. These steps in Figure 5-3 are further described in 23 detail in the following sections. 24 25 5.2.3.2.I. S election o f k ey data sets. 26 Based on the study selection criteria outlined in Section 2.3.2 (see Figure 2-3), EPA 27 selected five animal bioassays for use in the cancer dose-response assessment for TCDD (see 28 Table 2-6 and Section 2.4.2 for detailed study descriptions). Four of these studies (Della et al., 29 1987, 197405; Kociba et al., 1978, 001818; NTP, 1982, 594255; Toth et al., 1979, 197109), were 30 evaluated in the 2003 Reassessment, while one study (NTP, 2006, 543749) was published after 31 the 2003 Reassessment was released. The NTP (2006, 543749) study was specifically called out This document is a draftfor review purposes only and does not constitute Agency policy. 5-35 DRAFT--DO NOT CITE OR QUOTE 1 by the N A S (2006, 198441)) report as cancer bioassay data that E P A should evaluate prior to 2 com pleting its T C D D dose-response assessment. A s discussed below , E P A has chosen to 3 conduct dose-response modeling for a number o f tumor types from each of the sex/species 4 com binations in these studies in order to m axim ize the amount o f inform ation available to 5 support O S F derivation. Because tumors occurred in multiple sites in the exposed animals, each 6 tumor type w as considered separately (individual tumor models) and were also combined into 7 composite tumor incidence dose estimates (multiple tumor models). 8 Th e tumor incidence tables for these five bioassays are shown in Tab les 5-5 through 5-14 9 (see Section 2.4.2 for details o f these studies). The data in these tables are summarized from 10 each study' s reference publication and are the species/sex/tumor incidence data used for T C D D 11 dose-response m odeling in this report. E P A selected the anim al bioassay data sets in Tables 5-5 12 through 5-14 because they had been characterized by the study authors as having statistically 13 significant increases in tumor incidence by either a pair-w ise test between at least one treated 14 group and the controls or by a trend test showing increases in tumors w ith increases in dose. A n 15 exception w as made for cases w here statistical significance w as found in only one dose group 16 that w as not the highest dose group, and there w ere zero responses in every other dose group 17 including controls; these datasets w ere not modeled. F o r exam ple, in N T P (2006, 543749), E P A 18 notes that w hile the uterine tumors w ere statistically significant at 46 ng/kg using a pair-w ise 19 test, there w ere no uterine tumors in any other dose group, including the control and high dose 20 groups, and the trend test w as not significant; E P A excluded this tumor type from the analysis. 21 In addition, datasets w ith com bined tumors for the same site w ere given priority over subsets o f 22 tumors for that site. F o r example, in the N T P (1982, 594255) study on fem ale m ice, data on 23 com bined hepatocellular adenomas or carcinom as w ere modeled, but not data on hepatocellular 24 adenomas alone (not statistically significant) or on hepatocellular carcinomas alone (statistically 25 significant trend and high dose group). In the case o f the K o cib a et al. (1978, 001818) fem ale rat 26 combined hepatocellular adenomas and carcinom as only, E P A used data from a reanalysis o f the 27 pathology slides that w as published by Goodman and Sauer (1992, 197667) ; because the study 28 authors did not statistically analyze the revised tumor incidence data from their reanalysis, E P A 29 applied a F isc h e r' s E x a c t Test to evaluate the statistical significance o f those data. In the case o f 30 the N T P (2006, 543749) study only, information w as available regarding the length o f time that 31 the anim als stayed on test (105 w eeks); anim als w ho died w ithin the first year w ere censored This document is a draftfo r review purposes only and does not constitute Agency policy. 5-36 D R A FT-- DO NOT C IT E OR Q U O TE 1 from analysis in this document because animals who died within the first year were not 2 considered to have been alive long enough to develop tumors. Therefore, those animals were not 3 included in the denominators in Table 5-11. These adjusted incidence data were used in the 4 analysis of tumor dose-response for NTP (2006, 543749) in this document. The tumor incidence 5 data in Tables 5-5 through 5-14 include 6 7 nasal, tongue and adrenal tumors in males (Table 5-5), and liver, nasal and lung tumors in 8 females from the Kociba et al. (1978, 001818) 2-year study of Sprague-Dawley rats 9 (Table 5-6), 10 subcutaneous tissue, liver, adrenal and thyroid tumors in females (Table 5-7) and liver, 11 thyroid and adrenal tumors in males (Table 5-8), from the NTP (1982, 594255) 2-year 12 study of Osborne-Mendel rats, 13 subcutaneous tissue, hematopoietic system, liver and thyroid tumors in females 14 (Table 5-9), and lung and liver tumors in males, from the NTP (1982, 594255) 2-year 15 study of B6C3F1mice (Table 5-10), 16 liver, oral mucosa, pancreas and lung tumors in females from the NTP (2006, 543749) 217 year study of Sprague-Dawley rats (Table 5-11), 18 liver tumors in males from the Toth et al. (1979, 197109) 1-year study of Swiss/H/Riop 19 mice (Table 5-12), and 20 liver tumors in males (Table 5-13) and females from the Della Porta et al. (1987, 197405) 21 52-week study of B6C3F1mice (Table 5-14). 22 23 For each cancer endpoint, the reported (administered) doses from each study were converted, 24 where necessary, to average daily doses in ng/kg-day (e.g., doses administered 5 days/week were 25 adjusted by multiplying by 5 and dividing by 7 to get average daily doses). These doses were 26 then subjected to kinetic modeling to generate blood concentrations for use in TCDD dose- 27 response modeling. 28 29 5.2.3.2.2. D o se ad ju stm en t a n d extrapolation m eth ods f o r selec ted data sets. 30 5.2.3.2.2.I. D ose m e tr ic e stim a tio n f o r d o se -resp o n se m o d elin g . 31 Tables 5-5 through 5-14 show the blood concentrations that were used in TCDD dose- 32 response modeling of the animal bioassay data. Based on kinetic analysis (see Section 3), a 33 choice for whole blood concentration of TCDD was made for the purpose of dose extrapolation 34 between animals and humans. In order to estimate blood concentrations for each study selected, This document is a draftfor review purposes only and does not constitute Agency policy. 5-37 DRAFT--DO NOT CITE OR QUOTE 1 the Emond PBPK model was run using ACSLX software, version 2.5.0.6 (see Section 3). 2 Depending on the selected study, either rat or mouse versions of the model were used. In each 3 case, the simulation was performed using the exposure and observation durations, the body 4 weights, and the adjusted doses from the original studies. Details of PBPK model input 5 parameters are given for each study's m-file in Appendix C.2. In the case of Toth et al. (1979, 6 197109) study, which dosed the animals for a year and then followed up for the lifetime of the 7 animal, only the one-year simulation was performed. The m-files were used to run the 8 appropriate PBPK model to estimate time-averaged, maximum, and terminal (end of exposure) 9 blood concentration (see Appendix C.3). Other model simulated dose metrics such as 10 concentrations for liver, fat, Ah-receptor bound in liver, body burden, and the time at which the 11 maximum concentration was reached for each dose metric are also reported for illustrative 12 purposes in Appendix C.3. The complete results for each study modeled are shown in 13 Appendix C.3. 14 15 5.2.3.2.2.2. C a lcu la tio n o f h u m a n e q u iv a le n t d o se s (H E D s). 16 Human equivalent doses (ng/kg-day), corresponding to each BMDL (ng/kg) were 17 calculated using the Emond human PBPK model (see Section 3) and are denoted as BMDLHEDs. 18 The Emond human PBPK model was run for 70 years assuming a constant daily dose starting 19 from birth. The model concentrations were averaged over both the entire 70 year lifetime 20 (lifetime average) and over the five years surrounding the peak concentration (five-year average) 21 (see Section 3.3.1, describing first order body burden estimation). The human equivalent doses 22 were estimated by adjusting the daily dose model input until the time-averaged whole blood 23 concentration matched the associated alternative dose BMDL (derived earlier from animal PBPK 24 model). For animal studies which lasted longer than 540 days, the lifetime average was used; for 25 studies lasting less than 540 days, the five year average was used. The process was iterative and 26 continued until the modeled human concentration was within 1% of the BMDL. In general, 27 however, the concentrations matched to within 0.1%. 28 This document is a draftfor review purposes only and does not constitute Agency policy. 5-38 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.2.3. D o se-resp o n se m odelin g approach es f o r ro d en t bioassays. 2 5.2.3.2.3.I. M o d e lin 2 o f in d iv id u a l tu m ors. 3 EPA's BMDS Software, version 2.1 was used to estimate the BMDL01s for each of the 4 species/sex/tumor combinations, using the blood concentrations and incidence data shown in 5 Tables 5-5 through 5-14. Each data set was modeled using the multistage cancer model, and a 6 BMDL01 in blood concentration was estimated. The multistage model has been used by EPA in 7 the majority of its quantitative cancer assessments because it is statistically robust and able to 8 provide good fits to a wide range of dose-response patterns. It is also consistent with the 9 multistage nature of the carcinogenic process. The mathematical form of the multistage model is 10 11 P(d) = 1 - exp[-(q0 + q1d + q2d2 + ... + q d ) ] 12 (Eq. 5-6) 13 where 14 P(d) = lifetime excess risk (probability) of cancer at dose d 15 qi = parameters estimated in fitting the model, i = 1, ..., k. 16 17 To estimate the BMD01s and BMDL01s, BMDS was run with all parameters set to their 18 defaults; up to three degrees of freedom were specified for the dichotomous, multistage cancer 19 model; and a 95% confidence level. A 1% extra risk benchmark response (BMR) was used for 20 each tumor type, as this response level was judged to be sufficiently close to the observed 21 responses (see Section 5.2.3.2.6.11 for an expanded discussion). The BMDL01 (ng/kg) was then 22 converted to a BMDLhed (ng/kg-day) using the Emond human model, and an OSF in units of 23 (mg/kg-day)-1 was calculated by, OSF = 0.01/BMDLhed x106. Because of the nonlinearity of 24 blood concentration and ingested dose in the Emond Human PBPK model, the cancer risk is only 25 approximately linear with the TCDD blood concentration and low TCDD oral ingestion doses, 26 but is not linear with ingested TCDD at higher doses.40 Thus, to use these estimates in human 27 health risk assessment, risk-specific TCDD oral intake levels corresponding to the target risk 28 levels should be calculated, using a procedure similar to that for the slope factors based on 29 epidemiologic data (see Table 5-3). In the following sections, results are presented for the 40 T h is situ a tio n is a n a lo g o u s to th a t fo r th e c a n c e r risk m o d e lin g o f e p id e m io lo g ic d a ta fr o m th e C h e n g et a l. (2 00 6 ) analysis in Section 5.2.3.1 .2 .1 . This document is a draftfo r review purposes only and does not constitute Agency policy. 5-39 DRAFT--DO NOT CITE OR QUOTE 1 models that provided the best overall fit to the data as judged by comparison of likelihood ratios 2 for models that had an acceptable fit (chi-squared goodness of fit statisticp > 0.05). 3 4 5.2.3.2.3.2. M u ltip le tu m o r (B a yesia n ) m odels. 5 Statistically significant increased tumor incidences w ere observed at m ultiple sites in 6 m ale and/or female rats (K o cib a et al., 1978, 001818; N T P , 1982, 594255; N T P , 2006, 543749) 7 and male and female m ice (N TP , 1982, 594255) follow ing oral exposures to T C D D . W ith this 8 m ultiplicity o f tumors, the concern is that a potency or risk estimate based solely on one tumor 9 site (e.g., the most sensitive site) may underestimate the overall cancer risk associated with 10 exposure to this chem ical. R elevan t approaches in the 2005 C an ce r G u idelin es (U .S . E P A , 2005, 11 086237) for characterizing total risk include the follow ing: (1) analyze the incidence o f tum or 12 bearing anim als, or (2) com bine the potencies associated w ith significantly elevated tumors at 13 each site. The N R C (1994, 006424) concluded that an approach based on counts o f anim als w ith 14 one or more tumors (tumor-bearing anim als) w ould tend to underestimate overall risk w hen 15 tumor types occur independently, and thus an approach based on com bining the risk estimates 16 from each separate tumor type should be used. O n independence o f tumors, N R C (1994, 17 006424) stated " . . .a general assumption o f statistical independence o f tumor-type occurrences 18 w ithin anim als is not lik e ly to introduce substantial error in assessing carcinogenic potency." 19 B e cau se potencies are typ ically upper bound estimates, sum m ing such upper bound 20 estimates across tumor sites is likely to overstate the overall risk. Therefore, follow ing the 21 recom m endations o f the N R C (1994, 006424) and the 2005 C an ce r G u idelin es (U .S . E P A , 2005, 22 086237), a statistically valid upper bound on combined risk was derived, assuming 23 independence, in order to gain some understanding o f the overall risk resulting from tumors 24 occurring at multiple sites. In the case o f T C D D , tumors are thought to be independent across 25 the sites found in these three studies because: (1) they are in different organs and tissues, 26 specifically liver, lung, thyroid, subcutaneous tissue, oral cavity, tongue, pancreas, adrenal cortex 27 and the hematopoietic system; (2) different kinds o f tumors were found, even w ithin the same 28 organ (e.g., both cholangiocarcinom as and hepatocellular adenomas were found in female rat 29 livers in N T P (2006, 543749) ; and (3) the tumors found in these studies were not progressive 30 (i.e., they did not metastasize to other sites in the body). It is important to note that this estimate This document is a draftfor review purposes only and does not constitute Agency policy. 5-40 DRAFT--DO NOT CITE OR QUOTE 1 o f overall potency describes the risk o f developing tumors at any com bination o f the sites and is 2 not the risk o f developing tumors at all sites sim ultaneously. 3 F o r modeling individual tumor data, the multistage model is specified as shown in the 4 previous section (see E q . 5-6). U nder the assumption o f independence, the model for the 5 combined (or composite) tumor risk is still multistage, with a functional form that has the sum of 6 stage-specific multistage coefficients as the corresponding multistage coefficient. 7 8 Pc( d) = 1 - exp[ - ( l q 0 , + dLqIt + d1 l q 2 I + ... + dm' Lqmi) ] , for i = 1 ,..., k, (Eq. 5-7) 9 10 where k = total number of sites. 11 12 The resulting equation for fixed extra risk (BMR) is polynomial in dose (when logarithms 13 of both sides are taken) and can be solved in a straightforward manner for the combined BMD. 14 However, the current version of BMDS cannot estimate confidence bounds for this combined 15 BMD. 16 Therefore, a Bayesian approach to finding confidence bounds on the combined BMD was 17 implemented using WinBUGS (Spiegelhalter et al., 2003, 594261). WinBUGS software is freely 18 available and implements Markov Chain Monte Carlo (MCMC) computations. Use of 19 WinBUGS has been demonstrated for derivation of a distribution of BMDs for a single 20 multistage model (Kopylev et al., 2007, 194860) and is easily generalized (Kopylev et al., 2009, 21 198071) to derive the distribution of BMDs for the combined tumor load, following the NRC 22 (1994, 006424) methodology described above. The advantage of a Bayesian approach is that it 23 produces a distribution of BMDs that allows better characterization of statistical uncertainty. For 24 the current analysis, a diffuse (high variance or low tolerance) Gaussian prior restricted to be 25 nonnegative was used. The posterior distribution was based on three simulation chains with 26 50,000 burn-in (i.e., the initial 50,000 iterations were dropped) and a thinning rate of 20, 27 resulting in 150,000 interactions total. The median and 5th percentile of the posterior distribution 28 provided the BMD01 (central estimate) and BMDL01 (lower bound) for combined tumor load, 29 respectively. 30 The methodology above was applied to the statistically significant dose-response data 31 from Kociba et al. (1978, 001818), NTP (1982, 594255), and NTP (2006, 543749) (see This document is a draftfor review purposes only and does not constitute Agency policy. 5-41 DRAFT--DO NOT CITE OR QUOTE 1 Section 2.3.2 for data set selection criteria).41 As with the risk estimates generated for individual 2 tumor sites, the combined analysis used the internal dose metric, whole blood concentration (see 3 Section 3). For the combined tumors for each sex/species combination, a BMDL01 in blood 4 concentrations was estimated. The BMDL01 (ng/kg) was then converted to a BMDLHeD 5 (ng/kg-day) using the Emond human model, and an OSF in units of (mg/kg-day)-1 was 6 calculated by, OSF = 0.01/BMDLHed x 106. Because of the nonlinearity of blood concentration 7 and ingested dose in the Emond Human PBPK model, the cancer risk is linear only with the 8 TCDD blood concentration and low TCDD oral ingestion doses, but is not linear with ingested 9 TCDD at higher doses; a single OSF cannot represent the entire range of risks for oral ingestion. 10 Thus, to use these estimates in human health risk assessment, risk-specific TCDD oral intake 11 levels corresponding to the target risk levels should be calculated using a procedure similar to 12 that for the slope factors based on epidemiologic data (see Table 5-3). 13 14 5.2.3.2.4. R esu lts o f d ose-respon se m odelin g f o r ro d en t bioassays. 15 Table 5-15 presents the benchmark dose modeling results for both the individual tumors 16 and the combined tumors based on TCDD blood concentrations. The p-values in the table are 17 for a chi-square goodness of fit statistic with significance ofp > 0.05. Goodness of fit was 18 acceptable atp > 0.05 for all models. The difference in log likelihood (dLL) statistic documents 19 the difference in log likelihoods between stages of the models in cases where the stage is 20 above 1; it shows the difference between the stage in the table and the lower stage. For example, 21 for the NTP (2006, 543749) liver cholangiocarcinomas, twice the difference of 2.92 would be 22 >3.84, the test statistic from the assumed chi-square distribution,42 with p = 0.95, justifying the 23 choice of 3 stages over 2 stages. The best fitting multistage models include: a 1-stage (linear) 24 model for all of the individual tumor data sets from Kociba et al. (1978, 001818), NTP (1982, 25 594255), and Toth et al. (1979, 197109), for liver carcinomas in females in Della Porta et al. 26 (1987, 197405), as well as for the pancreatic and oral mucosa tumors in NTP (2006, 543749); a 41 Because only one turm or site was statistically sig nificantly elevated in both the D ella Porta et al. (1987, 197405) and To th et al. (1979, 197109) (i.e., only increased incideces o f liv e r tum ors were statistically sig nificant elevated in both studies), a m ulti-tu m or analysis was not conducted. 42The chi-square d istrib ution w ith 1 degree o f freedom is the correct d istrib ution only under standard conditions (e.g., no boundary parameters in n u ll hypothesis). Thus, the correct d istrib ution fo r the situation where the parameter o f interest is on the boundary, as happens w ith testing fo r the order o f the m ultistage model, and, possibly nuisance parameters (estimated parameters o f the m odel), is very d iffic u lt to derive (S e lf and Liang, 1987, 594398). Therefore the p-value o f chi-square w ith one degree o f freedom is used as the best available choice. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-42 DRAFT--DO NOT CITE OR QUOTE 1 2-stage model for the lung tumors in NTP (2006, 543749) and for liver carcinomas in males from 2 Della Porta et al. (1987, 197405); and a 3-stage model for the liver cholangiocarcinoma and liver 3 adenoma data sets from NTP (2006, 543749). The multi-stage model fit was not significant (p > 4 0.1) in the NTP (1982, 594255) study for lung tumors in the male mouse (p = 0.09), adrenal 5 cortex (p = 0.06)and thyroid follicular cell adenomas (p = 0.06) in male rats, and subcutaneous 6 tissue in female mice (p = 0.09), and was also not significant for liver carcinomas (p = 0.019) in 7 female mice in Della Porta et al. (1987, 197405). For the Toth et al. (1979, 197109) liver 8 tumors, the model fit to all of the data was poor, and the highest dose group was dropped in order 9 to achieve an acceptable fit (p = 0.29). The BMD01s and BMDL01s (ng/kg) were estimated from 10 these multistage models for the individual tumors. BMD01s and BMDL01s (ng/kg) were also 11 provided in Table 5-15 for the combined tumors for each sex/species combination within a study. 12 These were estimated from the distributions of BMD01s produced by the Bayesian MCMC 13 simulation (see Section 5.2.3.1.2.3.2). The BMD01s and BMDL01s (ng/kg) for the combined 14 tumors in Table 5-15 are the mean and lower 95% percentile values from these distributions, 15 respectively. 16 17 5.2.3.2.4.I. I n d iv id u a l tu m o r m o d els. 18 Table 5-16 shows the BMDLHEDs (ng/kg-day) that were estimated from the BMDL01s in 19 Table 5-15 using the Emond human model (see Section 5.2.3.1.2.2.2) and the OSFs calculated 20 by, OSF = 0.01/BMDLhed x 106to convert the OSF to (mg/kg-day)-1 units. BMDS results, 21 details of the model fits and dose-response graphics for all endpoints are shown in Appendix F. 22 Although only the blood concentration results are presented in this section, for comparison 23 purposes, Appendix F also provides modeling results for the studies' administered average daily 24 doses. Table 5-16 lists the OSFs in decreasing value. It can be seen that liver tumors in male 25 mice yield the highest slope factors; OSF values are 5.9 x106 and 5.2 x106 per mg/kg-day in 26 NTP (1982, 594255) and Toth et al. (1979, 197109), respectively. The OSFs for the new NTP 27 (2006, 543749) study in female rats are two orders of magnitude lower, ranging from 1.8 x104to 28 1.8 x105per mg/kg-day, representing the lowest OSFs for TCDD from the individual tumor 29 models. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 5-43 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.2.4.2. M u ltip le tu m o r (B a yesia n ) m odels. 2 Table 5-17 shows the BMDLhedS(mg/kg-day) that were estimated from the BMDL01s in 3 Table 5-15 using the Emond human model (see Section 5.2.3.1.2.2.2) and the OSFs calculated 4 by, OSF = 0.01/BMDLheD x106to convert the OSF to (mg/kg-day)-1 units. Table 5-17 lists the 5 OSFs in decreasing value. It can be seen that the combined liver and lung tumors in male mice 6 yield the highest OSF value of 9.4 x106 per mg/kg-day from NTP (1982, 594255), and the 7 combined adrenal, tongue and nasal tumors in male rats yield the lowest OSF value of 3.2 x105 8 from Kociba et al. (1978, 001818). The OSF for the combined liver, oral mucosa, lung, and 9 pancreatic tumors in female rats from the newer NTP (2006, 543749) study is 4.4 x105. 10 11 5.2.3.2.5. S u m m a ry evalu ation o f slo p e fa c to r estim ates f r o m ro d en t bioassays. 12 To estimate a range of candidate TCDD OSFs from the animal data, dose-response 13 modeling of the five chronic rodent bioassays identified in Section 2.4.3 was conducted. Dose14 response modeling was performed using whole blood concentrations, and BMDLheD values 15 (ng/kg-day) were derived for the 28 species/sex/endpoint data sets individually (see Table 5-16) 16 and for seven species/sex combined tumor data sets (see Table 5-17). 17 The highest OSFs that have been derived for these animal cancer bioassays using the 18 multistage models are from the multiple tumor analyses for NTP (1982, 594255; 2006, 543749) 19 and Kociba et al. (1978, 001818), presented in Table 5-17, and from the individual tumor 20 analyses for Toth et al. (1979, 197109) liver tumors and Della Porta et al. (1987, 197405) liver 21 carcinomas in male mice, presented in Table 5-16. The most sensitive species and sex is male 22 mice, for which the estimated BMDLheD for combined tumors is 1.1 x 10"3 ng/kg-day. This 23 result, which is derived under the assumption that multiple tumor types occur independently in 24 the exposed animals, is, as expected, lower than the BMDLheD for the most sensitive individual 25 tumor. 26 Based on these results, EPA believes that a credible value for the BMDLheD derived from 27 the animal studies lies in the range shown in Table 5-17 between 3.1 x 10-2 and 28 1.1 x 10-3 ng/kg-day. These values, which correspond to oral slope factor values of 3.2 x 105 29 and 9.4 x 106 per mg/kg-day, respectively, encompass the range at which elevated cancer risks 30 can be detected for the most sensitive species, sex, and endpoints in the animal bioassay data. This document is a draftfor review purposes only and does not constitute Agency policy. 5-44 DRAFT--DO NOT CITE OR QUOTE 1 A s noted above in Sections 5.2.3.1.2.2 and 5.2.3.1.2.3, the cancer mortality risk is strictly 2 linear only with T C D D blood concentration, such that a single O S F cannot represent the entire 3 range o f risks for oral ingestion. The O S F s shown in Tables 5-16 and 5-17 are based on H E D s 4 corresponding to the B M D L 01, w h ich are most representative o f low er human exposure levels, 5 including ambient exposures. F o r higher exposures, the risks increase at a slow er rate w ith 6 increasing dose and the corresponding O S F s are lower; in those cases, risk-specific doses can be 7 calculated as previously described (see Section 5.2.3.2.3.2). 8 9 5.2.3.2.6. Q ualitative u n certain ties in slope fa c to r estim ates f r o m ro d en t bioassays. 10 T h is section presents a qualitative discussion o f the uncertainties associated w ith 11 calculating the O S F for T C D D from chronic anim al bioassay data. D iscu ssio n s on the feasibility 12 o f conducting a quantitative uncertainty analysis for T C D D using dose-response methods are 13 provided in Section 6.4.2 o f this document. 14 15 5.2.3.2.6.1. Q u a lity o f s tu d ie s r e lie d u p o n f o r d e te rm in in g P O D . 16 E P A considers the overall quality and breadth o f the studies used for the cancer dose17 response analysis to be excellent. A ll o f the studies w ere published in the peer-reviewed 18 literature, and two o f them w ere conducted by N T P (1982, 594255; 2006, 543749) . 19 K o cib a et al. (1978, 001818) , D e lla Porta et al. (1987, 197405) and Toth et al. (1979, 197109) 20 are older studies, but appear to have been conducted according to good laboratory practice 21 standards. Th e control and dose group sample sizes w ere relatively large, ~ 4 0 -5 0 anim als or 22 more per group for all o f the studies. A ll five studies exposed the test anim als via the oral route 23 to T C D D alone, as w as stipulated in E P A ' s study inclusion criteria. C o llective ly , these five 24 studies reported development o f numerous cancer endpoints (tumors) in both sexes in two strains 25 o f rats (Sprague-D aw ley and Osborne-M endel) and two strains o f m ice (i.e., B 6 C 3 F 1, 26 Swiss/H/Riop). The overall high quality o f these studies and the strong, positive association 27 between T C D D exposure and cancer suggests that study quality is not a m ajor contributing factor 28 to uncertainty in the risk estimates. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 5-45 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.2.6.2. I n te r p r e ta tio n o f r e s u lts fr o m s tu d ie s r e lie d u pon f o r d e te rm in in g P O D . 2 A s discussed in Section 3.4.3.2.1, questions arose about the interpretation o f liver tumor 3 responses in fem ale rats in the K o cib a et al. (1978, 001818) study. Three re-evaluations o f the 4 slides have been reported (Goodm an and Sauer, 1992, 197667; K o cib a et al., 1978, 001818; 5 Squire, 1980, 594272) . The decision to use the Goodman and Sauer (1992, 197667) evaluation 6 w as based on their use o f the most current tumor classification procedures. The incidence of 7 hepatocellular adenomas and carcinomas (individually and combined), however, did vary 8 (sometimes w idely) for each dose group across the three evaluations. Although the state-of-the9 science is reflected in the Goodman and Sauer analysis, there is some uncertainty in the 10 interpretation o f any post-hoc analysis. N o issues have arisen w ith regard to the interpretation o f 11 the N T P (1982, 594255; 2006, 543749), D e lla Porta et al. (1987, 197405) or Toth et al. (1979, 12 197109) tumor identification and classification. 13 14 5.2.3.2.6.3. C o n siste n c y o f r e s u lts a c ro ss ch ro n ic r o d e n t bioassavs. 15 Th e existence of five high-quality chronic bioassays for T C D D increases confidence and 16 reduces uncertainty in the cancer O S F s . Considered together, these studies tested two species 17 and both sexes o f m ice and rats, and a w ide range o f w ell-characterized tumor types. A ll five 18 studies w ere consistent in observing increases (at some dose level) in rates of liv e r tumors (in 19 both species and sexes). W h ile tumors at other sites w ere observed (and those sites varied across 20 study, species, and sex), the liver tumors were consistently the most sensitive indicators of 21 carcinogenic response (w ith respect to B M D LH ED estimates). L u n g tumors w ere also 22 consistently observed across three o f the studies, in male m ice in the N T P (1982, 594255) study 23 and in fem ale rats in K o cib a et al. (1978, 001818) and N T P (2006, 543749) . A s discussed above, 24 the two most sensitive single-tumor endpoints as judged by BM D L01 values were associated with 25 elevated liver tumor risks, followed by lung, lymphoma or leukemia, thyroid and adrenal 26 cancers. The consistency of tumor types and sensitivities across endpoints and studies lends 27 confidence to the multistage m odeling results. 28 29 5.2.3.2.6.4. H u m a n re le v a n c e o f r o d e n t tu m o r data. 30 There is some concordance in the tumor responses observed in the rodent test species and 31 humans, how ever, the most sensitive tumor site in the anim als, the liver, has not been associated This document is a draftfor review purposes only and does not constitute Agency policy. 5-46 DRAFT--DO NOT CITE OR QUOTE 1 w ith cancer from T C D D exposures in humans. O n the other hand, lung cancer and leukem ia are 2 found both in the animal studies and in epidem iologic studies o f exposed workers. The 3 consistency across sex, species, and strains in the animal studies suggests that the occurrence of 4 several of these tumors, in particular, liver and lung tumors is not an idiosyncratic response of a 5 particular combination o f species, strain, or sex. A s discussed in Section 5.2.1, the likely A h R 6 related carcinogenic mechanism is credible for humans as w ell as for rodent species. 7 8 5.2.3.2.6.5. R e le v a n c e o f r o d e n t e x p o su re scen ario. 9 Three o f the five chronic rodent bioassays exposed the test anim als for ~2 years, the 10 m ajority o f their lifespans. Toth et al. (1979, 197109) exposed the anim als only for one year, but 11 they w ere kept on the study for a second year before they w ere evaluated for cancer. The D e lla 12 Porta et al. (1987, 197405) study also exposed the test anim als for one year, and a dosing error 13 occurred during the study. A t ages 31 to 39 w eeks, 41 m ale m ice and 32 fem ale m ice in the 14 2,500 ng/kg B W dose group w ere m istakenly adm inistered a single dose o f 25,000 ng/kg B W 15 T C D D . T C D D treatment for the 2,500 ng/kg B W dose group w as halted for 5 w eeks (beginning 16 the w eek after the 25,000 ng/kg B W dose w as adm inistered in error) and resumed until exposure 17 w as terminated at 57 w eeks. Thus, the large single dose and subsequent period without T C D D 18 exposure confounds the dose-response relationship for this study. In general, these lifetim e 19 bioassays in anim als have long been used by E P A to assess potential lifetim e exposures and 20 effects in humans. H ow ever, in the case o f T C D D , the h alf life o f T C D D in the body for rats, 21 m ice, and hum ans is very different (see Section 3). Th u s, there is a significant amount o f 22 uncertainty in the use o f rat and mouse data to develop O S F s for human cancer risk assessment 23 o f T C D D . 24 25 5.2.3.2.6.6. I m p a c t o f b a c k g r o u n d T C D D exposu res. 26 It is known that T C D D has been found in the feed used in animal bioassays, and that this 27 is a confounding factor, particularly in older studies. The effect o f T C D D in the diets o f test 28 species has the potential to be quite significant given the lo w levels o f T C D D at w h ich adverse 29 effects have been observed. Insofar as that is an issue, the risks associated with T C D D 30 exposures in the animal bioassays, and therefore the O S F s, w ould be biased high, w hich could be 31 the case for the N T P (1982, 594255) , D e lla Porta et al. (1987, 197405), K o cib a et al. (1978, This document is a draftfor review purposes only and does not constitute Agency policy. 5-47 DRAFT--DO NOT CITE OR QUOTE 1 001818) and Toth et al. (1979, 197109) studies. The im pact o f this issue is that the new er study, 2 N T P (2006, 543749) , accounted for T C D D exposures in the anim al feed. Thus, there is lik e ly to 3 be less uncertainty in the T C D D dose-response information presented in N T P (1982, 594255; 4 2006, 543749) than in the other four studies conducted before 1990. 5 6 5.2.3.2.6.7. C h o ice o f e n d p o in t f o r P O D d eriva tio n . 7 A s noted above, the liver tumor P O D s represent the most sensitive single-tumor endpoint 8 across the five cancer bioassays. Thus, the liver cancer endpoints must be seriously considered 9 for derivation o f a T C D D O S F . A s discussed in the previous section, E P A has also developed 10 B ay esia n dose-response estimates for com bined tumors, w h ich yield B M D L0 1 values slightly 11 low er than those for any individual tumor type. Although it is the most conservative choice to 12 select the low est com bined tumor P O D for O S F derivation, there are uncertainties associated 13 w ith the m ultiple tumor analysis. The assumption o f independence o f tumors across sites is 14 reasonable, particularly since the tumors from T C D D do not metastasize. H ow ever, the 15 independence assumption lacks hard evidence and needs further laboratory confirm ation. 16 17 5.2.3.2.6.8. C h o ice o f a n im a i-to -h u m a n e x tra p o la tio n m e th o d . 18 Th e analyses presented here have used the Em o n d human kinetic model for extrapolating 19 dose from anim als to hum ans (as discussed in Section 3.4.2). The rationale for this choice is that 20 the blood concentration metric most accurately reflects the concentration of T C D D in the various 21 tissues. A s discussed in Section 3.4.3.2.4, use o f the blood concentration dose m etric results in 22 critical dose estimates (H E D s) that are considerably low er (10- to more than 100-fold) than those 23 derived based on adm inistered dose. T h is does not reflect bias in the blood-based measure; 24 rather it is a reflection of the nonlinear biokinetics of T C D D in the body. E P A has also explored 25 the impacts of using other dose metrics, including AhR-bound T C D D concentration calculated 26 based on the Em ond model. A s discussed in Section 3.4.3.2.6.2, this also results in H E D 27 estimates much low er than those obtained based on administered dose. 28 29 5.2.3.2.6.9. C h o ice o f m o d e l f o r P O D a n d m o d e l u n c e r ta in ty f o r P O D d eriva tio n . 30 The bioassay-based cancer dose-response assessment in this section has used the 31 m ultistage model w h ich is the standard model choice for such assessm ents and has been the basis This document is a draftfor review purposes only and does not constitute Agency policy. 5-48 DRAFT--DO NOT CITE OR QUOTE 1 for most of EPA's cancer risk assessments. The multistage model is the standard because it is 2 the only available model form that allows for low-dose linearity while accommodating 3 curvilinearity at higher doses and can be readily implemented. 4 There is some model choice uncertainty associated with instances of lack of fit. When 5 the multistage model does not adequately describe the observed pattern of responses (typically 6 determined by examining the p-value for lack of fit), a decision must be made about possible 7 adjustments, including the dropping of higher dose groups thought to be less relevant to the 8 estimation of low-dose slopes. In this analysis, poorer fits (p-values less than 0.10) were 9 observed in five cases, four from NTP (1982, 594255) and one from Della Porta et al. (1987, 10 197405) (see Table 5-15). The lowest BMDL01 associated a lowp-value (p = 0.09) was for the 11 lung tumors in the NTP (1982, 594255) male mouse, the third lowest POD behind the liver 12 PODs in the individual tumor data sets. The other instances were for adrenal cortex and thyroid 13 follicular cell adenomas in male rats and for subcutaneous tissue in female mice in the NTP 14 (1982, 594255) study and for liver carcinomas in female mice in Della Porta et al. (1987, 15 197405). In those instances, the p-values were 0.06, 0.06, 0.09, and 0.019, respectively. These 16 poorly fit data sets provide OSF estimates that are uncertain and also contribute to uncertainty in 17 the combined tumor PODs from NTP (1982, 594255). The lowest BMDL01 in the combined 18 tumors is for the male mice combined liver and lung tumors, thus estimates from this sex/species 19 combination from NTP (1982, 594255) is highly uncertain and impacts its choice as a POD. 20 21 5.2.3.2.6.10. S ta tis tic a l u n c e r ta in ty in m o d e l fits. 22 Every model fit to a data set is associated with some inherent statistical uncertainty. For 23 this reason, bounds were calculated and used for OSF derivation (e.g., lower bounds on 24 benchmark doses, in this case the BMDL01s). Those bounds account for uncertainties associated 25 with finite samples o f test animals, both in terms o f the number o f dose groups and o f the 26 number of animals per dose group. Valid and accepted statistical procedures have been applied 27 to ascertain the impact o f those limitations on the estimates o f interest. That being the case, the 28 statistical uncertainties associated with finite samples have been adequately addressed. 29 This document is a draftfor review purposes only and does not constitute Agency policy. 5-49 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.2.6.11. C h oice o f r is k le v e l f o r P O D d eriva tio n . 2 The BMR level that has been used for the POD in deriving the cancer OSF is one percent 3 extra risk. A single BMR was chosen for consistency across studies. Also, a BMR of 1% was 4 judged to be near the range of the observations. For the TCDD animal cancer bioassay data, 5 although many of the first positive tumor incidence responses (relative to controls) are closer to 6 10% (some higher), some are as low as 2%. Furthermore, most of the BMD01 values are within a 7 factor of 3 of the lowest tested dose, and the BMDL01 values are generally less than a factor of 2 8 below the BMD. Table 5-18 presents a comparison of BMDs, BMDLs and slope factors for 1%, 9 5% and 10% BMRs from the multi-tumor analyses of NTP (1982, 594255; 2006, 543749) and 10 Kociba et al. (1978, 001818) and for selected single tumor data sets from Toth et al. (1979, 11 197109) and Della Porta et al. (1987, 197405). In Table 5-18, the choice of BMR has little or no 12 impact on the slope factors based on TCDD blood concentration for the combined or single 13 tumor incidences selected as representative of each study.43 In contrast, Table 5-19 presents a 14 comparison of Human Equivalent Dose BMDs, BMDLs and slope factors for 1, 5, and 10% 15 BMRs from these same datasets. Table 5-19 shows that, when converting the blood 16 concentration to the equivalent HED, a 2-fold to 4-fold decrease in the OSF is obtained when 17 using a BMR of 10% rather than 1%. This result is a consequence of the nonlinearity in the 18 Emond PBPK model at higher doses, where dose-dependent elimination of TCDD in the liver 19 results in a less-than-proportional increase in blood concentration relative to oral intake. At 20 lower exposure levels, blood concentration is proportional to oral intake. Therefore, EPA has 21 chosen the lower BMR of 1% as more representative of the low-dose risk. 22 23 5.2.3.3. E P A ' R esp o n se to th e N A S C om m en ts on C hoice o f R esp o n se L e v e l a n d 24 C h aracterization o f th e S ta tistic a l C on fiden ce A r o u n d L o w D o se M o d e l P rediction s 25 The NAS was concerned with the statistical power to determine the shape of the dose 26 response curve at low doses, well below observed dose-response information. EPA shares this 27 concern in that the shape of the dose-response curve in the low-dose region cannot be determined 28 with confidence when based on higher dose information. 43 T h is w ill g e n e r a lly b e th e ca se fo r m u ltis ta g e m o d e l fits w ith 1st-d e g re e c o e ffic ie n ts g re a te r th a n ze ro b e c a u s e the response at the B M D L is virtually linear at B M R s o f 10% or less. For m odel fits dom inated by higher-order coefficients, linearity o f response at the B M D L begins at low er B M R s. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-50 DRAFT--DO NOT CITE OR QUOTE 1 W hen tumor data are used for dose-response modeling, a P O D is obtained from the 2 modeled tumor incidences. W hen assessing carcinogenicity using a linear extrapolation 3 approach from a P O D , a balance must be struck between staying w ithin the range o f the 4 observations and obtaining a representative estimate o f the low-dose slope. Traditional cancer 5 bioassays, w ith approximately 50 anim als per group, can typically support m odeling down to an 6 increased incidence o f 1-10% ; epidemiologic studies, with larger sample sizes, below 1%. Fo r 7 the T C D D anim al cancer bioassay data, most o f the low -dose tumor incidence responses are 8 under 10% (relative to controls), w ith some as low as 2% . F o r comparison purposes, B M D s, 9 B M D L s and O S F s from the anim al cancer bioassay benchm ark dose m odeling assum ing 1, 5, 10 and 10% extra risk are shown in units o f blood concentrations and human equivalent doses in 11 Tab les 5-18 and 5-19, respectively. A fter evaluating the magnitude o f the uncertainty in 12 B M D L 01s against the im pact o f using B M D L 10s, E P A has chosen to use a 1% B M R in all cases, 13 determining that the uncertainty bounds on the B M D L0 1 values are reasonable. 14 In the analysis o f the anim al cancer bioassays presented in this document, the multistage 15 cancer model w as applied w ith a linear dose extrapolation to zero. E P A used a 1% excess risk 16 estimate, i.e., a B M D L 01, as the P O D for development o f candidate T C D D cancer oral slope 17 factors using a B a y esia n m ultitum or approach (see Section 5.2.3.2. The advantage o f a B a y esia n 18 approach is that it produces a distribution o f B M D s that allow s better characterization o f 19 statistical uncertainty. 20 Central tendency slope estimates and upper bound oral slope factor estimates are part of 21 the standard B M D S m ultistage cancer model and are included in each output file for the animal 22 bioassay single tumor analyses in Appendix F. Central tendency B M D s are also reported for the 23 results o f the anim al bioassay m ultitum or an alysis (see Tab le 5-15). Central tendency slope 24 estimates are given for all the qualifying epidem iological studies as w ell (see Tab les 5-1 and 25 5-4), where possible. 26 27 5.2.3.4. E P A ' R esp o n se to th e N A S C om m en ts on M o d e l F orm s f o r P red ictin g C an cer R isks 28 B elow the PO D 29 The N A S offered extensive comments on the cancer dose-response modeling in the 2003 30 Reassessm ent. Although epidemiologic and rodent bioassay data are useful for the evaluation of 31 the dose-response curve w ithin the range o f the observed response data, they have traditionally This document is a draftfor review purposes only and does not constitute Agency policy. 5-51 DRAFT--DO NOT CITE OR QUOTE 1 not been useful sources o f information for identifying a threshold or for estimating the shape of 2 the dose-response curve below the P O D . Rather, m echanistic toxicological data have been the 3 evidentiary sources o f choice for those types o f analyses. A s noted above, any quantitative 4 estimation o f carcinogenic risk associated with T C D D exposure requires low-dose extrapolation 5 o f experimental data. Unfortunately, the shape o f the dose-response curve in the lo w dose region 6 is unknown. 7 Several o f the analyses o f epidemiological cohort data evaluated the fit o f different dose- 8 response m odels to the data. Log -d ose m odels accentuate the im portance o f low -dose low - 9 magnitude responses and can yield im plausible results. The most relevant models used in these 10 studies are the untransformed-dose C o x regression models. Better results have been obtained in 11 the cohort analyses w hen the flattening o f the hazard-ratio curve is taken into account. The latter 12 has been m odeled explicitly by Steenland et al. (2001, 198589), w ho use a p iecew ise linear 13 model and im p licitly by Cheng et al. (2006, 523122) , w ho drop out a percentage o f the high-dose 14 response data and fit a linear model to the remainder. Im portantly, the analyses o f the 15 epidem iologic cohorts presented in Section 5.2.3.1 are lim ited to evaluation and reanalyses o f 16 published data as reported by the study authors. E P A does not have access to the raw data from 17 these epidem iologic studies and, therefore, could not conduct d e n o v o analyses. 18 19 5.2.3.4.1. C hoice o f extrapolation approach 20 5.2.3.4.1.1. T C D D a n d r e c e p to r th eo ry. 21 T C D D is considered to be a receptor-mediated carcinogen in anim als. N e arly all T C D D 22 experimental data are consistent with the hypothesis that the binding o f T C D D to the A h R is the 23 first step in a series o f biochem ical, cellular, and tissue changes that ultim ately lead to toxic 24 responses observed in both experimental animals and humans (Part II, Chapter 2 of the 2003 25 Reassessm ent). Ligand-receptor binding, like any bim olecular interaction, obeys the law o f mass 26 action as originally formulated by A .J. C lark (Lim bird, 1996, 594276) . The law o f mass action 27 predicts the fractional receptor occupancy at equilibrium as a function o f ligand concentration. 28 Fractional occupancy ( Y ) is defined as the fraction o f all receptors that are bound to ligand: 29 _ [TCDD - AhR ] _ [TCDD - AhR ] _ [TCDD ] 30 _ [A h R ]TOT _ [A h R ] + [TCDD - A h R ] _ [TC D D ] + K d (Eq. 5-8) This document is a draftfor review purposes only and does not constitute Agency policy. 5-52 DRAFT--DO NOT CITE OR QUOTE 1 where [TCDD] is the concentration of the ligand, [AhR] is the concentration of the receptor and 2 [TCDD-AhR] is the amount of liganded receptor. The equilibrium dissociation constant Kd 3 describes the affinity of the interaction and is the concentration of TCDD that results in 50% 4 receptor occupancy. This simple equation defines a rectangular hyperbola, which is the 5 characteristic shape of the vast majority of biological dose-response relationships. 6 In certain cases, no response occurs even when there is some receptor occupancy. This 7 suggests that there may be a threshold phenomenon that reflects the biological "inertia" of the 8 response (Ariens et al., 1960, 594279). In other cases, a maximal response occurs well before all 9 receptors are occupied, a phenomenon that reflects receptor "reserve" (Stephenson, 1956, 10 594280). Therefore, the law of mass action cannot by itself fully explain the effect or response 11 observed after TCDD interacts with AhR. The ligand-receptor complex is associated with a 12 signal transduction or effector system. In the case of the AhR, this effector system can be 13 considered to be the transcriptional machinery itself. The key feature of this formulation is that a 14 response is proportional, or a function of, the number of receptors occupied. 15 Furthermore, for a ligand such as TCDD that elicits multiple receptor-mediated effects, 16 one cannot assume that the binding-response relationship for a simple effect (such as enzyme 17 induction) will necessarily be identical to that for a different and more complex effect (such as 18 cancer). The cellular cascades of events leading to different complex responses (e.g., altered 19 immune function, developmental effects, or cancer) are different, and other rate-limiting events 20 likely influence the final biological outcome resulting in different dose-response curves. Thus, 21 even though TCDD binding to AhR is assumed to be the initial event leading to a spectrum of 22 biological responses, TCDD-AhR binding data may not always correlate with the dose-response 23 relationship observed for particular effects. 24 A receptor-based mechanism would predict that, except in cases where the concentration 25 of TCDD is already high (i.e., [TCDD]~Kd), incremental exposure to TCDD will lead to some 26 increase in the fractional occupancy of AhR. However, as discussed above, it cannot be assumed 27 that an increase in receptor occupancy will necessarily elicit a proportional increase in all 28 biological response(s), because numerous molecular events contributing to the biological 29 endpoint are integrated into the overall response. That is, the final biological response could be 30 considered as an integration of a series of interdependent dose-response curves with each curve 31 dependent on the molecular dosimetry for each particular step. Dose-response relationships that This document is a draftfo r review purposes only and does not constitute Agency policy. 5-53 DRAFT--DO NOT CITE OR QUOTE 1 will be specific for each endpoint must be considered when using mathematical models to 2 estimate the risk associated with exposure to TCDD. It remains a challenge to develop models 3 that incorporate all the complexities associated with each biological response as the modes of 4 action for various toxicological endpoints appear to vary greatly. For TCDD, extensive 5 experimental data from studies using animal and human tissues indicate that cell- or tissue6 specific factors determine the quantitative relationship between receptor occupancy and the 7 ultimate biological response. This would suggest that the parameters for each mathematical 8 model might only apply to a single biological response within a given tissue and species, making 9 extrapolation to other systems challenging. 10 11 5.2.3.4.I.2. L o w -d o s e ex tra p o la tio n : th r e s h o ld o r n o th re sh o ld ? 12 As indicated in the 2005 Cancer Guidelines,44 toxicity reference values for human 13 noncancer endpoints have historically been estimated based on a no-observed-adverse-effect 14 level (NOAEL) or lowest-observed-adverse-effect level (LOAEL) from animal bioassay studies. 15 This terminology suggests a biological population threshold beneath which no harm is 16 anticipated. Reference values such as the oral reference dose (RfD) or inhalation reference 17 concentration (RfC) are derived by applying uncertainty factors (UFs) to a POD. Depending on 18 the nature of available data and modeling choice, a POD can be selected from values other than 19 an NOAEL or LOAEL, such as an EDx, or a benchmark dose (BMD) or its BMDL. An RfD is 20 described as "likely to be without appreciable risk" but the probabilistic language has not as yet 21 been operationalized. There is no quantitative definition of "appreciable" and no mechanism to 22 compute risk as a function of dose, so as to ascertain that the risk is indeed not appreciable. The 23 risk at the RfD is not calculated, and it cannot be calculated within the current UF framework. 24 Instead, a hazard quotient is computed as the ratio of a given exposure to the RfD, or a margin of 25 exposure is estimated as the ratio of the POD to the human exposure level. 26 Cancer endpoints are predominantly thought to have no population biological threshold. 27 Although the terminology "threshold/nonthreshold" is still common in cancer dose-response 44A s s t a t e d i n t h e 2 0 0 5 C a n c e r G u i d e l i n e s ( U . S . E P A , 2 0 0 5 , 0 8 6 2 3 7 ) : " F o r e f f e c t s o t h e r t h a n c a n c e r , r e f e r e n c e v a lu e s h a v e b e e n d e s c rib e d a s b e in g b a s e d o n th e a s s u m p tio n o f b io lo g ic a l th re s h o ld s . T h e A g e n c y 's m o re c u rre n t g u id e lin e s f o r th e s e e ffe c ts (U .S . E P A , 1 9 9 6 , 5 9 4 3 9 9 ; U .S . E P A , 1 9 9 8 , 0 3 0 0 2 1 ) h o w e v e r, d o n o t u s e th is a s s u m p tio n , c itin g th e d iffic u lty o f e m p iric a lly d is tin g u is h in g a tru e th re s h o ld f r o m a d o s e -re s p o n s e c u rv e th a t is n o n lin e a r a t lo w d o se s." This document is a draftfo r review purposes only and does not constitute Agency policy. 5-54 DRAFT--DO NOT CITE OR QUOTE 1 discussions, the 2005 Cancer Guidelines propose a different terminology, whereby "nonlinear 2 models" are those whose dose-response slope is zero at or above zero. In the natural language, 3 and indeed in data analysis, it is difficult to distinguish the following situations: 4 5 The response approaches zero as dose goes to zero, versus 6 The response slope goes to zero as dose goes to zero (nonlinear model). 7 8 This use of "nonlinear" is acknowledged to be idiosyncratic.45 The NAS review (NAS, 9 2006, 198441) does not consistently apply the terminology from the 2005 Cancer Guidelines, nor 10 does it consistently distinguish the above two circumstances: " .. .the observed data are more 11 consistent with a sublinear response that approaches zero at low doses rather than a linear dose 12 response" (NAS, 2006, 198441). The point of a nonlinear model in the sense of the 2005 Cancer 13 Guidelines is that the response slope approaches zero. Both linear and nonlinear responses 14 approach zero at low dose (in the absence of background). Since the terms "linear," "sublinear," 15 and "nonlinear" invite confusion in this context, the following terminology is used in this 16 document: 17 18 ThresholdModel: There is some threshold dose T > 0 such that the probability of 19 response for any dose less than or equal to T is zero, and the probability is nonzero for 20 any dose greater than T. 21 Linear/Linear above ThresholdModel: For the linear model, the probability of response 22 is proportional to the dose. For the linear over threshold model, the probability of 23 response is zero for a dose below the threshold, and it is proportional to the excess dose 24 over the threshold otherwise. Note that under the EPA cancer guidelines, the linear 25 above threshold model is classified as a nonlinear model. 26 Nonlinear Model: Any model that is not linear. 27 Supralinear/ Supralinear above ThresholdModel: For the supralinear model, the slope of 28 the probability of response decreases as dose increases; in other words, the second 29 derivative of the response curve is negative. For the supralinear above threshold model, 45 F r o m t h e 2 0 0 5 C a n c e r G u i d e l i n e s ( U .S . E P A , 2 0 0 5 , 0 8 6 2 3 7 ) : " T h e t e r m ' nonlinear ' is u s e d h e r e i n a n a r r o w e r s e n s e t h a n it s u s u a l m e a n i n g i n t h e f i e l d o f m a t h e m a t i c a l m o d e lin g . I n th e s e c a n c e r g u i d e l i n e s , th e t e r m ' nonlinear ' re fe rs to th re sh o ld m o d e ls (w h ic h sh o w n o re sp o n se o v e r a ra n g e o f lo w d o se s th a t in c lu d e z e ro ) a n d so m e n o n th re sh o ld m o d e ls (e .g ., a q u a d ra tic m o d e l, w h ic h sh o w s so m e re s p o n s e a t a ll d o s e s a b o v e z e ro ). In th e se c a n c e r g u i d e l i n e s , a n o n l i n e a r m o d e l i s o n e w h o s e s l o p e i s z e r o a t ( a n d p e r h a p s a b o v e ) a d o s e o f z e r o .......... U s e o f n o n l i n e a r a p p ro a c h e s d o e s n o t im p ly a b io lo g ic a l th re sh o ld d o se b e lo w w h ic h th e re sp o n se is z e ro ." This document is a draftfo r review purposes only and does not constitute Agency policy. 5-55 DRAFT--DO NOT CITE OR QUOTE 1 the second derivative is negative above the threshold, and the response probability is zero 2 below the threshold. 3 Sublinear/Sublinear above ThresholdModel: For the sublinear model, the slope of the 4 probability of response increases as dose increases; in other words, the second derivative 5 of the response curve is positive. For the sublinear above threshold model, the second 6 derivative is positive above the threshold, and the response probability is zero below the 7 threshold. 8 Zero Slope at Zero Model: The slope of the response curve is zero at or above dose zero. 9 10 All of these models may be understood in an individual or population sense. According 11 to the 2005 Cancer Guidelines, the trigger for applying the basic RfD methodology for cancer 12 endpoints is sufficient evidence for the "zero slope at zero" model for the population. By 13 definition, any sublinear, supralinear, or linear model above the threshold is a zero slope at zero 14 ("ZS@Z") model. 15 The relation between individual and population models is not immediately evident. 16 Figure 5-4 shows dose-response curves of the probability of response vs. dose for different 17 models dose-response shapes. The left panel in Figure 5-4 shows a supralinear dose-response 18 curve; the rate of increase of the response probability goes down as dose increases, or in the strict 19 mathematical sense, the second derivative is negative. The middle panel shows a sublinear dose20 response curve; the second derivative is positive. In this case the slope at zero is zero (ZS@Z). 21 However, sublinearity, in the strict mathematical sense, by itself does not imply that the slope at 22 zero is zero. The probit dose-response model shown in the right graph is sublinear and has 23 positive slope at zero (the log-probit model is zero slope at zero). 24 If individuals in a population have different dose-response curves, then the population 25 dose-response curve is obtained by averaging all these dose-response curves over the population. 26 The shape of the population dose-response curve will generally be quite different from the 27 individual curves. Figure 5-5 is a simple depiction of the relationship of individual vs. 28 population dose response. The left panel in Figure 5-5 shows dose-response curves for seven 29 individuals, each with a supralinear dose-response curve above individual-specific thresholds. 30 Averaging these curves gives the dashed dose-response curve, which is nearly linear. The graph 31 on the right is similar, except that the individual dose-response curves are linear above individual 32 thresholds. The population curve is quadratic and zero slope at zero applies. This document is a draftfor review purposes only and does not constitute Agency policy. 5-56 DRAFT--DO NOT CITE OR QUOTE 1 Of course these are not the only possibilities; in general, the population dose-response 2 curve depends on (1) the distribution of individual thresholds in the neighborhood of zero, (2) the 3 dose-response curve for each individual, and (3) the dose metric. Under EPA's Cancer 4 Guidelines, the zero-slope-at-zero criterion applies strictly to ingested dose, but the other two 5 factors (distribution of individual thresholds and dose-response curve for each individual) need 6 to be established before a zero slope at zero dose can be established. Otherwise the default linear 7 extrapolation to zero approach applies. 8 On the nature or the distribution of individual thresholds, often referred to as the 9 population tolerance distribution, there is ongoing debate as to how receptor kinetics influence 10 the shape of that distribution. Even within an individual, there is a lack of consensus as to 11 whether receptor kinetics confer linear or sublinear attributes to downstream events, or whether 12 receptor kinetics, themselves, are linear, sublinear, or supralinear. Whatever the nature of the 13 form of receptor kinetics, it may have little or no influence on the ultimate population response. 14 The kinetics of receptors is in the domain of the individual, rather than the population. As 15 described previously, receptor kinetics are governed by the law of mass action, which leads to a 16 low-dose proportional response model, generally modeled by some form of Hill function, the 17 low-dose linear form being Michaelis-Menten kinetics. There is no a priori reason to believe 18 that the shape of the dose-response curve in an individual has any relationship to the shape of the 19 population response, particularly for quantal endpoints. Lutz and Gaylor (2008, 594297) present 20 an argument for considering the population response in terms of the more traditional tolerance 21 distribution, which is likely the result of more variable factors than the shape of receptor kinetics. 22 Perhaps more to the point, receptor activation is only the first of many events in the path to the 23 apical event (a tumor in this example). Because there are undoubtedly numerous additional 24 downstream events that must occur before the apical effect is observed, there are many 25 opportunities for interindividual variability to become manifest in the tolerance distribution. 26 Even at the first step, a more likely contributor to interindividual variability than the shape of the 27 response is the dose resulting in the response, as measured by the ED50 (Kmin the Michaelis 28 Menten formulation), which shifts the response curve. Factors that influence shifts in response 29 curves are generally modeled as normal or log-normal distributions and may confer a log-normal 30 shape on the population tolerance distribution, particularly if there are a number of dependent 31 sequential steps or distinct subpopulations (Hattis and Burmaster, 1994, 594301; Hattis et al., This document is a draftfor review purposes only and does not constitute Agency policy. 5-57 DRAFT--DO NOT CITE OR QUOTE 1 1999, 594299; Lu tz, 1999, 594298), although other distributions could be equally likely (Crum p 2 et al., 2010, 380192) . 3 T o see how the discussion over threshold/nonthreshold might play out for T C D D , 4 consider the equilibrium dissociation constant Kdfor T C D D , w hich measures the binding affinity 5 o f T C D D to the A h R . L o w e r values indicate higher binding affinity and (other things being 6 equal) greater risk. F o r H an/W istar rats, the value Kd = 3.9 is reported (standard deviation not 7 given); human values are reported as Kd = 9.6 7.8 (0.3 - 38.8 with 15 o f 67 donors without 8 detectable binding) (Connor and A ylw ard , 2006, 197632) . I f A h R binding is the rate-limiting 9 step for carcinogenesis, then the m ajority of a human population m ay be less susceptible than 10 H an/W istar rats, w hereas a population threshold, i f it exists, might still be w ell below the 11 H an/W istar rat threshold, given the large variab ility in the human Kd estimate (see also Section 12 6.4.2.9). The N A S contends that an A hR-m ediated mode o f action indicates a threshold dose13 response relation (N A S , 2006, 198441) . Presum ably, the value o f the threshold, i f it exists, 14 depends on the A h R binding affinity. A rguing for a population threshold in this case requires 15 two types of inform ation: 16 17 1. Th e distribution o f the individual thresholds induced by, among other things, the 18 individual Kdvalues; and 19 2. Th e dose-response function for values above the threshold induced by Kd. 20 21 W ithout this inform ation, the shape of the population dose-response curve cannot be 22 determined with any confidence and the default linear relationship applies; response probability 23 is modeled as a lin ear function of dose, for dose near zero. H ow ever, from the 2005 C a n ce r 24 Guidelines: "W hen adequate data on mode o f action provide sufficient evidence to support a 25 nonlinear mode o f action fo r the general population (em phasis added) and/or any subpopulations 26 o f concern, a different approach-- a reference dose/reference concentration that assumes that 27 nonlinearity-- is used." In current terminology, the reference dose methodology applies if there 28 is sufficient evidence supporting a "zero slope at zero" m odel; otherwise, the lin ear nonthreshold 29 model applies by default. 30 In principle, the choice between the above models could fall w ithin the purview o f dose31 response m odeling. H ow ever, standard statistical methods encounter w ell-kn ow n difficulties in This document is a draftfor review purposes only and does not constitute Agency policy. 5-58 DRAFT--DO NOT CITE OR QUOTE 1 detecting thresholds. W ithout going into detail, suffice to say that the maximum likelihood 2 estimate o f response probability when no responses are observed in a finite sample is always 3 zero. That said, some researchers have attempted to identify thresholds (A y lw a rd et al., 2003, 4 594305; M a ck ie et al., 2003, 594303) or nonlinearity (H oel and Portier, 1994, 198741) by means 5 o f parameter estimation o f appropriate models. A review o f 344 rodent bioassays on 315 6 chem icals led to the follow ing conclusion by Hoel and Portier (1994, 198741) : 7 8 W e have also found that the oft-held b elief that genotoxic compounds typically 9 follow a linear dose-response pattern and that nongenotoxic compounds follow a 10 nonlinear or threshold dose response pattern is not supported by the data. In fact 11 w e find the opposite w ith genotoxic compounds differing from linearity more 12 often than nongenotoxic compounds. 13 14 Th e choice between a linear and "zero slope at zero" model in current practice does not 15 fall under dose-response model fitting, it is made on the basis o f a structured narrative as set 16 forth in the 2005 C an ce r G u idelin es (U .S . E P A , 2005, 086237) : 17 18 In the absence o f sufficiently, scien tifically ju stifiab le mode o f action information, 19 E P A generally takes public health-protective, default positions regarding the 20 interpretation o f toxicologic and epidem iologic data: anim al tumor findings are 21 judged to be relevant to humans, and cancer risks are assum ed to conform w ith 22 low dose linearity. ... The linear approach is used when: (1) there is an absence of 23 sufficient inform ation on modes o f action or (2) the mode o f action inform ation 24 indicates that the dose-response curve at lo w dose is or is expected to be linear. 25 W here alternative approaches have significant biological support, and no 26 scientific consensus favors a single approach, an assessment may present results 27 using alternative approaches. A nonlinear approach can be used to develop a 28 reference dose or a reference concentration. 29 30 5.2.3.4.I.3. E x tra p o la tio n m e th o d . 31 Th e 2005 C an ce r G u idelin es (U .S . E P A , 2005, 086237) em phasize that the method used 32 to characterize and quantify cancer risk from a chem ical is determined by what is known about 33 the M O A o f the carcinogen and the shape o f the cancer dose-response curve. 34 The N A S w as critical o f E P A ' s decision to apply linear low-dose extrapolation for 35 T C D D cancer assessm ent in the 2003 R eassessm ent and encouraged the use o f a nonlinear 36 approach. The 2005 C an ce r G u idelin es state that a nonlinear approach should be used w hen This document is a draftfor review purposes only and does not constitute Agency policy. 5-59 DRAFT--DO NOT CITE OR QUOTE 1 "there are sufficient data to ascertain the mode o f action and conclude that it is not linear at low 2 doses and the agent does not demonstrate m utagenic or other activity consistent w ith linearity at 3 low doses." 4 Receptor modeling theory (as outlined in the 2003 Reassessm ent, Part II, Chapter 8) 5 indicates that exogenous compounds w hich operate through receptor binding m echanism s, such 6 as T C D D , w ill follow a linear dose-response binding in the 1 -1 0 % receptor occupancy region. 7 Th is theory has been supported by em pirical findings and suggests that the proximal biochem ical 8 effects (such as enzyme induction) and transcriptional reactions for T C D D may also follow 9 linear dose-response kinetics. M ore distal toxic effects could take any one of multiple forms 10 (i.e., linear, sublinear, supralinear or threshold) depending on (1) the toxic m echanism ; 11 (2) location on the dose-response curve; and (3) interactions w ith other processes such as 12 intracellular protein binding and cofactor induction/repression. 13 In the case o f T C D D , m any adverse effects experienced at lo w exposure levels have too 14 m uch data variab ility to distinguish on a statistical basis (goodness-of-fit) between dose-response 15 curve options, and whether the dose-response is linear, sublinear or supralinear. F o r tumor 16 responses, w ith the exception o f squamous cell carcinom a o f the oral m ucosa and adenomas or 17 carcinom as o f the pancreas, w h ich w ere fit w ith a linear m ultistage model, the tumor endpoints 18 in the N T P (2006, 543749) study using fem ale Sprague-D aw ley (S -D ) rats are all best fit w ith a 19 sublinear model (i.e., the m ultistage model fits to tumor incidence data w ere second or third 20 degree; see Table 5-15 and A ppendix F ). F o r all tumor incidence data from three o f the other 21 cancer bioassays that met the study inclusion criteria (K o cib a et al., 1978, 0 01818; N T P , 1982, 22 594255; Toth et al., 1979, 197109), the m ultistage model fit w as linear (first degree), w hen based 23 on either adm inistered dose or modeled blood concentrations (see A ppendix F ). F o r D e lla Porta 24 et al. (1987, 197405) , the fem ale liv e r carcinom as w ere linear (first degree), but the fem ale liv er 25 adenomas and the male liver carcinomas w ere best modeled using a second degree model (see 26 Table 5-15). 27 Another issue o f potential importance when evaluating the shape o f the dose-response 28 curve for low dose effects is the concept o f "interacting background." The concept o f interacting 29 background refers to a pathological process in the exposed population that shares a causal 30 intermediate with the toxicant being evaluated. On this issue, a recent N A S committee (N A S , 31 2009, 594307) contended that This document is a draftfor review purposes only and does not constitute Agency policy. 5-60 DRAFT--DO NOT CITE OR QUOTE 1 .. .the current E P A practice o f determining "nonlinear" M O A s does not account 2 for m echanistic factors that can create linearity at lo w dose. The dose-response 3 relationship can be linear at a lo w dose w hen an exposure contributes to an 4 existing disease process Crum p et al., 1976, 0 03 19 2 ; L u tz, 1990, 0 00399. E ffe cts 5 o f exposures that add to background processes and background endogenous and 6 exogenous exposures can lack a threshold if a baseline level o f dysfunction occurs 7 without the toxicant and the toxicant adds to or augments the background process. 8 Thus, even small doses may have a relevant biologic effect. That may be difficult 9 to measure because o f background noise in the system but may be addressed 10 through dose-response m odeling procedures. H um an variab ility w ith respect to 11 the individual thresholds for a nongenotoxic cancer m echanism can result in 12 linear dose-response relationships in the population (Lu tz, 2001, 053426; N A S , 13 2009, 5 94307. 14 15 A h R activation could be considered a causal intermediate in several disease processes. 16 R ecen t studies have linked A h R activation in the absence o f exogenous ligand to a multitude o f 17 biological effects, ranging from control o f m am m ary tum origenesis to regulation o f 18 autoimmunity (review ed in H ahn et al., 2009, 548725) . W h ile the level o f background activation 19 o f A h R by endogenous compounds (or exogenous compounds other than T C D D ) in the human 20 population is unknown, given the ubiquitous nature o f several o f the known endogenous and 21 exogenous A h R ligands, it is reasonable to assum e that a certain baseline level o f A h R activation 22 exists in the population. Th e degree to w h ich T C D D exposure augments this baseline level o f 23 A h R activation is unknown. 24 The 2005 Cancer Guidelines (U .S . E P A , 2005, 086237) recommend that the method used 25 to characterize and quantify cancer risk from a chem ical be determined by what is known about 26 the mode o f action o f the compound and the shape o f the cancer dose-response curve. The linear 27 approach is used if there is sufficient evidence supporting linearity or if the mode o f action is not 28 understood (U .S. E P A , 2005, 086237) . In the case o f T C D D , (1) the mode o f action o f T C D D - 29 induced carcinogenesis beyond potential A h R activation is unknown; (2) information is lacking 30 to determine the shape o f the dose-response curves at lo w doses for various adverse endpoints 31 (including cancer) in hum ans or experimental anim als; (3) there is undoubtedly a certain level o f 32 interacting background (i.e., A h R activation by endogenous ligands) in the human population; 33 (4) m any o f the rodent cancer dose-response relationships (K o cib a et al., 1978, 001818; N T P , 34 1982, 594255; Toth et al., 1979, 197109) are consistent w ith low-dose linearity (first degree 35 m ultistage model fit) w hen based on either adm inistered dose or modeled blood concentrations; This document is a draftfor review purposes only and does not constitute Agency policy. 5-61 DRAFT--DO NOT CITE OR QUOTE 1 and (5) higher human interindividual variab ility compared to experimental rodents w ill tend to 2 shift the shape of the dose-response towards linear (relative to rodents). None of these 3 suggestions of linearity, however, is conclusive (see next section for additional detail). The true 4 shape of the dose-response curve remains unknown. Therefore, in the absence of sufficient 5 evidence to the contrary or evidence to support nonlinearity, to estimate human carcinogenic risk 6 associated with T C D D exposure E P A assumed a linear low-dose extrapolation approach. 7 8 5.2.3.4.I.4. D iscu ssion o f lo w -d o s e lin e a rity . 9 A n y quantitative estimation of carcinogenic risk associated with T C D D exposure requires 10 low -dose extrapolation of high dose experimental and epidem iologic data. Unfortunately, 11 despite the availability of the extensive database on the biological effects of T C D D , the shape of 12 the dose-response curve in the low -dose region is not know n. T h is situation is not unique to 13 T C D D . F o r most carcinogens the available biological data do not provide sufficient m echanistic 14 inform ation to determine the shape of the dose-response relationship at doses below the levels 15 w here direct experimental or epidem iologic data are available. E P A ' s G u idelin es for Carcinogen 16 R is k A ssessm ent (2005, 086237) recognize this situation and describe approaches the A g en cy 17 uses for dose response assessm ent in cancer risk assessm ents depending on the available 18 scientific database. E P A ' s b asic approach m akes a distinction between " low -dose linear" and 19 "nonlinear" dose response patterns. T h is distinction is important to understand as it addresses 20 the potential response at lo w dose, not the em pirical pattern o f response seen in the available 21 (often high dose) tumor data. T o put matters sim ply, under a low -dose-linear model, the 22 estimated risk due to the carcinogen exposure is approximately proportional to the dose received 23 (at lo w dose). In mathematical terms, a low -dose-linear model is one w hose slope is greater than 24 zero at a dose o f zero (U .S . E P A , 2005, 086237; footnote, p. 1-11). Im portantly, a low-dose25 linear model need not be linear at higher doses, and this is consistent w ith upward curving 26 responses (e.g., linear-quadratic) and downward curving (plateauing) responses that may be seen 27 various cancer studies. In E P A ' s terminology a "nonlinear" dose-response, refers to situations 28 w here there is not a linear component in the response at low-dose. In this context, a " nonlinear 29 m odel" is one w hose slope is zero at (and perhaps above) a dose o f zero (ibid). N onlinear 30 response patterns can include threshold models where there is no response below a defined dose This document is a draftfor review purposes only and does not constitute Agency policy. 5-62 DRAFT--DO NOT CITE OR QUOTE 1 level, or other patterns w here response at lo w dose otherwise decreases rapidly as compared to a 2 low-dose-linear model. 3 A s stated in the previous section, the low-dose linear approach for the T C D D 4 carcinogenicity assessment in this document is based on E P A ' s scientific baseline inference 5 (" default") regarding dose-response modeling. E P A believes that the mode o f action is not 6 known, so is using the default linear extrapolation approach specified by E P A ' s cancer 7 guidelines. 8 Nonetheless, there are biological data on T C D D that help inform the appropriateness of 9 low-dose-linear risk extrapolation for this compound. Furthermore, there is utility in 10 sum m arizing scientific reasoning that supports the approach o f low -dose linearity as an 11 appropriate scientific baseline inference (" default") for carcinogen risk assessment. 12 Th e issues pertaining to low -dose linearity w ere discussed in the report o f a recent state- 13 of-the-science w orkshop on issues in low -dose risk extrapolation held by U .S . E P A and Johns 14 H o p kins R is k Scien ce and P u b lic P o licy Institute in 2007 (W hite et al., 2009, 622764) . T h is 15 report states: 16 17 Th e com plex m olecular and cellu lar events that underlie the actions o f agents that 18 lead to cancer and noncancer outcomes are lik e ly to be both linear and nonlinear. 19 A t the human population level, however, biological and statistical attributes tend 20 to smooth and linearize the dose-response relationship, obscuring thresholds that 21 might exist for individuals. M ost notable o f these attributes are population 22 variability, additivity to preexisting disease or disease processes, and background 23 exposure-induced disease processes; m easurem ent error also undoubtedly 24 contributes to this phenomenon. Th e linear appearance o f the population-level 25 dose-response function does not presume that the dose-response relationship is 26 necessarily linear for individuals (Lu tz, 1990, 000 39 9 ; 2001, 0 53426; L u tz et al., 27 2005, 087763), but may reflect a distribution o f individual thresholds. These 28 attributes are lik e ly to explain, at least in part, w hy exposure-response m odels o f 29 the relationship between cancer or noncancer health effects and exposure to 30 environmental toxicants w ith relatively robust human health effects databases at 31 ambient concentrations (e.g., ozone and particulate matter air pollution, lead, 32 secondhand tobacco smoke, radiation) do not exhibit evident thresholds, even 33 though the M O A s include nonlinear processes for key events N R C (2005); 34 U .S. E P A (2006, 088089; 2006, 157071; 2006, 090110) ; U .S. D H H S (2004, 35 056384) . 36 This document is a draftfor review purposes only and does not constitute Agency policy. 5-63 DRAFT--DO NOT CITE OR QUOTE 1 Original arguments in favor of low-dose linearity for carcinogen risk assessment 2 (including for ionizing radiation, as developed from human data) are based on the occurrence of 3 damage (often termed "hits") to DNA and the inference that resulting mutations would 4 contribute to cancer development. These arguments envisioned direct damage to DNA; 5 however, based on subsequent advances in mechanistic understanding, damage to DNA by 6 "secondary" reactive molecules (not just direct hits to DNA by radiation or other agents) is also 7 considered to play a major role. TCDD is not thought to produce DNA damage directly. 8 However, DNA damage may result subsequent to increased formation of reactive molecules 9 (reactive oxygen species (ROS) and metabolites of endogenous compounds). Thus, the presence 10 of low-dose linearity by this pathway would depend on whether such reactive molecules were 11 produced at low dose and whether that increased formation was proportional to dose. If that 12 were the case for TCDD, which is still unknown, arguments in favor of low-dose linearity 13 remain similar to those for direct-acting agents. 14 The kinetics of ligand receptor binding, and then the attachment of a receptor/ligand 15 complex to a promoter region of DNA are biochemical processes where low-dose linearity can 16 occur. Simple receptor binding interactions are often modeled using Michaelis-Menten 17 relationships which are linear at low dose. Thus, the early key events in a process of a receptor18 mediated toxicity pathway may often be expected to be low-dose linear. However, as in any 19 toxicity process, the ultimate shape of the dose-response relationship for an apical46 toxicity 20 endpoint will depend on all the processes involved, not just receptor kinetics. These issues were 21 considered by NRC (NAS, 2009, 594307) which included as an indication for non-threshold 22 dose response: "The fact that in receptor-mediated events, even at very low doses a chemical can 23 occupy receptor sites and theoretically perturb cell functions (such as signal transduction or gene 24 expression) or predispose the cell to other toxicants that bind to or modulate the receptor systems 25 (such as organochlorines and the aryl hydrocarbon receptor or endocrine disruptors and 26 hormonal binding sites)." The role of these factors for TCDD has not been fully elucidated. 27 Two other factors supporting low-dose linearity discussed in the workshop described by 28 White et al. (2009, 622764) are additivity to background processes (dose additivity) and the 29 magnitude of human heterogeneity. 46 A n apical endpoint is an observable outcome in a w hole organism, such as a clinical sign or pathologic state, that is indicative o f a disease state that can result fro m exposure to a toxicant (N A S, 2007). This document is a draftfo r review purposes only and does not constitute Agency policy. 5-64 DRAFT--DO NOT CITE OR QUOTE 1 Concerning dose additivity, Crump et al. (1976, 003192) argued in the context of a 2 carcinogenic response that if the carcinogenic process resulting from exposure to an exogenous 3 agent (e.g., TCDD) is already operant in causing background responses, then the effect of the 4 exposure is to augment this process in a dose-additive fashion. The additional response caused 5 by the exposure is expected to increase approximately linearly with exposure at low exposures 6 (i.e., be low-dose linear). The NRC Science and Decisions report (NAS, 2009, 594307) 7 examined the issue of additivity to background, in particular calling attention to a need for 8 systematic consideration of endogenous processes related to disease development as well as 9 additivity to other exogenous exposures.47 While the baseline activity (unexposed to exogenous 10 agents) of AhR is not well understood, the effects of exogenous agents need to be considered in 11 terms of how they add on to or modulate baseline physiological processes instead of considering 12 TCDD or other exogenous ligands to be "acting in a vacuum." 13 The issue of human heterogeneity relative to the rodents used in bioassays has been 14 discussed at length in the literature and will not be repeated here (see also relevant text in 15 Section 5.2.3.4.1.3). However, as discussed by NAS (2009, 594307), even in situations where 16 processes thought to be nonlinear are precursors to the development of cancer in test animals, a 17 different situation may result in humans: "However, given the high prevalence of those 18 background processes, and given the multitude of chemical exposure and high variability in 19 human susceptibility, the results may still be manifested as low-dose linear dose-response 20 relationships in the human population." The population dose-response will be influenced by 21 heterogeneities in the population that affect internal dose as well as response. First, even if there 22 is strong curvilinearity in the dose-response curve in the dose range of relevance to human 23 exposures, there may be large differences across individuals in the doses at which transitions in 24 the shape of the dose-response curve occur. Greater variability in response to exposures would 25 be anticipated in heterogeneous populations than in inbred laboratory species under controlled 26 conditions (due to, e.g., genetic variability, disease status, age, and nutrition). The effect of 27 increased heterogeneity will be a broadening of the dose-response curve (i.e., less rapid fall-off 28 of response with decreasing dose) in diverse human populations and, accordingly, a greater 47 It may be noted that w hen there are m ultip le exogenous exposures, it may be d iffic u lt to ascertain w hich exposure came firs t. H ow ever, the p oint is that if a com bination o f endogenous and exogenous factors is operative in causing biological response, then an additional sm all, dose additive, exposure can be predicted to cause a proportionate change in response. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-65 DRAFT--DO NOT CITE OR QUOTE 1 potential for risks from low-dose exposures (Lutz et al., 2005, 087763; Zeise et al., 1987, 2 060867). The degree to which heterogeneity must be increased to "linearize" sublinear 3 responses of varying degrees has not yet been established. 4 Interpreting the shape of animal bioassay dose-response model fits always involves 5 assumptions about the shape of the response in the unobserved range (i.e., low dose). Cancer 6 bioassays can provide relatively little information on actual dose-response patterns below the 7 point of departure. However, it is generally not possible to either exclude or affirm low-dose 8 linear components statistically based upon empirical modeling of the dose-response data.48 9 Dose-response modeling can, however, be useful in describing the size of a linear component in 10 the response that is compatible with study data. As an example, NRC (NAS, 2006, 198441) 11 advised EPA to examine the results of the NTP (2006, 543749) study as indicating nonlinearity 12 of the observed tumor response. Among the tumors seen in the NTP bioassay, the dose-response 13 shape for cholangiosarcoma is notably curvilinear in the dose range of the observed tumor 14 response. Figure 5-6 shows the multistage modeling of the cholangiosarcoma data from the NTP 15 bioassay. The BMDL is calculated at an extra risk of 0.01. Even though the MLE dose response 16 is nonlinear (1st-degree coefficient is zero), the dose-response curve pertaining to the statistical 17 upper bound on risk (calculated here as the 95% lower confidence bound on dose) is 18 approximately linear below the 0.01 benchmark level and roughly superposes on the EPA default 19 linear extrapolation (see Figure 5-6B). For the oral squamous cell carcinoma (SCC) tumor data 20 (plot not shown), the MLE dose-response curve itself displays low-dose linearity (1st-degree 21 coefficient is greater than zero) and the EPA low-dose linear extrapolation is indistinguishable 22 from the upper bound curve. These observations are consistent with the findings of 23 Subramaniam et al. (2006), that for the large majority of chemicals, straight line extrapolation of 24 risk from the BMDL provides slope factor values very similar to those obtained by using an 25 upper bound on the multistage model risk estimate. Furthermore, in this assessment, EPA has 26 chosen to derive oral slope factors based on combined tumor incidence whenever possible, 27 modeling them under an assumption of independence. A Bayesian analysis is used in this 28 document to develop PODs based on combined tumor risk across the significantly elevated 29 tumor types observed in this bioassay (see Section 5.2.3.2.3.2). As a result of this analysis, the 48 E P A policy is to allo w fo r low-dose line arity in the m odeling o f tum ors if a non-linear M O A has not been established. This document is a draftfor review purposes only and does not constitute Agency policy. 5-66 DRAFT--DO NOT CITE OR QUOTE 1 central estimate for the composite dose-response curve shows little curvilinearity and the MLE 2 dose-response curve is substantially linear below a 0.1 extra risk level (see Figure 5-7A and 3 5-7B; see also Section 5.2.3.2.6.11). 4 The results here provide a comparison of EPA's linear (straight line) dose-response 5 estimates with the degree of linearity seen in the fitted dose-response curves and the statistical 6 upper bounds on these curves. To do this the fitted model needs to allow for the possibility of 7 both curvilinearity at high dose and linearity at low dose. The multistage model has these 8 properties, which is among its advantages for application in carcinogen risk assessment. Most 9 other models commonly used to fit data in the observed range do not have this property.49 10 One other issue relative to the determination of linearity arises in the visual interpretation 11 of dose-response plots. The common practice of plotting receptor kinetics data on semi 12 logarithmic plots for scale convenience has unfortunately led to difficulties in the interpretation 13 of the shape of these relationships. An example is presented using the modeling study of Kohn 14 and Melnick (2002, 199104), which was cited by NRC (NAS, 2006, 198441) in its review of 15 EPA's dioxin assessment as an example of nonlinear behavior at low dose: "Response is a 16 function of the number of occupied and activated receptors, which typically exhibit steep dose17 response relationships. For example, Kohn and Melnick (2002, 199104) modeled the shape of 18 the dose-response relationship for receptor-mediated responses, using the estrogen receptor and 19 various xenoestrogens as a model receptor and ligands, respectively. The model included a 20 variety of assumptions with regard to receptor number, ligand binding affinity, and partial 21 agonist activities, yet in every instance clear sublinear responses were observed at low doses." 22 However, as shown in Figure 5-8, the apparent strong upward curvature of the low-dose 23 relationship is no longer seen when the results are plotted on an arithmetic scale. Instead, the 24 system may be seen as providing an example of close to linear behavior in the low-dose region. 25 49 The standard H ill models do not: A H ill m odel is only linear at low dose w hen the H ill parameter is equal to 1 (and in that case the H ill model is linear over the fu ll dose range u n til the high dose region o f " saturation" where the km parameter results in downward curvature). Thus, w hile the H ill m odel is a valuable to o l fo r fittin g data in the observed experim ental range, it is not help ful in illustratin g the potential fo r low-dose response. However, some have considered a dose-additive version o f the H ill model w hich w ould allow fo r low-dose line arity. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-67 DRAFT--DO NOT CITE OR QUOTE 1 5.2.3.4.I.5. Consideration o fnonlinear methods. 2 While the 2005 Cancer Guidelines deem linear extrapolation to be most appropriate for 3 TCDD, EPA has carefully considered the NAS recommendation to provide risk estimates using 4 both linear and nonlinear methods. 5 The 2005 Cancer Guidelines state 6 7 For cases where the tumors arise through a nonlinear mode of action, an oral 8 reference dose or an inhalation reference concentration, or both, should be 9 developed in accordance with EPA's established practice for developing such 10 values ... This approach expands the past focus of such reference values 11 (previously reserved for effects other than cancer) to include carcinogenic effects 12 determined to have a nonlinear mode of action. 13 14 In this section, EPA presents two illustrative examples of RfD development for 15 carcinogenic effects of TCDD. Each of these examples focuses on data derived from animal 16 bioassays as described in Section 2.4.2. 17 18 5.2.3.4.1.5.1. Illustrative RfDs based on tumorigenesis in experimental animals. 19 TCDD has been shown to be a multisite carcinogen in both sexes of several species of 20 experimental animals. It also has been shown to be carcinogenic to humans. Most of the 21 available quantitative human epidemiologic data related to TCDD carcinogenesis are for all 22 cancer mortality. Mortality is a frank effect and is generally considered to be inappropriate for 23 RfD development, therefore, the illustrative example below utilizes available evidence from 24 experimental animals. Table 5-20 presents candidate PODs and RfDs for TCDD carcinogenicity 25 based on combined tumor responses from the animal bioassays described in Section 2.4.2. The 26 PODs from the NTP (2006, 549255; 2006, 543749) and Kociba et al. (1978, 001818) animal 27 studies were derived from Bayesian multitumor dose-response modeling (as described in 28 Section 5.2.3.2, Table 5-17) using a BMR of 1%. Because only TCDD-induced liver tumors 29 were reported by Toth et al. (1979, 197109), the BMR of 1% (POD) from that study was 30 generated using a first degree linear multistage model (see Table 5-15). TCDD-induced liver 31 tumors were reported by Della Porta et al. (1987, 197405), with the male mouse producing the 32 lowest BMR of 1% (POD) using a second degree linear multistage model (see Table 5-15). 33 Following BMD modeling, BMDLheds were then estimated (see Tables 5-16 and 5-17) using the This document is a draftfor review purposes only and does not constitute Agency policy. 5-68 DRAFT--DO NOT CITE OR QUOTE 1 T C D D whole-blood-concentration dose m etric from the Em o n d model as described in Section 3. 2 Th e illustrative R f D s w ere derived by dividing the B M D L HEDs by appropriate uncertainty 3 factors. In each instance, a total U F o f 30 w as applied, com prising factors o f 3 for the 4 toxicodynam ic component o f the interspecies extrapolation factor ( U F A) and a factor o f 10 for 5 human interindividual variab ility ( U F H). 6 A s shown in Table 5-20, the illustrative R fD s for TC D D -induced tumors range from 7 3.6E-11 for liver and lung tumors in male m ice (N TP , 1982, 594255) to 1.0E-9 for adrenal 8 cortex, tongue and nasal/palate tumors in m ale rats (K o c ib a et al., 1978, 001818) . T h is 9 illustrative R fD range for T C D D tumorigenesis falls within the range o f candidate R fD s for 10 noncancer T C D D effects presented in Tab le 4-5. 11 12 5.2.3.4.I.5.2. Illustrative R fD s based on hypothesized key events in T C D D ' s M O A s for liver 13 and lung tumors. 14 A s described in Section 5.1, most evidence suggests that the m ajority o f toxic effects o f 15 T C D D are mediated by interaction w ith the A h R . E P A considers interaction w ith the A h R to be 16 a necessary, but not sufficient, event in T C D D carcinogenesis. Th e sequence o f key events 17 follow ing binding o f T C D D to the A h R and that ultim ately leads to the development o f cancer is 18 unknown. W h ile the mode o f action o f T C D D in producing cancer has not been elucidated for 19 any tumor type, the best characterized carcinogenic actions o f T C D D are in rodent liver, lung, 20 and thyroid. The hypothesized sequence o f events follow ing T C D D interaction w ith the A h R is 21 m arkedly different for each o f these three tumor types. A dditionally, no detailed hypothesized 22 mode o f action information exists for any o f the other reported tumor types. 23 Th e endpoints selected for this illustration w ere evaluated to provide insight into the 24 quantitative relationships between tumor development and precursor events in TC D D -induced 25 carcinogenesis. The endpoints described below may or may not be biologically adverse in 26 them selves; the intent herein w as to consider T C D D -in d u ced biochem ical and cellular changes 27 that could lead to subsequent tumor development. 28 In the following exercise, illustrative R fD s were derived for key events in T C D D ' s 29 hypothesized modes o f action in the liver and lung. N o appropriate dose-response data were 30 identified for key events in T C D D 's hypothesized M O A for thyroid tumors in a This document is a draftfor review purposes only and does not constitute Agency policy. 5-69 DRAFT--DO NOT CITE OR QUOTE 1 sex/species/strain that has been shown to develop thyroid tumors (i.e., female B 6 C 3 F 1 m ice and 2 m ale and fem ale Osborne-M endel rats (N T P , 1982, 594255)). 3 A s this is an illustrative exercise only, only studies that were originally identified in 4 Section 2 for potential noncancer dose-response modeling were evaluated here (see Section 2.4.2 5 for study details). There may be additional studies available in the literature that would further 6 inform the dose-response assessment o f these endpoints. 7 Additionally, for animal model consistency, only results from studies conducted in 8 female S-D rats are presented here. The majority o f the available information on T C D D 9 carcinogenicity (and T C D D carcinogenic precursor events) comes from studies conducted in 10 fem ale S -D rats and the most recent T C D D carcinogenicity study w as conducted in fem ale S-D 11 rats (N T P , 2006, 197605) . W h ile both K o cib a et al. (1978, 001818) and N T P (2006, 543749) 12 have conducted T C D D carcinogenicity studies in fem ale S -D rats, different substrains w ere used; 13 this difference in substrain m ay have resulted in the different carcinogenic responses reported 14 from the two studies. W h ile the carcinogenicity o f T C D D in fem ale S -D rats has been w ell 15 characterized, this anim al model does not exhibit the full suite o f tumor responses reported for 16 T C D D (for instance, fem ale S -D rats have not been shown to develop thyroid tumors). 17 A dditionally, the most sensitive single tumor response in fem ale S-D rats from N T P (2006, 18 543749) is squamous cell carcinom a o f the oral m ucosa (see Section 5.2.3.2), a tumor type for 19 w h ich no mode o f action inform ation exists. Therefore, the illustrative R fD s described below 20 may not be protective against all tumor types. 21 F o r each endpoint, P O D s for illustrative cancer R fD development w ere identified as 22 described for the noncancer R fD derivation in Section 4. B rie fly, for the endpoints identified 23 below , the N O A E L HEDs and/or L O A E L HEDs w ere determined based on E P A analysis o f the 24 original data presented by the study author (see Section 2.4.2 for details) and by application of 25 the Em o n d P B P K m odels as described in Section 3.3.4. B M D L HEDs w ere determined as 26 described in Section 4.2 for all data sets amenable to B M D modeling. M odeling outputs for the 27 endpoints are presented in Appendices E and G as noted in Table 5-21. The illustrative R fD s 28 were derived by dividing the P O D by appropriate uncertainty factors as indicated in Table 5-21. 29 5.2.3.4.1.5.2.1. Liver tumors. 30 Figure 5-9 presents one hypothesized mode of action for TC D D -induced liver tumors in 31 rats. T C D D activation o f the A h R leads to a variety o f changes in gene expression, including This document is a draftfo r review purposes only and does not constitute Agency policy. 5-70 D R A FT-- DO NOT C IT E OR Q U O TE 1 increased C Y P 1 A 1 m R N A and subsequent increases in C Y P 1 A 1 activity. These alterations in 2 gene expression are hypothesized to lead to hepatotoxicity, follow ed by compensatory 3 regenerative cellular proliferation and subsequent tumor development. The details o f the 4 m echanism o f TC D D -in d u ced hepatotoxicity have not been fully determined but both C Y P 5 induction and oxidative stress have been postulated to be involved (M aronpot et al., 1993, 6 198386; V ilu k se la et al., 2000, 198968) . A dditionally, oxidative D N A damage has been 7 im plicated in liv e r tumor promotion (U m em ura et al., 1999, 198001) . Th e enhanced cell 8 proliferation arising from either altered gene expression or hepatotoxicity, or both, could be the 9 principal factor leading to promotion o f hepatocellular tumors (W hysner and W illiam s, 1996, 10 197556) . 11 A dose-response relationship exists for T C D D -m ed ia ted hepatotoxicity, and this parallels 12 the dose-response relationship for tumor formation (or formation o f foci o f cellu lar alteration as a 13 surrogate o f tumor formation). H ow ever, the dose-response relationship for other 14 T C C D -in d u c e d responses such as enhanced gene expression is different from the dose-response 15 for tumor formation in terms o f both efficacy and potency (see Popp et al. (2006, 197074) for 16 review ). 17 A representative endpoint for each o f the hypothesized key events follow ing A h R 18 activation for T C D D -in d u c e d liv e r tumors w as identified and is shown in Fig u re 5-9. Illustrative 19 R f D s based on each representative endpoint are shown in Table 5-21. 20 21 5.2.3.4.I.5.2.2. Lung tumors. 22 F ar less is known about T C D D ' s mode o f action in the lung. Figure 5-10 presents two 23 hypothesized modes o f action for T C D D -in d u c e d lung tumors in rats. The first hypothesized 24 mode o f action o f T C D D in the lung involves disruption o f retinoid homeostasis in the liver. 25 Retinoic acids and their corresponding nuclear receptors, the R A R s and the R X R s , w ork together 26 to regulate cell growth, differentiation, and apoptosis. It is hypothesized that T C D D , through 27 activation o f the A h R , can affect parts o f the complex retinoid system and/or other signaling 28 systems regulated by, and/or cross-talking with, the retinoid system (reviewed in (N ilsson and 29 H akansson, 2002, 548746)). Th ese effects are then hypothesized to lead to lung tumor 30 development, how ever the mechanism s underlying this hypothesis are not well-defined. The 31 second hypothesized m echanism for the carcinogenic action o f T C D D in the lung is through This document is a draftfo r review purposes only and does not constitute Agency policy. 5-71 D R A FT-- DO NOT C IT E OR Q U O TE 1 induction of metabolic enzymes. Through activation of AhR and subsequent induction of 2 metabolizing enzymes (such as CYP1A1), TCDD may enhance bioactivation of other 3 carcinogens in lung (Tritscher et al., 2000, 197265). However, there are few studies to support 4 this hypothesis. 5 Representative endpoints could only be identified for two of the hypothesized key events 6 following AhR activation for TCDD-induced lung tumors. These endpoints are presented in 7 Figure 5-10. Illustrative RfDs based on each of these two representative endpoints are shown in 8 Table 5-21. There is insufficient information to form any conclusions on the quantitative 9 progression to tumorigenicity or on the relative protection afforded by preventing the key events 10 shown. 11 12 5.2.3.4.I.5.2.3. Limitations o f illustrative RfDs based on hypothesized key events in TCDD's 13 MOAsfor liver and lung tumors. 14 A trend for increasing RfD values that follows the progression of endpoints towards the 15 production of tumors is evident. However, there are a number of factors that prevent making 16 strong conclusions based on this exercise. These limitations include the following 17 18 This example addresses only two tumor types in one species, strain and sex (female S-D 19 rats), with little information available on the hypothesized mode of action for lung 20 tumors. No mode of action information is available for the most sensitive tumor type in 21 this animal model (squamous cell carcinoma of the oral mucosa). Therefore, it is 22 possible that the illustrative RfDs presented in this example would not be protective 23 against all tumor types in female S-D rats. Importantly, other animal models have been 24 shown to be more sensitive to TCDD-induced carcinogenesis based on combined tumor 25 analysis (see Section 5.2.3.2); an RfD based on tumorigenesis in this animal model may 26 not be protective against tumorigenesis in other, more sensitive, animal models (or, by 27 extension, in humans). 28 Several of the BMDLs are based on poorly-fitting models, such that the RfD is based on 29 a LOAEL (or LOEL), which is not a particularly good measure for comparison across 30 endpoints (e.g., LOAELs are dependent on dose spacing in bioassays). Furthermore, the 31 hepatotoxicity BMDL based on a dichotomous 10% BMR, is not directly comparable to 32 all the other BMDLs based on a continuous 1 standard-deviation BMR (Crump, 2002, 33 035681). In addition, as the earlier effects (CYP induction, cellular proliferation) are not 34 considered to be necessarily adverse in themselves, the BMR of 1 standard-deviation 35 from the mean may not be the best choice for determining a POD based on biological 36 signficance. The use of the 1 standard-deviation BMR for the illustrative examples is 37 primarily for comparison on an equal-magnitude-of-response basis across endpoints. This document is a draftfor review purposes only and does not constitute Agency policy. 5-72 DRAFT--DO NOT CITE OR QUOTE 1 The endpoints selected as representative of each hypothesized key event may not be the 2 most appropriate choices. These particular endpoints were chosen because they were the 3 most sensitive indicator (i.e., lowest POD) from the available data or were the only 4 available choice based on a lack of data for other effects related to the hypothesized key 5 event. 6 The optimum timing of these events may not be reflected in the endpoints selected. 7 Almost certainly, changes in gene expression are early events, such that a single 8 exposure should be relevant, as in the mRNA changes reported after a single TCDD 9 exposure (Vanden Heuvel et al., 1994, 594318), although it is not known whether the 10 magnitude of these changes would be altered after longer-term exposure, or whether 11 longer-term exposure would be more relevant to downstream events. Similarly, single 12 exposures for induction of CYP enzymes would seem to relevant as a measure of the 13 immediate effect, but it may be longer-term repeated CYP activity that is important for 14 longer-term downstream events; Table 5-21 shows a nominal order-of-magnitude 15 difference in effect levels for similar effect magnitudes (ca. 20-fold) from single 16 exposures (Kitchin and Woods, 1979, 198750) and long-term exposures (53-weeks; 17 NTP, 2006, 543749). The relevant exposure durations for oxidative stress and later 18 effects are longer term, so a measurement of oxidative stress at 90-days in a rodent may 19 be appropriate; Wyde et al. (2001, 198575) suggest that induction of 8-oxo-dG DNA 20 adducts are a result of longer-term oxidative stress because of the lack of effect of single 21 exposures. Hepatotoxicity and hepatocellular proliferation events would appear at 22 successively later times, but the effective exposure levels would depend heavily on the 23 endpoints chosen to represent those events and the time at which they were measured. 24 The toxic hepatopathy endpoint reported in NTP (2006, 543749), is a general measure of 25 mild to moderate liver toxicity, but is measured only at the end of the study when tumors 26 have already appeared. Hepatocyte hypertrophy, measured at 31weeks may be more 27 duration-relevant, but may not indicate actual hepatocellular toxicity. 28 The lowest of the tested doses may well be much higher, given that all animal diets are 29 contaminated to a certain extent by TCDD, resulting in initial TCDD body burdens in all 30 animals. Vanden Heuvel et al. (1994, 594318) reported TCDD liver concentrations in 31 control animals almost as high as for the low-dose group, which could equate to a 32 significant increase in the actual exposure experienced by the low-dose group. A similar 33 effect on the low-dose group (0.45 ng/kg) in Kitchin and Woods (1979, 198750) is 34 possible, although they did not report control animal tissue concentrations. Higher 35 exposure levels or longer-term exposures would not be affected to the same degree, as 36 administered TCDD levels would likely be large compared to initial body burden or low37 level feed stock exposure. 38 39 Given the limitations described above, establishing an unambiguous progression of 40 effects is extremely problematic given the lack of sufficient data. Identifying a RfD that could 41 be considered to be protective against tumorigenesis in humans based on these data and models 42 is subject not only to the determination of effective low doses for the RfDs in Table 5-21 but also This document is a draftfor review purposes only and does not constitute Agency policy. 5-73 DRAFT--DO NOT CITE OR QUOTE 1 to the determination of effective exposures that could be considered to be protective of all other 2 tumor types in female S-D rats as well as all other animal models. The latter would entail 3 identifying precursors that are sufficient in themselves for progression to tumorigenesis for all 4 tumor types. Given the disparate sequence of hypothesized key events following TCDD-induced 5 AhR activation for the tumor types for which some information is available, AhR 6 binding/activation is the only key event that is likely to be shared across tumor types. No 7 appropriate quantitative data on AhR binding/activation by TCDD in relevant animal models 8 were located; therefore, an illustrative RfD based on TCDD AhR activation could not be 9 developed. 10 Simon et al. (2009, 594321) present a similar analysis for the liver tumors observed in the 11 NTP (2006, 543749) study, showing a progression of effects from early biochemical events to 12 irreversible liver toxicity, culminating in tumorigenesis. While illustrative of the putative tumor13 promoting MOA for TCDD, the limitations of using such an approach within the context of an 14 assessment of the overall carcinogenic risk of TCDD as detailed above still apply. Simon and 15 colleagues also present RfDs for liver tumors and several precursor endpoints. All the RfDs 16 presented in Simon et al. (2009, 594321) are essentially equivalent and are 1 to 3 orders of 17 magnitude higher than the RfDs for equivalent endpoints presented in Table 5-21. These 18 discrepancies are partly due to the fact that the Emond PBPK models (Emond et al., 2004, 19 197315; Emond et al., 2005, 197317; Emond et al., 2006, 197316; see also Section 3.3.4) used in 20 this document predicts lower TCDD intakes for similar tissue concentrations than the CADM 21 kinetic model (Aylward et al., 2005, 197014; Carrier et al., 1995, 197618) used by Simon and 22 colleagues. However, a larger contributor to these discrepancies is the use of a chemical-specific 23 adjustment factor (CSAF) of 0.1 for the toxicodynamic component of the interspecies 24 uncertainty factor by Simon et al. (2009, 594321), while EPA used an uncertainty factor of 3. 25 EPA does not find that the in vitro evidence presented by Simon et al. in support of a CSAF of 26 0.1 for interspecies toxicodynamics meets the burden of proof necessary for a reduction in this 27 uncertainty factor. 28 This document is a draftfor review purposes only and does not constitute Agency policy. 5-74 DRAFT--DO NOT CITE OR QUOTE 1 5.3. DERIVATION OF THE TCDD ORAL SLOPE FACTOR AND CANCER RISK 2 ESTIMATES 3 EPA was able to derive candidate OSFs for all cancer mortality from human 4 epidemiologic studies as well as for individual and combined tumor incidence from rodent 5 cancer bioassays. Each of these studies was selected for TCDD dose-response modeling using 6 the study inclusion criteria outlined in Section 2. The derivation of these OSFs can be found for 7 the epidemiologic data in Section 5.2.3.1 and for the rodent bioassay data in Section 5.2.3.2. 8 The OSFs based on epidemiologic studies from three cohorts ranged from 3.75 x 105to 9 2.5 x 106 per mg/kg-day (see Tables 5-1 and 5-3). For the animal data, OSFs based on 10 individual tumors were developed for 28 study/sex/endpoint combinations, and the results ranged 11 from 1.8 x 104 to 5.8 x 106 per mg/kg-day (see Table 5-16). The OSFs based on combined 12 tumors were developed for 7 study/sex combinations, and the results ranged from 3.2 x 105to 13 9.4 x 106 per mg/kg-day (see Table 5-17). Figure 5-11 demonstrates the range of these OSFs in 14 units of per mg/kg-day. The human study OSFs are shown at the far left of the figure, and the 15 rodent endpoints are arranged by species to the right. For comparison with the other studies, the 16 OSF from Cheng et al. (2006, 523122) is based on a 1 x 10-6 risk level (Table 5-3). 17 As recommended by expert panelists at EPA's 2009 Dioxin Workshop (U.S. EPA, 2009, 18 522927) and in the 2005 Cancer Guidelines (U.S. EPA, 2005, 086237), EPA has chosen to give 19 higher consideration to the human epidemiologic data rather than the animal bioassay data in 20 developing an OSF for TCDD. Candidate OSFs derived from the human data are consistent with 21 the animal bioassay OSFs; specifically, the human OSFs fall within the same range as the animal 22 bioassay OSFs. Because all the human and animal studies were considered to be of high quality 23 and yielded similar ranges of OSFs, EPA has chosen to rely on the epidemiologic data for OSF 24 derivation. 25 The strengths and limitations of the five epidemiological studies meeting the inclusion 26 criteria for cancer dose-response modeling are summarized in Table 5-22. Among the human 27 studies, the occupational TCDD exposures in the NIOSH and Hamburg cohorts are assumed to 28 be reasonably constant over the duration of occupational exposure. In contrast, the TCDD 29 exposure patterns in the Seveso and BASF cohorts are associated with industrial accidents; as a 30 consequence, the exposure patterns are acute, high dose followed by low-level background 31 exposure. Such exposure patterns similar to those experienced by the BASF and Seveso cohorts This document is a draftfor review purposes only and does not constitute Agency policy. 5-75 DRAFT--DO NOT CITE OR QUOTE 1 have been shown to yield higher estimates o f risk w hen compared to constant exposure scenarios 2 w ith sim ilar total exposure magnitudes (K im et al., 2003, 199146; M urdoch and K re w sk i, 1988, 3 548718; M urdoch et al., 1992, 548719) . Thus, E P A has judged that the N IO S H and Ham burg 4 cohort response data are more relevant than the B A S F and Seveso data for assessing cancer risks 5 from continuous ambient T C D D exposure in the general population. 6 Th e N IO S H (Cheng et al., 2006, 523122; Steenland et al., 2001, 198589) and Ham burg 7 (B ech er et al., 1998, 197173) cohort studies report cum ulative T C D D levels in the serum for 8 cohort members. Th e most significant difference among the Cheng et al. (2006, 523122) 9 analysis and those o f Steenland et al. (2001, 198589) and B e ch e r et al. (1998, 197173) is the 10 method used to back-extrapolate exposure concentrations based on serum T C D D measurements. 11 Steenland et al. (2001, 198589) and B e ch er et al. (1998, 197173) back-extrapolated exposures 12 and body burdens using a first-order model w ith a constant half-life. In contrast, Cheng et al. 13 (2006, 523122) back-extrapolated body burdens using a kinetic m odeling approach that 14 incorporated concentration- and age-dependent elim ination kinetics. 15 Although all three of these are high-quality studies, the kinetic m odeling used by Cheng 16 et al. (2006, 523122) is judged to better reflect T C D D pharm acokinetics, as currently 17 understood, than the first-order m odels used by Steenland et al. (2001, 198589) and B e ch er et al. 18 (1998, 197173) . E P A believes that the representation o f physiological processes provided by 19 Cheng et al. (2006, 523122) is more realistic than the assumption o f sim ple first-order kinetics 20 and this outweighs the attendant modeling uncertainties. Furthermore, the use of kinetic 21 m odeling is consistent w ith recom m endations both by the N A S and the D io x in W orkshop panel. 22 H ow ever, as discussed in Section 3.3.2, the kinetic model that they employed does have 23 certain lim itations, including the fact that it has been calibrated based on a relatively small 24 number o f human subjects. In addition, their kinetic model does not allow body mass index 25 (B M I; and hence fat content) to vary w ith age, w hich m ay bias the model results. Nonetheless, 26 E P A prefers the increased technical sophistication of the dose estimates used in the cancer 27 m ortality risk estimates derived from Cheng et al. (2006, 523122) to those derived from 28 Steenland et al. (2001, 198589) . 29 EPA, therefore, has decided to use the results of the Cheng et al. (2006, 523122) 30 study for derivation of the TCDD OSF based on total cancer mortality as calculated by 31 EPA using data and models from the Cheng et al. (2006, 523122) study as described in This document is a draftfor review purposes only and does not constitute Agency policy. 5-76 DRAFT--DO NOT CITE OR QUOTE 1 Section 5.2.3.I.2. Although the OSF is only strictly defined for exposures above the 2 background exposure experienced by the NIOSH cohort, which was assumed to be 0.5 3 pg/kg-day TCDD, or 5 pg/kg-day total TEQ, EPA assumes that the slope (risk vs. blood 4 concentration) is the same below those background exposure levels as it is above. Table 5-3 5 shows the oral slope factors at specific target risk levels (OSFRLs) which range from 6 1.1 x 105to 1.3 x 106 per (mg/kg-day). EPA recommends the use of an OSF of 1 x 106 per 7 (mg/kg-day) when the target risk range is 10 to 10 . Although EPA prefers the human 8 data, EPA also presents a number of OSFs derived from rodent bioassays. Most of these 9 animal studies are of note, because in general they were well-designed and conducted. In 10 particular, the NTP (2006, 543749) study was recently conducted and represents the most 11 comprehensive evaluation of TCDD chronic rodent toxicity to date. 12 13 5.3.1. Uncertainty in Estimation of Oral Slope Factors from Human Studies 14 A fair amount o f uncertainty is associated w ith the estimation o f slope factor values and 15 cancer risk specific doses for T C D D based on the epidem iological studies. In some instances, 16 the influence o f a given factor is theoretically amenable to analysis, but such investigation is 17 lim ited by the availability o f sufficiently detailed data to support such an analysis. In other 18 cases, only very broad ranges can be placed on the uncertainty associated w ith a given feature o f 19 the analysis, or uncertainties must be discussed qualitatively. 20 The following four sources o f uncertainty are addressed in this section: uncertainty in 21 exposure estimates in the epidem iologic studies (see Section 5.3.1.1), uncertainty in the shape o f 22 the dose-response curve (see Section 5.3.1.2), uncertainty in extrapolating risks below exposure 23 levels in the reference population (see Section 5.3.1.3), uncertainty in cancer risk estimates 24 arising from background D L C exposure (see Section 5.3.1.4) and uncertainty in cancer risk 25 estimates arising from occupational coexposures to D L C s (see Section 5.3.1.5). Section 5.3.2 26 explores other sources o f uncertainty in the epidemiologic risk estimates including the use of 27 cancer mortality rather than cancer incidence data in the derivation o f the oral slope factor, 28 possible influences o f inter-individual variability in T C D D kinetics, and exposures to other 29 occupational carcinogens. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 5-77 DRAFT--DO NOT CITE OR QUOTE 1 5.3.1.1. U ncertainty in E x p o su re E stim ation 2 The major technical challenge within each of the epidemiological studies was developing 3 relevant and precise estimates of exposure. While Warner et al.(2002, 197489) collected blood 4 samples relatively close to the time of the Sevesso accident and could reasonably estimate peak 5 exposures based on these collected samples, in the case of the Becher et al. (1998, 197173), Ott 6 and Zober (1996, 198408), Steenland et al. (2001, 198589), and Cheng et al. (2006, 523122) 7 studies, the major exposure issues included the following 8 9 Selecting (an) appropriate dose metric(s) for dose-response modeling, 10 Estimating serum TCDD levels for the entire cohort based on measurements from a 11 smaller number of the subjects in the cohort collected long after the occupational 12 exposures had occurred, and then assigning exposures to the remaining members of the 13 cohort based on qualitative job classifications. 14 Estimating time-weighted average tissue doses (e.g., lipid-average serum concentration 15 over time) based on single samples taken at one point in time. (Except for the Becher et 16 al. (1998, 197173) analysis where one of the study strengths was their estimate of TCDD 17 half life, which utilized repeated measurements from a subset of their cohort). 18 19 In the Becher et al. (1998, 197173), Steenland et al. (2001, 198589), and Cheng et al. 20 (2006, 523122) studies, dose-response modeling was performed using ppt-years lipid-adjusted 21 serum concentration as the primary dose metric for TCDD; serum TCDD was the only direct 22 measurement of exposure or dose that was available. In addition, as discussed in Section 3.3.4, 23 serum concentration is a reasonable index of total tissue concentration (target organ dose), and 24 lipid-adjusted serum concentration provides a reasonable index of TCDD in the fatty components 25 of tissues. Ott and Zober (1996, 198408) used ng/kg body weight at the time of the accident as 26 the primary dose metric, and U.S. EPA (2003, 537122) later converted these to units of ppt-years 27 lipid-adjusted serum concentration. 28 The decision to use cumulative serum concentrations (ppt-years) as the primary dose 29 metric for carcinogenicity is based on the understanding that time weighted concentrations (over 30 a chronic exposure period) are the most appropriate dose measures for cancer risk assessment. 31 This may not be strictly true if cancer induction by TCDD is considered to be a "threshold 32 process." However, as discussed in Section 5.2, there are reasonable grounds to believe that the This document is a draftfor review purposes only and does not constitute Agency policy. 5-78 DRAFT--DO NOT CITE OR QUOTE 1 assumption o f low-dose linearity is reasonable for T C D D , especially when calculating 2 population risks where the effects o f interindividual variability must be taken into account. 3 In addition to the issue o f low -dose thresholds, the rationale for using cum ulative dose 4 m etrics also can fail at high doses if the adverse response in question in vo lves a step that is 5 saturable (e.g., where there is a m axim um level o f response that cannot be exceeded owing to a 6 rate-limited process). There is some evidence for such a phenomenon in the N IO S H cohort 7 where cancer risks in the highest exposure group (>50,000 ppt-years) appear to saturate, and the 8 response decreases at this level (Steenland et al., 2001, 198589) . Steenland et al. (2001, 198589) 9 suggest that the apparent saturation o f dose-response in this cohort m ay be due, at least partially, 10 to exposure m isclassificatio n among the highest exposed individuals, rather than to an actual 11 reduction in response per unit exposure. 12 Th e uncertainty associated w ith differences in the exposure patterns is important to 13 consider across the five epidem iologic studies. Steenland et al. (2001, 198589) , Cheng et al. 14 (2006, 523122), and B e ch er et al. (1998, 197173) studied cohorts exposed to elevated T C D D 15 levels over a long period o f time, w h ile Ott and Zob er (1996, 198408) and W arner et al.(2002, 16 197489) studied cohorts exposed to T C D D levels significantly above background at one point in 17 time but the exposures and lik e ly the T C D D body burdens declined significantly follow ing these 18 periods of elevated exposure. Both these chronic and acute exposures can be analyzed in terms 19 o f cum ulative exposure to T C D D . U se o f such a m etric requires an assumption that the " actual" 20 cancer potency associated w ith a cum ulative dose w here m uch of the dose is received at a single 21 point in time and then gradually elim inated w ould be sim ilar to the cancer potency of the same 22 cumulative dose received over a longer period of time and also gradually eliminated. W hile E P A 23 believes that such an assumption is not unreasonable, the experiment o f K im et al. (2003, 24 199146), w hich showed statistically significant increase in liver effects due to a peak T C D D 25 dose w hen compared to chronically-dosed Sprague-Daw ley rats administered the same levels of 26 T C D D when measured as a cum ulative dose, suggests that additional analyses of cum ulative and 27 peak T C D D dose measures m ay need to be conducted. 28 There are uncertainties associated w ith the approaches used to estimate T C D D exposures 29 in the members of the occupational epidem iologic studies for w hich no measurement data were 30 available. To impute T C D D levels for workers without measured samples, all four occupational 31 epidem iologic studies m atched w orkers for w hom m easured T C D D sam ples had never been This document is a draftfo r review purposes only and does not constitute Agency policy. 5-79 D R A FT-- DO NOT C IT E OR Q U O TE 1 reported to workers with measured TCDD levels based on job histories. The NIOSH cohort is 2 used to illustrate some of the uncertainties. In the NIOSH cohort, the subset of workers (roughly 3 5% of the total cohort) with blood serum data comprised surviving members of the cohort (in 4 1988), and therefore, their age distribution likely differed from the rest of the cohort. For each 5 worker in this subset, the following data were available: (1) job classification information, 6 (2) employment history, and (3) serum TCDD measures. All of the workers in this subset were 7 employed at a single plant where the work histories were less detailed than at other plants, and 8 many of the workers at this plant had the same job title and were employed during the same 9 calendar period. There is an assumption that workers with same job title and work history were 10 exposed to the same TCDD levels within a plant and across plants; this obviously does not 11 account for exposure heterogeneity. 12 Both Steenland et al. (2001, 198589) and Cheng et al. (2006, 523122) addressed the 13 potential for exposure measurement error in TCDD estimates and possible exposure 14 misclassification. For the highest exposure workers, Steenland et al. (2001, 198589) and Cheng 15 et al. (2006, 523122) found weak, "noisy," and/or negative exposure-response relationships. 16 Steenland et al. (2001, 198589) suggests that possible explanations for this observation include 17 the saturation of effects at the upper end of the dose-response curve, instability of the TCDD 18 exposure estimates based on the limited number of highly exposed individuals, and the increased 19 probability of exposure misclassification for workers whose job histories indicate the highest 20 exposures. As Steenland et al. (2001, 198589) reported, some of the highest exposures might 21 have been inaccurately estimated because they occurred in workers exposed to short-term, high 22 dose exposures during spill clean-up. Cheng et al. (2006, 523122) used sensitivity analyses to 23 examine this measurement error issue and evaluated the potential for exposure misclassification 24 by using ln-transformed TCDD ppt-years. The authors removed all observations with exposures 25 within the lower and upper 1, 2.5, or 5th percentiles of the TCDD ppt-year distribution and also 26 removed observations within just the upper 1, 2.5, or 5th percentile of TCDD ppt-years. These 27 sensitivity analyses yielded results similar to those reported in the primary analysis. An 28 additional concern is that exposure errors might distort the exposure distribution in the 29 population, which generally spreads the response out over a wider dose range. This serves to 30 increase the variance of the regression model, altering both the POD and the corresponding OSF. This document is a draftfor review purposes only and does not constitute Agency policy. 5-80 DRAFT--DO NOT CITE OR QUOTE 1 B e ch er et al. (1998, 197173) only considered w orkers from a single plant but their 2 analysis included w orkers employed in five different job locations w ithin the plant. The 3 influence o f w orker location on slope factor estimates does not appear to be further explored and 4 may represent a source of uncertainty. 5 To estimate long-term body burden metrics from the serum T C D D measurements, 6 Steenland et al. (2001, 198589) em ployed sim ple first order kinetic elim ination rate model w ith a 7 half-life o f 8.7 years. Lim itations o f this approach include (1) the average elimination half-life 8 among the study subjects may not be 8.7 years given differences between the study population 9 and the R anch Hand population from w hich the value w as estimated, (2) use o f a single-value 10 estimate fails to take into account the inherent variab ility in elim ination h a lf life among the 11 individual w orkers, and (3) it fails to take into account variations in elim ination kinetics 12 throughout the lifetim e o f the exposed w orker due to change in body fat, age, etc. The im pact o f 13 these potential sources o f bias on the estimates o f time-integrated body burden cannot be 14 quantitatively assessed. H ow ever, Steenland et al. (2001, 198589) noted that modest changes in 15 elim ination half-life (to 7.1 years) had only a very sm all im pact on risk estimates. 16 Cheng et al. (2006, 523122) estimated past body burdens using the C A D M approach 17 (described in Section 3) (A y lw a rd et al., 2005a, b) rather than a half-life estimate. A s noted 18 above, the incorporation o f concentration- and age-dependent elim ination into this approach has 19 significant advantages over the use o f a constant elim ination half-life. H ow ever, as discussed in 20 Section 3.3, the C A D M has only been subject to lim ited testing against human validation data 21 sets, so the degree to w h ich its advantages are realized in practice cannot be easily assessed. 22 There are no available human data in the low dose region, the region o f interest to this 23 assessment, to compare w ith the C A D M (or Em o n d ) model predictions. 24 B e ch er et al. (1998, 197173) developed h a lf life estimates based on m ultiple T C D D 25 blood m easures in 48 individuals from this cohort. These h a lf life estimates w ere then used to 26 b ack calculate T C D D concentrations at the end o f each w o rk er' s employment, accounting for 27 age and percentage o f body fat. T h is cohort-specific inform ation may provide a better exposure 28 estimate than Steenland et al. (2001, 198589) or Ott and Zob er (1996, 198408) w ho used sim ilar 29 kinetic approaches. H owever, the comparison o f the accuracy o f the exposure estimates across 30 the cohorts is not easily assessed. There are several assumptions and important uncertainties 31 involved in m odeling T C D D exposures in these cohorts. Th e study authors have invoked This document is a draftfo r review purposes only and does not constitute Agency policy. 5-81 D R A FT-- DO NOT C IT E OR Q U O TE 1 different kinetic assumptions when extrapolating measured levels o f T C D D in sera backward in 2 time to estimate higher chronic or peak dosage (i.e., there is uncertainty in these back3 calculations that includes assumptions regarding elimination kinetics). There is also uncertainty 4 in applying such estimates to other members o f the cohort based on sim ilar characteristics (e.g., 5 job category). 6 7 5.3.I.2. U ncertainty in S h a p e o f th e D o se-R esp o n se C urve 8 Another source o f uncertainty is the nature o f the dose-response curve in the low dose 9 region o f interest for risk assessm ent for environmental exposures (e.g., <1 pg/kg-day). The 10 epidem iologic data are based on occupational studies in w h ich exposures w ere often several 11 orders o f magnitude higher than environmental exposures. In these studies, data from the low 12 dose region are quite sparse, and only one study exam ined uncertainty due to the lo w dose 13 region. Steenland and Deddens (2003, 198587) attempted to analyze this region sp ecifically by 14 fitting threshold curves to the N IO S H data in w h ich there w as no extra risk from exposure until 15 some specific level. H ow ever, this model did not fit as w ell as m odels without a threshold. In 16 general, the usual assumption o f linearity in the lo w dose region seems reasonable w hen using 17 epidem iologic data given the la ck o f data in this region that precludes the rejection o f linearity. 18 There is uncertainty in the extrapolation o f the O S F to the lo w dose region (e.g., 19 <5 pg/kg-day). E P A developed the cancer assessm ent in this document assum ing the slope in the 20 low-dose region o f the dose-response curve is linear; the decision w as made due to the lack of 21 sufficient evidence to support an assumption o f nonlinearity as outlined in the E P A ' s Can cer 22 Guidelines (U .S . E P A , 2005, 086237) . Sim ilarly, there is uncertainty as to whether a threshold 23 exists for T C D D -in d u c e d toxicity leading to tum origenesis and the dose associated w ith such a 24 threshold, if it exists, is unknown. E P A chose to model this dose-response without a threshold 25 because there is insufficient evidence to support an assumption o f a threshold. 26 It also is noteworthy that the shapes o f the exposure-response in several o f these studies, 27 based on the published statistical models, is indicative o f a response that tends to tail off or 28 "plateau" at high cum ulative exposures to T C D D . T h is phenomenon has been seen in many 29 studies o f occupational carcinogens, and may reflect a number o f things including exhaustion of 30 people susceptible to cancer, saturation o f biological pathways w h ich are part o f the pathway to This document is a draftfor review purposes only and does not constitute Agency policy. 5-82 DRAFT--DO NOT CITE OR QUOTE 1 cancer, and increased error measurement of dose at high levels biasing dose-response towards 2 the null (Stayner et al., 2003, 054922). 3 4 5.3.I.3. U ncertainty in E x tra p o la tin g R isk s below R eferen ce P opu lation E x p o su re L evels 5 Another source of uncertainty in using human epidemiologic data is due to the lack of 6 completely unexposed populations; there are no human populations that have zero dioxin 7 exposure. The cancer exposure responses modeled in all epidemiologic cohorts, whether 8 primarily exposed via occupational or environmental exposures, can be evaluated with 9 confidence only above the lowest exposed group (i.e., the reference population). There are 10 substantial uncertainties associated with estimating cancer risks from background exposures of 11 TCDD and DLCs because these risks are aggregated in the overall background risk of the 12 referent population, to which outcomes of cohort subjects experiencing higher dioxin exposures 13 are compared. Therefore, the risk modeled from the epidemiologic data is unavoidably the 14 incremental risk above a background exposure to dioxins in the general environment (assumed to 15 be primarily from food intake). Typically, serum TCDD levels in the general populations in the 16 geographic locations and times at which the epidemiologic studies were undertaken have been 17 reported to be on the order of 5 to 20 ppt (Mocarelli et al., 1991, 199600)(WHO, 1998; Pinsky 18 and Lorber, 1998). Hence, the extra risks should be considered as those incurred by added 19 exposure above these background exposures, which then introduces uncertainty associated in the 20 cancer slope factor estimate at exposures below background levels. EPA assumes that the slope 21 of the risk curve below the background exposure experienced by the epidemiologic study cohorts 22 is the same as the (modeled) slope above those background exposure levels; data do not exist to 23 test this assumption. 24 Also, background TCDD/DLC exposures experienced by the epidemiologic study cohorts 25 have been estimated to be much larger (5 to 10-fold) than current background levels. Lorber et 26 al. (2009, 543766) estimate that current U.S. intake rates are roughly 0.58 pg TEQ/kg-day at the 27 50th percentile and suggest that human TEQ ingestion exposures likely peaked in the 1970's. 28 Steenland et al. (2001, 198589), presumably based in part on WHO (1998), estimated 29 background intake rates to be 5 pg TEQ/kg-day for the NIOSH cohort. As a result, the 30 assessment of cancer mortality risk at current background exposure levels is also subject to 31 extrapolation uncertainty. This document is a draftfor review purposes only and does not constitute Agency policy. 5-83 DRAFT--DO NOT CITE OR QUOTE 1 2 5.3.I.4. U ncertainty in C an cer R isk E stim ates A risin g f r o m B a ck g ro u n d D L C E x p o su re 3 None o f the slope factors presented in this document, whether based on epidemiologic 4 studies or animal bioassays, takes into account the impact o f background exposure to D L C s . 5 Background D L C exposure can be estimated for only one o f the animal cancer bioassays N T P 6 (2006, 543749) . Background T C D D and D L C exposure for the rats in the N T P (2006, 543749) 7 does not appear to have been significant, w ith respect to the magnitude o f administered doses 8 (see Section 5.3.2.1). H ow ever, given the trend towards low er exposures to T C D D in recent 9 years, the T C D D / D L C exposure m ay have been m uch higher in the older studies (e.g., K o c ib a et 10 al., 1978, 0 01818; N T P , 1982, 543764; Toth et al., 1979, 197109) . The im pact o f background 11 T C D D / D L C exposure on the cancer risk m odeling o f any o f the bioassay data w ould be to 12 increase the dose term associated w ith each response; consequently, increasing the magnitude o f 13 the B M D L , w ith a proportional reduction in the magnitude o f the slope factor, although the 14 effect w ould probably be sm all (see Section 5.3.2.1). Note that the shift in dose increases the 15 estimated lo w doses proportionately more than the higher doses, potentially obscuring the 16 relationship between dose and response in the lo w dose region. 17 Background dioxin exposure for the epidem iologic cohorts, however, could have been 18 substantial w ith respect to the T C D D exposures in the reference populations used in the 19 m odeling. A s an exam ple, the background dioxin intake the N IO S H cohort, w h ich is the basis 20 for the oral slope factor described previously in this section (5.3), w as estimated to be 21 0.5 pg/kg-day for T C D D and 10 tim es higher (5 pg/kg-day) for total T E Q (Steenland et al., 2001, 22 197433)(W H O , 1998). W H O (1998) estimated that T C D D comprised only about 5 to 10% of 23 total T E Q from exposure to D L C s in food, based on D L C exposure estimates and T E F s available 24 at that time. E sk e n a zi et al. (2004, 197160) estimated that T C D D w as 2 0 % o f total T E Q in the 25 serum o f the reference population in the Seveso W om en' s Health Study from measurements 26 taken in 1976. B ased on m ore recent estimates (Lo rb er et al., 2009, 543766), T C D D is about 27 10% o f total T E Q in human serum in the U nited States. Steenland et al. (2001, 198589) assumed 28 a (cumulating) background exposure o f 5-6 ppt T C D D and 50 ppt total T E Q per year in serum 29 for their analysis of the N IO S H cohort cancer mortality response. The resulting cumulative 30 background exposures, particularly for total T E Q , are large compared to the low er cumulative 31 occupational exposures over the life-tim e o f the cohort (birth to death or end o f follow -up). This document is a draftfor review purposes only and does not constitute Agency policy. 5-84 DRAFT--DO NOT CITE OR QUOTE 1 Crump et al. (2003, 197384), based on Steenland et al. (2001, 198589), assumed a cumulative 2 background serum concentration of 3,000 ppt-year for total TEQ (50 ppt/year x 60 years), which 3 is much larger than the lower NIOSH cohort occupational TCDD exposures. The latter, when 4 grouped in cumulative TCDD serum-concentration septiles Steenland et al. (2001, 197433), 5 range from 260 to 850 ppt-yr in the first few septiles. Conceivably, the much larger background 6 exposure could have a somewhat larger effect on the slope factor than for the relatively lower 7 background exposure in the animal bioassays. Because the Cheng et al. (2006, 523122) 8 modeling does not account for background TEQ, the resulting slope factor is biased high. None 9 of the published analyses of the NIOSH cohort data (Cheng et al., 2006, 523122; Crump et al., 10 2003, 197384; Steenland et al., 2001, 198589) present an analysis that addresses the effect of 11 background TEQ exposure on the modeled risk.50 Given the data and modeling results currently 12 available, the EPA could not find an approach for expressing the quantitative impact with any 13 accuracy or confidence. 14 15 5.3.I.5. U ncertainty in C an cer R isk E stim ates A risin g f r o m O ccu pation al D L C C oexposu res 16 The slope factor estimates are based on an assumption that occupational exposure was 17 entirely to TCDD, with no explicit consideration of the risk attributable to occupational DLCs. 18 Because TCDD typically occurs as a component of a mixture with other DLCs that are assumed 19 to affect cancer risk through dose addition, the assumption that the exposures are entirely TCDD 20 could lead to a positive bias in the slope factor estimates derived from these epidemiologic 21 studies, if the estimates are confounded by other exposures to DLCs and the TEQ dose is larger 22 than the fraction accounted for by TCDD alone. The magnitude of the potential bias can be 23 estimated in a general way through the estimation of risks for plausible mixtures of DLCs and 24 TCDD exposures in the cohort with the same composition as the Steenland et al. (2001, 198589) 25 and Cheng et al. (2006, 523122) studies, but the detailed data required to perform such an 26 analysis on the NIOSH cohort are not available. In addition to the slope factor estimated for 27 TCDD, Becher et al. (1998, 197173) also evaluated the slope based on TEQs. They found a 28 dose-response effect for TCDD but not for TEQ (excluding TCDD) which suggests that 29 confounding by DLCs did not occur. 50 Steenland et al. (2001, 197433) present a TE Q analysis but fo r a scenario where total TE Q is 10 tim es the TC D D exposure fo r both background and occupational exposure. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-85 DRAFT--DO NOT CITE OR QUOTE 1 5.3.2. Other Sources of Uncertainty in Risk Estimates from the Epidemiological Studies 2 Other aspects o f the Steenland et al. (2001, 198589) , Cheng et al. (2006, 523122) , B e ch er 3 et al. (1998, 197173) , and Ott and Zo b er (1996, 198408) studies that are not directly associated 4 w ith T C D D or D L C s may contribute uncertainty to the cancer slope factor estimates. This 5 section lists several o f these and discusses their potential directional bias in slope. General issues 6 associated with potential confounding effects also were discussed in the 2003 Reassessm ent 7 (U .S. E P A , 2003, 537122) . 8 A ll o f the studies that meet the criteria (with the exception o f W arner et al., 2002, 9 197489) measure cancer mortality rather than cancer incidence. Th is likely biases the slope 10 factor dow nward relative to a slope calculated for cancer incidence, the typical basis o f E P A 11 cancer slope factors. In the N IO S H cohort, roughly one-third o f the fatal cancers w ere identified 12 as lung cancer. B e cau se o f the high case m ortality rate associated w ith lung cancer during the 13 period o f cohort evaluation (e.g., the 5-year relative survival rates for lung cancer w ere less than 14 10% before 1973 and w ere less than15% before 1995 (H orner et al., 2009), the slope factor 15 estimated for cancer m ortality m ight not be m uch low er than that calculated for cancer incidence. 16 T h is assum es that the outcome o f a cancer incident (i.e., cancer m ortality) is independent o f 17 occupational T C D D exposure levels. Estim ation o f cancer incidence in the general population 18 associated w ith T C D D exposure w ould require assumptions related to the relative survival and 19 age-specific cancer risks in the exposed population compared to the N IO S H cohort or the 20 Hamburg cohort; insufficient data are available to support such an analysis. 21 Th e routes of T C D D exposures in the occupational cohorts include dermal and inhalation 22 exposures (Steenland et al., 1999, 197437), the U .S . population is assum ed to be prim arily 23 exposed through the intake of T C D D and D L C s in foods). G iv e n the persitence of T C D D in the 24 body, differences in exposure routes may not be significant, but route-specific effects can not be 25 precluded. The directional bias on the slope factor that is associated with this uncertainty is not 26 known. 27 Occupational exposures to other carcinogens could lead to uncertainty in the slope factor. 28 F o r exam ple, in addition to T C D D , the H am burg cohort w as also exposed to 29 hexachlorocyclohexane (H C H ), w hich IA R C classified as possibly carcinogenic to humans, and 30 lindane, w h ich E P A (2001) stated had " suggestive evidence o f carcinogenicity, but not sufficient 31 to assess human carcinogenic potential." W h ile such co-exposures w ould not bias the exposure This document is a draftfo r review purposes only and does not constitute Agency policy. 5-86 D R A FT-- DO NOT C IT E OR Q U O TE 1 metric (i.e., not dose additive), to the extent that they were correlated w ith T C D D exposure, the 2 cancer mortality risk attributed to T C D D w ould be overestimated,biasing the slope factor high 3 because all cancers are attributed to T C D D . T o exam ine this, Cheng et al. (2006, 523122) 4 assessed the im pact of possible confounding by conducting excluding individual plants in the 5 modeling. If the estimated cancer risks as a function of exposure did not change too much when 6 specific facilities w ere left out, then confounding w as deemed un likely. Cheng et al. (2006, 7 523122) likew ise found little variation in risks based on these analyses. 8 There is adequate evidence to believe age, gender, and body fat content all can have a 9 significant impact on elimination kinetics and consequent cancer risks associated with T C D D 10 exposure (U .S . E P A , 2003, 537122) . W h ile the authors evaluating the H am burg cohort 11 accounted for such im pacts in their kinetic analysis, interindividual kinetic differences w ere not 12 considered in evaluations o f other cohorts. 13 There m ay be gender differences that affect susceptibility to T C D D exposure. The 14 cohorts analyzed by Steenland et al. (2001, 198589), Cheng et al. (2006, 523122), Ott and Zober 15 (1996, 198408) and B e ch er et al. (1998, 197173) w ere com prised alm ost ex clu sively o f men. 16 T h is precluded system atically addressing differences between m ales and fem ales in these studies. 17 Further, because E P A could not develop an estimate from the W arner et al. (2002, 197489) 18 cohort, none o f the studies analyzed here for cancer dose-response contained a significant 19 percentage o f women. Thus, the generalizability o f the slope factor estimates to w om en is 20 uncertain. 21 F in a lly , o f these cancer cohorts only the Seveso cohort included children. Th e unique 22 sensitivities o f infants, toddlers, and children cannot be addressed based on information in the 23 occupational cohorts, although the increases in cancer risk in the Seveso cohort, to date, appear 24 to be modest. A sid e from differences in exposure patterns and body fat content, the unique 25 developmental status o f children may result in a substantially different profile o f cancer risks 26 (and magnitudes o f those risks) than can be addressed by sim ply compensating on the basis of 27 differences in body weight, food intake, etc. Further, because E P A could not develop an 28 estimate from the W arner et al. (2002, 197489) cohort, none o f the studies for cancer dose29 response analyzed contained a significant percentage o f women. Thus, the generalizability o f the 30 slope factor estimates to women and children is uncertain. This document is a draftfor review purposes only and does not constitute Agency policy. 5-87 DRAFT--DO NOT CITE OR QUOTE 1 A number of other factors are routinely evaluated in cancer epidemiology studies, but 2 appear likely to have little impact on the direction of the slope factor; however, they likely 3 increase overall variability either in the dose or response. These include smoking and lifestyle 4 factors. Intraindividual variation in TCDD kinetics and susceptibility also could affect the 5 relationship between exposure and cancer risk. In each of these cases, it is difficult to determine 6 the directional bias these factors introduce into the derivation of the slope factor, unless 7 somehow they are correlated with with occupational dioxin exposures. 8 9 5.3.2.I. E ffect o f A d d e d B a ck g ro u n d TE Q on T C D D D ose-R espon se 10 A source of uncertainty for TCDD dose-response modeling is the impact that background 11 exposures of TCDD and other DLCs might have on the modeling output. As mentioned 12 previously in Text Box 4-1, NTP (2006, 543749) presented measurements of TCDD in the fat of 13 control animals. To study the potential impact of background TCDD and total TEQ on the 14 cancer dose-response modeling for the NTP (2006, 197605) study, EPA has estimated 15 background levels of TCDD and TEQ (based on total TCDD, PeCDF and PCB126) from the 16 mixture study to serve as surrogates for background exposures in the TCDD-only study (limit of 17 detection too high for control level measurements). Background doses were estimated as: 18 19 Chemical%(B) = x D oseTcDD (Eq. 5-9) TCDDyfatTcDD ) 20 21 where 22 Chemicali(B) estimate of background exposure to Chemical i in ng/kg units of TCDD 23 blood concentrations at 105 weeks, for i = TCDD, PeCDF and PCB126. 24 Chemicali(fatMC) mean pg/g of Chemical i in the fat tissues of the control animals at 25 105 weeks in mixtures study (NTP, 2006, 543749). 26 TCDD(fatTCDD) mean pg/g of TCDD in the fat tissues of the 3 ng/kg dose group at 27 105 weeks in the TCDD study (NTP, 2006, 197605). 28 DoseTCDD 29 30 2.56 ng/kg TCDD blood concentration for the 3 ng/kg dose group in the TCDD study (from the Emond rat PBPK modeling of NTP, 2006, 197605) 31 TEFi 32 Toxicity Equivalence Factor for Chemical i (from Van den berg et al. (2006, 543769)). This document is a draftfor review purposes only and does not constitute Agency policy. 5-88 DRAFT--DO NOT CITE OR QUOTE 1 Assuming simple proportionality of blood TCDD concentrations between controls and 2 low-dose (3 ng/kg) animals, the TEF-adjusted ratio of each congener (Chemical i) in control 3 animal fat to low-dose-animal fat is multiplied by the modeled TCDD blood concentration for 4 the low-dose animals to obtain an equivalent background exposure in the dose metric (ng/kg 5 whole blood) used to calculate all the OSFs in this assessment. For total TEQ, the estimates 6 across the three congeners are summed. The total TEQ estimates are biased somewhat high 7 because they are based on terminal (2-year) measurements rather than representing lifetime 8 averages. Background exposures are then added to the modeled TCDD blood concentrations for 9 several different background exposure scenarios (see Table 5-23) prior to conducting 10 Benchmark-Dose (BMD) modeling. 11 BMD modeling was conducted for the cholangiocarcinoma endpoint in the TCDD study 12 (NTP, 2006, 197605). This was done for scenarios that added the following estimated TCDD or 13 TEQ background doses to the TCDD study doses: background TCDD only, total estimated TEQ, 14 twice the total TEQ and ten times the background TCDD (see Table 5-23). These doses may 15 bound the potential background exposures as TCDD has been thought to represent about 10% of 16 all TEQs at environmental levels (WHO, 1998). Table 5-24 shows that, as expected, adding to 17 the exposure term increases the BMDL (and decreases the OSF) and also shifts the shape of the 18 dose-response slope slightly towards sublinear (see Appendix I). However, at these background 19 exposure levels relative to the administered dose levels, there is very little quantitative impact on 20 the cancer dose-response modeling for the NTP (2006, 197605) study. Even with the most 21 extreme assumption that background TCDD is only 10% of total background TEQ, the BMDL 22 changes by only 12%. Assuming that background exposures were higher for older studies (e.g., 23 Kociba et al., 1978, 001818; NTP, 1982, 594255), the impact would be somewhat higher, but 24 unless the background exposures were substantially higher than the lower tested doses (ca. 25 1-10 ng/kg-day), a significant change in the dose-response modeling results would not be 26 expected.51 27 However, as discussed previously, background TEQ exposures were likely very high 28 with respect to the lower occupational TCDD exposure levels as reported in the epidemiologic 29 studies. Table 5-25 shows the relative increase in exposure levels (as cumulative serum TCDD 51 N ote that the situation is d ifferent fo r single-exposure studies where accumulated body burden fro m background exposures could be higher than the low est administered dose (see Tex B ox 4.1 in Section 4.4). This document is a draftfo r review purposes only and does not constitute Agency policy. 5-89 DRAFT--DO NOT CITE OR QUOTE 1 concentrations) for the NIOSH cohort septiles assuming that total background TEQ is 10 times 2 background TCDD and that 50 ppt TEQ per year is accumulated in serum (Crump et al., 2003, 3 197384; Steenland et al., 2001, 198589). Although definitive quantitative analyses have not yet 4 been published or designed, the impact on modeled TCDD risk from these studies could be 5 substantial. The expectation for the direction of the effect would be the same as for the animal 6 bioassays; adding to the exposure magnitude without changing the response would decrease the 7 unit risk. 8 9 5.3.3. Approaches to Combining Estimates from Different Epidemiologic Studies 10 Meta-analyses and pooled analyses are two common approaches for combining 11 epidemiologic study data. Meta-analyses are a useful way to combine epidemiologic data from 12 different studies and derive a common estimate of effect, particularly when there are a large 13 number of comparable studies that are fairly homogenous as to make them possible to combine. 14 A meta-analysis often involves a weighted average of effect measures, dose-response 15 coefficients, or ED01s. 16 Unlike a meta-analysis, a pooled analysis combines the original exposure and health 17 outcome data across multiple studies, enabling a fit of new models to the data which were not 18 used in the original publications. Whereas a pooled analysis of the four different cohorts 19 considered here would be useful to explore the functional form and fit of models (either 20 statistical or multistage) across all four cohorts, this would entail a lengthy undertaking and is not 21 being contemplated here, due in part to concerns about the confidence in the results of such an 22 undertaking. 23 24 5.3.3.1. The C rum p et al. (2 0 0 3 ,197384) M eta-an alysis 25 Crump et al. (2003, 197384) published a meta-analysis that incorporated data from the 26 three studies EPA used in the quantitative dose-response modeling presented in the 2003 27 Reassessment (U.S. EPA, 2003, 537122). These three study populations were the NIOSH 28 (Steenland et al., 2001, 197433), the Hamburg (Becher et al., 1998, 197173), and the BASF (Ott 29 and Zober, 1996, 198408) cohorts. The data for the NIOSH study included six additional years 30 of follow-up and improved TCDD exposure estimates that had not been applied to EPA's dose31 response modeling in the 2003 Reassessment. This study examined the relationship between This document is a draftfor review purposes only and does not constitute Agency policy. 5-90 DRAFT--DO NOT CITE OR QUOTE 1 T C D D exposure and all-cancer mortality. S M R statistics that had been used in all three studies 2 were applied. 3 Th e Crum p et al. (2003, 197384) analysis w as based on published data, and therefore, 4 selection of the dose m etric w as lim ited to how aggregated data had been presented in the 5 publications. F o r the N IO S H component o f the analysis, the exposure data were based on 6 w o rker-specific data and specific processes performed at each plant (Steenland et al., 2001, 7 197433) . The previous approach assigned workers that had broad categories o f exposure 8 duration with the same cumulative serum level, and did not take into account the particular plant 9 or the jo b assignm ent w ithin the plant. Th e Crum p et al. (2003, 197384) approach did take into 10 account w hen exposure occurred in relation to the follow -up interval. Th e T C D D exposure 11 m etric used w as a cum ulative serum lip id concentration ( C S L C ) . F o r the Ham burg cohort, 12 Crum p et al. (2003, 197384) used an average value from the exposure ranges provided in Flesch - 13 Janys et al. (1998, 197339) . F o r the B A S F cohort, arithmetic averages for the dose categories 14 w ere converted to T C D D C S L C intakes by dividing them by 0.25 (average body fat o f 2 5 % ) and 15 a decay rate that corresponded to a half-life o f 7 years. 16 Th e outcome variable for the dose-response m odeling w as all cancer m ortality, and 17 C S L C w as the independent variable. Crum p et al. (2003, 197384) performed a series o f trend 18 tests to determine the low est dose for w h ich a statistically significant trend in S M R could be 19 shown and all other lo w er doses. These tests also exam ined the highest dose in w h ich there w as 20 no statistically significant trend using data from this dose and all other low er doses. Estim ates of 21 ED10, ED05, and ED01 for T E Q w ith respect to the lifetim e probability of dying from cancer were 22 calculated. This calculation assumed a first-order elimination process with a half-life of 23 7.6 years, a 50% system ic uptake o f ingested dioxin, that dioxin concentration in serum lip id is a 24 suitable measure for dioxin concentration in all lipid, and that all dioxin is sequestered in lipid 25 (w hich comprises 25% o f body weight). Age-specific mortality rates in the presence o f dioxin 26 exposure were then generated. Life-table methodology w as used to calculate lifetim e risks of 27 cancer mortality. 28 B ased on the m odeling results, the hypothesis o f a baseline S M R o f 1.0 w as rejected, and 29 the linear model produced an S M R estimate o f 1.17 (95% C I = 1 .0 4-1.30) from these studies. 30 The dose-response curves for the three studies were not homogeneous. N am ely, the points from 31 the B A S F cohort fell below the predicted curve. B ecau se the heterogeneity w as not judged to be This document is a draftfo r review purposes only and does not constitute Agency policy. 5-91 D R A FT-- DO NOT C IT E OR Q U O TE 1 extreme by different statistical tests, however, the investigators used a common model in a 2 combined analysis of the data from the three studies. The linear model provided an adequate fit 3 of the data, and the slope associated with CSLC-ppt was 6.3 x 10-6 (95% CI = 8.8 x 10-7 to 4 1.3 x 10-5). Based on goodness of fit analysis, the preferred estimate of ED01 was 45 pg/kg-day, 5 which was six times higher than the estimate of 7.7 pg/kg-day derived by Steenland et al. (2001, 6 198589). 7 8 5.3.3.2. E P A ' D ecision N o t to C on du ct a M eta-an alysis 9 From a statistical perspective, meta-analyses may not be very reliable when applied to a 10 small number of studies. Crump et al. (2003, 197384) used only three studies. Had EPA 11 undertaken a meta-analysis for the studies that met its criteria, most of the weight would come 12 from the two large studies on the NIOSH and Hamburg cohorts. However, such an analysis 13 relies on an assumption of a normally distributed between-study effect. This normality 14 assumption cannot be assessed with only three observations, yet the meta-analysis estimate is 15 highly sensitive to this distributional assumption (Higgins et al., 2009, 594339). Because of this 16 limitation and the imprecision of the between-study variance estimate, statisticians often 17 recommend forgoing meta-analysis in favor of discussing the individual studies when few 18 studies are available (Cox, 2006, 594342; Higgins et al., 2009, 594339). Based on these 19 considerations, EPA decided not to undertake a meta-analysis in this document. 20 As noted previously, Crump et al. (2003, 197384) has conducted a meta-analysis of the 21 three cohorts considered here, i.e., the NIOSH, Hamburg, and BASF cohorts. However, Crump 22 et al. modeled SMR data in which the cohorts were compared to the general population, rather 23 than on internal exposure-response analyses as relied upon in this document. Their analysis 24 included a total of 15 different SMRs from the three studies. A prior analysis of the dose25 responses by Becher et al. (1998, 197173) was used (i.e., the categorical SMR analysis by 26 Flesch-Janys et al. (1998, 197339)). Additionally, a prior analysis of the NIOSH cohort 27 (Steenland et al., 1999, 197437) in which SMRs were calculated was used. Crump et al. (2003, 28 197384) found that a linear dose-response gave a good fit to the data, and used that for deriving 29 an ED01. However, they found that a supra-linear dose-response provided a better fit to the data, 30 but rejected the supra-linear model (a power model) because of an infinite slope at zero dose. In 31 the original publications by Becher et al. (1998, 197173) and Steenland et al. (2001, 198589), This document is a draftfor review purposes only and does not constitute Agency policy. 5-92 DRAFT--DO NOT CITE OR QUOTE 1 both observed a supra-linear dose-response trend. Crump et al. (2003, 197384) concluded that 2 the ED01 was 45 pg/kg-day, six times higher than the ED01 of 7.7 pg/kg-day calculated by 3 Steenland et al. (2001, 198589) using the same dietary units (pg/kg-day). This document is a draftfor review purposes only and does not constitute Agency policy. 5-93 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-1. Cancer slope factors calculated from Becher et al. (1998, 197173), 2 Steenland et al. (2001, 197433) and Ott and Zober (1996, 198408) from 2003 3 Reassessment Table 5-4 4 Study ED01 (LED01) (ng/kg) Cancer slope factor per ng/kg-day above background3 (UCL) Hamburg cohort Power model B e ch er et al. (1998, 197173) 6 (N .A .) 5.1 (N .A .) Hamburg cohort Additive model B e ch er et al. (1998, 197173) 18.2 (N .A .) 1.6 (N .A .) Hamburg cohort M ultiplicative model B e ch er et al. (1998, 197173) 32.2 (N .A .) 0.89 (N .A .) N IO S H cohort Piecew ise linear model Steenland et al. (2001, 198589) 18.6 (11.5) 1.5 (2.5) B A S F cohort, from Ott and Zober (1996, 198408), m ultiplicative 50.9 (25.0) 0.57 (1.2) 5 6 "Assumes 25% o f body w eight is lip id ; in humans 80% o f d ioxin dose is absorbed from the norm al 7 diet; the TC D D h a lf-life is 7.1 years in humans. Background a ll cancer m ortality rate calculated 8 through lifetab le analysis to 75 years. Summary results are fo r male a ll cancer risk, because the 9 male life tim e (to 75 years) a ll cancer risk is greater than fo r fem ales, leading to correspondingly 10 higher cancer slope factors. As detailed in Part III, Chapter 8, R elR isk(E D 01) = 0.99 + 11 0.01/R isk(0dose). Based on the manner in w hich the dose-response data were calculated using Cox 12 regression rate ratio analyses, risks are given as cancer slope factors fo r 1 pg/kg-day above 13 background, assumed 5 ppt TC D D in lipid. 14 U C L = upper confidence lim it. 15 16 Source: U .S. E P A (U .S. E P A , 2003, 537122; Part III, Chapter 5, Table 5-4) This document is a draftfor review purposes only and does not constitute Agency policy. 5-94 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-2. Cox regression coefficients and incremental cancer-mortality risk 2 for NIOSH cohort data 3 Model Cox regression coefficient estimate (ppt-year)-1 Incremental riska Steenland et al.(2001, 197433) (unlagged exposures) Piecewise linear 1.5 x 10 5 7.0 x 10 4 Cheng et al. (Cheng et al., 2006, 523122) (exposures lagged 15 years) Linear, lower 95% of observations 3.3 x 10 6b 1.2 x 10 4 Linear, full data 1.7 x 10-8 c 6.3 x 10-7 4 5 "Compared to internal reference population (low est exposure group),w ith a cancer m ortality rate o f 0.214; assumes 6 background exposure o f 5 ppt per year serum -lipid TC D D concentration. 7 bp < 0.05. 8 cp < 0.05. 9 dN o t statistically sig nificant (p > 0.05). 10 11 Source: Cheng et al. (2006, 523122; Table IV ). This document is a draftfor review purposes only and does not constitute Agency policy. 5-95 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-3. Comparison of fat concentrations, risk specific dose estimates and 2 associated oral slope factors based on upper 95thpercentile estimate of 3 regression coefficient of all fatal cancers reported by Cheng et al. (2006, 4 523122) for selected risk levels 5 Risk level (RL) 1 x 10-2 AUCRL (ppt-yr) 1.262 x 104 FATrl (ng/kg) 1.803 x 102 Equivalent oral slope Risk specific doseb factors (OSFrl) per (Drl) (ng/kg-day) (mg/kg-day) 8.79 x 10-2 1.1 x 1 0 5 5 x 10-3 6.432 x 103 9.189 x 101 3.14 x 10-2 1.6 x 105 1 x 10 -3 1.307 x 103 1.867 x 101 2.88 x 10-3 3.5 x 105 5 x 10-4 6.546 x 102 9.352 x 100 9.56 x 10-4 5.2 x 105 1 x 10 -4 1.311 x 10 2 1.873 x 100 1.29 x 10-4 7.8 x 105 5 x 10-5 6.558 x 101 9.368 x 10'1 5.52 x 10-5 9.1 x 1 0 5 1 x 10 -5 1.312 x 101 1.874 x 10-1 8.94 x 10-6 1.1 x 1 0 6 5 x 10-6 6.559 x 100 9.370 x 10'2 4.25 x 10-6 1.2 x 106 1 x 10 -6 1.312 x 100 1.874 x 10-2 8.08 x 10-7 1.2 x 106 5 x 10-7 6.559 x 10'1 9.370 x 10'3 4.00 x 10-7 1.3 x 106 1 x 10 -7 1.312 x 10-1 1.874 x 10-3 7.92 x 10-8 1.3 x 106 6 7 aBased on regression coefficient o f Cheng et al. (2006, 523122, Table III), excluding observations in the upper 5% 8 range o f the exposures; where reported P = 3.3 x 10-6 ppt-years and standard error = 1.4 * 10-6. Upper 95th 9 percentile estim ate o f regression coefficient (P95) calculated to be 6.04 x 10-6 = (3.3 x 1 0 -6) + 1.96 x ( 1 .4 x 10-6); 10 background cancer m ortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 523122) . 11 bTo calculate the extra cancer risk (E R ) and OSF fo r any TC D D daily oral intake (D ): 12 5. F o r D in ng/kg-d, loo k up the corresponding fa t concentration (ng/kg = ppt) fro m the conversion chart 13 (nongestational life tim e dose m etrics) in Appendix C.4.1. 14 6. Calculate the A U C in ppt-yrs by m ultip lying the fa t concentration by 70 years. 15 7. Calculate E xtra R isk (E R ) using the fo llo w in g equation: 16 E R = [exp(AUC x 6.04E-6) x 0.112 - 0.112] - 0.888. 17 8. Calculate the OSF (m g/kg-d)-1 = 1E6 x (E R - D ). 18 Exam ple fo r risk at the R fD : D = 7 x 10-4 ng/kg-d; fa t concentration = 6.93 ng/kg; 19 A U C = 70 years x 6.93 ppt = 485 ppt-year; 20 E R = exp(485 ppt-year x 6.04E-6 (ppt-yr)-1) x 0.112 - 0.112) - 0.888 = 3.7 x 10-4 21 O SF = 1E6 ng/mg x (3 .7 x -1 0 -4 7 x 1 0 -4 ng/kg-d) = 5.3 x 1 0 5 (m g/kg-d)-1. This document is a draftfor review purposes only and does not constitute Agency policy. 5-96 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-4. Comparison of fat concentrations, risk specific dose estimates and 2 associated central tendency slope estimates based on best estimate of 3 regression coefficient of all fatal cancers reported by Cheng et al. (2006, 4 523122) for selected risk levels 5 Risk level (RL) AUCrl, (ppt-yr) FATrl (ng/kg) Risk specific dose (Drl) (ng/kg-day) Central tendency slope estimates (mg/kg-day)-1 1 x 10-2 2.312 x 104 3.303 x 102 2.21 x 10-1 4.5 x 104 1 x 10-3 2.393x 103 3.419 x 101 6.97 x 10-3 1.4 x 105 1 x 10-4 2.402 x 102 3.431 x 100 2.74 x 10-4 3.7 x 105 1 x 10-5 2.403 x 101 3.432 x 10-1 1. 74 x 10-5 5.7 x 105 1 x 10-6 2.403 x 100 3.432 x 10-2 1.50 x 10-6 6.7 x 105 1 x 10-7 2.403 x 10-1 3.432 x 10-3 1.46 x 10-7 7.0 x 105 6 7 aBased on regression coefficient of Cheng et al (2006, 523122; Table III) excluding observations in the upper 5% 8 range (>252,950 ppt-year lipid adjusted serum TCDD) of the exposures; where reported P = 3.3 x 10-6 ppt-years; 9 background cancer mortality risk is assumed to be 0.112 as reported by Cheng et al. (2006, 523122). 10 11 12 Table 5-5. Kociba et al. (1978, 001818) male rat tumor incidence dataa and 13 blood concentrations for dose-response modeling 14 Morphology: topography Vehicle control Low dose Medium dose High dose (ng/kg) (ng/kg) (ng/kg) (ng/kg) 0 1.56 7.16 38.72 Stratified squamous cell carcinoma of hard palate or nasal turbinates 0/85 0/50 0/50 4/50b Stratified squamous cell carcinoma of tongue 0/85 1/50 1/50 3/50b Adenoma of adrenal cortex 0/85 0/50 2/50 5/50b 15 16 "Source: Kociba et al.(1978, 001818; Table 4). 17 bStatistically significant by Fischer Exact Test (p < 0.05). This document is a draftfor review purposes only and does not constitute Agency policy. 5-97 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-6. Kociba et al. (1978, 001818) female rat tumor incidence dataa and 2 blood concentrations for dose-response modeling 3 Morphology: topography Vehicle control (ng/kg) 0 Low dose (ng/kg) 1.55 Medium dose (ng/kg) 7.15 High dose (ng/kg) 38.56 Hepatocellular adenoma(s) or carcinoma(s) 2/86 1/50 9/5 0a 18/45b Stratified squamous cell 0/86 0/50 1/50 4/49b carcinoma of hard palate or nasal turbinates Keratinizing squamous cell 0/86 0/50 0/50 7/49b carcinoma of lung 4 5 aSource: Kociba et al. (1978, 001818; Table 5). Incidence for Hepatocellular adenomas or carcinomas is from 6 Goodman and Sauer (Goodman and Sauer, 1992, 197667; Table 1); EPA calculated statistical significance as the 7 study authors did not provide this. 8 bStatistically significant by Fischer Exact Test (p < 0.05). 9 10 11 Table 5-7. NTP (1982, 594255) female rat tumor incidence dataa and blood 12 concentrations for dose-response modeling 13 Vehicle control Low dose Medium dose High dose Morphology: topography (ng/kg) 0 (ng/kg) 1.96 (ng/kg) 5.69 (ng/kg) 29.75 Subcutaneous tissue: fibrosarcoma 0/75 2/50 3/50 4/49b Liver: neoplastic nodule or hepatocellular carcinoma 5/75c 1/49 3/50 14/49b Adrenal: cortical adenoma, or carcinoma or adenoma, NOS 11/73c 9/49 5/49 14/46b Thyroid: follicular-cell adenoma 3/73c 2/45 1/49 6/47 14 15 "Source: NTP (1982, 594255; Table 10). 16 bStatistically significant by Fischer Exact Test (p < 0.05). 17 Statistically significant trend by Chochran-Armitage test (p < 0.05). This document is a draftfor review purposes only and does not constitute Agency policy. 5-98 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-8. NTP (1982, 594255) male rat tumor incidence dataa and blood 2 concentrations for dose-response modeling 3 Vehicle control Low dose Medium dose High dose Morphology: topography (ng/kg) 0 (ng/kg) 1.96 (ng/kg) 5.70 (ng/kg) 29.87 Liver: neoplastic nodule or hepatocellular carcinoma 0/74b 0/50 0/50 3/50 Thyroid: follicular-cell adenoma or carcinoma 1/69b 5/48c 8/5 0c 11/50c Adrenal cortex: adenoma 6/72 9/50 12/49b 9/49 4 5 aSource: N TP(1982, 594255; Table 9). 6 bS tatistically significant trend by Chochran-Arm itage test (p < 0.05). 7 cS tatistically significant by Fischer Exact Test (p < 0.05). 8 9 10 Table 5-9. NTP (1982, 594255) female mouse tumor incidence dataa and 11 blood concentrations for dose-response modeling 12 Vehicle control Low dose Medium dose High dose (ng/kg) (ng/kg) (ng/kg) (ng/kg) Morphology: topography 0 1.95 5.84 32.06 Subcutaneous tissue: fibrosarcoma 1/74b 1/50 1/48 5/47c Hematopoietic system: lymphoma or leukemia 18/74b 12/50 13/48 20/47c Liver: hepatooellular adenoma or carcinoma 3/73b 6/50 6/48 11/47c Thyroid: follicular-cell adenoma 0/69b 3/50 13 14 "Source: N TP (1982, 594255; Table 15). 15 bS tatistically significant trend by Chochran-Arm itage test (p < 0.05). 16 S ta tis tic a lly significant by Fischer Exact Test (p < 0.05). 1/47 5/46c This document is a draftfor review purposes only and does not constitute Agency policy. 5-99 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-10. NTP (1982, 594255) male mouse tumor incidence dataa and 2 blood concentrations for dose-response modeling 3 Morphology: topography Vehicle control Low dose Medium High dose (ng/kg) (ng/kg) dose (ng/kg) (ng/kg) 0 0.77 2.27 11.24 Lung: alveolar/bronchiolar adenoma 10/71b 2/48 4/48 13/50 or carcinoma Liver: hepatocellular adenoma or carcinoma 15/73b 12/49 13/49 27/50c 4 5 aSource: N TP (1982, 594255; Table 14). 6 bS tatistically significant trend by Chochran-Arm itage test (p < 0.05). 7 S ta tis tic a lly significant by Fischer Exact Test (p < 0.05). 8 9 10 Table 5-11. NTP (2006, 197605) female rat tumor incidence dataa and blood 11 concentrations for dose-response modelingb 12 System: morphology: topography Vehicle control (ng/kg) 0 Low dose (ng/kg) 2.56 Low-med dose (ng/kg) 5.69 Median dose (ng/kg) 9.79 Med-high High dose dose (ng/kg) (ng/kg) 16.57 29.70 Liver: 0/49c 0/48 0/46 1/50 4/49 25/53c cholangiocarcinoma Liver: hepatocellular adenoma 0/49c 0/48 0/46 0/50 1/49 13/53c Oral mucosa: squamous cell carcinoma 1/49c 2/48 1/46 0/50 4/49 10/53c Pancreas: adenoma 0/48c 0/48 0/46 0/50 0/48 3/51 or carcinoma Lung: cystic keratinizing epithelioma 0/49c 0/48 0/46 0/49 0/49 9/52c 13 14 "Source: N TP (2006, 197605; Table A3a). 15 bIncidence adjusted fo r anim als <365 days on study. 16 S ta tis tic a lly sig nificant by Poly-3 Test (p < 0.05). 17 18 This document is a draftfor review purposes only and does not constitute Agency policy. 5-100 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-12. Toth et al. (1979, 197109) male mouse tumor incidence dataa and 2 blood concentrations for dose-response modeling 3 Morphology: topography Vehicle control (ng/kg) 0 Low dose (ng/kg) 0.57 Medium dose (ng/kg) 14.21 High dose (ng/kg) 91.21 Liver tumors 7/38 13/44 21/44b 13/43 4 5 "Source: Toth et al. (1979, 197109; Table 1). 6 bStatistically significant by Chi2Test (p < 0.01). 7 8 9 Table 5-13. Della Porta et al. (1987, 197405) male mouse tumor incidence 10 dataa and blood concentrations for dose-response modeling 11 Morphology: topography Vehicle control (ng/kg) 0 Low dose (ng/kg) 38.00 High dose (ng/kg) 67.77 Hepatocellular carcinoma 5/43 15/51b 33/50b 12 13 "Source: Della Porta et al. (1987, 197405; Table 4). 14 bStatistically significant by Chi2Test (p < 0.05). 15 16 17 Table 5-14. Della Porta et al. (1987, 197405) female mouse tumor incidence 18 dataa and blood concentrations for dose-response modeling 19 Vehicle control Low dose High dose Morphology: topography (ng/kg) 0 (ng/kg) 37.59 (ng/kg) 66.97 Hepatocellular adenoma 2/49 4/42b 11/48b Hepatocellular carcinoma 1/49 12/42b 9/4 8b 20 21 aSource: Della Porta et al. (1987, 197405; Table 4). 22 bStatistically significant by Chi2Test (p < 0.05). This document is a draftfor review purposes only and does not constitute Agency policy. 5-101 DRAFT--DO NOT CITE OR QUOTE Table 5-15. Comparison of multi-stage modeling results across cancer bioassays using blood concentrations This document is a draftfo r review purposes only and does not constitute Agency policy. 5-102 Study Species Sex D ella Mouse M ale Porta et al. Fe m ale (1987, 197405) Kociba Rat et al. (1978, 001818) M ale Fe m ale NTP Rat (1982, 594255) Fe m ale M ale Morphology: topography H epatocellular carcinoma H epatocellular adenoma H epatocellular carcinoma S tratified squamous cell carcinom a o f hard palate or nasal turbinates S tratified squamous cell carcinoma o f tongue Adenoma o f adrenal cortex Combined tum ors Bayesian analysis H epatocellular adenoma(s) o r carcinoma(s) S tratified squamous cell carcinom a o f hard palate or nasal turbinates K eratinizing squamous cell carcinoma o f lung Combined tum ors Bayesian analysis Subcutaneous tissue: fibrosarcom a L ive r: neoplastic nodule o r hepatocellular carcinoma Adrenal: cortical adenoma, o r carcinoma or adenoma, NOS Thyroid : fo llic u la r-c e ll adenoma Combined tum ors Bayesian analysis L ive r: neoplastic nodule o r hepatocellular carcinoma Thyroid : fo llic u la r-c e ll adenoma o r carcinoma Adrenal cortex: adenoma Combined tum ors Bayesian analysis Multi-stage modeling:a stage, GoF _p-value, LL difference 2, p = 0.52 2, p = 0.86 1. p = 0.019 1, p = 0.81 1, p = 0.47 1, p= 0.78 1, p = 0.24 1, p = 0.97 1, p = 0.63 1, p = 0.18 1, p = 0.22 1, p = 0.34 1, p = 0.57 1, p = 0.85 1, p = 0.06 1, p = 0.06 BMD01 (ng/kg) 7.14 14.49 2.30 5.76 6.09 3.25 1.57 0.70 4.51 3.14 0.51 3.13 1.17 1.61 3.38 0.46 6.14 1.21 3.98 0.74 BMDL01 (ng/kg) 1.17 2.34 1.54 2.79 2.60 1.85 0.96 0.50 2.34 1.79 0.37 1.38 0.74 0.81 1.55 0.31 2.70 0.70 1.22 0.44 D RA FT: DO NOT C IT E OR Q U O TE Table 5-15. Comparison of multi-stage modeling results across cancer bioassays using blood concentrations (continued) This document is a draftfo r review purposes only and does not constitute Agency policy. 5-103 Specie Study s Sex N TP Mouse Female (1982, 594255) cont. Morphology: topography Subcutaneous tissue: fibrosarcom a Hem atopoietic system: lym phom a o r leukem ia L ive r: hepatocellular adenoma o r carcinoma Thyroid : fo llic u la r-c e ll adenoma NTP Rat (2006, 197605) Toth et Mouse al. (1979, 197109) M ale Fe m ale M ale Combined tum ors Bayesian analysis Lung: alveolar/bronchiolar adenoma or carcinoma L ive r: hepatocellular adenoma o r carcinoma Combined tum ors Bayesian analysis L ive r: cholangiocarcinom a L ive r: hepatocellular adenoma O ral mucosa: squamous cell carcinoma Pancreas: adenoma o r carcinoma Lung: cystic keratinizing epitheliom a Combined tum ors Bayesian analysis Live r: tum ors Multi-stage modeling:11 stage, GoF _p-value, LL difference 1, p = 0.93 1, p = 0.98 1, p = 0.34 1, p = 0.09, no im provem ent w ith higher orders 1, p = 0.09 1, p = 0.93 3, p = 0.99, d LL = 2.93 3, p = 0.93, d LL = 2.10 1, p = 0.27 1, p= 0.64 2, p = 0.51, d LL = 3.55 1, p = 0.29 BMD01 (ng/kg) 3.40 1.14 1.49 3.05 BMDL01 (ng/kg) 1.69 0.61 0.83 1.44 0.44 0.29 2.53 0.41 0.21 0.14 0.16 0.11 7.57 4.13 10.22 6.53 2.20 1.39 10.52 4.63 8.30 5.24 1.18 0.78 0.37 0.21 *A nalysis uses a chi-square goodness o f fit statistic fo r differences in the log likelihood s (p > 0.05). D RA FT: DO NOT C IT E OR Q U O TE 1 Table 5-16. Individual tum or points of departure and slope factors using 2 blood concentrations 3 Study Tumor Site (Sex/Species) b m d l hed OSF (ng/kg-day) (per mg/kg-day) N TP (1982, 594255) L ive r: adenoma o r carcinoma (m ale m ice) 1 .7 E -0 3 5.8E+6 To th et al. (1979, 197109) L iv e r tum ors (m ale m ice) 1 .9 E -0 3 5.2E+6 N TP , (1982, 594255)a Lung: adenoma o r carcinoma (m ale m ice) 8 .7 E -0 3 1.1E+6 Kociba et al. (1978, 001818) L ive r: adenoma o r carcinoma (fem ale rats) 1 .2 E -0 2 8.6E+5 N TP (1982, 594255) Hem atopoietic: lym phom a o r leukem ia (fem ale m ice) 1 .6 E -0 2 6.4E+5 N TP (1982, 594255)a Thyroid : fo llic u la r cell adenoma (m ale rats) 1 .9 E -0 2 5.2E+5 N TP (1982, 594255) L ive r: neoplastic nodule o r hepatocellular carcinoma (fem ale rats) 2 .1 E -0 2 4.8E+5 N TP (1982, 594255) Adrenal: cortical adenoma o r carcinoma or adenoma, NOS (fem ale rats) 2 .4 E -0 2 4.1E+5 N TP (1982, 594255) L ive r: adenoma o r carcinoma (fem ale m ice) 2 .5 E -0 2 4.0E+5 D ella Porta et al. (1987, 197405) H epatocellular carcinoma (m ale mice) 3 .1 E -0 2 3.2E+5 N TP (1982, 594255)a Adrenal cortex: adenoma (m ale rats) 4 .5 E -0 2 2.2E+5 D ella Porta et al. (1987, 197405)a H epatocellular carcinoma (fem ale m ice) 4 .9 E -0 2 2.0E+5 N TP (1982, 594255) Subcutaneous fibrosarcom a (fem ale rats) 5 .4 E -0 2 1.8E+5 N TP (2006, 197605) O ral mucosa: squamous cell carcinoma (fem ale rats) 5 .5 E -0 2 1.8E+5 N TP (1982, 594255)a Thyroid : adenoma (fem ale m ice) 5 .7 E -0 2 1.7E+5 N TP (1982, 594255) Thyroid : fo llic u la r cell adenoma (fem ale rats) 6 .5 E -0 2 1.5E+5 N TP (1982, 594255) Subcutaneous fibrosarcom a (fem ale m ice) 7 .4 E -0 2 1.4E+5 Kociba et al. (1978, 001818) Lung: carcinoma (fem ale rats) 8 .0 E -0 2 1.2E+5 Kociba et al. (1978, 001818) Adenoma o f adrenal cortex (m ale rats) 8 .5 E -0 2 1.2E+5 D ella Porta et al. (1987, 197405) H epatocellular adenoma (fem ale m ice) 9 .4 E -0 2 1.1E+5 Kociba et al. (1978, 001818) Nasal/Palate: carcinoma (fem ale rats) 1 .2 E -0 1 8.2E+4 Kociba et al. (1978, 001818) Tongue: carcinoma (m ale rats) 1 .4 E -0 1 7.0E+4 N TP (1982, 594255) L ive r: neoplastic nodule o r hepatocellular carcinoma (m ale rats) 1 .5 E -0 1 6.6E+4 Kociba et al. (1978, 001818) Nasal/Palate: carcinoma (m ale rats) 1 .6 E -0 1 6.3E+4 N TP (2006, 197605) L ive r: cholangiocarcinom a (fem ale rats) 2 .9 E -0 1 3.5E+4 N TP (2006, 197605) Pancreas: adenoma o r carcinoma (fem ale rats) 3 .4 E -0 1 2.9E+4 N TP (2006, 197605) Lung: cystic keratinzing epitheliom a (fem ale rats) 4.1E -01 2.4E+4 N TP (2006, 197605) L ive r: hepatocellular adenoma (fem ale rats) 5 .6 E -0 1 1.8E+4 This document is a draftfor review purposes only and does not constitute Agency policy. 5-104 DRAFT--DO NOT CITE OR QUOTE 1 2 T able 5-17. M ultiple tu m o r points of d e p a rtu re and slope factors using blood 3 concentrations 4 Study Sex/species: tumor sites BMDLh e d (ng/kg-day) OSF (per mg/kg-day) N T P (1 9 8 2 , 5 9 4 2 5 5 ) M a le m ice: liv e r a d e n o m a a n d c a rc in o m a , lu n g 1 .1 E -0 3 9 .4 E + 6 N T P (1 9 8 2 , 5 9 4 2 5 5 ) F e m a le m ice: liv e r a d e n o m a a n d c a rc in o m a , th y ro id a d e n o m a , su b c u ta n e o u s fib ro sa rc o m a , a ll ly m p h o m a s 5 .3 E -0 3 1 .9 E + 6 N T P (1 9 8 2 , 5 9 4 2 5 5 ) F e m a le rats: liv e r n e o p la sitc n o d u le s, liv e r a d e n o m a a n d c a rc in o m a , th y ro id fo llic u la r c e ll a d e n o m a ,a d re n a l c o rte x a d e n o m a o r c a rc in o m a 5 .7 E -0 3 1 .8 E + 6 K o c ib a e t al. (1 9 7 8 , 001818) F e m a le rats: liv e r a d e n o m a c a rc in o m a , o ra l ca v ity , lu n g 7 .3 E -0 3 1 .4 E + 6 N T P (1 9 8 2 , 5 9 4 2 5 5 ) M a le rats: th y ro id fo llic u la r c e ll a d e n o m a , a d re n a l co rtex ad e n o m a 9 .6 E -0 3 1 .0 E + 6 N T P (2 0 0 6 , 1 9 7 6 0 5 ) F e m a le rats: liv e r c h o la n g io c a rc in o m a , h e p a to c e llu la r a d e n o m a , o ra l m u c o sa sq u a m o u s c e ll c a rc in o m a , lu n g c y stic k e ra tin iz in g e p ith e lio m a , p a n c re a s ad e n o m a , c a rc in o m a 2 .3 E -0 2 4 .4 E + 5 K o c ib a e t al. (1 9 7 8 , 001818) M a le rats: a d re n a l c o rte x a d e n o m a , to n g u e ca rc in o m a , n a sa l/p a la te c a rc in o m a 5 3 .1 E -0 2 3 .2 E + 5 This document is a draftfor review purposes only and does not constitute Agency policy. 5-105 DRAFT--DO NOT CITE OR QUOTE Table 5-18. Comparison of cancer BMDs, BMDLs, and slope factors for combined or selected individual tumors for 1, 5, and 10% extra risk This document is a draftfo r review purposes only and does not constitute Agency policy. 5-106 Study Species Sex Kociba Rat (1978, 001818)a Female Male NTP (1982, 594255)a Rat Female Male Mouse Female Male NTP Rat (2006, 197605)a Female Della Porta Mouse et al. (1987, 197405)b Male Female Toth et al., Mouse Male (1979 197109)c BMD01 (ng/kg) 4.9E-01 1.5E+00 4.4E-01 6.9E-01 4.3E-01 1.5E-01 1.1E+00 7.1E+00 2.3E+00 3.7E-01 B M D L 01 SF01 BMD05 (ng/kg) (ng/kg) 1 (ng/kg) 3.8E-01 2.7E-02 2.5E+00 9.6E-01 1.0E-02 7.2E+00 3.2E-01 4.5E-01 3.0E-01 1.1E-01 3.2E-02 2.2E-02 3.4E-02 9.4E-02 2.2E+00 3.5E+00 2.1E+00 7.7E-01 7.8E-01 1.3E-02 4.8E+00 1.2E+00 8.5E-03 1.4E+01 1.5E+00 6.5E-03 1.0E+01 2.1E-01 4.8E-02 1.9E+00 BMDL05 (ng/kg) 1.9E+00 4.8E+00 1.6E+00 2.2E+00 1.5E+00 5.4E-01 3.6E+00 5.0E+00 6.8E+00 1.1E+00 SF05 BMD10 (ng/kg) 1 (ng/kg) 2.7E-02 4.9E+00 1.0E-02 1.5E+01 3.2E-02 2.2E-02 3.4E-02 9.4E-02 4.4E+00 6.9E+00 4.3E+00 1.5E+00 1.4E-02 8.2E+00 1.0E-02 2.0E+01 7.3E-03 2.1E+01 4.7E-02 3.9E+00 BMDL10 (ng/kg) 3.8E+00 9.6E+00 3.2E+00 4.5E+00 3.0E+00 1.1E+00 6.6E+00 9.7E+00 1.4E+01 2.2E+00 SF10 (ng/kg) 1 2.7E-02 1.0E-02 3.2E-02 2.2E-02 3.4E-02 9.4E-02 1.5E-02 1.0E-02 7.1E-03 4.6E-02 "Combined tumors, Bayesian analysis. bHepatocellular carcinomas for both males and females. "Hepatocellular carcinomas. TCDD blood concentrations from Emond rodent PBPK models. SF = BMR ^ BMDLbmr, where BMR = 0.01, 0.05, or 0.10. D RA FT: DO NOT C IT E OR Q U O TE Table 5-19. TCDD human-equivalent dose (HED) BMDs, BMDLs, and oral slope factors (OSF) for 1, 5, and 10% extra risk This document is a draftfo r review purposes only and does not constitute Agency policy. 5-107 Study Species Sex BMD01 BMDL01 OSF01 BMD05 BMDL05 OSF05 BMD10 BMDL10 OSF10 (ng/kg-d) (ng/kg-d) (ng/kg-d) 1 (ng/kg-d) (ng/kg-d) (ng/kg-d) 1 (ng/kg-d) (ng/kg-d) (ng/kg-d) 1 Kociba Rat (1978, 001818)a Female 1.1E-02 7.4E-03 Male 5.9E-02 3.1E-02 1.4E+00 3.3E-01 1.3E-01 8.6E-02 6.6E-01 3.6E-01 5.8E-01 1.4E-01 3.8E-01 2.59E-01 4.0E-01 1.8E+00 9.7E-01 1.0E-01 NTP (1982, 594255)a Rat Mouse Female Male Female 9.7E-03 1.9E-02 9.1E-03 5.8E-03 9.7E-03 5.4E-03 1.7E+00 1.0E+00 1.9E+00 1.1E-01 2.2E-01 1.1E-01 6.6E-02 1.1E-01 6.0E-02 7.6E-01 4.5E-01 8.3E-01 3.2E-01 6.2E-01 3.0E-01 1.9E-01 3.3E-01 1.8E-01 5.2E-01 3.1E-01 5.7E-01 Male 1.9E-03 1.2E-03 8.3E+00 2.2E-02 1.3E-02 3.8E+00 6.4E-02 3.8E-02 2.7E+00 NTP Rat (2006, 197605)a Female 4.1E-02 2.3E-02 4.4E-01 3.6E-01 2.4E-01 2.1E-01 7.9E-01 5.7E-01 1.8E-01 Della Porta Mouse et al. (1987, 197405)b Male Female 5.2E-01 9.2E-02 3.1E-02 4.9E-02 3.2E-01 2.0E-01 1.7E+00 3.8E-01 1.1E+00 6.0E-01 1.3E-01 8.3E-02 2.8E+00 2.9E+00 1.0E+00 1.7E+00 1.0E-01 5.9E-02 Toth et al. Mouse Male (1979, 197109)c 5.1E-03 1.9E-03 5.3 E+00 6.7E-02 2.7E-02 1.9E+00 2.0E-01 8.5E-02 1.2 E+00 aCombined tumors, Bayesian analysis. bHepatocellular carcinomas for both males and females. cHepatocellular carcinomas. HEDs from Emond human PBPK model corresponding to blood concentration BMDs and BMDLs in Table F3-1. OSF = BMR ^ BMDLbmr, where BMR = 0.01, 0.05, or 0.10. D RA FT: DO NOT C IT E OR Q U O TE Table 5-20. Illustrative RfDs based on tumorigenesis in experimental animals This document is a draftfo r review purposes only and does not constitute Agency policy. Study Species, strain (sex) Protocol NTP (1982, 594255) Mouse, B6C3F1, 2-year gavage; male n = 50 Toth et al. Mouse, Swiss/ (1979, 197109) H/Riop, male 1-year gavage (1-year average); n = 38-44 NTP (1982, 594255) Mouse, B6C3F1, 2-year gavage; female n = 50 NTP (1982, 594255) Rat, Osborne- 2-year gavage; Mendel, female n = 50 Kociba et al. Rat, S-D, female (1978, 001818) NTP (1982, 594255) Rat, OsborneMendel, male Della Porta et al. Mouse, B6C3F1, (1987, 197405) male NTP (2006, 197605) Rat, S-D, female 2-year dietary; n = 50 2-year gavage; n = 50 1-year gavage; n = 40-50 2-year gavage; n = 53 Kociba et al. Rat, S-D, male (1978, 001818) 2-year dietary; n = 50 Endpoint Liver adenoma and carcinoma, lung Liver tumors Liver adenoma and carcinoma, thyroid adenoma, subcutaneous fibrosarcoma, all lymphomas Liver neoplasitc nodules, thyroid follicular cell adenoma, liver adenoma and carcinoma, adrenal cortex adenoma or carcinoma Liver adenoma carcinoma, oral cavity, lung Thyroid follicular cell adenoma, adrenal cortex adenoma Hepatocellular carcinoma Liver cholangiocarcinoma, hepatocellular adenoma, oral mucosa squamous cell carcinoma, lung cystic keratinizing epithelioma, pancreas adenoma, carcinoma Adrenal cortex adenoma, tongue carcinoma, nasal/palate carcinoma aBMR = 0.01. bUF = 30; UFa = 3, UFH= 10. BMDLHEDa (ng/kg-day) 1.1E-3 RfDb (mg/kg-day) 3.6E-11 1.9E-3 6.3E-11 5.3E-3 5.7E-3 1.7E-10 1.9E-10 7.3E-3 9.6E-3 3.1E-02 3.1E-2 2.4E-10 3.2E-10 1.0E-9 1.0E-9 3.1E-2 1.0E-9 5-108 D RA FT: DO NOT C IT E OR Q U O TE Table 5-21. Illustrative RfDs based on hypothesized key events in TCDD's MOAs for liver and lung tumors This document is a draftfo r review purposes only and does not constitute Agency policy. 5-109 Key event NO(A)ELh e d LO(A)ELh e d Endpoint and exposure duration (ng/kg-day) (ng/kg-day) BMDLh e d 3 (ng/kg-day) Liver tumors Changes in gene expression CYP1A1 mRNA, 1 day 1.8E-05 3.4E-04 2.3E-03c (Appendix H) Changes in gene expression Benzo(a)pyrene hydroxylase (BPH) activity (CYP1A1), 1 day 9.2E-04 6.0E-03 4.6E-04cd (Appendix H) EROD (CYP1A1), 53 weeks none 1.4E-01 9.5E-03c (Appendix H) Oxidative stress DNA single-strand breaks, 90 days none 3.3E-02 2.2E-02c (Appendix H) TBARS, 90 days - - 4.4E-02 (Appendix H) Cytochrome C reductase, 90 days - - 8.8E-02 (Appendix H) Hepatotoxicity Toxic hepatopathy, 2 years none 1.4E-01 1.8E-01c (Appendix E) Hepatocyte hypertrophy, 31 weeks 9.3E-02 3.3E-01 8.8E-03 (Appendix E) Hepatocellular proliferation Labeling index, 31 weeks none 1.4E-01 6.6E-02c (Appendix H) Lung tumors Metabolic enzyme induction EROD (CYP1A1), 53 weeks none 1.4E-01 2.9E-04c (Appendix H) Retinoid homeostatsis Hepatic retinol and retinyl palmitate, 90 days none 1.1E+00 1.7E-01c (Appendix E) aBMR for continuous endpoints-- 1 standard deviation; for quantal endpoints-- 10%. bBolded NOAEL, LOAEL, or BMDL is selected POD; poorly-fitting BMDLs above the LOAEL not used. cPoor BMD model fit or no good model fit. dCould be higher depending on the effect of background exposureoor (see Section 5.3.2.1). eUF = 30; UFa = 3; UFH= 10. fUF = 300; UFA = 3; UFH = 10; UFL = 10. RfDb (mg/kg-day) Study 6E-13d,e 2E-11d,e 3E-10e Vanden Heuvel et al. (1994, 594318) Kitchin and Woods (1979, 198750) NTP (2006, 197605) 7E-10e 2E-09e 3E-09e 5E-09f Hassoun et al. (2000, 197431) Hassoun et al. (2000, 197431) Hassoun et al. (2000, 197431) NTP (2006, 197605) 3E-10e NTP (2006, 197605) 2E-09e NTP (2006, 197605) 1E-11e 6E-09e NTP (2006, 197605) Van Birgelen et al. (1995, 198052) D RA FT: DO NOT C IT E OR Q U O TE 1 Table 5-22. Comparison of principal epidemiological studies 2 Strengths Weaknesses Study C u m u la tiv e T C D D le v e ls in th e se ru m w e re e stim a te d o n a n in d iv id u a l-le v e l b a sis fo r th e 3 ,5 3 8 w o rk e rs. E v a lu a te d e ffe c t o f la g p e rio d s (0 a n d 15 y ears). M e a su re d a n d b a c k -e x tra p o la te d T C D D c o n c e n tra tio n s to re fin e a n d q u a n tify jo b e x p o su re m a tric e s, w h ic h w e re th e n u s e d to e stim a te d io x in c u m u la tiv e d o se fo r e a c h m e m b e r o f th e ir e n tire co h o rt. In te rn a l c o h o rt c o m p a riso n s (C o x re g re ssio n m o d e l). B a c k g ro u n d e x p o su re e stim a te d . E x p o su re to o th e r c h lo rin a te d h y d ro c a rb o n s (d io x in lik e c o m p o u n d s). E x tra p o la tio n o f d o se fro m a sm all su b se t (ro u g h ly 5 % , n = 170) o f th e co h o rt. N IO S H co h o rt S te e n la n d e t al. (2 0 0 1 , 1 9 7 4 3 3 ) S e ru m fa t o r b o d y fa t le v e ls o f T C D D w ere b a c k -c a lc u la te d u sin g a sim p le first-o rd e r m o d e l. H a lf-life o f T C D D is v a ria b le b u t sim u la te d a s a co n sta n t. C h a n g e s in th e lip id fra c tio n o f b o d y w e ig h t o r p re se n c e /a b se n c e o f g e n e tic d iffe re n c e s in h u m a n s th a t a lte r th e d istrib u tio n a n d m e ta b o lism o f T C D D w e re n o t c o n sid e re d . S e ru m lip id le v e ls o f T C D D in 198 8 w e re m e a su re d o n ly a t o n e o f th e e ig h t p la n ts in th e stu d y . N o fo llo w -u p m e a su re s. T h e e stim a te s o f d o se a re b a se d o n b lo o d sam p les ta k e n d ec ad es a fte r ex p o su re. C u m u la tiv e T C D D le v e ls in th e se ru m w e re e stim a te d o n a n in d iv id u a l-le v e l b a sis fo r th e 3 ,5 3 8 w o rk e rs. T C C D d o se e stim a te s w e re sim u la te d w ith a k in e tic m o d e l th a t in c lu d e d c o n sid e ra tio n s o f e x p o su re in te n sity a n d a g e -d e p e n d e n t b o d y w e ig h t a n d fa t le v e ls. E x tra p o la tio n o f d o se fro m a sm all su b se t (ro u g h ly 5 % , n = 170) o f th e co h o rt. T h e a u th o rs re p o rte d th e C A D M m o d e l p ro v id e d a n im p ro v e d fit o v e r th e o n ec o m p a rtm e n ta l m o d e l, b u t n o e v id e n c e w a s rep o rted reg ard in g an y fo rm a l te st o f sta tistic a l sig n ific a n c e . N IO S H co h o rt C h e n g e t al. (2 0 0 6 , 5 2 3 1 2 2 ) E v a lu a te d e ffe c t o f la g p e rio d s (0 a n d 15 y ears). B a c k g ro u n d e x p o su re e stim a te d . S tra tifie d risk e stim a te s fo r sm o k in g a n d n o n sm o k in g . R a c e a n d a g e ad ju stm en ts. S e ru m lip id le v e ls o f T C D D in 198 8 w e re m e a su re d o n ly a t o n e o f th e e ig h t p la n ts in th e stu d y . N o fo llo w -u p m e a su re s. T h e e stim a te s o f d o se a re b a se d o n b lo o d sam p les ta k e n d ec ad es a fte r ex p o su re. E x p o su re to o th e r c h lo rin a te d h y d ro c a rb o n s (d io x in lik e c o m p o u n d s). In te rn a l c o h o rt n o te d a n in v e rse -d o se re sp o n se fo r h ig h -e x p o su re g ro u p s a n d th u s e x c lu d e d th e d a ta re su ltin g in stro n g e r a sso c ia tio n s. N o c o n s id e ra tio n fo r re c e n t e x p o su re s to T C D D , c h a n g e s in th e lip id fra c tio n o f b o d y w e ig h t o r p re se n c e /a b se n c e o f g e n e tic d iffe re n c e s in h u m a n s th a t a lte r th e d istrib u tio n a n d m e ta b o lism o f T C D D c o u ld c a u se m isc la ssific a tio n . This document is a draftfor review purposes only and does not constitute Agency policy. 5-110 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-22 Comparison of principal epidemiological studies (continued) Strengths Repeated TC D D measures in serum in 48 individuals. Used to estimate h a lf-life fo r study cohort. Took into account the age and body fa t percentage o f the workers. Measured and back-extrapolated TC D D concentrations to quantify exposures fo r the rem aining cohort members using 5 d ifferent w orking areas o f the plant. Evaluated effect o f lag periods up to 20 years. M u ltip le statistical models used to evaluate fatal cancer slope estimates. Background exposure estimated. Weaknesses Exposure to other chlorinated hydrocarbons (d io xin like compounds), H C H , and lindane. Extrap olation o f dose from a sm all subset (roughly 4%, n = 1,189) o f the cohort. Serum fa t or body fa t levels o f TC D D were back-calculated using a sim ple first-ord er model. Presence/absence o f genetic differences in humans that alter the d istrib ution and m etabolism o f TC D D were not considered. Serum lip id levels o f TC D D fo r only 275 w o rke rs. Study Becher et al. (1998, 197173); H am b urg C oh ort B oth internal and external analyses. Adjustm ent fo r age, B M I, and smoking. B oth cancer incidence and cancer m ortality data available, although results somewhat discordant, w ith steeper dose-response seen fo r cancer m ortality. Acute dose due to accident may not be O tt and Zober comparable to chronic dose accumulated over (1996, 198408) a long tim e, as in m ost environm ental exposures. R ela tive ly sm all num ber o f cancer deaths compared to N IO SH and Ham burg cohorts (n = 31). Serum TC D D levels measured 30 years after accident, requiring extrapolation back in tim e to estimate cum ulative dose over tim e. Serum TC D D levels measured only on a sample o f the cohort (138 out o f 243), requiring assumptions about sim ila rities in exposure scenario fo r other w orkers to estimate th e ir exposure This document is a draftfor review purposes only and does not constitute Agency policy. 5-111 DRAFT--DO NOT CITE OR QUOTE Table 5-22 Comparison of principal epidemiological studies (continued) Strengths TC D D levels measured in a ll 891 members o f this fem ale cohort. M ost TC D D measurements based on observed levels in stored serum at the tim e o f the accident in 1976, no extrapolation needed to estimate past levels. Internal analyses. Evaluates fem ale cancer incidence, other studies evaluate male cancer m ortality. Presumed adjustm ent fo r age and potential breast cancer confounders (15 o f 21 cancers were breast cancer). Weaknesses Study Acute dose due to accident may not be W arner et al. comparable to chronic dose accumulated over (2002, 197489) a long tim e, w hich is typ ical o f m ost environm ental exposures. D id not evaluate d ifferent lag periods. N ot clear if any adjustm ent fo r confounders. Sm all number o f cancers (n = 21). Doses know n in 1976, require assumptions about excretion over tim e to estimate cum ulative dose (9 year h a lf life assumed), presumed m etric o f prim ary interest. No more recent TC D D concentration data used. Reported log10transform ation o f the exposure estimates in th e ir regression analysis. 1 2 3 Table 5-23. Added background TEQ exposures to blood TCDD/TEQ 4 concentrations in ratsa 5 Background TEQ added None Est. TCDD onlyb Est. TEQc 2x Est. TEQd 10X Est. TCDDe 0 0.064 0.19 0.38 0.64 2.56 2.62 2.75 2.94 3.20 5.69 5.75 5.88 6.07 6.33 9.79 9.85 9.98 10.1 10.5 16.6 16.7 16.8 17.0 17.2 29.7 29.8 29.9 30.1 30.3 6 7 "Background exposures estimated fro m N TP (2006, 543749); rat TC D D concentrations fro m N TP (2006, 197605)). 8 '"Estim ated fro m TC D D fa t concentration measurements in N TP (2006, 543749). 9 cEstim ated fro m combined TC D D , PeCDF, and PCB-126 fa t concentration measurements in N TP (2006, 543749). 10 dAssumes that measured congeners comprise 50% o f actual TE Q exposure. 11 eAssumes that TC D D comprises 10% o f total background TEQ exposure. This document is a draftfor review purposes only and does not constitute Agency policy. 5-112 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-24. Effect of added background TEQ exposure on BMDL01 for 2 cholangiocarcinomas in rats (NTP, 2006, 197605) 3 Background TEQa Added exposure (ng/kg blood TEQ) BMDLoib (ng kg blood) Nonec 0 4.14 Est. TCDD only 0.064 4.19 Est. TEQ 0.19 4.30 2* Est. TEQ 0.38 4.45 10x Est. TCDD 0.64 4.65 4 5 "S c e n a rio s a s in T a b le 5 -2 0 . 6 '"M u l t i s t a g e m o d e l r e s u l t s f r o m B M D S v e r s i o n 2 . 1 . 1 ( s e e A p p e n d i x I f o r m o d e l i n g d e t a i l s ) . 7 cS a m e r e s u lt a s f o r th e s in g le tu m o r m o d e lin g p r e s e n te d p r e v io u s ly in th is s e c tio n . 8 9 10 Table 5-25. NIOSH cohort septile data with added TEQ background3 11 Septile TCDD serum level TCDD + background TEQ Relative increase (ppt-yr) (ppt-yr) (%) 1 260 2,960 1,040 2 402 3,102 770 3 853 3,553 320 4 1,895 4,595 140 5 4,420 7,120 60 6 12,125 14,825 20 7 59,838 62,538 5 12 13 a S e p t i l e d a t a f r o m S t e e n l a n d e t a l . ( 2 0 0 1 , 1 9 7 4 3 3 ) ; c u m u l a t i v e b a c k g r o u n d T E Q e s t i m a t e f r o m C r u m p e t a l . 14 ( 2 0 0 3 , 1 9 7 3 8 4 ) ; b o t h b a s e d o n e s t i m a t e s b y W H O ( 1 9 9 8 ) . This document is a draftfor review purposes only and does not constitute Agency policy. 5-113 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 5-1. Mechanism of altered gene expression by AhR. The regulation of 3 gene expression by T C D D in mammalian cells requires binding o f the xenobiotic 4 to the aryl hydrocarbon receptor (A h R ). The A h R is part o f a multi-protein 5 complex that includes heat shock proteins and various kinases and other post 6 translational m odifying factors. Upon ligand binding, the A h R heterodimerizes 7 w ith the aryl hydrocarbon receptor nuclear translocator (A rnt) and binds to dioxin 8 response elements ( D R E s ) found in target genes. Alternatives to D RE-dependent 9 gene expression exist whereby the A h R complex associates with other 10 transcription factors and results in a cross-talk between these systems. The 11 culm ination o f regulation o f A h R targets genes (both increases ad decreases in 12 transcription) results in an alteration in cellu lar phenotypes, including changes in 13 intracellular m etabolism and changes in cell cy cle regulation. This document is a draftfor review purposes only and does not constitute Agency policy. 5-114 DRAFT--DO NOT CITE OR QUOTE Liver Lung Thyroid This document is a draftfo r review purposes only and does not constitute Agency policy. 5-115 TCDD 1 AhR \. Changes in Gene Expression '' Hepatotoxicity TCDD 1 i AhR 1 1 Oxidative Stress l Co-carcinogens Changes in j / GeneExirEe s s ion .... * t Metabolic Retinoid Homeostasis (CEynpzsy, mCOeXs2) Toxicity TCDD 1 AhR I Hepatic UGT1 Liver Decreased T4 i IncreasedTTSH J Cellular Proliferation Adenoma and Carcinoma \ Proliferation Adenoma and Carcinoma (7 Proliferation Thyroid Adenoma and V Carcinoma J Figure 5-2. TCDD's hypothesized modes of action in site-specific carcinogenesis. See text for details. In each instance, the solid arrows depict pathways that are w ell-established and are associated w ith low uncertainty. The dashed arrows represent connections that are less established and are associated with higher uncertainty. D RA FT: DO NOT C IT E OR Q U O TE 1 2 Figure 5-3. EPA's process to select and identify candidate OSFs from key 3 animal bioassays for use in the cancer risk assessment of TCDD. 4 For each cancer study that qualified for TCDD dose-response assessment using the study inclusion criteria, 5 EPA first selected the species/sex/tumor combinations with statistically significant increases in tumor 6 incidence by either a pair-wise test between the treated group and the controls or by a trend test showing 7 increases in tumors with increases in dose. Next, EPA used an animal kinetic model to estimate blood 8 concentrations corresponding to the study average daily administered doses for use in dose response 9 modeling. BMDL01's were then estimated for the blood concentrations by, (1) using the linearized 10 multistage model for each species/sex/tumor combination within each study, and (2) using the linearized 11 multistage model within a Bayesian Markov Chain Monte Carlo framework that assumes independence of 12 tumors and modeling all tumors together for each species/sex combination within each study. Using the 13 human kinetic model, human equivalent doses (BMDLHEDs) were then estimated for each of the BMDL01s 14 and oral slope factors were calculated by OSF = 0.01/BMDLhed. The lowest OSF for a species/sex 15 combination for either a single tumor type or all tumors combined was selected as a candidate OSF for 16 TCDD risk assessment. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-116 D R A F T -- D O N O T C IT E O R Q U O T E 1 2 3 Figure 5-4. Dose-response model shape This document is a draftfor review purposes only and does not constitute Agency policy. 5-117 DRAFT--DO NOT CITE OR QUOTE 1 2 3 Figure 5-5. Comparison of individual and population dose-response curves; 4 a simple illustration. 5 This document is a draftfor review purposes only and does not constitute Agency policy. 5-118 DRAFT--DO NOT CITE OR QUOTE 1 A. Full response range 2 Multistage C ancer Model with 0.95 Confidence Level 09:48 04/22 2010 3 B . Low -dose region C holangiosarcom a low dose Blood Conc (ng/kg) 4 Figure 5-6. Multistage benchm ark dose modeling of NTP (2006, 197605) 5 cholangiosarcoma data. This document is a draftfor review purposes only and does not constitute Agency policy. 5-119 DRAFT--DO NOT CITE OR QUOTE 1 A. Full response range Composite Risk of Dioxin in NTP (2006a) female rats 2 EXPERIMENTAL DOSE RANGE 3 B . Low -dose region Composite Risk of Dioxin in NTP (2006a) female rats around BMD01 4 DOSE 5 Figure 5-7. Multistage benchm ark dose modeling of NTP (2006, 197605) 6 combined tumor data. This document is a draftfor review purposes only and does not constitute Agency policy. 5-120 DRAFT--DO NOT CITE OR QUOTE 1 A. 2 Kohn and Melnick (2002) Figure 5 on log Scale (KR.X=3300) 3 4 5 6 7 8 9 10 log10 Concentration Xin nM 11 12 B . Kohn and Melnick (2002) Figure 5 on Arithmetic Scale (KR.X=3300) 13 14 15 16 17 18 19 20 21 Concentration X in nM 22 23 Figure 5-8. Estrogen receptor-mediated response-modeling plot from Kohn 24 and Melnick (2002, 199104): low-dose region shown. This document is a draftfor review purposes only and does not constitute Agency policy. 5-121 DRAFT--DO NOT CITE OR QUOTE Liver TCDD Ah1R genCehaenxgperesssion Osxtidreastsive Hepatotoxicity CYP1A1 mRNA, 1 day Vanden Heuvel et al., 1994 EROD (CYP1A1), 53 weeks NTP, 2006 TBARS, 90 days Hassoun et al., 2000 Toxic hepatopathy, 2 years NTP, 2006 Hperpoalitfoecraetlilounlar Labeling index, 31 weeks NTP, 2006 cAadrecainnnodommaa 1 2 Figure 5-9. Representative endpoints for each of the hypothesized key events 3 following AhR activation for TCDD-induced liver tumors. This document is a draftfor review purposes only and does not constitute Agency policy. 5-122 DRAFT--DO NOT CITE OR QUOTE Lung Hepatic retinol, 90 days Van Birgelen et al., 1995a TCDD Ah1R genCehaenxgperesssinioCnocarcinogens -- Retinoid homeostatsis Metabolic (Cyepnsz,yCmOesX2) Toxicity EROD (CYP1A1), 53 weeks NTP, 2006 Proliferation Adenoma and carcinoma 1 2 3 Figure 5-10. Representative endpoints for two hypothesized key events 4 following AhR activation for TCDD-induced lung tumors. 5 This document is a draftfor review purposes only and does not constitute Agency policy. 5-123 DRAFT--DO NOT CITE OR QUOTE 1.0E+07 Cancer Slope Factors for 2,3,7,8-TCDD Human Mouse Rat This document is a draftfo r review purposes only and does not constitute Agency policy. 5-124 D RA FT: DO NOT C IT E OR Q U O TE O 03 QoO- -a 1.0E+06 QOI- ^<QD. to cm' 1.0E+05 ^ a.At e iV n v Ai ' / '4? / / 4 4 4' ? ^ "' ? '> ^ ^ O^' /it>50" xx< f <>s- ' x t .A A -Oi xr c^ f 41 x>r nr5_5ONP T <*>&&7J>b'' JF&, &r ^ & & &U \& \ *A V r'T 9r K-x> \5j^y ^ A A x E<c r N ry j y x<?' <v K&V^' 'N A &v' A E 9f Figure 5-11. Candidate oral slope factor array. 1 6. FEASIBILITY OF QUANTITATIVE UNCERTAINTY ANALYSIS 2 FROM NAS EVALUATION OF THE 2003 REASSESSMENT 3 4 5 6.1. INTRODUCTION 6 Th is section focuses on the third area for improvement in the 2003 Reassessm ent that w as 7 identified by the National Academ y o f Sciences (N A S ) review committee (N A S , 2006, 198441), 8 i.e., im proving transparency, thoroughness, and clarity in quantitative uncertainty analysis. 9 Although the N A S committee summarized the shortfalls in the 2003 Reassessm ent categorically, 10 the elaborations w ithin their report often contain the qualification "i f possible" and do not take a 11 position w ith regard to the feasibility o f m any o f its suggestions. W ith appreciation for the 12 extent o f inform ation available for dioxin, the goal o f this section is to circum scribe the 13 feasibility o f a data-driven quantitative uncertainty analysis for T C D D dose-response 14 assessment. Fo llo w in g b rie f highlights o f the evolution o f quantitative uncertainty analysis for 15 such applications, this section lays out definitions o f key terms, review s E P A ' s position 16 regarding cancer and noncancer endpoints, sum m arizes the N A S critique, and evaluates the 17 feasibility o f quantitative uncertainty analysis for T C D D w ithin the fram ew ork o f E P A ' s 18 noncancer R f D and cancer slope factor dose-response methodologies. 19 20 6.1.1. Historical Context for Q uantitative Uncertainty Analysis 21 Th e b asic methods o f probabilistic risk assessm ent (P R A ) w ere developed in the 22 aerospace program in the 1960s, and they found their first full-scale application in the 23 U .S . N u clear Regulatory C o m m issio n ' s (U .S . N R C ' s) Reactor Safety Study o f 1975-- including 24 accident consequence analysis and uncertainty analysis (U .S . N R C , 1975, 543729) . T h is study, 25 com m only referred to as the Rasm ussen Report after its lead author, is considered to be the first 26 modern P R A . In the aftermath o f the 1979 Three M ile Island accident, a new generation of 27 P R A s appeared in w hich some o f the methodological problems o f the 1975 study were avoided. 28 These advances w ere reflected in the C o m m issio n ' s Fault Tree Handbook (U .S . N R C , 1981, 29 543730) and P R A guide (U .S. N R C , 1983, 543732), w hich shored up and standardized much of 30 the risk assessment methodology. A n extensive chapter o f the latter w as devoted to uncertainty 31 and sensitivity analysis. These documents form ed the basis for standards and guidelines This document is a draftfor review purposes only and does not constitute Agency policy. 6-1 DRAFT--DO NOT CITE OR QUOTE 1 established by other agencies, including the U .S . Department of Energy (U .S. D O E , 1992, 2 543733) and National Aeronautics and Space Administration (N A S A , 2002, 543734) . 3 In 1991, a set o f U .S . N R C studies known as N U R E G 1150 used structured expert 4 judgment to quantify uncertainty and set new standards for uncertainty analysis, in particular 5 w ith regard to expert elicitation (U .S . N R C , 1991, 543736) . Th is w as follow ed by a joint 6 U .S.-European U nion (E U ) program for quantifying uncertainty in accident consequence models. 7 Expert judgment methods were further elaborated in those evaluations, as w ell as screening, 8 dependence modeling and sensitivity analysis (E C , 2009, 543738) . Studies building off of this 9 w ork have performed a large-scale uncertainty analysis o f European consequence models and 10 provided extensive guidance on identifying important variables; selecting, interview ing and 11 com bining experts; propagating uncertainty; inferring distributions on model parameters; and 12 com m unicating results, as documented by G oossens et al. (1996, 548727; 1997, 543752; 1998, 13 548726; 2001, 548730; 2001, 548731; 2001, 548732; 2001, 548735; 2001, 548737; 2001, 14 548738; 2001, 548734) and others (B ro w n et al., 1997, 543739; H arper et al., 1995, 2 02 31 7 ; 15 2002, 198124) . 16 Th e N ational R esearch C o u n cil (N R C ) has been a persistent vo ice in urging the 17 government to enhance its risk assessm ent methodology beginning w ith its report on risk 18 assessm ent in the federal government (N R C , 1983, 194806) . The C o u n c il' s 1989 report, 19 Improving Risk Communication, inveighed against m inim izing the existence o f uncertainty and 20 noted the importance o f considering the distribution o f exposure and sensitivities in a population 21 (N R C , 1989, 000858) . Th e issue o f uncertainty w as a clear concern in subsequent reports, 22 including those assessing human exposure to airborne pollutants (N R C , 1991, 037823) . B u ild in g 23 on its evaluation o f Issues in Risk Assessment (N R C , 1993, 078637), the landm ark study Science 24 and Judgment in Risk Assessment (N R C , 1994, 006424) gathered m any o f these themes in a plea 25 for quantitative uncertainty analysis as "the only w ay to combat the false sense o f certainty 26 w h ich is caused by a refusal to acknowledge and (attempt to) quantify the uncertainty in risk 27 predictions." A subsequent report, Estimating the Public Health Benefits o fProposed Air 28 Pollution Regulations (N R C , 2002, 035312) , identified three barriers to the broad acceptance o f 29 recent E P A health benefit analyses: (1) the large amount o f uncertainty inherent in these 30 analyses, (2) the manner in w hich E P A deals with this uncertainty, and (3) " ... projected health 31 benefits are often reported as absolute num bers o f avoided death or adverse health outcomes This document is a draftfo r review purposes only and does not constitute Agency policy. 6-2 D R A F T -- D O N O T C I T E O R Q U O T E 1 without a context of population size or total numbers of outcomes." The Council encouraged 2 EPA to "explore alternative options for incorporating expert judgment into its probabilistic 3 uncertainty analyses." 4 In an early 2009 report, Science and Decisions: Advancing Risk Assessment, the NRC 5 committee on improving risk analysis encouraged EPA to harmonize approaches for cancer and 6 noncancer dose-response assessment (NRC, 2009, 194810), which involves uncertainty issues 7 discussed in this section. Even more recently, EPA released a draft white paper, Using 8 Probabilistic Methods to Enhance the Role o fRisk Analysis in Decision Making (U.S. EPA, 9 2009, 522927). Although not focused specifically on quantitative uncertainty analysis, there is 10 overlap with the issues treated here, and relevant insights are anticipated from ongoing efforts in 11 this area. 12 13 6.1.2. Definition of Terms 14 For purposes of this study, the following definitions are adopted:52 15 16 Uncertainty Characterization. This consists of a Structured Uncertainty Narrative and, if 17 the uncertainty is supported by quantitative models, Quantitative Uncertainty Analysis. 18 Structured Uncertainty Narrative. This identifies the assumptions conditional on which 19 uncertainty is to be characterized and delineates the type of arguments with supporting 20 evidence that buttress these assumptions. 21 Quantitative Uncertainty Analysis. This is a quantification of the uncertainty attending 22 the use of quantitative models. It applies to a mathematical model of physical 23 phenomena, some of whose parameter values are not known with certainty. A joint 24 distribution is assigned to uncertain model parameters and propagated through the model 25 to yield a joint distribution over the model output. Thus, a quantitative uncertainty 26 analysis always has a joint distribution over model outputs as its result. 27 Joint Distribution/Marginal Distribution. For a set of uncertain quantities, a joint 28 distribution is an assignment of probabilities (or probability densities) for each possible 29 combination of values of these quantities. Each uncertain quantity has a marginal 30 distribution, that is, an assignment of probabilities (or probability densities) to each 31 possible value of that quantity. Assigning a marginal distribution to each quantity is not 32 equivalent to assigning a joint distribution to the set of quantities, unless the quantities 33 are independent; in this case the joint distribution is just the product of the margins. 52M a n y o f t h e s e d e f i n i t i o n s a r e s t a n d a r d t e r m s i n p r o b a b i l i t y a n d s t a t i s t i c s , a s d e s c r i b e d i n S a l t e l l i e t a l . ( 2 0 0 0 , 5 4 3 7 5 6 ), C o x (2 0 0 6 , 5 9 4 3 4 2 ), K u ro w ic k a a n d C o o k e (2 0 0 6 , 5 4 3 7 5 8 ), a n d N R C (2 0 0 7 , 5 4 3 7 4 8 ) ; so m e a re re fle c te d in c u rre n t A g e n c y p ra c tic e (U .S . E P A , 2 0 0 9 , 5 2 2 9 2 7 ) . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-3 DRAFT--DO NOT CITE OR QUOTE 1 Qualitative/Informal Uncertainty Analysis. This assembles the arguments and evidence 2 and provides an assessment of their plausibility in terms of verbal modifiers. The 3 meaning of verbal modifiers such as "likely/unlikely" or "plausible/implausible" in the 4 natural language53 is indeterminate and context dependent. The way in which these 5 qualifiers combine in the natural language requires critical attention from a quantitative 6 viewpoint. (For example, if A is likely and B is likely and C is likely, is A and B and C 7 likely?) It is sometimes claimed that the probability formalism does not capture the way 8 people reason with uncertainty, and many alternatives have been proposed.54 9 This is not the place to discuss foundational issues, except to remark that the practitioner 10 wishing to depart from the standard probability formalism should carefully explore the 11 whole range of alternatives and critically examine the operational meaning of the 12 primitive notions. 13 Sensitivity Analysis. If a quantitative model uses "nominal values" (approximations of the 14 real values) for various input parameters, a sensitivity analysis is performed by choosing 15 different values for these parameters and re-running the model to assess the impact of 16 changes in these parameters on model output. Applicable methods include one- and 17 two-at-a-time methods, design of experiments and Morris's method (Saltelli et al., 2000, 18 543756). They aim at estimating first- and perhaps higher-order effects with a minimal 19 number of model runs, by systematically varying the nominal values. In large 20 uncertainty analyses, sensitivity analysis is used to screen variables for in-depth 21 uncertainty quantification, and thus is part of a quantitative uncertainty analysis 22 (Kurowicka and Cooke, 2006, 543758). As a note, the NAS committee report (NRC, 23 2006) does not distinguish between uncertainty and sensitivity analysis. In fields which 24 have not developed a tradition in uncertainty quantification, the spread of values 25 generated by a sensitivity analysis is sometimes presented as a representation of 26 uncertainty (Murphy et al., 2004, 543741). The question of whether this is or is not the 27 case is moot so long as the uncertainty on model input parameters is not quantified. 28 Systematically varying input values is not the same as sampling input parameter values 29 from their uncertainty distributions. In any event, a systematic approach to parameter 30 variation is essential; simply choosing a few values of interest and generating different 31 output is of limited scientific benefit and inevitably raises questions of selection bias. 32 That said, if alternative values are commonly used and therefore recommend themselves, 33 then running these through the models can help sensitize users to parameter variations 34 and their impacts on model outputs. 53N a tu r a l la n g u a g e d e n o t e s a n y d i s c o u r s e i n w h i c h t h e m e a n i n g o f t h e w o r d s i s n o t f o r m a l i z e d ; r a t h e r , t h e s e w o r d s a re ju s t " as th e y c o m e in o ff th e stre e t" w ith w h a te v e r m e a n in g a p a rtic ip a n t m a y g iv e th e m . 54B e f o r e t h e a d v e n t o f p e r s o n a l c o m p u t e r s , v a r i o u s s h o r t h a n d t e c h n i q u e s w e r e d e v e l o p e d f o r c o m p u t i n g s y s t e m r i s k . I n c o n tr o l th e o r y , s c h e m e s o f `in te r v a l p r o b a b i lit ie s ' w e r e p r o p o s e d w h ic h c o u ld b e p r o p a g a te d th r o u g h a s y s te m to y ie ld b o u n d s o n sy ste m re lia b ility . W h e re a s th e se b o u n d s o rig in a lly re fle c te d a c c u ra c y o f sh o rth a n d a p p ro x im a tio n s o f c o m p le x fo rm u la e , th e ir o ffsp rin g h a v e b e e n p ro p o se d as q u a n tific a tio n s o f u n c e rta in ty . A lte rn a tiv e n o tio n s o f u n c e rta in ty a re a lso p ro p o se d w ith th e g o a l o f sim p lify in g th e a sse ssm e n t a n d c o m p u ta tio n a l b u rd e n o r c a p tu rin g p u ta tiv e fe a tu re s o f u n c e rta in ty w h ic h a re o v e rlo o k e d in p ro b a b ility th e o ry . T h e se in c lu d e p o ssib ility th e o ry , fu z z y n u m b e rs, q u a lita tiv e a lg e b ra , im p re c ise p ro b a b ilitie s , b e lie f fu n c tio n s , c e rta in ty fa c to rs, a n d th e lik e . N o n m o n o to n ic re a so n in g sy ste m s a tte m p t to c a p tu re re a so n in g a b o u t k n o w le d g e , o r re a so n in g fro m p a rtia l k n o w le d g e ; th e y in c lu d e d e fa u lt lo g ic , d e fe a sib le lo g ic , a b d u c tiv e lo g ic , a n d a u to e p iste m ic lo g ic , to n a m e a fe w . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-4 DRAFT--DO NOT CITE OR QUOTE 1 Cognitive Uncertainty. This concerns uncertainty regarding what is the case. Not 2 knowing "what is the case" may be conceived as uncertainty over the set of all 3 possibilities, sometimes expressed as `uncertainty over the set of possible worlds.' 4 Uncertainty over possible worlds may be represented formally as probability; that is, the 5 uncertainty of a given situation is represented as a number between zero and one, and the 6 uncertainty of either of two mutually exclusive situations is the sum of the uncertainties 7 of each situation.55 Two interpretations or operationalizations of the probability 8 formalism are current: the objective or frequentist interpretation and the subjective or 9 Bayesian interpretation. These interpretations are not mutually exclusive, as subjective 10 probabilities can and often do track relative frequencies. 11 Volitional Uncertainty. This concerns uncertainty regarding what to do. In the natural 12 language, being unsure which course of action to choose is also called "uncertainty." 13 Insofar as uncertainty on the best course of action can be translated into a claim about the 14 state of the world, volitional uncertainty can be translated into cognitive uncertainty. For 15 example, a regulatory body charged with setting a speed limit is obliged to make a 16 decision. The decision may be cautious or reckless, well or poorly motivated, wise or 17 foolish; but it cannot be true or false. Since the decision makes no claim about the state 18 of the world, it cannot be uncertain in the cognitive sense. The uncertainty cannot be 19 analyzed by sampling from some distribution. However, if the decision is based on the 20 claim that the chosen speed limit minimizes accidents while maintaining a prescribed 21 traffic volume, that claim may be uncertain and may be subjected to quantitative 22 uncertainty analysis. A discretionary decision of a regulatory body may entrain cognitive 23 uncertainty, but it becomes amenable for quantitative uncertainty analysis only when it is 24 linked to a claim about the state of the world. 25 Aleatoric/Epistemic Uncertainty. This terminology has become standard in the technical 26 uncertainty analysis literature, and it has been called Variability/Uncertainty in some 27 areas, particularly dealing with human populations. A variable whose uncertainty is 28 aleatoric for a given population takes different, uncertain, values for each member of the 29 population. If its uncertainty is epistemic, it takes the same uncertain value for all 30 members of the population. Issues involving uncertainty and variability or epistemic and 31 aleatory uncertainty translate into issues of dependence, when conducting a quantitative 32 uncertainty analysis (see Section 6.1.3.3). In its Science and Judgment report, NRC 33 (1994, 006424) correctly remarks that "the amount of variability is generally itself an 34 uncertain parameter." It is natural to ask whether a given uncertainty is aleatoric or 35 epistemic, whereas it is awkward to ask whether this uncertainty is uncertain or 36 variable--which explains the preference for the epistemic/aleatoric terminology. 37 55T h e s e a r e k n o w n c o l l e c t i v e l y a s K o l m o g o r o v ' s p r o b a b i l i l t y a x i o m s . T h e a d d i t i v i t y o f p r o b a b i l i t y f o r e x c l u s i v e a lte rn a tiv e s sta te s, e .g ., th a t th e p ro b a b ility o f a n u n s e e n o b je c t b e in g re d o r g re e n is th e s u m o f th e p ro b a b ility th a t it is re d a n d th e p ro b a b ility th a t it is g re e n . T h is o f c o u rse a ssu m e s th a t " re d " a n d " g re e n " a re c le a rly d e fin e d , su c h th a t n o th in g c a n b e sim u lta n e o u sly re d a n d g reen . M a n y a lte rn a tiv e re p re se n ta tio n s o f u n c e rta in ty c o n te s t th is a d d itiv ity p ro p e rty . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-5 DRAFT--DO NOT CITE OR QUOTE 1 6.1.3. Key Elements of a Quantitative Uncertainty Analysis 2 The uncertainty propagation can be performed by some rough estimation, as for example 3 the delta method (Oehlert, 1992, 543742), or in rare cases it can be performed analytically, as in 4 simple error propagation.56 Most often, however, it will be performed using Monte Carlo 5 simulation. A joint distribution is assigned to the parameters of a quantitative model and then 6 propagated through the model by sampling repeatedly from this joint distribution, computing 7 model output and generating a distribution of model output. Every uncertainty analysis is 8 conditional on initial assumptions. A "complete" uncertainty analysis is an unattainable goal; the 9 best that can be done in practice is to identify and motivate the assumptions that are used. This 10 section is not a how-to guide, but a to-do list of key elements of any quantitative uncertainty 11 analysis.57 12 13 6.1.3.1. Q uan titative M o d el 14 The starting point of any quantitative uncertainty analysis is a mathematical model or 15 procedure for calculating quantities of interest. A structured narrative explains the choice of 16 quantitative models. If some values of input parameters for this calculation are not known with 17 certainty, then the question arises: "What is the uncertainty attending the use of this model?" 18 This is the question a quantitative uncertainty analysis seeks to answer. 19 20 6.1.3.2. M a rg in a l D istribu tion s o ver M o d e l P a ra m eter 21 If the model parameters are directly measurable with sampling error, then the sampling 22 distribution may itself be used in the quantitative uncertainty analysis. If the model parameters 23 are fit to data that are sampled from a known or hypothesized distribution, then by resampling 24 this distribution and refitting the model, distributions over the model parameters may be 25 constructed. Physically-based simulation models, such as pharmacokinetic models or 26 environmental transport models, may be solved analytically if equilibrium reaction rates (the 56S im p l e m e a s u r e m e n t e r r o r is o f t e n r e p r e s e n t e d b y a d d i n g a n o r m a lly d i s t r i b u t e d r a n d o m v a r i a b l e w i t h m e a n z e r o to a " tru e " v a lu e . I f se v e ra l m e a su re m e n ts a re p e rfo rm e d in su c c e ssio n , a n d th e e rro rs o n e a c h m e a su re m e n t are a s s u m e d to b e in d e p e n d e n t, th e n th e e rro r in d u c e d b y a d d in g th e m e a s u re m e n t re s u lts is a lso a n o rm a lly d is trib u te d ra n d o m v a ria b le w h o se m e a n is z e ro a n d w h o se v a ria n c e is th e s u m o f th e v a ria n c e s o n th e in d iv id u a l m e asu rem en ts. 57T h e s e k e y e l e m e n t s o f q u a n t i t a t i v e u n c e r t a i n t y a n a l y s i s a r e d i s c u s s e d i n m a n y p u b l i c a t i o n s s u c h a s S a l t e l l i e t a l . (2 0 0 0 , 5 4 3 7 5 6 ), C o x (2 0 0 6 , 5 9 4 3 4 2 ), K u ro w ic k a a n d C o o k e (2 0 0 6 , 5 4 3 7 5 8 ), N R C (2 0 0 7 , 5 4 3 7 4 8 ) an d E P A (2 0 0 9 , 522927). This document is a draftfor review purposes only and does not constitute Agency policy. 6 -6 DRAFT--DO NOT CITE OR QUOTE 1 transfer coefficients) are constant. If these rates are not constant, as when concentrations are 2 near saturation levels, then simulating the pharmacokinetics or transport is indicated. Structured 3 expert judgment has been applied for uncertainty quantification within the engineering 4 community since the time of the Rasmussen Report (U.S. NRC, 1975, 543729). More recently, 5 this approach has been "test-driven" by EPA in assessing health effects of fine particulates 6 (Walker et al., 1999, 198615), and its potential application has been identified in the Agency's 7 Guidelinesfor Carcinogen Risk Assessment, commonly referred to as the Cancer Guidelines 8 (U.S. EPA, 2005, 086237).58 9 10 6.I.3.3. D ep en d en ce betw een P a ra m eter U ncertainties: A lea to ric a n d E p iste m ic (U ncertain ty 11 a n d V ariability) 12 Two uncertain quantities are independent if knowledge about one of them does not alter 13 our uncertainty regarding the other. The quantities are dependent if they are not independent. 14 The role of dependence modeling in quantitative uncertainty analysis must be addressed. To 15 illustrate, cigarette smoking and body fat are both thought to influence biomarkers for toxic 16 response to dioxin exposure, such as ethoxyresorufin-O-deethylase (EROD) activity (Pereg et al., 17 2002, 199797). In an individual sampled at random from a target population, both percent body 18 fat and whether (and how much) he or she smokes are uncertain.59 However, these uncertainties 19 are not independent, inasmuch as smokers tend to have less body fat (Vanni et al., 2009, 20 543754). 21 Issues involving uncertainty and variability, or epistemic and aleatory uncertainty, 22 translate into issues of dependence when conducting a quantitative uncertainty analysis. For 23 example, a constant used to estimate the biokinetic behavior of dioxin may be uncertain. If it is 24 believed to be the same for every member of the population, the uncertainty is termed 58T h e E P A ( 2 0 0 5 , 0 8 6 2 3 7 ) c a n c e r g u i d e l i n e s s t a t e : " I n m a n y o f t h e s e s c i e n t i f i c a n d e n g i n e e r i n g d i s c i p l i n e s , re se a rc h e rs h a v e u s e d rig o ro u s e x p e rt e lic ita tio n m e th o d s to o v e rc o m e th e la c k o f p e e r-re v ie w e d m e th o d s a n d d a ta ...." T h e se c a n c e r g u id e lin e s are fle x ib le e n o u g h to a c c o m m o d a te th e u se o f e x p e rt e lic ita tio n to c h a ra c te riz e c a n c e r risk s, as a c o m p le m e n t to th e m e th o d s p re s e n te d in th e c a n c e r g u id e lin e s. A c c o rd in g to N R C (2 0 0 2 , 0 3 5 3 1 2 ) , th e rig o ro u s u se o f e x p e rt e lic ita tio n fo r th e a n a ly se s o f risk s is c o n s id e re d to b e q u a lity sc ie n c e ." 59B e c a u s e d i o x i n s g e n e r a l l y d i s t r i b u t e t o b o d y f a t / l i p i d , t h e p e r c e n t b o d y f a t i s o f t e n u s e d t o e s t i m a t e b o d y b u r d e n ; a d e fa u lt v a lu e o f 2 5 % is c o m m o n (C o n n o r a n d A y lw a rd , 2 0 0 6 , 1 9 7 6 3 2 ) . H o w e v e r, b o d y fa t p e rc e n ta g e v a rie s w id e ly b e tw e e n in d iv id u a ls , f ro m a m in im u m e s s e n tia l le v e l (e .g ., 2 % f o r m e n , 1 0 % f o r w o m e n ) to o b e s ity (e .g ., 3 8 % o r m o re f o r m e n , 4 2 % f o r w o m e n ). C o n s id e rin g th a t c u rre n t e s tim a te s s u g g e s t 3 0 % o f th e U .S . p o p u la tio n a re o b e se , a n u n c e rta in ty a n a ly sis o f d io x in risk in th is p o p u la tio n sh o u ld sa m p le in d iv id u a ls fro m th e ir g e n d e r/b o d y fa t d is trib u tio n a n d c o rre la te th is w ith o th e r k n o w n o r s u sp e c te d c o v a ria te s in flu e n c in g to x ic re sp o n se (su c h a s d ie t, sm o k in g , n a tu ra l a n d e n d o g e n o u s lig a n d s, d ise a se , a n d a g e ). This document is a draftfo r review purposes only and does not constitute Agency policy. 6-7 DRAFT--DO NOT CITE OR QUOTE 1 "epistemic." In a quantitative uncertainty analysis, this factor would be sampled from its 2 uncertainty distribution on each Monte Carlo run and used for all members of the population. 3 Body fat, in contrast, is aleatoric. We do not sample one value from the body fat distribution and 4 use this value for all members of the population on each Monte Carlo run. Instead we sample a 5 body fat value for each individual on each run. Because body fat is correlated with other 6 relevant variables (e.g., smoking, gender, age, and socioeconomic status), all of these variables 7 should be sampled in a manner that reflects their dependences. Kinetic constants whose 8 uncertainty is epistemic are completely correlated across individuals: if the value is 0.5 for one 9 individual, it is 0.5 for everyone. Body fat values vary from individual to individual, and they 10 are correlated through a host of other variables. 11 12 6.I.3.4. M o d e l U ncertainty 13 All models, being idealizations, are false; on this there is no uncertainty to quantify. 14 However, the choice of model may constrain the ability to represent uncertainty in observable 15 phenomena. Thus, in a linear low-dose model, uncertainty over a cancer slope factor may be 16 quantified, but uncertainty regarding changes in slope at distinct low-dose regimes cannot be 17 captured. When the model choice imposes severe and potentially unwelcome constraints on 18 uncertainty quantification, this must be addressed. Distributions over model parameters may be 19 selected and evaluated based on their ability to reflect uncertainty distributions over observable 20 phenomena predicted by the models.60 In such cases, the uncertainty propagated through the 21 quantitative model is not strongly model-dependent. In other cases, multiple model alternatives 22 may be applied, whose "probability of being the true model" is known or assumed. Since 23 different models can always be regarded as specializations of more general models, the 24 distinction between parameter and model uncertainty is sometimes more apparent than real. For 25 example, as illustrated in the EPA Benchmark Dose Software (BMDS) (U.S. EPA, 2000, 26 052150), the multistage and Weibull dose-response models both contain the model Pr(x) = y + 27 (1 - y) (1 - e-p1x) as a submodel, to which they collapse if other parameters are zero (multistage) 28 or one (Weibull). Recalling that the function 1/(1 + x) is first-order equivalent to (1 - x) for 60 S u c h t e c h n i q u e s w e r e f i r s t u s e d o n a l a r g e s c a l e i n t h e U . S . N R C - E U j o i n t u n c e r t a i n t y a n a l y s i s o f c o n s e q u e n c e m o d e ls fo r a c c id e n ts a t n u c le a r p o w e r p la n ts, se e G o o sse n s e t al. (1 9 9 6 , 5 4 8 7 2 7 ; 2 0 0 1, 5 4 8 7 3 7 ; 2 0 0 1 , 5 4 8 7 3 8 ; 2 0 0 1 , 5 4 8 7 3 1 ; 2 0 0 1 , 5 4 8 7 3 2 ; 2 0 0 1 , 5 4 8 7 3 5 ) (B o c k e t a l., 1 9 9 8 , 5 4 8 7 5 2 ) . This document is a draftfo r review purposes only and does not constitute Agency policy. 6 -8 DRAFT--DO NOT CITE OR QUOTE 1 small x, the same may be said for logistic models as well. In this case, these models could easily 2 be parameterized within one family, rendering the choice between them a choice of parameter 3 values. Similarly, the choice between sub-, supra-, and linear models is sometimes reduced to 4 parameter estimation within a more general class of model (Hoel and Portier, 1994, 198741). 5 In other cases, the reduction of model uncertainty to parameter uncertainty is less natural. 6 For example, according to the "chemoprotection model" of Greenlee et al. (2001, 015400), 7 dioxin's binding to the aryl hydrocarbon receptor (AhR) inhibits proliferation in tumor cells and 8 thus suppresses mammary tumors. Dose-dependent protection and cancer induction can both be 9 true, each applying to different tissues. Although mathematical models exhibiting these twin 10 features have been suggested (e.g., Kohn and Melnick, 2002, 199104), these models are not yet 11 readily estimable from data, and the choice between them is referred to the structured narrative. 12 13 6.I.3.5. S am plin g M eth o d 14 All sampling on a computer is "pseudo random." Significant issues arise in choosing a 15 method for sampling high-dimensional distributions with dependence. If evaluating the 16 quantitative model is very time consuming, various "quasi random" schemes may be applied, 17 including Latin hypercube sampling, importance sampling, and Hammersley sampling. These 18 methods involve trade-offs between economy and accuracy of the dependence modeling. 19 20 6.I.3.6. M e th o d f o r E x tra ctin g a n d C om m u n icatin g R esu lts 21 When a large quantitative uncertainty analysis has been performed, the method for 22 identifying important contributors and communicating this information to users is not 23 straightforward. Analysts have proposed many ways to quantify the uncertainty contribution of 24 one variable, or set of variables, on others,61 and the analyst's decision at this juncture may 25 strongly impact the "take-home" message from the study. An importance measure that averages 61A f e w e x a m p l e s m a y s u f f i c e . T h e s t a n d a r d P e a r s o n c o r r e l a t i o n c o e f f i c i e n t m e a s u r e s t h e l i n e a r d e p e n d e n c e b e tw e e n tw o v a ria b le s, p o sitiv e o r n e g a tiv e . T h e ra n k o r S p e a rm a n c o rre la tio n c o e ffic ie n t m e a su re s th e m o n o to n e d e p e n d e n c e . T h e c o rre la tio n ra tio m e a su re s th e (u n sig n e d ) v a ria n c e c o n trib u tio n o f a n e x p la n a to ry v a ria b le o n a ta rg e t v a ria b le . T h e re g re ssio n c o e ffic ie n t m e a su re s th e e x p e c te d c h a n g e in sta n d a rd (n o t n a tu ra l!) u n its o f a ta rg e t v a ria b le , p e r sta n d a rd u n it c h a n g e in a n e x p la n a to ry v a ria b le , a n d a ssu m e s th is e x p e c te d c h a n g e is in d e p e n d e n t o f th e v a lu e s o f th e e x p la n a to ry v a ria b le s. M u ltip le c o rre la tio n m e a su re s th e c o rre la tio n b e tw e e n a g iv e n v a ria b le a n d its b e s t lin e a r p re d ic to r b a s e d o n a n o th e r se t o f v a ria b le s. T h e p a rtia l c o rre la tio n o f tw o v a ria b le s g iv e n a se t o f o th e r v a ria b le s is th e ir c o rre la tio n a fte r d is c o u n tin g th e in flu e n c e o f th e o th e r v a ria b le s . T h e c o rre la tio n ra tio , m u ltip le c o rre la tio n , a n d th e re g re s s io n c o e ffic ie n t a re n o t sy m m e tric ; th e c o rre la tio n ra tio a n d m u ltip le c o rre la tio n a re a lw a y s n o n -n e g a tiv e (K u ro w ic k a a n d C o o k e , 2 0 0 6 , 5 4 3 7 5 8 ; S a lte lli e t a l., 2 0 0 0 , 5 4 3 7 5 6 ) . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-9 DRAFT--DO NOT CITE OR QUOTE 1 over an entire sample space may obscure the features of real interest. For example, the drivers of 2 cancer induction at low doses might be different from the drivers at high doses. If the drivers of 3 low-dose cancer induction are of interest, then importance measures that average over all doses 4 should not be considered. 5 6 6.2. EPA APPROACHES FO R ORAL CANCER AND NONCANCER ASSESSMENT 7 Different types of toxicity information have historically been used in EPA's oral cancer 8 and noncancer dose-response assessments, although efforts to harmonize these approaches are 9 ongoing. For oral exposures, noncancer endpoints are commonly assessed using the RfD 10 methodology to derive "an estimate (with uncertainty spanning perhaps an order of magnitude) of 11 a daily oral exposure to the human population (including sensitive subgroups) that is likely to be 12 without an appreciable risk of deleterious effects during a lifetime." In contrast, cancer 13 endpoints are commonly assessed using a dose-response function with the probability of excess 14 risk above background modeled as a linear function of dose, for doses down to zero. The RfD 15 method relies on a POD. The cancer dose-response method uses a POD if the linear model is 16 chosen. From the Cancer Guidelines, cancer endpoints can also be assessed using the RfD 17 methodology if the proof burden is satisfactorily met (as described in Section 5.2.3.4.1.2). 18 Toxicity reference values have typically been derived for human noncancer endpoints 19 based on a no-observed-adverse-effect level (NOAEL) or lowest-observed-adverse-effect level 20 (LOAEL) from animal bioassay studies. This terminology suggests a biological population 21 threshold beneath which no harm is anticipated. Reference values such as the oral RfD or 22 inhalation reference concentration are derived by applying uncertainty factors (UFs) to a POD. 23 Depending on the nature of available data and modeling choice, a POD can be selected from 24 values other than a NOAEL or LOAEL, such as an EDx (effective dose eliciting x percent 25 response), or a benchmark dose (BMD) or its lower confidence bound (BMDL). The BMD is 26 the dose that induces a benchmark response (BMR), which is often chosen to represent a 5 or 27 10% increase in excess risk above background. The POD is divided by one or more uncertainty 28 factors that represent knowledge gaps (see Section 6.4.1.2 for details on specific types of UFs). 29 An RfD is described as "likely to be without appreciable risk" but the probabilistic 30 language has not as yet been operationalized. A quantitative definition of "appreciable" has not 31 been articulated, and methods to compute risks above the RfD as a function of dose have not This document is a draftfo r review purposes only and does not constitute Agency policy. 6-10 DRAFT--DO NOT CITE OR QUOTE 1 been designated for use by the EPA; thus, it is not current practice to ascertain that the risk is 2 indeed not appreciable. In addition, different participants in discussions over 3 threshold/nonthreshold models for dioxin may have different perspectives regarding how to 4 define "appreciable risk." Under the current POD/UF framework, dose-response functions are 5 not provided for calculating the actual risk at or above the RfD. Instead, to provide a "risk 6 indicator" for use in screening for health hazards, a hazard quotient (HQ) is computed as the 7 ratio of a given oral exposure to the RfD, or a margin of exposure (MOE) is estimated as the 8 ratio of the POD to the human exposure level. 9 For the cancer endpoint, an oral cancer slope factor may be derived for human health risk 10 assessment, typically based on tumor incidence data from an animal bioassay or on cancer 11 incidence or deaths from an epidemiologic study. In the EPA Cancer Guidelines, cancer is 12 predominantly thought to have no population biological threshold and a linear extrapolation to 13 zero is applied from the POD based on extra risk above background, i.e., the probability of the 14 endpoint decreases linearly in dose from the POD to zero or to a population background level. In 15 the absence of sufficient information on the cancer mode of action (MOA), the linear model is 16 applied as a default. The linear model also can be applied when there is sufficient MOA 17 evidence supporting this choice for low-dose cancer induction. Cancer endpoints could also be 18 evaluated using a "nonlinear" model. In this case, the proof burden clearly rests on the nonlinear 19 model; there must be sufficient evidence to override the health-protective default or 20 scientifically-based choice of a linear model, as described in the Cancer Guidelines. These 21 Guidelines state, "When adequate data on mode of action provide sufficient evidence to support 22 a nonlinear mode of actionfor the generalpopulation (emphasis added) and/or any 23 subpopulations of concern, a different approach--a reference dose/reference concentration that 24 assumes that nonlinearity--is used." In current terminology, the RfD methodology applies to the 25 cancer endpoint if there is sufficient evidence supporting a "zero slope at zero" model; 26 otherwise, the linear nonthreshold model applies by default. (See Section 5.2.3 for a detailed 27 discussion of linear vs. nonlinear extrapolations below the observed data, population vs. 28 individual thresholds, and how the Cancer Guidelines are applied in choosing dose-response 29 model forms for risk assessment.) 30 This document is a draftfor review purposes only and does not constitute Agency policy. 6-11 DRAFT--DO NOT CITE OR QUOTE 1 6.3. HIGHLIGHTS OF NAS REVIEW COMMENTS ON UNCERTAINTY 2 QUANTIFICATION FOR THE 2003 REASSESSMENT 3 The N A S (2006, 198441; 2006, 543760) identified a number o f uncertainty 4 characterization issues for the 2003 Reassessment; key sources of uncertainty for w hich 5 quantification is suggested are highlighted in Table 6-1. The discussion in this section focuses 6 on comments related to dose response. 7 There are several nuances in the N A S position relative to the need for substantial 8 improvement in transparency, thoroughness, and clarity in quantitative uncertainty analysis for 9 the 2003 Reassessment. These nuances concern whether the nonlinear model (note that the N A S 10 committee uses " sublinear" and "nonlinear" interchangeably) is scien tifically better supported 11 than the linear model, and i f the sublinear model is better supported, whether this is based on 12 data or on apodictic knowledge (know ledge without uncertainty) o f the M O A . The N A S 13 committee does not distinguish between individual and population dose-response models; 14 how ever the criteria from the E P A C an ce r G u idelin es clearly apply to population models. 15 A ssu m in g that the A hR-m ediated M O A im p lies a threshold for each individual, the step to a 16 population "zero slope at zero" model requires the follow ing, as identified and discussed in detail 17 in Section 5.2.3.: 18 19 1. Th e distribution o f the individual thresholds induced by the M O A , and 20 2. The dose-response function for values above the thresholds. 21 22 T h is information can either come from data or from known information o f the M O A , but 23 the burden o f proof clearly rests on the nonlinear model. T h is section sum m arizes the N A S 24 committee's overall positions. Responses to specific suggestions are given in Section 6.4 and 25 summarized in Section 6.5. Several excerpts o f specific comments from N A S (2006, 198441) 26 illustrate key issues. 27 The N A S committee favors the nonlinear model with a threshold: 28 This document is a draftfor review purposes only and does not constitute Agency policy. 6-12 DRAFT--DO NOT CITE OR QUOTE 1 .. .the committee concludes that, although it is not possible to scientifically prove 2 the absence of linearity at low doses, the scientific evidence, based largely on 3 mode of action, is adequate to favor the use of a nonlinear model that would 4 include a threshold response over the use of the default linear assumption. 5 (p. 122) 6 7 The committee does not state whether the threshold applies to the population, or whether each 8 individual has his/her own threshold. 9 The NAS also comments on whether the nonlinear model should be used to compare with 10 the linear default: 11 12 Because the committee concludes that the data support the hypothesis that the 13 dose-response relationship for dioxin and cancer is sublinear, it recommends that 14 EPA include a nonlinear model for cancer risk estimates but also use the current 15 linear models for comparative purposes. (p. 16) 16 17 The committee does not suggest what the (population/individual) threshold might be, nor how it 18 might be supported on the basis of data. Rather, the apodictic knowledge that there is a 19 (population/individual) threshold places the dioxin risk assessment within the RfD framework, 20 using a HQ or MOE as the basis for indicating the potential risks from exposure. The committee 21 further asks for a quantitative characterization of the range of uncertainty: 22 23 The committee determined that the available data support the use of a nonlinear 24 model, which is consistent with receptor-mediated responses and a potential 25 threshold, with subsequent calculations and interpretation of MOEs. EPA's sole 26 use of the default assumption of linearity and selection of ED01 as the only POD 27 to quantify cancer risk does not provide an adequate quantitative characterization 28 of the overall range of uncertainties associated with the final estimates of cancer 29 risk. (p. 24) 30 31 Regarding the Cancer Guidelines' requirement of sufficient evidence to use a nonlinear 32 approach for cancer risk assessment, the committee indicates that quantitative evidence will not 33 decide the linearity/nonlinearity (nonthreshold/threshold) issue, but knowledge (without 34 uncertainty) of the MOA should be used: 35 This document is a draftfor review purposes only and does not constitute Agency policy. 6-13 DRAFT--DO NOT CITE OR QUOTE 1 Quantitative evidence of nonlinearity below the point of departure (POD), the 2 ED0i62 will never be available because the POD is chosen to be at the bottom end 3 of the available dose-response data. ... EPA should give greater weight to 4 knowledge about the mode of action and its impact on the shape of the 5 dose-response relationship. (p. 178) 6 7 The comment continues, with the committee implicitly acknowledging that there is no 8 evidence arguing against linearity, but that the lack of evidence should not justify using the linear 9 model. 10 The committee considers that the absence of evidence that argues against linearity 11 is not sufficient justification for adopting linear extrapolation, even over a dose 12 range of one to two orders of magnitude or to the assumption of linearity through 13 zero, which would not normally be applied to receptor-mediated effects. (p. 178) 14 15 In addition, the committee recommended that EPA explore both linear and nonlinear 16 approaches to TCDD cancer assessment: 17 18 On the whole, the committee concluded that the empirical evidence supports a 19 nonlinear dose response below the ED01, while acknowledging that the possibility 20 of a linear response cannot be completely ruled out. The Reassessment 21 emphasizes the lack of such nonlinear models, hence its adoption of the approach 22 of linear extrapolation below the POD level. Although this approach remains 23 consistent with the cancer guidelines...., EPA should acknowledge the qualitative 24 evidence of a nonlinear dose response in a more balanced way, continue to fill in 25 the quantitative data gaps, and look for opportunities to incorporate mechanistic 26 information as it becomes available. The committee recommends adopting both 27 linear and nonlinear methods of risk characterization to account for the 28 uncertainty of dose-response relationship shape below ED01 (p. 72). 29 30 In this document, EPA has applied its own guidance on cancer risk assessment and 31 adopted linearity (and an assumption of no threshold) as a health-protective default approach in 32 the absence of sufficient evidence of MOA involving a threshold for all tumors resulting from 33 TCDD exposures (volitional uncertainty). (Note that the NAS report appears to view the 34 absence of evidence as imposing a burden of proof on the linear model [cognitive uncertainty]; 35 see Sections 5.2.3.4.1.2 and 6.2 regarding the burden of proof.) In addition, the NAS 36 committee's request to apply nonlinear methods for the cancer assessment is addressed, in 62 E e f f e c t i v e d o s e ( E D ) i s t h e d o s e c o r r e s p o n d i n g t o a X % i n c r e a s e ( i n t h i s c a s e a 1 % ) i n a n a d v e r s e e f f e c t s u c h a s a c o n c e r e n d p o in t, re la tiv e to th e c o n tro l resp o n se. This document is a draftfo r review purposes only and does not constitute Agency policy. 6-14 DRAFT--DO NOT CITE OR QUOTE 1 Section 5.2.3.4.1.4 of this document. That evaluation describes the application of nonlinear 2 methods to TCDD data and presents two illustrative examples of RfD development for 3 carcinogenic effects: one based on tumorigenesis in experimental animals, and the other on 4 hypothesized key events in TCDD's MOAs for liver and lung tumors. 5 The thrust of the NAS remarks regarding transparency, thoroughness and clarity in 6 quantitative uncertainty analysis relevant to dose-response can be summarized as follows: 7 8 1. The uncertainty of cancer risks due to dioxin exposure should be quantified. 9 2. Dioxin cancer risk should be treated either as a threshold phenomenon, thus following the 10 basic RfD methodology, or should be modeled using a sublinear dose-response function 11 below the observed data, with the linear model used for comparison. 12 3. The POD should be subjected to quantitative uncertainty analysis. 13 A similar point of view has been indicated by others.63 Detailed suggestions regarding specific 14 improvements for quantitative uncertainty analysis in the 2003 Reassessment are outlined in the 15 next section and summarized in Section 6.5. 16 17 6.4. FEASIBILITY OF CONDUCTING A QUANTITATIVE UNCERTAINTY 18 ANALYSIS FO R TCDD 19 This section focuses on uncertainty analysis for TCDD dose response, which involves a 20 range of issues as highlighted in Table 6-1. 21 22 6.4.1. Feasibility of Conducting a Quantitative Uncertainty Analysis under the RfD 23 Methodology 24 This discussion applies to all noncancer endpoints of TCDD, and to cancer endpoints 25 insofar as they fall under the RfD methodology. An RfD is obtained through the following steps: 26 27 1. Choose a POD, then 28 2. Apply uncertainty factors (UFs) to account for knowledge shortfalls. 29 63F o r e x a m p l e , f r o m P o p p e t a l . ( 2 0 0 6 , 1 9 7 0 7 4 ) . " O v e r a l l , t h e e v i d e n c e i n d i c a t e s t h a t ( 1 ) T C D D c a u s e s c a n c e r v i a a re c e p to r-m e d ia te d p ro c e s s ; (2 ) th is d o s e -re s p o n s e is n o n -lin e a r; a n d (3 ) a th re sh o ld re g io n e x ists fo r T C D D -in d u c e d c a n c e r b e lo w w h ic h a d v e rse e ffe c ts a re u n lik e ly to o c c u r." This document is a draftfo r review purposes only and does not constitute Agency policy. 6-15 DRAFT--DO NOT CITE OR QUOTE 1 The method of uncertainty factors harkens back to the engineering practice of safety 2 factors (Lehman and Fitzhugh, 1954, 003195). To illustrate, if the reference load for an 3 engineered structure is X, then engineers might design the structure to withstand load 3X, using a 4 safety factor of 3 to create a margin of safety. If the structure functions in a corrosive 5 environment, another factor could be multiplied to guarantee safety for that condition, and 6 another factor could be applied for heat, another for vibrations, and so on. The choice of values 7 is simply based on good engineering practice, i.e., reflecting what has worked in the past. 8 Although safety factors are still common in engineering, they are giving way to probabilistic 9 design in many applications. The reason is that compounding safety factors quickly leads to 10 overdesigning. Compounding safety margins for spaceflight systems may render them too heavy 11 to fly. As our understanding of a system increases, it becomes possible to guarantee the requisite 12 safety by leveraging our scientific understanding of the materials and processes. That of course 13 requires formulating clear probabilistic safety goals and developing the techniques to 14 demonstrate compliance. 15 The engineering community has never sought to account for uncertainty by treating 16 safety factors as random variables and assigning them (marginal) distributions; such an approach 17 would not counteract the overdesigning inherent in safety factors. Many authors, including the 18 recent national committee for Science and Decisions (NRC, 2009, 194810), have advocated just 19 such a probabilistic approach to the apparent "overdesigning" of the RfD when multiple UFs are 20 used in its derivation. 21 The NAS committee that evaluated the 2003 Reassessment does not discuss how to 22 perform uncertainty analysis. But their call for substantial improvement in quantitative 23 uncertainty analysis with TCDD falling under the RfD framework entails examining the 24 feasibility of quantitative uncertainty analysis within this framework. (Note that the EPA 25 Integrated Risk Information System (IRIS) database uses uncertainty factors without 26 probabilistic interpretations; some context for this is offered in Section 6.4.1.2.) 27 28 6.4.1.1. F easibility o f C on du ctin g a Q u an titative U ncertainty A n a lysis f o r th e P o in t o f 29 D epartu re 30 The POD plays a role in both the noncancer RfD methodology and the cancer 31 dose-response methodology. The POD can be selected from various options, such as a NOAEL This document is a draftfor review purposes only and does not constitute Agency policy. 6-16 DRAFT--DO NOT CITE OR QUOTE 1 or LOAEL, a BMDL, or an EDx. The feasibility of quantitative uncertainty analysis for each of 2 these three options is considered below. 3 By definition, the NOAEL is the highest of the tested doses in a toxicological experiment 4 that is judged not to have caused an adverse effect (with dose expressed as a dose rate, in 5 mg/kg-day). A quantitative uncertainty analysis for a NOAEL or LOAEL encounters the 6 following problem. In an experiment involving a small, positive response, the probability of 7 seeing no response can be calculated using a binomial model with the number of exposed 8 animals and the observed number of responses. However, in an experiment with no response, 9 the probability o f having observed a response cannot be calculated without assuming a response 10 probability. Such an assumption could not be based on observed data. The probability of a 11 higher NOAEL or higher LOAEL can be computed, but not that of a lower NOAEL or LOAEL. 12 In other words, the probability that an experiment with a positive result may have yielded a null 13 response can be estimated, but not the probability that an experiment with a null response might 14 have yielded a positive response.64 15 In addressing uncertainty quantification for a BMDL or EDx, two questions must be 16 distinguished regarding the response: 17 18 1. What is the distribution of possible doses that causes an x% increase over background? 19 2. What is the distribution for possible values of increase over background caused by a 20 given dose? 21 22 The BMD is defined as the dose that realizes a BMR. It is an estimate from bioassay data 23 that requires choosing a BMR and fitting a dose-response curve. The BMR, being a choice, is 24 not amenable to quantitative uncertainty analysis, but the choice can be motivated in a structured 25 narrative. The BMDL is the lower confidence limit on the dose that realizes a BMR (e.g., 95%) 26 that can be found based on the uncertainty in the parameters of the dose-response relationship. 27 Thus, the BMDL is addressed to the first question above, and represents in this case the 28 95% lower confidence band of the distribution of possible doses causing an x% increase over 29 background. In the standard approach, the uncertainty captured by the BMDL is sampling 64T h e p r o b a b i l i t y a s s o c i a t e d w i t h a n u l l r e s p o n s e i s o f t e n e s t i m a t e d b y f i t t i n g a d o s e - r e s p o n s e m o d e l t o t h e d a t a . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-17 DRAFT--DO NOT CITE OR QUOTE 1 uncertainty conditional on the truth of the dose-response model. Different models might fit the 2 data equally well yet lead to different BMDLs. 3 The BMDL is also influenced by the constraints imposed on the parameter fitting. 4 Suppose that the slope is expected to be greater than one, and that the maximum likelihood 5 estimate of the slope is slightly greater than one. Since the constraint is not binding, the 6 constrained and unconstrained model would have the same Akaike Information Criterion and 7 would be equivalent in this sense. However, computing the BMDL with the slope constraint can 8 lead to very different values than without this constraint. In the latter case, slope values less than 9 one contribute to the uncertainty in the dose causing the BMR (see Cooke, 2009, 543763). 10 The EDx can also be taken as a POD. It is similar in spirit to the BMD; however, as used 11 here, the term EDxapplies when the dose causing an x% extra risk over background has actually 12 been observed, not estimated from a fitted dose-response model.65 The observations are subject 13 to sample fluctuations, and if the experiment on which the EDx is based were repeated, different 14 values might be found. It is helpful to consider a numerical example. Suppose a background 15 response rate of 10% is established based on many observations of nonexposed individuals. In a 16 given experiment, involving say 100 individuals given dose d, 14 individuals responded. The 17 percent increase x over background (extra risk) is found by solving: 18 19 14/100 = 10/100 + x x 90/100, or x = 4.4%. 20 21 We conclude that d = ED44. We may assume that if the experiment were repeated with 100 new 22 individuals sampled independently from the whole population, the response would be given by a 23 binomial distribution with parameters (14, 100). The number of responses might be greater or 24 smaller than four, there is a 16% chance of observing 10 or fewer responses. The response to 25 dose d would not be distinguished from the background in that case, and a higher dose would be 26 used for the POD. 27 The uncertainty analysis of EDx as the POD involves addressing the second question 28 above, without a quantitative dose-response model. A quantitative uncertainty analysis is 29 hampered, however, by the possibility that dose d would produce a response less than or equal to 65T h i s d e f i n i t i o n o f E D x i s a d o p t e d t o d i s t i n g u i s h t h e m o d e l e d r e s p o n s e ( B M D ) a n d t h e o b s e r v e d r e s p o n s e ( E D x) , a n d it is m o re re stric tiv e th a n u s a g e s c o m m o n in th e lite ra tu re . This document is a draftfo r review purposes only and does not constitute Agency policy. 6-18 DRAFT--DO NOT CITE OR QUOTE 1 the background, in which case the POD is indeterminate--another experiment with a different 2 dose would be chosen as the POD. A true quantitative uncertainty analysis of EDx as the POD 3 would thus require a full bioassay experimental design, with binomial sampling of response rates 4 at each dose level in the assay. Absent that, quantitative uncertainty analysis is not possible. 5 The interplay of choice and estimation ingredients in the POD depends on the type of 6 POD. The main features of the above discussion are captured in Table 6-2. This table notes that 7 the BMDL captures the uncertainty caused by sampling fluctuations given that the dose-response 8 model is true. Other methods are available to compute the BMDL using (1) model-independent, 9 observable uncertainty; (2) nonparametric Bayesian dose-response models; or (3) Bayesian 10 model averaging (Cooke, 2009, 543763). Only the EDx can be subject to a quantitative 11 uncertainty analysis, and then only if a full bioassay data set is available. 12 13 6.4.I.2. F easibility o f C on du ctin g a Q u an titative U ncertainty A n a lysis w ith U ncertainty 14 F actors 15 Uncertainty factors are chosen based on a structured narrative characterizing knowledge 16 shortfalls involving the following issues: 17 18 1. Interspecies extrapolation (UFA: from animal data to humans). 19 2. Intraspecies extrapolation (UFH: to account for human interindividual variability, 20 considering sensitive subgroups). 21 3. LOAEL to NOAEL extrapolation (UFL: to estimate the dose corresponding to no adverse 22 effect, from a reported LOAEL). 23 4. Subchronic to chronic extrapolation (UFS: to estimate effects of chronic exposures, from 24 a subchronic study). 25 5. Database deficiency (UFD: to extrapolate from an incomplete data set, e.g., in terms of 26 endpoints assessed or study design, i.e., from a poor to a sufficient or rich data context). 27 28 The standard chronic RfD can represent a sensitive human (H) response to a toxic 29 substance under chronic (C) exposure conditions. Suppose a BMDL POD were based on animal 30 (A) data from a subchronic (S) study. The database for that chemical could be rich (R), e.g., 31 involving multiple (and at least sensitive) species/strains, both sexes, multiple life stages, with 32 multiple endpoints observed under sound study designs. Or the data could be poor (P), with 33 limited measurements from only a subchronic animal study (ASP) forming the basis for This document is a draftfor review purposes only and does not constitute Agency policy. 6-19 DRAFT--DO NOT CITE OR QUOTE 1 estimating a general reference value for humans (including sensitive subgroups) under chronic 2 exposure conditions. For that case, the UF method would be applied as follows: 3 4 RfD ________a s p _______ UFa XUFS XUFd XUFh 5 (Eq. 6-1) 6 w here U F A, U F S, U F D, and U F Hare the uncertainty factors for extrapolating from anim als to 7 hum ans ( U F A), subchronic to chronic exposure conditions ( U F S), without adequate endpoint 8 coverage ( U F D), and considering sensitive human subpopulations ( U F H). It is possible to assign 9 distributions to the U F s in E q . 6-1, and to perform a M onte C a rlo analysis to produce a 10 quantitative uncertainty distribution over the dose or value lik ely to be without appreciable risk 11 to hum ans for chronic exposures. M an y authors have proposed such an approach,66 and the 12 recent N R C (2009, 194810) report on science and decisions encourages E P A to m ove in this 13 direction. 14 Th e idea o f using a M onte C a rlo analysis to develop quantitative uncertainty distributions 15 for the R f D is sim ple, but the data on w h ich the U F s are based and the assum ptions that w ould 16 need to be made should be further explored. F o r exam ple, it is assum ed that the extrapolation 17 from subchronic to chronic exposure ( U F S) is the same whether applied to anim als or humans, 18 and whether applied to sufficient (rich) or deficient (poor) data contexts. Swartout et al. (1998, 19 093460) noted "W ithin the current R f D methodology, U F S does not consider differences among 20 species, endpoints, or severity o f effects; the same factor is applied in all cases." In addition, due 21 to the paucity o f relevant human data, the same authors suggested the use o f other endpoints as 22 surrogates in estimating the extrapolation from anim als to hum ans, U F A. Further, few data exist 66There has been considerable work on giving a probabilistic interpretation of the UFs, including by Abdel-Rahman and Kadry (1995), Vermeire et al. (1999), Baird et al. (1996), Swartout et al. (1998, 093460), Slob and Pieters (1998, 087256), Evans and Baird (1998), Calabrese and Gilbert (1993), Calabrese and Baldwin (1995), Hattis et al. (2002, 548720), Kang et al. (2000, 548722), and Pekelis et al. (2003, 548723). These evaluations can be considered to frame what might be called a random chem ical approach. Several authors adduce properties based on log normal distributions. Insightful studies by Kodell and Gaylor (1999;)(Gaylor and Kodell, 2000, 548724) suggest that uncertainty factors are independent log normal variables. Combining uncertainty factors involves multiplying the median values, and combining the "error factors" according to the formula KSxH= exp[1.6449 x where aS2, &h are the variances of ln(UFS) and ln(UFH). Thus UFSx UFHis a lognormal variable with Median(UFS x UFh) = Median(UFS) x Median(UFH), and 95thpercentile given by Median(UFSx UFH) x KSXH. If USand UH each have an error factor or 10, then the error factor of UFSx UFHis not 100 but 25.95. Several authors suggest that multiplying uncertainty factors might over-protect. Recent proposals from the National Research Council reflect the random chemical concept, and they inherit its problems (NRC, 2009, 194810). This document is a draftfo r review purposes only and does not constitute Agency policy. 6-20 DRAFT--DO NOT CITE OR QUOTE 1 in humans to accurately portray the interindividual variability represented by UFH. Much of the 2 data gathered to date on distributions of UFs have aggregated across other extrapolations; that is, 3 data from subchronic to chronic ratios are aggregated over different species and different data 4 contexts. Finally, it may be noted that an important issue is the data on which empirical 5 distributions of response ratios are based. Brand et al. (1999, 007629; 2001, 543765) studied the 6 sampling behavior of response ratios and raised concerns with regard to their informativeness. 7 Detailed analyses of the data underlying a Monte Carlo uncertainty analysis of Eq. 6-1 8 would afford the possibility of verifying at least some of the assumptions and numerical 9 estimations such an analysis must make. Even if the assumption that the same UFSis applicable 10 for all species, endpoints, and effect severities is thought to be biological plausible, the question 11 of whether a given set of chemicals reflects this assumption, and hence they are suitable for a 12 Monte Carlo analysis of Eq. 6-1, can only be decided by data evaluation. Data are the ultimate 13 arbiter of whether quantitative uncertainty analysis with uncertainty factors, as currently 14 envisioned, has sufficient evidentiary support. 15 16 6.4.I.3. U ncertainty R edu ction U sing Q u an titative D a ta f o r S p ecies E xtrapolation 17 Expressing dose in units of exposure that are more closely related to target tissue than to 18 contact with administered feed (or an environmental medium) can reduce uncertainty in 19 extrapolations of dose, route or species. This concept underlies EPA's establishment of the 20 Inhalation Reference Concentration Methodology (U.S. EPA, 1994, 006488). Under this 21 method, species differences in tissue exposure for inhalation toxicants serve as the basis for 22 interspecies adjustments of dose. Likewise, the International Programme on Chemical Safety 23 (IPCS) has established guidance for chemical-specific adjustment factors (IPCS, 2005), which 24 also uses a measure of internal exposure (dose) to normalize (e.g., make equivalent) the dose 25 between species. Certain more recent IRIS values also reflect such an approach, with 26 data-derived extrapolation factors replacing default adjustments. Under such approaches, the 27 relationship between external exposure and target tissue exposure is determined in each species, 28 and the applied doses are normalized on the basis of the same level of the internal tissue 29 exposure. One distinction between the two approaches is that the IPCS (2005) approach is 30 based on the attainment of the same levels of the toxicant in the blood (the central compartment) 31 rather than in the actual target tissue (a consideration based in part on the fact that typically the This document is a draftfo r review purposes only and does not constitute Agency policy. 6-21 DRAFT--DO NOT CITE OR QUOTE 1 only data available to evaluate a human toxicokinetic model will be venous blood 2 concentrations, rather than concentrations in a responding tissue or organ). Further, it has been 3 shown that species differences in internal dosimetry are more a function of species differences 4 in blood solubility than differences in tissue solubility--that is, once distributed to blood, 5 species differences in tissue exposure are less likely to be based on species differences in tissue 6 solubility. 7 The approach to development of interspecies extrapolation factors for inter- and 8 intraspecies extrapolation of effective dose for the oral RfD for dioxin, which is described in 9 Sections 3 and 4 of this document, is in agreement with both of these approaches. All tissues in 10 the body are exposed to dioxin via the bloodstream. Even in instances where the specific target 11 tissues for observed effects may be other than the tissue where the effect is observed (e.g., 12 effects mediated through the endocrine system), this biologically-based approach remains valid 13 and reduces uncertainty in dose extrapolation. The approach to extrapolation of dosimetry--on 14 the basis of circulating levels of dioxin in blood--makes optimal use of human 15 exposure-response data, human biomonitoring data, and toxicokinetic modeling to estimate 16 equivalent exposures for humans and test species without requiring that the target tissue be 17 conclusively identified. The decision to base animal-to-human extrapolation on circulating 18 levels of dioxin in blood, as predicted by a well-evaluated PBPK model, reduces some potential 19 sources of uncertainty. 20 21 6.4.I.4. C onclusion on F easibility o f Q u an titative U ncertainty A n a lysis w ith th e R fD 22 A p p ro a ch 23 A quantitative uncertainty analysis of the POD is not feasible for PODs based on 24 NOAELs or LOAELs. For the BMDL, such an analysis is not appropriate because the BMDL is 25 already a quantile from an uncertainty distribution of the BMD. However, this uncertainty 26 distribution can be obtained in different ways that capture different aspects of uncertainty. 27 Quantitative uncertainty analysis is feasible if the POD is based on the EDx (as defined above) 28 and is supported by a full set of bioassay data. A quantitative uncertainty analysis based on a 29 probabilistic interpretation of uncertainty factors in their present form invokes strong 30 assumptions. The data on which the distributions of uncertainty factors are based could be used 31 to check at least some of these assumptions. This document is a draftfor review purposes only and does not constitute Agency policy. 6-22 DRAFT--DO NOT CITE OR QUOTE 1 6.4.2. Feasibility of Conducting a Quantitative Uncertainty Analysis for TCDD under the 2 Dose-Response Methodology 3 Quantitative uncertainty analysis starts w ith a mathematical model and seeks to quantify 4 the uncertainty attending the use o f this model. Dose-response relations are mathematical 5 models expressing the probability o f response as a mathematical function o f dose. F o r several 6 decades, the uncertainty attending the use o f dose-response models has been an abiding concern 7 in many sectors, including the chem ical and nuclear industries as w ell as the public health sector. 8 G iv e n a set o f anim al bioassay data, quantifying dose-response uncertainty m ay be approached in 9 different w ays. The differences reflect different types o f uncertainty that are captured. A recent 10 evaluation enumerates the follow ing possible m ethodologies (B ussard et al., 2009, 543770) : 11 12 Benchm ark Dose Modeling (BMD): Choose the `best' model, and assess 13 uncertainty assum ing this model is true. Supplemental results can compare 14 estimates obtained w ith different m odels, and sensitivity analyses can investigate 15 other m odeling issues. 16 Probabilistic Inversion with Isotonic Regression (PI-IR ): D efine 17 m odel-independent ` observational' uncertainty, and look for a model that captures 18 this uncertainty by assum ing the selected model is true and providing for a 19 distribution over its parameters. 20 Non-Parametric Bayes (NPB): Choose a prior mean response (potency) 21 curve (potentially a "non-inform ative prior") and a precision parameter to express 22 prior uncertainty over all increasing dose-response relations, and update this prior 23 distribution w ith the bioassay data. 24 Bayesian Model Averaging (BMA) (as considered here): Choose an 25 initial set o f models, and then estimate the parameters o f each model with 26 m axim um likelihood. U se classical methods to estimate parameter uncertainty, 27 given the truth o f the model. Determ ine a probability weight for each model 28 using the B ayes Information Criterion, and use these weights to average the model 29 results. 30 31 Th e first o f the above methods in volves standard classical statistical methods and 32 captures sam pling uncertainty conditional on the truth o f the m odel used. Th e other methods are 33 " exotic" in the sense that they attempt to capture uncertainty that is not conditional on the truth 34 o f a given model. A ll have been subjected to peer review and published, but they do not enjoy 35 the w ide usage o f the standard classical methods. Th e B a y esia n models in volve subjective 36 choices o f prior distributions. Insofar as the final result is largely independent o f the choice of This document is a draftfor review purposes only and does not constitute Agency policy. 6-23 DRAFT--DO NOT CITE OR QUOTE 1 prior, these methods conform to the current starting point o f focusing on data-driven methods 2 and not appealing to structured expert judgment. (Structured expert judgment can also be 3 considered an exotic method; an explanation o f this approach falls outside the scope o f this 4 report.) 5 A quantitative uncertainty analysis of T C D D capturing uncertainty in extrapolating data 6 from animal bioassays to human reference values together with consideration o f epidemiological 7 data from studies o f workers (routine exposures) or the general public (including dietary 8 exposures and those reflecting discrete poisonings or accidental releases) would raise many 9 issues. The major issues are summarized below. 10 11 6.4.2.I. F easibility o f Q uan titatively C h aracterizin g th e U ncertainties E n co u n tered w h en 12 D eterm in in g A p p ro p ria te Types o f S tu d ies (E pidem iological, A n im a l, B oth, a n d 13 O ther) 14 Th e risk assessor must choose the data set(s) that w ill serve as a starting point for 15 dose-response m odeling. W ith respect to T C D D , a w ealth o f anim al bioassay data exist in the 16 scientific literature, across species ranging from rats, m ice, guinea pigs, and hamsters to m ink, 17 dogs and m onkeys, and a variety o f tissues, organs, and systems. In addition, a considerable 18 amount o f human data is available from clinical/case reports, accidental releases, and 19 occupational exposures, including epidem iological data for several cohorts. A s detailed in 20 Sections 2, 4 and 5, some o f the m ain sources o f uncertainty in the T C D D epidem iological data 21 include the healthy w orker effect, confounding and exposure m isclassification. Epidem iolo gical 22 data are usually attended w ith large uncertainties regarding the doses actually received by 23 individuals. Th e difficulty in characterizing individual-level exposures largely stems from 24 having limited internal measures o f T C D D exposure, as biomonitoring data may only be 25 available for one point in time or on a subset o f the exposed population. Although there is little 26 direct evidence o f strong confounding in the cohorts o f T C D D and dioxin-like compounds, some 27 o f the confounders that have been evaluated in a few o f the epidemiological studies include 28 gender, body mass index, age, cigarette and alcohol consumption, and hair and eye color 29 (B a cca relli et al., 2005, 197053; 2006, 197036; E sk e n a zi et al., 2002, 197168; 2002, 197164; 30 Pereg et al., 2002, 199797) . A s discussed in Section 5 on T C D D carcinogenicity, an additional 31 lim itation o f the epidem iological evidence includes the la ck o f organ specificity, as m any o f the This document is a draftfor review purposes only and does not constitute Agency policy. 6-24 DRAFT--DO NOT CITE OR QUOTE 1 studies have shown associations between T C D D exposure and all-cause mortality. W ith 2 disagreement in the literature over the nature, scope, and quality of the epidem iological data for 3 T C D D , given the lack of precedent for a multisite carcinogen without particular sites 4 predominating, some have urged caution in the interpretation of the epidem iological data based 5 on sm all relative risks Popp et al. (2006, 197074) . 6 Despite these uncertainties, the E P A Cancer Guidelines express a clear preference for 7 epidem iological studies over anim al data. Th e question here is w hether quantitative uncertainty 8 analyses based on either a collection of bioassay data or on several epidem iological studies can 9 be combined in some overall uncertainty assessment. D iverse human studies are sometimes 10 com bined into a m eta-analysis, and the issues arising in this regard are instructive. A prim ary 11 challenge of m eta-analytical approaches is com bining heterogeneous effects that may result from 12 studies of different populations, study designs or analytical techniques. The question of whether 13 uncertainty arising from com bining such different studies can be taken into account in 14 quantitative uncertainty an alysis is sim ilar to that of accounting for uncertainty due to m issing 15 covariates in C o x regression (see Section 6.4.2.2). 16 E x istin g standard statistical tools are insufficient to address this issue, as they quantify 17 uncertainty in model parameters estimated from data. H ow ever, exotic methods, such as 18 B a y esia n methods, probabilistic inversion, or structured expert judgment m ay be applicable. 19 Th ese methods can be applied w hen a quantitative model p r e d i c t s other phenomena, even though 20 these phenomena could not be used to estimate the model. The question o f whether such 21 methods could rem ain sufficiently tethered to data, or whether structured expert judgment is 22 unavoidable, is a subject for future research. 23 24 6.4.2.2. U ncertainty in T C D D E x p o su re/D o se in E p id em io lo g ica l S tu dies 25 Uncertainties in epidemiological studies arise from a variety o f study characteristics. 26 There are many types o f epidemiological study designs w hich determine the data structure, 27 including intervention trials, case-control studies, cohort studies and cross-sectional studies. A 28 variety o f mathematical m odels some o f these can be used to analyze epidemiological data; some 29 o f these includes C o x proportional hazard, Poisson regression, linear and logistic regression. 30 The model outputs are based on different measures o f association such as rate ratios, risk ratios, 31 odds ratios, and standardized mortality ratios (S M R s, ratio o f observed to expected deaths). This document is a draftfo r review purposes only and does not constitute Agency policy. 6-25 D R A FT-- DO NOT C IT E OR Q U O TE 1 Exposure uncertainties often concern back-casted exposures based on current serum lipid 2 concentrations, estimated/self reported dietary habits, fish consumption, placenta lipid 3 concentrations, and other measures. 4 Uncertainty in exposure is often dealt with by coarsely grouping a cohort into exposed 5 and unexposed groups. The output o f such a study can be coarse grained in a sim ilar w ay; 6 instead o f computing dose-dependent risk estimates, standard m ortality ratios might be used to 7 compare the exposed and unexposed groups. Packages computing the outputs routinely produce 8 confidence intervals that reflect sampling fluctuations (e.g., can indicate the potential for chance 9 to explain the association), assuming truth o f the model. Additional uncertainty could be 10 factored in w ith exotic methods. A significant issue in epidem iological studies is the effect o f 11 omitted covariates. Omitted covariates in C o x regression w ill bias the estimates o f effects o f 12 included covariates. I f the omitted covariates are independent o f the included covariates, the bias 13 is toward zero in absolute value (Bretagnolle and H uber-C arol, 1988, 543772) ; i f the omitted 14 covariates are not independent, little can be inferred. 15 W ith regard to individual studies, it m ight be possible to identify sp ecific opportunities 16 for uncertainty quantification. T h is is illustrated here using the study o f Steenland et al. (2001, 17 198589) o f m ore than 3,500 m ale w orkers exposed to TC D D -con tam in ated products at eight 18 U .S . chem ical plants. E a c h w orker w as assigned an exposure score based on an estimated level 19 o f contact w ith T C D D , the degree o f T C D D contamination o f product at each plant over time, 20 and the fraction o f a workday in contact w ith the product. F o r 170 workers, the serum T C D D 21 levels w ere also measured. Th e serum levels w ere back-extrapolated to the last time o f exposure 22 using a constant biological half life, and regressed on the exposure scores. This regression 23 model w as used to predict the dose in all w orkers, and predicted dose w as correlated w ith cancer 24 m ortality. Fig u re 6-1 shows a scatter plot o f back-casted versus predicted T C D D serum levels 25 for the 170 w orkers on w hich the regression w as based. 26 G iven a predicted T C D D level, the uncertainty on the back-casted T C D D value could be 27 inferred from such data by various techniques. A key question is whether the actual cancer 28 mortalities among 170 back-casted workers are randomly placed in the conditional distribution 29 given predicted T C D D . Imagine, in other words, that the mortalities among the 170 back-casts 30 are colored red in Figure 6-1. A t any given level o f T C D D prediction, are the red points evenly 31 distributed, or are they shifted to the right? In principle, the correlation between mortality and This document is a draftfo r review purposes only and does not constitute Agency policy. 6-26 D R A FT-- DO NOT C IT E OR Q U O TE 1 back-casted TCDD level, given the predicted level, could be estimated. This amounts to 2 estimating heteroscedasticity in the regression model.67 Then, for each of the 3,538 workers, 3 given his predicted TCDD level, we could sample a back-casted TCDD level, appropriately 4 correlating with mortality, and recompute the dose response analysis. Repeating this many times 5 we could build up a distribution for excess lifetime cancer mortality risk. 6 It is instructive to step through similar issues with regard to biological half life, 7 background and body fat. The Steenland et al. (2001, 197433) analysis assumed a constant 8 TCDD biological half life (8.7 years). A distribution over this half life could plausibly be 9 developed from published sources. Assuming this half life is constant for all workers, but 10 uncertain (epistemic uncertainty), this distribution could easily supplement the previous 11 distribution: first sample a half life (to be applied to all workers), then estimate the regression 12 model for this half life, and sample back-casted TCDD levels given each worker's exposure 13 score, taking account of correlation with mortality. This works if the half life uncertainty is 14 epistemic. However, since the half life is estimated from data, it is more reasonable to suppose 15 that the half life varies from worker to worker (aleatoric uncertainty). Here again the correlation 16 with mortality must be taken into account, indeed it seems reasonable to suppose that the 17 256 cancer deaths involved workers with longer half lives. However, there is no way ex post of 18 determining the biological half life in the deceased workers. 19 The potential impact of uncertainty regarding background exposure and body fat is likely 20 similar to the uncertainty of estimating the half life of TCDD. Steenland et al. (2001, 197433) 21 held the background level constant at the median level (6.1 ppt, range 2.0 to 19.7) for 22 79 nonexposed workers from whom blood was also drawn (see also Section 6.4.2.4). The full 23 distribution of TCDD levels for these nonexposed workers could be used as well. Is it 24 reasonable to suppose that responsive workers (i.e., those exhibiting the response) have 25 background levels that are sampled randomly from this distribution, or might they not plausibly 26 come from the high end of the distribution? The analysis also assumed a constant percentage of 27 body fat (30%), whereas body fat percentage varies in the general population, e.g., for men this 28 has been reported to range from 2 to 38% or more (see Footnote in Section 6.1.3.3). The body 67 H e t e r o s c e d a s t i c i t y o c c u r s w h e n t h e v a r i a n c e o f t h e d e p e n d e n t v a r i a b l e i n a r e g r e s s i o n a n a l y s i s v a r i e s a c r o s s t h e d a ta , v io la tin g th e a ss u m p tio n o f e q u a l v a ria n c e c o m m o n ly u s e d in m a n y re g re s s io n m o d e ls. This document is a draftfo r review purposes only and does not constitute Agency policy. 6-27 DRAFT--DO NOT CITE OR QUOTE 1 fat distribution in the worker population could have been ascertained, but again the question 2 arises, are the responsive workers sampled randomly from this distribution? 3 These three factors, variable half life, variable background, and variable body fat 4 percentage, might combine to make the effective dose level among the responsive workers 5 significantly higher than would appear in a study that assumes these factors to be constant. 6 However, such concerns cannot be addressed in a quantitative uncertainty analysis, unless cancer 7 mortality can be correlated with these variables. In an optimal study design, this information 8 could be retrieved from the data. However, in most observational epidemiological studies such 9 data are not available, and it might be possible to estimate these correlations in some other 10 defensible manner, in which case the effect of exposure uncertainty could be quantified and 11 propagated. Such an analysis would involve substantial effort and should not be undertaken 12 under assumptions that are themselves implausible. Protocols for epidemiological studies do not 13 currently require such uncertainty quantification. In any event, Steenland et al. (2001, 197433) 14 should be recognized for conscientiously identifying these key issues. 15 16 6.4.2.3. U ncertainty in T oxicity E q u iva len ce (TE Q ) E xposu res in E p id em io lo g ica l S tu dies 17 Toxicity equivalence factors (TEFs) are used to infer the health effects of dioxin-like 18 compounds based on their relative potencies compared to TCDD. These factors are not known 19 with certainty, and the question arises whether uncertainty in TEFs can be incorporated into a 20 quantitative uncertainty analysis. The process of deriving TEFs applied by the World Health 21 Organization (WHO, 2005, 198739) is described in Van den Berg et al. (2006, 543769). 22 Distributions of relative potencies (REPs) were developed from the scientific literature, with 23 preference for in vivo studies, as supplemented by in vitro studies. An expert panel used a 24 consensus process to select a TEF value for each congener, in half log steps "Thus, the TEF is a 25 central value with a degree of uncertainty assumed to be at least half a log, which is one order 26 of magnitude. However, it should be realized that TEF assignments are usually within the 50th to 27 75th percentile of the REP distribution, with a general inclination toward the 75th percentile in 28 order to be health protective" (Van den Berg et al., 2006, 543769) (see Figure 6-2 of this 29 document). 30 The WHO considers the uncertainty in TEFs to span one order of magnitude (presumably 31 log uniformly distributed). It would be tempting to use the distributions in Figure 6-2 to quantify This document is a draftfo r review purposes only and does not constitute Agency policy. 6-28 DRAFT--DO NOT CITE OR QUOTE 1 uncertainty in the TEFs in a quantitative uncertainty analysis. However, the issue of dependence 2 in this case is daunting. For example, should values of 1,2,3,7,8,-pentachlorodibenzofuran and 3 2,3,4,7,8-pentachlorodibenzofuran be sampled independently? The choice of dependence 4 structure will have a large effect. As described by (Van den Berg et al., 2006, 543769), the 5 differences in REPs reflect differences in dosing regimens, species, endpoints, mechanisms, and 6 calculation methods. In a quantitative uncertainty analysis one must insure that these are not 7 double counted. 8 9 Reasons for significant differences in REPs for the same congener can be caused 10 by the use of different dosing regimens (acute vs. subchronic), different endpoints, 11 species, and mechanisms (e.g., tumor promotion caused by at least two different 12 mechanisms as for mono-ortho-substituted PCBs), as well as different methods 13 used for calculating REPs. Thus, different methodological approaches used in 14 different studies clearly provide uncertainties when deriving and comparing REPs. 15 If future study designs to derive REPs were more standardized and similar, the 16 variation in REPs when using the same congener, endpoint, and species might be 17 expected to be smaller (Van den Berg et al., 2006, 543769). 18 19 Although the TEFs themselves and the distributions underlying them are based on expert 20 judgment, it is possible to incorporate these into a quantitative uncertainty analysis; however, it 21 is not simply a matter of taking the distributions in Figure 6-2 to predict the results, with 22 uncertainty, of exposure to dioxin-like compounds. The issues of dependence and double 23 counting must first be addressed. Inasmuch as the distributions are the result of expert judgment, 24 this would reasonably involve structured expert judgment as well. (Procedures for this type of 25 assessment have been developed and applied, and it would entail a significant level of effort.) 26 27 6.4.2.4. U ncertainty in B a ck g ro u n d F e e d E x p o su res in B ioassays 28 TCDD is not produced intentionally but rather is formed as a byproduct of volcano 29 eruptions, forest fires, manufacturing of steel and certain chemicals (including some pesticides 30 and paints), pulp and paper bleaching, exhaust emissions, and incineration. It enters the food 31 supply primarily via aerial transport and deposition of emissions, and it bioaccumulates in animal 32 fat. In general, food of animal origin contributes to about 80% of the overall human exposure. 33 For example, Schecter et al. (1997, 198396) measured dioxins in pooled food samples collected 34 in 1995 from supermarkets across the United States. Reported as parts per trillion (ppt) toxicity This document is a draftfor review purposes only and does not constitute Agency policy. 6-29 DRAFT--DO NOT CITE OR QUOTE 1 equivalences (TEQs), fresh water fish had the highest level (1.43); followed by butter (1.07); 2 hotdog/bologna (0.54); ocean fish (0.47); cheese (0.40); beef (0.38); eggs (0.34); ice cream 3 (0.33); chicken (0.32); pork (0.32); milk (0.12); and vegetables, fruits, grains, and legumes 4 (0.07). More recent exposure studies indicate dietary levels have decreased over time. Values 5 reported for the early 2000s by Lorber et al. (2009, 543766), in ppt TEQ, are: fish (0.33); beef 6 (0.12); dairy, other than milk (0.079); eggs (0.06); pork (0.036); poultry (0.018); other meat 7 (0.058); and milk (0.012). 8 These results illustrate that a person's dietary intake of dioxins depends on the relative 9 intake of foods with high or low levels of contamination, and human background levels will vary 10 accordingly. The same applies to experimental animals in bioassays, although in those cases the 11 background intake can in principle be controlled. Some of the effects of TCDD and other AhR 12 agonists in enhancing the early initiation stages of cancers are considered to occur as a result of 13 prenatal exposures that are not included in the standard National Toxicology Program (NTP) 14 bioassay protocol (Brown et al., 1998, 051311; Muto et al., 2001, 548713). Further, to enhance 15 reproducibility and keep statistical fluctuations to a minimum, the standard NTP assays are 16 deliberately run on groups of animals that are relatively uniform genetically, fed uniform diets, 17 and have the minimum possible exposures to toxicants other than the agent(s) being tested. This 18 tends to reduce the potential for observing the consequences of potential interactive effects that 19 might occur in the diverse human population with its variety of dietary and other exposures to a 20 wide range of potentially interacting substances and conditions. 21 A critical question is the extent to which the background exposure influences the 22 dose-response curve, and how this background should be taken into account. One idea, 23 articulated in the recent NRC (2009, 194810) report on science and decisions, involves an 24 "interacting background."68 This can be implemented by computing a virtual dose B which, 25 according to the selected dose-response model, would explain a chosen fraction of the 26 background response. If the chosen model for dose 5 is f(5), the model can be adapted to 68" E f f e c t s o f e x p o s u r e s t h a t a d d t o b a c k g r o u n d p r o c e s s e s a n d b a c k g r o u n d e n d o g e n o u s a n d e x o g e n o u s e x p o s u r e s c a n la c k a th re sh o ld if a b a s e lin e le v e l o f d y s fu n c tio n o c c u rs w ith o u t th e to x ic a n t a n d th e to x ic a n t a d d s to o r a u g m e n ts th e b a c k g ro u n d p ro c e s s . T h u s, e v e n s m a ll d o s e s m a y h a v e a re le v a n t b io lo g ic e ffe c t. T h a t m a y b e d iffic u lt to m e a su re b e c a u se o f b a c k g ro u n d n o ise in th e sy ste m b u t m a y b e a d d re sse d th ro u g h d o se -re sp o n se m o d e lin g p ro ced u re s" (N R C , 20 0 9 ). This document is a draftfo r review purposes only and does not constitute Agency policy. 6-30 DRAFT--DO NOT CITE OR QUOTE 1 account for an interacting background by writing f*(5) = f(5 + B) - f(B). This can alter the 2 model's behavior at zero dose. 3 For example, if f(5) = 57(5n + EC50n), the derivative d(f)/d(5) is n5n-1EC50n/(5n + EC50n)2, 4 which goes to zero as 5 ^ 0 , if n > 1. However, replacing 5 with (5 + B) evidently changes the 5 derivative at zero to nBn-1EC50n/(Bn + EC50n). This model is not yet estimable from data, as we 6 have no way of choosing from the available animal data the fraction of background response to 7 be explained by the model when applied to humans (although judgments could be made if we 8 had better information about the details of the processes that are involved in causing various 9 human health effects). However, as a conceptual model, it serves to remind us that the manner 10 of accounting for background exposures can influence a model's behavior in the low-dose 11 region. (Note that sensitivity analyses can be done showing the consequences of assuming 12 different amounts of interacting background within the context of a specific nonlinear model.) 13 14 6.4.2.5. F easibility o f Q uan tifyin g th e U ncertainties E n co u n tered W hen C h oosin g S p ecific 15 S tu d ies a n d S u bsets o f D a ta (e.g ., S p ecies a n d G en d er) 16 Species, strain, gender, life stage, and other characteristics of experimental animals are 17 selected for a given study based on previous knowledge (e.g., of the species sensitivity, 18 availability of strains having little genetic variation for the endpoints in question, relevance of 19 the MOA, and degree to which the endpoints are similar for humans). Many other decisions are 20 made in designing a bioassay study; will the animals be sacrificed at the termination of the study 21 (if not a lifetime study), or will they be allowed to live out their natural lives? What dosing 22 regimen should be applied? How will the animals be fed and handled? Although such questions 23 may engender uncertainty in the minds of the experimenters, and reviewers; such uncertainty is 24 not amenable for quantitative uncertainty analysis unless and until there are quantitative models, 25 with parameters estimable from data, that can predict the effect of these choices on the response 26 function. 27 28 6.4.2.6. F easibility o f Q uan tifyin g th e U ncertainties E n co u n tered w h en C h oosin g S p ecific 29 Endpoints fo r D ose-Response M odeling 30 Standard experimental protocols guide the selection of exposure/dosing conditions for a 31 given bioassay, including the amount, delivery vehicle, route, timing, dosing frequency and This document is a draftfor review purposes only and does not constitute Agency policy. 6-31 DRAFT--DO NOT CITE OR QUOTE 1 duration, and dose spacing. The goal is to find the dose range where the experimental animals 2 begin to respond adversely, to help anchor the lower end of the dose-response relationship, and 3 to avoid multiple experiments in which all or none of the animals respond. A common 4 recommendation is that the dose levels be chosen such that the increments in probability of 5 response are roughly equal. Hence, the choice of endpoint, dose spacing, and number of animals 6 should be made with these factors in mind. Of particular importance is the number of animals at 7 each dose level in relation to the choice of endpoint and probability of response. Using more 8 animals at the lower dose levels increases the probability of seeing some animals respond; on the 9 other hand, it will give higher weight to the low-dose responses in model fitting and uncertainty 10 quantification. Including many low-dose groups in a study with no expected response can 11 produce a bias in the event of model mis-specification (see Text Box 6-1). The conclusion with 12 regard to the feasibility of this quantitative uncertainty analysis echoes that of the previous 13 paragraph: such uncertainty is not amenable for quantitative analysis unless and until there are 14 quantitative models, with parameters estimable from data, that predict the effect of these choices 15 on the response function. 16 17 6.4.2.7. F easibility o f Q uan tifyin g th e U ncertainties E n co u n tered w h en C h oosin g a S p ecific 18 D o se M etric (T ra d e -O ff betw een C on fiden ce in E stim a te d D o se a n d R eleva n ce o f 19 M O A ) 20 The concept of dose is not straightforward. To review, the Cancer Guidelines provide the 21 following taxonomy: 22 23 Exposure is contact of an agent with the outer boundary of an organism. 24 Exposure concentration is the concentration of a chemical in its transport or 25 carrier medium at the point of contact. 26 Dose is the amount of a substance available for interaction with metabolic 27 processes or biologically significant receptors after crossing the outer boundary of 28 an organism. 29 Potential dose is the amount ingested, inhaled, or applied to the skin. 30 Applied dose is the amount of a substance presented to an absorption barrier and 31 available for absorption (although not necessarily having yet crossed the outer 32 boundary of the organism). 33 This document is a draftfor review purposes only and does not constitute Agency policy. 6-32 DRAFT--DO NOT CITE OR QUOTE 1 Text Box 6-1. Model Mis-Specification and Maximum Likelihood Estimation. T h e m a x im u m lik e lih o o d e stim a te (M L E ) is w id e ly u s e d in sta tistic s b e c a u se o f its a ttra c tiv e p ro p e rtie s: I f th e t r u e m o d e l g e n e r a t i n g t h e d a t a i s f r o m t h e c l a s s w h o s e p a r a m e t e r s a r e b e i n g e s t i m a t e d , then u n d e r r e g u l a r i t y c o n d itio n s, th e e x p e c te d M L E c o n v e rg e s to th e tru e v a lu e , a n d its v a ria n c e c o n v e rg e s to ze ro . T h e c a v e a t a g a in st w h a t is c a lle d " m is-sp e c ific a tio n " is v e ry im p o rta n t a n d e a sily o v e rlo o k e d . A n illu s tra tio n c a n b e e x tra c te d fro m th e N T P (2 0 0 6 a ) d a ta fo r fe m a le ra t tu m o r in c id e n c e o f c h o la n g io c a rc in o m a , re p re se n ta tiv e o f th e d a ta w h ic h p e rs u a d e d th e N A S c o m m itte e th a t th e c a n c e r d o se re sp o n se fo r d io x in w a s " su b lin e a r." NTP (2006a) Female Rat Tumor Incidence Data for Cholangiocarcinoma Blood concentration (ng/kg) Number exposed Number responding Relative frequency 2.56 48 0 0 5.69 46 0 0 9.79 50 1 0 .0 2 16.57 49 4 0 .0 8 29.70 53 25 0 .4 7 T h e H ill m o d e l w ith M L E in th is c a se h a s z e ro slo p e a t ze ro . T h e d e fa u lt L in e a r L o w D o se (L L D ) m o d e l fits a H ill m o d e l to d o se s w ith p o sitiv e re sp o n se s, b u t it e x tra p o la te s lin e a rly fro m th e lo w e s t o b se rv e d n o n z e ro re sp o n se fre q u e n c y . B o th m o d e ls h a v e th e sam e tw o p a ra m e te rs, b u t th e p a ra m e te r v a lu e s o f th e H ill m o d e l u s e d in th e L L D m o d e l a re d iffe re n t fro m th o se in H ill m o d e l fit to a ll d o se s, in c lu d in g d o se s w ith z e ro re sp o n se . A lth o u g h th e n u ll re sp o n se s a re e x p e c te d o n th e L L D m o d e l, th e H ill m o d e l h a s g re a te r lo g lik e lih o o d sin c e it g iv e s h ig h e r p ro b a b ility to th e n u ll re sp o n se s (se e b e lo w ). NTP (2006a) Female Rat Tumor Incidence Data for Cholangiocarcinoma: Low-Dose Linear and Hill Models Blood concentration (ng/kg) 2.56 5.69 9.79 16.57 Number exposed 48 46 50 49 Response probability: Linear Low Dose (LLD) 0 .0 0 5 0 .0 1 2 0 .0 1 4 0 .0 9 Response probability: Hill model 0 .0 0 0 0 9 0 .0 0 1 7 0 .0 1 3 0 .0 9 Probability of cohort null response: LLD 0 .7 7 0 .5 8 Probability of cohort null response: Hill 0 .9 9 0 .9 2 Log Likelihood LLD Hill 2 .4 6 2 .1 6 29.70 53 0 .4 7 0 .4 7 S u p p o se , fo r th e sak e o f illu stra tio n , th a t th e d a ta w e re g e n e ra te d w ith th e re sp o n se p ro b a b ilitie s fro m th e L L D m o d e l. T h e H ill m o d e l w o u ld b e m is-sp e c ifie d in th is c a se , as th e m o d e l g e n e ra tin g th e d a ta is n o t a H ill m o d e l. B e c a u s e o f th e sm a ll c o h o rt siz e , th e p ro b a b ility o f n u ll re s p o n s e s is su c h th a t th e H ill m o d e l h a s g re a te r lik e lih o o d th a n th e L L D m o d e l w ith p ro b a b ility (b a s e d o n b o o ts tra p p in g ) a b o u t 0 .4 3 , e v e n th o u g h th e la tte r, b y c o n stru c tio n , is th e tru e m o d e l. A v e ra g in g o v e r m a n y s im u la te d re sp o n se s fro m th e L L D m o d e l, th e H ill m o d e l u n d e re s tim a te s th e re s p o n s e p ro b a b ilitie s f o r d o s e s 2 .5 6 a n d 5 .6 9 b y fa c to rs o f 7 .5 a n d 2 .1 re s p e c tiv e ly . In th e e v e n t o f s u c h m issp e c ific a tio n , th e b ia s in th e H ill m o d e l w o u ld b e a g g ra v a te d b y in c lu d in g m o re 5 0 -ra t e x p e rim e n ts w ith d o se s lo w e r th a n 2 .5 6 . 2 This document is a draftfor review purposes only and does not constitute Agency policy. 6-33 DRAFT--DO NOT CITE OR QUOTE 1 Absorbed dose is the amount crossing a specific absorption barrier (e.g., the 2 exchange boundaries of skin, lung, and digestive tract) through uptake processes. 3 Internal dose is a more general term, used without respect to specific absorption 4 barriers or exchange boundaries. Delivered dose is the amount of the chemical 5 available for interaction by any particular organ or cell 6 7 Due to their greater causal proximity to the affected organs, using the absorbed dose or 8 internal dose would yield statistically more powerful results and enable more precise predictions 9 than potential dose. If it is not possible to measure these or they were not measured during the 10 conduct of the study (as is commonly the case), then other available dose metrics, such as 11 potential dose or exposure, are used. Due to toxicokinetic variability, different individuals 12 receiving the same exposure may not have the same absorbed dose. Hence, use of either 13 exposure or exposure concentration adds variability to the predicted results. The dose metric 14 should be selected that (1) has the most proximate possible causal relation to the production of an 15 adverse health endpoint, and (2) can be readily related to the units of (external) exposure that 16 will be the basis for assessing human exposures. 17 18 6.4.2.8. F easibility o f Q uan tifyin g th e U ncertainties E n co u n tered W hen C h oosin g M o d el 19 T ype a n d F orm 20 The EPA (2009, 522927)draft white paper on probabilistic methods notes: "There is no 21 consensus on any one well-accepted general methodology for dealing with model uncertainty, 22 although there are various examples of efforts to do so." Model uncertainty was introduced in 23 Section 6.1.3.4. Many statistical techniques are available to evaluate model adequacy or to 24 choose a "best" model. Although it is tempting to qualify such deliberations as "uncertainty that 25 a model is true," one must remember that all models, being idealizations, are false. Ultimately, 26 one is interested in uncertainty with regard to observable phenomena, not with regard to models. 27 Models are merely tools for describing the phenomena. Nonetheless, the choice of a model 28 constrains the ways in which uncertainty can be represented, so the question is how to deal with 29 these constraints. A recent study of uncertainty modeling in dose response (Cooke, 2009, 30 543763) addresses precisely this issue and provides technical details to frame possible options. 31 Before exploring exotic approaches to model uncertainty (i.e., those not yet widely used 32 in dose-response analyses), one feature in the standard statistical treatment of uncertainty must This document is a draftfor review purposes only and does not constitute Agency policy. 6-34 DRAFT--DO NOT CITE OR QUOTE 1 be appreciated. Consider a model based on experimental data, typically bioassay data, in which 2 a certain number of study subjects are exposed to varying doses of a test substance, and in which 3 the numbers of subjects exhibiting a response are tallied. Values for the parameters in the model 4 are chosen by the principle of maximal likelihood: those values are chosen which render the data 5 as likely as possible. According to standard practice, a model is chosen that best fits the data 6 according to one of the accepted criteria, such as reduced R2, or the Akaike Information 7 Criterion. There might be many incompatible models that are nearly as good. 8 One can ask the following: If the experiments on which the model is based were repeated, 9 sampling the same number of experimental subjects from the distribution posited by the model, 10 how much could our parameter estimates change? This is described by a joint distribution over 11 the model's parameters, which captures sampling uncertainty under the assumption that the 12 model is true. Now, all models are false, and as our sample sizes grow the lack of fit in the 13 model becomes increasingly apparent. At the same time, the sample fluctuations in parameter 14 estimates--assuming the model is true--become smaller and smaller. In very large 15 epidemiological studies, standard statistical methods can produce razor-thin confidence bands in 16 this way, which fail to capture experts' uncertainty regarding observable phenomena.69 17 The exotic methods sketched in the beginning of Section 6.4.2 may be viewed as attempts 18 to deal with this feature. Probabilistic inversion methods were deployed on a large scale in the 19 joint U.S. NRC-EU uncertainty analyses noted in Section 6.1. Distributions over model 20 parameters are intended to capture an antecedently defined uncertainty over observable 21 phenomena predicted by the model. This method was applied in dispersion and deposition 22 modeling and further environmental transport models (including uptake) for radionuclides. In 23 most cases, the observable uncertainty was based on structured expert judgment, but it has also 24 been based on binomial uncertainty in bioassay studies. A potential drawback is that it may not 25 prove possible to capture the observable uncertainty in this way with a classically best-fitting 26 model, and new models may be required. 27 Nonparametric Bayesian methods arose in the biomedical and reliability fields. They 28 start with a prior distribution over all nondecreasing dose-response functions, and update these 69S e e , f o r e x a m p le , T u o m is to e t a l. ( 2 0 0 8 , 5 4 8 7 1 5 , T a b l e 6 ) f o r a c o m p a r i s o n o f e x p e r t s ' u n c e r t a i n t y i n h e a l t h e ffe c ts o f fin e p a rtic u la te s w ith u n c e rta in tie s d e riv e d fro m s a m p lin g u n c e rta in ty fro m la rg e e p id e m io lo g ic a l stu d ie s. A lth o u g h th e e x p e rts g e n e ra lly a g re e w ith th e stu d ie s ' c e n tra l e stim a te s, th e ir u n c e rta in ty b a n d s a re o fte n m u c h w id e r th a n th o se su rro u n d in g th e p u b lish e d e stim a te s. This document is a draftfo r review purposes only and does not constitute Agency policy. 6-35 DRAFT--DO NOT CITE OR QUOTE 1 w ith observations from a bioassay. N o further assumptions regarding parametric form are 2 introduced, but the prior distribution rem ains important for doses outside the range of 3 observations. B a y esia n model averaging starts w ith a prior distribution over a set o f candidate 4 m odels, and updates this distribution w ith bioassay data. Th e method is flexib le and intuitive, 5 though attenuation o f the effect o f the prior on the posterior must be verified. 6 A ll these approaches represent attempts to capture " extramodel uncertainty," that is, 7 uncertainty that is not conditional on the truth o f the model. T h is is an active research area, and 8 improvements in methods for capturing extramodel uncertainty in quantitative uncertainty 9 analysis are anticipated. A m ajor effort w ith regard to T C D D dose-response w ould be indicated 10 w hen the strengths and w eakness of the exotic methods are w ell understood. 11 12 6.4.2.9. T h resh old M O A f o r C ancer 13 Th e N A S committee avers that know ledge of the A h R binding M O A im p lies that there is 14 a response threshold for T C D D cancer induction. The differences between individual and 15 population thresholds are not discussed, but the follow ing two possibilities are distinguishable: 16 17 1. Th e threshold is the same for each individual; since human variab ility in A h R binding 18 affinity is rather large (see Section 5.2.3.3), this entails that the threshold is not affected 19 by the binding affinity. 20 2. Th e threshold varies across individuals and is related to the individual A h R binding 21 affinity. 22 23 These two positions are different. A s shown in Section 5.2.3 it is quite possible that each 24 individual in a population has a threshold, whereas the population dose-response relation is 25 linear. B ecau se the N A S committee does not distinguish w h ich of these positions it holds, the 26 feasibility o f quantitative uncertainty analysis is examined here for both. 27 28 i. Quantitative uncertainty analysis concerns a mathematical model. In case (1), this model 29 w ould show how the existence of the A h R binding w ould induce a threshold, 30 independently of the strength of the binding. A ssessin g the feasibility of quantitative 31 uncertainty analysis m ust aw ait the elaboration o f such a model. 32 ii. In case (2), it must be shown that the distribution o f thresholds, and the dose-response 33 function above the threshold, are able to induce a population "zero slope at zero dose" 34 (Z S @ Z ) model. R ecall, the burden o f proof is on this (Z S @ Z ) model. Scoping the This document is a draftfor review purposes only and does not constitute Agency policy. 6-36 DRAFT--DO NOT CITE OR QUOTE 1 population variability with regard to AhR-mediated mechanisms in general, and dioxin 2 sensitivity in particular, is an active area of research. It involves phenotyping human 3 AhR-mediated responsiveness and relating this to polymorphisms in the human 4 population. Harper et al. (2002, 198124) report that a 10-fold variation in binding 5 affinity of AhR for TCDD in human placental tissue did not reveal any polymorphisms, 6 suggesting that the relation between phenotypical and genotypical variation is tenuous. 7 Tuomisto et al. (1999, 548717) demonstrate large variations in efficacy in two rat strains 8 whose binding affinity is similar (Long-Evans, Kd = 3.4, Han/Wistar, Kd = 3.9 (as also 9 discussed in Connor and Aylward, 2006, 197632)), and they also show that this variation 10 is endpoint-specific. The responses in both strains are similar for cytochrome P450 11 (CYP)1A1 induction, but very dissimilar for thymus atrophy, serum bilirubin, and 12 mortality. Toide et al., (2003, 548792) suggest that common biochemical measures of 13 EROD activity might be mediated by CYP1B1 and CYP1A2. The differences in serum 14 bilirubin at doses around 10 pg/kg are about a factor of 30. Han/Wistar rats seldom die at 15 this dose, while mortality of Long Evans rats is about 50%. The mechanisms are not 16 understood. 17 18 Although the mass action dose-response model does not have a threshold, it is possible 19 that certain enzymes block the receptor binding, and until these are overwhelmed, no response 20 occurs. The availability of such enzymes may vary from individual to individual, and may or 21 may not covary with the dissociation constant, Kd. Pursuing these lines of research may result in 22 a convincing demonstration of a population (ZS@Z) model. Such a model would express the 23 individual threshold in terms of parameters that could be estimated with uncertainty from the 24 data. 25 26 6.4.2.10. F easibility o f Q uan tifyin g th e U ncertainties E n co u n tered w h en S electin g th e B M R 27 The NAS committee explicitly requested that the uncertainty attending the choice of a 28 BMR be quantified. Although selecting relevant alternative values for the BMR may provide 29 information of interest, it does not constitute a quantitative analysis of uncertainty. The 30 alternative values must be sampled from some uncertainty distribution. Since this concerns 31 volitional uncertainty, there is no underlying distribution from which to sample, unless the 32 choice of BMR is related to some claim about the state of the world. 33 However, in response to the NAS concerns, this document provides some limited 34 quantitative comparisons of BMR choices. BMDs, BMDLs and OSFs from the animal cancer 35 bioassay benchmark dose modeling assuming 1, 5, and 10% extra risk are compared in units of 36 blood concentrations and human equivalent doses in Tables 5-18 and 5-19, respectively. In This document is a draftfor review purposes only and does not constitute Agency policy. 6-37 DRAFT--DO NOT CITE OR QUOTE 1 addition, M L E and upper bound slope factor estimates based on Cheng et al. (2006, 523122) are 2 presented (see Tab les 5-3 and 5-4). F o r the noncancer effects, key anim al study P O D s 3 (ng/kg-day) are shown based on different dose metrics: administered dose, first-order body 4 burden H E D , and blood concentration (see Tables 4-3 and 4-4). 5 6 6.5. CONCLUSIONS REGARDING THE FEASIBILITY OF QUANTITATIVE 7 UNCERTAINTY ANALYSIS 8 In this section the main conclusions regarding the feasibility o f quantitative uncertainty 9 analysis are sum m arized in relation to specific suggestions made by the N A S committee (see 10 Section 6.5.1). Fo llo w in g this, a suggested research agenda for m oving forward in this area is 11 provided (see Section 6.5.2). 12 13 6.5.1. Summary of NAS Suggestions and Responses 14 O n page 130 o f their report (N A S , 2006, 198441), N A S m akes specific suggestions 15 regarding uncertainty quantification. These are reformatted and presented in italics below. 16 Fo llo w in g each suggestion, a sum m ary o f the discussion in this section is given, w ith reference 17 to the section in w h ich it is addressed. 18 19 EPA should have addressed quantitatively thefollowing sources o f uncertainty: 20 21 Basisfor risk quantification: 22 1. bioassay data, 23 2. occupational cohort data. 24 25 Response: (1) Classical statistical methods yield distributions on model parameters 26 w hich reflect sample fluctuations, assuming that the model is true. T h is type of 27 uncertainty is taken into account in the B M D L . Exo tic methods can account for 28 uncertainty w h ich is not conditional on the truth o f a model, at least for bioassay data 29 (see Section 6.4.2). (2) Fo r epidemiological data, the dose reconstruction often involves 30 assumptions w hich may support data driven uncertainty analysis, if sufficient data can 31 be retrieved. Ex a m p les discussed above include back-casted T C D D level, biological 32 h alf life, body fat and background (see Section 6.4.2.2). Uncertainty in the choice of 33 bioassay data sets or choice o f occupational cohort data sets is volitional, and is not 34 quantified by sampling an input distribution. To be amenable for quantitative 35 uncertainty analysis, the choice must be linked to a statement about the state o f the 36 world (see Section 6.1.1). This document is a draftfo r review purposes only and does not constitute Agency policy. 6-38 D R A FT-- DO NOT C IT E OR Q U O TE 1 Epidemiology data to use: 2 1. risk estimate developed with data aggregatedfrom all suitable studies, 3 2. risk estimate or estimates developed using each study individually. 4 Factors affecting extrapolationfrom occupational to generalpopulation cohorts, 5 including differences in baseline health status, age distribution, the healthy worker 6 survivor effect, and background exposures. 7 8 Response: (1) Quantitative uncertainty analysis based on meta-analysis data poses 9 challenges owing to differences in study protocols. Exotic methods might take us further, 10 the question is whether the restriction to data driven methods (as opposed to expert 11 judgment or Bayesian methods) could be maintained (see Sections 6.4.2.2 and 6.4.2.3). 12 (2) If the general population is characterized by distributions over known confounders 13 whose coefficients are estimated from the epidemiological studies, then uncertainty over 14 these coefficients can be extracted with the methods mentioned in Section 6.4.2.I. 15 Uncertainty due to missing covariates is intractable for data driven uncertainty analysis 16 (see Section 6.4.2.2). 17 Bioassay data to use: 18 1. risk estimate developed with the single data set implying the greatest risk (that is, 19 single study, tumor site, gender), 20 2. risk estimate developed with multiple data sets satisfying an a priori set o f 21 selection criteria. 22 23 Response: (1) Uncertainty in choice of data sets is volitional and is not quantified by 24 sampling an input distribution. To be amenable for quantitative uncertainty analysis the 25 choice must be linked to a statement about the state of the world (see Section 6.1.1). 26 (2) The issue here is similar to the meta-analysis addressed in (2.a). 27 Dose-response model: 28 1. linear dose response, 29 2. nonlinear dose. 30 31 Response: (1) When low dose extrapolation is done using a linear model by default, the 32 uncertainty is volitional. To be amenable for quantitative uncertainty analysis, the choice 33 must be linked to a statement about the state of the world (see Section 6.1.1). The EDx as 34 POD for the linear extrapolation can be subjected to quantitative uncertainty analysis, if 35 based on sufficient bioassay data. (2) With respect to nonlinear dose response, it is 36 possible that human thresholds exist, and that the distribution of thresholds can be 37 characterized in the human population. In as much as the mechanisms for this are not yet 38 understood, there is no quantitative model expressing threshold as a function of 39 parameters which could be estimated, with uncertainty, from data. This currently limits 40 the application of uncertainty quantification (see Section 6.4.2.9). This document is a draftfor review purposes only and does not constitute Agency policy. 6-39 DRAFT--DO NOT CITE OR QUOTE 1 Dose metric: 2 1. average daily intake, 3 2. area under the blood concentration-time curve, 4 3. lifetime average body burden, 5 4. peak body burden, 6 5. other. 7 8 Response: (1-5) The dose m etric is chosen to m axim ize causal proxim ity to the endpoint, 9 w hile maintaining the link to measured exposure (see Section 6.4.2.7). There may be 10 uncertainty w ith regard to w h ich m etric is optim al. I f an inappropriate m etric is chosen 11 in a bioassay study, this w ould be expressed in noisier responses w h ich w ould tend to 12 suppress the dependence of endpoint on dose. A data driven quantitative uncertainty 13 analysis of dose m etric w ould require a m athematical model expressing endpoints as a 14 function, inter alia, of dose m etric, w ith parameters estimated from data. 15 Dose metric-- biological measure: 16 1. free dioxin, 17 2. bound dioxin. 18 19 Response: (1 - 2 ) Th e issue is whether all T C D D available for A h R binding, or only the 20 bound T C D D , should be used as a dose m etric. Binding affinity is determined by more 21 factors than genetic polym orphism s and these other factors are poorly understood (see 22 Section 6.4.2.9). A quantitative uncertainty analysis must await the formulation o f a 23 quantitative model expressing binding affinity in terms o f parameters w h ich can be 24 estimated from data. 25 POD: 26 1. EDlo, 27 2. ED0 5 , 28 3. ED0 1 29 30 Response: (1 -3) Uncertainty in choosing a P O D is volitional. Uncertainty in the value 31 of an E D x can be quantified in a data driven manner if sufficient bioassay data is at hand 32 (see Section 6.4.1.1). 33 Valuefrom ED distribution to use: 34 1. ED, 35 2. lower confidence bound valuefor the ED (LED), 36 3. upper confidence boundfor the ED (UED). 37 This document is a draftfor review purposes only and does not constitute Agency policy. 6-40 DRAFT--DO NOT CITE OR QUOTE 1 Response: (1-3) Given that uncertainty on the POD is quantified, a distribution of the 2 slopes of a linear low dose extrapolation is readily derived, and hence a distribution of a 3 risk specific dose. 4 Where alternative assumptions or methodologies could not be ruled out as implausible or 5 unreasonable, EPA could have estimated the corresponding risks and reported the 6 impact of these alternatives on the risk assessment results. The potential impacts of four 7 sources o f uncertainty are discussed below. 8 1. Thefull range ofplausible parameter valuesfor the dose-response functions used 9 to characterize the dose-response relationshipfor the three occupational cohort 10 studies selected by EPA (Becher et al., 1998, 197173; Ott and Zober, 1996, 11 198408; Steenland et al., 2001, 197433)). 12 2. Use o f other points o f departure, not just the ED01 (or LED01), to develop a CSF. 13 3. Alternative dose-responsefunctionalforms as well as goodness o ffit o f all 14 models, especially at low doses. 15 4. Uncertainty introduced by estimation o f occupational exposures. 16 17 Response: (1) The study of Steenland et al. (2001, 197433) was selected to illustrate the 18 possibilities and limitations of quantitative uncertainty analysis for this type of study (see 19 Section 6.4.2.2). (2) The possibilities for uncertainty quantification with regard to the 20 POD are discussed in Section 6.4.1.1 and in the POD bullet above. (3) Goodness of fit at 21 any measured dose is evaluated in standard packages. There may be different models 22 with comparable goodness of fit at observed doses which differ strongly at doses outside 23 the measured range. Extra model uncertainty, that is, uncertainty which is not conditional 24 on the truth of any given model, is addressed by the exotic methods (see Section 6.4.2). 25 (4) The feasibility of quantifying uncertainty in occupational exposure is study specific. 26 The example of Steenland et al. (2001, 197433) was discussed in some detail (see 27 Section 6.4.2.2). In general, the problem is not so much quantifying the exposure 28 uncertainty, but in quantifying the dependence between the endpoints and the exposure 29 uncertainty. 30 31 6.5.2. How Forw ard? Beyond RfDs and Cancer Slope Factors to Development of 32 Predictive Hum an Dose-Response Functions 33 Uncertainty quantification is an emerging area in science. There are many examples of 34 highly vetted and peer-reviewed uncertainty analyses based on structured expert judgment. 35 Under this process, experts in effect synthesize a wide diversity of information in generating 36 their subjective probability distributions. Where considerable data exist for an environmental 37 pollutant, such as for the well-studied TCDD, it is natural to ask whether these extensive data can 38 be leveraged more directly in uncertainty quantification. This is an area where research could be 39 focused. The requisite knowledge does not yet exist, but there are promising lines of attack. It is This document is a draftfor review purposes only and does not constitute Agency policy. 6-41 DRAFT--DO NOT CITE OR QUOTE 1 therefore not a question of convening blue-ribbon panels to reveal the proper approach; instead 2 multiple approaches should be encouraged, to try out new ideas and share experiences. 3 An important idea that has been pioneered in Europe is to organize bench-test exercises 4 where different approaches are applied to a common problem. This focuses the discussion on 5 real issues and builds a community of capable practitioners. Such initiatives have proven much 6 more productive than simply supporting individual researchers to explore their ideas. 7 Areas for which bench-test exercises might be appropriate include: 8 9 Testing "exotic" methods for capturing model uncertainty; 10 Combining bioassay and epidemiological data for uncertainty quantification; 11 Assessing applicability of structured expert judgment, e.g., for low-dose extrapolation; 12 and 13 Conducting dependence modeling, dependence inference, and dependence elicitation 14 (such as with regard to TEFs). 15 16 Looking beyond compounds for which considerable data exist, there will always be a 17 need to evaluate new substances. The target will be a simple method that: 18 19 1. Can yield predictions of toxicological indicators with uncertainty via a valid probabilistic 20 mechanism; 21 2. Could evolve from approaches based on similarities (such as a random chemical model) 22 under which the new substance could be seen as a random sample from a reference 23 distribution of chemicals considered sufficiently similar, e.g., in terms of structure, 24 physicochemical properties, and biological activity (potency); and 25 3. Is consistent with current risk assessment science and approaches, peer-reviewed and 26 accepted as EPA policy. 27 28 This last feature is important because advancements in risk assessment approaches should 29 extend logically from current methodology based on data analysis and scientific methods. For 30 example, the discussion surrounding uncertainty factors suggests that a probabilistically valid 31 inference system could substantially differ from the current system. Nonetheless, to meld with 32 current practice, it must initialize on the current system and allow this system to evolve in a 33 measured fashion. Ideally, methodological changes should be undertaken in a forum where such 34 issues are being addressed and not within an assessment of a single chemical. This document is a draftfo r review purposes only and does not constitute Agency policy. 6-42 DRAFT--DO NOT CITE OR QUOTE 1 Additional research topics relevant to dioxin that could further inform health assessments 2 include population variability of biokinetic constants, threshold mechanisms for the mass action 3 model, and low-frequency polymorphisms (e.g., less than 1%). Further data and improved 4 methodologies in these areas, combined with developments illustrated elsewhere in this report, 5 will help reduce uncertainties and strengthen our understanding of potential health implications 6 of environmental contaminants. This document is a draftfor review purposes only and does not constitute Agency policy. 6-43 DRAFT--DO NOT CITE OR QUOTE 1 Table 6-1. Key sources of uncertainty 2 Selection of endpoint and of species/strain, gender, life stage, other subject characteristics - critical effect - sensitivity (e.g., species, life stage) _______- human relevance______________________________________________________ Selection of key study(ies): human data and bioassays (strength, inclusion criteria) - epidemiological studies, clinical/case reports (exposure estimate) - adequacy of study design, statistical power (exposure term, histopathology) - human relevance of bioassays (TK, MOA, endpoint) _______- data uncertainty, confidence in data; database deficiencies______________________ Use of TK, dosimetry; body burden; species differences, cross-species extrapolation - bioavailability, dose dependence - half life, life stage, body fat, other compartments, age, other factors - body burden (peak, steady state, lifetime average) - physiologically-based pharmacokinetic (PBPK) modeling _______- scaling (human equivalents), adjustments (default and nondefault; with TD)________ Selection of dose metric - intake (averaging time) - background (what place on the dose-response curve) - free vs. receptor-bound TCDD - tissue-specific concentration _______- lipid-normalized level__________________________________________________ Selection of POD - selection (e.g., NOAEL/LOAEL, BMDL, ED01, 05, 10) - derivation method (e.g., BMD) - choice of model form (e.g., Hill, Weibull) _______- statistical uncertainty at/confidence in POD_________________________________ Selection of dose-response model (e.g., biologically based, multistage) and of BMR - biological plausibility, MOA - model type and form, alternative functional forms - range of plausible parameter values _______- goodness of fit, especially at low doses_____________________________________ Selection of low-dose extrapolation approach - linear/nonlinear - threshold/nonthreshold Human population variability - subpopulations (e.g., occupational, general public, sensitive groups) - polymorphisms - life stage, other features _______- individual vs. population threshold________________________________________ Characterization of risk/effect - adversity of effect (vs. in normal range of variation and adaptation) - uncertainty factors (TK; TD; chemical-specific vs. default; justification) - consistency of methods for endpoints with common MOA - back-extrapolation from occupational data _______- MOE, RfD; beyond a point estimate for SF_________________________________ 3 PBPK = physiologically-based pharmacokinetic; SF = slope factor; TD = toxicodynamic; 4 TK = toxicokinetic. (Other acronyms are as defined elsewhere within this section.) This document is a draftfor review purposes only and does not constitute Agency policy. 6-44 DRAFT--DO NOT CITE OR QUOTE 1 Table 6-2. PODs and amenability for uncertainty quantification 2 POD Data profile Choice Uncertainty quantification LOAEL Experimental dose Choose set of level from set of exposure-response exposure-response data measurements No NOAEL Experimental dose Choose set of level from set of exposure-response exposure-response data measurements No BMDL Estimate from bioassay data Choose BMR, choose dose-response relation No, the BMDL is a quantile of an uncertainty distribution assuming that the dose-response model is true EDx 3 Estimate from set of Choose bioassay Yes, if full bioassay data are exposure-response data experiments to estimate EDx available This document is a draftfor review purposes only and does not constitute Agency policy. 6-45 DRAFT--DO NOT CITE OR QUOTE 1 2 10000.0 : 10 0 0 ,0 ' i 1 1 tto.o ) J i )-- Hfi ft . T 1U , U SO* w. m# * Predicted TCDD i.o 0 - 1 : i ' i n r mm ----- j-----!--i--rrmTj---- 1--r r m n j ---- 1 i--r r r i r r r --------- m i : -- 0,1 i.o 1 0 .0 100,0 1000,0 10000,0 100000.0 3 Back-Extrapolated TCDD 4 5 6 Figure 6-1. Back-casted vs. predicted TCDD serum levels for a worker 7 subset. 8 9 Source: Steenland et al. (2001, 197433). This document is a draftfor review purposes only and does not constitute Agency policy. 6-46 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 6-2. Distribution of in vivo unweighted REP values in the 2004 3 database. 4 5 Source: Van den Berg et al. (2006, 543769), reprinted with permission from Haws 6 et al. (2006, 198416). 7 This document is a draftfor review purposes only and does not constitute Agency policy. 6-47 DRAFT--DO NOT CITE OR QUOTE 1 REFERENCES 2 3 4 Abbott BD; BirnbaumLS; Diliberto JJ (1996). Rapid Distribution of 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) 5 to Embryonic Tissues in C57BL/6NMice and Correlationwith Palatal Uptakein Vitro. Toxicol Appl Pharmacol, 6 141: 256-263. 155093 7 AbrahamK; Geusau A; Tosun Y; Helge H; Bauer S; Brockmller J (2002). Severe 2,3,7,8-tetrachlorodibenzo-p8 dioxin (TCDD) intoxication: insights into the measurement of hepatic cytochrome P450 1A2 induction. Clin 9 Pharmacol Ther, 72: 163-174. 197034 10 AbrahamK; Knoll A; Ende M; Ppke O; Helge H (1996). Intake, fecal excretion, andbody burden of 11 polychlorinated dibenzo-p-dioxins and dibenzofurans inbreast-fed and formula-fed infants. Pediatr Res, 40: 671 12 679.548782 13 AbrahamK; Krowke R; Neubert D (1988). Pharmacokinetics andbiological activity of 2,3,7,8-tetrachlorodibenzo14 p-dioxin. 1. Dose-dependent tissue distribution and induction of hepatic ethoxyresorufin O-deethylase in rats 15 following a single injection. Arch Toxicol, 62: 359-368. 199510 16 Ailhaud G (2006). Adipose tissue as a secretory organ: from adipogenesis to the metabolic syndrome. CR Biol, 329: 17 570-577. 549255 18 Aittomaki A; Lahelma E; Roos E; Leino-Aijas P; Martikainen P (2005). Gender differences in the association of age 19 with physical workload and functioning. Br Med J, 62: 95-100. 197139 20 Akhmedkhanov A, Revich B, Adibi JJ, Zeilert V, Masten SA, Patterson DG Jr, NeedhamLL, Toniolo P (2002). 21 Characterization of dioxin exposure in residents of Chapaevsk, Russia. J Expo Anal Environ Epidemiol, 12: 409 22 417.197140 23 Akhtar FZ; Garabrant DH; KetchumNS; Michalek JE (2004). Cancer in US Air Force veterans of the Vietnamwar. 24 J Occup Environ Med, 46: 123. 197141 25 Alaluusua S; Calderara P; Gerthoux PM; Lukinmaa PL; Kovero O; NeedhamL; Patterson Jr DG; Tuomisto J; 26 Mocarelli P (2004). Developmental dental aberrations after the dioxin accident in Seveso. Environ Health Perspect, 27 112: 1313-1318. 197142 28 Alvarez-Pedrerol M; Ribas-Fito N; Torrent M; Carrizo D; Garcia-Esteban R; Grimalt JO; Sunyer J (2008). Thyroid 29 disruption at birth due to prenatal exposure to beta-hexachlorocyclohexane. Environ Int, 34: 737-740. 594407 30 Amin S; Moore RW; Peterson RE; Schantz SL (2000). Gestational and lactational exposure to TCDD or coplanar 31 PCBs alters adult expression of saccharin preference behavior in female rats. Neurotoxicol Teratol, 22: 675-682. 32 197169 33 Andersen ME; BirnbaumLS; Barton HA; Eklund CR (1997). Regional hepatic CYP1A1 and CYP1A2 induction 34 with 2,3,7,8-tetrachlorodibenzo-p-dioxin evaluated with a multicompartment geometric model of hepatic zonation. 35 Toxicol Appl Pharmacol, 144: 145-155. 197172 36 Andersen ME; Mills JJ; Gargas ML; Kedderis L; BirnbaumLS; Neubert D; Greenlee WF (1993). Modeling 37 receptor-mediated processes with dioxin: Implications for pharmacokinetics and risk assessment. Risk Anal, 13: 25 38 36. 196991 39 Anderson LM; Beebe LE;Fox SD; Issaq HJ; Kovatch RM (1991). Promotion of mouse lung tumors by 40 bioaccumulated polychlorinated aromatic hydrocarbons. Exp Lung Res, 17: 455-471. 201761 This document is a draftfor review purposes only and does not constitute Agency policy. R-1 DRAFT--DO NOT CITE OR QUOTE 1 AnderssonP; McGuire J; Rubio C; Gardin K; Whitelaw ML; Pettersson S; Hanberg A; Poellinger L (2002). A 2 constitutively active dioxin/aryl hydrocarbon receptor induces stomach tumors. PNAS, 99: 9990-9995. 197101 3 Ariens EJ; van RossumJM; Koopman PC (1960). Receptor reserve and threshold phenomena. I. Theory and 4 experiments with autonomic drugs tested on isolated organs. Arch Int Pharmacodyn Ther, 127: 459-478. 594279 5 Armstrong BG (1995). Comparing standardized mortality ratios. Ann Epidemiol, 5: 60-64. 594397 6 ATSDR (1998). Toxicological profile for chlorinated dibenzo-p-dioxins (CDDs). Agency for Toxic Substances and 7 Disease Registry. Atlanta, GA.http://www.atsdr.cdc.gov/toxprofiles/tp104.pdf. 197033 8 Aylward L; Kirman C; Cher D; Hays S (2003). Re: analysis of dioxin cancer threshold. Environ Health Perspect, 9 111: A510. 594305 10 Aylward LL; Bodner KM; Collins JJ; Hays SM (2007). Exposure reconstruction for a dioxin-exposed cohort: 11 Integration of serum sampling data and work histories. , 69: 2063-2066. 197175 12 Aylward LL; Bodner KM; Collins JJ; Wilken M, McBride D; Burns CJ; Hays SM; Humphry N (2009). TCDD 13 exposure estimation for workers at a New Zealand 2,4,5-T manufacturing facility based on serum sampling data. J 14 Expo Sci Environ Epidemiol, TBA: 1-10. 197187 15 Aylward LL; Brunet RC; Carrier G; Hays SM; Cushing CA; NeedhamLL; Patterson DG; Gerthoux PM; Brambilla 16 P; Mocarelli P (2005). Concentration-dependent TCDD elimination kinetics in humans: Toxicokinetic modeling for 17 moderately to highly exposed adults from Seveso, Italy, and Vienna, Austria, and impact on dose estimates for the 18 NIOSH cohort. J Expo Anal Environ Epidemiol, 15: 51-65. 197114 19 Aylward LL; Brunet RC; Starr TB; Carrier G; Delzell E; Cheng H; Beall C (2005). Exposure reconstruction for the 20 TCDD-exposed NIOSH cohort using a concentration- and age-dependent model of elimination. Risk Anal, 25: 945 21 956.197014 22 Aylward LL; Goodman JE; Charnley G; Rhomberg LR (2008). A margin-of-exposure approach to assessment of 23 noncancer risks of dioxins based on human exposure and response data. Environ Health Perspect, 116: 1344-1351 . 24 197068 25 Aylward LL; Hays SM; Karch NJ; Paustenbach DJ (1997). Relative susceptibility of animals and humans to the 26 cancer hazard posed by 2,3,7,8-tetrachlorodibenzo-p-dioxin using internal measures of dose. Environ Sci Tech, 31: 27 1252. 594365 28 Baccarelli A; Giacomini SM; Corbetta C; Landi MT; Bonzini M; Consonni D; Grillo P; Patterson DG; Pesatori AC; 29 Bertazzi PA (2008). Neonatal thyroid function in Seveso 25 years after maternal exposure to dioxin. PLoS Med, 5: 30 e161. 197059 31 Baccarelli A; Hirt C; Pesatori AC; Consonni D; Patterson DG Jr; Bertazzi PA; Dolken G; Landi MT (2006). t(14;18) 32 translocations in lymphocytes of healthy dioxin-exposed individuals from Seveso, Italy. Carcinogenesis, 27: 2001 33 2007.197036 34 Baccarelli A; Mocarelli P; Patterson DG Jr; Bonzini M; Pesatori AC; Caporaso N; Landi MT (2002). Immunologic 35 effects of dioxin: new results from Seveso and comparisonwith other studies. Environ Health Perspect, 110: 1169 36 1173.197062 37 Baccarelli A; Pesatori AC; Consonni D; Mocarelli P; Patterson DG Jr; Caporaso NE; Bertazzi PA; Landi MT 38 (2005). Health status and plasma dioxin levels in chloracne cases 20 years after the Seveso, Italy accident. Br J 39 Dermatol, 152: 459-465. 197053 This document is a draftfor review purposes only and does not constitute Agency policy. R-2 DRAFT--DO NOT CITE OR QUOTE 1 Baccarelli A; Pesatori AC; Masten SA; Patterson DG Jr; NeedhamLL; Mocarelli P; Caporaso NE; Consonni D; 2 Grassman JA; Bertazzi PA; Landi MT (2004). Aryl-hydrocarbon receptor-dependent pathway and toxic effects of 3 TCDD in humans: a population-based study in Seveso, Italy. Toxicol Lett, 149: 287-293. 197045 4 Bang KM; Kim JH (2001). Prevalence of cigarette smoking by occupation and industry in the United States. AmJ 5 Ind Med, 40: 233-239. 197081 6 Banks YB; BirnbaumLS (1991). Absorption of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) after low dose dermal 7 exposure. Toxicol Appl Pharmacol, 107: 302-310. 548742 8 Banks YB; Brewster DW; BirnbaumLS (1990). Age-related changes in dermal absorption of 2,3,7,8 9 tetrachlorodibenzo-p-dioxin and 2,3,4,7,8-pentachlorodibenzofuran. Fundam Appl Toxicol, 15: 163-173. 548741 10 Baron JM; Zwadio-Klarwasser G; Jugert F; Hamann W; Rbben A; Mukhtar H; Merk HF (1998). Cytochrome P450 11 1B1: A major P450 isoenzyme in human blood monocytes and macrophage subsets. BiochemPharmacol, 56: 1105 12 1110.548791 13 Barouki R; Coumoul X; Fernandez-Salguero PM (2007). The aryl hydrocarbon receptor, more than a xenobiotic14 interacting protein. FEBS J, 581: 3608-3615. 543778 15 Bastomsky CH (1977). Enhanced thyroxine metabolismand high uptake goiters in rats after a single dose of 2,3,7,8 16 tetrachlorodibenzo-p-dioxin. Endocrinology, 101: 292-296. 548760 17 Bates MN; Buckland SJ; Garrett N; Ellis H; NeedhamLL; Patterson DG Jr; Turner WE; Russell DG (2004). 18 Persistent organochlorines in the serum of the non-occupationally exposed New Zealand population. Chemosphere, 19 54: 1431-1443. 197113 20 Becher H; Flesch-Janys D; Kauppinen T; Kogevinas M; SteindorfK; Manz A; Wahrendorf J (1996). Cancer 21 mortality in German male workers exposed to phenoxy herbicides and dioxins. Cancer Causes Control, 7: 312-321. 22 197121 23 Becher H; Steindorf K; Flesch-Janys D (1998). Quantitative cancer risk assessment for dioxins using an 24 occupational cohort. Environ Health Perspect, 106: 663-670. 197173 25 Beebe LE; Anver MR; Riggs CW; Fornwald LW; Anderson LM (1995). Promotion of N-nitrosodimethylamine26 initiated mouse lung tumors following single or multiple low dose exposure to 2,3,7,8- tetrachlorodibenzo-p-dioxin. 27 Carcinogenesis, 16: 1345-1349. 548754 28 Bell DR; Clode S; Fan MQ; Fernandes A; Foster PM; Jiang T; Loizou G; MacNicoll A; Miller BG; Rose M; Tran L; 29 White S (2007). Relationships between tissue levels of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), mRNAs and 30 toxicity in the developing male Wistar(Han) rat. Toxicol Sci, 99: 591-604. 197050 31 Bell DR; Clode S; Fan MQ; Fernandes A; Foster PM; Jiang T; Loizou G; MacNicoll A; Miller BG; Rose M; Tran L; 32 White S (2007). Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the developing male Wistar(Han) rat. II: Chronic 33 dosing causes developmental delay . Toxicol Sci, 99: 224-233. 197041 34 Bernert JT; Turner WE; Patterson DG; NeedhamLL (2007). Calculation of serumtotal lipid concentrations for the 35 adjustment of persistent organohalogen toxicant measurements in human samples. Chemosphere, 68: 824-831. 36 594270 37 Bertazzi A; Pesatori AC; Consonni D; Tironi A; Landi MT; Zocchetti C (1993). Cancer incidence in a population 38 accidentally exposed to 2,3,7,8-tetrachlorodibenzo-para-dioxin. Epidemiology, 4: 398-406. 192445 39 Bertazzi PA; Consonni D; Bachetti S; Rubagotti M; Andrea Baccarelli A; Zocchetti C; Pesatori1 AC (2001). Health 40 effects of dioxin exposure: a 20-year mortality study. Am J Epidemiol, 153: 1031-1044. 197005 This document is a draftfo r review purposes only and does not constitute Agency policy. R-3 DRAFT--DO NOT CITE OR QUOTE 1 Bertazzi PA; Zocchetti C; Guercilena S; Consonni D; Tironi A; Landi MT; Pesatori AC (1997). Dioxin exposure 2 and cancer risk: A 15-year mortality study after the "Seveso accident". Epidemiology, 8: 646-652. 197097 3 Bertazzi PA; Zocchetti C; Pesatori AC; Guercilena S; Sanarico M; Radice L (1989). Ten-year mortality study of the 4 population involved in the Seveso incident in 1976. Am J Epidemiol, 129: 1187-1200. 197013 5 BirnbaumLS (1986). Distribution and excretion of 2,3,7,8-tetrachlorodibenzo-p-dioxin in congenic strains of mice 6 which differ at the Ah locus. Drug Metab Dispos, 14: 34-40. 548749 7 Blankenship A; Matsumura F (1997). 2,3,7,8-Tetrachlorodibenzo-p-dioxin-induced activation of a protein tryosine 8 kinase, pp60src, in murine hepatic cytosol using a cell-free system. Mol Pharmacol, 52: 667-675. 543751 9 Bock KW (1994). Aryl hydrocarbon or dioxin receptor: biologic and toxic responses. Rev Physiol Biochem 10 Pharmacol, 125: 1-42. 543755 11 Bock KW; Gschaidmeier H; Heel H; Lehmkoster T; Mnzel PA; Raschko F; Bock-Hennig B (1998). AH receptor12 controlled transcriptional regulation and function of rat and human UDP-glucuronosyltransferase isoforms. Adv 13 Enzyme Regul, 38: 207-222. 548752 14 Bodner K; Collins J; Bloemen L; Carson M (2003). Cancer risk for chemical workers exposed to 2,3,7,8 15 tetrachlorodibenzo-p-dioxin. Occup Environ Med, 60: 672-675. 197135 16 Bond GG; McLaren EA; Brenner FE; Cook RR (1989). Incidence of chloracne among chemical workers potentially 17 exposed to chlorinated dioxins. J Occup Environ Med, 31: 771-774. 064967 18 Bond GG; WetterstroemNH; Roush GJ; McLaren EA; Lipps TE; Cook RR (1988). Causespecific mortality among 19 employees engaged in the manufacture, formulation, or packaging of 2,4-dichlorophenoxyacetic acid and related 20 salts. Occup Environ Med, 45: 98-105. 197183 21 Boverhoff DR; Burgoon LD; Tashiro C; ChittimB; Harkema JR; Jump DB; Zacharewski TR (2005). Temporal and 22 dose-dependent hepatic gene expression patterns in mice provide new insights into TCDD-mediated hepatotoxicity. 23 Toxicol Sci, 85: 1048-1063. 594260 24 Bowman RE; Schantz SL; Gross ML; Ferguson SA (1989). Behavioral effects in monkeys exposed to 2,3,7,8 25 TCDD transmitted maternally during gestation and for four months of nursing. Chemosphere, 18: 235-242. 543745 26 Bowman RE; Schantz SL; Weerasinghe NCA; Gross ML; Barsotti DA (1989). Chronic dietary intake of 2,3,7,8 27 tetrachlorodibenzo-p-dioxin (TCDD) at 5 or 25 parts per trillion in the monkey: TCDD kinetics and dose-effect 28 estimate of reproductive toxicity. Chemosphere, 18: 243-252. 543744 29 Brand KP; Catalano PJ; Hammitt JK; Rhomberg L; Evans JS (2001). Limitations to empirical extrapolation studies: 30 the case of BMD ratios. Risk Anal, 21: 625-640. 543765 31 Brand KP; Rhomberg L; Evans JS (1999). Estimating noncancer uncertainty factors: are ratios NOAELs 32 informative? Risk Anal, 19: 295-308. 007629 33 Bretagnolle J; Huber-Carol C (1988). Effects of omitting covariates in Cox's model of survival data. , 15: 125-138. 34 543772 35 Brouwer A; Morse DC; Lans MC; Schuur AG; Murk AJ; Klasson-Wehler E; Bergman A; Visser TJ (1998). 36 Interactions of persistent environmental organohalogens with the thyroid hormone system: Mechanisms and possible 37 consequences for animal and human health. Toxicol Ind Health, 14: 59-84. 201801 This document is a draftfor review purposes only and does not constitute Agency policy. R-4 DRAFT--DO NOT CITE OR QUOTE 1 Brown J; Goossens LH; Kraan BCP (1997). Probabilistic accident consequence uncertainty study: food chain 2 uncertainty assessment. U.S. Nuclear Regulatory Commission; Commission of the European Communities. 3 Washington, DC; Brussels, Belgium. NUREG/CR-6523, EUR 16771, SAND97-0335. 543739 4 Brown NM; Manzolillo PA; Zhang J-X; Wang J; Lamartiniere CA (1998). Prenatal TCDD and predisposition to 5 mammary cancer in the rat. Carcinogenesis, 19: 1623-1629. 051311 6 Budinsky RA; Paustenbach D; Fontaine D; Landenberger B; Starr TB (2006). Recommended relative potency 7 factors for 2,3,4,7,8 pentachlorodibenzofuran: The impact of different dose metrics. Toxicol Sci, 91: 275-285. 8 594248 9 Buelke-Sam J; Holson JF; Nelson CJ (1982). Blood flow during pregnancy in the rat: II Dynamics of and litter 10 variability in uterine flow. Teratology, 26: 279-288. 020478 11 Buelke-SamJ; Nelson CJ; Byrd RA; Holson JF (1982). Blood flow during pregnancy in the rat: I Flow patterns to 12 maternal organs. Teratology, 26: 269-277. 020477 13 Bueno de Mesquita HB; Doornbos G; Van der Kuip DA; Kogevinas M; WinkelmannR (1993). Occupational 14 exposure to phenoxy herbicides and chlorophenols and cancer mortality in The Netherlands. , 23: 289-300. 196993 15 Burleson GR; Lebrec H; Yang YG; Ibanes JD; Pennington KN; BirnbaumLS (1996). Effect of 2,3,7,8 16 tetrachlorodibenzo-p-dioxin (TCDD) on influenzavirus host resistance in mice. Fundam Appl Toxicol, 29: 40-47. 17 196998 18 Bussard D; Preuss P; White P (2009). Conclusions. In RM Cooke (Ed.),Uncertainty modeling in dose response: 19 bench testing environmental toxicity (pp. 217-224). New York, NY: John Wiley & Sons, Inc. 543770 20 Calvo RM; Jauniaux E; Gulbis B; Asuncion M; Gervy C; Contempre B; Morreale De Escobar G (2002). Fetal 21 tissues are exposed to biologically relevant free thyroxine concentrations during early phases of development. J Clin 22 Endocrinol Metab, 87: 1768-1777. 051690 23 Cantoni L; Salmona M; Rizzardini M (1981). Porphyrogenic effect of chronic treatment with 2,3,7,8 24 tetrachlorodibenzo-p-dioxin in female rats. Dose-effect relationship following urinary excretion of porphyrins. 25 Toxicol Appl Pharmacol, 57: 156-163. 197092 26 Carrier G; Brunet RC; Brodeur J (1995). Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and 27 dibenzofurans in mammalians, including humans. I. Nonlinear distribution of PCDD/PCDF body burdenbetween 28 liver and adipose tissues. Toxicol Appl Pharmacol, 131: 253-266. 197618 29 Carrier G; Brunet RC; Brodeur J (1995). Modeling of the toxicokinetics of polychlorinated dibenzo-p-dioxins and 30 dibenzofurans in mammalians, including humans: II. Kinetics of absorption and disposition of PCDDs/PCDFs . 31 Toxicol Appl Pharmacol, 131: 267-276. 543780 32 CDC (2004). The health consequences of smoking: A report of the Surgeon General. Centers for Disease Control 33 and Prevention, U.S. Department of Health and Human Services. Washington, DC. 056384 34 Cesana GC; de Vito G; Ferrario M; Sega R; Mocarelli P (1995). Trends of smoking habits in northern Italy (1986 35 1990). The WHO MONICA Project in Area Brianza, Italy. MONICA Area Brianza Research Group. Eur J 36 Epidemiol, 11: 251-258. 594366 37 Checkoway H; Pearce N; Crawford-Brown DJ (1989). Research methods in occupational epidemiology. 027173 38 Cheng H; Aylward L; Beall C; Starr TB; Brunet RC (2006). TCDD exposure-response analysis and risk assessment. 39 Risk Anal, 26: 1059-1071. 523122 This document is a draftfor review purposes only and does not constitute Agency policy. R-5 DRAFT--DO NOT CITE OR QUOTE 1 Chevrier J; Eskenazi B; Bradman A; Fenster L; Barr DB (2007). Associations between prenatal exposure to 2 polychlorinated biphenyls and neonatal thyroid-stimulating hormone levels in a Mexican-American population, 3 Salinas Valley, California. Environ Health Perspect, 115: 1490-1496. 594408 4 Chiaro CR; Morales JL; Prabhu KS; Perdew GH (2008). Leukotriene A4 metabolites are endogenous ligands for the 5 AH receptor. Biochemistry, 47: 8445-8455. 543771 6 Choi BC (1992). Definition, sources, magnitude, effect modifiers, and strategies of reduction of the healthy worker 7 effect. J Occup Med, 34: 979-988. 594250 8 Chu I; Lecavalier P; Hkansson H; Yagminas A; Valli VE; Poon P; Feeley M (2001). Mixture effects of 2,3,7,8 9 tetrachlorodibenzo-p-dioxin and polychlorinated biphenyl congeners in rats . Chemosphere, 43: 807-814. 521829 10 Clark GC; Tritscher A; Maronpot R; Foley J; Lucier G (1991). Tumor promotionby TCDD in female rats. In 11 Banbury Report 35: biological basis for risk assessment of dioxin and related compounds (pp. 389-404). Cold 12 Spring Harbor, NY: Cold Spring Harbor Laboratory. 594378 13 Clegg LX; Li FP; Hankey BF; Chu K; Edwards BK (2002). Cancer survival among US whites and minorities: a 14 SEER (Surveillance, Epidemiology, and End Results) Programpopulation-based study. Arch Intern Med, 162: 15 1985-1993. 594267 16 Clewell HJ; Gentry PR; Covington TR; Sarangapani R; Teeguarden JG (2004). Evaluation of the potential impact of 17 age- and gender-specific pharmacokinetic differences on tissue dosimetry. Toxicol Sci, 79: 381-383. 056269 18 Cohen SM; Boobis AR; Meek ME; Preston RJ; McGregor DB (2006). 4-Aminobiphenyl and DNA reactivity: Case 19 study within the context of the 2006 IPCS Human Relevance Framework for Analysis of a cancer mode of action for 20 humans. Crit Rev Toxicol, 36: 803-819. 197621 21 Cole P; Trichopoulos D; Pastides H; Starr T; Mandel JS (2003). Dioxin and cancer: A critical review. Regul Toxicol 22 Pharmacol, 38: 378-388. 197626 23 Collins JJ; Bodner K; Aylward LL; Wilken M; Bodnar CM (2009). Mortality rates among trichlorophenol workers 24 with exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Am J Epidemiol, 170: 501-506. 197627 25 Connor KT; Aylward LL (2006). Human response to dioxin: Aryl hydrocarbon receptor (AhR) molecular structure, 26 function, and dose-response data for enzyme induction indicate an impaired human AhR. J Toxicol Environ Health 27 B Crit Rev, 9: 147-171. 197632 28 Consonni D; Pesatori AC; Zocchetti C; Sindaco R; D'Oro LC; Rubagotti M; Bertazzi PA (2008). Mortality in a 29 population exposed to dioxin after the Seveso, Italy, accident in 1976: 25 years of follow-up. AmJ Epidemiol, 167: 30 847-858. 524825 31 Cooke RM (2009). Uncertainty modeling in dose response: bench testing environmental toxicity. New York, NY: 32 Wiley, John & Sons, Inc. 543763 33 Cooper GS; Klebanoff MA; Promislow J; Brock JW; Longnecker MP (2005). Polychlorinated biphenyls and 34 menstrual cycle characteristics. Epidemiology, 16: 191-200. 594401 35 Cox DR (2006). Combination of data. In Kotz S; Read CB; Balakrishnan N et al. (Ed.),Encyclopedia of statistical 36 sciences (pp. 1074-1081). Hoboken: Wiley. 594342 37 CroftonKM; Craft ES; Hedge JM; Gennings C; Simmons JE; CarchmanRA; Carter WH Jr; DeVito MJ (2005). 38 Thyroid-hormone-disrupting chemicals: Evidence for dose-dependent additivity or synergism. Environ Health 39 Perspect, 113: 1549-1554. 197381 This document is a draftfor review purposes only and does not constitute Agency policy. R-6 DRAFT--DO NOT CITE OR QUOTE 1 Croutch CR; Lebofsky M; SchrammKW; Terranova PF; Rozman KK (2005). 2,3,7,8-Tetrachlorodibenzo-p-dioxin 2 (TCDD) and 1,2,3,4,7,8-hexachlorodibenzo-p-dioxin (HxCDD) alter body weight by decreasing insulin-like growth 3 factor I (IGF-I) signaling. Toxicol Sci, 85: 560-571. 197382 4 Crump K (2002). Critical issues in benchmark calculations from continuous data. Crit Rev Toxicol, 32: 133-153. 5 035681 6 Crump Kenny S; Chiu Weihsueh A; SubramaniamRavi P (2010). Issues in using humanvariability distributions to 7 estimate low-dose risk. Environ Health Perspect, 118: 387-393. 380192 8 Crump KS; Canady R; Kogevinas M (2003). Meta-analysis of dioxin cancer dose response for three occupational 9 cohorts. Environ Health Perspect, 111: 681-687. 197384 10 Crump KS; Hoel DG; Langley CH; Peto R (1976). Fundamental carcinogenic processes and their implications for 11 low dose risk assessment. Cancer Res, 36: 2973-2979. 003192 12 D'Amico M; Agozzino E; Biagino A; Simonetti A; Marinelli P (1999). Ill-defined and multiple causes on death 13 certificates--a study of misclassification in mortality statistics. Eur J Epidemiol, 15: 141-148. 197389 14 DeCaprio AP; McMartin DN; O'Keefe PW; Rej R; Silkworth JB; Kaminsky LS (1986). Subchronic oral toxicity of 15 2,3,7,8-tetrachlorodibenzo-p-dioxin in the guinea pig: Comparisons with a PCB-containing transformer fluid 16 pyrolysate. Fundam Appl Toxicol, 6: 454-463. 197403 17 DeKoning EP; Karmaus W (2000). PCB exposure in utero and via breast milk. A review. J Expo Anal Environ 18 Epidemiol, 10: 285-293. 548801 19 Della Porta G; Dragani TA; Sozzi G (1987). Carcinogenic effects of infantile and long-term 2,3,7,8 20 tetrachlorodibenzo-p-dioxin treatment in the mouse. Tumori, 73: 99-107. 197405 21 Denison MS; Nagy SR (2003). Activation of the aryl hydrocarbon receptor by structurally diverse exogenous and 22 endogenous chemicals. Annu Rev Pharmacol Toxicol, 43: 309-334. 197226 23 DeVito MJ; Ma X; Babish JG; Menache M; BirnbaumLS (1994). Dose-response relationships in mice following 24 subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin: CYP1A1, CYP1A2, estrogen receptor, and protein 25 tyrosine phosphorylation. Toxicol Appl Pharmacol, 124: 82-90. 197278 26 Diliberto JJ; Akubue PI; Luebke RW; BirnbaumLS (1995). Dose-response relationships of tissue distribution and 27 induction of CYP1A1 and CYP1A2 enzymatic activities following acute exposure to 2,3,7,8-tetrachlorodibenzo-p28 dioxin (TCDD) in mice. Toxicol Appl Pharmacol, 130: 197-208. 197309 29 Diliberto JJ; Burgin DE; BirnbaumLS (1997). Role of CYP1A2 in hepatic sequestration of dioxin: Studies using 30 CYP1A2 knock-out mice. BiochemBiophys Res Commun, 236: 431-433. 548755 31 Diliberto JJ; Burgin DE; BirnbaumLS (1999). Effects of CYP1A2 on Disposition of 2,3,7,8-Tetrachlorodibenzo-p32 dioxin, 2,3,4,7,8-Pentachlorodibenzofuran, and 2,2',4,4',5,5'-Hexachlorobiphenyl in CYP1A2 Knockout and Parental 33 (C57BL/6N and 129/Sv) Strains of Mice. Toxicol Appl Pharmacol, 159: 52-64. 143713 34 Diliberto JJ; DeVito MJ; Ross DG; BirnbaumLS (2001). Subchronic Exposure of [3H]- 2,3,7,8-tetrachlorodibenzo35 p-dioxin (TCDD) in female B6C3F1 mice: Relationship of steady-state levels to disposition and metabolism. 36 Toxicol Sci, 61: 241-255. 197238 37 Diliberto JJ; Jackson JA; BirnbaumLS (1996). Comparison of 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) 38 Disposition Following Pulmonary, Oral, Dermal, and Parenteral Exposures to Rats. Toxicol Appl Pharmacol, 138: 39 158-168. 143712 This document is a draftfor review purposes only and does not constitute Agency policy. R-7 DRAFT--DO NOT CITE OR QUOTE 1 Dolwick KM; Schmidt JV; Carver LA; SwansonHI; Bradfield CA (1993). Cloning and expression of a human Ah 2 receptor cDNA. Mol Pharmacol, 44: 911-917. 543762 3 Dragan YP; Schrenk D (2000). Animal studies addressing the carcinogenicity of TCDD (or related compounds) with 4 an emphasis on tumour promotion. Food Addit Contam, 17: 289-302. 197243 5 Dunson DB; Baird DD (2001). A flexible parametric model for combining current status and age at first diagnosis 6 data. Biometrics, 57: 396-403. 197248 7 EC (2009). Nuclear energy library: Archives. Retrieved 17-JUL-09, from http://cordis.europa.eu/fp58 euratom/src/lib_docs.htm. 543738 9 Ema M; Ohe N; Suzuki M; Mimura J; Sogawa K; Ikawa S; Fujii-Kuriyama Y (1994). Dioxinbinding activities of 10 polymorphic forms of mouse and human arylhydrocarbon receptors. J Biol Chem, 269: 27337-27343. 197313 11 Emond C; BirnbaumLS; DeVito MJ (2004). Physiologically based pharmacokinetic model for developmental 12 exposures to TCDD in the rat. Toxicol Sci, 80: 115-133. 197315 13 Emond C; BirnbaumLS; DeVito MJ (2006). Use of a physiologically based pharmacokinetic model for rats to study 14 the influence ofbody fat mass and induction of CYP1A2 on the pharmacokinetics of TCDD. Environ Health 15 Perspect, 114: 1394-1400. 197316 16 Emond C; Michalek JE; BirnbaumLS; DeVito MJ (2005). Comparison of the use of a physiologically based 17 pharmacokinetic model and a classical pharmacokinetic model for dioxin exposure assessments. Environ Health 18 Perspect, 113: 1666-1668. 197317 19 Eskenazi B; Mocarelli P; Warner M; Chee WY; Gerthoux PM; Samuels S; NeedhamLL; Patterson DG Jr (2003). 20 Maternal serum dioxin levels and birth outcomes in women of Seveso, Italy. Environ Health Perspect, 111: 947-953. 21 197158 22 Eskenazi B; Mocarelli P; Warner M; NeedhamL; Patterson DG Jr; Samuels S; Turner W; Gerthoux PM; Brambilla 23 P (2004). Relationship of serum TCDD concentrations and age at exposure of female residents of Seveso, Italy. 24 Environ Health Perspect, 112: 22-27. 197160 25 Eskenazi B; Mocarelli P; Warner M; Samuels S; Vercellini P; Olive D; NeedhamL; Patterson D; Brambilla P 26 (2000). Seveso Women's Health Study: A study of the effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on 27 reproductive health. Chemosphere, 40: 1247-1253. 197162 28 Eskenazi B; Mocarelli P; Warner M; Samuels S; Vercellini P; Olive D; NeedhamLL; Patterson DG, Jr.; Brambilla 29 P; Gavoni N; Casalini S; Panazza S; Turner W; Gerthoux PM (2002). Serumdioxin concentrations and 30 endometriosis: A cohort study in Seveso, Italy. Environ Health Perspect, 110: 629-634. 197164 31 Eskenazi B; Warner M; Marks AR; Samuels S; Gerthoux PM; Vercellini P; Olive DL; NeedhamL; Patterson D Jr; 32 Mocarelli P (2005). Serumdioxin concentrations and age at menopause. Environ Health Perspect, 113: 858-862. 33 197166 34 Eskenazi B; Warner M; Mocarelli P; Samuels S; NeedhamLL; Patterson DG Jr; Lippman S; Vercellini P; Gerthoux 35 PM; Brambilla P; Olive D (2002). Serumdioxin concentrations and menstrual cycle characteristics. Am J 36 Epidemiol, 156: 383-392. 197168 37 Eskenazi B; Warner M; Samuels S; Young J; Gerthoux PM; NeedhamL; Patterson D; Olive D; Gavoni N; 38 Vercellini P; Mocarelli P (2007). Serumdioxin concentrations and risk ofuterine leiomyoma in the Seveso 39 Women's Health Study. Am J Epidemiol, 166: 79-87. 197170 This document is a draftfor review purposes only and does not constitute Agency policy. R-8 DRAFT--DO NOT CITE OR QUOTE 1 Fattore E; Trossvik C; Hakansson H (2000). Relative potency values derived from hepatic vitamin A reduction in 2 male and female Sprague-Dawley rats following subchronic dietary exposure to individual polychlorinated dibenzo3 p-dioxin and dibenzofuran congeners and a mixture therof. Toxicol Appl Pharmacol, 165: 184-194. 197446 4 Fernandez-Salguero PM; Hilbert DM; Rudikoff S; Ward JM; Gonzalez FJ (1996). Aryl-hydrocarbon receptor5 deficient mice are resistant to 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced toxicity. Toxicol Appl Pharmacol, 140: 6 173-179. 197650 7 Ferriby LL; Knutsen JS; Harris M; Unice KM; Scott P; Nony P; Haws LC; Paustenbach D (2007). Evaluation of 8 PCDD/F and dioxin-like PCB serumconcentration data from the 2001-2002 National Health and Nutrition 9 Examination Survey of the United States population. J ExpoSci Environ Epidemiol, 17: 358-371. 548789 10 Fielden MR; Brennan R; Gollub J (2007). A gene expressionbiomarker providesearly prediction and mechanistic 11 assessment of hepatic tumor induction by nongenotoxic chemicals. Toxicol Sci, 99: 90-100. 197298 12 Fingerhut MA; Halperin WE; Marlow DA; Piacitelli LA; Honchar PA; Sweeney MH; Greife AL; Dill PA; 13 Steenland K; Suruda AJ (1991). Cancer mortality in workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin. N Engl 14 J Med, 324: 212-218. 197301 15 Fingerhut MA; Halperin WE; Marlow DA; Piacitelli LA; Honchar PA; Sweeney MH; Greife AL; Dill PA; 16 Steenland K; Suruda AJ (1991). Mortality of U.S. workers employed in the production of chemicale contaminated 17 with 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). U.S. Department of Health and Human Services. Cincinnati, OH. 18 197375 19 Fisher JW; Whittaker TA; Taylor DH; Clewell HJ III; Andersen ME (1989). Physiologically based pharmacokinetic 20 modeling of the pregnant rat: a multiroute exposure model for trichloroethylene and its metabolite, trichloroacetic 21 acid. Toxicol Appl Pharmacol, 99: 395-414. 065288 22 Flesch-Janys D (1997). Analyses of exposure to polychlorinated dibenzo-p-dioxins, furans, and 23 hexachlorocyclohexane and different health outcomes in a cohort of former herbicide-producing workers in 24 Hamburg, Germany. Teratog Carcinog Mutagen, 17: 257-264. 197305 25 Flesch-Janys D; Becher H; Gurn P; Jung D; Konietzko J; Manz A; Papke O (1996). Elimination of polychlorinated 26 dibenzo-p-dioxins and dibenzofurans in occupationally exposed persons. J Toxicol Environ Health, 47: 363-378. 27 197351 28 Flesch-Janys D; Berger J; Gurn P; Manz A; Nagel S; Waltsgott H; Dwyer JH (1995). Exposure to polychlorinated 29 dioxins and furans (PCDD/F) and mortality in a cohort of workers from a herbicide-producing plant in Hamburg, 30 Federal Republic of Germany. Am J Epidemiol, 142: 1165-1175. 197261 31 Flesch-Janys D; Gurn P; Jung D; Konietzko J; Manz A; Papke O (1994). First results of an investigation of the 32 elimination of polychlorinated dibenzo-p-dioxins and dibenzofurans (PCDD/F) in occupationally exposed persons. , 33 21: 93-99. 197372 34 Flesch-Janys D; SteindorfK; Gurn P; Becher H(1998). Estimation of the cumulated exposure to polychlorinated 35 dibenzo-p-dioxins/furans and standardized mortality ratio analysis of cancer mortality by dose in an occupationally 36 exposed cohort. Environ Health Perspect, 106: 655-662. 197339 37 Flodstrom S; Ahlborg UG (1991). Promotion of hepatocarcinogenesis in rats by PCDDs and PCDFs. In Gallo MA; 38 Scheuplein RJ; van der Heijden (Ed.),Banbury Report 35: biological basis for risk assessment of dioxin and related 39 compounds (pp. 405-414). Cold Spring Harbor, NY: Cold Spring Harbor Laboratory. 548728 40 Fox TR; Best LL; Goldsworthy SM; Mills JJ; Goldsworthy TL (1993). Gene expression and cell proliferation in rat 41 liver after 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure. Cancer Res, 53: 2265-2271. 197344 This document is a draftfor review purposes only and does not constitute Agency policy. R-9 DRAFT--DO NOT CITE OR QUOTE 1 Franc MA; Pohjanvirta R; Tuomisto J; Okey AB (2001). Persistent, low-dose 2,3,7,8-tetrachlorodibenzo-p-dioxin 2 exposure: effect on aryl hydrocarbon receptor expression in a dioxin-resistance model. Toxicol Appl Pharmacol, 3 175: 43-53. 197353 4 Franczak A; Nynca A; Valdez KE; Mizinga KM; PetroffBK (2006). Effects of acute and chronic exposure to the 5 aryl hydrocarbon receptor agonist 2,3,7,8-tetrachlorodibenzo-p-dioxin on the transition to reproductive senescence 6 in female Sprague-Dawley rats. Biol Reprod, 74: 125-130. 197354 7 Fretland AJ; Safe S; Hankinson O (2004). Lack of antagonismof 2,3,7,8-tetrachlorodibenzo-p-dioxin's (TCDDs) 8 induction of cytochrome P4501A1 (CYP1A1) by the putative selective aryl hydrocarbon receptor modulator 6-alkyl9 1,3,8-trichlorodibenzofuran (6-MCDF) in the mouse hepatoma cell line Hepa-1c1c7. ChemBiol Interact, 150: 161 10 170.197357 11 Fritz W; Lin TM; Safe S; Moorea RW; Peterson RE (2009). The selective aryl hydrocarbon receptor modulator 6 12 methyl-1,3,8-trichlorodibenxofuran inhibits prostate tumor metastasis in TRMP mice. BiochemPharmacol, 77: 13 1151-1160. 594372 14 Fujii-Kuriyama Y; Ema M; Mimura J;Matsushita N; Sogawa K (1995). Polymorphic forms of the Ah receptor and 15 induction of the CYP1A1 gene. Pharmacogenetics, 5 (S): 149-153. 543727 16 Funatake CJ; Dearstyne EA; Steppan LB; Shepherd DM; Spanjaard ES; Marshak-Rothstein A; Kerkvliet NI (2004). 17 Early consequences of 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure on the activation and survival of antigen18 specific T cells. Toxicol Sci, 82: 129-142. 197267 19 Gasiewicz TA; Henry EC; Collins LL (2008). Expression and activity of aryl hydrocarbon receptors in development 20 and cancer. Crit Rev EukaryotGene Expr,18: 279-321. 473406 21 Gaylor DW; Kodell RL (2000). Percentiles of the product of uncertainty factors for establishing probabilistic risk 22 doses. Risk Anal, 20: 245-250. 548724 23 Ge NL; Elferink CJ (1998). A direct interactionbetween the aryl hydrocarbon receptor and retinoblastoma protein: 24 linking dioxin signaling to the cell cycle. J Biol Chem, 273: 22708-22713. 197702 25 Geusau A; AbrahamK; Geissler K; Sator MO; Stingl G; Tschachler E (2001). Severe 2,3,7,8-tetrachlorodibenzo-p26 dioxin (TCDD) intoxication: Clinical and laboratory effects. Environ Health Perspect, 109: 865-869. 197444 27 Geusau A; Schmaldienst S; Derfler K; (2002). Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) intoxication: 28 Kinetics and trials to enhance elimination in two patients. Arch Toxicol, 76: 316-325. 594259 29 Geyer H; Scheunert I; Korte F (1986). Bioconcentration potential of organic environmental chemicals in humans. 30 Regul Toxicol Pharmacol, 6: 313-347. 064899 31 Geyer HJ; Scheuntert I; Rapp K; Kettrup A; Korte F; GreimH; Rozman K (1990). Correlationbetween acute 32 toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) and total body fat content in mammals. Toxicology, 65: 97 33 107.197700 34 Geyer HJ; SchrammKW; Scheunert I; Schughart K; Buters J; Wurst W; GreimH; Kluge R; Steinberg CE; Kettrup 35 A; Madhukar B; Olson JR; Gallo MA (1997). Considerations on genetic and environmental factors that contribute to 36 resistance or sensitivity of mammals including humans to toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) 37 and related compounds. Ecotoxicol Environ Saf, 36: 213-230. 543768 38 Gielen JE; Nebert DW (1971). Aryl hydrocarbon hydroxylase induction in mammalian liver cell culture. I. 39 Stimulation of enzyme activity in nonhepatic cells and in hepatic cells by phenobarbital, polycyclic hydrocarbons, 40 and 2,2-bis(p-chlorophenyl)-1,1,1-trichloroethane. J Biol Chem, 246: 5189-5198. 543775 This document is a draftfor review purposes only and does not constitute Agency policy. R-10 DRAFT--DO NOT CITE OR QUOTE 1 Goodman DG; Sauer RM (1992). Hepatotoxicity and carcinogenicity in female Sprague-Dawley rats treated with 2 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD): a pathology working group reevaluation. Regul Toxicol Pharmacol, 3 15: 245-252. 197667 4 Goossens LH; Harrison JD; Harper FT; Kraan BCP; Cooke RM; Hora SC (1998). Probabilistic accident 5 consequence uncertainty assessment: uncertainty assessment for internal dosimetry. U.S. Nuclear Regulatory 6 Commission; Commission of the European Communities. Washington, DC; Brussels-Luxembourg. NUREG/CR7 6571, EUR 16773, SAND98-0119. 548726 8 Goossens LH; Kraan BCP; Cooke RM; Ehrhardt J; Fischer F; Hasemann I; Brown J; Jones JA; Smith JG (2001). 9 Nuclear science and technology: Probabilistic accident consequence uncertainty assessment using Cosyma: 10 Uncertainty from the food chain module. European Commission. Luxemborg. EUR 18823EN. 548737 11 Goossens LH; Kraan BCP; Cooke RM; Ehrhardt J; Fischer F; Hasemann I; Jones JA; Brown J; Khursheed A; 12 Phipps A (2001). Probabilistic accident consequence uncertainty assessment using Cosyma: Uncertainty fromthe 13 dose module. European Commission. Luxemborg. EUR 18825EN. 548738 14 Goossens LH; Kraan BCP; Cooke RM; Jones J; Brown J; Ehrhardt J; Fischer F; Hasemann I (2001). Overall 15 uncertainty analysis. European Commission. Luxemborg. EUR 18826EN. 548731 16 Goossens LH; Kraan BCP; Cooke RM; Jones J; Ehrhardt J (2001). Nuclear science and technology: 17 countermeasures uncertainty assessment. European Commission. Luxemborg. EUR 18821EN. 548732 18 Goossens LH; Kraan BCP; Cooke RM; Jones JA; Ehrhardt J; Fischer F; Hasemann I (2001). Uncertainty fromthe 19 early and late health effects module. European Commission. Luxemborg. EUR 18824EN. 548735 20 Goossens LHJ; Cooke RM; Kraan BCP (1996). Evaluation of weighting schemes for expertjudgment studies. Delft 21 University of Technology. Delft, The Netherlands. 548727 22 Goossens LHJ; Kraan BCP; Cooke RM; Boardman J; Jones JA; Harper FT; Young ML; Hora SC (1997). 23 Probabilistic accident consequence uncertainty analysis: uncertainty assessment for deposited material and external 24 doses. Office for Official Publications of the European Communities. Washington, DC; Brussels-Luxembourg. 25 NUREG/CR-6526, EUR 16772, SAND97-2323. 543752 26 Goossens LJH; Kraan BCP; Cooke RM; Jones J; Brown J; Ehrhardt J; Fischer F; Hasemann I (2001). Methodology 27 and processing techniques. European Commission. Luxembourg. EUR 18827EN. 548730 28 Goossens, LH; Kraan, BCP; Cooke, RM; Jones JA; Ehrhardt J; Fischer F; Hasemann I (2001). Probabilistic accident 29 consequence uncertainty assessment using Cosyma: Uncertainty fromthe atmospheric dispersion and deposition 30 module. European Commission. Luxemborg. EUR 18822EN. 548734 31 GrahamMJ; Lucier GW; Linko P; Maronpot RR; Goldstein JA (1988). Increases in cytochrome P-450 mediated 32 17-estradiol 2-hydroxylase activity in rat liver microsomes after both acute administration and subchronic 33 administration of 2,3,7,8-tetrachlorodibenzo-p-dioxin in a two-stage hepatocarcinogenises model. Carcinogenesis, 9: 34 1935-1941. 594375 35 Grassman JA; NeedhamLL; Masten SA; Patterson D; Portier CJ; Lucier GW; Walker NJ (2000). Evidence of 36 hepatic sequestration of dioxin in humans? An examination of tissue levels and CYP1A2 expression. , 48: 87-90. 37 548762 38 Greenlee WF; Hushka LJ; Hushka DR (2001). Molecular basis of dioxin actions: evidence supporting 39 chemoprotection. Toxicol Pathol, 29: 6-7. 015400 This document is a draftfor review purposes only and does not constitute Agency policy. R-11 DRAFT--DO NOT CITE OR QUOTE 1 Greer MA; Goodman G; Pleus RC; Greer SE (2002). Health effects assessment for environmental perchlorate 2 contamination: The dose response for inhibition of thyroidal radioiodine uptake in humans. Environ Health Perspect, 3 110: 927-937. 051202 4 Guess HA; Hoel DG (1977). The effect of dose on cancer latency period. J Environ Pathol Toxicol, 1: 279-286. 5 197464 6 Haarmann-Stemmann T; Bothe H; Abel J (2009). Growthfactors, cytokines and their receptors as downstream 7 targets of arylhydrocarbon receptor (AhR) signaling pathways. BiochemPharmacol, 77: 508-520. 197874 8 Haddow JE; Palomaki GE; Allan WC; Williams JR; Knight GJ; Gagnon J; O'Heir CE; Mitchell ML; Hermos RJ; 9 Waisbren SE; Faix JD; Klein RZ (1999). Maternal thyroid deficiency during pregnancy and subsequent 10 neuropsychological development of the child. N Engl J Med, 341: 549-555. 002176 11 Hahn ME (2002). Aryl hydrocarbon receptors: Diversity and evolution. ChemBiol Interact, 141: 131-160. 099302 12 Hahn ME; Allan LL; Sherr DH (2009). Regulation of constitutive and inducible AHR signaling: complex 13 interactions involving the AHR repressor. BiochemPharmacol, 77: 485-497. 548725 14 Hahn MW (2009). Distinguishing Among Evolutionary Models for the Maintenance of Gene Duplicates. J Hered, 15 100: 605-617. 477460 16 Hakk H; Diliberto JJ; BirnbaumLS (2009). The effect of dose on 2,3,7,8-TCDD tissue distribution, metabolism and 17 elimination in CYP1A2 (-/-) knockout and C57BL/6Nparental strains of mice. Toxicol Appl Pharmacol, 241: 119 18 126.594256 19 Harper N; Connor K; Steinberg M; Safe S (1995). Immunosuppressive activity of polychlorinated biphenyl mixtures 20 and congeners: nonadditive (antagonistic) interactions. Fundam Appl Toxicol, 27: 131-139. 202317 21 Harper PA; Wong JY; Lam MS; Okey AB (2002). Polymorphisms in the human AH receptor. ChemBiol Interact, 22 141: 161-187. 198124 23 Harrad S; Wang Y; Sandaradura S; Leeds A (2003). Human dietary intake and excretion of dioxin-like compounds. 24 J Environ Monit, 5: 224-228. 197324 25 Hassoun EA; Al-Ghafri M; Abushaban A (2003). The role of antioxidant enzymes in TCDD-induced oxidative 26 stress invarious brain regions of rats after subchronic exposure. Free Radic Biol Med, 35: 1028-1036. 198726 27 Hassoun EA; Li F; Abushaban A; Stohs SJ (2000). The relative abilities of TCDD and its congeners to induce 28 oxidative stress in the hepatic and brain tissues of rats after subchronic exposure. Toxicology, 145: 103-113. 197431 29 Hassoun EA; Wang H; Abushaban A; Stohs SJ (2002). Induction of oxidative stress following chronic exposure to 30 TCDD, 2,3,4,7,8-pentachlorodibenzofuran, and 2,3',4,4',5-pentachlorobiphenyl. J Toxicol Environ Health A Curr 31 Iss, 65: 825-842. 543725 32 Hassoun EA; Wilt SC; Devito MJ; Van Birgelen A; AlsharifNZ; BirnbaumLS; Stohs SJ (1998). Induction of 33 Oxidative Stress in Brain Tissues of Mice after Subchronic Exposure to 2,3,7,8-Tetrachlorodibenzo-p-dioxin. , 42: 34 23-27. 136626 35 Hattis D; Baird S; Goble R (2002). A straw man proposal for a quantitative definition of the RfD. Drug Chem 36 Toxicol, 25: 403-436. 548720 37 Hattis D; Banati P; Goble R (1999). Distributions of individual susceptibility among humans for toxic effects--for 38 what fraction of which kinds of chemicals and effects does the traditional 10-fold factor provide how much 39 protection? Ann N Y Acad Sci, 23: 117-142. 594299 This document is a draftfo r review purposes only and does not constitute Agency policy. R-12 DRAFT--DO NOT CITE OR QUOTE 1 Hattis D; Burmaster DE (1994). Assessment ofvariability and uncertainty distributions for practical risk analyses. 2 Risk Anal, 14: 713 - 730. 594301 3 Hattis D; Ginsberg G; Sonawane B; Smolenski S; Russ A; Kozlak M; Goble R (2003). Differences in 4 pharmacokinetics between children and adults- II. Childrens variability in drug elimination half-lives and in some 5 parameters needed for physiologically-based pharmacokinetic modeling. Risk Anal, 23: 117-142. 548773 6 Haws LC; Su SH; Harris M; Devito MJ; Walker NJ; Farland WH; Finley B; BirnbaumLS (2006). Development of a 7 refined database of mammalian relative potency estimates for dioxin-like compounds. Toxicol Sci, 89: 4-30. 198416 8 Henck JM; New MA; Kociba RJ; Rao KS (1981). 2,3,7,8-Tetrachlorodibenzo-p-dioxin: acute oral toxicity in 9 hamsters. Toxicol Appl Pharmacol, 59: 405-407. 543779 10 Henriksen GL; KetchumNS; Michalek J; Swaby JA (1997). Serumdioxin and diabetes mellitus in veterans of 11 Operation Ranch Hand. Epidemiology, 8: 252-258. 197645 12 Hertz-Picciotto I (1995). Epidemiology and quantitative risk assessment: a bridge from science to policy. AmJ 13 Public Health, 85: 484-491. 065678 14 Higgins JPT; Thompson SG; Spiegelhalter DJ (2009). Re-evaluation of random-effects meta analysis. , 172: 137 15 159.594339 16 Hochstein MS, Jr.; Render JA; Bursian SJ; Aulerich RJ (2001). Chronic toxicity of dietary 2,3,7,8 17 tetrachlorodibenzo-p-dioxin to mink. Vet Hum Toxicol, 43: 134-139. 197544 18 Hoel DG; Portier CJ (1994). Nonlinearity of dose-response functions for carcinogenicity. Environ Health Perspect 19 Suppl, 102 (Suppl 1): 109-113. 198741 20 Hglund M; SehnL; Connors JM; Gascoyne RD; Siebert R; Sll T; Mitelman F; Horsman DE (2004). Identification 21 of cytogenetic subgroups and karyotypic pathways of clonal evolution in follicular lymphomas. Genes 22 Chromosomes Cancer, 39: 195-204. 199130 23 Hojo R; Stern S; Zareba G; Markowski VP; Cox C; Kost JT; Weiss B (2002). Sexually dimorphic behavioral 24 responses to prenatal dioxin exposure. Environ Health Perspect, 110: 247-254. 198785 25 Hooiveld M; Heederik DJ; Kogevinas M; Boffetta P; NeedhamLL; Patterson DG Jr; Bueno-de-Mesquita HB 26 (1998). Second follow-up of a Dutch cohort occupationally exposed to phenoxy herbicides, chlorophenols, and 27 contaminants. Am J Epidemiol, 147: 891-901. 197829 28 Huff JE (1992). 2,3,7,8-TCDD: Apotent and complete carcinogen in experimental animals. Chemosphere, 25: 173 29 176. 548757 30 Huff JE; Salmon AG; Hooper NK; Zeise L (1991). Long-termcarcinogenesis studies on 2,3,7,8-tetrachlorodibenzo31 p-dioxin and hexachlorodibenzo-p-dioxins . Cell Biol Toxicol, 7: 67-94. 197981 32 Hurst CH; Abbott BD; DeVito MJ; BirnbaumLS (1998). 2,3,7,8-Tetrachlorodibenzo-p-dioxin in Pregnant Long 33 Evans Rats: Disposition to Maternal and Embryo/Fetal Tissues. , 45: 129-136. 134516 34 Hurst CH; DeVito MJ; BirnbaumLS (2000). Tissue disposition of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) in 35 maternal and developing long-evans rats following subchronic exposure . Toxicol Sci, 57: 275-283. 198806 36 Hurst CH; DeVito MJ; Setzer RW; BirnbaumLS (2000). Acute administration of 2,3,7,8-tetrachlorodibenzo-p37 dioxin (TCDD) in pregnant Long Evans rats: association of measured tissue concentrations with developmental 38 effects. Toxicol Sci, 53: 411-420. 199045 This document is a draftfor review purposes only and does not constitute Agency policy. R-13 DRAFT--DO NOT CITE OR QUOTE 1 Hutt KJ; Shi Zhanquan; Albertini DF; Petroff BK (2008). The environmental toxicant 2,3,7,8-tetrachlorodibenzo-p2 dioxin disrupts morphogenesis of the rat pre-implantation embryo. BMC Developmental Biology, 8: 1-12. 198268 3 IARC (1997). IARC monographs on the evaluation of carcinogenic risks to humans. International Agency for 4 Research on Cancer. Lyon, France. 537123 5 Ikeda M; Tamura M; Yamashita J; Suzuki C; Tomita T (2005). Repeated in utero and lactational 2,3,7,8 6 tetrachlorodibenzo-p-dioxin exposure affects male gonads in offspring, leading to sex ratio changes in F2 progeny. 7 Toxicol Appl Pharmacol, 206: 351-355. 197834 8 ILSI (1994). Physiological parameter values for PBPK models. Risk Science Institute. Washington, DC. 046436 9 Institute of Medicine (1994). Veterans and Agent Orange. Washington, DC: National Acadmies Press. 594376 10 Institute of Medicine (2006). Veterans and Agent Orange: update 2000. Washington, DC: National Academies 11 Press. 594374 12 Ishihara K; Warita K; Tanida T; Sugawara T; Kitagawa H; Hoshi N (2007). Does paternal exposure to 2,3,7,8 13 tetrachlorodibenzo-p-dioxin (TCDD) affect the sex ratio of offspring. J Vet Med Sci, 69: 347-352. 197677 14 James WH (1995). What stabilizes the sex ratio? Ann Hum Genet, 59: 243-249. 197722 15 Jorgensen N; Andersen AG; Eustache F; Irvine DS; Suominen J; Petersen JH; Andersen AN; Auger J; Cawood EH; 16 Horte A; Jensen TK; Jouannet P; Keiding N; Vierula M; Toppari J; Skakkebaek NE (2001). Regional differences in 17 semen quality in Europe. Hum Reprod, 16: 1012-1019. 594402 18 Kang HK; Dalager NA; NeedhamLL; Patterson DG Jr; Lees PS; Yates K; Matanoski GM (2006). Health status of 19 Army Chemical Corps Vietnamveterans who sprayed defoliant in Vietnam. Am J Ind Med, 49: 875-884. 199133 20 Kang SH; Kodell RL; Chen JJ (2000). Incorporating model uncertainties along with data uncertainties in microbial 21 risk assessment. Regul Toxicol Pharmacol, 31: 68-72. 548722 22 Kattainen H; Tuukkanen J; Simanainen U; Tuomisto JT; Kovero O; Lukinmaa P-L; Alaluusua S; Tuomisto J; 23 Viluksela M (2001). In Utero/Lactational 2,3,7,8-Tetrachlorodibenzo-p-dioxin Exposure Impairs Molar Tooth 24 Development in Rats . Toxicol Appl Pharmacol, 174: 216-224. 198952 25 Kauppinen T; Kogevinas M; Johnson E; Becher H; Bertazzi PA; Bueno de Mesquita HB; Coggon D; GreenL; 26 Littorin M; Lynge E Mathews J; Neuberger M; Osman J; Pannett B; Pearce N; Winkelmann R; Saracci R (1993). 27 Chemical exposure in manufacture of phenoxy herbicides and chlorophenols and in spraying of phenoxy herbicides. 28 Am J Ind Med, 23: 903-920. 594388 29 Keller JM; Huet-Hudson Y; Leamy LJ (2008). Effects of 2,3,7,8-tetrachlorodibenzo-p-dioxin on molar development 30 among non-resistant inbred strains of mice: A geometric morphometric analysis. GrowthDevelopment and Aging, 31 71: 3-16. 198033 32 Keller JM; Huet-Hudson YM; Leamy LJ (2007). Qualitative effects of dioxin on molars vary among inbred mouse 33 strains. Arch Oral Biol, 52: 450-454. 198526 34 Keller JM; Zelditch ML; Huet YM; Leamy LJ (2008). Genetic differences in sensitivity to alterations of mandible 35 structure caused by the teratogen 2,3,7,8-tetrachlorodibenzo-p-dioxin. Toxicol Pathol, 36: 1006-1013. 198531 36 Kerger BD; Leung H-W; Scott P; Paustenbach DJ; NeedhamLL; Patterson DG Jr; Gerthoux PM; Mocarelli P 37 (2006). Age- and concentration-dependent elimination half-life of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Seveso 38 children. Environ Health Perspect, 114: 1596-1602. 198651 This document is a draftfor review purposes only and does not constitute Agency policy. R-14 DRAFT--DO NOT CITE OR QUOTE 1 K erger B D ; Leung H W ; Scott P K ; Paustenbach D J (2007). Refinem ents on the age-dependent half-life m odel for 2 estim ating child body burdens o f polychlorodibenzodioxins and dibenzofurans. Chem osphere, 67: S272-S278. 3 548784 4 K etchum N S ; M ichalek JE ; Burton JE (1999). Serum dioxin and cancer in veterans o f O peration R anch H and. A m J 5 Epidem iol, 149: 630-639. 198120 6 K im A H ; K o h n M C ; N yska A ; W alker N J (2003). A rea under the curve as a dose m etric for prom otional responses 7 follo w in g 2,3,7,8-tetrachlorodibenzo-p-dioxin exposure. T o xico l A p p l Pharm acol, 191: 12-21. 199146 8 K itch in K T ; W oods JS (1979). 2,3,7,8-T etrachlorodibenzo-p-dioxin (T C D D ) effects on hepatic m icrosom al 9 cytochrom e P-448-m ediated enzym e activities. T oxicol A p p l Pharm acol, 47: 537-546. 198750 10 K o cib a R J; K eeler P A ; Park C N ; G eh rin g P J (1976). 2,3,7,8-T etrachlorodibenzo-p-dioxin (T C D D ): R esults o f a 13 11 w eek oral toxicity study in rats. T o xicol A p p l Pharm acol, 35: 553-574. 198594 12 K ociba R J; K eyes D G ; B eyer JE ; Carreon R M ; W ade C E ; Dittenber D A ; K alnins R P ; Frauson L E ; Park C N ; 13 Barnard SD ; H um m el R A ; H um iston C G (1978). Results o f a tw o-year chronic toxicity and oncogenicity study o f 14 2,3,7,8-tetrachlorodibenzo-p-dioxin in rats. T o xico l A p p l Pharm acol, 46: 279-303. 001818 15 K ogevinas M ; Becher H ; B enn T; Bertazzi P A ; B offetta P; Bueno-de-M esquita H B ; C oggon D ; C olin D ; Flesch16 Janys D ; Fingerhut M ; Green L ; Kauppinen T ; LJttorin M ; Lynge E ; M athew s JD ; Neuberger M ; Pearce N ; Saracci 17 R (1997). Cancer m ortality in w orkers exposed to phenoxy herbicides, chlorophenols, and dioxins an expanded and 18 updated international cohort study . A m J Epidem iol, 145: 1061-1075. 198598 19 K oh n M C ; Lucier G W ; Clark G C ; Sew all C ; Tritscher A M ; Portier C J (1993). A m echanistic m odel o f effects o f 2 0 D io xin on gene expression in the rat liver . T o xicol A p p l Pharm acol, 120: 138-154. 198601 21 K oh n M C ; M eln ick R L (2002). B iochem ical origins o f the non-m onotonic receptor-m ediated dose-response. Journal 2 2 o f M olecular Endocrinology, 29: 113-123. 199104 23 K o h n M C ; Sew all C H ; Lucier G W ; Portier C J (1996). A m echanistic m odel o f effects o f dioxin on thyroid 2 4 horm ones in the rat. T o xico l A p p l Pharm acol, 165: 29-48. 022626 2 5 K oh n M C ; W alker N J; K im A H ; Portier C J (2001). Physiological m odeling o f a proposed m echanism o f enzym e 2 6 induction by T C D D . T oxicology, 162: 193-208. 198767 2 7 K ollu ri S K ; W eiss C ; K o ff A ; G ttlicher M (). p27(Kip1) induction and inhibition o f proliferation by the 2 8 intracellular A h receptor in developing thym us and hepatom a cells. G enes D e v , 13: 1742-1753. 548721 2 9 K op ylev L ; C h en C ; W hite P (2007). Tow ards quantitative uncertainty assessm ent for cancer risks: central estim ates 3 0 and probability distributions o f risk in dose-response m odeling. R egu l T o xicol Pharm acol, 49: 203-207. 194860 31 K op ylev L ; Jo hn F ox J; C hen C (2009). Com bining risks from several tum ors using M arkov C hain M onte Carlo. In 3 2 R M C oo ke (Ed.),U ncertainty M o d elin g in D o se R esponse (pp. 197-205). H oboken , N J: Jo h n W iley & Sons. 198071 33 K reu zer P E ; Csanady G A ; B au r C ; K essler W ; Ppke O ; G reim H ; F ilser J G (1997). 2,3,7,8-T etrachlorodibenzo-p 3 4 dioxin (T C D D ) and congeners in infants. A toxicokinetic m odel o f hum an lifetim e body burden by T C D D w ith 3 5 special em phasis on its uptake by nutrition. A rch T o xico l, 71: 383-400. 198088 3 6 K rishnan K ; A ndersen M E (1991). Interspecies scaling in pharm acokinetics. In A R escingo; A Thakkur (E d.),N ew 3 7 trends in pharm acokinetics (pp. 203-226). N ew Y o rk , N Y : Plenum Press. 548799 This document is a draftfor review purposes only and does not constitute Agency policy. R-15 DRAFT--DO NOT CITE OR QUOTE 1 Krowke R; Chahoud I; Baumann-Wilschke I; Neubert D (1989). Pharmacokinetics and biological activity of 2,3,7,8 2 tetrachlorodibenzo-p-dioxin 2: pharmacokinetics in rats using a loading-dose/maintenance-dose regime with high 3 doses. Arch Toxicol, 63: 356-360. 198808 4 Kuchiiwa S; Cheng SB; Nagatomo I; Akasaki Y; Uchida M; Tominaga M; Hashiguchi W; Kuchiiwa T (2002). In 5 utero and lactational exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin decreases serotonin-immunoreactive neurons 6 in raphe nuclei of male mouse offspring. Neurosci Lett, 317: 73-76. 198355 7 Kurowicka D; Cooke RM (2006). Uncertainty analysis with high dimensional dependence modelling. West Sussex, 8 England: John Wiley & Sons. 543758 9 LaKind JS; Berlin CM; Park CN; Naiman DQ; Gudka NJ (2000). Methodology for characterizing distributions of 10 incremental body burdens of 2,3,7,8-TCDD and DDE frombreast milk in North American nursing infants. J Toxicol 11 Environ Health A Curr Iss, 59: 605-639. 198094 12 Lakshmanan MR; Campbell BS; Chirtel SJ; Ekarohita N; Ezekiel M (1986). Studies on the mechanismof absorption 13 and distribution of 2,3,7,8-tetrachlorodibenzo-p-dioxin in the rat. J Pharmacol Exp Ther, 239: 673-677. 548729 14 Landi MT, Consonni D, Patterson DG Jr, NeedhamLL, Lucier G, Brambilla P, Cazzaniga MA, Mocarelli P, 15 Pesatori AC, Bertazzi PA, Caporaso NE.. (1998). 2,3,7,8-Tetrachlorodibenzo-p-dioxinplasma levels in Seveso 20 16 years after the accident. Environ Health Perspect, 106: 273-277. 594409 17 Landi MT; Bertazzi PA; Baccarelli A; Consonni D; Masten S; Lucier G; Mocarelli P; NeedhamL; Caporaso N; 18 Grassman J (2003). TCDD-mediated alterations in the AhR-dependent pathway in Seveso, Italy, 20 years after the 19 accident. Carcinogenesis, 24: 673-680. 198362 20 Larsen JC (2006). Risk assessments of polychlorinated dibenzo-p-dioxins, polychloriniated dibenzofurans, and 21 dioxin-like polychlorinated biphenyls in food. Mol Nutr Food Res, 50: 885-896. 548744 22 Latchoumycandane C; Chitra C; Mathur P (2002). Induction of oxidative stress in rat epididymal sperm after 23 exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. Arch Toxicol, 76: 113-118. 197839 24 Latchoumycandane C; Chitra KC; Mathur PP (2002). The effect of 2,3,7,8-tetrachlorodibenzo-p-dioxin on the 25 antioxidant systemin mitochondrial and microsomal fractions of rat testis. Toxicology, 171: 127-135. 198365 26 Latchoumycandane C; Chitra KC; Mathur PP (2003). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) induces 27 oxidative stress in the epididymis and epididymal spermof adult rats. Arch Toxicol, 77: 280-284. 543746 28 Latchoumycandane C; Mathur PP (2002). Effects ofvitamin E on reactive oxygen species-mediated 2,3,7,8 29 tetrachlorodibenzo-p-dioxin toxicity in rat testis. J Appl Toxicol, 22: 345-351. 197498 30 Lawrence GS; Gobas FAPC (1997). A pharmacokinetic analysis of interspecies extrapolation in dioxin risk 31 assessment. Chemosphere, 35: 427-452. 199072 32 Lean MEJ; Han TS; Deurenberg P (1996). Predicting body composition by densitometry from simple 33 anthropometric measurements. Am J Clin Nutr, 63: 4-14. 548770 34 Lee DJ; Fleming LE; Arheart KL; LeBlanc WG; Caban AJ; Chung-Bridges K; Christ SL; McCollister KE; Pitman T 35 (2007). Smoking rate trends in U.S. occupational groups: the 1987 to 2004 National Health Interview Survey. J 36 Occup Environ Med, 49: 75-81. 594391 37 Lehman AJ; Fitzhugh OG (1954). 100-fold margin of safety. , 18: 33-35. 003195 38 Leo A; Hansch C; Elkins D (1971). Partition coefficients and their uses. ChemRev, 71: 557-558. 019600 This document is a draftfor review purposes only and does not constitute Agency policy. R-16 DRAFT--DO NOT CITE OR QUOTE 1 Leung H-W; Poland A; Paustenbach DJ; Murray FJ; Andersen ME (1990). Pharmacokinetics of [125I]-2-iodo-3,7,82 trichlorodibenzo-p-dioxin in mice: analysis with a physiological modeling approach. Toxicol Appl Pharmacol, 103: 3 411-419. 192833 4 Leung HW; Kerger BD; Paustenbach DJ (2006). Elimination half-lives of selected polychlorinated dibenzodioxins 5 and dibenzofurans in breast-fed human infants. J Toxicol Environ Health A Curr Iss, 69: 437-443. 548779 6 Leung HW; Ku RH; Paustenbach DJ; Andersen ME (1988). Aphysiologically based pharmacokinetic model for 7 2,3,7,8-tetrachlorodibenzo-p-dioxin in C57BL/6J and DBA/2J mice. Toxicol Lett, 42: 15-28. 198815 8 Li B; Liu HY; Dai LJ; Lu JC; Yang ZM; Huang L (2006). The early embryo loss causedby 2,3,7,8 9 tetrachlorodibenzo-p-dioxin may be related to the accumulation of this compound in the uterus. Reprod Toxicol, 21: 10 301-306. 199059 11 Li CY; Sung FC (1999). A review of the healthy worker effect in occupational epidemiology. Occup Med (Lond), 12 49: 225-9. 198427 13 Li X; Johnson DC; Rozman KK (1997). 2,3,7,8-Tetrachlorodibenzo-p-dioxin (TCDD) increases release of 14 luteinizing hormone and follicle-stimulating hormone from the pituitary of immature female rats in vivo and in vitro. 15 Toxicol Appl Pharmacol, 142: 264-269. 199060 16 Limbird LE (1996). Cell surface receptors: a short course on theory and method. 594276 17 Longnecker MP; Gladen BC; Patterson DG; Rogan WJ (2000).Polychlorinatedbiphenyl (PCB) exposure in relation 18 to thyroid hormone levels in neonates. Epidemiology, 11: 249-254. 201463 19 Lorber M; Patterson D; Huwe J; Kahn H (2009). Evaluation of background exposures of Americans to dioxin-like 20 compounds in the 1990s and the 2000s . Chemosphere, 77: 640-651. 543766 21 Lorenzen A; Okey AB (1991). Detection and characterization of Ah receptor in tissue and cells from human tonsils. 22 Toxicol Appl Pharmacol, 107: 203-214. 198397 23 Lucier GW (1991). Humans are a sensitive species to some of the biochemical effects of structural analogs of 24 dioxin. Environ Toxicol Chem, 10: 727-735. 198691 25 Lucier GW; Rumbaugh RC; McCoy Z; Hass R; Harvan D; Albro P (1986). Ingestion of soil contaminated with 26 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) alters hepatic enzyme activities in rats. Fundam Appl Toxicol, 6: 364 27 371.198398 28 Lucier GW; Tritscher A; Goldsworthy T; Foley J; Clark G; Goldstein J; Maronpot R (1991). Ovarian hormones 29 enhance 2,3,7,8-tetrachlorodibenzo-p-dioxin-mediated increases in cell proliferation and preneoplastic foci in a two30 stage model for rat hepatocarcinogenesis. Cancer Res, 51: 1391-1397. 199007 31 Lutz WK (1990). Dose-response relationship and low dose extrapolation in chemical carcinogenesis. 32 Carcinogenesis, 11: 1243-1247. 000399 33 Lutz WK (1999). Dose-response relationships in chemical carcinogenesis reflect differences in individual 34 susceptibility. Hum Exp Toxicol, 18: 707-712. 594298 35 Lutz WK (2001). Susceptibility differences in chemical carcinogenesis linearize the dose-response relationship: 36 threshold doses canbe defined only for individuals. DNA Repair (Amst), 482: 71-76. 053426 37 Lutz WK; Gaylor DW (2008). Letter to the editor. Dose-response relationships for cancer incidence reflect 38 susceptibility distributions. ChemRes Toxicol, 21: 971-972. 594297 This document is a draftfor review purposes only and does not constitute Agency policy. R-17 DRAFT--DO NOT CITE OR QUOTE 1 Lutz WK; Gaylor DW; Conolly RB; Lutz RW (2005). Nonlinearity and thresholds in dose-response relationships for 2 carcinogenicity due to sampling variation, logarithmic dose scaling, or small differences in individual susceptibility. 3 Toxicol Appl Pharmacol, 207: S565-S569. 087763 4 Mackie D; Liu J; Loh Y-S; Thomas V (2003). No evidence of dioxin cancer threshold. Environ Health Perspect, 5 111: 1145-1147. 594303 6 Mally A; Chipman JK (2002). Non-genotoxic carcinogens: Early effects on gapjunctions, cell proliferation and 7 apoptosis in the rat. Toxicology, 180: 233-248. 198098 8 Manchester DK; Gordon SK; Golas CL; Roberts EA; Okey AB (1987). Ah receptor in human placenta: stabilization 9 by molydate and characterization ofbinding of 2,3,7,8-tetrachlorodibenzo-p-dioxin, 3-methylcholanthrene, and 10 benzo(a)pyrene. Cancer Res, 47: 4861-4868. 198054 11 Manz A; Berger J; Dwyer JH; Flesch-Janys D; Nagel S; Waltsgott H (1991). Cancer mortality among workers in 12 chemical plant contaminated with dioxin. Lancet, 338: 959-964. 199061 13 Markowski VP; Zareba G; Stern S; Cox C; Weiss B (2001). Altered operant responding for motor reinforcement and 14 the determination ofbenchmark doses following perinatal exposure to low-level 2,3,7,8-tetrachlorodibenzo-p15 dioxin. Environ Health Perspect, 109: 621-627. 197442 16 Maronpot RR; Foley JF; Takahashi K; Goldsworthy T; Clark G; Tritscher A; Portier C; Lucier G (1993). Dose 17 response for TCDD promotion of hepatocarcinogenesis in rats initiated with DEN: histologic, biochemical, and cell 18 proliferation endpoints. , 101: 643-642. 198386 19 Maronpot RR; Montgomery CA; Boorman GA; McConnell EE (1986). National Toxicology Program nomenclature 20 for hepatoproliferative lesions of rats. Toxicol Pathol, 14: 263-273. 013967 21 Maronpot RR; Pitot HC; Peraino C (1989). Use of rat liver altered focus models for testing chemicals that have 22 completed two-year carcinogenicity studies. Toxicol Pathol, 17: 651-652. 548778 23 Maruyama W; Yoshida K; Tanaka T; Nakanishi J (2002). Determination of tissue-blood partition coefficients for a 24 physiological model for humans, and estimation of dioxin concentration in tissues. Chemosphere, 46: 975-985. 25 198448 26 Matsumoto Y; Ide F; Kishi R; Akutagawa T; Sakai S; Nakamura M; Ishikawa T; Fujii-Kuriyama Y; Nakatsuru Y 27 (2007). Aryl hydrocarbon receptor plays a significant role in mediating airborne particulate-induced carcinogenesis 28 in mice. Environ Sci Tech, 41: 3775-3780. 548748 29 McBride DI, Collins JJ, Humphry NF, Herbison P, Bodner KM, Aylward LL, Burns CJ, Wilken M (2009). 30 Mortality in workers exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin at a trichlorophenol plant in New Zealand. J 31 Occup Med, 51: 1049-56. 198490 32 McBride DI; Burns CJ; Herbison GP; Humphry NF; Bodner K; Collins JJ (2009). Mortality in employees at a New 33 Zealand agrochemical manufacturing site. Occup Med (Lond), 59: 255-263. 197296 34 McEwen LN, Kim C, Haan M, Ghosh D, Lantz PM, Mangione CM, Safford MM, Marrero D, Thompson TJ, 35 Herman WH; TRIAD Study Group (2006). Diabetes reporting as a cause of death: results fromthe Translating 36 Research Into Action for Diabetes (TRIAD) study. Diabetes Care, 29: 247-253. 594400 37 McMichael AJ (1976). Standardized mortality ratios and the "healthy worker effect": scratchingbeneath the surface. 38 J Occup Environ Med, 18: 165-168. 073484 39 McMillan BJ; Bradfield CA (2007). The aryl hydrocarbon receptor sans xenobiotics: endogenous function in genetic 40 model systems. Mol Pharmacol, 72: 487-498. 543777 This document is a draftfo r review purposes only and does not constitute Agency policy. R-18 DRAFT--DO NOT CITE OR QUOTE 1 McNulty WP; Nielsen-Smith KA; Lay JO Jr; Lippstreu DL; Kangas NL; Lyon PA; Gross ML (1982). Persistence of 2 TCDD in monkey adipose tissue. Food Chem Toxicol, 20: 985-986. 543782 3 Michalek JE; Pavuk M (2008). Diabetes and cancer inveterans of Operation Ranch Hand after adjustment for 4 calendar period, days of spraying, and time spent in Southeast Asia. J Occup Environ Med, 50: 330-340. 199573 5 Michalek JE; Pirkle JL; NeedhamLL; Patterson DG Jr; Caudill SP; Tripathi RC; Mocarelli P (2002). 6 Pharmacokinetics of 2,3,7,8-tetrachlorodibenzo-p-dioxin in Seveso adults and veterans of operation Ranch Hand. J 7 Expo Anal Environ Epidemiol, 12: 44-53. 199579 8 Michalek JE; Pirkle JL; Caudill SP; Tripathi RC; Patterson DG Jr; NeedhamLL (1996). Pharmacokinetics of TCDD 9 in veterans of Operation Ranch Hand: 10-year follow-up. J Toxicol Environ Health, 47: 209-220. 198893 10 Micka J; Milatovich A; Menon A; Grabowski GA; Puga A; Nebert DW (1997). Human Ah receptor (AHR) gene: 11 Localization to 7p15 and suggestive correlation ofpolymorphismwith CYP1A1 inducibility. Pharmacogenetics, 7: 12 95-101. 548797 13 Miettinen HM; Sorvari R; Alaluusua S; Murtomaa M; Tuukkanen J; Viluksela M (2006). The Effect of Perinatal 14 TCDD exposure on caries susceptibility in rats. Toxicol Sci, 91: 568-575. 198266 15 Milbrath MO; Wenger Y; Chang CW; Emond C; Garabrant D; Gillespie BW; Jolliet O (2009). Apparent half-lives 16 of dioxins, furans, and polychlorinated biphenyls as a function of age, body fat, smoking status, and breast-feeding. 17 Environ Health Perspect, 117: 417-425. 198044 18 Mocarelli P (2001). Seveso: a teaching story. Chemosphere, 43: 391-402. 197002 19 Mocarelli P; Needham LL; Marocchi A; Patterson DG Jr; Brambilla P; Gerthoux PM; Meazza L; Carreri V 20 (1991). Serumconcentrations of 2,3,7,8-tetrachlorodibenzo-p-dioxin and test results from selected residents of 21 Seveso, Italy . J Toxicol Environ Health A Curr Iss, 32: 357-366. 199600 22 Mocarelli P; Brambilla P; Gerthoux PM; Patterson Jr DG; Needham LL (1996). Change in sex ratio with exposure 23 to dioxin. Lancet, 348: 409.197637 24 Mocarelli P; Gerthoux PM; Ferrari E; Patterson Jr DG; Kieszak SM; Brambilla P; Vincoli N; Signorini S; 25 Tramacere P; Carreri V; Sampson EJ; Turner WE (2000). Paternal concentrations of dioxin and sex ratio of 26 offspring. Lancet, 355: 1858-1863. 197448 27 Mocarelli P; Gerthoux PM; Patterson DG Jr; Milani S; Limonata G; Bertona M; Signorini S; Tramacere P; Colombo 28 L; Crespi C; Brambilla P; Sarto C; Carreri V; SampsonEJ; Turner WE; Needham LL (2008). Dioxin exposure, from 29 infancy through puberty, produces endocrine disruption and affects human semen quality . Environ Health Perspect, 30 116: 70-77. 199595 31 Monson RR (1986). Observations on the healthy worker effect. J Occup Environ Med, 28: 425-433. 001410 32 Morreale de Escobar G; Obregon MJ; Escobar del Ray F (2000). Is neuropsychological development related to 33 maternal hypothyroidismorto maternal hypothyroxinemia? J ClinEndocrinol Metab, 85:3975-3987. 019231 34 Moser GA; McLachlan MS(2001). The influence of dietary concentrationon the absorption and excretion of 35 persistent lipophilic organic pollutants in the human intestinal tract. Chemosphere, 45: 201-211. 198045 36 Muller A; De La Rochebrochard E; Labb-Declves C; Jouannet P; Bujan L; Mieusset R; Le Lannou D; Guerin JF; 37 Benchaib M; Slama R; Spira A (2004). Selectionbias in semen studies due to self-selection ofvolunteers. Hum 38 Reprod, 19: 2838-2844. 594403 This document is a draftfor review purposes only and does not constitute Agency policy. R-19 DRAFT--DO NOT CITE OR QUOTE 1 M urdoch D J; K rew ski D (1988). Carcinogenic risk assessm ent w ith tim e-dependent exposure patterns. R isk A n al, 8: 2 521-530. 548718 3 M urdoch D J; K rew ski D ; W argo J (1992). Can cer risk assessm ent w ith interm ittent exposure. R isk A n al, 12: 569 4 5 7 7 .548719 5 M urphy JM ; Sexton D M ; Barnett D N ; Jones G S ; W ebb M J; Collins M ; Stainforth D A (2004). Q uantification o f 6 m odeling uncertainties in a large ensem ble o f clim ate change sim ulations. N ature, 430: 768-772. 543741 7 M urray F J; Sm ith F A ; N itschke K D ; H um iston C G ; K o cib a R J; Schw etz B A (1979). Three-generation reproduction 8 study o f rats given 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) in the diet . T o xico l A p p l Pharm acol, 50: 241-252. 9 197983 10 M uto T; W akui S; Im ano N ; N akaaki K ; H ano H ; Furusato M ; M asaoka T (2001). In-utero and lactational exposure 11 o f 3,3',4,4',5-pentachlorobiphenyl m odulate dim ethlbenz[a]anthracene-induced rat m am m ary carcinogenesis. J 12 T oxicol Pathol, 4: 213-224. 548713 13 M yers JE ; Thom pson M L (1998). M eta-analysis and occupational epidem iology. O ccup M ed (Lond), 48: 99-101. 14 594395 15 N agel S; B erger J; Flesch-Janys D ; M an z A ; O llroge I (1994). M ortality and cancer m ortality in a cohort o f fem ale 16 w orkers o f a herbicide producing plant exposed to polychlorinated dibenzo-p-dioxins and furans. Inform B iom et 17 Epidem iol M ed B iol, 25: 32-38. 594369 18 N A S (2006). H ealth risks from dioxin and related com pounds. Retrieved 0 9-FEB -1 0 , from 19 http://w w w .nap.edu/w ebcast/w ebcast_detail.php?w ebcast_id=328. 543760 2 0 N A S (2006). H ealth risks from dioxin and related com pounds: Evaluation o f the E P A reassessm ent. N ational 21 A cadem y o f Scien ce. W ashin gton, D C .http://w w w .nap.edu/catalog.php?record_id=11688. 198441 2 2 N A S (2009). Tow ard a unified approach to dose-response assessm ent: the need for an im proved dose-response 23 fram ew ork. N ational A cadem ics Press. W ashington D C . 594307 2 4 N A S A (2002). Probabilistic risk assessm ent procedures guide for N A S A m anagers and practitioners. N ational 25 Aeronautics and Space Adm inistration. W ashington, D C . 543734 2 6 N ebert D W ; Petersen D D ; Fornace A J Jr (1990). C ellular responses to oxidative stress: the [Ah] gene battery as a 2 7 paradigm . Environ H ealth Perspect, 88: 13-25. 548756 2 8 N ebert D W ; Peterson D D ; Puga A (1991). H um an A h locus polym orphism and cancer: Inducibility o f C Y P IA 1 and 2 9 other genes b y co m b u stion products and d ioxin . P h arm acogen etics, 1: 6 8 -7 8 . 543728 3 0 N eedham L L ; Barr D B ; Caudill SP; Pirkle JL ; Turner W E ; Osterloh J; Jones R L ; Sam pson E J (2005). 31 Concentrations o f environm ental chem icals associated w ith neurodevelopm ental effects in the U S population. 3 2 N eurotoxicology, 26: 531-545. 594295 33 N eedham L L ; Gerthoux P M ; Patterson Jr D G ; B ram billa P; Prikle JL ; Tram acere P L ; Turner W E ; Beretta c; 3 4 Sam pson E J; M o carelli P (1994). H alf-life o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in serum o f Seveso adults: interim 35 report. , 21: 81-85. 200030 3 6 N essel C S ; A m oruso M A ; U m breit T H ; M eeker R J; G allo M A (1992). Transpulm onary uptake and bioavailability 3 7 o f 2 ,3 ,7 ,8 -T C D D from respirable soil particles. Chem osphere, 25: 29-32. 548743 3 8 N ilsson C B ; H kansson H (2002). The retinoid signaling system - a target in dioxin toxicity. Crit R ev T o xicol, 32: 39 211-232. 548746 This document is a draftfo r review purposes only and does not constitute Agency policy. R -20 D R A F T -- D O N O T C IT E O R Q U O T E 1 N ishim ura N ; Yonem oto J; N ishim ura H ; Ikushiro S; Tohyam a C (2005). D isruption o f thyroid horm one 2 hom eostasis at w eaning o f H oltzm an rats by lactational but not in utero exposure to 2,3,7,8-tetrachlorodibenzo-p3 dioxin. T oxicol Sci, 85: 607-614. 197860 4 N iskar A ; N eedham L L ; R ubin C ; Turner W E ; M artin C A ; Patterson D G Jr; H astyL ; W ong L Y ; M arcus M (2009). 5 Serum dioxin, polychlorinated biphenyls, and endom etriosis: A case-control study in Atlanta. Chem osphere, 74: 6 944-949. 548802 7 N ohara K ; Fujim aki H ; Tsukum o S; U shio H ; M iyabara Y ; K ijim a M ; Tohyam a C ; Yonem oto J (2000). The effects 8 o f perinatal exposure to low doses o f 2,3,7,8-tetrachlorodibenzo-p-dioxin on im m une organs in rats. T o xico lo gy, 9 154: 123-133. 200027 10 N oh ara K ; Izu m i H ; Tam ura S; N agata R ; T ohyam a C (2002). E ffe ct o f low -dose 2,3,7,8-tetrachlorodibenzo-p11 dioxin (T C D D ) on influenza A virus-induced m ortality in m ice. T o xicology, 170: 131-138. 199021 12 N o lan K J; Sm ith F A ; H efn er J G (1979). E lim ination and tissue distribution o f 2,3,7,8-tetrachlorodibenzo-p-dioxin 13 (T C D D ) in fem ale guinea pigs follow in g a single oral dose. T oxicol A ppl Pharm acol, 48: 162. 543785 14 N R C (1983). R isk assessm ent in the federal governm ent: M anaging the process. N ational A cadem y Press. 15 W ashington, D C . 194806 16 N R C (1989). Im proving risk com m unication. W ashington, D C : N ational A cadem y Press. 000858 17 N R C (1991). H um an exposure assessm ent for airborne pollutants: advances and opportunities. W ashington, D C : 18 N ational Academ ies Press. 037823 19 N R C (1993). Issues in risk assessm ent. Com m ittee on R isk Assessm ent M ethodology, N ational Research Council. 2 0 W ashin gton, D C .http://w w w .nap.edu/catalog.php?record_id=2078. 078637 21 N R C (1994). Science and judgm ent in risk assessm ent. N ational R esearch Council; N ational A cadem y Press. 2 2 W ashington, D C . 006424 23 N R C (2002). Estim ating the public health benefits o f proposed air pollution regulations. W ashington, D C : N ational 2 4 A cadem y o f Sciences. 035312 2 5 N R C (2007). Scientific review o f the proposed risk assessm ent bulletin from the O ffice o f M anagem ent and Budget. 2 6 N ation al R esearch C o u n cil. W ashington, D C .http://w w w .nap.edu/catalog.php?record_id=11811. 543748 2 7 N R C (N ational Research Council) (2009). Science and decisions: advancing risk assessm ent. N ational A cadem y 2 8 Press. W ashington, D C . 194810 2 9 N T P (1982). Carcinogenesis bioassay o f B IS(2-chloro-1-m ethylethyl) ether ( 70% ) (C A S no. 108-60-1) containing 3 0 2-chloro-1-m ethylethyl(2-chloropropyl) ether ( 30% ) (C A S no. 83270-31-9) in B 6 C 3 F1 m ice (gavage study). 31 N ational T oxicology Program . Research Triangle Park, N C and Bethesda, M D . N TP-81-55. 200870 3 2 N T P (1982). N T P T echnical Report on carcinogenesis bioassay o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in O sborne33 M endel rats and B 6 C 3 F1 m ice (gavage study). Public H ealth Service, U .S . Departm ent o f H ealth and H um an 3 4 Services, N ational Toxicology Program . Research Triangle Park, N C . 543764 3 5 N T P (1982). N T P T echnical Report on carcinogenesis bioassay o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in O sborne3 6 M endel rats and B 6 C 3 F1 m ice (gavage study). Public H ealth Service, U .S . Departm ent o f H ealth and H um an 3 7 Services; N T P T R 209. N IE H S . Research Triangle Park, N C . 594255 3 8 N T P (2006). N T P tech nical report on the to xico lo gy and carcinogenesis studies o f 2,3,7,8-tetrachlorodibenzo-p3 9 dioxin (T C D D ) in fem ale harlan Sprague-D aw ley rats. N ational T oxicology Program . R T P , N C . 06-4468. 197605 This document is a draftfo r review purposes only and does not constitute Agency policy. R -21 D R A F T -- D O N O T C IT E O R Q U O T E 1 N T P (2006). T o xicolo gy and carcinogenesis studies o f a m ixture o f 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) 2 (C A S N o . 1746-01-6), 2,3,4,7,8-pentachlorodibenzofuran (P eC D F ) (C A S N o . 57117-31-4), and 3,3',4,4',5 3 pentachlorobiphenyl (P C B 126) (C A S N o . 57465-28-8) in fem ale H arlan Sprague-D aw ley rats (gavage studies). 4 Public H ealth Service, U .S . Departm ent o f H ealth and H um an Services, tional T oxicology Program . Research 5 T riangle Park, N C .http://n tp.niehs.n ih.gov/in dex.cfm ?objectid=070B 7300-0E 62-B F12-F4C3E3B 5B 645A 92B . 6 543749 7 O ehlert G W (1992). A note on the delta m ethod. A m Stat, 46: 27-29. 543742 8 O hsako S; M iyabara Y ; N ishim ura N ; K urosaw a S; Sakaue M ; Ishim ura R ; Sato M ; Takeda K ; A ok i Y ; Sone H ; 9 T ohyam a C ; Yonem oto J (2001). M aternal exposure to a low dose o f 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) 10 suppressed the developm ent o f reproductive organs o f m ale rats: D ose-dependent increase o f m R N A levels o f 5a11 reductase type 2 in contrast to descrease o f androgen receptor in the pubertal ventral prostate. T o xicol Sci, 60: 132 12 1 4 3 .198497 13 O k ey A B ; R id d ick D S ; H arper P A (1994). The A h receptor: M ediator o f the toxicity o f 2,3,7,8-tetrachlorodibenzo14 p -dioxin (T C D D ) and related com pounds. T o xico l Lett, 70: 1-22. 548759 15 O lso n JR ; H olscher M A ; N eal R A (1980). T oxicity o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in the G old en Syrian 16 ham ster . T oxicol A ppl Pharm acol, 55: 67-78. 197976 17 O lson JR ; M cG arrigle B P ; G igliotti P J; K um ar S; M cR eyn olds JH (1994). H epatic uptake and m etabolism o f 18 2 ,3,7,8-tetrach lorodiben zo-p-dioxin and 2,3,7,8-tetrachlorodibenzofuran. F u n d am A p p l T o x ico l, 22: 631-640. 19 198008 2 0 O tt M G ; M esserer P ; Z o b er A (1993). A ssessm ent o f past occupational exposure to 2,3,7,8-tetrachlorodibenzo-p21 dioxin using blood lipid analyses. Int A rch O ccu p Environ H ealth, 65: 1-8. 594322 2 2 O tt M G ; O lson R A ; C o o k R R ; B on d G G (1987). Cohort m ortality study o f chem ical w orkers w ith potential 23 exposure to the higher chlorinated dioxins. J O ccu p Environ M ed , 29: 422-429. 064994 2 4 O tt M G ; Zober A (1996). Cause specific m ortality and cancer incidence am ong em ployees exposed to 2 ,3 ,7 ,8 2 5 T C D D after a 1953 reactor accident. O ccup Environ M ed , 53: 606-612. 198408 2 6 O tt M G ; Zober A (1996). M orbidity study o f extruder personnel w ith potential exposure to brom inated dioxins and 2 7 furans. II. Results o f clinical laboratory studies. O ccu p Environ M ed , 53: 844-846. 198101 2 8 Ppke O ; B all M ; L is A (1994). P C D D /P C D F in hum ans, a 1993-update o f background data. Chem osphere, 29: 2 9 2355-2360. 198279 3 0 Pekelis M ; N ico lich M J; Gauthier JS (2003). Probabilistic fram ew ork for the estim ation o f the adult and child 31 toxicokinetic intraspecies uncertainty factors. R isk A n al, 23: 1239-1255. 548723 3 2 Percy C ; Stanek E III; G loeckler L (1981). A ccuracy o f cancer death certificates and its effect on cancer m ortality 33 statistics. A m J Public H ealth, 71: 242-250. 004891 3 4 Pereg D ; D ew ailly ; Poirier G G ; A yotte P (2002). Environm ental exposure to polychlorinated biphenyls and 35 placental C Y P 1 A 1 activity in Inuit w om en from northern Q ubec. Environ H ealth Perspect, 110: 607-612. 199797 3 6 Pesatori A C ; Consonni D ; Bachetti S; Zocchetti C ; B onzini M ; B accarelli A ; Bertazzi P A (2003). Short- and long3 7 term m orbidity and m ortality in the population exposed to dioxin after the "Seveso accident". Ind H ealth, 41: 127 38 138. 197001 3 9 Pesatori A C ; Zocchetti C ; G uercilena S; Consonni D ; Turrini D ; Bertazzi P A (1998). D io xin exposure and non4 0 m alignant health effects: A m ortality study. O ccu p Environ M ed , 55: 126-131. 523076 This document is a draftfo r review purposes only and does not constitute Agency policy. R -22 D R A F T -- D O N O T C IT E O R Q U O T E 1 Piacitelli L A ; Sw eeney M H ; Fingerhut M A ; Patterson D G ; Turner W E ; Connally L B ; W ille K K ; Tom pkins B 2 (1992). Serum levels o f P C D D S and P C D F S am ong w orkers exposed to 2 ,3 ,7 ,8 -T C D D contam inated chem icals. 3 Chem osphere, 25: 251-254. 197275 4 Pipe N G ; Sm ith T ; H alliday D ; Edm onds C J; W illiam s C ; Coltart T M (1979). Changes in fat, fat-free m ass and body 5 w ater in hum an norm al pregnancy. B r J Obstet G ynaecol, 86: 929-940. 548786 6 Pirkle JL ; W olfe W H ; Patterson D G ; N eedham L L ; M ichalek JE ; M iner JC ; Peterson M R ; Phillips D L (1989). 7 Estim ates o f the h alf-life o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in V ietn am Veterans o f O peration R an ch H and. J 8 T oxicol Environ H ealth, 27: 165-171. 197861 9 Pitot H ; Goldsw orthy T ; Cam pbell H ; Poland A (1980). Quantitative evaluation o f the prom otion by 2 ,3 ,7 ,8 10 tetrachlorodibenzo-p-dioxin o f hepatocarcinogenesis from diethylnitrosam ine. Can cer R es, 40: 3616-3620. 197885 11 Pohjanvirta R ; Tuom isto J (1994). Short-term toxicity o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in laboratory anim als: 12 E ffects, m echanism s, and anim al m odels. Pharm acol R ev, 46: 483-549. 543767 13 Pohjanvirta R ; Tuom isto L ; Tuom isto J (1989). The central nervous system m ay be involved in T C D D toxicity. 14 T oxicology, 58: 167-174. 548766 15 Poiger H ; Schlatter C (1986). Pharm acokinetics o f 2 ,3 ,7 ,8 -T C D D in m an. Chem osphere, 15: 1489-1494. 197336 16 P oland A ; G lo v er E (1980). 2,3,7,8-tetrachlorodibenzo-p-dioxin: segregation o f toxicity w ith the A h locus. M o l 17 Pharm acol, 17: 86-94. 543761 18 Poland A ; G lover E (1990). Characterization and strain distribution pattern o f the m urine A h receptor specified by 19 the A hd and Ahb-3 alleles. M o l Pharm acol, 38: 306-312. 543759 2 0 Poland A ; Palen D ; G lover E (1982). Tum our prom otion by T C D D in skin o f H R S /J hairless m ice. N ature, 300: 21 271-273. 199756 2 2 Poland A ; Palen D ; G lover E (1994). A nalysis o f the four alleles o f the m urine aryl hydrocarbon receptor. M o l 23 Pharm acol, 46: 915-921. 198439 2 4 Popp JA ; Crouch E ; M cC on n ell E E (2006). A W eight-of-evidence analysis o f the cancer dose-response 2 5 characteristics o f 2,3,7,8-tetrachlorodibenzodioxin (T C D D ). T o xico l S ci, 89: 361-369. 197074 2 6 Potter C L ; M oore R W ; Inhorn SL ; H agen T C ; Peterson R E (1986). Thyroid status and therm ogenesis in rats treated 2 7 w ith 2,3,7,8-tetrachlorodibenzo-p-dioxin. T o xico l A p p l P harm acol, 84: 45-55. 548771 2 8 Potter C L ; Sipes IG ; R u ssell D H (1983). H ypothyroxinem ia and hypotherm ia in rats in response to 2 ,3 ,7 ,8 2 9 tetrachlorodibenzo-p-dioxin adm inistration. T o xicol A p p l Pharm acol, 69: 89-95. 548769 3 0 P ou lin P ; T h eil F P (2001). Prediction o f pharm icokinetics prior to in v ivo studies. 1. m echanism -based prediction o f 31 volum e o f distribution. J Pharm Sci, 91: 129-156. 594269 3 2 Puga A ; N ebert D W ; Carier F (1992). D io xin induces expression o f c-fos and c-jun proto-oncogenes and a large 33 increases in transcription factor A P -1 . T oxicol A p p l Pharm acol, 55: 67-78. 543784 3 4 R am adoss P; Perdew G H (2004). U se o f 2-azido-3-[125I]iodo-7,8-dibrom odibenzo-p-dioxin as a probe to determ ine 35 the relative ligand affinity o f hum an versus m ouse aryl hydrocarbon receptor in cultured cells. M o l Pharm acol, 66: 3 6 129-136. 198824 This document is a draftfor review purposes only and does not constitute Agency policy. R-23 DRAFT--DO NOT CITE OR QUOTE 1 R am sey JC ; H efn er JG ; K arbow ski R J; B raun W H ; G eh rin g P J (1982). The in vivo biotransform ation o f 2 ,3 ,7 ,8 2 tetrachlorodibenzo-p-dioxin (T C D D ) in the rat. T o xico l A p p l Pharm acol, 65: 180-184. 548750 3 R ao M S ; Subbarao V ; Prasad JD ; Scarpelli D G (1988). Carcinogenicity o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in 4 the Syrian golden ham ster. Carcinogenesis, 6: 1677-1679. 199032 5 Reddy M ; Y a n g R ; Clew ell H J; A ndersen M E (2005). Physiologically based pharm acokinetic m odeling: Science 6 and applications. H oboken, N ew Jersey: John W iley & Sons. 594251 7 R evich B ; A ksel E ; U shakova T ; Ivanova I; Z h uchenko N ; K lyu ev N ; Brodsky B ; Sotskov Y (2001). D io xin 8 exposure and public health in Chapaevsk, Russia. Chem osphere, 43: 951-966. 199843 9 R evich B ; Sergeyev O ; Zeilert V ; H auser R (2005). Chapaevsk, Russia: 40 years o f dioxins exposure on the hum an 10 health and 10 years o f R ussian ? U S A epidem iological studies. Presented at A lm aty 2005, A lm aty, K azakhstan. 11 198777 12 R ier SE ; Coe C L ; Lem ieux A M ; M artin D C ; M orris R ; Lu cier G W ; Clark G C (2001). Increased tum or necrosis 13 factor-alpha production by peripheral blood leukocytes from T C D D -exp o sed rhesus m onkeys. T o xicol Sci, 60: 327 14 3 3 7 .543773 15 R ier SE ; M artin D C ; B ow m an R E ; B ecker JL (1995). Im m unoresponsiveness in endom etriosis: Im plications o f 16 estrogenic toxicants. Environ H ealth Perspect, 103: 151-156. 198566 17 R ier SE ; M artin D C ; B ow m an R E ; D m ow ski W P; B ecker JL (1993). Endom etriosis in Rhesus M onkeys (M acaca 18 m ulatta) F o llo w in g C hron ic Exposure to 2,3,7,8-T etrachlorodibenzo-p-dioxin . Fundam A p p l T o xico l, 21: 433-441. 19 199987 2 0 R ier SE ; Turner W E ; M artin D C ; M orris R ; Lucier G W ; Clark G C (2001). Serum levels o f T C D D and dioxin-like 21 chem icals in Rhesus m onkeys chronically exposed to dioxin: Correlation o f increased serum P C B levels w ith 2 2 endom etriosis. T o xicol Sci, 59: 147-159. 198776 23 Roberts E A ; G olas C L ; O key A B (1986). A h receptor m ediating induction o f aryl hydrocarbon hydroxylase: 2 4 D etection in hum an lu ng by binding o f 2,3,7,8-[H ]tetrachlorodibenzo-p-dioxin. Can cer R es, 46: 3739-3743. 198780 2 5 Roberts E A ; Shear N H ; O k ey A B ; M anchester D K (1985). The A h receptor and d ioxin toxicity: From rodent to 2 6 hum an tissues . Chem osphere, 14: 661-674. 198706 2 7 Rohde S; M oser G A ; Papke O ; M cLach lan M S (1999). Clearance o f P C D D /Fs v ia the gastrointestinal tract in 2 8 occupationally exposed persons. Chem osphere, 38: 3397-3410. 548764 2 9 R oth W L ; Ernst S; W eber L W D ; K erescen L ; R ozm an K K (1994). A pharm acodynam ically responsive m odel o f 3 0 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) transfer betw een liver and fat at low and high doses. T o xico l A p p l 31 Pharm acol, 127: 151-162. 198063 3 2 Rothm an K J (1986). M odern epidem iology. 046091 33 R o y T ; H am m erstrom K ; Schaum J (2008). Percutaneous absorption o f 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) 3 4 from soil. J T oxicol Environ H ealth A Curr Iss, 71: 1509-1515. 548747 3 5 R o zm an K K (2000). The role o f tim e in toxicology or H aber's c x t product. T o xico lo gy, 149: 35-42. 548758 3 6 R yan JJ; A m irova Z ; Carrier G (2002). Sex Ratios o f Children o f Russian Pesticide Producers Exposed to D ioxin . 3 7 Environ H ealth Perspect, 110: A699-A701. 198508 This document is a draftfor review purposes only and does not constitute Agency policy. R-24 DRAFT--DO NOT CITE OR QUOTE 1 R yan JJ; Schecter A (2000). Exposure o f Russian phenoxy herbicide producers to dioxin. J O ccu p Environ M ed , 42: 2 861-870. 594412 3 Saltelli A ; C han K ; Scott E M (2000). Sensitivity analysis. England: John W iley & Sons Ltd. 543756 4 Sandau C D ; A yotte P; D ew ailly E ; D u ffe J; Norstrom R J (2002). Pentachlorophenol and hydroxylated 5 polychlorinated biphenyl m etabolites in um bilical cord plasm a o f neonates from coastal populations in Q ubec. 6 Environ H ealth Perspect, 110: 411-417. 594406 7 Santostefano M J; Johnson K L ; W hisnant N A ; Richardson V M ; D evito M J; Birnbaum L S (1996). Subcellular 8 localization o f T C D D differs betw een the liver, lungs, and kidneys after acute and subchronic exposure: 9 Species/dose com parison and possible m echanism . Fundam A p p l T oxicol, 34: 365-375. 594258 10 Santostefano M J; W ang X ; Richardson V M ; R oss D G ; D eV ito M J; Birnbaum L F (1998). A pharm acodynam ic 11 analysis o f T C D D -In d u ced Cytochrom e 450 gene expression in m ultiple tissues: D ose and tim e-dependent effects. 12 T oxicol A ppl Pharm acol, 151: 294-310. 200001 13 Saracci R ; K ogevinas M ; Bertazzi P A ; Bueno de M esquita B H ; Coggon D ; Green L M ; Kauppinen T; L'A bb K A ; 14 Littorin M ; Lyn ge E ; M athew s JD ; N euberger M ; O sm an J; Pearce N ; W inkelm ann R (1991). Can cer m ortality in 15 w orkers exposed to chlorophenoxy herbicides and chlorophenols. Lancet, 338():: 1027-1032. 199190 16 Sauer R M (1990). 2,3,7,8-T etrachlorodibenzo-p-dioxin in sprague-daw ley rats. P A T H C O , IN C . M arylan d. 198829 17 Schantz S L ; B o w m an R E (1989). Learn in g in m onkeys exposed perinatally to 2,3,7,8-tetrachlorodibenzo-p-dioxin 18 (T C D D ). N eurotoxicol Teratol, 11: 13-19. 198104 19 Schantz SL ; Laughlin N K ; V an Valkenberg H C ; B ow m an R E (1986). M aternal care by rhesus m onkeys o f infant 2 0 m onkeys exposed to either lead or 2,3,7,8-tetrachlorodibenzo-P-dioxin. N eu rotoxicology, 7: 637-650. 088206 21 Schantz SL ; Seo B W ; M oshtaghian J; Peterson R E ; M oore R W (1996). E ffects o f gestational and lactational 2 2 exposure to T C D D or coplanar P C B s on spatial learning. N eurotoxicol Teratol, 18: 305-313. 198781 23 Schecter A ; Cram er P; B oggess K ; Stanley J; O lson JR (1997). Levels o f D ioxin s, D ibenzofurans, P C B and D D E 2 4 congeners in pooled food sam ples collected in 1995 at superm arkets across the U nited States. Chem osphere, 34: 25 1437-1447. 198396 2 6 Schw artz M ; A p p el K E (2005). Carcinogenic risks o f dioxin: m echanistic considerations. R egu l T o xicol Pharm acol, 2 7 43: 19-34. 543737 2 8 Seidel SD ; W inters G M ; Rogers W J; Ziccardi M H ; L i V ; K eser B ; D enison M S (2001). A ctivation o f the A h 2 9 receptor signaling pathw ay by prostaglandins. J B iochem M o l T o xicol, 15: 187-196. 543776 3 0 S e lf S G ; L ia n g K Y (1987). A sym p totic properties o f m axim um likelihood estim ators and likelihood ratio tests under 31 nonstandard conditions. J A m Stat A ssoc, 82: 605-610. 594398 3 2 Seo B W ; L i M H ; H ansen L G ; M oore R W ; Peterson R E ; Schantz S L (1995). E ffects o f gestational and lactational 33 exposure to coplanar polychlorinated biphenyl (P C B ) congeners or 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) on 3 4 thyroid horm one concentrations in w eanling rats. T o xicol Lett, 78: 253-262. 197869 35 Sew all C ; Lucier G ; Tritscher A ; Clark G (1993). T CD D -m ediated changes in hepatic epiderm al grow th factor 3 6 receptor m ay be a critical event in the hepatocarcinogenic action o f T C D D . Carcinogenesis, 14: 1885-1893. 197889 3 7 Sew all C H ; Flagler N ; V anden H euvel JP ; Clark G C ; Tritscher A M ; M aronpot R M ; Lucier G W (1995). Alterations 3 8 in thyroid fun ction in fem ale Sp rague-D aw ley rats fo llo w in g chronic treatm ent w ith 2,3,7,8-tetrachlorodibenzo-p3 9 dioxin. T oxicol A ppl Pharm acol, 132: 237-244. 198145 This document is a draftfo r review purposes only and does not constitute Agency policy. R -25 D R A F T -- D O N O T C IT E O R Q U O T E 1 Shi Z ; V ald ez K E ; T in g A Y ; Franczak A ; G u m S L ; P etroff B K (2007). O varian endocrine disruption underlies 2 prem ature reproductive senescence follow in g environm entally relevant chronic exposure to the aryl hydrocarbon 3 receptor agonist 2,3,7,8-tetrachlorodibenzo-p-dioxin. B io l R eprod, 76: 198-202. 198147 4 Shu H ; Teitelbaum P; W ebb A S ; M arple L ; B runck B ; D ei R ossi D ; M urray F J; Paustenbach D (1988). 5 B ioavailability o f soil-bound T C D D : D erm al bioavailability in the rat. Fundam A p p l T o xico l, 2: 335-343. 548739 6 Siem iatycki J; W acholder S; D ew ar R ; Cardis E ; Greenw ood C ; Richardson L (1988). D egree o f confounding bias 7 related to sm oking, ethnic group, and socioeconom ic status in estim ates o f the associations betw een occupation and 8 cancer. J O ccup M ed, 30: 617-625. 198556 9 Sikov M (1970). Radiation biology o f the fetal and juvenile m am m al. Science, 167: 1640-1641. 594274 10 Sim anainen U ; H aavisto T; Tuom isto JT ; Paranko J; Toppari J; Tuom isto J; Peterson R E ; V iluksela M (2004). 11 Pattern o f m ale reproductive system effects after in utero and lactational 2,3,7,8-tetrachlorodibenzo-p-dioxin 12 (T C D D ) exposure in three differentially T C D D -sen sitive rat lines Pattern o f m ale reproductive system effects after 13 in utero and lactational 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) exposure in three differentially T C D D -sen sitive 14 rat lines. T o xicol Sci, 80: 101-108. 198948 15 Sim anainen U ; Tuom isto JT ; Pohjanvirta R ; Syrjala P; Tuom isto J; V ilu ksela M (2004). Postnatal developm ent o f 16 resistance to short-term high-dose toxic effects o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in T C D D -resistan t and 17 sem iresistant rats. T o xicol A p p l Pharm acol, 196: 11-19. 198106 18 Sim anainen U ; Tuom isto JT ; Tuom isto J; V ilu ksela M (2002). Structure-Activity relationships and dose responses 19 ofPolychlorinated D ibenzo-p-dioxins for short-term effects in 2,3,7,8-Tetrachlorodibenzo-p-dioxin-R esistant and 2 0 sensitive rat strains. T o xico l A p p l Pharm acol, 181: 38-47. 201369 21 Sim anainen U ; Tuom isto JT ; Tuom isto J; V iluksela M (2003). D ose-response analysis o f short-term effects o f 2 2 2,3,7,8-tetrachlorodibenzo-p-dioxin in three differentially susceptible rat lines. , 187: 128-136. 198582 23 Sim on T ; A ylw ard L L ; K irm an C R ; R ow lands JC ; B udinsky R A (2009). Estim ates o f cancer potency o f 2 ,3 ,7,8 2 4 tetrachlorodibenzo(p)dioxin using linear and non-linear dose-response m odeling and toxicokinetics. T oxicol Sci, 25 112: 490-506. 594321 2 6 Slezak B P ; H atch G E ; D eV ito M J; Diliberto JJ; Slade R ; Crissm an K ; H assoun E ; Birnbaum L S (2000). O xidative 2 7 stress in fem ale B 6 C 3 F 1 m ice fo llo w in g acute and subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin 2 8 (T C D D ). T oxicol Sci, 54: 390-398. 199022 2 9 Slob W ; Pieters M N (1998). A probabilistic approach for deriving acceptable hum an intake lim its and hum an health 3 0 risks from toxico logical studies: general fram ew ork. R isk A n al, 18: 787-798. 087256 31 Sm art J; D aly A (2000). V ariation in induced C Y P 1 A 1 levels: Relationship to C Y P 1 A 1 , A h receptor, and G ST M 1 3 2 polym orphism s. Pharm acogenetics, 10: 11-24. 548794 33 Sm ialow icz R J; B u rgin D E ; W illiam s W C ; D iliberto JJ; Setzer R W ; B irnbaum L S (2004). C Y P 1 A 2 is not required 3 4 fo r 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced im m unosuppression. T o xico lo gy , 197: 15-22. 110937 35 Sm ialow icz R J; D eV ito M J; W illiam s W C ; Birnbaum L S (2008). Relative potency based on hepatic enzym e 3 6 induction predicts im m unosuppressive effects o f a m ixture o f P C D D S / P C D F S and P C B S . T o xicol A p p l Pharm acol, 37 227: 477-484. 198341 3 8 Sm ith A H ; Fisher D O ; Pearce N ; Chapm an C J (1982). Congenital defects and m iscarriages am ong N ew Zealand 2, 3 9 4, 5-T sprayers. A rch Environ H ealth, 37: 197-200. 198586 This document is a draftfor review purposes only and does not constitute Agency policy. R-26 DRAFT--DO NOT CITE OR QUOTE 1 Sm ith A H ; Lopipero P (2001). Invited com m entary: how do the Seveso findings affect conclusions concerning 2 T C D D as a hum an carcinogen? A m J Epidem iol, 153: 1045-1047. 198585 3 Sp iegelh alter D ; Thom as A ; B est N ; G ilk s W (2003). B U G S 0.5 B ay esian inference u sin g G ib b s sam plin g m anual, 4 version ii. M R C B iostatistics U n its, Institute o f Public H ealth, Cam bridge. 594261 5 Squire R A (1980). P athologic evaluations o f selected tissues from the D o w Chem ical T C D D and 2 ,4 ,5 -T rat studies. 6 U .S . Environm ental Protection A gency. W ashington D C . 594272 7 Squire R A (1990). P athologic evaluations o f selected tissues from the D o w Chem ical T C D D and 2 ,4 ,5 -T rat studies. 8 Subm itted to Carcinogen A ssessm ent G roup, U .S . Environm ental Protection A gen cy. W ashington, D C . 548781 9 Starr T B (2003). Significant issues raised by m eta-analyses o f cancer m ortality and dioxin exposure. Environ H ealth 10 Perspect, 111: 1443-1447. 594271 11 Staskal D F ; D iliberto JJ; D eV ito M J; Birnbaum L S (2005). Inhibition o f hum an and rat C Y P 1 A 2 by T C D D and 12 dioxin-like chem icals. T oxicol Sci, 84: 225-231. 198276 13 Stayner L ; B ailer A J; Sm ith R ; G ilbert S; R ice F; K uem pel E (1999). Sources o f uncertainty in dose-response 14 m odeling o f epidem iological data for cancer risk assessm ent. A n n N Y A cad Sci, 895: 212-222. 198654 15 Stayner L ; Steenland K ; D osem eci M ; H ertz-Picciotto I (2003). A ttenuation o f exposure-response curves in 16 occupational cohort studies at high exposure levels. Scand J W ork Environ H ealth, 29: 317-324. 054922 17 Steenland K ; Calvert G ; K etchum N ; M ichalek J (2001). D io xin and diabetes m ellitus: an analysis o f the com bined 18 N IO S H and R anch H and data. O ccu p E nviron M ed , 58: 641-648. 198589 19 Steenland K ; D eddens J (2003). D ioxin : Exposure-response analyses and risk assessm ent. Ind H ealth, 41: 175-180. 2 0 198587 21 Steenland K ; D eddens J; P iacitelli L (2001). R isk assessm ent for 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) based 2 2 on an epidem iologic study. A m J Epidem iol, 154: 451-458. 197433 23 Steenland K ; Piacitelli L ; D eddens J; Fingerhut M ; C hang L I (1999). Cancer, heart disease, and diabetes in w orkers 2 4 exposed to 2,3,7,8-tetrachlorodibenzo-p-dioxin. J N atl C an cer Inst, 91: 779-786. 197437 2 5 Stellm an SD ; Stellm an JM (1986). Estim ation o f exposure to A gent O range and other defoliants am ong A m erican 2 6 troops in Vietnam : a m ethodological approach. A m J Ind M ed, 9: 305-321. 594380 2 7 Stephenson R P (1956). A m odification o f receptor theory. B r J Pharm acol, 11: 379-393. 594280 2 8 Su gita-K onish i Y ; K obayashi K ; N aito H ; M iu ra K ; Suzuki Y (2003). E ffect o f lactational exposure to 2 ,3 ,7 ,8 2 9 tetrachlorodibenzo-p-dioxin on the susceptibility to Listeria infection. B iosci Biotechnol B iochem , 67: 89-93. 30 198375 31 Sw artout JC ; Price P S; D ourson M L ; Carlson-Lynch H L ; K eenan R E (1998). A probabilistic fram ew ork for the 3 2 reference dose (probabilistic R fD ). R isk A n al, 18: 271-282. 093460 3 3 t' M annetje A ; M c L e a n D ; C h e n g S; B o ffetta P ; C o lin D ; Pearce N (2005). M ortality in N ew Zealan d w orkers 3 4 exposed to phenoxy herbicides and dioxins. O ccu p Environ M ed , 62: 34-40. 197593 35 Takem oto K ; N akajim a M ; Fujiki Y ; K atoh M ; G onzalez F J; Y o k o i T (2004). R ole o f the aryl hydrocarbon receptor 3 6 and Cyp1b1 in the antiestrogenic activity o f 2,3,7,8-tetrachlorodibenzo-p-dioxin. A rch T o xico l, 78: 309-315. 37 543753 This document is a draftfor review purposes only and does not constitute Agency policy. R-27 DRAFT--DO NOT CITE OR QUOTE 1 Teeguarden JG ;, D ragan Y P ; Sin gh J; V aughan J; X u Y H ; Goldsw orthy T ; H C Pitot H C (1999). Quantitative 2 analysis o f dose- and tim e-dependent prom otion o f four phenotypes o f altered hepatic fo ci by 2 ,3 ,7 ,8 3 tetrachlorodibenzo-p- dioxin in fem ale Sprague-D aw ley rats. T o xicol Sci, 51: 211-223. 198274 4 Thiess A M ; Frentzel-B eym e R (1977). M ortality study o f persons exposed to dioxin follow in g an accident w hich 5 occurred in the B A S F on 17 N ovem ber 1953. Presented at Proceedings o f the 5th International Conference 6 M edichem , 1977, San Francisco, C A . 594302 7 Thiess A M ; Frentzel-B eym e R ; L in k R (1982). M ortality study o f persons exposed to dioxin in a trichlorophenol8 process accident that occurred in the B A S F A G on N ovem ber 17, 1953. A m J Ind M ed , 3: 179-189. 064999 9 T ian Y ; K e S; D enison M S ; R abson A B ; G allo M A (1999). A h Receptor and N F -k B Interactions, a Potential 10 M echanism for D io xin Toxicity. J B iol Chem , 274: 510-515. 198378 11 Toide K ; Yam azaki JH ; N agashim a R ; Itoh K ; Iw ano S; Takahashi Y ; W atanabe S; K am ataki T (2003). A ryl 12 hydrocarbon hydroxylase represents C Y P 1 B 1 and not C Y P 1 A 1 , in hum an freshly isolated w hite cells: Trim odal 13 distribution o f Japanese population according to induction o f C Y P 1 B 1 m R N A by environm ental dioxins. Cancer 14 Epidem iol Biom arkers Prev, 12: 219-222. 548792 15 Toth K ; So m fai-R elle S; Sugar J; B en ce J (1979). Carcinogenicity testing o f herbicide 2 ,4 ,5 16 trichlorophenoxyethanol containing dioxin and o f pure dioxin in Sw iss m ice. Nature, 278: 548-549. 197109 17 Tritscher A M ; M ahler J; Portier C J; Lu cier G W ; W alker N J (2000). Induction o f lung lesions in fem ale rats 18 fo llo w in g chronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. T o xico l P athol, 28: 761-769. 197265 19 Tuom isto JT ; V ilu ksela M ; Pohjanvirta R ; Tuom isto J (1999). The A H receptor and a novel gene determ ine acute 2 0 toxic responses to T C D D : segregation o f the resistant alleles to different rat lines. T o xico l A p p l Pharm acol, 155: 71 21 81. 548717 2 2 Tuom isto JT ; W ilson A M ; Evans JS ; Tainio M (2008). U ncertainty in m ortality response to airborne fine particulate 23 matter: com bining European air pollution experts. Reliab E n g Syst Saf, 93: 732-744. 548715 2 4 U .S . D O E (1992). D O E standard: H azard categorization, and accident analysis techniques for com pliance w ith D O E 2 5 O rder 5480.23, nuclear safety analysis reports. U .S . D epartm ent o f E nergy. W ashin gton, D C . D O E -S T D -1 0 2 7 -9 2 . 2 6 http://w w w .hss.energy.gov/nuclearsafety/ns/techstds/standard/std1027/s1027cn1.pdf. 543733 2 7 U .S . E P A (1994). M ethods for derivation o f inhalation reference concentrations and application o f inhalation 2 8 dosim etry. Environm ental Criteria and Assessm ent O ffice, O ffice o f H ealth and Environm ental Assessm ent, O ffice 2 9 o f Research and D evelopm ent, U .S . Environm ental Protection A gency. Research Triangle Park, N C . EPA/600/83 0 90/066F. http://cfpub.epa.gov/ncea/cfm /recordisplay.cfm ?deid=71993. 006488 31 U .S . E P A (1996). Colum bus w aste-to-energy m unicipal incinerator D io xin soil sam pling project. U .S . E P A 32 R E G IO N 5. Chicago, IL . 905R96018. 33 http ://n ep is.ep a.gov/E xe/Z yN E T .exe/2 0 0 0 P C X X .T X T ?Z yA ctio n D = Z yD o cu m en t& C lien t= E P A & In dex= 1 9 9 5 + T h ru 3 4 +1999& D ocs=& Q uery=colum bus+w aste-to35 energy+m unicipal+incinerator& Tim e=& EndTim e=& SearchM ethod=3& TocRestrict=n& Toc=& TocEntry=& Q Field 3 6 =pubnum ber% 5E% 22905R96018% 22& Q FieldYear=& Q FieldM onth=& Q FieldD ay=& U seQ Field=pubnum ber& Int 37 Q FieldO p=1& ExtQ FieldO p=1& Xm lQ uery=& File=D % 3A % 5Czyfiles% 5CIndex% 20D ata% 5C95thru99% 5CTxt% 5 3 8 C 0 00 0 0 0 1 7 % 5 C 2 0 0 0 P C X X .txt& U ser= A N O N Y M O U S& P assw ord = an o n ym o u s& SortM eth o d = h % 7 C 3 9 & M axim um D ocum ents=10& FuzzyD egree=0& Im ageQ uality=r75g8/r75g8/x150y150g16/i425& D isplay=p% 7Cf& D 4 0 efSeekPage=x& SearchB ack=ZyActionL& B ack=ZyActionS& B ackD esc=Results% 20page& M axim um Pages=1& Zy 41 Entry=1& SeekPage=x. 198087 4 2 U .S . E P A (1996). Proposed guidelines for carcinogen risk assessm ent. R isk Assessm ent Forum . U .S . Environm ental 43 Protection A gency. W ashington, D .C .. 594399 This document is a draftfo r review purposes only and does not constitute Agency policy. R -28 D R A F T -- D O N O T C IT E O R Q U O T E 1 U .S . E P A (1998). G uidelines for neurotoxicity risk assessm ent. Federal R egister 63(93):26926-26954. N ational 2 Center for Environm ental Assessm ent; O ffice o f Research and D evelopm ent; U .S . Environm ental Protection 3 A gency. W ashington, D C . EPA/630/R-95/001Fa. 4 http://oaspub.epa.gov/eim s/eim scom m .getfile?p_dow nload_id=4555.030021 5 U .S . E P A (2000). Benchm ark dose technical guidance docum ent [external review draft]. R isk Assessm ent Forum , 6 U .S . Environm ental Protection A gency. W ashington, D C . EPA/630/R-00/001. 7 http://w w w .epa.gov/raf/publications/benchm ark-dose-doc-draft.htm . 052150 8 U .S . E P A (2003). Exposure and hum an health reassessm ent o f 2 ,3,7,8 tetrachlorodibenzo-p dioxin (T C D D ) and 9 related com pounds [N A S review draft]. U .S . Environm ental Protection A gen cy, N ational Center for Environm ental 10 A ssessm ent. W ashin gton, D C . EPA/600/P 00/001. http://w w w .epa.gov/nceaw w w 1/pdfs/dioxin/nas-review /. 537122 11 U .S . E P A (2005). Guidelines for carcinogen risk assessm ent, Fin al Report. R isk Assessm ent Forum , U .S . 12 Environm ental Protection A gency. W ashington, D C . EPA/630/P-03/001F. 13 http://cfpub.epa.gov/ncea/cfm /recordisplay.cfm ?deid=116283. 086237 14 U .S . E P A (2006). A ir quality criteria for lead, in 2 V o lum es. O ffice o f H ealth and Environm ental Assessm ent, 15 Environm ental Criteria and Assessm ent O ffice, O ffice o f Research and D evelopm ent, U .S . Environm ental 16 Protection A gency. Research Triangle Park, N C . EPA-600/R-5/144aF-bF. 090110 17 U .S . E P A (2006). A ir quality criteria for ozone and related photochem ical oxidants. E P A . D C . 088089 18 U .S . E P A (2006). Provisional Assessm ent o f Recent Studies on H ealth Effects o f Particulate M atter Exposure. U .S . 19 Environm ental Protection A gency. Research Triangle Park, N C . 157071 2 0 U .S . E P A (2008). 2,3,7,8 Tetrachlorodibenzo-p d ioxin (T C D D ) dose response studies: prelim inary literature search 21 results and request for additional studies. U .S . Environm ental Protection A gen cy. W ashington, D C . EPA/600/R2 2 08/119. 519261 23 U .S . E P A (2008). Fram ew ork for application o f the toxicity equivalence m ethodology for polychlorinated dioxins, 2 4 furans, and biphenyls in ecological risk assessm ent. U .S . Environm ental Protection A gen cy. W ashington, D C . 2 5 E PA /100/R 08/004. http://w w w .epa.gov/raf/teffram ew ork/index.htm . 543774 2 6 U .S . E P A (2009). Integrated risk inform ation system (IR IS). Retrieved 2 4 -JU N -0 9 , from 2 7 http://cfpub.epa.gov/ncea/iris/index.cfm . 192196 2 8 U .S . E P A (2009). Sum m ary o f U .S . E P A dioxin w orkshop: February 18-20, 2009. U .S . Environm ental Protection 2 9 A gency. N ational Center for Environm ental Assessm ent. Cincinnati, O H . EPA/600/R-09/027. 543757 3 0 U .S . E P A (2009). U sin g probabilistic m ethods to enhance the role o f risk analysis in decision-m aking w ith case 31 study exam ples. U .S . Environm ental Protection A gency. W ashington, D C . W ashington, D C . EPA/100/R-09/001. 32 522927 33 U .S . N R C (1975). R eactor safety study-an assessm ent o f accident risks in U .S . com m ercial nuclear pow er plants. 3 4 U .S . N uclear Regulatory Com m ission. R ockville, M D . N U R EG -75/014 (W A SH -1400). 3 5 http://w w w .nrc.gov/reading-rm /doc-collections/nuregs/staff/sr75-014/. 543729 3 6 U .S . N R C (1981). Fault tree handbook. U .S . N uclear Regulatory Com m ission. W ashington, D C . N U R E G -0 4 9 2 . 3 7 http://w w w .nrc.gov/reading-rm /doc-collections/nuregs/staff/sr0492/. 543730 3 8 U .S . N R C (1983). A guide to the perform ance o f probabilistic risk assessm ents for nuclear pow er plants. U .S . 3 9 N u clear R egulatory C om m ission . W ashin gton , D C . N U R E G / C R -2 3 0 0 . http://w w w .nrc.gov/reading-rm /doc4 0 collections/nuregs/contract/cr2300/. 543732 This document is a draftfor review purposes only and does not constitute Agency policy. R-29 DRAFT--DO NOT CITE OR QUOTE 1 U .S . N R C (1991). Severe accident risks: an assessm ent for five U .S . nuclear pow er plants. U .S . N uclear Regulatory 2 C om m ission . W ashin gton, D C . N U R E G -1 1 5 0 . http://w w w .nrc.gov/reading-rm /doc-collections/nuregs/staff/sr1150/. 3 543736 4 U m em ura T; K a i S; H asgaw a R ; Sai K ; K urokaw a Y ; W illiam s G M (1999). Pentachlorophenol (PCP) produces liver 5 oxidative stress and prom otes but does not initiate hepatocarcinogenesis in B 6 C 3 F1 m ice. Carcinogenesis, 20: 1115 6 1 1 2 0 .198001 7 V an B irgelen A P ; Sm it E A ; K am pen IM ; Groeneveld C N ; Fase K M ; V an der K o lk J; Poiger H ; V an den B erg M ; 8 K oem an JH ; Brouw er A (1995). Subchronic effects o f 2 ,3 ,7 ,8 -T C D D or P C B s on thyroid horm one m etabolism : use 9 in risk assessm ent. Eur J Pharm acol, 293: 77-85. 197096 10 V an den B erg M ; Birnbaum L ; B osveld A T ; Brunstrm B ; Cook P; Feeley M ; G iesy JP ; H anberg A ; H asegaw a R ; 11 Kennedy SW ; K ubiak T; Larsen JC ; van Leeuw en F X ; Liem A K ; N olt C ; Peterson R E ; Poellinger L ; Safe S; 12 Schrenk D ; Tillitt D ; Tysklind M ; Younes M ; W aern F; Zacharew ski T (1998). T oxic equivalency factors (TEFs) for 13 P C B s, P C D D s, P C D F s for hum ans and w ildlife. Environ H ealth Perspect, 106: 775-792. 198345 14 V anden H euvel JP ; Clark G C ; K oh n M C ; Tritscher A M ; Greenlee W F; Lucier G W ; B ell D A (1994). D ioxin 15 responsive genes: exam ination o f dose-response relationships using quantitative reverse transcriptase-polym erase 16 chain reaction. Cancer R es, 54: 62-68. 197551 17 V anden H euvel JP ; Clark G C ; Tritscher A ; Lucier G W (1994). A ccum ulation o f polychlorinated dibenzo-p-dioxins 18 and dibenzofurans in liver o f control laboratory rats. Fundam A p p l T o xicol, 23: 465-469. 594318 19 V an ni H ; K azeros A ; W ang R ; H arvey B G ; Ferris B ; D e B ishnu P; Carolan B J; H bner R H ; O 'Connor T P ; Crystal 2 0 R G (2009). Cigarette sm oking induces overexpression o f a fat-depleting gene A Z G P 1 in the hum an airw ay 21 epithelium . Chest, 135: 1197-1208. 543754 2 2 van B irgelen A P ; van den B erg M (2000). Toxicokinetics. Food A ddit Contam , 17: 267-273. 523248 23 V a n B irgelen A P ; V a n der K o lk J; Fase K M ; B o l I; Poiger H ; Brouw er A ; V an den B erg M (1995). Subchronic 2 4 dose-response study o f 2,3,7,8-tetrachlorodibenzo-p-dioxin in fem ale Sprague-D aw ley rats. T o xico l A p p l 2 5 Pharm acol, 132: 1-13. 198052 2 6 V a n D e n H ove M F ; Beckers C ; D evlieger H ; D e Zegher F ; D e N ayer P (1999). H orm one synthesis and storage in 2 7 the thyroid o f hum an preterm and term newborns: effect o f thyroxine treatment. B iochim ie, 81: 563-570. 016478 2 8 V an den B erg M ; Birnbaum L S ; D enison M ; D e V ito M ; Farland W ; Feeley M ; Fiedler H ; H akansson H ; H anberg 2 9 A ; H aw s L ; R ose M ; Safe S; Schrenk D ; Tohyam a C ; Tritscher A ; Tuom isto J; Tysklind M ; W alker N ; Peterson R E 3 0 (2006). The 2005 W orld H ealth O rganization reevaluation o f hum an and m am m alian toxic equivalency factors for 31 dioxins and dioxin-like com pounds. T oxicol Sci, 93: 223-241. 543769 3 2 V an den B erg M ; de V room E ; O lie K ; H utzinger O (1986). B ioavailability o f P C D D s and P C D F s o f fly ash after 33 sem i-chronic oral ingestion by guinea pig and Syrian golden ham ster. Chem osphere, 15: 519-533. 543781 3 4 V a n der M o len G W ; K ooijm an B A ; W ittsiepe J; Schrey P; Flesch-Janys D ; Slob W (2000). Estim ation o f dioxin and 35 furan elim ination rates w ith a pharm acokinetic m odel. J E xp o A n al Environ E pidem iol, 10: 579-585. 548777 3 6 V a n der M olen G W ; K ooijm an S A L M ; M ichalek JE ; Slob W (1998). The estim ation o f elim ination rates o f 3 7 persistent com pounds: A re-analysis o f 2,3,7,8-tetrachlorodibenzo-p-dioxin levels in V ietnam veterans. 38 Chem osphere, 37: 1833-1844. 548765 3 9 V a n der M olen , G ; K ooijm an A ; Slob W (1996). A generic toxicokinetic m odel for persistent lipophilic com pounds 4 0 in hum ans: A n application to T C D D . Fundam A p p l T o xicol, 31: 83-94. 548768 This document is a draftfor review purposes only and does not constitute Agency policy. R-30 DRAFT--DO NOT CITE OR QUOTE 1 V ilu ksela M ; B ager Y ; Tuom isto JT ; Scheu G ; U n kila M ; Pohjanvirta R ; Flodstrm S; K osm a V M ; M ki2 Paakkanen J; Vartiainen T; K lim m C ; Schram m K W ; W rngrd L ; Tuom isto J (2000). Liver tum or-prom oting 3 activity o f 2,3,7,8-tetrachlorodibenzo-p-dioxin (T C D D ) in T C D D -sen sitive and T C D D -resistan t rat strains. Cancer 4 Res, 60: 6911-6920. 198968 5 V o s JG , M oore JA , Z in k l J G (1973). E ffe ct o f 2,3,7,8-tetrachlorodibenzo-p-dioxin on the im m une system o f 6 laboratory anim als. Environ H ealth Perspect, 5: 149-162. 198367 7 W alker N J; Portier C J; L a x SF ; Crofts F G ; L i Y ; Lucier G W ; Sutter T R (1999). Characterization o f the dose8 response o f C Y P 1 B 1 , C Y P 1 A 1 , and C Y P 1 A 2 in the liver o f fem ale Sprague-D aw ley rats follow in g chronic 9 exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin. T o xico l A p p l P harm acol, 154: 279-286. 198615 10 W alker N J; Tritscher A M ; Sills R C ; Lucier G W ; Portier C J (2000). H epatocarcinogenesis in fem ale Sprague11 D aw ley rats fo llo w in g discontinuous treatm ent w ith 2,3,7,8-tetrachlorodibenzo-p-dioxin. T o xico l S ci, 54: 330-337. 12 198733 13 W ang S L ; Su P H ; Jo n g SB ; G uo Y L ; Chou W L ; Ppke O (2005). In utero exposure to dioxins and polychlorinated 14 biphenyls and its relations to thyroid function and grow th horm one in new borns. E nviron H ealth Perspect, 113: 15 1645-1650. 198734 16 W ang X ; Santostefano M J; D eV ito M J; Birnbaum L S (2000). Extrapolation o f a P B P K m odel for dioxins across 17 dosage regim en, gender, strain, and species. T o xico l S ci, 56: 49-60. 198738 18 W ang X ; Santostefano M J; Evans M V ; Richardson V M ; Diliberto JJ; Birnbaum L S (1997). Determ ination o f 19 param eters responsible for pharm acokinetic behavior o f T C D D in fem ale Sprague-D aw ley rats. T o xicol A p p l 2 0 Pharm acol, 147: 151-168. 104657 21 W are JH ; Spengler JD ; N eas L M ; Sam et JM ; W agner G R ; Coultas D ; O zkaynak H ; Schw ab M (1993). Respiratory 2 2 and irritant health effects o f am bient volatile organic com pounds: the K anaw ha County health study. A m J 23 Epidem iol, 137: 1287-1301. 004687 2 4 W arner M ; Eskenazi B ; M ocarelli P; Gerthoux P M ; Sam uels S; N eedham L ; Patterson D ; B ram billa P (2002). 2 5 Serum dioxin concentrations and breast cancer risk in the seveso w om en's health study. Environ H ealth Perspect, 2 6 110: 625-628. 197489 2 7 W arner M ; Eskenazi B ; O live D L ; Sam uels S; Q u ick-M iles S; V ercellini P; Gerthoux P M ; N eedham L ; Patterson 2 8 D G Jr; M ocarelli P (2007). Serum dioxin concentrations and quality o f ovarian function in w om en o f seveso. 2 9 Environ H ealth Perspect, 115: 336-340. 197486 3 0 W arner M ; Sam uels S; M ocarelli P; Gerthoux P M ; N eedham L ; Patterson D G Jr; Eskenazi B (2004). Serum dioxin 31 concentrations and age at m enarche. Environ H ealth Perspect, 112: 1289-1292. 197490 3 2 W eber R ; Schm itz H -J; Schrenk D ; H agenm aier H (1997). M etabolic degradation, inducing potency, and 33 m etabolites o f fluorinated and chlorinated-fluorinated dibenzodioxins and dibenzofurans. Chem osphere, 34: 29-40. 34 548753 3 5 W en d lin g JM ; O rth R G ; P oiger H (1990). D eterm ination o f [3H ]-2,3,7,8-tetrachlorodibenzo-p-dioxin in hum an 3 6 feces to ascertain its relative m etabolism in m an. A n al Chem , 62: 796-800. 548751 3 7 W hite K L Jr; Lysy H H ; M cC ay JA ; Anderson A C (1986). M odulation o f serum com plem ent levels follow in g 3 8 exposure to polychlorinated dibenzo-p-dioxins. T oxicol A ppl Pharm acol, 84: 209-219. 197531 3 9 W hite R H ; Cote I; Zeise L ; F o x M ; D om inici F; Burke T A ; W hite P D ; H attis D B ; Sam et JM (2009). State-of-the4 0 Science W orkshop Report: Issues and Approaches in Low -D ose--R esponse Extrapolation for Environm ental H ealth 41 R isk Assessm ent. Environ H ealth Perspect, 117: 283-287. 622764 This document is a draftfo r review purposes only and does not constitute Agency policy. R -31 D R A F T -- D O N O T C IT E O R Q U O T E 1 W H O (1978). International Classification o f D iseases: N inth R evision. G eneva, Sw itzerland: W orld H ealth 2 O rganization. 594329 3 W H O (1988). A ssessm ent o f the health risk o f dioxins: re evaluation o f the tolerable daily intake (T D I). W H O 4 European Centre for Environm ental H ealth and International Program m e on Chem ical Safety. G eneva, Sw itzerland. 5 594278 6 W H O (2005). C hem ical-specific adjustm ent factors for interspecies differences and hum an variability: guidance 7 docum ent for use o f data in dose/concentration-response assessm ent. W orld H ealth O rganization. G eneva, 8 Sw itzerland. H arm onization Project D ocum ent N o. 2. 198739 9 W hysner J; W illiam s G M (1996). 2,3,7,8-T etrachlorodibenzo-p-dioxin m echanistic data and risk assessm ent: gene 10 regulation, cytotoxicity, enhanced cell proliferation, and tum or prom otion. Pharm acol Ther, 71: 193-223. 197556 11 W ittsiepe J; Erlenkm per B ; W elge P; H ack A ; W ilhelm M (2007). B ioavailability o f P C D D /F from contam inated 12 soil in young Goettingen m inipigs. Chem osphere, 67: S355-S364. 548736 13 W ong T K ; D om in B A ; B ent P E ; B lanton T E ; Anderson M W ; Philpot R M (1986). Correlation o f placental 14 m icrosom al activities w ith protein detected by antibodies to rabbit cytochrom e P-450 isozym e 6 in preparations 15 from hum ans exposed to polychlorinated biphenyls, quaterphenyls, and dibenzofurans. Cancer R es, 46: 999-1004. 16 548795 17 W oods C G ; Burns A M ; Bradford B U ; Ross P K ; K osyk O ; Sw enberg JA ; Cunningham M L ; R usyn I (2007). W Y 18 14,643-induced cell proliferation and oxidative stress in m ouse liver are independent o f N A D P H oxidase. T o xico l 19 Sci, 98: 366-374. 543735 2 0 W yde M E ; Cam bre T ; Lebetkin M ; Eldridge S R ; W alker N J (2002). Prom otion o f altered hepatic fo ci by 2 ,3 ,7 ,8 21 Tetrachlorodibenzo-p-dioxin and 17h-estradiol in m ale Sprague-D aw ley rats. T o xicol Sci, 68: 295-303. 197009 2 2 W yd e M E ; E ldridge S R ; L u cier G W ; W alker N J (2001). R egu lation o f 2,3,7,8-tetrachlorodibenzo-p-dioxin-induced 23 tum or prom otion by 17 beta-estradiol in fem ale Sprague-D aw ley rats. T o xicol A p p l Pharm acol, 173: 7-17. 198575 2 4 Y a n g JZ ; A garw al S K ; Foster W G (2000). Subchronic exposure to 2,3,7,8-tetrachlorodibenzo-p-dioxin m odulates 2 5 the pathophysiology o f endom etriosis in the cynom olgus m onkey. T o xicol Sci, 56: 374-381. 198590 2 6 Y o u ak im S (2006). R isk o f cancer am ong firefighters: A quantitative review o f selected m alignancies. A rch Environ 2 7 O ccup H ealth, 61: 223-231. 197295 2 8 Z a ck JA ; G affey W R (1983). A m ortality study o f w orkers em ployed at the M onsanto Com pany plant in N itro, W est 2 9 V irginia. Environ Sci R es, 26: 575-591. 548783 3 0 Z a ck JA ; Suskind R R (1980). The m ortality experience o f w orkers exposed to tetrachlorodibenzodioxin in a 31 trichlorophenol process accident. J O ccup Environ M ed , 22: 11-14. 065005 3 2 Zareba G ; H ojo R ; Zareba K M ; W atanabe C ; M arkow ski V P ; B aggs R B ; W eiss B (2002). Sexually dim orphic 33 alterations o f brain cortical dom inance in rats prenatally exposed to T C D D . J A p p l T o xicol, 22: 129-137. 197567 3 4 Zeise L ; W ilson R ; Crouch E A C (1987). D ose-response relationships for carcinogens: a review . Environ H ealth 35 Perspect, 73: 259-308. 060867 3 6 Zober A ; M esserer P; H uber P (1990). Thirty-four-year m ortality follow -up o f B A S F em ployees exposed to 2 ,3 ,7,8 3 7 T C D D after the 1953 accident. Int A rch O ccup Environ H ealth, 62: 139-157. 197604 This document is a draftfor review purposes only and does not constitute Agency policy. R-32 DRAFT--DO NOT CITE OR QUOTE 1 Zober A ; O tt M G ; M esserer P (1994). M orb id ity fo llo w up study o f B A S F employees exposed to 2,3,7, 8 2 tetrachlorodibenzo-p-dioxin (TC D D ) after a 1953 chem ical reactor incident. Occup E n viro n Med, 51: 479-486. 3 197572 4 Zober A ; Papke O (1993). Concentrations o f PCDDs and PCDFs in human tissue 36 years after accidental d ioxin 5 exposure. Chemosphere, 27: 413-418. 197602 6 Zober A ; S c hilling D ; O tt M G ; Schauwecker P; Riem ann JF; M esserer P (1998). Helicobacter p ylo ri infection: 7 prevalence and clin ica l relevance in a large company. J Occup E n viro n M ed, 40: 586-594. 594300 8 A ltekruse, SF; Kosary, C L; Krapcho, M ; et al., eds. (2010) SEER Cancer Statistics R eview , 1975-2007.N ational 9 Cancer Institute. Bethesda, M D , based on Novem ber 2009 SEER datasubmission, posted to the SEER web site, 10 2010. A vailable online at http://seer.cancer.gov/csr/1975_2007/. 11 Auso, E ; Lavado-A utric, R ; Cuevas, E ; et al. (2004) A moderate and transient deficiency o f m aternal thyroid 12 function at the beginning o f fetal neocorticogenesis alters neuronal m igration. Endocrinology 145:4037-4047. 13 B aird, SJS; Cohen, JT; Graham, JD, et al. (1996) Noncancer risk assessment: a probabilistic alternative to current 14 practice. H um an Ecol R isk Assess 2:79-102. 15 Calabrese, EJ; G ilbert, CE. (1993) Lack o f to ta l independence o f uncertainty factors (U fs): Im plications fo r the size 16 o f the total uncertainty factor. Reg T o xicol Pharm acol 17:44-51. 17 Calabrese, EJ; B aldw in, L A . (1995) A toxicological basis to derive generic interspecies uncertainty factors fo r 18 application in human and ecological risk assessment. H um an Ecol R isk Assess 1(5):555-564. 19 C alvo, R M ; Jauniaux, E ; G ulbis, B ; et al. (2002) Fetal tissues are exposed to b iolog ically relevant free thyroxine 2 0 concentrations during early phases o f development. J C lin Endocrinol Metab 87(4): 1768-1777. 21 Chan, S; Franklyn, JA ; K ilb y , M D . (2005) M aternal thyroid hormones and fe ta l brain development. C urr O pinion 2 2 Endocrinol Diab 12:23-30. 23 Chu, I; V a lli, V E ; Rousseaux, CG. (2007) Combined effects o f 2,3,7,8-tetrachlorodibenzo-pdioxin and 2 4 polychlorinated biphenyl congeners in rats. To xic o l E n viro n Chem 89(1):71-87. 25 Cook, R R . (1981) D io xin , chloracne, and soft tissue sarcoma. Lancet 1:618-619. 2 6 Crump, K S ; C hiu, W A ; Subramanian, RP. (2010) Issues in using human v a ria b ility distributions to estimate lo w 2 7 dose risk. E n viro n H ealth Perspect 118(3):387-393. 2 8 Delange, F; Bourdoux, P; Erm ans, A M . (1985) Transient disorders o f thyroid function and regulation in preterm 2 9 infants. In: Delange, F; Fisher, D ; M alvaux, P; eds. Pediatric Thyroidology.Basel, S. Karger. pp 369-393. 3 0 D ella Porta, G; Dragani, T A ; Sozzi, G. (1987) Carcinogenic effects o f in fa n tile and long-term 31 2,3,7,8-tetrachlorodibenzo-p-dioxin treatm ent in the mouse. Tum ori 73: 99-107. 3 2 Denison, M S ; Nagy, SR. (2003) A c tivation o f the a ryl hydrocarbon receptor by structurally diverse exogenous and 33 endogenous chemicals. A nnu R ev Pharm acol T o xic o l 43:309-334. 3 4 Evans, JS; B aird, SJS. (1998) Accounting fo r m issing data in noncancer risk assessment. H um an Ecological R isk 35 Assess 4:291-317. 3 6 Geusau, A ; Abraham, K ; Geissler, K ; et al. (2001) Severe 2,3,7,8-tetrachlorodibenzo-p-dioxin (TC D D ) intoxication: 3 7 clinical and laboratory effects. E nviron H ealth Perspect 109(8):865-869. This document is a draftfor review purposes only and does not constitute Agency policy. R-33 DRAFT--DO NOT CITE OR QUOTE 1 G linoer, D ; D elange, F . (2000) The potential repercussions o f m aternal, fetal, and neonatal hypothyroxinem ia on the 2 progeny. Thyroid 10(10):871-887. 3 H aavisto, T E ; M y lly m ak i, S A ; A dam sson, N A ; et al. (2006) The effects o f m aternal exposure to 4 2,3,7,8-tetrachlorodibenzo-p-dioxin on testicular steroidogenesis in infantile m ale rats. Int J A ndrol 2 9:313-322. 5 H ahn, M E . (2002) A ry l hydrocarbon receptors:diversity and evolution. C hem -B iol Interact 141:131-160. 6 H orner M J; R ies L A G ; K rapcho M ; et al.; eds. (2009) S E E R Cancer Statistics R eview , 1975-2006. Betheda, M D : 7 N ation al C an cer Institute. A vailab le online at http://seer.cancer.gov/csr/1975_2006/, based on N ovem ber 2008 8 S E E R data subm ission, posted to the S E E R w eb site, 2009. 9 H utt, K J; Sh i, Z ; A lb ertin i, D F ; et al. (2008) The environm ental toxicant 2,3,7,8-tetrachlorodibenzo-p-dioxin 10 disrupts m orphogenesis o f the rat pre-im plantation em bryo. B M C D evelopm ental B iology 8 :1-1 2 . 11 IO M (Institute o f M edicine). (1994) Veterans, and A gen t O range: health effects o f herbicides used in V ietnam . 12 W ashington, D C : N ational A cadem y Press. 13 IP C S (International Program m e on Chem ical Safety). (2005) Chem ical-specific adjustm ent factors for interspecies 14 differences and hum an variability: guidance docum ent for use o f data in dose/concentration-response assessm ent. 15 harm onization project D ocum ent N o. 2. W orld H ealth O rganization, G eneva, Sw itzerland. 16 K ahn , P C ; G ochfeld , M ; N ygren, M ; et al. (1988) D ioxin s and dibenzofurans in blood and adipose tissue o f A gent 17 O range-exposed V ietnam veterans and m atched controls. JA M A 259:1661-1667. 18 K an g, H K ; D alager, N A ; N eedham , L L ; et al. (2006) H ealth status o f A rm y Chem ical Corps V ietnam veterans w ho 19 sprayed defoliant in Vietnam . A m er J Indust M ed 49:875-884. 2 0 K od ell, R L ; G aylor, D W . (1999) Com bining uncertainty factors in deriving hum an exposure levels o f 21 noncarcinogenic toxicants. A nnals N ew Y o rk A cadem y o f Sciences 895:188-195. 2 2 K rishnan, K ; Andersen, M . (2007) Physiologically based pharm acokinetic m odelling in toxicology. In Principles 2 3 a n d m e th o d s o f to x ic o lo g y ( A .W .H a y e s , E d .) , 5th e d ., p p . 2 3 1 -2 9 1 . C R C P re ss , N e w Y o r k . 2 4 Landi, M T ; B ertazzi, P A ; B accarelli, A ; et al. (2003) T C D D -m ed iated alterations in the AhR -dependent pathw ay in 2 5 Seveso, Italy, 20 years after the accident. Carcinogenesis 24:673-680. 2 6 Lavad o-A u tric, R ; A u so , E ; G arcia-V elasco , JV ; et al. (2003) Early m aternal hypothyroxinem ia alters histogenesis 2 7 and cerebral cortex cytoarchitecture o f the progeny. J C lin Invest 111:1073-1082. 2 8 Lu tz, W K ; G aylor, D W ; C on olly, R B , et al. (2005) N onlinearity and thresholds in dose-response relationships for 2 9 carcinogenicity due to sam pling variation, logarithm ic dose scaling, or sm all differences in individual susceptibility. 3 0 T oxicol A p p l Pharm acol 207(Suppl. 2):565-569. 31 M orreale de Escobar, G ; O bregon, M J; et al. (2000) Is neuropsychological developm ent related to m aternal 3 2 hypothyroidism or to m aternal hypothyroxinem ia? J C lin Endocrinol M etab 85(11):3975-3987. 33 N A S (N ational A cadem y o f Sciences), ed. (2005) H ealth Im plications o f Perchlorate Ingestion. W ashington D C : 3 4 N ational Research C ouncil o f the N ational Academ ies. 35 N A S (N ational A cadem y o f Sciences). (2007) Toxicity testing in the 21st century. A vision and a strategy. Report 3 6 o f the Com m ittee on Toxicity Testing and Assessm ent o f Environm ental A gents. N ational Research Council o f The 3 7 N ational A cadem ies. W ashington, D C : The N ational Academ ies Press. A vailable online at 3 8 w w w .nap.edu/catalog/11970.htm l. This document is a draftfor review purposes only and does not constitute Agency policy. R-34 DRAFT--DO NOT CITE OR QUOTE 1 N avarro, C; Chirlaque, M D ; Torm o, M J; et al. (2006) V a lid ity o f se lf reported diagnoses o f cancer in a m ajor 2 Spanish prospective cohort study. J Epidem iol Comm H ealth 60: 593-599. 3 Needham, L L ; G erthoux, P M ; Patterson, D G , Jr; et al. (1997) Serum d ioxin levels in Seveso, Ita ly, population in 4 1976. Teratog Carcinog M utagen 17:225-240. 5 N R C (N ational Research C ouncil). (2005) H ealth risks fro m exposure to low levels o f ionizing radiation: B E IR V II. 6 W ashington, D C: N ational Academy Press (as cited by W hite et al., 2008). 7 N TP (N ational Toxicology Program ). (2006a) N TP technical report on the toxicology and carcinogenesis studies o f 8 2,3,7,8-tetrachlorodibenzo-p-dioxin (TC D D ) (C AS No. 1746-01-6) in fem ale H arlan Sprague-Dawley rats (Gavage 9 Studies). N atl T o xicol Program Tech Rep 521. Public H ealth Service, N ational Institute o f H ealth, U .S. Departm ent 10 o f H ealth and Hum an Services, Research Triangle Park, NC. 11 O kura, Y ; Urban, L H ; M ahoney, D W ; et al.(2004) Agreem ent between self-report questionnaires and medical 12 record data was substantial fo r diabetes, hypertension, m yocardial infa rctio n and stroke but not fo r heart failure. 13 J C linic Epidem iol 57: 1096-110 14 Patterson, D ; Ham pton, L ; Lapeza, CR, Jr; et al. (1987) H ig h resolution gas chrom atographic/high resolution mass 15 spectrometer analysis o f human serum on a w hole-w eight and lip id basis fo r 2,3,7,8-tetrachlorodibenzo-p-dioxin. 16 A nal Chem 59:2000-2005. 17 Patterson, D G , Jr.; W ong, L -Y ; Turner, W E ; et al. (2009) Levels in the U .S. population o f those persistent organic 18 pollutants (2003-2004) included in the Stockholm C onvention o r in other Long-Range Tran boundary A ir P o llu tio n 19 Agreements. E n viro n Sci Technol 43 (4 ):1 2 1 1 -1218. 2 0 Pesonen, SA; H aavisto, T E ; V iluksela, M ; et al. (2006) Effects o f in utero and lactational exposure to 21 2,3,7,8-tetrachlorodibenzo-p-dioxin (TC D D ) on rat fo llic u la r steroidogenesis. Reprod T o xic o l 22:521-528. 2 2 P ito t et al. 1991 pg 5-36. N ot listed in reference section and not in H ERO . Maybe P ito t and Dragan (available in 23 HERO )? 2 4 Poiger, M ; Schlatter, C. (1986) Pharm acokinetics o f 2,3,7,8-TC D D in man. Chemosphere 15:9-12. 25 Puga, A ; Barnes, SJ; D alton, TP ; et al. (2000) A rom atic hydrocarbon receptor interaction w ith the retinoblastom a 2 6 protein potentiates repression o f E2F-dependent transcription and cell cycle arrest. J B io l Chem 275:2943-2950. 2 7 Rigon, F; Bianchin, L ; Bernasconi, S; et al. (2010) Update on age at menarche in Ita ly: tow ard the leveling o ff o f 2 8 the secular trend. J Adolesc H ealth 46(3):238-244. 2 9 R ovet, JF. (2002) Congenital hypothyroidism : an analysis o f persisting deficits and associated factors. C hild 30 Neuropsychol 8(3):150-62. 31 Royland, J; Parker, J; G ilbert, M E . (2008) A genomic m icroarray analysis o f hippocampus and neocortex fo llo w in g 3 2 modest reductions thyroid hormone during development. J N euroendocrinol 12:1319-13 33 Safe, SH. (1986) Com parative toxicology and mechanism o f action o f polychlorinated dibenzo-p-dioxins and 3 4 dibenzofurans. A nnu R ev Pharm acol T o xic o l 26:371-379. 35 Savin, S; C vejic, D ; Nedic, O et al. (2003) Thyroid hormone synthesis and storage in the thyroid gland o f human 3 6 neonates. J. Pediatr Endocrinol Metab 16:521-528.Schantz, SL; Bowm an, R E. (1989) Learning in monkeys 3 7 exposed perinatally to 2,3,7,8-tetrachloridibenzo-p-dioxin (TC D D ). N eurotoxicol Teratol 11:13-19. 38 Sharlin, D S; Tighe, D ; et al. (2008) The balance between oligodendrocyte and astrocyte production in m ajor w hite 3 9 m atter tracts is line arly related to serum total thyroxine. Endocrinology 149(5):2527-2536. This document is a draftfo r review purposes only and does not constitute Agency policy. R -35 D R A F T -- D O N O T C IT E O R Q U O T E 1 Sharlin, D S ; G ilbert, M E ; Taylor, M ; et al. (2010). The nature o f the com pensatory response to low thyroid horm one 2 in the developing brain. J N euroendocrinol. 22(3):153-165. 3 Slam a, R ; Eutache, F; D ucot, B ; et al. (2002) Tim e to pregnancy and sem en param eters: a cross-sectional study 4 am ong fertile couples from four European cities. H um an Repro 17:503-515. 5 Subram aniam , R P ; W hite, P; C oglian o, V J. (2006) Com parison o f cancer slope factors using different statistical 6 approaches. R isk A n al. 26(3):825-830. 7 Sw an, SH ; B razil, C ; D robnis, E Z ; et al. (2003) G eographic differences in sem en quality o f fertile U .S . m ales. 8 Environ H ealth Perspect 111(4):414-20. 9 U .S . E P A (Environm ental Protection A gen cy). (1994) M ethods for derivation o f inhalation reference concentrations 10 and application o f inhalation dosim etry. O ctober. O ffice o f H ealth and Environm ental Assessm ent, Environm ental 11 Criteria and Assessm ent O ffice, W ashington, D C . EPA/600/8-90/066F. 12 U .S . E P A (Environm ental Protection A gen cy). (2001) Evaluation o f the carcinogenic potential o f lindane. Final 13 Report. Cancer assessm ent docum ent. Cancer Assessm ent R eview Com m ittee, H ealth Effects D ivision, O ffice o f 14 Pesticide Program s, W ashington, D C . A vailable online at 15 http://w w w .lindane.com /pdf/EP A _Cancer_A ssessm ent_of_Lindane2001.pdf. 16 Verm eire, T ; Stevenson, H ; Pieters, M N ; et al. (1999) Assessm ent factors for hum an health risk assessm ent: a 17 discussion paper. C rit R ev T oxiocol 29(5):439-490. 18 W alker, N J; M iller, B D ; K ohn, M C ; et al. (1998). D ifferences in kinetics o f induction and reversibility o f 19 T C D D -in d u ced changes in cell proliferation and C Y P 1 A 1 expression in fem ale Sprague-D aw ley rat liver. 2 0 Carcinogenesis 19:1427-1435. 21 W alker, N J; Tritscher, A M ; Sills, R C ; et al. (2000) H epatocarcinogenesis in fem ale Sprague-D aw ley rats follow in g 2 2 discontinuous treatm ent w ith 2,3,7,8-tetrachlorodibenzo-p-dioxin T o xico l S ci 54:330-337. 23 W are, J. (1993) A ppendix C : Script for personal interview SF-36 adm inistration. In: W are, JE , Jr; Snow , K K ; 2 4 K osin sk i, M ; et al., eds. SF-36 health survey m anuals and interpretation guide. B oston , M A : N im rod Press. 25 W arner, M ; Eskenazi, B . (2005) T C D D and puberty: W arner and Eskenazi Respond. Environ H ealth Perspect 2 6 113:A18-A18. 2 7 W hite, R H ; C ote, I; Z eise, L ; et al. (2009) State-of-the-science w orkshop report: issues and approaches in low -dose2 8 response extrapolation for environm ental health risk assessm ent. E nviron H ealth Perspect 117(2):283-287. 2 9 W hitlock, JP . (1999) Induction o f cytochrom e P4501A 1. A nnu R ev Pharm acol T o xicol 39:103-125. 3 0 W H O (W orld H ealth O rganization). (1994) Indicators for assessing iodine deficiency disorders and their control 31 through salt iodization. G en eva: W orld H ealth O rganization. W H O /N U T /94.6 W H O /N U T /94.6. 3 2 W H O (W orld H ealth O rganization). (2007) Assessm ent o f iodine deficiency disorders and m onitoring their 33 elim ination. Geneva: W H O Press. 3 4 W ijchm an, JG ; D e W o lf, B ; G raaff, R ; et al. (2001) V ariation in sem en param eters derived from com puter aided 35 sem en analysis w ithin donors and betw een donors. J A ndrol 22(5):773-780. 3 6 W yrobek, A J; Gordon, L A ; W atchm aker, G ; et al. (1982) H um an sperm m orphology testing: description o f a 3 7 reliable m ethod and its statistical pow er. In: B redges, B A ; Butterw orth, B E ; W einstein, IB ; eds. B anbury Report 3 8 Indicators o f G enotoxic Exposure. C old Spring H arbor, N Y : C old Spring Laboratory. pp. 527-54 This document is a draftfor review purposes only and does not constitute Agency policy. R-36 DRAFT--DO NOT CITE OR QUOTE 1 Zoeller, R T ; R ovet, J. (2004). T im in g o f thyroid horm one action in the developing brain: clinical observations and 2 experim ental findings. J N euroendocrinol 16(10):809-818. 3 Zeise, L ; W ilson, R ; Crouch, E .A . (1987) D ose-response relationships for carcinogens: a review . Environ H ealth 4 Perspect 73:259-306. 5 This document is a draftfor review purposes only and does not constitute Agency policy. R-37 DRAFT--DO NOT CITE OR QUOTE