Document x53exJ8ren0xOva0rp8LJdk2J

FILE NAME: Scotts (SCTS) DATE: 2011 Aug DOC#: SCTS010 DOCUMENT DESCRIPTION: Toxicological Review of Libby Amphibole Asbestos DRAFT--DO NOT CITE OR QUOTE EPA/635/R-11/002A www.epa.gov/iris *>EPA TOXICOLOGICAL REVIEW PLAINTIFFS EXHIBIT SCO-25 OF LIBBY AMPHIBOLE ASBESTOS In Support of Summary Information on the Integrated Risk Information System (IRIS) August 2011 (Note: This document is an assessment o f the noncancer and cancer health effects associated with the inhalation route of exposure only) NOTICE This document is an External Review draft. This information is distributed solely for the purpose of predissemination 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 any Agency determination or policy. It is being circulated for review of its technical accuracy and science policy implications. U.S. Environmental Protection Agency Washington, DC DISCLAIMER This document is a preliminary draft for review purposes only. This document is distributed solely for the purpose of predissemination 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 any Agency determination or policy. Mention of trade names or commercial products does not constitute endorsement or recommendation for use. This document is a draftfor review purposes only and does not constitute Agency policy. iii DRAFT--DO NOT CITE OR QUOTE CONTENTS--TOXICOLOGICAL REVIEW OF LIBBY AMPHIBOLE ASBESTOS LIST OF TABLES.......................................................................................................................... x LIST OF FIGURES.....................................................................................................................xiii LIST OF ABBREVIATIONS AND ACRONYMS.................................................................... xiv FOREWORD.............................................................................................................................xviii AUTHORS, CONTRIBUTORS, AND REVIEWERS............................................................... xix 1. INTRODUCTION................................................................................................................... 1-1 1.1. RELATED ASSESSMENTS.................................................................................... 1-3 1.1.1. IRIS Assessment for Asbestos (U.S. EPA, 1988a)................................... 1-3 1.1.2. EPA Health Assessment for Vermiculite (1991b)..................................... 1-4 1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH ASSESSMENT.................................................................................................... 1-4 2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY, USE, AND EXPOSURE POTENTIAL.................................................................................................................... 2-1 2.1. HISTORICAL BACKGROUND.............................................................................. 2-1 2.2. GEOLOGY AND MINERALOGY OF LIBBY AMPHIBOLE ASBESTOS........2-3 2.2.1. Silicate Minerals........................................................................................2-4 2.2.1.1. Mineralogy and Structure of Amphiboles...................................2-5 2.2.1.2. Amphibole Morphology.............................................................. 2-7 2.2.2. Vermiculite.............................................................................................. 2-12 2.2.3. The Mineralogy of Libby Amphibole Asbestos......................................2-14 2.2.3.1. Mineralogy................................................................................ 2-14 2.2.3.2. Morphology of the Libby Amphibole Asbestos.......................2-17 2.2.3.3. Dimensional Characteristics of Libby Amphibole Asbestos.... 2-18 2.3. EXPOSURE POTENTIAL.....................................................................................2-20 2.3.1. Libby Community...................................................................................2-21 2.3.2. Communities near Vermiculite Expansion and Processing Plants.........2-24 2.3.3. Exposures from Zonolite and Vermiculite for Homeowners, Contractors, and Other Populations.......................................................2-24 3. FIBER TOXICOKINETICS.................................................................................................... 3-1 3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT.............................. 3-3 3.2. CLEARANCE........................................................................................................... 3-8 3.2.1. Inhalation................................................................................................... 3-8 3.2.1.1. Respiratory Tract........................................................................ 3-8 3.2.1.2. Pleural Cavity and Extrapulmonary Sites................................. 3-13 3.2.2. Ingestion.................................................................................................. 3-14 3.2.3. Dermal Contact........................................................................................ 3-14 3.3. SUMMARY............................................................................................................ 3-15 4. HAZARD IDENTIFICATION OF LIBBY AMPHIBOLE ASBESTOS...............................4-1 4.1. STUDIES IN HUMANS--EPIDEMIOLOGY...................................................... 4-1 4.1.1. Studies of Libby, MT Vermiculite Mining Operation Workers.............. 4-2 This document is a draftfor review purposes only and does not constitute Agency policy. iv DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 4.1.1.1. Description of Mining and Milling Operations.......................... 4-3 4.1.1.2. Exposure Estimation...................................................................4-4 4.1.1.3. Cancer Mortality Risk...............................................................4-14 4.1.1.4. Noncancer Effects: Respiratory and Cardiovascular D isease....................................................................................... 4-20 4.1.2. Libby, MT Community Studies...............................................................4-28 4.1.2.1. Geographic Mortality Analysis.................................................4-28 4.1.2.2. Community Screening--Respiratory Health............................ 4-29 4.1.2.3. Other Reports of Asbestos-Related Disease Among Libby, MT Residents.............................................................................4-35 4.1.2.4. Summary of Respiratory Health Effects in Libby, MT Community Studies.................................................................... 4-35 4.1.3. Marysville, OH Vermiculite Processing Plant Worker Studies..............4-36 4.1.3.1. Summary of Marysville, OH Vermiculite Processing Plant Worker Studies........................................................................... 4-44 4.1.4. Community Studies from Other Vermiculite Processing Plants.............4-44 4.1.4.1. Summary of Community Studies from Other Vermiculite Processing Plants.......................................................................4-46 4.1.5. Case Reports............................................................................................ 4-48 4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN ANIMALS--ORAL, INHALATION AND OTHER ROUTES OF EXPOSURE....................................................................................................... 4-48 4.2.1. Oral.......................................................................................................... 4-49 4.2.2. Inhalation.................................................................................................4-49 4.2.3. Intratracheal Instillation Studies.............................................................. 4-54 4.2.4. Injection/Implantation Studies.................................................................4-56 4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite Asbestos................................................................................................. 4-58 4.3. OTHER DURATION OR ENDPOINT-SPECIFIC STUDIES..............................4-59 4.3.1. Immunological.........................................................................................4-59 4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF ACTION..........................................................................................4-62 4.4.1. Inflammation and Immune Function.......................................................4-63 4.4.2. Genotoxicity............................................................................................ 4-67 4.4.3. Cytotoxicity and Cellular Proliferation...................................................4-69 4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS...........................................4-71 4.5.1. Pulmonary Effects...................................................................................4-71 4.5.1.1. Pulmonary Fibrosis (Asbestosis).............................................. 4-71 4.5.1.2. Other Nonmalignant Respiratory Diseases...............................4-73 4.5.2. Pleural Effects.........................................................................................4-73 4.5.3. Other Noncancer Health Effects (Cardiovascular Toxicity, Autoimmune Effects).............................................................................4-74 4.5.4. Libby Amphibole Asbestos Summary of Noncancer Health Effects.....4-75 This document is a draftfor review purposes only and does not constitute Agency policy. v DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 4.5.5. Mode-of-Action Information (Noncancer)............................................ 4-76 4.6. EVALUATION OF CARCINOGENICITY........................................................... 4-77 4.6.1. Summary of Overall Weight of Evidence..............................................4-77 4.6.1.1. Synthesis of Human, Animal, and Other Supporting Evidence..................................................................................... 4-78 4.6.2. Mode-of-Action Information...................................................................4-80 4.6.2.1. Description of the Mode-of-Action Information.....................4-80 4.6.2.2. Application of the Age-Dependent Adjustment Factors..........4-81 4.7. SUSCEPTIBLE POPULATIONS.......................................................................... 4-81 4.7.1. Influence of Different Lifestages on Susceptibility................................4-82 4.7.1.1. Lifestage Susceptibility.............................................................4-82 4.7.2. Influence of Gender on Susceptibility.....................................................4-88 4.7.3. Influence of Race or Ethnicity on Susceptibility....................................4-89 4.7.4. Influence of Genetic Polymorphisms on Susceptibility.........................4-89 4.7.5. Influence of Health Status on Susceptibility........................................... 4-91 4.7.6. Influence of Lifestyle Factors on Susceptibility...................................... 4-92 4.7.7. Susceptible Populations Summary.......................................................... 4-93 5. EXPOSURE-RESPONSE ASSESSMENT........................................................................... 5-1 5.1. ORAL REFERENCE DOSE (RID).......................................................................... 5-1 5.2. INHALATION REFERENCE CONCENTRATION (Rl'C).................................... 5-1 5.2.1. Choice of Principal Study and Critical Effect..........................................5-1 5.2.1.1. Evaluation of Candidate Studies and Selection of Critical Study............................................................................................ 5-3 5.2.1.2. Evaluation of Exposure Paradigm in Candidate Studies...........5-3 5.2.1.3. Evaluation of Exposure Assessment in Candidate Studies......5-12 5.2.1.4. Selection of Critical Effect........................................................ 5-15 5.2.2. Evaluation of Radiographic Lesions as Potential Critical Effects..........5-16 5.2.2.1. Health Effects of Parenchymal Changes as Small Opacities Viewed on Standard Radiographs.............................................. 5-16 5.2.2.2. Health Effects of Diffuse Pleural Thickening (DPT) Viewed on Standard Radiographs.............................................. 5-17 5.2.2.3. Health Effects of Localized Pleural Thickening (LPT) Viewed on Standard Radiographs.............................................. 5-18 5.2.3. Methods of Analysis................................................................................ 5-23 5.2.3.1. Exposure Data and Choice of Exposure Metric.......................5-23 5.2.3.2. Data Sets for Modeling Analyses............................................. 5-26 5.2.3.3. Statistical Modeling of the Sub-cohort..................................... 5-29 5.2.4. RfC Derivation--Including Application of Uncertainty Factors (L i's)...................................................................................................... 5-35 5.2.5. Alternative Analyses of the Full Marysville Cohort............................... 5-39 5.2.6. Previous Reference Concentration (RfC) Derivation.............................. 5-41 This document is a draftfor review purposes only and does not constitute Agency policy. vi DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 5.3. UNCERTAINTIES IN THE INHALATION REFERENCE CONCENTRATION (RfC)............................................................................... 5-41 5.3.1. Uncertainty in the Exposure Reconstruction........................................... 5-41 5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural Thickening............................................................................................. 5-44 5.3.3. Uncertainty Due to Time From First Exposure......................................5-45 5.3.4. Uncertainty in Background Rate of Localized Pleural Thickening........ 5-45 5.3.5. Uncertainty in Model Functional Form and Lagged Exposure.............. 5-46 5.3.6. Uncertainty Due to Effect of Smoking...................................................5-46 5.3.7. Sensitivity Analysis: Derivation of a POD for Lifetime Exposure From the Cumulative Exposure Metric.................................................. 5-48 5.3.8. Sensitivity Analysis for Choice of Critical Effect and Selection of Benchmark Response (BMR)................................................................ 5-50 5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT........................................... 5-51 5.4.1. Overview of Methodological Approach.................................................. 5-51 5.4.2. Choice of Study/Data--with Rationale and Justification........................ 5-53 5.4.2.1. Description of the Libby Worker Cohort.................................5-54 5.4.2.2. Description of Cancer Endpoints.............................................. 5-56 5.4.2.3. Description of Libby Amphibole Asbestos Exposures............5-58 5.4.2.4. Description of Libby Worker Cohort Work Histories............. 5-65 5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM) and Work Histories......................................................... 5-66 5.4.3. Exposure-Response Modeling................................................................. 5-73 5.4.3.1. Modeling of Mesothelioma Exposure Response in the Libby Worker Cohort................................................................. 5-74 5.4.3.2. Mesothelioma Mortality Analysis in the Libby Worker Cohort........................................................................................ 5-76 5.4.3.3. Modeling of Lung Cancer Exposure Response in the Libby Worker Cohort........................................................................... 5-78 5.4.3.4. Lung-Cancer Mortality Analysis in the Libby Worker Cohort........................................................................................ 5-80 5.4.3.5. Summary of Mesothelioma and Lung Cancer Analysis of Libby Worker Cohort................................................................. 5-81 5.4.3.6. Analysis of Sub-Cohort of Employees Hired After 1959......... 5-83 5.4.4. Exposure Adjustments and Extrapolation Methods................................ 5-97 5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality..................................... 5-97 5.4.5.1. Unit Risk Estimates for Mesothelioma Mortality..................... 5-97 5.4.5.2. Unit Risk Estimates for Lung-Cancer mortality..................... 5-100 5.4.5.3. IUR Derivation for Combined Mesothelioma and Lung-Cancer Mortality............................................................ 5-102 5.4.5.4. Applications of the Combined Mesothelioma and Lung-cancer mortality IUR to Partial Lifetime Environmental Exposure Scenarios......................................... 5-109 This document is a draftfor review purposes only and does not constitute Agency policy. vii DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) 5.4.6. Uncertainties in the Cancer Risk Values............................................. 5-111 5.4.6.1. Sources of Uncertainty............................................................ 5-111 5.4.6.2. Summary................................................................................. 5-129 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND EXPOSURE RESPONSE...............................................................................................6-1 6.1. HUMAN IIAZARD POTENTIAL........................................................................... 6-1 6.1.1. Exposure.................................................................................................... 6-1 6.1.2. Fiber Toxicokinetics.................................................................................. 6-3 6.1.3. Noncancer Health Effects in Humans and Laboratory Animals............... 6-4 6.1.4. Carcinogenicity in Humans and Laboratory Animals............................... 6-6 6.1.5. Susceptible Populations............................................................................. 6-7 6.1.6. Mode-of-Action Information..................................................................... 6-7 6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard.................................. 6-8 6.2. EXPOSURE RESPONSE........................................................................................6-9 6.2.1. Noncancer/Inhalation................................................................................ 6-9 Uncertainty and Sensitivity Analyses for RfC Derivation..................... 6-12 6.2.2. Cancer/Inhalation.................................................................................... 6-16 6.2.2.1. Background and Methods......................................................... 6-16 6.2.3. Modeling of Mesothelioma Exposure Response.....................................6-19 6.2.4. Unit Risk Estimates for Mesothelioma Mortality................................... 6-20 6.2.5. Modeling of Lung Cancer Exposure Response....................................... 6-21 6.2.6. Unit Risk Estimates for Lung-Cancer Mortality.....................................6-22 6.2.7. IUR Derivation Based on Combined Mesothelioma and Lung-Cancer Mortality from Exposure to Libby Amphibole Asbestos................................................................................................. 6-24 6.2.7.1. Comparison with Other Published Studies of Libby, MT Workers Cohort.......................................................................... 6-25 6.2.8. Sensitivity Analysis................................................................................. 6-25 6.2.8.1. Sensitivity Analysis of Influence of High Exposures in Early 1960s on Model F it.......................................................... 6-25 6.2.8.2. Analysis of Potential Confounding of Lung Cancer Results by Smoking in the Sub-cohort................................................... 6-26 6.2.9. Uncertainty in the Cancer Risk Values................................................... 6-26 6.3. APPLICATION OF THE LIBBY AMPHIBOLE ASBESTOS RfC AND IUR. .. 6-28 6.3.1. Sites and Materials.................................................................................. 6-28 6.3.2. Exposure Units for Libby Amphibole Asbestos...................................... 6-29 6.3.3. Applications to Early Lifetime and Partial Lifetime Environmental Exposure Scenarios for IU R.................................................................. 6-30 6.3.4. Applications to Lifetime and Partial Lifetime Environmental Exposure Scenarios for RfC................................................................... 6-31 7. REFERENCES........................................................................................................................ 7-1 This document is a draftfor review purposes only and does not constitute Agency policy. viii DRAFT--DO NOT CITE OR QUOTE CONTENTS (continued) APPENDIX A: SUMMARY OF EXTERNAL PEER-REVIEW AND PUBLIC COMMENTS AND DISPOSITION............................................................... A-l APPENDIX B: PARTICLE SIZE DISTRIBUTION DATA FOR LIBBY AMPHIBOLE STRUCTURES OBSERVED IN AIR AT THE LIBBY ASBESTOS SUPERFUND SITE......................................................................................... B-l APPENDIX C: CHARACTERIZATION OF AMPHIBOLE FIBERS FROM ORE ORIGINATING FROM LIBBY, MT; LOUISA COUNTY, VA; AND PALABORA, REPUBLIC OF SOUTH AFRICA.......................................... C-l APPENDIX D: ANALYSIS OF SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN ANIMALS AND MECHANISTIC STUDIES......................................................................................................... D -l APPENDIX E: EVALUATION OF EXPOSURE RESPONSE DATA FOR DISCRETE PLEURAL THICKENING IN WORKERS FROM THE MARYSVILLE. 011COIIORT....................................................................... E-l APPENDIX F : MARYSVILLE, OH WORKER OCCUPATIONAL EXPOSURE RECONSTRUCTION.......................................................................................F - 1 APPENDIX G: EXTRA RISK AND UNIT RISK CALCULATION.................................... G-l APPENDIX H: GLOSSARY OF ASBESTOS TERMINOLOGY..........................................11-1 This document is a draftfor review purposes only and does not constitute Agency policy. ix DRAFT--DO NOT CITE OR QUOTE LIST OF TABLES 1-1. Derivation of the current IRIS inhalation unit risk for asbestos from the lifetime risk tables in the AAHAU...............................................................................................1-4 2- 1 Properties of vermiculite................................................................................................ 2-13 2- 2. Air sampling results for asbestos from ZonoliteVAI in three homes........................... 2-26 3- 1. Factors influencing fiber depositionand clearancein the respiratory system....................3-4 4-1. Exposure assessment methodologies used in evaluations of Libby, MT and Marysville, OH worker cohorts.......................................................................................4-5 4-2. Source of primary samples for fiber measurements at the Libby mining and milling operations............................................................................................................ 4-6 4-3. Dimensional characteristic of fibers from air samples collected in the vermiculite mill and screening plant, Libby, M T .............................................................................4-10 4-4. Respiratory (lung) cancer mortality and exposure-response analyses based on studies of the vermiculite mine workers in Libby, M T ................................................ 4-11 4-5. Mesothelioma mortality risk based on studies of the vermiculite mine workers in Libby, M T...................................................................................................................... 4-18 4-6. Nonmalignant respiratory mortality studies of the vermiculite mine workers in Libby, M T...................................................................................................................... 4-21 4-7. Chest radiographic studies of the Libby, MT vermiculite mine workers.....................4-24 4-8. Cancer mortality and nonmalignant respiratory disease mortality in the Libby, MT community..................................................................................................................... 4-30 4-9. Pulmonary function and chest radiographic studies in the Libby, MT community...... 4-31 4-10. Pulmonary function and chest radiographic studies of the Marysville, OH vermiculite processing plant workers............................................................................ 4-38 4-11. Prevalence of pleural radiographic abnormalities according to quartiles of cumulative fiber exposure in 280 participants...............................................................4-40 4-12. Prevalence of pleural thickening in 280 participants according to various cofactors.... 4-41 4-13. Description of study areas in ATSDR health consultations evaluating cancer incidence and mortality.................................................................................................. 4-45 4-14. Incidence and mortality results for potential asbestos-related cancers (by cancer site) in communities in the vicinity of vermiculite-processing facilities (with ATSDR health consultations evaluating potentialpathways of exposure)..................... 4-47 4-15. In vivo data following exposure to Libby Amphibole asbestos.................................... 4-50 4-16. In vivo data following exposure to tremolite asbestos................................................... 4-52 4-17. Autoimmune-related studies in the Libby, MT community.......................................... 4-61 4-18. In vitro data following exposure to Libby Amphibole asbestos.................................... 4-64 4- 19. In vitro data following exposure to tremolite asbestos.................................................. 4-65 5-1. Summary of candidate principal studies on Libby Amphibole asbestos for reference concentration (RfC) derivation........................................................................ 5-4 5-2. Summary of rationale for identifying candidate principal studies on Libby Amphibole asbestos for RfC development...................................................................... 5-8 5-3. Distribution of cases and time from first exposure (7) for cohort of Marysville workers........................................................................................................................... 5-28 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) 5-4. Candidate models for association between cumulative Libby Amphibole asbestos exposure in the Marysville sub-cohort and localized pleural thickening.....................5-33 5-5. Modeling of alternative endpoints in the Marysville worker cohort members examined in 2002-2005................................................................................................5-51 5-6. Demographic and exposure characteristics of the Libby worker cohort......................5-55 5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of operations through 1982 [Amandus et al. (1987a); Table V II]....................................5-60 5-8. Demographic and exposure characteristics of the subset of the Libby worker sub-cohort hired after 1959............................................................................................ 5-65 5-9. Comparison of univariate model fit of various exposure metrics for mesothelioma mortality in the full Libby worker cohort (n = 1,871).................................................. 5-77 5-10. Comparison of model fit of exposure metrics for mesothelioma mortality in the sub-cohort hired after 1959. Only the model fits with information weights greater than 0.010 are shown..................................................................................................... 5-85 5-11. Mesothelioma mortality exposure metrics fits, slopes, and credible intervals.............5-86 5-12. Model fit comparison for different exposure metrics and lung-cancer mortality associated with Libby Amphibole asbestos, controlling for age, gender, race, and date of birth.................................................................................................................... 5-88 5-13. Lung-cancer mortality exposure metrics fits, slopes, and confidence intervals for all retained metrics from Table 5-12. Subset of lung cancer models with lagged exposures that yielded statistically significant model fit (p < 0.05) and exposure metric fit (p < 0.05) to the epidemiologic data.............................................................5-92 5-14. Sensitivity analysis of model fit comparison for different exposure metrics and lung-cancer mortality associated with Libby Amphibole asbestos. Estimated exposure intensities for all jobs during 1960-1963 were reduced by 50%.................. 5-94 5-15. Sensitivity analysis of model fit comparison for different exposure metrics and lung-cancer mortality associated with Libby Amphibole asbestos, controlling for age, gender, race, and date of birth. Estimated exposure intensities for all jobs during 1960-1963 were reduced by 50%. Lung cancer models presented include those with statistically significant multivariate model /i-value and nonzero lag in exposure..........................................................................................................................5-95 5-16. Mesothelioma mortality exposure metrics unit risks..................................................... 5-98 5-17. Adjusted for underascertainment unit risks for the sub-cohort hired after 1959 corresponding to the different metrics........................................................................... 5-99 5-18. Unit risks for subset of lung cancer models with lagged exposures that yielded statistically significant model fit (p < 0.05) and exposure metric fit (p < 0.05) to the epidemiologic data................................................................................................. 5-101 5-19. Reasonable upper bound and lowest information criteria estimates of central risks and unit risks, per fibers/cc, for mesothelioma mortality, lung-cancer mortality, and the IUR for the combined mortality risk from mesothelioma and lung cancer .... 5-104 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) 5-20. Lung cancer regression results from different analyses of cumulative exposure in the cohort of workers in Libby, MT. All analyses used NIOSH-collected exposure data but used different cohort definitions, lengths of follow-up, and lengths of exposure lags to account for cancer latency...............................................5-106 5- 21. Mesothelioma regression results from different analyses of cumulative exposure in the cohort of workers in Libby, MT. All analyses used NIOSH-collected exposure data but different cohort definitions, lengths of follow-up, and lengths of exposure lags to account for cancer latency...............................................................5-110 6-1. Reasonable upper bound and lowest information criteria estimates of central risks and unit risks, for mesothelioma mortality, lung-cancer mortality, and the IUR for the combined mortality risk from mesothelioma and lung cancer (IURs are presented in the units of excess cancers per fibers/cc, continuous lifetime exposure)......................................................................................................................6-24 1 This document is a draftfor review purposes only and does not constitute Agency policy. xii DRAFT--DO NOT CITE OR QUOTE 1 LIST OF FIGURES 2-1. Vermiculite mining operation on Zonolite Mountain, Libby, Montana......................... 2-1 2-2. Expanded vermiculite and vermiculite attic insulation (VAI) shown in place between ceiling joists...................................................................................................... 2-2 2-3. Nationwide distribution of Libby ore by county.............................................................. 2-3 2-4. Structure of the silicate minerals, illustrating silicate subclasses by the linking of the basic silicon tetrahedron into more complex structures............................................2-6 2-5. Cross-section of amphibole fibers showing the silicon tetrahedrons that make up each double-chain plate...................................................................................................2-8 2-6. Comparison of crystalline forms amphibole minerals......................................................2-9 2-7. Vermiculite ore sample...................................................................................................2-13 2-8. Mineralogy of Libby Amphibole asbestos structures from samples taken from the Zonolite Mountain site...................................................................................................2-15 2-9. Solution series linking tremolite, winchite, and richterite amphibole fibers.................. 2-16 2-10. Scanning electron microscope image of amphibole mineral structures from the Libby, MT mine.............................................................................................................2-17 2-11. Fiber morphology of amphibole asbestos from the Libby, MT mine viewed under a transmission electron microscope................................................................................2-18 2- 12. Particle size of fibers in Libby ore and Libby air........................................................... 2-19 3- 1. General scheme for fiber deposition, clearance, and translocation of fibers from the lung and GI tract......................................................................................................... 3-2 4-1. Location of 28 sites included in the Phase 1 community evaluations conducted by AT SDR........................................................................................................................... 4-45 5- 1. Estimated and measured exposure concentrations in Marysville, OH facility...............5-24 5-2. Graph of observed and estimated prevalence of localized pleural thickening calculated using the Michaelis-Menten model with 10-year lagged exposure...............5-35 5-3. Plot of the NIOSH job-exposure matrix for different job categories over tim e.............5-59 5-4. Histogram showing the number of workers who experienced each incremental number of different jobs among the 880 workers hired after 1959...............................5-66 This document is a draftfor review purposes only and does not constitute Agency policy. xiii DRAFT--DO NOT CITE OR QUOTE 2-D 3-D AAHAU AIC AM APC AT SDR BMC BMCL BMD BMDL BMI BMR cc CDF CE CHEEC Cl COPD COX-2 CYP DHE DIC DLCO DNA ECX ECSOD EDX EPA EPMA FEV1 FVC GSH LIST OF ABBREVIATIONS AND ACRONYMS two-dimensional three-dimensional Airborne Asbestos Health Assessment Update Akaike information criterion amosite antigen-presenting cells Agency for Toxic Substances and Disease Registry benchmark concentration lower 95% confidence limit of the benchmark concentration benchmark dose lower 95% confidence limit of the benchmark dose body mass index benchmark response cubic centimeter cumulative distribution frequency cumulative exposure cumulative human equivalent exposure for continuous exposure confidence interval chronic obstructive pulmonary disease cyclooxygenase-2 cytochrome P450 dehydroergosterol deviance information criterion single breath carbon monoxide diffusing capacity deoxyribonucleic acid effective concentration extracellular superoxide dismutase energy dispersive x-ray analysis U.S. Environmental Protection Agency electron probe microanalysis forced expiratory volume in one second forced vital capacity glutathione This document is a draftfor review purposes only and does not constitute Agency policy. xiv DRAFT--DO NOT CITE OR QUOTE GST HAEC HKNM HO HPRT HRCT HTE IARC ICD ICRP IFN IH IL ILO IRIS IUR JEM LAA LECX LDH MCMC MESA MIP-2 MnSOD MOA mppcf MPPD MSHA NAT2 NCHS NDI Nf2 NIEHS LIST OF ABBREVIATIONS AND ACRONYMS (continued) glutathione S-transferase human airway epithelial cells human pleural mesothelial cells heme oxygenase hypoxanthine-guanine phosphoribosyltransferase high resolution computed tomography hamster tracheal epithelial International Agency for the Research on Cancer International Classification of Diseases International Commission Radiological Protection interferon industrial hygiene interleukin International Labour Organization Integrated Risk Information System inhalation unit risk job exposure matrix Libby Amphibole asbestos lowest effective concentration lactate dehydrogenase Markov chain Monte Carlo Mining Enforcement and Safety Administration macrophage inflammatory protein-2 manganese superoxide dismutase mode of action million particles per cubic foot multipath particle dosimetry Mine Safety and Health Administration N-acetyl-transferase 2 National Center for Health Statistics National Death Index neurofibromatosis 2 National Institute of Environmental Health Sciences This document is a draftfor review purposes only and does not constitute Agency policy. xv DRAFT--DO NOT CITE OR QUOTE LIST OF ABBREVIATIONS AND ACRONYMS (continued) NIOSH NMRD NRC NTP NVSR Oggl OSHA PARP PBS PCM PCMe PHS PM2.5 POD RfC RID RNA ROS RPM RR RT-PCR RTW SE SEER SEM SH SHE SHHF SIR SMR SOD SPF SRR National Institute for Occupational Safety and Health non-malignant respiratory disease National Research Council National Toxicology Program National Vital Statistics Report 8-oxoguanine-DNA-glycosylase 1 Occupational Safety and Health Administration poly(ADP-ribose)polymerase phosphate buffered saline phase contrast microscopy phase contrast microscopy equivalent Public Health Service particulate matter 2.5 pm diameter or less point of departure reference concentration reference dose ribonucleic acid reactive oxygen species rat pleural mesothelial relative risk reverse transcription polymerase chain reaction residence time-weighted Standard Error Surveillance, Epidemiology, and End Results scanning electron microscopy spontaneously hypertensive Syrian hamster embryo spontaneously hypertensive-heart failure standardized incidence ratio standardized mortality ratio superoxide dismutase specific-pathogen-free standardized rate ratio 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 (continued) SSA/Ro52 STEM SV40 TEM TLC TSFE TWA UCL UF USGS VAI WHO WKY XRCC1 XRD autoantibody marker for apoptosis scanning transmission electron microscopy simian virus 40 transmission electron microscopy total lung capacity time since first exposure time-weighted average upper confidence limit uncertainty factor United States Geological Survey vermiculite attic insulation World Health Organization Wistar-Kyoto rat X-ray repair cross complementing protein 1 X-ray diffraction This document is a draftfor review purposes only and does not constitute Agency policy. xvii DRAFT--DO NOT CITE OR QUOTE FOREWORD The purpose of this Toxicological Review is to provide scientific support and rationale for the hazard and dose-response assessment in IRIS pertaining to chronic inhalation exposure to Libby Amphibole asbestos, a unique mixture of asbestos fibers originating from the vermiculite mine near Libby, MT. It is not intended to be a comprehensive treatise on the agent or toxicological nature of Libby Amphibole asbestos. The purpose of this document is to establish a Libby Amphibole asbestos-specific reference concentration to address noncancer health effects and to characterize the carcinogenic potential and establish an inhalation unit risk for Libby Amphibole asbestos-related lung cancer and mesothelioma mortality. The intent of Section 6, Major Conclusions in the Characterization o fHazard and Exposure Response, is to present the significant conclusions reached in the derivation of the reference dose, reference concentration, and cancer assessment where applicable, and to characterize the overall confidence in the quantitative and qualitative aspects of hazard and dose response by addressing the quality of data and related uncertainties. The discussion is intended to convey the limitations of the assessment and to aid and guide the risk assessor in the ensuing steps of the risk assessment process. For other general information about this assessment or other questions relating to IRIS, the reader is referred to EPA's IRIS Hotline at (202) 566-1676 (phone), (202) 566-1749 (fax), or hotline.iris@epa.gov (e-mail address). This document is a draftfor review purposes only and does not constitute Agency policy. xviii DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS CHEMICAL MANAGERS/AUTHORS Thomas F. Bateson, ScD, MPH National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Robert Benson, PhD Region 8 Office of Partnerships and Regulatory Assistance U.S. Environmental Protection Agency Denver, CO Danielle DeVoney, PhD, DABT, PE Captain, U.S. Public Health Service National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC AUTHORS Krista Yorita Christensen, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Glinda Cooper, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Rebecca Dzubow, MPH, MEM Office of Children's Health Protection U.S. Environmental Protection Agency Washington, DC Maureen R. Gwinn, PhD, DABT National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC This document is a draftfor review purposes only and does not constitute Agency policy. xix DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued) Leonid Kopylev, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC CONTRIBUTING AUTHORS David Berry, PhD Region 8 U.S. Environmental Protection Agency Denver, CO Malcolm Field, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Patricia Sullivan, ScD Division of Respiratory Disease Studies National Institute for Occupational Safety and Health Morgantown, WV CONTRIBUTORS David Bussard National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Samantha J. Jones, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Babasaheb Sonawane, PhD National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC Paul White National Center for Environmental Assessment U.S. Environmental Protection Agency Washington, DC This document is a draftfor review purposes only and does not constitute Agency policy. xx DRAFT--DO NOT CITE OR QUOTE AUTHORS, CONTRIBUTORS, AND REVIEWERS (continued) CONTRACTOR SUPPORT William Brattin, PhD Syracuse Research Corporation Denver, CO ECFlex, Inc., Fairborn, OH Heidi Glick Debbie Kleiser Crystal Lewis Lana Wood IntelliTech Systems, Inc., Fairborn, OH Cris Broyles Kathleen Secor REVIEWERS This document has been provided for review to EPA scientists and interagency reviewers from other federal agencies and White House offices. This document is a draftfor review purposes only and does not constitute Agency policy. xxi DRAFT--DO NOT CITE OR QUOTE 1 1. INTRODUCTION 2 This document presents background information and justification for the Integrated Risk 3 Information System (IRIS) Summary of the hazard and exposure-response assessment of Libby 4 Amphibole asbestos,1*a mixture of amphibole fibers identified in the Rainy Creek complex and 5 present in ore from the vermiculite mine near Libby, MT. IRIS Summaries may include oral 6 reference dose (RfD) and inhalation reference concentration (RfC) values for chronic and other 7 exposure durations, and a carcinogenicity assessment. This assessment reviews the potential 8 hazards, both cancer and noncancer health effects, from exposure to Libby Amphibole asbestos 9 and provides quantitative information for use in risk assessments: an RfC for noncancer and an 10 inhalation unit risk addressing cancer risk. Libby Amphibole asbestos-specific data are not 11 available to support RfD or cancer slope factor derivations for oral exposures. 12 An RfC is typically defined as "an estimate (with uncertainty spanning perhaps an order 13 of magnitude) of a continuous inhalation exposure to the human population (including sensitive 14 subgroups) that is likely to be without an appreciable risk of deleterious effects during a 15 lifetime." In the case of Libby Amphibole asbestos, the RfC is expressed in terms of the lifetime 16 exposure in units of fibers per cubic centimeter of air (fibers/cc) in units of the fibers as 17 measured by phase contrast microscopy (PCM). The inhalation RfC for Libby Amphibole 18 asbestos considers toxic effects for both the respiratory system (portal-of-entry) and for effects 19 peripheral to the respiratory system (extrarespiratory or systemic effects) that may arise after 20 inhalation of Libby Amphibole asbestos. In this assessment, the estimates of hazard are derived 21 from modeling cumulative exposures from human data, and thus for exposures of less than a 22 lifetime the risk assessor should calculate a lifetime average concentration to compare to the 23 RfC. 24 The carcinogenicity assessment provides information on the carcinogenic hazard 25 potential of the substance in question, and quantitative estimates of risk from inhalation 26 exposures are derived. The information includes a weight-of-evidence judgment of the 27 likelihood that the agent is a human carcinogen and the conditions under which the carcinogenic 28 effects may be expressed. Quantitative risk estimates are derived from the application of a low- 1 The term "Libby Amphibole asbestos" is used in this document to identify the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfor review purposes only and does not constitute Agency policy. 1-1 DRAFT--DO NOT CITE OR QUOTE 1 dose extrapolation procedure from human data. An inhalation unit risk (IUR) is typically 2 defined as a plausible upper bound on the estimate of cancer risk per pg/nv air breathed for 3 70 years. For Libby Amphibole asbestos, the RFC is expressed as a Lifetime Daily Exposure in 4 fibers/cc (in units of the fibers as measured by PCM), and the IUR is expressed as cancer risk per 5 fibers/cc (in units of the fibers as measured by PCM). 6 Development of these hazard identification and exposure-response assessments for Libby 7 Amphibole asbestos has followed the general guidelines for risk assessment as set forth by the 8 National Research Council (1983). U.S. Environmental Protection Agency (EPA) Guidelines 9 and Risk Assessment Forum technical panel reports that may have been used in the development 10 of this assessment include the following: Guidelinesfor the Health Risk Assessment o f Chemical 11 Mixtures (U.S. EPA. 1986c). Guidelinesfor Mutagenicity Risk Assessment (U.S. EPA. 1986b). 12 Recommendationsfor and Documentation o fBiological Valuesfor Use in Risk Assessment (U.S. 13 EPA. 1988b). Guidelinesfor Developmental Toxicity'Risk Assessment (U.S. EPA. 1991a). 14 Interim Policyfor Particle Size and Limit Concentration Issues in Inhalation Toxicity (U.S. EPA, 15 1994a). Methodsfor Derivation o fInhalation Reference Concentrations and Application o f 16 Inhalation Dosimetry (U.S. EPA. 1994b). Use o f the Benchmark Dose Approach in Health Risk 17 Assessment (\JAs. EPA. 1995). Guidelinesfor Reproductive Toxicity Risk Assessment (U.S. EPA. 18 1996). Guidelinesfor Neurotoxicity Risk Assessment (U.S. EPA. 1998). Science Policy Council 19 Handbook. Risk Characterization (U.S. EPA. 2000c). Benchmark Dose Technical Guidance 20 Document (U.S. EPA. 2000a). Supplementary Guidancefor Conducting Health Risk Assessment 21 o f Chemical Mixtures (U.S. EPA. 2000d). A Review o f the Reference Dose and Reference 22 Concentration Processes (U.S. EPA. 2002). Guidelinesfor Carcinogen Risk Assessment (U.S. 23 EPA. 2005a). Supplemental Guidancefor Assessing Susceptibilityfrom Early-Life Exposure to 24 Carcinogens (U.S. EPA. 2005b). Science Policy Council Handbook: Peer Review (US. EPA. 25 2006d). and A Frameworkfor Assessing Health Risks o fEnvironmental Exposures to Children 26 (U.S. EPA. 2006b). 27 The literature search strategy employed for this assessment is based on EPA's National 28 Center for Environmental Assessment's Health and Environmental Research Outline database 29 tool (which includes PubMed, MEDLINE, Web of Science, JSTOR, and other literature 30 sources). The key search terms included the following: Libby Amphibole, tremolite, asbestos, 31 richterite, winchite, amphibole, and Libby, MT. The relevant literature was reviewed through This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 1-2 DRAFT--DO NOT CITE OR QUOTE 1 July 2011. Any pertinent scientific information submitted by the public to the IRIS Submission 2 Desk was also considered in the development of this document. 3 4 1.1. RELATED ASSESSMENTS 5 1.1.1. IRIS Assessment for Asbestos (U.S. EPA, 1988a) 6 The IRIS assessment for asbestos was posted online in IRIS in 1988 and includes an IUR 7 of 0.23 excess cancers per 1 fiber/cc (TJ.S. EPA. 1988a) (this unit risk is given in units of the 8 fibers as measured by PCM). The IRIS IUR for general asbestos is derived by estimation of 9 excess cancers for a continuous lifetime exposure and is based on the central tendency--not the 10 upper bound--of the risk estimates (U.S. EPA. 1988a) and is applicable to exposures across a 11 range of exposure environments and types of asbestos (CAS Number 1332-21-4). Although 12 other cancers have been associated with asbestos (e.g., laryngeal, stomach, ovarian) (Straif et al.. 13 2009), the IRIS IUR for asbestos accounts for only lung cancer and mesothelioma. Additionally, 14 pleural and pulmonary effects from asbestos exposure (e.g., localized pleural thickening, 15 asbestosis, and reduced lung function) are well documented, though, currently, there is no RfC 16 for these noncancer health effects. 17 The derivation of the unit risk for general asbestos is based on the Airborne Asbestos 18 Health Assessment Update (AAHAU) (U.S. EPA. 1986a). The AAHAU provides various cancer 19 potency factors and mathematical models of lung cancer and mesothelioma mortality based on 20 synthesis of data from occupational studies and presents estimates of lifetime cancer risk for 21 continuous environmental exposures (0.0001 fiber/cc and 0.01 fiber/cc) (U.S. EPA. 1986a) (see 22 Table 6-3). For both lung cancer and mesothelioma, life-table analysis was used to generate risk 23 estimates based on the number of years of exposure and the age at onset of exposure. Although 24 various exposure scenarios were presented, the unit risk is based on a lifetime continuous 25 exposure from birth. The final asbestos IUR is 0.23 excess cancer per 1 fiber/cc continuous 26 exposure2 and was established by the EPA Carcinogen Risk Assessment Verification Endeavor 27 workgroup and posted on the IRIS database in 1988 (U.S. EPA. 1988a) (see Table 1-1). 28 'An IUR o f 0.23 can be interpreted as a 23% increase in lifetime risk o f dying from mesothelioma or lung cancer with each 1 fiber/cc increase in continuous lifetime exposure. This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 1-3 DRAFT--DO NOT CITE OR QUOTE 1 Table 1-1. Derivation of the current IRIS inhalation unit risk for asbestos 2 from the lifetime risk tables in the AAHAU 3 Excess deaths per 100,000a Gender Mesothelioma Lung cancer Total Risk Unit risk Female 183 35 218.5 2.18 x 10 Male 129 114 242.2 2.42 x 10 All 156 74 230.3 2.30 x 10 0.23 4 5 aData are for exposure at 0.01 fibers/cc for a lifetime. 6 AAHAU = Airborne Asbestos Health Assessment Update. 7 Source: U .S.E P A (1988a). 8 9 10 1.1.2. EPA Health Assessment for Vermiculite (1991b) 11 An EPA health assessment for vermiculite reviewed available health data, including 12 studies on workers who mined and processed ore with no significant amphibole fiber content. 13 The cancer and noncancer health effects observed in the Libby, MT worker cohort were not seen 14 in studies of workers exposed to vermiculite from mines with similar exposure to vermiculite but 15 much lower exposures to asbestos fibers. Therefore, it was concluded that the health effects 16 observed from the materials mined from Zonolite Mountain near Libby, MT, were most likely 17 due to amphibole fibers not the vermiculite itself (U.S. EPA. 1991b). At the time, EPA 18 recommended the application of the IRIS 1TIRfor asbestos fibers (0.23 per fiber/cc) in 19 addressing potenti al risk of the amphibole fibers entrai ned in vermicul ite mined in Libby, MT. 20 21 1.2. LIBBY AMPHIBOLE ASBESTOS-SPECIFIC HUMAN HEALTH ASSESSMENT 22 Libby Amphibole asbestos is a complex mixture of amphibole fibers--both 23 mineralogically and morphologically (see Section 2.2). The mixture primarily includes 24 tremolite, winchite, and richterite fibers with trace amounts of magnesioriebeckite, edenite, and 25 magnesio-arfvedsonite. These fibers exhibit a complete range of morphologies from prismatic 26 crystals to asbestiform fibers (Meeker et al.. 2003). Epidemiologic studies of workers exposed to 27 Libby Amphibole asbestos fibers indicate increased lung cancer and mesothelioma, as well as 28 asbestosis, and other nonmalignant respiratory diseases (Larson et ah, 2010b; Larson et al.. 29 2010a; Moolgavkar et al.. 2010; Rohs et a f. 2008; Sullivan. 2007; McDonald et al. 2004. 2002; This document is a draftfor re\'iew purposes only and does not constitute utgencvpolicy. 1-4 DRAFT--DO NOT CITE OR QUOTE 1 Amandus et al., 1988; Amandus et aL 1987a; Amandus and Wheeler, 1987; Amandus et al, 2 1987b: McDonald et a).., 1986a: McDonald et al., 1986b: Lockev et al, 1984). 3 The IRIS database has an IUR3for asbestos based on a synthesis of 14 epidemiologic 4 studies that included occupational exposure to chrysotile, amosite, or mixed mineral exposures 5 (chrysotile, amosite, crocidolite) (U.S. EPA, 1988a. 1986a). There is some uncertainty in 6 applying the resulting IUR for asbestos to exposure environments and minerals different from 7 those analyzed in the AAHAU (U.S. EPA. 1986a). There is currently no RfC, RfD, or oral slope 8 factor derived for asbestos on the IRIS database. 3For purposes o f this document, tenned "IRIS IUR." This document is a draftfor review purposes only and does not constituteAgency policy. 1-5 DRAFT--DO NOT CITE OR QUOTE 1 2. LIBBY AMPHIBOLE ASBESTOS: GEOLOGY, USE, AND EXPOSURE POTENTIAL 2 2.1. HISTORICAL BACKGROUND 3 The term Libby Amphibole asbestos4 refers to various mineral forms of amphibole 4 asbestos found in the rocks and ore of Zonolite Mountain, 6 miles northeast of Libby, MT (see 5 Figure 2-1). Zonolite Mountain contains a large vermiculite deposit that has been mined since 6 the early 1920s for various commercial uses. Vermiculite miners, mill workers, and those 7 working in the processing plants were exposed to these amphibole fibers, which remain within 8 the vermiculite ore and product. As amphibole asbestos is present in the geological deposit from 9 which the vermiculite ore was being mined, workers were exposed to asbestos fibers during 10 various activities such as extracting ore from the mine, transporting ore and waste rock, milling 11 operations, and shipping the final product (Meeker et al., 2003; Amandus et al.., 1987a; 12 McDonald et al.. 1986a). Mortality and morbidity studies on the mine and mill workers from 13 Libby have reported adverse health effects in these workers including lung cancer, mesothelio 14 ma, nonmalignant respiratory disease (NMRD; e.g., asbestosis), and pleural abnormalities 15 (McDonald et al.., 2004; Amandus and Wheeler. 1987; Amandus et al... 1987b; McDonald et al., 16 1986a; McDonald et al.. 1986b);Sullivan. 2007, 709497;Larson, 2010, 711560;Moolgavkar, 17 2010, 709457}. Pleural abnormalities and 18 signs of interstitial fibrosis have also been 19 reported in workers exfoliating and 20 processing expanded Libby vermiculite in 21 other facilities (Rohs et al.., 2008; Lockev et 22 al,, 1984V 23 The primary commercial product from 24 the Zonolite mining operation was vermiculite 25 concentrate, which is produced by screening 26 and grading the ore to enrich for the raw . . ii-i Figure 2-1. Vermiculite mining operation on 27 vermiculite mineral. The unexpanded mineral Zonolite Mountain, Libby, Montana. 4The term "Libby Amphibole asbestos" is used in this document to identity the mixture of amphibole mineral fibers of varying elemental composition (e.g., winehite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfor review purposes only and does not constitute Agency policy. 2-1 DRAFT--DO NOT CITE OR QUOTE 1 exhibits a sheetlike structure that is seen in related minerals (e.g., mica) (see Figure 2-2). (b) 2 3 Figure 2-2. Expanded vermiculite (a) and veriniculite attic insulation (b) 4 (VAI) shown in place between ceiling joists. 5 6 T When heated to approximately 150JC, the vermiculite mineral expands like popcorn into 8 a light porous material. This process of expanding the mineral ore is termed "exfoliation" or 9 "popping" and occurs when the silicate sheets within the ore are rapidly dehydrated by applying 10 high heat. Libby Amphibole asbestos fibers were released during the energetic and other kinetic 11 processing of the ore and vermiculite concentrate, potentially exposing workers. 12 A portion of the vermiculite concentrate was exfoliated in Libby, MT and either used 13 locally or packaged and shipped for use elsewhere. However, most of the vermiculite 14 concentrate was transported across the country and elsewhere to expansion plants where it was 15 exfoliated and distributed. The Agency for Toxic Substances and Disease Registry (ATSDR. 16 2008b) has surveyed 28 of these facilities, identifying potential community exposures both to 17 amphibole asbestos fibers from the vermiculite concentrate before exfoliation, during exfoliation, 18 and during processing and in waste rock from the processing plants (see Section 4.1.4 and 19 Figure 2-3). Vermiculite from the Libby, MT mine was used commercially from the 1920s to 20 1990, and a review of company records from 1964-1990 indicates that approximately 21 6,109,000 tons of vermiculite concentrate was shipped to over 200 facilities (ATSDR. 2008b). 22 Expanded vermiculite from the Libby, MT site was used in numerous consumer and construction 23 products: including attic insulation, packing material, and soil conditioners, and in the production This document is a draftfor review purposes only and does not constitute Agency policy. 2-2 DRAFT--DO NOT CITE OR QUOTE Total tonnage by county 30D.GDD w more 2DD.0DDto 200,900 o 100,000 to 168,900 1 o t to 00,999 2 Figure 2-3. Nationwide distribution of Libby ore by county (in tons). Data 3 on the distribution of ore are based on approximately 80,000 invoices that EPA 4 obtained from W.R. Grace that document shipments of vermiculite ore made from 5 the Libby mine between 1964 to 1990. EPA tabulated this shipping information 6 in a database. 7 8 Source: U.S. GAO (2007). 9 10 n of gypsum wall board. There is also potential for exposure to Libby Amphibole asbestos in these 12 products (see Section 2.4). 13 14 2.2. GEOLOGY AND MINERALOGY OF LIBBY AMPHIBOLE ASBESTOS 15 A large vermiculite deposit is located on Zonolite Mountain, northeast of Libby, MT, 16 within a geologic unit known as the Rainy Creek complex. Geologic processes within the Rainy 17 Creek complex have resulted in the formation of fibrous amphiboles adjacent to igneous 18 intrusions into the complex (veins and dikes of alkaline granite, pegmatite, and quartz) 19 (Boettcher et ah. 1996). The amphibole fibers identified fall within the tremolite-richterite- This document is a draftfor re\'iew purposes only and does not constitute utgencvpolicy. 2-3 DRAFT--DO NOT CITE OR QUOTE 1 magnesioriebecktite solid solution series (e.g., winchite, richterite, and tremolite) (Meeker et a l , 2 2003). An appropriate understanding of the mineralogy and geology of these materials is helpful 3 in defining the mineral fibers in Libby Amphibole asbestos. 4 Geological terms provide fiber and mineral definitions based on habit of formation and 5 fiber morphology. Conversely, the analytical methods that have been used to count fibers in air 6 samples, in both historical and current exposure environments, define microscopic fibers based 7 on dimensional characteristics and mineralogy (depending on the analytical method). Current 8 analytical methods do not have specific procedures for determining fiber morphology at the 9 microscopic level. Because the human and experimental animal data on adverse health effects of 10 asbestos rely on available analytical methods to document exposure, these definitions are 11 relevant to determining what constitutes a fiber for this health assessment. Therefore, available 12 data on the fiber morphology and fiber-size distribution of Libby Amphibole asbestos are 13 presented in the following sections. 14 15 2.2.1. Silicate Minerals 16 Silicate minerals are basically made up of oxygen and silicon, two of the most abundant 17 elements in the Earth's crust. Approximately 25% of known minerals and 40% of the common 18 minerals are silicates. Silicate minerals are hard, infusible, and have very low solubility in strong 19 mineral acids. Specific gravity ranges from fairly light to intermediate, luster is commonly 20 glassy, and most crush to a light powder even when the bulk specimen is black prior to crushing. 21 Silicates chiefly occur as components of rocks, segregations in rocks, or crystals lining cavities 22 in rocks. Most hard silicates are primary minerals (i.e., mineral forms that have not undergone 23 oxidative weathering). Secondary silicates have undergone oxidative weathering and contain 24 water of hydration (Dana et al,, 1977). Silicate minerals can be defined by chemical structure, 25 crystal structure, trace minerals, and habit of formation. 26 The basic chemical unit of silicate crystalline structure is the [SiCL]4- tetrahedron-shaped 27 anionic group. The basic unit consists of four oxygen molecules at the apices of a regular 28 tetrahedron surrounding and coordinated with one silicon ion (Si1 ) at the center. The chemistry 29 is such that the oxygen molecules can bond to another silicon ion and, therefore, link one 30 [S1O4]1 tetrahedron to another, and then another, and so forth by the process of polymerization . 31 The silicates can form as single tetrahedrons, double tetrahedrons, chains, sheets, rings and This document is a draftfor re\'iew purposes only and does not constitute utgencvpolicy. 2-4 DRAFT--DO NOT CITE OR QUOTE 1 framework structures (see Figure 2-4). More complex three-dimensional structures tectosilicates 2 (frameworks) may also form mineral fibers (e.g., erionite). 3 Each subclass of silicates has many mineral members. Specific minerals are defined by 4 the structure, chemistry, and morphology of the mineral. The minerals of interest in this 5 assessment are various forms of amphiboles (double-chain inosilicates) and vermiculite (a 6 phyllosilicate) (see Figure 2-4). 7 8 2.2.1.1. Mineralogy and Structure o fAmphiboles 9 The mineralogy of amphiboles is important to understanding which mineral forms are 10 present in the Libby vermiculite mine, and, therefore, considered to be Libby Amphibole 11 asbestos. Amphibole minerals are double-chain inosilicates, meaning the chemical building 12 block for amphiboles is connected chains of the silicon tetrahedron (see Figure 2-4c). 13 Amphiboles form when edge-shared octahedra link two of the double-chain [SiCL]4- plates (see 14 Figure 2-4d). The specific cations between the two double-chain plates define the elemental 15 composition of the mineral, while the ratio of these cations in each location is used to classify 16 amphiboles within a solid-solution series. The cation sites are designated as A, B, and C in 17 Eq. 2-1, which shows the general chemical formula for double-chain inosilicate amphiboles. The 18 Libby Amphibole asbestos is a complex mixture of mineral forms defined by the cation ratios in 19 each site (further discussed in Section 2.2.3). 20 21 22 ^0-1^2(^578022(014, F, 0)2 Eq. 2-1 23 24 where: 25 26 A = Na, K 27 B = Na, Li, Ca, Mn, Fe2+, Mg 28 C = Mg, Fe2+, Mn, Al, Fe3+, Ti 29 T= Si, Al. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 2-5 DRAFT--DO NOT CITE OR QUOTE (a) N esosilicates or single tetra h ed ro n . The single tetrahedron com prises four oxygen m olecules covalently bound to the silicon, at th e ce n te r o f th e | Si 0 414 -tetrahedron. (b) In o silic a te s [ino (gr.) = th re a d ] - Single-chain silicates. Chain silicates are realized by linking |S i0 4|4 -tetrahedrons in a w ay to form continuous chains. They can be re p resen te d b y a c o m p o sitio n o f [ S i 0 3]2 . A typical exam ple is diopside C aM g| SnOg], in w hich the "endless" chains are also held together by Ca2 and M g"'ions. (c) Inosilicates - D ouble-chain silicates. Tw o silicate chai ns o f the inosilicates are linked at the com ers, form ing double-chains and yielding |Si :0 | ions, as realized in the trem olite-ferro-actinolite series C a2(M g,Fe)5Si80 2 2 (0 H )2. D o u b le-ch ain silicates are com m only grouped w ith the single-chain inosilicates. (d) P h y llo silic a te s \phyllo (gr.) = sheet] o r sheet silicates. These are form ed if the d o u b le-c h ain in o silicate [Si40 11]6 cltM ls are linked to form continuous sheets with the ch e m ic al fo rm u la [S i20 5]2 . E x am p les o f sheet silicates include chrysotile M g 3S i2O s(O H ) an d v erm ic u lite [(M g, F e ,A )3(A l,S i)2 O 10(O H ) 2 4 H 20 ] , 1 2 3 Figure 2-4. Structure of the silicate minerals, illustrating silicate subclasses 4 by the linking of the basic silicon tetrahedron (a) into more complex 5 structures (b, c, or d). 6 This document is a draftfo r review purposes only and does not constitute Agency policy. 2-6 DRAFT--DO NOT CITE OR QUOTE 1 The mineral subgroup within amphiboles is determined by the elemental composition. 2 3 4 Tremolite subgroup (Ca amphiboles) 5 Anthophyllite subgroup (Fe-Mg-Li orthoamphiboles) 6 Richterite subgroup (Ca-Na amphiboles) 7 Cummingtonite (Fe-Mg-Li clinoamphiboles) 8 9 10 A solid solution series includes a continuum of minerals with different cation 11 composition for each site. Solid solution series are defined by their end-members, where mineral 12 terminology can change as the proportion of cations changes within the crystalline structure. For 13 example, a solid solution series for the cation Site A will have one end-member with 100% 14 sodium ions and one end-member with 100% potassium ions. This series would include all 15 intervening ratios. Because each cation site has multiple possibilities, the elemental composition 16 of the amphibole silicates can be quite complex. It is the complexity of the amphiboles that 17 historically has given rise to a proliferation of mineral names with no systematic basis 18 (Hawthorne. 1981). Currently, amphiboles are identified by a clear classification scheme based 19 on crystal chemistry that uses well-established names based on the basic mineralogy, with 20 prefixes and adjective modifiers indicating the presence of substantial substitutions that are not 21 essential constituents of the end-members (Leake et al.. 1997). The mineral classification system 22 does not designate certain amphibole mineral as asbestos. However, some mineral designations 23 have traditionally been considered asbestos (e.g., tremolite, anthophyllite.) Other commercial 24 forms of asbestos were known by trade names (i.e., amosite) rather than mineralogtcal 25 terminology (i.e., an amphibole mineral in the cummingtonite-grunerite solid solution series). 26 27 2 .2 .I .2 . Amphibole Morphology 28 Mineral morphology is a function of the structural form of the silicate and the geologic 29 habit of formation, weathering and other mechanical processes. This discussion will focus on 30 morphology with respect to amphibole minerals. 31 The basic crystal structure of amphibole mineral is formed by the binding of a series of 32 double-chain plates (see Figure 2-5). Where the conditions are suitable, these crystals may form This document is a draftfor re\'iew purposes only and does not constitute utgencvpolicy. 2-7 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 2-5. Cross-section of amphibole fibers showing the silicon 3 tetrahedrons (A) that make up each double-chain plate (shown along the 4 fiber axis). Cations (shown as the darkened dots) occur between the plates 5 forming the basic fiber. 6 7 Source: Kroschwitz and Seidel (2010). 8 9 10 as elongated particles. The morphology of the elongated crystal structure is a function of the 11 temperature, pressure, local stress field and solution chemistry conditions during 12 crystallization-- habit o fformation. Thus, morphology at this level is described in terms of the 13 crystal forms which result from different habits of formation. Individual amphibole structures 14 may be described as acicular, prismatic, or a fibrous. A fiber would be an elongated crystal with 15 parallel sides, where acicular crystals are "needlelike" in appearance and prismatic crystals may 16 have several non parallel faces (e.g., varied, faceted faces). Asbestiform morphology is present 17 where the habit of formation allows crystals to form very long individual fibrils and fibers which 18 may become visible to the naked eye (see Figure 2-6). Thus, the amphibole crystalline staicture 19 may result in a range of particle morphologies, including fibers. Where conditions are not 20 conducive to the formation of individual fibers and particles, the amphibole is described as 21 massive--appearing as a solid contiguous sample. Mechanical forces that break amphibole 22 crystals along the cleavage plane create smaller pieces or cleavage fragments. These fragments This document is a draftfo r review purposes only and does not constitute Agency policy. 2-8 DRAFT--DO NOT CITE OR QUOTE 1 2 3 Figure 2-6. Comparison of crystalline forms amphibole minerals. Panel A 4 shows a specimen identified as an amphibole mineral in the 5 cummingtonite-grunerite solid solution series, although crystalline in form, 6 the habit of formation did not favor formation of individual particles and 7 fibers, hence its appearance as `massive'. Panel B shows an amphibole mineral 8 with very similar elemental composition but formed in a habit where very long 9 fibers were allowed to form--hence the asbestiform appearance. 10 11 Source: Adapted from Bailey (2006). 12 13 14 may be elongated, but differ from the crystals described above as at least one face of the 15 structure is the cleavage plane--not the face of a formed crystal. 16 With respect to classifying mineral field samples, geologists applied descriptive terms 17 appropriate for viewing samples simply or at low magnification (e.g., field glass). The geologic 18 terms for fiber morphology for classification of field samples is based on the macroscopic 19 appearance of the crystals and libers (e.g., acicular "needle-like in form") (AGL 2005). In this 20 framework, asbestos and asbestiform fibers are defined as long, slender, hair-like fibers visible to 21 the naked eye (see Figure 2-6). This is a hallmark of commercially mined asbestos which is 22 sought after for numerous applications because of its high tensile strength, heat resistance and in 23 some cases, can be woven. Although these terms were used to describe fibers in hand samples 24 and identify commercially valuable asbestos they are only applicable at the macroscopic level. It 25 is important to realize that material defined as commercial asbestos, mined, milled, and 26 manufactured into products not only contained these visible fibers, but many smaller fibers and 27 single crystals which were not visible to the naked eye (Dement and Harris. 1979). As further This document is a draftfor w fe * purposes only and does not constitute Agency policy. 2-9 DRAFT--DO NOT CITE OR QUOTE 1 explained in Section 3, only these smaller fibers can enter the lung and transport to the pleura 2 where the health effects of asbestos are best characterized. Therefore, for the purposes of this 3 assessment (i.e., examining the health effects of asbestos fibers), consideration must be given to 4 how these microscopic fibers are defined. For this purpose, terms intended for describing field 5 samples may need to set aside, or redefined when applied at the microscopic level. 6 Currently there are several technologies commonly used to view and identify mineral 7 structures at high magnification using light microscopes or electron microscopy. As standard 8 analytical methods were developed for counting mineral fibers, structures and matrices using 9 these instruments, analytical definitions to describe fibers and structures were developed. Phase 10 contrast microscopy (PCM) was developed to detect fibers in occupational settings and has been 11 widely used to assess worker exposure (see Text Box 2-1). The definition of a PCM-fiber is 12 based purely on its dimensions. The standardization of the PCM method (i.e., NIOSH 7400) and 13 its importance in applying health standards in occupational settings, results the common usage of 14 the term `fiber' to refer to those objects counted in the PCM analytical method (NIOSH. 1994a). 15 However, this method cannot define the material or morphology of the viewed fiber. Thus 16 PCM-fibers may be any material, and if they are mineral 17 fibers may be any fiber morphology. If the nature of the 18 fiber needs to be defined, NIOSH Method 7402 employs Text Box 2-1. Fibers Viewed by Light Microscopy The collection o f fibers on an air filter, and 19 electron microscopy to determine if the fibers viewed by visually counted under a phase contrast microscope (PCM), was first described in 20 PCM are mineral fibers, and can establish the mineral 1934 by the Dutch physicist Frits Zernike. The specification o f a fiber as >5 pm in 21 composition (NIOSH. 1994b). This method does not length and length-to-diameter ratio (i.e., aspect ratio) o f at least 3:1 resulted from this 22 recount the fibers, but, rather, it identifies what proportion method. As a light microscope technique, the PCM method cannot distinguish mineral 23 of the fibers are mineral fibers, with an elemental fibers from other fibers. 24 composition consistent with asbestos, which is then used The U.S. Public Health Service developed and tested a standard air sampling method 25 to adjust the PCM-fiber count. Although the PCM-fiber based on PCM detection (i.e.. National Institute for Occupational Safety and Health 26 definition was not based on either mineralogy or an [NIOSH] Method #7400). The NIOSH method specifies the analyst count fibers 27 understanding of which fibers might be biologi cally 28 relevant, this definition has become the basis of existing >5 pm in length with an aspect ratio of at least 3:1. Results from PCM analysis are reported as fibers per cubic centimeter o f air 29 health standards (e.g., MSHA. 2008; OSHA. 1994; U.S. (fibers/cc.) 30 EPA, 1988a). This document is a draftfor w ife* purposes only and does not constitute Agency policy. 2-10 DRAFT--DO NOT CITE OR QUOTE 1 Electron microscopy can view objects at much higher magnification and can be coupled 2 with other techniques which can identify the mineralogy (see Text Box 2-2). X-ray diffraction 3 (XRD) may be used with the above techniques to differentiate crystalline structure of minerals in 4 solid materials and provides information on the availability of the total mineral present. Thus, 5 XRD can determine the mineral composition of the material analyzed, identifying its solid 6 solution series and classifying the mineral per standardized nomenclature for amphibole minerals 7 (see Section 2.1.1.1). 8 With the advent of the use of electron microscopy to identify mineral particles, there has 9 been an attempt to resolve the traditional dimensional fiber defmition(s), by describing the 10 particles examined by electron microscopy and 11 X-ray diffraction in terms that are both Text Box 2-2. Minerals Viewed by Electron 12 geologically and mineralogically relevant. Microscopy Electron microscopy employs electrons-- rather 13 Structures viewed by electron microscopy may that light-- to visualize the specimen. Furthermore, instead o f using glass lenses to focus the light 14 be described as having parallel sides, and wavelengths, electromagnetic lenses are used to focus electrons on the sample. The analytical 15 considered `fibers'. Where long, thin, curving techniques included in electron microscopy for asbestos testing are TEM, scanning electron 16 fibers are viewed they may be described as microscopy (SEM), and scanning transmission electron microscopy (STEM). TEM produces 17 `asbestiform'. Structures with nonparallel sides two-dimensional (2-D) images that generally use a 18 can be considered acicular or prismatic, magnification factor o f about 500 to 500,000x. SEM produces tliree-dimensional (3-D) images that 19 depending on their proportions. Thus, the generally result in about 10 to 300,000x magnification. STEM can produce both 2-D and 20 descriptive terms used by geologists have 3-D images that generally result in about 10 to 21 migrated into the analytical field. However, the 500,000x magnification. The ISO 10312 method for analyzing air filters, 22 habit of formation of a single structure viewed enumerates structures much smaller than the PCM fibers with a minimum length requirement of 23 by electron microscopy cannot be determined, 0.5 pm. Additionally, structures with an aspect ratio o f at least 5:1 are considered fibers, rather than 3:1, 24 and, while descriptive, these terms may not 25 correlate to the geologic and commercial as with PCM analysis. The ISO 10312 method also defines other structures (fiber bundles, clusters, and matrices) that are included in the structure count. Therefore, the term "structure" rather than "fiber" is 26 definitions of these terms. Therefore, the use of used when presenting air sampling results from the 27 these definitions to describe individual particles I(sS/Occ)1a0r3e1r2epmorettehdo.d where structures per cc of air 28 viewed by TEM can be problematic (Meeker et 29 al,, 2003). Important characteristics such as crystal structure and surface chemistry cannot be 30 adequately categorized solely with visually determined definitions developed for the 31 classification of field samples. This document is a draftfor re\'iew purposes only and does not constitute utgencvpolicy. 2-11 DRAFT--DO NOT CITE OR QUOTE 1 The definition of `fiber' and the appropriate application of other morphological terms is 2 an area of ongoing debate. From a public health and regulatory perspective, a PCM-fiber is the 3 fiber of interest (where confirmed as a mineral fiber with an elemental composition consistent 4 with asbestos). There is no requirement for a PCM-fiber to be asbestiform, and, in fact, the 5 method explicitly includes fibers with fairly low aspect ratios (i.e., as low as 3:1). Electron 6 microscopy identified a much broader range of fibers (having much greater resolution) and can 7 provide more specific identification of both mineralogy and the form of the structure. 8 9 2.2.2. Vermiculite 10 Vermiculite is the mineralogical name given to hydrated laminar 11 magnesium-aluminum-ironsilicate, which resembles mica in appearance [see Figure 2-7; (Mg, 12 Fe,A)3(Al,Si)2 0 io(OH)2 *41420] (AGf 2005). Vermiculite is in the clay mineral group of the 13 phyllosilicates, which also includes kaolinite and montmorillonite. Mica, talc, and serpentine 14 (e.g., chrysotile asbestos) minerals are other well-known sheet silicates. These sheet-like 15 structures are produced by rings of tetrahedrons that are linked to other rings by shared oxygen 16 ions in a two-dimensional plane (see Figure 2-4d). The silicate sheet can extend broadly, and the 17 layered appearance of the mineral reflects this sheet-like structure. The symmetry of these 18 minerals is controlled primarily by the symmetry of the rings, which is usually altered to a lower 19 symmetry by other ions and other layers. Typically, crystals of this subclass are flat, platy, and 20 book-like, as in the mica group, and the sheets are then connected to each other by layers of 21 cations. These cation layers are weakly bonded and often have water molecules and other 22 neutral atoms or molecules trapped between the sheets. When subjected to heat, vermiculite has 23 the unusual property of exfoliating or expanding into "worm-like" pieces. The term vermiculite 24 is derived from the Latin vermiculare, which means to breed worms (The Vermiculite 25 Association, http://www.vermiculite.org). Vermiculite exfoliation occurs at approximately 26 150C, producing a lightweight and highly absorbent material (AGF 2005). Additional 27 properties of vermiculite are listed in Table 2-1. Vermiculite ore is shown in Figure 2-7. This document is a draftfor w ife* purposes only and does not constitute Agency policy. 2-12 DRAFT--DO NOT CITE OR QUOTE 1 Table 2-1 Properties of vermiculite 2 Mineral class/subclass Mineral silicates/phyllosilicate Chemical formula (Mg, Fe,A)3(Al,Si)2O10(OH)2 4H 20 Crystal habit o f formation Clay, scaly, aggregate Hardness (Mohs scale) 203 Cleavage Perfect Specific gravity 2.4 2.7 3 4 5 6 Figure 2-7. Vermiculite ore sample. Brinton's Quarry, near West Chester, 7 Chester County, Pennsylvania, USA. 8 9 Source: Micaceous vermiculite book (http://www.excaliburmineral.com/cdintro.htm) 10 Jeff Weissman/Photographic Guide to Mineral Species. n 12 13 Vermiculite is mined across the world, including the United States (Virginia, South 14 Carolina, and Montana); South Africa; Uganda; China; Brazil; Russia; India; and Australia 15 (BGS, 2011). The specific mineralogy and geologic formati on habit of vermiculite deposits 16 vary, and although amphibole minerals are consistent with the ultramafic rock formations 17 (composed chiefly of ferromagnesian igneous rock) that bear vermiculite, not all vermiculite 18 deposits contain amphibole asbestos. 19 This document is a draftfor review purposes only and does not constitute Agency policy. 2-13 DRAFT--DO NOT CITE OR QUOTE 1 2.2.3. The Mineralogy of Libby Amphibole Asbestos 2 2.2.3.1. Mineralogy 3 The amphibole mineral fibers within the vermiculite ore and product have historically 4 been reported as a sodium-rich tremolite (Amandus et ah. 1987a; McDonald et ah, 1986a; Leake- 5 1978; Boettcher. 1966; Larsen. 1942). More recently, various research groups have 6 characterized the more specific mineralogical composition of amphiboles from the Rainy Creek 7 deposit near Libby, MT (Gunter and Sanchez, 2009; Sanchez et al, 2008; Meeker et ah, 2003; 8 Wylie and Verkouteren, 2000; Ross et al.., 1993; Moatamed et al.. 1986). 9 EPA requested that the U S. Geological Survey (USGS) design and conduct a study to 10 identify the amphibole minerals in the Libby vermiculite mine. Accordingly, USGS personnel 11 collected samples from different areas of the mine in an attempt to identify the range of materials 12 present both geographically, as well as collecting material which represented different habits of 13 formation (Meeker et al.. 2003). Figure 2-8 shows data from 30 samples across the mine. The 14 mineral composition of each structure determines its mineral identity (Leake et al.. 1997). Here, 15 the U.S. Geological Survey (USGS) used two different techniques to identify the mineral 16 composition of each structure (energy dispersive X-ray analysis [EDS] and electron probe 17 microanalysis [EPMA]). Similar mineral composition was determined by the two methods (see 18 Figure 2-8). Most amphibole structures are classified as winchite (84%), with lesser amounts 19 classified as richterite (11%) and tremolite (6%) (Meeker et al.. 2003). based on the current 20 mineralogical nomenclature by Leake (1997). There are also trace amounts of 21 magnesioriebeckite, edenite, and magnesio-arfvedsonite present in Libby Amphibole asbestos 22 (Meeker et al.. 2003). All of these minerals are within the mineral solid solution series for 23 tremolite-richterite- magnesioriebecktite. All of the amphiboles found at the mine site, with the 24 possible exception of magnesioriebeckite, can occur in fibrous habit. It was observed these 25 amphibole materials--even when originally present as massive material--can produce abundant, 26 extremely fine fibers by gentle abrasion or crushing (Meeker et al.. 2003). 27 This document is a draftfor review purposes only and does not constitute Agency policy. 2-14 DRAFT--DO NOT CITE OR QUOTE Edenite 0.5 Richterite A* xEDS OEPMA EDS EPMA Mean Error (1a) x x Magnesioarfvedsonite 0.5 1 1.5 Meeker et al fig. 6 Na(B) 1 2 Figure 2-8. Mineralogy of Libby Amphibole asbestos structures from samples taken 3 from the Zonolite Mountain site. An evaluation of the textural characteristi cs shows 4 the material to include a complete range of morphologies from prismatic crystals to 5 fibers. Each data point represents the cation composition (number of occupied sites) for a 6 single fiber. The X-axis shows the number of sites occupied by Na, and the Y-axis shows 7 the number of sites occupied by Na or K. The data shown are a composite of the analysis 8 fibers taken from 30 different field samples from various locations within the mine. 9 10 Notes; EDS is energy dispersive X-ray analysis; EPMA is electron probe microanalysis. 11 Source: Meeker et al. (2003). 12 13 14 Figure 2-9 shows the compositional variations between the predominate minerals found 15 in the Libby Amphibole asbestos (winchite, richterite, and tremolite). Although each structure 16 has as discrete mineral composition, when viewed as a population, fall within solid solution 17 series shown in Figure 2-8. For example, tremolite is one end-member of the solid solution 18 series. As calcium decreases and sodium increases, the fibers transition to richterite. Similarly, 19 as fibers have decreased magnesium and calcium with respect to tremolite, they are defined as 20 winchite. The sodium content that distinguishes these amphiboles has been redefined over time 21 in the International Mineralogical Association's mineral classification system, most recently in 22 1997 (Leake et al.. 1997; Leake. 1978). As a result, some amphibole fibers previously defined as 23 tremolite prior to the new classification system are currently considered winchite based on 24 chemical composition (Leake et al.. 1997). This document is a draftfo r review purposes only and does not constitute Agency policy. 2-15 DRAFT--DO NOT CITE OR QUOTE Decreasing My 1 2 Figure 2-9. Solution series linking tremolite, winchite, and richterite 3 amphibole fibers. 4 5 Source: Meeker et al. (2003). 6 7 8 The mineral composition of the fibers present is not classifiable to one distinct named 9 mineral category, but, rather, the composition spans several solid-solution series. However, 10 there seems to be a consistency in the range of elemental composition found within this material . 11 Libby Amphibole asbestos is not only made up of the end-members of these solid soluti on series, 12 but the spectrum of minerals along the solid solution series shown. Although the majority of 13 structures analyzed fell within these solid solution series, traces of other minerals were 14 identified. The term "Libby Amphibole" is used in this document to identify the mixture of 15 amphibole minerals, of varying elemental compositions (e.g., winchite, richterite, and tremolite), 16 which have been identified in the rocks and ore of the vermiculite mine near Libby, MT, and are 17 characteristic of the elongated structures commingled with the vermiculite mined at this location 18 (Meeker et al.. 2003) (i.e., present in the ore vermiculite concentrate and processed materials). 19 Libby Amphibole Asbestos refers to those elongated structures of the Libby Amphibole mineral 20 mixture, which have been identified as amphibole fibers or structures, and have been associated This document is a draftfo r review purposes only and does not constitute Agency policy. 2-16 DRAFT--DO NOT CITE OR QUOTE 1 with health effects consistent with asbestos exposure (i.e., asbestosis, pleural abnormalities, lung 2 cancer and mesothelioma) (ATSDR. 2008b). 3 4 2.2.3.2. Morphology o f the Libby Amphibole Asbestos 5 Mineral samples taken from the mine include veins of asbestiform amphibole and various 6 fiber morphologies in surrounding rock (Meeker et al.. 2003). A sample viewed by scanning 7 electron microscope from the Zonolite Mountain mine illustrates the broad range of size and 8 morphologies for the mineral structures (see Figure 2-10). The USGS has described fibers 9 (including asbestiform), acicular and prismatic structures, and curved fibers all within the 10 minerals from the mine (Meeker et al.. 2003). As individual fibrils and fiber bundles are viewed 11 under greater magnification under a transmission electron microscope, the range of fiber 12 morphologies can be more clearly seen (see Figure 2-11). 13 14 15 16 Figure 2-10. Scanning electron microscope image of amphibole mineral 17 structures from the Libby, MT mine. An evaluation of the textural 18 characteristics shows the material to include a range of morphologies from 19 prismatic crystals to fibers. Acicular and prismatic crystals, fibers bundles and 20 curved fibers are all present. 21 22 Source: Meeker et al. (2003). 23 This document is a draftfor review purposes only and does not constitute Agency policy. 2-17 ' DRAFT--DO NOT CITE OR QUOTE MEEKER ET Al_: 1HECOMPOSITION OF AMI1UBOLEN FROM THE KAINV CHEEK COMPLEX 1 Figure 2-11. Fiber morphology of amphibole asbestos from the Libby, MX 2 mine viewed under a transmission electron microscope. 3 4 Source: Meeker et al. (2003). 5 6 7 2.2.3.3. Dimensional Characteristics o f Libby Amphibole Asbestos 8 Cumulative particle-size-distribution frequencies (CDF) were developed for Libby ore 9 Grade 3, and Libby ore Grade 3 expanded by EPA Region 8 using the procedure described in 10 detail in Appendix C. As shown in Figure 2-12, the particle-size-distribution frequency for the 11 Libby Grade 3 ore, and the Libby Grade 3 ore expanded were similar to the 12 particle-size-distribution frequency in the ambient air monitoring samples in Libby, MT. Data 13 from ambient air monitoring in Libby are presented in Appendix B. The data to construct the 14 This document is a draftfor review purposes only and does not constitute Agency policy. 2-18 DRAFT--DO NOT CITE OR QUOTE Particle Size Distributions of LA Particles Libby #3 Ore (N = 320), Width (um) 1 2 Figure 2-12. Particle size (length, width, aspect ratio) of fibers in Libby ore 3 and Libby air. 4 5 CDF = cumulative distribution frequency; LA = Libby Amphibole. 6 Source: U.S. EPA (2010b) (Provided as Appendix B.) This document is a draftfor review purposes only and does not constitute Agency policy. 2-19 DRAFT--DO NOT CITE OR QUOTE 1 plot in Figure 2-11 are described in Appendices B and C. There are slight shifts towards longer 2 and thicker fibers in the ore samples compared to the air samples, with the aspect ratios being 3 almost identical in the ore and air samples. However, all of these differences are minor, and the 4 majority of these fibers are respirable. 5 Mineralogical characterization of the fibers from the Libby ore Grade 3 and the expanded 6 product using energy dispersive X-ray analysis (EDS) and selected area electron diffraction 7 (SAED) provided further confirmation of the similarity between the fibers from the Libby 8 Grade 3 ore and Libby Amphibole asbestos (methodology described in Section 2.3; see also 9 Appendix B). EDS spectra yielded an elemental fingerprint with sodium and potassium peaks 10 that were highly consistent with values reported for the winchite-richerite solution series 11 described for the Libby, MT ores (Meeker et aL 2003). 12 Based on these data, it is reasonable to conclude that the fibers from the Libby Grade 3 13 ore and expanded ore are similar in physical and mineralogical characteristics to the Libby 14 Amphibole asbestos fibers found in air samples from Libby, MT. The O.M. Scott facility in 15 Marysville, OEI used Libby Grade 3 ore from about 1959 to 1980 (Moatamed et ah. 1986; 16 Lockev et aL 1984). Therefore, the exposure and health effects information from the 17 Marysville, OH facility may be used to derive an RfC that can be applied to the Libby 18 community and other sites that received vermiculite ore from Libby, MT. 19 The Marysville, OH facility also used venniculate ore from Virginia, South Africa, and 20 South Carolina. The Virgini a and South African ores were tested for the presence of fibers as 21 described in Appendix C.5 As described in Appendix E, the Virginia and South African ores 22 released only a small quantity of amphibole fibers. EPA was unable to obtain an ore sample 23 from South Carolina. However, vermiculite ore from the Enoree mine in South Carolina is 24 known to contain amphibole fibers (see Appendix C) (U.S. EPA. 2000b; McDonald et aL 1988). 25 26 2.3. EXPOSURE POTENTIAL 27 Although the occurrence of Libby Amphibole asbestos is limited to a relatively small 28 geographic area, the potential for exposure to it has been greatly enhanced by the historical 29 mining, milling, and distribution of vermiculite operations in Libby, MT. Additionally, material 5 Dr. Lockey, University o f Cincinnati, obtained samples o f the Virginia and South Africa ores from the Marysville, OH facility in 1980 and supplied these ores to the EPA for analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 2-20 DRAFT--DO NOT CITE OR QUOTE 1 was sent to processing plants across the nation where plant workers and community contacts may 2 have been exposed. Lastly, consumer products containing vermiculite mined near Libby contain 3 Libby Amphibole asbestos, and consumers may be exposed to Libby Amphibole asbestos while 4 using the products. For example, asbestos-contaminated vermiculite attic insulation from Libby 5 remains in homes today across North America, where there is the potential for residential 6 exposures. This section summarizes the potential for current exposures to the Libby Amphibole 7 asbestos in vermiculite in the Libby community, other communities potentially impacted by 8 processing plants, and from in-place Libby vermiculite attic insulation. Historical exposures for 9 the workers in Libby, MT, and other facilities are discussed in Section 4.1, where data are 10 available. 11 There are also lifestyle, activity, and lifestage factors, which may influence one's 12 exposure potential to asbestos. For example, children may spend more hours outside and engage 13 in activities that impact exposure level compared to adults (U.S. EPA, 2006b; NRC 1993). In 14 general, children inhale more air per unit body weight (TJ.S. EPA. 2006b) and spend more time 15 outdoors than adults (Bateson and Schwartz. 2008; NRC 1993). which could have resulted in 16 increased inhalation exposure to Libby Amphibole asbestos in children compared with adults. In 17 contrast, some adult activity patterns, such as gardening and home repair, may also result in 18 increased exposures where Libby Amphibole asbestos may be present. Thus for the various 19 environments where people may be exposed to Libby Amphibole asbestos, the potential 20 activiti es and pathways of exposure are di scussed below, and where avail able, exposure 21 measurements are given for various exposure environments and activities. 22 23 2.3.1. Libby Community 24 The Libby community (the towns of Libby, Troy, and surrounding residences) defines the 25 area that may have been directly and indirectly impacted by mining/milling-activities. Many 26 individuals who worked in the mine lived in the surrounding areas. Facilities in the community 27 may have residual contamination from past milling and transport activities. Additionally, 28 expanded vermiculite, waste stoner rock (the waste material from exfoliation), and other 29 materials all potentially containing Libby Amphibole asbestos may have been transported off site 30 to residences and recreational areas. Taken together, there are numerous potential exposure 31 pathways for community residents, both historical and current. This document is a draftfor review purposes only and does not constitute Agency policy. 2-21 DRAFT--DO NOT CITE OR QUOTE 1 During plant operations, individuals may have been exposed to materials inadvertently 2 transported from the workplace to vehicles, homes, and other establishments, typically on the 3 clothing, shoes, and hair of workers. This transport of material may result in "take-home 4 exposure" for the workers, their families, and other coresidents. The magnitude of these 5 exposures was not measured, so the levels to which individuals in the home might have been 6 exposed are not known. Based on studies of other industrial take-home exposures, individuals 7 doing laundry and cleaning house (often women) can be exposed to materials on workers' 8 clothing. Also, children who play on the floor might be more exposed than adults to dust from 9 take-home exposures (Kelly et ah. 2006). The community health screening studies from Libby 10 showed that men were more likely to have both occupational and nonoccupational exposures, 11 while women were more likely to have household contact with exposed workers (Peipins et al., 12 2003; ATSDR. 2001b). There could also be gender differences in types of activities (e.g., 13 household chores such as laundry and cleaning) or in intensity or duration of occupational and 14 recreational activities (Peipins et al.. 2003). 15 Expanded vermiculite, as a finished product, was used as a soil amender and for attic 16 insulation. Community members may have been exposed and are possibly still exposed to these 17 consumer products. In a survey of Libby resi dents conducted by ATSDR in 2000-2001, almost 18 52% reported using vermiculite for gardening, 8 .8% used vermiculite around the home, and 19 51% reported handling vermiculite attic insulation (Peipins et al.. 2003). As vermiculite ore, 20 waste Stoner rock, and product were present in the community; numerous activities may have 21 resulted in exposure. Individuals also reported exposures from the following activities: 22 participating in recreational activities along Rainy Creek Road, the road leading to the mine 23 (67%); playing at the ball field near the expansion plant (66%); playing in the vermiculite piles 24 (34%); heating the vermiculite to make it expand/pop (38%); or other activities in which there 25 was contact with vermiculite (31%) (Peipins et al.. 2003). Memoranda from Christopher Weis 26 (Weis. 2001a) state that asbestos mineral fibers were detected in outdoor sources (yard soil, 27 garden soil, driveway material, and assorted mine-waste materials) and indoor sources (dust and 28 vermiculite insulation) in Libby (Weis. 2001a. b). 29 EPA has conducted more recent exposure sampling in the Libby community. Air 30 samples were taken in the community during activities considered appropriate for various 31 potential exposure scenarios. Personal air monitors were placed on the investigator conducting This document is a draftfor review purposes only and does not constitute Agency policy. 2-22 DRAFT--DO NOT CITE OR QUOTE 1 the activity, and a second air sample was taken from a fixed location (area sample). Asbestos 2 fibers were collected on filters and counted by two different laboratory methods: (1) PCM and 3 (2) TEM. Although TEM analysis can count smaller fibers, results are shown here for PCM size 4 fibers used to estimate risk, called PCM equivalent fibers (PCMe); 5 EPA continues to conduct air monitoring in the Libby community to support clean-up 6 and risk assessment activities. Ambient air monitoring conducted in 2006/2007 at 18 locations 7 across the area indicated that low levels of asbestos fibers are occasionally detected in the air, 8 even with no localized disturbance of asbestos-contaminated material (U.S. EPA. 2009b). Fibers 9 were counted by TEM, and structures67>0.5 pm in length and with an aspect ratio >3 were 10 included (measured in structures per cc of air, s/cc). Average ambient air levels for the various 11 sampling locations ranged from 8 * U f6 s/cc to 1.9 x 10 5 s/cc (U.S. EPA. 2009b). Both 12 ambient and activity-based air monitoring have been completed in five community schools (U.S. 13 EPA. 2010c). Outdoor activities conducted that were considered relevant to children's exposures 14 at the schools included playing sports, using playground equipment, and mnning/walking in 15 outdoor areas. Outdoor activities to assess exposure of the school maintenance workers included 16 digging/raking, power sweeping parking lots, and mowing and edging school lawns. 17 Additionally, ambient air samples were taken in each school (i.e., classrooms, cafeteria, 18 gymnasium, and hallways). Asbestos PCMe fibers were detected by TEM analysis in 5 of 19 63 outdoor activity-based samples, ranging from 0.0022 to 0.039 s/cc. No PCMe fibers were 20 detected in indoor air samples. Elowever, 2 of 50 indoor area samples detected TEM asbestos 21 structures not considered to be PCMe fibers (5.1 > 10 1s/cc and 5.9 * 10 4 s/cc), which are 22 within the range of analytical sensitivity for the indoor air samples (U.S. EPA. 2010c). It should 23 be noted that indoor air sampling did not include any activity-based sampling to assess student or 24 employee exposures. 25 6 These PCM equivalent fibers (PCMe fibers) are defined as those fibers viewed on TEM that meet the PCM analytical requirements: >5 pm in length and an aspect ratio o f at least 3:1. Although the PCM methodology does not specify a minimum fiber width, current PCM analytical methods reliably detect fibers o f 0.25 pm in width (IPCS. 1986). which EPA employs to define PCMe fibers (U.S. EPA. 2008). 7A single fiber, fiber bundle, cluster, or matrix as defined in the TEM analytical method ISO 10312. This document is a draftfor review purposes only and does not constitute Agency policy. 2-23 DRAFT--DO NOT CITE OR QUOTE 1 2.3.2. Communities near Vermiculite Expansion and Processing Plants 2 Vermiculite from the Libby, MT mine was used commercially from the 1920s to 1990, 3 and a review of company records available from (1964-1990) indicates approximately 4 6,109,000 tons of vermiculite concentrate was shipped to over 200 facilities (ATSDR. 2008b). 5 The 2008 ATSDR (2008b) Summary Report on the 28 Libby vermiculite expansion and 6 processing facilities stated that household residents were exposed by contact with vermiculite 7 from the workers' clothes, shoes, and hair. Workers' personal vehicles likely contained 8 vermiculite dust from the facility emissions and from vermiculite that fell from their clothing and 9 hair on the drive home after work. The O.M. Scott Company (Marysville, OH) reported that 10 company policy was to launder work clothes for their employees and to make showers available 11 for use after work. These procedures, when implemented, should greatly reduce exposure 12 potential via household contact (ATSDR. 2005b). Whether other facilities made these services 13 available or how frequently they might have been used is unknown. 14 Communities near the expansion plants were subjected to some of the same exposure 15 pathways as for the Libby community. The 2008 ATSDR Summary Report observed that 16 individuals in the community could have been exposed through multiple avenues, such as living 17 near the plant and breathing emi ssions from the facility, disturbing waste-rock piles, having 18 direct contact with waste rock brought home, and living with indoor dust containing asbestos 19 brought in from outdoor sources (ATSDR. 2008b). 20 21 2.3.3. Exposures from Zonolite and Vermiculite for Homeowners, Contractors, and Other 22 Populations 23 Vermiculite was most notably used as attic insulation, as a soil amender for gardening, 24 and in the manufacturing of gypsum wallboard. EPA conducted a study to estimate the potential 25 for exposure to asbestos in homes containing VAI. Air samples were taken to define exposure 26 levels in the homes under various conditions: no activity (e.g., ambient air), as well as during 27 simulated remodeling activities and removal of the VAI (Versar. 2003). Samples were taken in 28 the living space of the homes as well as the attic space. 29 Air samples were collected in five occupied homes where Zonolite VAI was in place 30 (asbestos detected from trace levels to 1.54% by bulk analysis); no fibers were detected in the air 31 samples above 0.0016 PCMe fibers/cc in these homes. However, the air samples were taken This document is a draftfor review purposes only and does not constitute Agency policy. 2-24 DRAFT--DO NOT CITE OR QUOTE 1 when the homes were empty, and there was no disturbance of the VAI or entry/exit into the attic 2 space. Therefore, EPA conducted a number of simulations under controlled conditions to 3 estimate exposures when VAI is disturbed during normal activities (e.g., moving boxes in an 4 attic), remodeling, and removal of the VAI. Structures were built within safe containment to 5 simulate attic space above living space, and VAI was installed in the simulated attics. 6 Remodeling activities resulted in personal exposures ranging from 0.50 to 1.841 fibers/cc PCMe. 7 Stationary samples of the attic air ranged from 0.008 to 0.203 fibers/cc PCMe. For those 8 simulations that included sampling in the `living space' below the attic, asbestos fibers ranged 9 from 0.001 to 0.25 fibers/cc PCMe during renovations and from 0.001 to 0.035 fibers/cc PCMe 10 in the living space after renovations were complete (Versar. 2003). These data indicate that 11 exposures to asbestos fibers may occur when disturbing Libby Amphibole asbestos-containing 12 VAI in homes. 13 A second study on potential exposures to Zonolite VAI was conducted by an 14 environmental firm hired by attorneys representing individuals with VAI in their homes (Ewing 15 et al., 2010). This study was conducted in three homes containing Zonolite VAI, and air samples 16 were taken, representing ambient conditions (no disturbance of VAI), remodeling, activity in the 17 attic, and removal of the VAI by various methods (see Table 2-2). Disturbance of the 18 asbestos-containing VAI resulted in airborne asbestos levels, both in the personal air monitors 19 and area samples (Ewing et al.. 2010). This document is a draftfor review purposes only and does not constitute Agency policy. 2-25 DRAFT--DO NOT CITE OR QUOTE 1 Table 2-2. Air sampling results for asbestos from Zonolite VAI in 2 three homes 3 Personal samples Area samples Activity PCMa (fibers/cc) TEMb (PCMe, s/cc) TEM (PCMe, s/cc) No activity NSC NS <0.003 Cleaning items in the 1.54 <0.42 0.07 attic Cleaning storage area in 2.87 2.58 0.47 the attic Cutting a hole in the 5.80 1.32 0.52 ceiling below the VAI VAI removal (various 2.9-12.5d 0.98-10.3 0.53-1.47 methods) 4 5 aAir sampling results reported as fibers analyzed by phased contrast microscopy (PCM). 6 bAir sampling results reported as structures, PCMe as analyzed by transmission electron microscope (TEM). 7 CNS-- not sampled, personal samples were not taken for background levels. 8 dRange o f results for three different removal methods (shop vacuum, homeowner method, and 9 manufacturer-recommended method). 10 11 Source: Ewing et al. (2010). This document is a draftfor review purposes only and does not constitute Agency policy. 2-26 DRAFT--DO NOT CITE OR QUOTE 1 3. FIB ER TO X IC O K IN ETIC S 2 There are no published data on the toxicokinetics of Libby Amphibole asbestos.8 3 However, to help inform the reader as to the expected toxicokinetics of Libby Amphibole 4 asbestos, this section contains a general summary description of toxicokinetics of fibers. A more 5 detailed discussion of fiber toxicokinetics is beyond the scope of this document and is reviewed 6 elsewhere (NIOSH, 2011; 1CRP. 19941. 7 The principal components of fiber toxicokinetics in mammalian systems are 8 (1) deposition at the lung epithelial surface, and (2 ) clearance from the lung due to physical and 9 biological mechanisms (including both translocation from the lung to other tissues [including the 10 pleura]), and elimination from the body (see Figure 3-1). 11 Libby Amphibole asbestos includes fibers with a range of mineral compositions 12 including amphibole fibers primarily identified as richterite, winchite, and tremolite (see 13 Section 2.2). Although the fiber size varies somewhat from sample to sample, a large percentage 14 (-45%) is less than 5 pm long in bulk samples examined from the Libby mine site (Meeker et al.., 15 2003). Limited data from air samples taken in the workplace also document a large percentage 16 of fibers (including both respirable9 fibers as well as fibers <5 pm-long) (see Section 4.1.1.2 and 17 Table 4-3). The importance of the size of fibers and how they deposit following inhal ation is 18 described below. Due to a lack of data specific to Libby Amphibole asbestos, these deposition 19 steps are discussed for general forms of asbestos. The main route of human exposure to mineral 20 fibers is through inhalation, although other routes of exposure play a role. Exposure of 21 pulmonary tissue to fibers via the inhalation route depends on the fiber concentration in the 22 breathing zone, the physical (aerodynamic) characteristics of the fibers, and the anatomy and 23 physiology of the respiratory tract. Ingestion is another pathway of human exposure and occurs 24 mainly through the swallowing of material removed from the lungs via mucociliary clearance or 25 drinking water contaminated with asbestos, or eating, drinking, or smoking in 26 asbestos-contaminated work environments (Coridie. 1983). Handling asbestos can result in 27 8The term "Libby Amphibole asbestos" is used in this document to identify the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby. MT. It is further described in Section 2.2. 9 Respirable fibers are those that can be inhaled into the lower lung where gas exchange occurs and are defined by their aerodynamic diameter (da< 3 pm; NIOSH) (2011). This document is a draftfor review purposes only and does not constitute Agency policy. 3-1 DRAFT--DO NOT CITE OR QUOTE 1 2 Figure 3-1. General scheme for fiber deposition, clearance, and translocation 3 of fibers from the lung and GI tract. General scheme for fiber deposition 4 (heavy arrows), clearance (light dotted arrows), and translocation (light arrows). 5 Diagram of Bignon et al. (1978) derived from International Commission on 6 Radiological Protection (ICRP) lung model by the Task Group on Lung 7 Dynamics (1966), as cited in ICRP (1994). 8 9 Source: ICRP (19941. 10 11 12 heavy dermal contact and exposure. Asbestos fibers could become lodged in the skin, producing 13 a callus or corn--but generally with no serious health effects (Lockev et al.. 1984). Because few 14 studies have examined the deposition and clearance of fibers foll owing ingestion of or dermal 15 exposure to fibers, the focus of this section is on the main route of exposure: inhalation. 16 Studies useful for assessing the relationship between airborne fiber concentrations and 17 respiratory disease must involve meani ngful measurements of environmental exposure and an 18 understanding of how to apply these measurements to the target tissue dose. Ti ssue dose is a 19 more specific measure than external dose, and it is determined both by fiber characteristics of the 20 exposure environment and the exposed population. Dose to the lung is a function of airway 21 anatomy, lung volume, ventilation rate, and clearance from the lung, as well as the fiber's 22 physical and chemical characteristics (U.S. EPA. 2004: Oberdorster. 1991). Many studies have This document is a draftfor review purposes only and does not constitute Agency policy. 3-2 DRAFT--DO NOT CITE OR QUOTE 1 examined the role of these physical and chemical characteristics in asbestos-induced disease in 2 the lung and are reviewed in more depth elsewhere (NIQSH. 2011; ATSDR. 2001a; Mvoio and 3 lakava. 2001; Witschi and Last 1996; Lippmann. 1990; Merchant. 1990; Yu et al., 1986; Griffis 4 et aI . 1983; Harris and Fraser. 1976; Harris and Timbrell. 1975). Factors influencing dose to 5 other tissues in the body (e.g., pleura, peritoneum, stomach, and ovaries) are not as well known, 6 but they are discussed below where data are available. 7 8 3.1. DEPOSITION OF FIBERS IN THE RESPIRATORY TRACT 9 The deposition of fibers in the respiratory tract is dependent on the aerodynamic 10 properties of the fiber (length, width, and density) and the anatomy and physiology of the 11 respiratory tract (NIQSH. 2011; ATSDR. 2004. 2001a; Mvoio and Takaya. 2001; Witschi and 12 Last. 1996; Yu et aL 1986; Griffis et al.. 1983; Harris and Fraser.. 1976; Harris and Timbrell. 13 1975). The aerodynamic diameter of fibers is mostly determined by the geometric diameter and 14 density. In general, thicker fibers are deposited in the upper airways; thinner fibers are carried 15 deeper into the airways and alveolar regions. Fibers with aerodynamic diameters less than 16 approximately 3 pm meet the physical criteria necessary for deposition in the terminal 17 bronchioles and beyond to the alveoli. The site of fiber deposition within the respiratory tract 18 has implications related to lung retention and surface dose of fibers. 19 The respiratory tract encompasses the extrathoracic region (nasal passages, pharynx, and 20 larynx), thoracic region (the conducting airways [trachea bronchi, bronchioles]), and the 21 gas-exchange region of the lung (respiratory bronchioles, alveolar ducts, and alveoli). A full 22 review of the anatomy and architecture of the respiratory tract is beyond the scope of this 23 document but has been reviewed by ICRP (ICRP. 1994). 24 Fiber deposition occurs by five mechanisms: impaction, interception, sedimentation, 25 diffusion, and electrostatic precipitation (see Table 3-1): 26 27 28 1. Impaction: The momentum of the fiber causes it to directly impact the airway 29 surface as the airflow changes direction. This is the predominant method of 30 deposition in the nasopharyngeal region where airflow is swift and larger 31 fibers/particles are present. This document is a draftfor review purposes only and does not constitute Agency policy. 3-3 DRAFT--DO NOT CITE OR QUOTE Table 3-1. Factors influencing fiber deposition and clearance in the respiratory system This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 3-4 Size of fiber (aerodynaini c diameter) Area of deposition in respiratory system Predominant method of deposition Mechanisms for fiber retention 5-30 pm Extrathoracic Region (nasopharyngeal region nasal passages, pharynx, larynx) Impaction Epithelial cell uptake 1-5 pm Thoracic Region (trachea, bronchial and bronchiolar region) Sedimentation Epithelial cell , impaction, uptake interception Physical clearance Mucous flow (mucociliary apparatus into gastrointestinal tract) Macrophage: phagocytosis and transport Mucociliary apparatus Macrophage: phagocytosis and transport Dissolution Target tissue for translocation Not measured, although dissolution can occur, removal from mucous flow is fairly quick and likely predominant Gastrointestinal tract Nasal-associated lymphoid tissue, lymph system Mucous Macrophage Gastrointestinal tract Mucosa-associated lymphoid tissue, lymph system 2 pm or less Gas-Exchange Region (respiratory bronchioles, alveolar ducts, alveoli) Diffusion Epithelial cell uptake Translocation to other target tissues Macrophage: phagocytosis and transport Lung surfactant Macrophage Asbestos bodies Pleura Gastrointestinal tract Mucosa-associated lymphoid tissue, lymph system Pleura Source: Adapted from Witsclii and Last (2001) in Casarett and Doull's Toxicology: The Basic Science o fPoisons, 6th edition, p. 515. DRAFT--DO NOT CITE OR QUOTE 1 2 . Interception: A special case of impaction where the edge of the fiber touches the 2 airway surface and is prevented from continuing along the airway. This 3 mechanism is important in the conducting airways (trachea and bronchi), where 4 the airflow is slower and laminar flow along the airway surface is conducive to 5 interception. 6 3. Sedimentation: Gravitational forces and air resistance cause fibers/particles to 7 settle out of the air column onto the airway surface. For sedimentation to occur, 8 air flow velocities must be low to allow the particle/fiber to settle, and this is a 9 predominant mechanism to the smaller conducting airways. 10 4. Diffusion: This method of deposition is predominant in the alveolar region where 11 air movement is negligible. Diffusion occurs from interactions of the fibers with 12 the movement of air molecules; this Brownian motion increases with decreasing 13 fiber size (<0.5-pm diameter). 14 5. Electrostatic Precipitation: A special case of diffusion in which fiber motion 15 towards the airway surface is a function of static charge between the fiber and 16 airway surface. As with classic diffusion, this primarily occurs in the 17 gas-exchange region where airflow is negligible and electrostatic forces can 18 predominate. 19 20 21 Aerodynamic diameter (also called aerodynamic equivalent diameter) of fibers accounts 22 for the dimensional properties that influence the movement of the fiber's center of gravity 23 through the airways, so aerodynamic diameter is important in all depositional mechanisms. The 24 aerodynamic diameter is the diameter of a unit density (1 g/cm ) sphere that has the same 25 gravitational settling velocity as the particle of interest. Since the aerodynamic diameter informs 26 the deposition patterns of fibers, it is used in dosimetric modeling to determine the expected fiber 27 deposition in the respiratory tract. Impaction and interception, however, are also heavily 28 influenced by fiber length. Where the physical length of the fiber greatly exceeds the 29 aerodynamic diameter, impaction and interception can be underpredicted by modeling the center 30 of gravity of the fiber. Sedimentation is related to the mass of the fiber, as well as the 31 aerodynamic diameter, but generally occurs at lower velocities in smaller airways. Diffusion 32 occurs from interactions of the fibers with the movement of air molecules; this Brownian motion 33 increases with decreasing fiber size (<0.5-pm diameter). Electrostatic precipitation occurs when 34 fiber charges induce opposite charges on the airway surfaces and the fiber is drawn to the airway 35 walls (Lippmann. 1990). This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-5 DRAFT--DO NOT CITE OR QUOTE 1 For high aspect ratio fibers, like asbestos, the shape factor often approaches one and the 2 equation reduces to the aerodynamic diameter that is approximately equal to the nominal fiber 3 diameter.10 Therefore, in employing the information from Table 3-1 to high aspect ratio fibers, 4 one may get an idea of the depositional characteristic of fibers from the nominal diameter. By 5 definition, fibers have a greater aspect ratio than particles and as discussed, high aspect ratio 6 fibers may act significantly different than other particles with respect to some mechanisms of 7 deposition (e.g., impaction, interception, and electrostatic precipitation). Therefore, the 8 depositional characteristics of fibers are not characterized completely by aerodynamic diameter. 9 No equivalent depositional model, however, is yet available for fibers in the dimensional range 10 of asbestos that takes into consideration the increased sedimentation and impaction for high 11 aspect ratio particles. 12 Fibers enter the respiratory tract along with airflow through the nasal and oral passages. 13 The nasal passage, from the nostril to the pharynx, serves as a filter for some fibers with 14 diameters 5-30 pm. Clumps of fibers also could deposit in these regions. Many animal species, 15 including rats and mice, are obligate nose breathers, meaning that fibers pass only through the 16 nasal passages, and, therefore, are always subject to nasopharyngeal filtering. Humans, 17 monkeys, and dogs, among other species, breathe both orally and nasally (oronasal). Therefore, 18 larger fibers and clumps of fibers can bypass the upper respiratory tract filtering and be inhaled 19 directly into the larynx/traehea, especially during exertion (e.g., exercise or work), which may 20 further alter deposition by increased turbulence in the airways. This distinction is important 21 when comparing results of inhalation studies conducted in different species. 22 The conducting airways beyond the nasopharyngeal region include the trachea and 23 bronchi, which serially bifurcate into airways of decreasing internal diameters. The aerodynamic 24 diameter of fibers that can deposit in the tracheobronchial region is in the range of 1-5 pm. 25 Fibers with aerodynamic diameter <1 pm can deposit in the bronchioles and the alveoli (1CRP. 26 1994). 27 Generally, fibers with aerodynamic characteristics conducive to deposition in the 28 bronchioles and alveoli can cause pulmonary fibrosis and associated disease by either retention 29 in the alveoli or penetration into the peribronchiolar space. All fibers having an aerodynamic 10The physical properties o f a fiber that determine its aerodynamic transport are combined and defined as the aerodynamic diameter; one such property is the shape factor (ICRP. 1994). This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-6 DRAFT--DO NOT CITE OR QUOTE 1 diameter that is less than approximately 2 pm, which includes Libby Amphibole asbestos, meet 2 the physical criteria necessary for deposition in the deeper regions of the respiratory tract at the 3 level of the terminal bronchioles or alveoli. 4 Deposition of fibers in the alveolar region of the lung is consistent with radiological 5 findings in humans of fibrosis in the lower lung fields at early stages of disease. Deposition of 6 fibers in the alveoli can become limited when fiber length approaches 40 pm (Morgan et ah. 7 1978). Alveolar deposition of fibers with high aspect ratios and length ranging from less than 8 1 pm to greater than 200 pm long, however, has been recorded (Morgan et al.. 1978). In all 9 documented observations of fibers collected from either healthy or diseased individuals, short 10 fibers (<5 pm) were present in substantially greater numbers in lung tissue than were long fibers 11 (>5 pm) (Churg, 1982). Although information is limited on how fibers get to the pleura, fibers 12 observed in pleural tissue from mesothelioma cases are more likely to be short (<5 pm) (Suzuki 13 et al.. 2005). These observations could be due in part to the increased deposition of smaller 14 fibers or the breakage of larger fibers over time (Bernstein et al.. 1994; Davis. 1994). 15 The lung and nasal depositional differences are due in part to differences in airway 16 structure and breathing patterns across lifestages (i.e., children, adults), changing the depositional 17 pattern of different fiber sizes, possibly altering the site of action, and potentially resulting in 18 differential clearance and health effects (see Section 4.7). 19 Modeling of fiber deposition has been examined for various fiber types (e.g., refractory 20 ceramic fibers, chrysotile asbestos) (Sturm. 2009; Zhou et al.. 2007; Lentz et al.., 2003; Dai and 21 Yu. 1998; Yu et al... 1997; Coin et al.. 1992). but not for Libby Amphibole asbestos. In general, 22 the pattern of deposition for fibers is expected to have some similariti es to the well-studied 23 deposition pattern for essentially spherical particles (reviewed in ICRP) (1994). For example, 24 the multipath particle dose model (Brown et al.. 2005; Jarabek et al., 2005) uses information on 25 the physical properties of the particles (length and width [also called bivariate distribution] and 26 density), the anatomy and architectural features of the airways, airflow patterns that influence the 27 amount and the location of the deposition of the particles, and dissolution and clearance 28 mechanisms that are operative to estimate the retained dose in the target tissue. 29 This document is a draftfor re\'iew purposes only and does not constitute Agencypolicy. 3-7 DRAFT--DO NOT CITE OR QUOTE 1 3.2. CLEARANCE 2 3.2.1. Inhalation 3 3.2.1.1. Respiratory Tract 4 Once fibers deposit on the surface of the respiratory tract, they may be removed (cleared) 5 from the lungs in several ways--including physical clearance, dissolution, phagocytosis, or 6 encapsulation. Some of these mechanisms, such as dissolution of the fibers or removal via the 7 mucociliary apparatus, can result in the fibers being cleared from the body (see Figure 3-1). 8 Other clearance mechanisms may remove fibers from the surface of the respiratory tract but 9 result in transport of the fibers to other tissues by translocation. Translocation of fibers from the 10 terminal bronchioles and alveoli into the peribronchiolar space, lymph nodes, and pleura has 11 been implicated in disease causation (e.g., pleural plaques, mesothelioma) (Dodson et al.. 2001). 12 In human studies, the translocation of asbestos fibers following inhalation has been observed to 13 varying degrees throughout the pulmonary and extrapulmonary tissues of the respiratory system 14 (Dodson et al.. 2005; Dodson et al. 2001; Kohyama and Suzuki. 1991; Suzuki and Kohyama, 15 1991; Armstrong et al.. 1988). as well as other organs, including the brain, kidney, liver 16 (Miserocchi et al.. 2008). and ovaries (Langseth et al.. 2007). In many cases, the type of fiber 17 was not defined, and the individual exposure information is not available. Fibers that are not 18 cleared may remain at the epithelial surface or enter the parenchymal tissue of the lung. 19 Berry (1999) provided a revi ew of the animal toxicity literature specifically for fiber 20 clearance. There are limited data on clearance patterns based on autopsy studies in humans. 21 Two studies estimated clearance half-life for amphibole asbestos (~20 years) as compared with 22 chrysotile asbestos (~10 years) (Finkelstein and Dufresne. 1999; Churg and Vedah 1994); in 23 evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the half-life 24 for crocidolite to be between 5 and 10 years. Generally, studies have focused on determining the 25 size and type of asbestos retained in specific tissues (Suzuki et al.. 2005; McDonald et al.. 2001; 26 Suzuki and Yuen. 2001; Dumortier et al.. 1998; Gibbs et al... 1991; Dodson et al.. 1990) and did 27 not discuss changes in fiber content since exposure. Sebastien et al. (1980) concluded that lung 28 fiber burden could not be used as an accurate reflection of pleural fiber burden. 29 This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-8 DRAFT--DO NOT CITE OR QUOTE 1 3.2.1.1.1. Physical clearance o f fibers 2 Fibers deposited in the nasal passages can be removed by physical clearance. When 3 breathing occurs through the nose, many fibers are filtered by the turbulent airflow in the nasal 4 passages, impacting against the hairs and nasal turbinates, as well as becoming entrained in 5 mucus in the upper respiratory tract where they can be subsequently removed by mucociliary 6 action or reflexive actions such as coughing or sneezing. The mucociliary escalator removes 7 fibers through ciliary movement of the sticky mucus lining (Wanner et al.. 1996; Churg et al., 8 1989). Fibers removed from the conducting airways through this mechanism are coughed out or 9 swallowed and enter the digestive tract where they may adversely affect the gastrointestinal 10 tissue, enter the blood stream, or be excreted. Clearance of fibers via mucociliary action is rapid 11 and is usually complete within minutes or hours. FTowever, the mucociliary escalator extends 12 only down to the level of the terminal bronchioles and not to the alveoli. Therefore, particles 13 that reach the alveolar region of the lung cannot be cleared through this process. Fibers can also 14 translocate due to physical forces associated with respiration (Davis. 1989). 15 Some fibers are not cleared from the lung, leading to an accumulation with time (Case et 16 al.. 2000; Finkelstein and Dufresne. 1999; Jones et ah. 1988). The fibers that remain in the lung 17 may undergo a number of processes including translocation, dissolution, fragmentation, splitting 18 along the longitudinal axis, or encapsulation with protein and iron. Available data indicate 19 prolonged clearance from the lung of long (>5 pm) or short amphibole fibers (Coin et al.. 1994; 20 Tossavainen e ta l. 1994). The prolonged clearance times for long amphibole fibers have led 21 some investigators to conclude that long versus short amphibole fibers are predominant in the 22 cause of disease despite the relatively small numbers of these longer fibers in comparison to 23 short fibers (Mossman et al,, 2011; ATSDR. 2003b). Flowever, others argue that fibers of all 24 lengths induce pathological responses and urge caution in excluding, based on their length, any 25 population of fibers from consideration as possibly contributing to the disease process (Aust et 26 al.. 2011; Dodson et al.. 2003). Respirable-sized fibers of Libby Amphibole asbestos have been 27 identified in air samples from activity-based sampling from Libby, MT, and in airborne fibers 28 suspended from both Libby vermiculite concentrate and in the exfoliated product from that 29 concentrate. Based on fibers counted by the TEM analytical method (ISO 10312), the majority 30 of counted fibers are respirable (see Figure 2-12). 31 This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-9 DRAFT--DO NOT CITE OR QUOTE 1 3.2.1.1.2. Dissolution o f fibers 2 Dissolution, or the chemical breakdown of fibers, is another method of removal of fibers 3 from the lung. This process varies, depending on the chemical composition of the fibers, as well 4 as the physiological environment. Dissolution can occur in the lung's extracellular fluids or in 5 the macrophage phagolysosome. Studies performed in vitro to determine dissolution rate of 6 fibers attempt to mimic the extracellular lung fluids and macrophage-phagolysosome system to 7 understand the length of time that fibers remain in the system (Rendall and Du Toit 19941. 8 Studies have shown that dissolution occurs more rapidly for chrysotile fibers than for amphiboles 9 (Coffin et al.. 1983). Fibers can also be physically diminished through splitting or breakage. 10 These smaller fragments are then more easily removed by phagocytosis or translocation. 11 12 3.2.1.1.3. Removal o ffibers through phagocytosis 13 The principal clearance pathway for insoluble fibers deposited in the alveoli is through 14 phagocytosis by macrophages. Alveolar macrophages that have phagocytized insoluble fibers 15 migrate to the bronchoalveolar junctions where they enter onto the mucociliary escalator for 16 removal (Green. 1973). Alternatively, alveolar macrophages that have phagocytized insoluble 17 fibers can also migrate through the epithelial wall into the interstitial space and enter the 18 lymphatics (Green. 1973). 19 Alveolar macrophage cells engulf and transport deposited particles to the mucociliary 20 escalator or through the alveolar epithelium to the interstitial tissues, where they are removed or 21 translocated by the blood or lymphatics. Durable fiber impaction in these deeper regions also 22 stimulates activation of alveolar macrophage cells. In vitro and in vivo studies clearly indicate 23 that macrophage cells play a role in the translocation of fibers (Dodson et ak. 2000a; Castranova 24 etal.., 1996; Brody et al.. 1981; Bmilon et ah. 19791. These studies have demonstrated the 25 presence of asbestos fibers in cell cytoplasm where they can be transported in association with 26 cytoskeletal elements to the proximity of the cell nucleus. Small chrysotile fibers can also 27 penetrate the nuclear membrane (Malorni et al.. 19901. 28 A number of processes can disrupt the normal phagocytic function of the alveolar 29 macrophages. These processes include death or dysfunction of macrophages due to phagocytosis 30 of an excessive number of parti cles (often termed "overload") or highly reactive particles or an 31 attempted phagocytosis of fibers of lengths that exceed the dimensional capacity of the This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-10 DRAFT--DO NOT CITE OR QUOTE 1 macrophage (often termed "frustrated phagocytosis") (NIOSH, 2 0 1 1 ). All of these processes can 2 induce inflammatory and fibrogenic responses. Limited inhalation laboratory animal studies 3 exist at nonoverloading concentrations of fibers or particles; therefore, there is insufficient 4 information to determine mechanisms at these lower doses reviewed in Mossman et al. (2011). 5 6 3.2.1.1.4. Encapsulation o ffibers 7 Fibers that are too large to be easily engulfed by the alveolar macrophage can stimulate 8 the formation of "asbestos bodies." Asbestos bodies are fibers that, during prolonged residence 9 in the lung, have become coated with proteins, iron and calcium oxalate. Due to their iron 10 content, histological stains for iron have long been used to identify them in tissue; thus, they are 11 sometimes called "ferruginous bodies." The mechanisms that result in the formation of asbestos 12 bodies are poorly understood, although most appear to be formed around amosite fibers (Dodson 13 et al.. 1996). The iron in the coating, however, is derived from the asbestos fiber, cells, or 14 medium surrounding the fiber and can remain highly reactive (Lund et aL 1994; Ghio et ah. 15 1992). Asbestos bodies can remain in the lung throughout the lifetime of the exposed individual. 16 Asbestos bodies comprise a minor portion of the overall fiber burden of the lung, and, after the 17 fiber is fully coated, these fibers might or might not participate directly in asbestos disease. The 18 presence of iron in the coating, however, could provide a source for catalysis of reactive oxygen 19 species similar to that observed with fibers. 20 21 3.2.1.1.5. Translocation to extrapulmonary tissues 22 Clearance from one tissue may involve translocation to another tissue. For example, 23 following fiber deposition in the respiratory tract, fibers may then clear via translocation to 24 extrapulmonary tissues like the pleura. The specific mechanism and translocation route depend 25 both on fiber characteristics and the tissue of deposition. Whether or not fibers are translocated 26 appears to depend on their physical-chemical characteristics, including two-dimensional size 27 (length and width); durability; solubility; and reactivity. This translocation is aided by high 28 durability and an inflammation-induced increase in permeability but is hindered by fibrosis. 29 Deposition occurs in the respiratory tract as described above; translocation from the respiratory 30 tract may, in turn, lead to fibers `depositing' in extrapulmonary sites. This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-11 DRAFT--DO NOT CITE OR QUOTE 1 Apparent translocation of fibers throughout the respiratory tract is evident from 2 experimental animal research done by several investigators following exposure by both 3 intrapleural injection and inhalation (Miserocchi et al.. 2008; Holt, 1982; Smith et a).., 1980; 4 Bignon et ah, 1979; Smith et a.L, 1979; Smith and Hubert, 1974). The data from most studies 5 show that fibers can--and do--translocate among tissues and organs and move by both 6 physiological and physical mechanisms (Holt, 1983; Holt 1982; Cook and Olson. 1979). 7 Conflicting results from another study, however, indicate no evidence of fiber translocati on from 8 the central to peripheral compartments following inhalation exposure in rats, although this could 9 be due to the short duration of the study (29 days postexposure) (Coin et al., 1992). 10 Translocation of fibers to extrapulmonary tissues has been studied in multiple studies; 11 however, the mechanism is still unknown. This was more recently reviewed by Miserocchi et al. 12 (2008). Fibers have been measured in extrapulmonary tissues including pleural plaques and 13 mesothelial tissue (i.e., pleural or peritoneal) in miners, brake workers, insulation workers, and 14 shipyard workers (Roggli et al., 2002; Dodson et al,, 2000b; Churg, 1994; Kohyama and Suzuki, 15 1991). These studies found fibers at all locations analyzed, with increased levels of amphibole 16 as compared to chrysotile in the parenchyma when subjects were exposed to a mixture of both 17 fiber types. Amphibole fibers, however, were less prevalent in the pleura and mesothelial tissues 18 (Kohyama and Suzuki, 1991; Sebastien. et al, 1989; Armstrong et al., 1988; Churg, 1988; 19 Bignon et al,, 1979). Few studies have examined the size distribution of fibers translocated to 20 specific tissues. For example, one early study suggested that the longer amphibole fibers 21 predominate in the lung while shorter chrysotile fibers are found in the pleura (Sebastien et al., 22 1980); others showed that the fiber-length distribution was the same by fiber type regardless of 23 location (Kohyama and Suzuki, 1991; Bignon et al., 1979). 24 Transplacental transfer of both asbestos (chrysotile, tremolite, actinolite, and 25 anthophyllite) and nonasbestos fibers has been shown to occur in humans, as measured in the 26 placenta and in the lungs of stillborn infants (Hague et al., 1998; Hague et al,, 1996; Hague et al., 27 1992; Hague and Kanz, 1988). It is hypothesized that maternal health might influence the 28 translocation of fibers, as some of the mothers had preexisting health conditions (e.g., 29 hypertension, diabetes, or asthma) (Hague et al,, 1992). This group also measured transplacental 30 translocation in a mouse study and observed early translocation of crocidolite fibers through the 31 placenta in animals exposed via tail-vein injection (Hague et al., 1998). These studies did not This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-12 DRAFT--DO NOT CITE OR QUOTE 1 evaluate the source or levels of exposure, only the presence of fibers in the body during early 2 lifestages in mice and humans. 3 Sebastien et al. (1980) found chrysotile was the predominant fiber in parietal pleura of 4 autopsy cases, while the amphibole fibers found in the lungs ranged from 0 to 100% (mean 5 56%). Bignon et al. (1979) found similar distributions but also found increased amphibole fibers 6 in the associated lymph nodes. In this study, chrysotile and amphibole fibers were found 7 together in the lung parenchyma and alveolar spaces. Other studies show fewer amphibole fibers 8 at the site of diseased tissue in the pleura and mesothelial tissue than chrysotile (Kohyama and 9 Suzuki. 1991; Chum. 1988). Sebastien et al. (1989) examined fiber types in lungs of chrysotile 10 textile and mining workers from South Carolina and Quebec, respectively, to better understand 11 the unknown reason for differences in disease risk in each cohort. Both groups were exposed to 12 similar material, yet the South Carolina cohort had a much greater risk of respiratory cancer. 13 This study examined only lungs, although some of those exposed had nonpulmonary cancers. 14 Overall, the number of tremolite fibers retained in the lungs was higher than that of chrysotile 15 fibers retained in the lungs in both cohorts. Size distribution showed that most fibers measured 16 were 5.8-8.0 pm long, although measurements were not made for anything smaller than this. 17 Tremolite fibers had a greater mean diameter in both cohorts (0.35 pm) as compared to 18 chrysotile (0.10 pm), while chrysotile had more "Stanton" fibers (25.2-31.8%) as compared to 19 tremolite (5.9-6.3%). Stanton fibers are defined as >8 pm long and <0.25 pm in diameter 20 (Stanton et al. (1981)1. reviewed in Appendix D). 21 22 3.2.1.2. Pleural Cavity and Extrapulmonary Sites 23 Studies have demonstrated fiber clearance from the respiratory tract may lead to 24 translocation to the pleural cavity and extrapulmonary sites. For example, in a study comparing 25 fiber burden in the lung, thoracic lymph nodes, and pleural plaques, Dodson et al. (1990) 26 observed that the average-length fiber found in the lung (regardless of type) was longer than 27 those found in the lymph nodes or plaques. Most fibers at all three sites were short (<5 pm). A 28 later study by this group (i.e.. Dodson et al.. 2000b) examined tissue from 20 individuals with 29 mesotheliomas, most with known asbestos exposures. Seventeen of the cases (85%) had 30 asbestos fibers in at least one other extrapul monary site. The most prevalent type of asbestos in 31 the mesentery was amosite, and the second most prevalent was chrysotile. Tremolite was also This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-13 DRAFT--DO NOT CITE OR QUOTE 1 found, to some degree, in the mesentery and omentum, and in the lung. Dodson et al. (2005) 2 examined parenchymal lung tissue from a cohort of 54 mesothelioma patients and determined 3 the presence of asbestos in all patients analyzed. However, very little information is known 4 about the specific mechanisms of fiber clearance and/or translocation from the pleural cavity and 5 extrapulmonary sites, although many studies examining these tissues have observed fibers in 6 multiple tissue sites [reviewed in Aust et al. (2011). Case et al. (2011)1. Following intrapleural 7 injection of fibers in rats, Bignon et al. (1979) used transmission electron microscopic evaluation 8 following serial sacrifice to monitor migration of fibers from the pleural cavity to the lung 9 parenchyma. 10 11 3 .2 .2 . I n g e s ti o n 12 Although ingestion is a potential route of exposure, limited research has examined 13 clearance (e.g., translocation) of fibers following ingestion, and no clearance studies are 14 available specific to Libby Amphibole asbestos. An early study to examine the tissue response 15 to asbestos fibers is not truly representative of a natural ingestion exposure, as the researchers 16 directly injected a suspension of amosite fibers into the duodenal wall (Meek and Grasso. 1983). 17 This study, however, also examined oral ingestion of amosite in healthy animals and those with 18 gastrointestinal ulcers to determi ne if translocati on of fibers occurs through ulcers. Following 19 injection of amosite, granulomatous lesions were observed. Ingestion of the same material 20 resulted in no such lesions or in any other histopathological changes in either healthy or 21 compromised rats. Thus, no translocation was observed from either the healthy or the 22 compromised rat gastrointestinal tracts in this study. A later International Agency for Research 23 on Cancer study (Truhaut and Chouroulinkov. 1989) examined the effects of chrysotile and 24 crocidolite ingestion in Wistar rats. No translocation was observed. No further studies have 25 been found on clearance or translocation of fibers from the gastrointestinal tract. 26 27 3 .2 .3 . D e r m a l C o n t a c t 28 No studies of dermal clearance or translocation have been reported in the published 29 literature. 30 This document is a draftfor review purposes only and does not constitute Agencypolicy. 3-14 DRAFT--DO NOT CITE OR QUOTE 1 3.3. SUMMARY 2 Although oral and dermal exposure to fibers does occur, inhalation is considered the main 3 route of human exposure to mineral fibers, and, therefore, it has been the focus of more fiber 4 toxicokinetic analyses. Exposure to Libby Amphibole asbestos is presumed to be through all 5 three routes of exposure; this assessment specifically focuses on the inhalation pathway of 6 exposure. Generally, fiber deposition in the respiratory tract is fairly well defined based on fiber 7 dimensions and density, although the same cannot be said for fiber translocation to 8 extrapulmonary sites (e.g., pleura). The deposition location within the pulmonary and 9 extrapulmonary tissues plays a role in the clearance of the fibers from the organism. 10 Fiber clearance from the respiratory tract can occur through physical and biological 11 mechanisms. Limited mechanistic information is available on fiber clearance mechanisms in 12 general, and no information specific to clearance of Libby Amphibole asbestos fibers is 13 available. Fibers have been observed in various pulmonary and extrapulmonary tissues 14 following exposure, suggesting translocation occurs to a variety of tissues. Studies have also 15 demonstrated fibers may be cleared through physical mechanisms (coughing, sneezing) or 16 through dissolution of fibers. 17 Multiple fiber characteristics (e.g., dimensions, density, and durability) play a role in the 18 toxicokinetics of fibers. For this reason, careful attention has been paid to these fiber 19 characteristics when analyzing research studies on Libby Amphibole asbestos and asbestiform 20 tremolite, an amphibole fiber that comprises part of Libby Amphibole asbestos (see 21 Appendix D). No toxicokinetic data are available specific to Libby Amphibole asbestos, 22 tremolite, richterite, or winchite. When available, this information is presented in the discussion 23 of each study in relation to the toxic endpoints described. 24 This document is a draftfor review purposes only and does not constitute Agency policy. 3-15 DRAFT--DO NOT CITE OR QUOTE 1 4. HAZARD IDENTIFICATION OF LIBBY AVIPMIBOIT ASBESTOS 2 Several human studies are available that provide evidence for the hazard identification of 3 Libby Amphibole asbestos.11 This discussion focuses primarily on data derived from studies of 4 people exposed to Libby Amphibole asbestos--either at work or in the community. The adverse 5 health effects in humans are supported by the available Libby Amphibole asbestos experimental 6 animal and laboratory studies. Libby Amphibole asbestos contains winchite (84%), with lesser 7 amounts of richterite (11%) and tremolite (6%) with trace amounts of magnesioriebeckite, 8 edenite, and magnesio-arfvedsonite (Meeker et al., 2003) (see Section 2.2.3 for a more complete 9 discussion). Adverse health effects from tremolite exposure have been reported in both human 10 communities and laboratory animals; these effects are consistent with the human health effects 11 reported for Libby Amphibole asbestos. Studies examining the health effects of exposure to 12 winchite or richterite alone were not available in the published literature. The presentation of 13 noncancer and cancer health effects provides a comprehensive review of adverse health effects 14 observed from exposures to Libby Amphibole asbestos. 15 16 4.1. STUDIES IN HUMANS--EPIDEMIOLOGY 17 The Libby Amphibole asbestos epidemiologic database includes studies conducted in 18 occupational settings examining exposures to workers and community-based studies, which can 19 include exposures to workers, exposures to family members of workers, and exposures from 20 environmental sources. Occupational epidemiology studies exist for two worksites where 21 workers were exposed to Libby Amphibole asbestos. These worksites include the mine and mill 22 at the Zonolite Mountain operations near Libby, MT, and a vermiculite processing plant in 23 Marysville, OH. Worker cohorts from each site and the study results are described in 24 Section 4.1.1. Community-based studies include community health consultations for Libby, MT 25 conducted by the Agency for Toxic Substances and Disease Registry (ATSDR), including an 26 evaluation of cancer mortality data, and a health screening of current and former area 27 residents--including workers--that collected medical and exposure histories, chest X-rays, and 28 pulmonary function tests (ATSDR, 2001b, 2000) (see Section 4.1.2). ATSDR, in conjunction 11The term "Libby Amphibole asbestos" is used in this document to identity the mixture of amphibole mineral fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-1 DRAFT--DO NOT CITE OR QUOTE 1 with state health departments, also conducted health consultations for 28 other communities 2 around vermiculite processing plants that were potentially exposed to Libby Amphibole asbestos 3 (see Section 4.1.4). These health consultations consisted of analyses of cancer incidence or 4 mortality data; results from nine of these studies are currently available. 5 No occupational studies are available for exposure to tremolite, richterite, or winchite 6 mineral fibers individually or as a mixture exposure, other than Libby Amphibole asbestos. 7 Communities, however, have been exposed to tremolite and other mineral fibers from natural 8 soils and outcroppings. Tremolite asbestos-containing soil has been used in whitewash in 9 interior wall coatings in parts of Turkey and Greece. Studies in these areas published as early as 10 1979 reported an increased risk of pleural and peritoneal malignant mesothelioma (Sichletidis et 11 al.. 1992; Bails et aL 1987; Langer et aL 1987; Baris et al.. 1979). More recent studies of 12 communities exposed to tremolite and chrysotile fibers report excess lung cancer and 13 mesothelioma (1.3- and 6.9-fold, respectively) (Hasanoglu et al.. 2006). Other studies reported 14 pleural anomalies in residents exposed to naturally occurring asbestos, which includes actinolite, 15 tremolite, and anthophyllite (Metintas et al.. 2005; Zeren et al.. 2000). Clinical observations 16 include a bilateral increase in pleural calcification accompanied by restrictive lung function as 17 the disease progresses, a condition known as "Metsovo lung," named after a town in Greece 18 (Constantopoulos et al.. 1985). In one community, the prevalence of pleural calcification was 19 46% (of 268 residents), increasing with age to 80% in residents over 70 (Langer et al., 1987). 20 Both tremolite and chrysotile were identified in bronchoalveoiar lavage fluid of 65 residents 21 from different areas of Turkey who were environmentally exposed (Dumortier et al., 1998). The 22 health effects observed in communities with environmental and residential exposure to tremolite 23 are consistent with health effects documented for workers exposed to commercial forms of 24 asbestos. 25 26 4.1.1. Studies of Libby, MT Vermiculite Mining Operation Workers 27 Several studies of mortality from speci fic di seases among workers in the Libby, MT 28 mining operations have been conducted, beginning in the 1980s with the studies by McDonald 29 et al. (1986a) and Amandus and Wheeler. (1987). McDonald et al. (2004, 2002) published an 30 update with mortality data through 1999, and Sullivan (2007) updated the cohort originally 31 described by Amandus and Wheeler (1987) (referred to in this assessment as the Libby worker This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-2 DRAFT--DO NOT CITE OR QUOTE 1 cohort) with mortality data through 2001. Additionally, Larson et al. (2010b) reconstructed a 2 worker cohort and analyzed mortality through 2006 in this same study population, while another 3 study examined changes in lung abnormalities using X-rays taken between 1955 and 2004 of 88 4 workers (Larson et al., 2010a). 5 6 4.1.1.1. Description o f Mining and Milling Operations 7 The vermiculite mining and milling operations have been described in considerable detail 8 (ATSDR. 2000; Amandus et al.. 1987a). An open-pit vermiculite mine began limited operations 9 in 1923, and production increased rapidly between 1940 and 1950. This mine is located on 10 Zonolite Mountain, several miles east of Libby (ATSDR. 2000). The Kootenai River runs 11 between the town and the mine. The mining and milling operations continued until 1990 12 (ATSDR. 2008b. 2000). 13 The drilling and blasting procedures used in the strip-mining operations generated 14 considerable dust exposures, although the mining operations had lower intensity exposures 15 compared to the milling operations. Amandus et al. (1987a) noted that in 1970, a new drill with 16 a dust-control bagging system aimed at limiting workplace exposure was introduced to the 17 mining operations. Another aspect of the operations was the loading of ore for railroad 18 shipment. From 1935-1950, railroad box cars were loaded at a station in Libby. In 1950, the 19 loading stati on was moved to a loading dock on the Kootenai River, 7 miles east of town. Tank 20 cars were used from 1950-1959 and then switched to enclosed hopper cars in 1960. 21 The milling operations used a screening or sifting procedure to separate vermiculite 22 flakes from other particles and increase the concentration of vermiculite from approximately 23 20% in the bulk ore to 80-95% in the resulting product. A dry mill began operating in 1935, and 24 a wet mill began operating in the 1950s in the same building as the dry mill. One of the primary 25 changes in the conditions in the dry mill was the installation of a ventilation fan in 1964. 26 Exposure to asbestos inside the mill was estimated to be 4.6 times higher preceding this 27 installation (McDonald et al.. 1986a). This ventilation fan resulted in higher amphibole fiber 28 exposures in the mill yard until 1968, when the exhaust stack for the fan was moved. Other 29 changes to the milling operations in the 1970s included replacement of hand bagging and sewing 30 with an automatic bagging machine (1972), pressurization of the skipper control room used for 31 transferring the ore concentrate from the mill to a storage site (1972), and construction of a new This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-3 DRAFT--DO NOT CITE OR QUOTE 1 wet mill (1974). Closing of the old dry and wet mills in 1976 had a substantial impact on 2 exposures at the worksite. In 1974, a new screening plant used to size-sort the ore concentrate 3 was constructed at the loading dock near the river. Two processing plants operated within the 4 town of Libby (ATSDR. 2001b). These expansion or exfoliation plants heated the ore 5 concentrate, resulting in additional release of the Libby Amphibole asbestos fibers in the area. 6 7 4.1.1.2. Exposure Estimation 8 In the early 1980s, two research groups conducted parallel studies of the mortality 9 experienced by workers in the Libby mining and milling operations. One study was undertaken 10 by the National Institute of Occupational Safety and Health (NIOSH) (Amandus et aU 1987a; 11 Amandus and Wheeler. 1987; Amandus et al,, 1987b) and the other by researchers from McGill 12 University (McDonald et aU 1986a; McDonald et ak, 1986b). The exposure assessment 13 procedures used by the two groups relied on the same exposure measurements and used similar 14 assumptions in creating exposure estimates for specific job activities and time periods (see 15 Table 4-1). In brief, available air sampling data were used to construct a job-exposure matrix 16 assigning daily exposures (8-hour time-weighted average) for identified job codes based on 17 sampling data for specific locations and activities. Varying job codes and air exposures were 18 used for different time periods as appropriate to describe plant operations. Individual exposure 19 metrics (e.g., cumulative exposure) were calculated using the work history of each individual in 20 the study in conjunction with the plant job-exposure matrix. The specific study details for the 21 Libby, MT worker cohort are described in more detail below, with differences between the 22 research groups highlighted 23 Before 1970, exposure estimates were based on midget impinger samples taken primarily 24 in the dry mill by state and federal inspectors. Total dust samples were measured as million 25 particles per cubic foot (mppcf) by the midget impinger method. Amandus et al. (1987a) 26 describe the period during which most of the midget impinger measurements were made as 27 1962-1967, and McDonald et al. (1986a) describe this period as 1962-1969, with a few This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-4 DRAFT--DO NOT CITE OR QUOTE Table 4-1. Exposure assessment methodologies used in evaluations of Libby, MT (see Section 4.1.1) and Marysville, OH (see Section 4.1.2) worker cohorts Asbestos fiber quantification and job-exposure Operation and study cohort classification Studies using methodology Libby, MT mining and milling operations; NIOSH cohort Exposure based on phase-contrast microscopy of fibers >5 pm long and aspect ratio >3:1 (1967-1982), and midget impinger data (1956-1969). Samples assigned to 25 "occupation locations" to estimate exposures for specific jobs and time periods 1945-1982. Membrane-filter measurement to impinger conversion ratio: 4.0 fibers/cc per mppcf. Cumulative exposure reported in units o f liber-years (equivalent to the unit o f flbers/cc-years EPA is using for all studies). Amandus et al. (1987a: 1987b): Amandus and Wheeler (1987) Libby, MT mining and milling operations; NIOSH cohort Modification to Amandus et al. (1987a) iob classification: laborers and "unknown" jobs assigned weighted-average exposure for all unskilled jobs in work area (if known) during calendar time period, rather than lower mill yard exposure. Weights based on the number o f workers assigned to unskilled jobs during same calendar time period. Sullivan (2007); Mooleavkar et al. (2010) Libby, MT mining and milling operations; ATSDR cohort assembled from W.R. Grace & Co. records Extension o f Amandus et al. (1987a) exposure data, with additional application o f exposure estimates to job titles from early 1980s through 1993. Larson et al. (2010b: 2010a) Libby, MT mining and milling operations; McGill University cohort Similar to Amandus et al. (1987a). except with 28 "occupation locations," and conversion ratio = 4.6 for dry mill pre- and post 1964. Cumulative exposure reported in units o f fibers/ml-years (equivalent to the unit o f fibers/cc-years EPA is using for all studies). McDonald et al. (2004. 2002; 1986a: 1986b) Marysville, OH fertilizer production facility using Libby, MT vermiculite Libby, MT vermiculite ore used in the plant from around 1960 to 19804 Industrial hygiene monitoring began 1972 (based on fibers >5-pm long, diameter <3 pm, aspect ratio >3:1). Breathing zone samples used after 1976. Fiber analysis by PCM. Lockev et al. (1984); Rolls et al. (2008) 1 2 'Rolls et al. (2008) use 1963 as the beginning date o f the use o f Libby, MT vermiculite at the Marysville, OH plant, 3 based on information from ATSDR (2008b. 2005b). Lockey et al. (1984) used 1957 as the beginning date. 4 Subsequent to these publications, additional infonnation was used to conclude that the beginning date for use of 5 Libby vermiculite ore was 1959 (see Appendix F). 6 7 NIOSH = National Institute for Occupational Safety and Health; PCM = phase contrast microscopy. 8 This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-5 DRAFT--DO NOT CITE OR QUOTE 1 additional measures in earlier years.12 The number of samples available before 1970 was 2 336 (Amandus et al.. 1987a). Membrane-fdter air samples for fibers, taken at various locations 3 within the operations, began in 1967, and data are available from company records as well as 4 State and Federal Agencies (see Table 4-2). Stationary and short-term (i.e., 20-minute to less 5 than 4-hour) measurements were primarily used prior to 1974. The number of membrane-filter 6 samples available was 4,116. Air samples collected through membrane filters were analyzed by 7 phase contrast microscopy (PCM) to visually count fibers greater than >5-pm long and having an 8 aspect ratio >3:1 (Amandus et al,, 1987a).13 PCM methods from the 1960s allowed reliable 9 characterization of fibers with widths greater than approximately 0.4 pm (Amandus et al,, 1987a: 10 Rendall and Skikne, 1980). Further standardization of the PCM method provides better 11 visualization of thinner fibers, and 0.25 pm width is considered the limit of resolution for fiber 12 width OPCS. 19861. 13 14 15 Table 4-2. Source of primary samples for fiber measurements at the Libby 16 mining and milling operations 17 Source Unit of measurement Years Number of samples State o f Montana mppcf 1956-1969 336 NIOSH fibers/ccb 1967-1968 48 M ESAM SHAcd frbers/cc 1971-1981 789 Company records fibers/cc 1970-1982 3,279 18 19 M illion particles per cubic foot o f air, sampled by a midget impinger apparatus and examined by light 20 microscopy. 21 'Fibers per cc o f air drawn through a filter and examined under a phased contrast light microscope. Objects 22 >5 p and with an aspect ratio >3 were reported as fibers (see Section 2 for details). 23 M ESA: U.S. Mining and Enforcement and Safety Administration (former name o f MSHA). 24 dMSHA: U.S. Mining and Safety Administration. 25 26 Source: Amandus et al. (1987a). 27 28 '"Amandus et al. (1987a) indicates that one sample was available from 1942, and additional samples were available after 1956; McDonald et al. (1986a) indicates that addi tional samples were available from 1944,1956, and 1958. "Amandus et al. (1987a) indicate (page 12, 4lnfull paragraph) that fibers > 5-pm long and with an aspect ratio >3 were measured. The actual value o f the aspect ratio used by Amandus et al. could have been >3 because the criterion for the NIOSH recoimnended exposure limit is based on an aspect ratio o f >3, but EPA is reporting here the information that was in the Amandus et al. a 987a) publication. This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-6 DRAFT--DO NOT CITE OR QUOTE 1 The samples taken from specific work locations within the plant were used to estimate 2 exposures in specific jobs and time periods based on professional consideration of temporal 3 changes in facilities, equipment, and job activities. The analysis by McDonald et al. (1986a) was 4 based on 28 occupation locations, while the work of Amandus et al. (1987a) was based on 5 25 occupation locations. These were defined to categorize tasks and locations across the mining, 6 milling, and shipping operations to group like tasks, with respect to exposure potential, for 7 evaluation. Both research groups established similar location operations for the Libby cohort. 8 For the years after 1968, data from filter samples were available for all locations, and NIOSH 9 researchers used the average (arithmetic mean) exposure when more than one sample was 10 available for a given location or job task and time period. McDonald et al. (1986a) used an 11 alternative procedure described by Oldham (1965) to estimate the mean of log-normal 12 distributions. 13 For exposures occurring prior to 1968, different procedures had to be used to estimate 14 exposures at the various locations because measures from sample filters were not available from 15 this earlier period. McDonald et al. (1986a) estimated pre-1968 exposure measurements for 16 26 location operations; assumptions were made and estimates based on data from later years or 17 related operations, although these assumptions are not stated by the authors. McDonald et al. 18 (1986a) did recognize the uncertainty in these calculations, and, for four areas, (drilling, ore 19 loading, river dock, and bagging plant), provided high and low estimates. Amandus et al. 20 (1987a) interviewed company employees, considered relative exposure levels between locations 21 post 1968 employing best available judgment to estimate task specific exposure levels. 22 Amandus et al. (1987a) expanded the procedures described in McDonald et al. (1986a) to 23 estimate pre-1968 exposures for four location operations (drilling, ore loading, river dock, and 24 bagging plant). "Low" and "high" estimates were generated using different assumptions; the 25 detailed results for the various assumptions were not presented, but the differences between them 26 were described by the authors as "slight," and the results presented were based on the high 27 estimate of exposure. Their decisions and specific assumptions are detailed (Amandus et al.. 28 1987a). The authors acknowledge there is uncertainty in exposure estimates prior to 1968 for 29 many of these locations. They do note that variability in sample results for the midget impinger 30 was low and that, in general, sample variability was low for fiber air-sampling results for areas 31 where the greatest numbers of employees worked (mill, service area, loading and bagging). This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-7 DRAFT--DO NOT CITE OR QUOTE 1 To estimate dry mill exposures prior to 1967, when fiber counts from phase contrast 2 microscopy air samples began to be used to measure exposures, Amandus et al. (1987a) 3 established a conversion factor from total dust counts (mmpcf) to fiber counts (fibers/cc). The 4 conversion ratio was based on a comparison of 336 impinger samples taken in 1965-1969 and 5 81 filter samples taken in 1967-1971. Both sets of samples were taken in the dry mill. Using 6 different subsets of the samples (i.e., different years) resulted in ratios that ranged from 7 1.9 fibers/cc: 1.0 mppcf to 11.5 fibers/cc: 1.0 mppcf. The ratio based on the average fiber counts 8 from air samples (1967-1971) to the average total dust measurements in sample years 9 1965-1969 was 4.0 fibers/cc: 1.0 mppcf. This was the ratio used in the analyses in the NIOSH 10 studies (Amandus et al.. 1987a: Amandus and Wheeler, 1987: Amandus et al.. 1987b) because it 11 allowed for the use of the greatest amount of data from overlapping time periods, while 12 controlling for the reduced exposure levels after 1971 where fiber count based on phase contrast 13 microscopy--but not midget impinger data--were available. This dust-to-fiber conversion 14 factor was only used to estimate exposures in the dry mill. The resulting exposure 15 concentrations of 168 fibers/cc in 1963 and all prior years and 35.9 fibers/cc in 1964-1967 were 16 applied to dry mill exposures (Amandus et al.. 1987a). 17 McDonald et al. (1986a) used a different procedure, based on the estimated reduction in 18 dust exposure with the installation of the ventilation system in 1964. Rather than develop a 19 direct dust-to-fiber conversion factor, they observed that total dust levels dropped approximately 20 4.6-fold after the installation of ventilation in the dry mill. Therefore, exposures in the dry mill 21 prior to 1965 were calculated as 4.6 times the fiber exposures measured by PCM between 1970 22 and 1974 (22.1 fibers/cc) resulting in estimated dry mill exposures of 101.5 fibers/cc prior to 23 1965 (McDonald et al,, 1986a). 24 Exposure estimates for each location operation derived from sampling data and history of 25 changes in control measures were used to develop a job-exposure matrix that estimated exposure 26 in fibers/cc for each job code during several calendar time periods. Jobs were mapped to 27 operation/location based on estimated time spent in different job tasks, thus estimating an 8-hour 28 time-weighted average exposure for each job during several calendar time periods. Job histories 29 from date of first employment to 1982 were used with the job-exposure matrix to develop 30 cumulative exposure estimates for each worker. 31 This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-8 DRAFT--DO NOT CITE OR QUOTE 1 4.1.1.2.1. Characteristics o f historical fiber exposures 2 The resulting exposure estimates presented by both research groups, and the job-exposure 3 matrices used in calculating cumulative exposure for the cohort are based on fiber counts by 4 phase contrast microscopy analysis of air filters. As discussed in Section 2 (see Text Box 2-1), 5 phase contrast microscopy analysis does not distinguish between fiber mineralogy or 6 morphology and all fibers >5 pm in length with an aspect ratio of 3:1 or greater are included. 7 Both researcher groups analyzed fibers available at the facility in order to identify the mineral 8 fibers in the air samples. 9 Transmission electron microscopy14 (TEM) analysis of airborne asbestos fibers indicated 10 a range of fiber morphologies--including long fibers with parallel sides, needlelike fibers, and 11 curved fibers (McDonald et al.. 1986a). Of the fibers examined by TEM, >62% were >5 pm in 12 length and a wide range of dimensional characteristic were noted: length (1-70 pm), width 13 (0.1-2 pm), and aspect ratios from 3-100. Energy dispersive spectroscopy used to determine the 14 mineral analysis indicated that the fibers were in the actinolite-tremelite solid-solution series, but 15 sodium rich (McDonald et al.. 1986a). This analysis is consistent with the current understanding 16 of amphibole asbestos found in the Libby mine (see Section 2.2.3). 17 At the time of their study, when exposure concentrations were reduced to generally less 18 than 1 fiber/cc, Amandus et al. (1987a) obtained eight air filters from area air samples collected 19 in the new wet mill and screening plant (provided by the mining company). These samples were 20 analyzed by phase contrast microscopy using the appropriate analytical method for the time 21 (NIOSE1 Physical and Chemical Analytical Method No, 239). From early method development 22 through current PCM analytical techniques, the Public Elealth Service, Occupational Safety and 23 Flealth Administration and NIOSH methods have defined a fiber by PCM analysis as having an 24 aspect ratio >3:1 (NIOSEf 1994a: Edwards and Lynch, 1968). Amandus et al. (1987a) reported 25 the dimensional characteristics of the fibers from these filters including aspect ratio, width, and 26 length (see Table 4-3). Data for 599 fibers from the 8 area air samples collected in the wet mill 27 and screening plant are provided. These data are limited in one sense by the minimum diameter ^Transmission electron microscopy (TEM) utilizes a high-energy electron beam to irradiate the sample. This allows visualization o f structures much smaller than can been seen under light microscopy. TEM instruments may be fitted with two supplemental instruments that allow for a more complete characterization o f structure than is possible under light microscopy: energy dispersive spectroscopy (EDS) and selected area electron diffraction (SAED). .. This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-9 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-3. Dimensional characteristic of fibers from air samples collected in 2 the veriniculite mill and screening plant, Libby, MTa 3 Fiber length (pm) Fiber width (pm) Aspect ratio Range Total counted Percent (%) Range Total counted Percent (%) Range Total counted Percent (%) 4.98-7.04 54 9 0.44-0.62 406 68 5-10 24 4 7.04-9.96 109 18 0.62-0.88 151 25 10-20 176 29 9.96-14.08 107 18 0.88-1.24 27 5 20-50 305 51 14.08-19.91 111 19 1.24-1.76 14 2 50-100 84 14 19.91-28.16 90 15 1.76-2.49 0 0 >100 10 2 28.16-39.82 65 11 >2.49 1 0 39.82-66 46 8 66-88 10 2 >88 7 1 4 5 aFibers were viewed and counted by Phase Contrast Microscopy. 6 7 Source: Amandus et al. (1987a). 8 9 10 and length cutoffs (>4.98-pm long, >0.44-pm wide, aspect ratio >3.0).15 Even with these greater 11 than 10:1, with 16% greater than 50:1 aspect ratio. Only 7% of the fibers had a width greater 12 than 0.88 pm, with one fiber reported of the 559 with a width greater than 1.76. It should be 13 noted that as N10SH was examining PCM visible fibers, these data do not give the full fiber-size 14 di stribution of Libby Amphiboie asbestos fibers (see Section 2.2.3). 15 16 4.1.1.2.2. Descriptions o f cohorts 17 The cohort studi es conducted in the 1980s were similar in terms of exposure assessment 18 (as described in the previous section, Table 4-1), and other aspects of the study design (see 19 Table 4-4). Both studies included workers who had worked for at least 1 year. Amandus and 20 Wheeler (1987) included men hired before 1970 (n = 575), with follow-up through 21 December 31, 1981. McDonald et al. (1986a) included men hired before 1963 (n = 406) with 22 follow-up through 1983. A later analysis (McDonald et al., 2004) extended this follow-up 23 through 1999. 15 See footnote 3, page 4 -6 . This document is a draftfor review purposes only and does not constitute Agencypolicy. 4-10 DRAFT--DO NOT CITE OR QUOTE Table 4-4. R espiratory (lung) cancer m ortality and exposure-response analyses based on studies of the vermiculite mine workers in Libby, M Ta This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-11 Reference(s) Amandus and Wheeler (1987) Inclusion criteria and design details Standardized mortality ratio (SMR) (95% Cl) Men, hired before 1970, worked at least one year, follow-up through 1982 (;? = 575); 161 deaths (159 with death certificates). Mean duration: 8.3 years (0 worked less than 1 year). Mean fiber-years: 200.3. 12 female workers not included in this analysis. No exclusions: All cancer (// = 38) SMR: 1.3 (0.9, 1.8) Lung (;? = 20) SMR: 2.2 (1.4. 3.4) 20 or more years since first hire (latencv): Lung (;? = 12) SMR: 2.3 (/?<0.05) Exposure-response analyses-- lung cancer No exclusions: Cumulative Exposure 0.0-49 fibers/cc-yrs 50-99 fibers/cc-yrs 100-399 fibers/cc-yrs >400 fibers/cc-yrs n SMR (95% CI)b 6 1.5 (not reported) 2 1.6 (not reported) 2 1.1 (not reported) 10 5.8 (not reported, but p < 0.01) 20 or more years since first hire (20-year latency) Cumulative Exposure n SMR(95% CI)b 0.0-49 fibers/cc-yrs 2 0.85 (not reported) 50-99 fibers/cc-yrs 2 2.3 (not reported) 100-399 fibers/cc-yrs 1 1.1 (not reported) >400 fibers/cc-yrs 7 6.7 (not reported, but p < 0.01) McDonald et al. (2004); McDonald et al. (1986a) Men, hired before 1963, worked at least one year (;? = 406); follow-up through 1999 (McDonald et al.. 2004): 165 deaths before July 1983 (163 with death certificates); 120 deaths July 1983-- 1998 coded by nosologists using ICD-8 classifications; cause o f death for deaths from 1983-1998 obtained from National Death Index. Mean duration: 8.7 years (0 worked less than 1 year). Meanfiber-yrs: 144.6. Respiratory (;/ = 44) SMR: 2*4 (1.7, 3.2) In a linear regression analysis o f data with at least 20 years latency, the results per fiber-year were: beta (standard error) = 0.60 (0.13) and 0.58 (0.08) for threshold and nonthreshold models. Using a survival (Cox) model, the corresponding estimate is 0.11 (0.04). All estimates are statistically significant (p < 0.05). Excludingfirst 10 years o f follow-up: Cumulative Exposure n 0.0-11.6 fibers/cc-yrs 5 11.7-25.1 fibers/cc-yrs 9 25.2-113.7 fibers/cc-yrs 10 >113.8 fibers/cc-yrs ' 163 per 100 fibers/cc-yrs increase RR (95% CI)d 1.0 (referent) 1.7(0.58,5.2) 1.9(0.63,5.5) 3.2 (1.2,8.8) 0.36 (0.03,1.2) (p = 0.02) Similar patterns were reported for analyses o f intensity and residence-weighted exposure, but results not presented in paper. DRAFT--DO NOT CITE OR QUOTE This document is a draftfo r re\'iew purposes only and does not constitute Agency policy. Table 4-4. R espiratory (lung) cancer m ortality and exposure-response analyses based on studies of the vermiculite mine workers in Libby, M Ta (continued) Reference(s) Sullivan (2007) Larson et al. (2010b) Inclusion criteria and design details White men, enumerated in 1982, alive in 1960 or hired after 1960, worked at least 1 day, follow-up 1960-2001 (;? = 1,672); 767 deaths (95% with known cause o f death). Mean duration: 4.0 years (808, ~50% worked less than 1 year). Median fibers/cc-years: 8.7. Underlying cause o f death data from death certificates or National Death Index-Plus. Standardized mortality ratio (SMR) (95% Cl) 15 ye a r exposure lag: All cancer (n = 202) SMR: 1.4 (1.2, 1.6) Lung (;? = 89) SMR: 1.7 (1.4, 2.1) Inclusion criteria not described (;? - 1,862); follow-up through 2006; 952 deaths (80% with known cause of death). Median duration: 0.8 years; Median fibers/cc-yr = 4.3. Immediate and underlying cause of death data (i.e., multiple cause o f death) from death certificates or National Death Index-Plus. Lung (;? = 104) SMR: 1.6 (1.3, 2.0) Exposure-response analyses-- lung cancer 15 year exposure lag: Cumulative Exposure n SMR (95% CI)b SRR (95% CIV 0.0-4.49 fibers/cc-yrs 19 1.5 (0.9, 2.3) 1.0 (referent) 4.5-22.9 fibers/cc-yrs 24 1.6 (1.1, 2.5) 1.1 (0.6, 2.0) 23.0-99.0 fibers/cc-yrs 23 1.8 (1.1, 2.7) 1.4 (0.7, 2.7) >100 fibers/cc-yrs 23 1.9 (1.2, 2.9) 1.5 (0.8, 2.8) linear trend test (/?< 0.001) Duration n <1 year 41 1-9.9 years 34 >10 years 14 20 year exposure lag: Cumulative Exposure n 0.0-<1.4 fibers/cc-yrs 19 1.4 to <8.6 fibers/cc-yrs 20 8.6 to <44.0 fibers/cc-yrs 21 >44.0 fibers/cc-yrs 38 Per 100 fibers/cc-yrs increase SMR (95% CI)b 1.6 (1.1, 2.1) 1.7 (1.1, 2.3) 2.5 (1.4, 4.3) SRR (95% CIV 1.0 (referent) 1.1 (0.7, 1.8) 1.8 (0.9, 3.4) SMR (95% CI)bRR (95% CI V (not reported) 1.0 (referent) (not reported) 1.1 (0.6, 2.1) (not reported) 1.7 (1.0, 3.0) (not reported) 3.2 (1.8, 5.3) 1.11 (1.05, 1.18) (p = 0.006) 4-12 DRAFT--DO NOT CITE OR QUOTE Tncludes miners, millers, and processors; workers in the screening plant, loading docks, and expansion plants; and office workers. bSMR based on external referent group. Tn Sullivan (2007), the SRR is a ratio o f sums o f weighted rates in which the weight for each stratum-specific rate is the combined person-years for the observ ed cohort across all duration (or cumulative level o f exposure) categories. The Life-Table Analysis System provides the SRR for each duration (or cumulative level o f exposure) group compared to the referent group. The cutoff points for the categories are specified by the user. Taylor-series-based confidence interv als are given for each specific SRR. dIn McDonald et a t (2004). the RR is based on Poisson analysis using an internal referent group. Tn Larson et a t (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group. SMR = standardized mortality ratio. Cl = confidence interval, SRR = standardized rate ratio. RR = relative risk. 1 A more recent analysis of the Libby, MT workers expanded the cohort to include all 2 workers, regardless of duration of employment ('Sullivan. 2007). The total sample 3 (n = 1,672 white men) included 808 workers who had worked for less than 1 year. These 4 short-term workers had been excluded from the previous studies in Table 4-4. Analyses 5 presented in the report were based on follow-up from 1960-2001. This beginning point was 6 chosen because comparison rates for asbestosis, an outcome of interest, were not available before 7 1960 in the NIOSH Life-Table Analysis System, the analytic software used in the analysis 8 (Sullivan. 2007). Few deaths had occurred before 1960 (95 men dead or lost to follow-up before 9 1960 were excluded), so this exclusion criterion would not be expected to result in a substantial 10 loss of outcomes. Because mesothelioma was not coded separately until 1999, the mesothelioma 11 risk analysis is based on data from 1999-2001. 12 In the study by Sullivan (2007). comparison rates for standardized mortality ratio (SMR) 13 analyses were calculated from U.S. population cause-specific mortality data (limited to white 14 males) and adjusted for age and calendar year of follow-up (using 5-year groups). McDonald 15 et al. (2004) also used comparison rates from the U.S. population and included additional 16 analyses for the category of respiratory cancers using Montana population rates. 17 Larson et al. (2010b) reconstructed a worker cohort based on company records and 18 analyzed mortality risks through 2006. This study included 1862 workers; inclusion and 19 exclusion criteria are not stated, and, thus, it is not clear whether this analysis excluded females 20 or specific ethnic groups. The exposure assessment methodology was based on the methods 21 described by Amandus et al. (1987a)--without the modification used by Sullivan (2007). 22 Multiple causes of death (i.e., from any mention on the death certificate) were used, rather than 23 underlying cause of death. Because multiple causes of death are used, more than one cause of 24 death can be coded for an individual. 25 The studies of the Libby worker cohort by Amandus and Wheeler (1987). Sullivan 26 (2007). and Larson et al. (2010b) defined lung-cancer mortality based on more specific causes of 27 death codes compared to the broader classification of "all respiratory cancer" used by McDonald 28 et al. (2004; 1986a). For example, the International Classification of Diseases (ICD) codes used 29 for deaths due to cancers of the trachea, bronchus, and lung occurring during the applicable years 30 in the NIOSH cohort in Sullivan (2007) were ICD-7 162.0-162.1, 162.8, 163,1CD-8 162, and 31 ICD-9 162. In the first McDonald et al. (1986a) analysis, ICD-8 codes 160-163 for respiratory This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-13 DRAFT--DO NOT CITE OR QUOTE 1 cancer were used, which also included cancer of the larynx (ICD-8 code 161) and some types of 2 "other" respiratory cancers (ICD-8 code 160). The updated follow-up for 1999 included ICD-9 3 codes 160-165 for respiratory cancer, adding the "other" respiratory cancer group (ICD-9 codes 4 164 and 165). In the national Surveillance, Epidemiology, and End Results (SEER) cancer data 5 from 2003-2007, the age-adjusted mortality rate for cancer of the larynx was 1.2, compared to 6 52.5 per 100,000 person-years for lung and bronchial cancer (NCI. 2 0 1 1 ). Thus, these additional 7 categories (larynx and "other" respiratory cancers) represent a relatively small proportion of 8 respiratory cancers, but they could be a source of some misclassification of the outcome if these 9 other cancers are not related to asbestos exposure. 10 The classification of mesothelioma was more difficult because of the lack of a unique 11 ICD code for mesothelioma prior to the 10th revision, implemented in the United States in 1999. 12 The updated NIOSH study by Sullivan (2007) identified 15 deaths for which mesothelioma was 13 mentioned on the death certificate. Only two of these deaths occurred between 1999 and 2001; 14 these were coded using the ICD-10 mesothelioma coding (C45). Larson et al. (2010b) classified 15 all death certificates listing mesothelioma as ICD-10 code C45. The updated McGill study 16 (McDonald et al.. 2004) (with analysis through 1998) noted that the classification of 17 mesothelioma was based on a nosologist's review of death certificates; only 5 of the 12 cases 18 classified as mesothelioma had a cause of death listed as pleural cancer (ICD-9 code 163). 19 20 4.1.1.3. Cancer Mortality Risk 21 4.1.1.3.1. Lung cancer 22 The results within and among the papers in these two sets of studies (Larson et al.. 2010b; 23 Sullivan. 2007; McDonald et al... 2004; Amandins and Wheeler. 1987; McDonald et al.. 1986a) 24 show similar effects in terms of the increased risk seen for lung (or respiratory) cancer (see 25 Table 4-4). Exposure-response analyses from these studies demonstrated increasing mortality 26 with increasing exposure, using categorical and continuous measures of exposure, different lag 27 peri ods, and different exposure metrics. Because of the congruence in results and overl apping of 28 study participants among these studies, the most recent studies are discussed in detail below. 29 The analysis of McDonald et al. (2004) is limited to 406 male workers who were hired 30 before 1963 and who were employed for at least 1 year. The mean duration of work was 31 8.7 years. Cause of death data were obtained from the National Death Index for deaths from This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-14 DRAFT--DO NOT CITE OR QUOTE 1 1983 to 1998 and were based on ICD-8 coding by a nosologist using death certificates obtained 2 for deaths before 1983. Expected rates were based on age-, race- and sex- specific rates. A total 3 of 44 deaths due to respiratory cancers were observed, for an SMR = 2.4 (95% confidence 4 interval [Cl]: 1.7, 3.2). A pattern of increasing mortality with increasing cumulative exposure 5 was seen, with relative risks (RRs) of 1.0 (referent), 1.7, 1.9, and 3.2 in categories of 0.0-11.6, 6 11.7-25, 25.2-113.7, and >113.8 fibers/cc-years, respectively (see Table 4-4). The estimated 7 linear increase in RR of respiratory cancer risk per 100 fibers/cc-years cumulative exposure was 8 0.36 (95% Cl: 0.03, 1.2) (p = 0.02). McDonald et al. (2004) reported that similar results were 9 obtained with measures of exposure intensity and measures of residence-weighted exposure, but 10 the data were not presented in the paper. 11 Sullivan (2007) included 1,672 white male workers who were alive in 1960 or hired after 12 1960. There was no minimum duration of employment required for inclusion in this analysis, 13 and approximately 50% of the cohort (n = 808) had worked less than 1 year. Mortality follow-up 14 was conducted through 2001, with 767 identified deaths. The exposure assessment protocol was 15 based on that described by Amandus et al. (1987a). with a modification to the estimated intensity 16 of exposure to laborers and to those with "unknown" jobs. Sullivan (2007) assigned 17 weighted-average exposure for all unskilled jobs in a department (if known) during a calendar 18 time period, rather than lower mill yard exposure used by Amandus et al. (1987a). The weights 19 are based on the number of workers assigned to unskilled jobs during the same calendar time 20 period. In the Sullivan (2007) follow-up, SMRs, using underlying cause-of-death data (based on 21 death certificates) obtained through the National Death Index and from individual states, and 22 expected mortality based on national age-, race-, and sex-specific rates, were calculated. Using a 23 15-year exposure lag, SMRs were increased for lung cancer (n = 89, SMR =1.7, 95% Cl: 1.4, 24 2.1) and for all cancer mortality (n = 202, SMR = 1.4, 95% Cl: 1.2, 1.6) (see Table 4-4). 25 Additionally, an internal referent group was used for analyses of risk in relati on to cumulative 26 exposure and duration. The results of these internal analyses are presented as standardized rate 27 ratios (SRR) for white men, controlling for age group. Increasing risks across categories of 28 cumulative exposure and duration were observed with both types of analyses, indicating a 29 positive exposure-response relationship. The SMR estimates for lung-cancer mortality were 1.5, 30 1.6, 1.8, and 1.9 in the 1- to 4.49-, 4.5- to 22.9-, 23.0- to 99.0-, and >100 fibers/cc-year exposure 31 categories, respectively. The SRR estimates were 1.0, 1.1., 1.4, and 1.5, respectively, across This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-15 DRAFT--DO NOT CITE OR QUOTE 1 these same exposure categories (see Table 4-4). For comparison to the earlier work by 2 McDonald et al. (1986a). an SMR was provided for all respiratory cancer in those employed at 3 least 1 year (SMR 2.0, 95% Cl: 1.5-2.5). For the full cohort employed at least 1 day, the SMR 4 for all respiratory cancer was 1.7 (95% Cl: 1.4-2.1) (Sullivan. 2007). 5 Amandus and Wheeler (1987) provide some information on the smoking history of a 6 sample of 161 male workers employed during 1975-1982 with at least 5 years of employment in 7 the Libby cohort study and compari son data based on surveys conducted in the United States 8 from 1955-1978. Among the workers, 35% were current smokers, and 49% were former 9 smokers. This smoking information was obtained from questionnaires the company 10 administered to workers after 1975. Assuming the definitions are similar to those of the national 11 surveys, however, the prevalence of current smokers is similar in the worker cohort compared to 12 the U.S. white male population data (ranging from 37.5-41.9% current smokers between 1975 13 and 1978). The only year in this range with data on former smokers in the national survey is 14 1975, and, at that time, the prevalence of former smokers in the population data was 29.2%, 15 about 20% lower than among the workers. Using an estimated RR of lung cancer of 14 among 16 smokers, Amandus and Wheeler (1987) estimated that the difference in smoking rates between 17 workers and the comparison population could have resulted in a 23% increase in the observed 18 risk ratio and commented that the increased risk observed in the lower dose range 19 (<50 fiber-years) could be the result of confounding by smoking status. 20 Smoking patterns in the U.S. population changed considerably over the period 21 corresponding to the data reported by Amandus and Wheeler (1987). In the National Health 22 Interview Surveys conducted between 1974 and 1983, the prevalence of smoking in males 23 age 20 and older decreased from 42.1 to 35.5% (HHS. 1990). In addition, the prevalence of 24 former smokers can depend on the definition used. Based on 1986 survey data, the percentage of 25 adults age 17 and older classified as former smokers varied between 14.7 and 25.8% using 26 different definitions for time since last smoked (e.g., from quitting 5 or more years ago to 27 quitting within the past 3 months) (HHS. 1990). Thus, given the lack of information pertaining 28 to the period in which smoking information was collected and the specifics of the sources that 29 were used, EPA concludes there is considerable uncertainty regarding the evidence for 30 differences in smoking rates between the workers and the external comparison population. This document is a draftfor review purposes only and does not Constitute Agency policy. 4-16 DRAFT--DO NOT CITE OR QUOTE 1 Larson et al. (2010b) evaluated multiple causes of death, and, therefore, more than one 2 cause of death can be coded for an individual. A total of 104 lung or bronchus cancer deaths 3 were observed, for an SMR of 1.6 (95% Cl: 1.3, 2.0) using an external comparison of United 4 States cause of death data from 1960 to 2002 (Larson et al,, 2010b). A higher risk was seen in 5 the higher cumulative exposure categories using Cox proportional hazards modeling with an 6 internal referent group: relative risk 1.0 (referent), 1.1 (95% CI: 0.6, 2.1), 1.7 (95% Cl: 1.0, 3.0), 7 and 3.2 (95% CI: 1.8, 5.3) respectively, for <1.4 (referent), 1.4 to <8 .6 , 8.6 to <44.0 and >44.0 8 fibers/cc-years. Larson et al. (2010b) used data from a health screening program conducted in 9 Libby by ATSDR in 2000-2001 (described in Section 4.1.2.2) pertaining to smoking history to 10 estimate that the proportion of smokers ranged from 50% to 66% in the unexposed group 11 (defined as exposure <8.6 fibers/cc-years) and between 66% and 85% among the exposed 12 (defined as >8.6 fibers/cc-years). Larson et al. (2010b) used these estimates in a Monte Carlo 13 simulation to estimate the potential bias in lung cancer risks that could have been introduced by 14 differences in smoking patterns. The bias-adjustment factor (RRunadjusted/RRadjusted = 1.3) reduced 15 the overall RR estimate for lung cancer from 2.4 to 2.0. 16 17 4.1.1.3.2. Mesothelioma 18 Data pertaining to mesothelioma risk from the available studies are summarized in 19 Table 4-5. McDonald et al. (2004) presented dose-response modeling of mesothelioma risk 20 based on 12 cases. Using Poisson regression, the mesothelioma mortality rate across increasing 21 categories of exposure was compared to the rate in the lowest exposure category. Note that the 22 referent group was also at excess risk of dying from mesothelioma; that is, one to three cases of 23 mesothelioma were observed in the referent group, depending on the exposure index. Three 24 exposure indices were used in analysis: average intensity over the first 5 years of employment, 25 cumulative exposure, and residence-weighted cumulative exposure. Because of the requirement 26 for 5 years of employment data, 199 individuals (including three mesothelioma cases) were 27 excluded from the analysi s of average intensity. The residence-weighted cumulative exposure 28 was based on the summati on of exposure by year, weighted by years since the exposure. This 29 metric gives greater weight to exposures that occurred a longer time ago. Although evidence of 30 an excess risk of dying from mesothelioma was seen in all groups, there was little evidence of 31 increasing RR with increasing average intensity or cumulative exposure. For the This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-17 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-5. Mesothelioma mortality risk based on studies of the vermiculite 2 mine workers in Libby, MTa 3 ' Reference(s) Inclusion criteria and design details Results Amandus and Wheeler (1987) Men, hired before 1970, worked at least 1 year, follow-up through 1982 (n = 575); 161 deaths (159 with death certificates). Mean duration: 8.3 years (0 worked less than 1 year). Mean fiber-years: 200.3. Twelve female workers not included in this analysis. 2 mesothelioma deaths observed (hired in 1946, 33 years latency, exposure >300 fibers/cc-years); 1.2% o f all deaths McDonald et al. (2004); McDonald et al. (1986a) Men, hired before 1963, worked at least 1 year (;? = 406), follow-up through 1999 (McDonald et al.. 2004); 165 deaths before Julv 1983 (163 with death certificates); 120 deaths from July 1983-1998 coded by nosologists using ICD-8 classifications; cause o f death for deaths from 1983-1998 obtained from National Death Index. Mean duration: 8.7 years (0 worked less than 1 vear). Mean fiber-vrs: 144.6. 12 mesothelioma deaths observed: 4.2% o f all deaths Excluding first 10years o ffollow-up: Cumulative Exposure n RR (95% CI)b 0.0-11.6 fibers/cc-yrs 1 1.0 (referent) 11.7- 25.1 fibers/cc-yrs 4 3.7 (0.41,33.5) 2 5 .2 - 113.7 fibers/cc-yrs 3 3.4 (0.35,33.2) >113.8 fibers/cc-yrs ' 4 3.7 (0.41,33.2) per 100 fibers/cc-yrs increase 0.10 (<0, 1.81) (p > 0.20) Intensitv Category n RR (95% CI)b 0.0-11.6 fibers/cc-yrs 1 1.0 (referent) 11.7- 25.1 fibers/cc-yrs 4 3.4 (0.37,30.9) 2 5 .2 - 113.7 fibers/cc-yrs 2 2.3 (0.21,26.1) >113.8 fibers/cc-yrs ' 2 2.1 (0.19,23.9) per 100 fibers/cc-yrs increase 0.02 (<0, 1.08) (/) > 0.20) Residence-weighted n RR (95% CI)b 0.0-25.1 fibers/cc-yrs 3 1.0 (referent) 2 5 .2 - 113.7 fibers/cc-yrs 4 1.57 (0.35,7.07) >113.8 fibers/cc-yrs ' 5 1.95(0.41,8.51) per 100 fibers/cc-yrs increase 0.03 (<0, 6.4) (/) > 0.20) Sullivan (2007) White men, enumerated in 1982, alive in 1960 or hired after 1960, worked at least 1 day, follow-up 1960-2001 (n = 1,672); 767 deaths (95% with known cause o f death). Mean duration: 4.0 years (808, ~50% worked less than 1 year). Median fibers/cc-years: 8.7. Underlying cause o f death data from death certificates or National Death Index-Plus. SMR analysis limited to 1999-2001 because this is the period for which comparison data from ICD-10 are available. 15 mesothelioma deaths observed; 2% o f all deaths jV= 2 for 1999-2001: SMR: 15.1 (95% Cl: 1.8, 54.4) Pleural (n = 4) SMR: 23.3 (95% Cl: 6.3, 59.5) This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-18 DRAFT--DO NOT CITE OR QUOTE Table 4-5. M esothelioma m ortality risk based on studies of the verm iculite mine workers in Libby, M Ta (continued) Reference(s) Inclusion criteria and design details R esu lts Larson et al. (2010b) Inclusion criteria not described (;? = 1,862); follow-up through 2006; 952 deaths (80% with known cause o f death). Median duration: 0.8 years; Median fibers/cc-yr = 4.3. Immediate and underlying cause of death data (i.e.. multiple causes of death) from death certificates or National Death Index-Plus. 19 mesothelioma deaths observed SMR: 94.8 (95% CL 57, 248) 20 year exposure lag: Cumulative Exposure n <1.4 fibers/cc-yrs 1 1.4 to <8.6 fibers/cc-yrs 2 8.6 to <440 fibers/cc-yrs 5 >44.0 fibers/cc-yrs ' 11 per 100 fibers/cc-yrs increase ' RR (95% CI) 1.0 (referent) 1.9(0.31,13.6) 4.5 (0.8, 24.6) 17.1(3.7,78.1) 1.15 (1.03, 1.28) (p = 0.0134) 1 2 includes miners, millers, and processors; workers in the screening plant, loading docks, and expansion plants; and 3 office workers. 4 bIn McDonald et al. (2004), the RR is based on Poisson analysis using an internal referent group. 5 cIn Larson et al. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group. 6 7 SMR = standardized mortality ratio. Cl = confidence interval. SRR = standardized rate ratio, RR = relative risk. 8 9 10 residence-weighted cumulative exposure, an RR of 1.57 was observed among those with 11 500.1-1,826.8 fibers/cc-years exposure, and an RR of 1.95 was observed among workers with 12 higher residence-weighted cumulative exposure. Sullivan (2007) identified 15 deaths from 13 mesothelioma through a manual review of death certificates, with 14 classified as "pleural or 14 unspecified," and 1 classified as "peritoneal." Only two of these deaths occurred between 1999 15 and 2001, the period for which comparison data using the ICD-10 classification criteria were 16 available. Based on these two mesothelioma deaths, the SMR was 14.1 (95% Cl: 1.8, 54.4). 17 Larson et al. (2010b) identified 19 mesothelioma deaths (coding any mention of mesothelioma 18 on the death certificate as the ICD-10 classification of C45). Comparison data were based on 19 multiple-causes-of-death data (1960 to 2002). The SMR for mesothelioma was 94.8 (95% Cl: 20 57.0, 148.0), and an increasing risk was seen across quartiies of exposure (see Table 4-5). The 21 comparison rates for the SMR analysis are based on multiple cause of death data for the U.S. 22 population from 1960-2002; only a small portion of this period included the ICD-10 coding 23 scheme for mesothelioma. Thus, the expected rates could be underestimated, biasing the effect 24 estimates upward 25 This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-19 DRAFT--DO NOT CITE OR QUOTE 1 4.1.1.3.3. O ther cancers 2 Larson et al. (2010b) presented data on cancers other than respiratory tract and 3 mesothelioma. The category of malignant neoplasms of digestive organs and peritoneum 4 included 39 observed deaths, for an SMR of 0.8 (95% Cl: 0.6, 1.1). No risk in relation to 5 asbestos exposure was seen with a 20-year lag. The potential for underascertainment of specific 6 causes of death should be noted, however, given the 10% loss to follow-up and missing cause of 7 death data for 9% of the identified deaths. 8 9 4.1.1.3.4. Summary o f cancer mortality risk in Libby, M T vermiculite mining operation 10 workers 11 The studies conducted in the 1980s (Amandus and Wheeler. 1987: McDonald et al.. 12 1986a) as well as the extended follow-up studies published in more recent years (Larson et al., 13 2010b: Sullivan. 2007: McDonald et al.. 2004) provide evidence of an increased risk of 14 lung-cancer mortality and of mesothelioma mortality among the workers in the Libby 15 vermiculite mining and processing operations. The lung cancer analyses using an internal 16 referent group in the larger follow-up studies (Larson et al.. 2010b: Sullivan. 2007: McDonald et 17 al,, 2004) observed increasing risks with increasing cumulative exposure exposures when 18 analyzed using quartiles or as a continuous measure. Increased risks are also seen in the studies 19 reporting analyses using an external referent group (i.e., standardized mortality ratios) (Sullivan. 20 2007: Amandus and Wheeler, 1987: McDonald et al.. 1986a). 21 22 4.1.1.4. Noncancer Effects: Respiratory and Cardiovascular Disease 23 4.1.1.4.1. Asbestosis and other nonmalignant respiratory disease mortality 24 The studies described previously also reported noncancer mortality data, with a specific 25 focus on respiratory diseases (see Table 4-6). In Sullivan (2007), the SMR for asbestosis 26 (ICD-9 code 501) was 166 (based on n = 22, underlying cause of death compared to a U.S. white 27 male referent group). In Larson et al. (2010b), the SMR was 143 (95% Cl: 111, 181), based on 28 69 observed asbestosis-related deaths using multiple-causes-of-death data. Increasing 29 cumulative exposure was observed to increase the risk for asbestosis mortality in both of these 30 analyses (see Table 4-6). A two- to threefold increase was also seen for other categories of 31 nonmalignant respiratory disease in Larson et al. (2010b), with an SMR of 2.4 (95% Cl: 2.2, 2.6) This document is a draftfor review purposes only and does not Constitute Agency policy. 4-20 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-6. Nonmalignant respiratory mortality studies of the vermiculite 2 mine workers in Libby, MTa 3 ' Reference(s) Respiratory disease (SMR, 95% Cl) Dose-response analyses: Nonmalignant respiratory diseases and asbestosis Amandas and Wheeler (1987) (NIOSH) No exclusions: Nonmalignant respiratory diseases (a = 20) SMR: 2.4 (1.5, 3.8) 20 yea r latency: Nonmalignant respiratory diseases ( = 12) SMR: 2.5 (p < 0.05) N o exclusions'. Nonmalignant respiratory diseases Cumulative Exposure n SMR(95%CI)b 0.0-49 fibers/cc-yrs 8 2.2 (not reported) 50-99 fibers/cc-yrs 2 1.7 (not reported) 100-399 fibers/cc-yrs 3 1.8 (not reported) >400 fibers/cc-yrs 10 4.0 (not reported, but p < 0.01) 20 or more years since first hire diseases Cumulative Exposure n 0.0-49 fibers/cc-yrs 7 50-99 fibers/cc-yrs 2 100-399 fibers/cc-yrs 0 >400 fibers/cc-yrs 3 (latency) '. Nomnalignant respiratory SMR (95%CI)b 3.3 (not reported, but p < 0.05) 2.8 (not reported) 0 (not reported) 2.8 (not reported) McDonald et al. (2004); McDonald et al. (1986a) (McGill) Nonmalignant respiratory diseases (n 51) SMR: 3.1 (2.3,4.1) Excluding first 10 years o f follow-up'. Nonmalignant respiratory' diseases Cumulative Exposure n RR(95%CI)Q 0.0-11.6 fibers/cc-yrs 5 1.0 (referent) 11.7-25.1 fibers/cc-yrs 13 2.5 (0.88,7.2) 25.2-113.7 fibers/cc-yrs 14 2.6 (0.93,7.3) >113.8 fibers/cc-yrs ' 19 3 .1 (1 .2 ,8.4) per 100 fibers/cc-yrs - 0.38 (0.12, 0.96) (p = 0.0001) Sullivan (2007) (NIOSH) 15yea r exposure lag: Asbestosis (// = 22) SMR: 166 (104, 251) Nonmalignant respiratory diseases ( = 111) SMR: 2.4 (2.0, 2.9) Chronic obstructive pulmonary disease (;? = 53) SMR: 2.2 (1.7, 2.9) Other nonmalignant respiratory diseases (n = 19) ' SMR: 2.7 (1.6,4.2) 15year exposure lag: Asbestosis Cmnulative Exposure n 0.0-49.9 fibers/cc-yrs 3 50.0- 249.9 fibers/cc-yrs 8 >250 fibers/cc-yrs ' 11 linear trend test SM R (95% C I)b 37 (7.5,122) 213 (91.6,433) 749 (373,1,368) SR R (95% C I)C 1.0 (referent) 7.3 (1.9,28.5) 25.3 (6.6,96.3) (p < 0.01) 15 year exposure lag'. Nomnalignant respiratory diseases Cmnulative Exposure n SMR(95%CI)b SR R (95% C I)C 0.0-4.49 fibers/cc-yrs 18 1 .8 (1 4 ,2 .8 ) 1.0 (referent) 4.5-19.9 fibers/cc-yrs 24 2.0 (1.3,3.0) 1.2 (0.6,2.3) 2 0 .0 - 84.9 fibers/cc-yrs 26 2.2 (1.5,3.3) 1.5 (0.8,2.9) 85.0- 299.9 fibers/cc-yrs 20 2.6 (1.6,4.0) 14(0.7,2.7) >300 fibers/cc-yrs ' 23 4.8 (3 4 ,7 .3 ) 2.8 (1.3,5.7) This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-21 DRAFT--DO NOT CITE OR QUOTE Table 4-6. N onm alignant respiratory m ortality studies of the veriniculite mine workers in Libby, M Ta (continued) Reference(s) Respiratory disease (SMR, 95% Cl) Dose-response analyses: Nonmalignant respiratory diseases and asbestosis Larson et al. (2010b) Asbestosis ( = 69) SMR: 143 (111, 181) Nonmalignant respiratory diseases (n = 425) SMR: 2.4 (2.2,2.6) Chronic obstructive pulmonary disease (n = 152) SMR: 2.2 (1.9,2.6) Other nonmalignant respiratory (;? = 120) SMR: 2.8 (2.3 3.4) 20 year exposure lag: Asbestosis Cumulative Exposure n SMR (95% ClT RR (95% CI)e <1.4 fibers/cc-yrs 4 (not reported) 1.0 (referent) 1.4- <8.6 fibers/cc-yrs 8 (not reported) 2.8 (1.0, 7.6) 8 6 - <44.0 fibers/cc-yrs 25 (not reported) 8.0 (3.2, 19.5) >44.0 fibers/cc-yrs 32 (not reported) 11.8(4.9, 28.7) Per 100 fibers/cc-yrs increase 1.18(1.12, 1.23) O < 0.001) 2 0 year exposure lag'. Nonmalignant respiratory diseases Cumulative Exnosure n SMR (95% ClT RR (95% CIV3 <1.4 fibers/cc-yrs 43 (not reported) 1.0 (referent) 1.4- <8.6 fibers/cc-yrs 46 (not reported) 1.4 (0.9, 2.1) 8 6 - <44.0 fibers/cc-yrs 56 (not reported) 1.8 (1.3, 2.7) >44.0 fibers/cc-yrs 58 (not reported) 2.5 (1.7, 3.6) Per 100 fibers/cc-yrs increase 1.08 (1.03, 1.13) (p = 0.0028) 1 2 aIncludes miners, millers, and processors; workers in the screening plant, loading docks, and expansion plants; and 3 office workers. 4 bSMR based on external referent group. 5 cIn Sullivan (2007), the SRR is a ratio o f sums o f weighted rates in which the weight for each stratum-specific rate is 6 the combined person-years for the observed cohort across all duration (or cumulative level o f exposure) categories. 7 The Life-Table Analysis System provides the SRR for each duration (or cumulative level o f exposure) group 8 compared to the referent group. The cutoff points for the categories are specified by the user. Taylor-series-based 9 confidence intervals (Rothman, 1986) are given for each specific SRR. 10 dIn McDonald et al. (2004). the RR is based on Poisson analysis using internal referent group. 11 eIn Larson et al. (2010b). the RR is based on Cox proportional hazards modeling using an internal referent group. 12 ' ' ' ' ' ' 13 SMR = standardized mortality ratio. Cl = confidence interval, SRR = standardized rate ratio, RR = relative 14 risk. 15 16 17 for all nonmalignant respiratory disease, and SMR = 2.8 (95% Cl: 2.3, 3.4) for diseases other 18 than asbestosis, chronic obstructive pulmonary disease, and silicosis. These results are similar to 19 the nonmalignant respiratory disease mortality data from studies of this cohort using underlying 20 cause-of-death data. A markedly higher risk of nonmalignant respiratory disease mortality was 21 also observed in the cumulative exposure category of >300 or >400 ftbers/cc-years, respectively 22 in Sullivan (2007) and Amandus and Wheeler (1987). Larson et al. (2010b) used a Monte Carlo 23 simulation to estimate the potential bias in nonmalignant respiratory disease risk that could have 24 been introduced by differences in smoking patterns between exposed and unexposed workers in This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-22 DRAFT--DO NOT CITE OR QUOTE 1 the cohort. The bias-adjustment factor (RRunadjusted/RRadjusted = 1.2) reduced the overall RR 2 estimate for nonmalignant respiratory mortality from 2.1 to 1.8 . 3 4 4.1.1.4.2. Radiographic abnormalities 5 Respiratory disease risk is also evidenced by chest radiographs showing pleural and 6 parenchymal abnormalities in the Libby, MT worker cohorts (see Table 4-7). Two of these 7 studies were conducted in the 1980s and were based on X-rays of a subset of workers taken for 8 either an annual workplace screening (Amandus et al.. 1987b) or as part of a study examination 9 (McDonald et al.. 1986b). The subset of McDonald et al. (1986b) included 164 workers 10 currently employed at the Libby facility, 80 former employees, and 47 area residents without 11 known dust exposure. The subset selected by Amandus et al. (1987b) included workers with at 12 least 5 years tenure who had worked at Libby at some time during 1975-1982. The most recent 13 X-ray film for each worker, which NIOSH obtained from the Libby hospital that performed the 14 screening, was independently read by three qualified readers using the International Labor Office 15 (ILO) classification system. For the analysis, the classification indicating pleural abnormalities 16 by at least two of the three readers was used to determine the presence of pleural abnormalities, 17 while the median reading was used to determine the profusion category of small opacities. In the 18 McDonald et al. (1986b) study, all three readings agreed for about 90% of the chest X-rays that 19 showed evidence of pleural calcification, obliteration of the costophrenic angle, and pleural 20 thickening on the diaphragm. Similarly, all three readings agreed for about 80% of chest X-rays 21 that showed evidence of small opacities, pleural plaques, or diffuse thickening. Amandus et al. 22 (1987b) provided a more detailed breakdown of the correspondence between readers for the 23 rating of small opacities (by category). The prevalences of any opacities (category 1/0 or more) 24 were 10, 16, and 10% for Readers A, Et, and C. This difference among raters was similar to that 25 seen in other studies. Other design details are described in Table 4-7. 26 Although both research groups utilized the ILO 1980 guidelines, McDonald et al (1986b) 27 reported pleural thickening on the chest wall (both pleural plaques and diffuse) but excluding 28 other sites. Amandus et al (1987b) report "any pleural change" (both pleural plaques and 29 diffuse, defined as ".. .any unilateral or bilateral pleural change, which included pleural plaque, 30 diffuse pleural thickening of the chest wall, diaphragm or other site, but excluded costophrenic This document is a draftfor review purposes only and does not Constitute Agency policy. 4-23 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-7. Chest radiographic studies of the Libby, MT vermiculite mine 2 workers 3456789 Reference(s) Inclusion criteria and design details Results McDonald et al. (1986b) Men employed on July 1, 1983 (n = 164). Fonner male employees living within 200 miles; hired before 1963 (;? = 80), worked at least 1 year (80 participants from 110 eligible); 43 had a previous X-ray. Men without known dust exposure (n = 47); X-rays taken for other reasons (mostly employment related) at same place during study period; 24 had a previous X-ray. Data from nine women employed on July 1, 1983 not included in this report. Pleural thickening o f the chest wall observed in 15.9% o f current employees and 52.5% of past employees. Small opacities (> l/0) observed in 9.1% of current employees and 37.5% o f past employees. Both abnormalities increased with age. Age-adjusted and age-stratified (>60 years old) analyses showed increasing risk o f both abnonnalities with increasing cumulative exposure. Amandus et al. (1987b) Men, employed during 1975-1982 with at least 5 years tenure (n = 191); 184 with previous chest X-rays; 121 with smoking questionnaires. Annual radiographs taken since 1964; most recent radiograph evaluated. Mean employment duration: 14 years. Mean fiber-years: 123 (all workers), 119 (workers with radiographs). Pleural thickening o f the chest wall observed in 13%. Small opacities (> l/0) observed in 10%. Both abnonnalities increased with increasing cumulative exposure. Whitehouse (2004) n = 123 (86 fonner employees o f W.R. Grace & Co., 27 family members o f employees, and 10 Libby residents with only environmental exposures). Average age: 66 years; 80% males. Fifty-six patients had interstitial abnormalities at profusion category 0/1 or 1/0. Chest X-rays and/or HRCT scans: pulmonary function tests (FVC, TLC, and DLCO). Average yearly loss (n = 123): FVC 2.2% ' TLC 2.3% DLCO 3.0% Larson et al. (2010a) Men with 2 or more X-rays spanning a period o f 4 or more years. Most recent X-ray read independently by each of 3 NIOSH B-readers; each series o f X-rays (for a given participant) then read by the panel for a consensus determination o f time o f first appearance of the detectable abnormality (;? = 84). Latency (time from hire to observed change), median (25th. 75thpercentile) years: Localized pleural thickening 8.6(1.4,14.7) Any pleural calcification 17.5 (8.1,24.2) Diffuse pleural thickening 27.0 (10.7, 29.8) 4 5 DLCO = single breath carbon monoxide diffusing capacity; FVC = forced vital capacity; TLC = total lung capacity, 6 HRCT = high resolution computed tomography. 7 ' ' ' 8 9 This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-24 DRAFT--DO NOT CITE OR QUOTE 1 angle obliteration.. which included all sites as well as a second category of "pleural 2 thickening of the chest wall." 3 Amandus et al. (1987b) reported pleural thickening of the chest wall in 13% and small 4 opacities (>l/0) in 9.1% of current employees. Similar data were reported by McDonald et al. 5 (1986b), with 15.9 and 10% with pleural thickening of the chest wall and small opacities, 6 respectively. In both studies, prevalence of these abnormalities increased with increasing 7 cumulative exposure. McDonald et al. (1986b) also included 80 former employees in their 8 study. The prevalence of pleural thickening of the chest wall (52.5%) and small opacities 9 (37.5%) was higher in these workers compared with current workers. These groups differed by 10 age, however, with only one of the 80 former workers < age 40 years compared with 80 of 11 164 current workers. Within the age category 40 to 59 years, the prevalences of chest wall 12 pleural thickening were 20.3 and 40.0% in current and former employees, respectively, and, in 13 the >60-years age group, the prevalences were 40.0 and 61.2%, respectively. The authors 14 attribute these differences in prevalence rates in current compared with former employees to 15 differences in cumulative exposure. Among the 47 area residents without known dust exposure 16 in an occupational setting in the study by McDonald et al. (1986b). the prevalence of pleural 17 thickening was 8.5% (n = 4), and the prevalence of small opacities was 2.1% (ft = 1). 18 Both Amandus et al. (1987b) and McDonald et al. (1986b) provided categorical 19 exposure-response data as well as logistic models for various endpoints (e.g., small opacities, 20 pleural calcification, pleural thi ckening of the chest wall, and "any pleural change"). In 21 McDonald et al. (1986b). exposure and age were both predictive of pleural thickening along the 22 chest wall, and the regression coefficient for cumulative exposure (fibers-years/cc) was 23 0.0024 per unit increase in cumulative exposure for the log odds of the presence of pleural 24 thickening, adjusting for age and smoking. Exposure, age, and smoking status were all 25 predictive of small opacities, with a beta of 0.0035 per unit increase in cumulative exposure. In 26 contrast, although categorical analysis reported by Amandus et al. (1987b) indicated a positive 27 exposure response relationship for both "any pleural change" and pleural thickening along the 28 chest wall, exposure was not a significant predictor in regression analysis controlling for age 29 (regardless of smoking status). The estimated relationship between exposure and prevalence of 30 small opacities in Amandus et al. (1987b) was similar to that reported by McDonald et al. 31 (1986b). This document is a draftfor review purposes only and does not Constitute Agency policy. 4-25 DRAFT--DO NOT CITE OR QUOTE 1 Whitehouse (2004) examined changes in pulmonary function measures in 123 patients 2 seen in a pulmonary disease practice serving the Libby, MT area, with a mean follow-up time of 3 35 months. This study population included 86 former employees of W.R. Grace & Co., 4 27 family members of employees, and 10 Libby residents with only environmental (i.e., 5 nonoccupational, nonfamily-related) exposures. The average age at the time of the first 6 pulmonary study was 66 years, and 80% were male. Chest X-rays or high resolution computed 7 tomography scans revealed no evidence of interstitial changes in 67 (55%) of the 123 patients, 8 and 56 patients (45%) were found to have interstitial changes at profusion category 0/1 or 1/0. 9 Pulmonary function tests included forced vital capacity (FVC), total lung capacity (TLC), and 10 the single breath carbon monoxide diffusing capacity (DLCO). The average yearly loss was 11 2.2% for FVC, 2.3% for TLC, and 3.0% for DLCO. The subset of 94 patients who experienced a 12 loss of FVC was characterized as the group with worsening lung function. Among this group, 13 the average yearly loss was 3.2% for FVC, 2.3% for TLC, and 3.3% for DLCO. 14 Larson et al. (2010a) analyzed data from a subset of workers for whom pleural and/or 15 parenchymal abnormalities were seen on the most recently available X-ray and who had one or 16 more previous X-rays covering a span of at least 4 years available for comparison. Three 17 NIOSH B-readers independently reviewed the most recent of the available X-rays for each 18 individual in the study using ILO criteria (ILO. 2002). If pleural or parenchymal abnormalities 19 consistent with asbestos exposure were seen by each of the readers, the full series of X-rays for 20 that participant was evaluated to identify the time at which changes were first seen. For this set 21 of analyses, the readers worked as a consensus panel, examining each of the available X-rays in 22 reverse chronological order to determine the latency (i.e., length of time between first exposure, 23 as measured by date of hire and observed abnormality), and the degree of progression by type of 24 abnormality. Stored X-rays were found for 184 workers, and 84 were included in the analysis. 25 Exclusi ons were based on the following: 76 did not have at least two X-rays over the span of at 26 least 4 years, 20 declined to participate, unanimous classification of the most recent X-ray was 27 not reached for 3, and 1 worker did not have any detectable abnormality. Localized pleural 28 thickening was seen in 83 of these 84 workers who were known to have had pleural and/or 29 parenchymal abnormalities at a median latency of 8.6 years. Any pleural calcification was seen 30 in 37 workers, with a median latency of 17.5 years, and diffuse pleural thickening was seen in 31 12 workers (median latency: 27.0 years). The latency period increased with increasing profusion This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-26 DRAFT--DO NOT CITE OR QUOTE 1 categories, from a median of 18.9 years for >1/0, 33.3 years for progression to >2/1, and 2 36.9 years for progression to 3/2. 3 4 4.1.1.4.3. Cardiovascular-related mortality 5 Larson et al. (2010b) presents data on mortality due to cardiovascular diseases, with 6 SMRs of 0.9 (95% Cl: 0.9, 1.0) seen for heart disease (n = 552) and 1.4 (95% Cl: 1.2, 1.6) seen 7 for circulatory system diseases (n = 258). Deaths due to heart diseases were further categorized 8 into ischemic heart disease (n = 247) and other heart disease (n = 120, for pericarditis, 9 endocarditis, heart failure, and ill-defined descriptions and complications of heart disease), with 10 SMRs of 0.7 (95% Cl: 0.6, 0.8) and 1.5 (95% 1.2, 1.8), respectively. Circulatory diseases 11 included hypertension without heart disease (n = 42), with an SMR of 1.7 (95% Cl: 1.2, 2.4) and 12 diseases of arteries, veins, or lymphatic vessels (n = 136), SMR = 1.6 (95% Cl: 1.4, 2.0). The 13 combined category of cardiovascular-related mortality resulted in modestly increased risks 14 across quartiles of exposure, with RR of 1.0 (referent), 1.3 (95% Cl: 1.0, 1.6), 1.3 (95% Cl: 1.0, 15 1.6), and 1.5 (95% Cl: 1.1, 2.0) with exposure groups of <1.4, 1.4 to <8 .6 , 8.6 to <44.0, and 16 >44.0 fibers/cc-years, respectively. Larson et al. (2010b) used a Monte Carlo simulation to 17 estimate the potential bias in cardiovascular disease risk that could have been introduced by 18 differences in smoking patterns between exposed and unexposed workers in the cohort. The 19 bias-adjustment factor (RRUnadjusied/RRadjusted 1.1) reduced the overall RR estimate from 1.6 to 20 1.5. Because Larson et al. (2010b) analyzed multiple causes of death, the observed association 21 between exposure and cardiovascular disease-related mortality may reflect, at least in part, a 22 consequence of an underlying respiratory disease. 23 24 4.1.1.4.4. Summary o f noncancer risk in Libby, M T vermiculite mining operation workers 25 The risk of mortality related to asbestosis and other forms of nonmalignant respiratory 26 disease is elevated in the Libby vermiculite mining and processing operations, with increasing 27 risk seen with increasing exposure to Libby Amphibole asbestos fibers in studies conducted in 28 the 1980s (Amandus and Wheeler. 1987: McDonald et al.. 1986a) and in the extended follow-up 29 studies published in more recent years (Larson et al.. 2010b: Sullivan. 2007: McDonald et al.. 30 2004). The analyses using an internal referent group in the larger follow-up studies (Larson et This document is a draftfor review purposes only and does not Constitute Agency policy. 4-27 DRAFT--DO NOT CITE OR QUOTE 1 al., 2010b; Sullivan, 2007; McDonald et al., 2004)16observed increasing risks with increasing 2 cumulative exposure exposures when analyzed using tertiles or quartiles, or as a continuous 3 measure. Increased risks are also seen in the studies reporting analyses using an external referent 4 group, i.e., standardized mortality ratios (Sullivan. 2007; Amandus and Wheeler. 1987; 5 McDonald et al.. 1986a). Radiographic evidence of small opacities (evidence of parenchymal 6 damage) and pleural thickening (both discrete and diffuse) has also been shown in studies of 7 Libby workers (Larson et al.. 2010a; Whitehouse, 2004; Amandus et al., 1987b; McDonald et al.. 8 1986b). 9 10 4.1.2. Libby, MT Community Studies 11 In addition to worker exposures, the operations of the Zonolite Mountain mine are 12 believed to have resulted in both home exposures and community exposures. Potential pathways 13 of exposure (discussed below) range from release of airborne fibers into the community, 14 take-home exposure from mine workers (e.g., clothing), and recreational activities including 15 gardening and childhood play activities. Due to a potential for a broader community concern, 16 ATSDR conducted several studies and health actions responding to potential asbestos 17 contamination in the Libby, MT area. 18 19 4.1.2.1. Geographic Mortality Analysis 20 ATSDR conducted a location-specific analysis of mortality risks and a community health 21 screening for asbestos in the Libby area (see Table 4-8). The mortality analysis was based on 22 death certificate data from 1979-1998, with geocoding of current residence at time of death. The 23 six geographic areas used in the analysis were defined as the Libby city limits (1.1 square miles 24 around the downtown); the extended boundary of Libby (2.2 square miles around the 25 downtown); the boundary based on air modeling (16 square miles, based on computer modeling 26 of asbestos fiber distribution); the medical screening boundary (25 square miles, including the 27 town of Libby and areas along the Kootenai River); the Libby valley (65 square miles); and 28 central Lincoln County (314 square miles, based on a 10-mile radius around downtown Libby) 29 (ATSDR. 20001. 16See also reanalysis o f Sullivan (2007) data by Moolgavar et al. (2010). This document is a draftfo r review purposes only and does not Constitute A gency policy. 4-28 DRAFT--DO NOT CITE OR QUOTE 1 The 1990 population estimates were 2,531, 3,694, 4,300, 6,072, 8,617, and 9,512, 2 respectively, for these six areas. Age-standardized SMRs were calculated using underlying 3 cause-of-death information obtained from death certificates issued during the study period for 4 413 of 419 identified decedents, and Montana and U.S. populations were used as reference 5 groups. Increased SMRs were observed for both asbestosis and pulmonary circulation diseases 6 (see Table 4-8). The SMR for lung cancer ranged from 0.9-1.1 and 0.8-1.0 in the analyses for 7 each of the six geographic boundaries using Montana and U.S. reference rates, respectively. In 8 addition, four deaths due to mesothelioma were observed during the study period. These 9 analyses did not distinguish between deaths among workers and deaths among other community 10 members. 11 12 4.1.2.2. Community Screening-- Respiratory Health 13 The ATSDR community health screening was conducted from July-November 2000 and 14 July-September 2001 with 7,307 total participants (ATSDR. 2001b) (see Table 4-9). Eligibility 15 was based on residence, work, or other presence in Libby for at least 6 months before 1991. The 16 total population eligible for screening is not known; the population of Libby, MT in 2000 was 17 approximately 10,000. In addition to a standardized interview regarding medical history, 18 symptoms, work history, and other potential exposures, clinical tests included spirometry (forced 19 expiratory volume in one second [FEV1] and FVC) and chest X-rays (for participants aged 20 18 years and older). Moderate to severe restriction (defined by the researchers as FVC <70% 21 predicted value) was observed in 2 .2% of the men and 1.6% of women but was not observed in 22 individuals less than age 18. 23 Two board-certified radiologists (B readers) examined each radiograph, and a third reader 24 was used in cases of disagreement. Readers were aware that the radiographs were from 25 participants in the Libby, MT health screening but were not made aware of exposure histories 26 and other characteristics (Peipins et al.. 2004a; Price. 2004; Peipins et al.. 2003). The 27 radi ographs revealed pl eural abnormalities in 17.9% of parti cipants, with prevalence increasing 28 with increasing number of "exposure pathways" (defined on the basis of potenti al work and 29 residential exposure to asbestos within Libby and from other sources) (see Table 4-9). Detailed 30 results of an analysis excluding the former Libby workers cohort were not presented, but the 31 authors noted that the relationship between number of exposure pathways and increasing This document is a draftfor review purposes only and does not Constitute Agency policy. 4-29 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-8. Cancer mortality and nonmalignant respiratory disease mortality 2 in the Libby, MT community 3 Reference(s) Inclusion criteria and design details Results ATSDR (2000) 1979-1998, underlying cause o f death from death certificates; geocoding of street locations (residence at time of death) within six geographic boundaries (ranging from 2,532 residents in Libby city limits to 9,521 in central Lincoln County in 1990). Inquiries to postmaster were required because of P.O. Box address for 8% (n = 32); information on 47 o f 91 residents of elderly care facilities resulted in reclassification o f 16 o f 47 (34%) to nonresidents o f Libby. U.S. Census data corresponding to the same six geographic boundaries of Libby, MT. 419 decedents identified, 418 death certificates obtained, 413 with geocoding. Age-standardized SMRs based on Montana and U.S. comparison rates. Asbestosis SMRs were somewhat higher using the U.S. referent group. but choice o f referent group had little difference on SMRs for most diseases. Four deaths from mesothelioma observ ed in the study area. Lung cancer (n = 82) SMR (95% Cl) Comparison area (Montana reference rates): Libby city limits 1.1 (0.8, 1.5) Extended Libby boundary' 1.1 (0.8, 1.5) Air modeling 1.0 (0.8, 1.4) Medical screening 0.9 (0.7. 1.2) Libby valley 0.9 (0.7. 1.2) Central Lincoln County 0.9 (0.7, 1.1) Pancreatic cancer (n = 10) SMR (95% Cl) Comparison area (Montana reference rates): Libby city limits 1.0 (0.5, 2.1) Extended Libby boundary' 0.9 (0.4, 1.7) Air modeling 0.7 (0.3, 1.4) Medical screening 0.7 (0.3, 1.2) Libby valley 0.6 (0.3, 1.0) Central Lincoln County 0.5 (0.3, 1.0) Asbestosis (/? = 11) SMR (95% Cl) Comparison area (Montana reference rates): Libby city limits 40.8 (13.2.95.3) Extended Libby boundary' 47.3 (18.9.97.5) Air modeling 44.3 (19.1,87.2) Medical screening 40.6 (18.5. 77.1) Libby valley 38.7 (19.3,69.2) Central Lincoln County 36.3 (18.1,64.9) Comparison area (U.S. reference rates): Libby city limits 63.5 (20.5, 148) Extended Libby boundary' 74.9 (30.0, 154) Air modeling 71.0 (30.6, 140) Medical screening 66.1 (30.2, 125) Libby valley 63.7 (31.7, 114) Central Lincoln County 59.8 (29.8, 107) Pulmonary circulation (;? = 14) SMR (95% Cl) Comparison area (Montana reference rates): Libby city limits 2.3 (1.1, 4.4) Extended Libby boundary 1.9 (0.9, 3.7) Air modeling 1.8 (0.9, 3.3) Medical screening 1.6 (0.8, 2.9) Libby valley 1.6 (0.9, 2.7) Central Lincoln County 1.5 (0.8, 2.5) This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-30 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-9. Pulm onary function and chest radiographic studies in the Libby, 2 MT community 3 Reference(s) Inclusion criteria and design details Results Peipins et al. (2003): ATSDR (2001b) Resided, worked, attended school, or participated in other activities in Libby for at least 6 months before 1991 (including mine employees and contractors). Health screening between July and November 2000. Conducted interviews (n = 6,149, 60% o f Libby residents based on 2000 Census data) and chest X-rays (n = 5,590, 18 years and older), and determined spirometry-- forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC1), and ratio (FEV1/FVC). 19 "exposure pathways" including Libby mining company work, contractor work, dust exposure at other jobs. vermiculite exposure at other jobs, potential asbestos exposure at other jobs or in the military, cohabitation with Libby mining company worker, and residential and recreational use o f vermiculite. Chest X-rays read by 1980 ILO classifications (3 views; posterior-anterior, right- and left- anterior oblioue). Pcini ns et al. (2003) similar to (ATSDR. 2001b) excent longer screening period (July-November 2000 and July-Scptcmbcr 2001). Conducted interviews (n = 7.307) and chest X-rays (/? = 6,668). Peinins (2003) and ATSDR (2001b): Pleural abnormalities seen in 17.9% o f participants; increasing prevalence with increasing number o f exposure pathways (6.7% among those with no specific pathways, 34.6% among those with 12 or more pathways). ATSDR (2001b): Moderate-to-severe FVC1 restriction (FVC <70% predicted): 2.2% o f men >17 years old; 1.6% of women >17 years old; 0.0% of men or women <18 years old. Also includes data on self-reported lung diseases and symptoms. Weill et al. (2011) Participants in the ATSDR community health screening (see first row in table). Analysis limited to ages 25 to 90 years, excluding individuals with history o f other asbestosrelated work exposures, with spirometry, consensus reading o f chest X-ray, smoking data, and exposure pathway data (n = 4,397). Analysis based on five exposure categories: (1) W.R. Grace worker, (2) other vermiculite worker (contractor work), (3) other dusty occupation, (4) household (combination o f three household categories), and (5) environmental ("no" to work and household exposures in Categories 1-6). Chest X-rays read by 1980 ILO classifications (frontal view). Profusion DPT/ >1/0 Plaque CAO Prevalence (%), ages 25 to 40 years: 1) W.R. Grace 0.0 20.0 5.0 2) Other 0.8 0.8 0.0 3) Dusty 0.0 3.8 0.4 4) Household 0.0 2.2 0.0 5) Environment 0.0 0.4 0.0 Prevalence (%), ages 41 to 50 years: 1) W.R. Grace 0.0 26.2 5.0 2) Other 0.5 7.8 1.0 3) Dusty 0.0 2.8 0.9 4) Household 0.0 11.1 0.4 5) Environment 0.0 1.9 0.2 Prevalence (%), ages 51 to 60 years: 1) W.R. Grace 3.2 34.9 3.2 2) Other 0.6 13.7 0.6 3) Dusty 0.6 12.6 0.0 4) Household 1.0 20.1 1.5 5) Environment 0.0 7.7 0.9 Prevalence (%), ages 61 to 90 years: 1) W.R. Grace 11.1 45.7 8.6 2) Other 0.6 24.8 8.5 3) Dusty 1.1 21.9 3.3 4) Household 2.4 38.3 5.7 5) Environment 1.3 12.7 2.2 This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-31 DRAFT--DO NOT CITE OR QUOTE Table 4-9. Pulm onary function and chest radiographic studies in the Libby, MT community (continued) Reference(s) Inclusion criteria and design details Results Vinikoor et al. (2010) Participants in the ATSDR community health screening (see first row in table). Analysis limited to n = 1,003 ages 10-29 years at time o f health screening (<age 18 in 1990 when the mining/milling operations closed). Excluded if worked for W.R. Grace, or for a contractor o f W.R. Grace, exposed to dust at other jobs, or exposed to vermiculite at other jobs. Exposure characterized by 6 activities (never, sometimes, or frequently participated in 1-2 or >3 activities). Analysis o f history o f respiratory symptoms and spirometry data (obstructive, restrictive, or mixed). Little difference across exposure levels in prevalence of physician-diagnosed lung disease or abnormal spirometry. Odds Ratio (95% Cl) seen between >3 activities and Usual cough 2.93 (0.93, 9.25) Shortness o f breath 1.32 (0.51, 3.42) Bloody phlegm 1.49 (0.41, 5.43) 1 2 OR = odds ratio; DPT = diffuse pleural thickening; CAO = costophrenic angle obliteration. 3 4 5 prevalence of pleural abnormalities was somewhat attenuated with this exclusion. The 6 prevalence of pleural anomalies decreased from approximately 35% to 30% in individuals with 7 12 or more exposure pathways when these workers were excluded from the analysis. Among 8 indivi duals with no definable exposure pathways, the prevalence of pleural anomalies was 6.7%, 9 which is higher than reported in other population studies (Peipins et ah, 2004a; Price. 2004). The 10 direct comparability between study estimates is difficult to make; the possibility of over- or 11 underascertainment of findings from the X-rays based on knowledge of conditions in Libby was 12 not assessed in this study. No information is provided regarding analyses excluding all potential 13 work-related asbestos exposures. 14 Weill et al. (2011) used the ATSDR community health screening data to analyze the 15 prevalence of X-ray abnormalities in relation to age, smoking history, and types of exposures. 16 From the 6,668 participants with chest X-rays, 1,327 individuals with a history of 17 asbestos-related work (other than with the Grace mining or related vermiculite operations) were 18 excluded, along with 817 excluded based on age (<25 or >90 years) or lack of spirometric data, 19 smoking data, or exposure pathway data. An additional 127 were excluded because a consensus 20 agreement (2 out of 3 readers) was not reached regarding the X-ray findings, leaving n = 4,397 in 21 the analysis. Analysis was based on five exposure categories: (1) Grace worker (n = 255), 22 (2) other vermiculite worker (e.g., secondary contractor worker for Grace or other jobs with 23 vermiculite exposure (n = 664), (3) other dusty occupation (e.g., plumber, dry wall finisher, This document is a draftfor review purposes only and does not constitute Agency policy. 4-32 DRAFT--DO NOT CITE OR QUOTE 1 carpenter, roofer, electrician, welder, shipyard work or ship construction or repair (// = 831), 2 (4) household, including household with other vermiculite or dusty work (lived with a Grace 3 worker combination of three household categories) (n = 880), and (5) environmental ("no" to 4 work and household exposures in Categories 1-4) (n = 1,894). The frontal views (posterior- 5 anterior) of the chest X-rays were used in this analysis [in contrast to the use of frontal and 6 oblique views in Peipins et al. (2003)1. As expected, lung function (FEVi, FVC, and FEVi/FVC) 7 was lower among ever smokers compared with never smokers (within each age group) and 8 decreased with age (within each smoking category). The prevalence of X-ray abnormalities 9 (plaques, or diffuse pleural thickening, and/or costophrenic angle obliteration) also generally 10 increased with age (divided into 25-40, 41-50, 51-60, and 61-90 years) within each of the 11 exposure categories (see Table 4-9), with the highest prevalence seen among Grace workers. For 12 a given age, the prevalence among those with environmental exposure only (i.e., no household or 13 occupational exposures) was similar to the prevalence among those with non-Grace occupational 14 or household exposures in the next youngest age category. The prevalence among the household 15 contact category was similar or higher than the prevalence among the other vermiculite and dusty 16 job categories. This household contact category includes individuals who lived with a Grace 17 worker with no personal history of vermiculite or dust work (n = 594) and those who also had a 18 history of other vermiculite (n = 114) or dusty (n = 172) jobs. The authors noted the prevalence 19 rates were similar among these groups, and so the analysis was based on the combination of 20 these three groups. Mean FVCs (SE) percentage predicted were 78.76 (3.64), 82.16 (3.34), 21 95.63 (0.76), and 103.15 (0.25), respectively, in those with diffuse pleural thickening and/or 22 costophrenic angle obliteration, profusion >1/0 , other pleural abnormalities, and no pleural 23 abnormalties. The strongest effects of diffuse pleural thickening and/or costophrenic angle 24 obliteration on FVC were seen among men who had never smoked (-23.77, p < 0.05), with 25 smaller effects seen among men who had smoked (-9.77, p < 0.05) and women who had smoked 26 (-6.73,p < 0.05). 27 Vinikoor et al. (2010) used the 2000-2001 health screening data to examine respiratory 28 symptoms and spirometry results among 1,224 adolescents and young adults who were 18 years 29 or younger in 1990 when the mining/milling operations closed. At the time of the health 30 screening, the ages in this group ranged from 10 to 29 years. Exclusion criteria for this analysis 31 included previous work for W.R. Grace, work for a contractor of W.R. Grace, exposure to dust at This docum ent is a draftfo r review purposes only and does not constitute A gency policy. 4-33 DRAFT--DO NOT CITE OR QUOTE 1 other jobs, or exposure to vermiculite at other jobs. The total number of exclusions was 221, 2 leaving 1,003 in the analysis. The potential for vermiculite exposure was classified based on 3 responses to questions about six activities (handling vermiculite insulation, participation in 4 recreational activities along the vermiculite-contaminated gravel road leading to the mine, 5 playing at the ball fields near the expansion plant, playing in or around the vermiculite piles, 6 heating the vermiculite to "pop" it, and other activities involving vermiculite). The medical 7 history questionnaire included information on three respiratory symptoms: usually have a cough 8 (n = 108, 10.8%); troubled by shortness of breath when walking up a slight hill or when hurrying 9 on level ground (n = 145, 14.5%); coughed up phlegm that was bloody in the past year 10 (n = 59, 5.9%). A question on history of physician-diagnosed lung disease (// = 51, 5.1%) was 11 also included. The spirometry results were classified as normal in 896 (90.5%), obstructive in 12 62 (6.3%), restrictive in 30 (3.0%), and mixed in 2 (0.2%). Information on smoking history was 13 also collected in the questionnaire: 15.8% and 7.3% were classified as current and former 14 smokers, respectively. Approximately half of the participants lived with someone who smoked. 15 The analyses adjusted for age, sex, personal smoking history, and living with a smoker. For 16 usually having a cough, the odds ratios (ORs) were 1.0 (referent), 1.88 (95% Cl: 0.71, 5.00), 17 2.00 (95% Cl: 0.76, 5.28) and 2.93 (95% Cl: 0.93, 9.25) for never, sometimes, frequently 18 participated in 1-2 activities, and frequently participated in >3 activities, respectively. For 19 shortness of breath, the corresponding ORs across those exposure categories were 1.0 (referent), 20 1.16 (95% Cl: 0.55, 2.44), 1.27 (95% Cl: 0.61, 2.63) and 1.32 (95% Cl: 0.51, 3.42), and for 21 presence of bloody phlegm in the past year the ORs were 1.0 (referent), 0.85 (95% Cl: 0.31, 22 2.38), 1.09 (0.41, 2.98), and 1.49 (95% Cl: 0.41, 5.43). For history of physician-diagnosed lung 23 disease and abnormal spirometry results, there was little difference in the odds ratios across the 24 exposure categories: for lung disease, the ORs were 1.0 (referent), 1.95 (95% Cl: 0.57, 6.71), 25 1.51 (95% Cl: 0.43, 5.24) and 1.72 (95% Cl: 0.36, 8.32) for the categories of never, sometimes, 26 frequently participated in 1-2 activities, and frequently participated in >3 activities, respectively. 27 For abnormal spirometry (i.e., obstructive, restrictive, or mixed, n = 94 cases), the ORs were 28 1.0 (referent), 1.34 (95% Cl: 0.60, 2.96), 1.20 (95% Cl: 0.53, 2.70) and 1.33 (95% Cl: 0.42, 29 4.19) across these exposure groups. 30 Two other studies examining autoimmune disease and autoantibodies in residents of 31 Libby, Montana are described in Section 4.3. This document is a draftfor review purposes only and does not constitute Agency policy. 4-34 DRAFT--DO NOT CITE OR QUOTE 1 4.1.2.3. Other Reports o f Asbestos-Related Disease Among Libby, M T Residents 2 Whitehouse et al. (2008) recently reviewed 11 cases of mesothelioma diagnosed between 3 1993 and 2006 in residents in or around Libby, MT {n = 9) and in family members of workers in 4 the mining operations (n = 2). Three cases were men who might have had occupational asbestos 5 exposure through construction work (Case 1), working in the U.S. Coast Guard and as a 6 carpenter (Case 5), or through railroad work involving sealing railcars in Libby (Case 7). One 7 case was a woman whose father had worked at the mine for 2 years; although the family lived 8 100 miles east of Libby, her exposure may have come through her work doing the family 9 laundry, which included laundering her father's work clothes. The other seven cases 10 (four women, three men) had lived or worked in Libby for 6-54 years, and had no known 11 occupational or family-related exposure to asbestos. Medical records were obtained for all 12 11 patients; pathology reports were obtained for 10 of the 11 patients. The Centers for Disease 13 Control estimated the death rate from mesothelioma, using 1999 to 2005 data, as approximately 14 14 per million per year (CPC, 2009). approximately five times higher than the rate estimated by 15 Whitehouse et al. (2008) for the Libby area population based on the estimated population of 16 9,500 for Lincoln County and 15 years (or 150,000 person-years) covered by the analysis. 17 Whitehouse et al. (2008) stated that a W.R. Grace unpublished report of measures taken in 1975 18 indicated that exposure levels of 1.1 fibers/cc were found in Libby, and 1.5 fibers/cc were found 19 near the mill and railroad facilities. Because the mining and milling operations continued to 20 1990, and because of the expected latency period for mesothelioma, Whitehouse et al. (2008) 21 suggests that additional cases can be expected to occur within this population. 22 23 4.1.2.4. Summary o f Respiratory Health Effects in Libby, M T Community Studies 24 The geographic-based mortality analysis of 1997-1998 mortality data indicates that 25 asbestosis-related mortality is substantially increased in Libby, MT, and the surrounding area, 26 with rates 40 times higher compared with Montana rates and 60-70 times higher compared with 27 U.S. rates (ATSDR. 2000). These data provide evidence of the disease burden within the 28 community; however, because this analysis did not distinguish between deaths among workers 29 and deaths among other community members, it is not possible based on these data to estimate 30 the risk of asbestos-related mortality experienced by residents who were not employed at the 31 mining or milling operations. The community health screening studies provide more detailed This docum ent is a draftfo r review purposes only and does not Constitute A gency policy. 4-35 DRAFT--DO NOT CITE OR QUOTE 1 information regarding exposure pathways in addition to occupation (ATSDR, 2001b). Data from 2 the ATSDR community health screening study indicate that the prevalence of pleural 3 abnormalities, identified by radiographic examination, increases substantially with increasing 4 number of exposure pathways (Peipins et al., 2003). In addition, the prevalence of some 5 self-reported respiratory symptoms among 10 to 29-year-old adolescents and young adults was 6 associated with certain exposure pathways. These participants were < age 18 in 1990 when the 7 mining/milling operati ons closed (Vinikoor et al.. 2010). A better understanding of the 8 community health effects and the examination of the potential progression of adverse health 9 effect in this community would benefit from additional research to establish the clinical 10 significance of these findings. The observation by Whitehouse et al. (2008) of cases of 11 mesothelioma among individuals with no direct occupational exposure to the mining and milling 12 operations indicates the need for continued surveillance for this rare cancer. 13 14 4.1.3. Marysville, OH Vermiculite Processing Plant W orker Studies 15 Libby vermi culite was used in the production of numerous commerci al products, 16 including as a potting soil amender and a carrier for pesticides and herbicides. A Marysville, OH 17 plant that used Libby vermiculite in the production of fertilizer beginning around 1960 to 1980 is 18 the location of the two related studies described in this section . 19 The processing facility had eight main departments, employing approximately 20 530 workers, with 232 employed in production and packaging of the fertilizer and 99 in 21 maintenance; other divisions included research, the front office, and the polyform plant (Lockev. 22 1985). Six departments were located at the main facility (trionizing, packaging, warehouse, 23 plant maintenance, central maintenance, and front offices). Research and development and a 24 polyform fertilizer plant were located separately, approximately one-quarter mile from the main 25 facility. In the trionizing section of the plant, the vermiculite ore was received by rail or truck, 26 unloaded into a hopper, and transported to the expansion furnaces. After expansion, the 27 vermiculite was blended with other materials (e.g., urea, potash, herbicides), packaged, and 28 stored. Changes to the expander type and dust-control measures began in 1967, with substantial 29 improvement in dust control occurring throughout the 1970s. 30 Information about exposure assessment at the Marysville, OH plant is summarized in the 31 final row of Table 4-1. Industrial hygiene monitoring at the plant began in 1972. Lockey et al This document is a draftfo r review purposes only and does not Constitute Agency policy. 4-36 DRAFT--DO NOT CITE OR QUOTE 1 (1984) noted that the limited availability of data that would allow for extrapolation of exposures 2 for earlier time periods possibly resulted in the underestimation of exposures before 1974.17 3 Task-level air samples were conducted, and measurements were determined using scanning 4 electron microscopy and transmission electron microscopy (based on particles >5-um-long, 5 <3-pm-diameter, and >3:1 aspect ratio). 6 Based on measurements and knowledge of plant operations, three categories of exposure 7 levels were defined. Group I was considered to be the nonexposed group and consisted of the 8 chemical processing, research, and front office workers. The chemical process plant was about a 9 quarter mile from the main vermiculite facility, but the same chemicals were used in both 10 locations. The 8-hour time-weighted average vermiculite exposure in this group, both before and 11 after 1974, was estimated as 0.049 fiber/cc (based on a single stationary sample taken outside the 12 main facility), which was characterized as similar to the background levels in the community. 13 Group II was the "low exposure" category and included central maintenance, packing, and 14 warehouse workers. The 8-hour time-weighted average vermiculite exposures in this group were 15 estimated as approximately 0.1-0.4 fibers/cc before 1974 and 0.03-0.13 fibers/cc in and after 16 1974. Group III was the "highest exposure" category, and included vermiculite expanders, plant 17 maintenance, and pilot plant workers. The 8 -hour time-weighted average vermiculite exposures 18 in this group were approximately 1.2 1.5 fibers/cc before 1974 and 0.2-0.375 fibers/cc in and 19 after 1974. Cumulative fiber exposure indexes, expressed as fibers-year/cc, were derived for 20 each worker from available industrial hygiene data and individual work histories. Those with 21 less than 1 fiber/cc-year were assumed to be equivalent to a community population (in terms of 22 exposure) and were used as the comparison group. The estimated cumulative exposure for the 23 work force, including Group I workers, ranged from 0.01 to 28.1 fibers/cc-years using an 8-hour 24 workday and an assumed 365 days of exposure per year.18 Exposure was assumed to occur from 25 1957 to 1980 in this study. Exposure after work hours was assumed to be zero. 26 The first study of pulmonary effects in the Ohio plant workers was conducted in 1980 27 and involved 512 workers (97% of the 530 workers previously identified with past vermiculite 28 exposure) (Lockev et ah. 1984) (see Table 4-10). Physical examination (for detection of ' Subsequent exposure assessment efforts by this team of investigators are described in Appendix F. 18Lockey et al. (1984) reported the maximum value for this group as 39.9 fibers/cc-years. but this estimate was later corrected to exclude work from 1947 to 1956, prior to the use of vermiculite at the plant. Information provi ded in personal communication from J. Lockey to Robert Benson, U.S. EPA, June 7, 2011. This docum ent is a draftfo r review purposes only and does not Constitute A gency policy. 4-37 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-10. Pulmonary function and chest radiographic studies of the 2 Marysville, OH veriniculite processing plant workers 3 ' ' ' Reference(s) Inclusion criteria and design details Results Lockey et al. (1984) ; Lockev (1985) a 1980,77 = 512 (from 530 identified employees with past veriniculite exposure; nonparticipants included 9 refusals and 9 unavailable due to illness or vacation). Smoking history, work history at the plant, and other asbestos and fiber mineral work history data were collected. Chest exam (rales), nail clubbing, spirometry, forced vital capacity, forced expiratory volume, single-breath carbon monoxide diffusing capacity, and chest X-rays (available for 502 participants) were analyzed. Mean employment duration: 10.2 years1' Three exposure groups, based on jobs and area: Mean cumulative exposure456789102314567 Group I 0.45 fibers/cc-years Group II 1.13 fibers/cc-years Group III 6.16 fibers/cc-years Cumulative fiber exposure related to history of pleuritic chest pain and shortness of breath. No relation between cumulative exposure and forced vital capacity, forced expiratory volume, or diffusing capacity. Pleural thickening in 10 workers (2%); bilateral, small opacities in 1 (0.2%). Abnormality (combined outcomes) increased with increasing cumulative exposure. Rohs et al. (2008) 2002-2005, interviews and chest X-rays conducted, 77 = 298; 280 with interviews and readable chest X-rays (from 431 workers in the 1980 study group, of which. 513 were alive in 2004; 151 living nonparticipants included 49 refusals, 76 located but did not respond, 8 not located but presumed alive, and 18 missing either X-ray or interview). Age, smoking, asbestos exposure measure (at this plant), and other asbestos exposure data used to compare participants and nonparticipants. Libby, MT veriniculite ore used in the plant from 1963-1980. Pleural abnonnalities in 80 workers (28.7%). Small opacities (>l/0) in 8 workers (2.9%). Increasing risk of pleural abnormalities with increasing cumulative fiber exposure: odds ratios (adjusting for date of hire, body mass index) by exposure quartile were 1.0 (referent), 2.7, 3.5, and 6.9. 4 5 aLockey et aL (1984) is the published paper based on the unpublished thesis (Lockcv. 1985). 6 C alculated based on stratified data presented in Table 2 of Lockey et at. (1984). 7 cRolis et al. (2008) identified one additional eligible worker from the original 512 employees identified in Lockey 8 et al. (1984). 9 10 11 pulmonary rales and nail clubbing), spirometry, and chest-X-rays were performed, and 12 information pertaining to smoking history, work history at the plant, and other relevant work 13 exposures was collected using a trained interviewer. Radiographs were read independently by 14 two board-certified radiologists (B-readers), with a reading by a third reader when the initial 15 two readings did not agree. The number of workers within each exposure group was 112, 206, 16 and 194 in Groups I, II, and III, respectively. Approximately 44% were current smokers, 17 20% former smokers, and 35% lifetime nonsmokers, but smoking history (i.e., smoking status, This docum ent is a draftfo r review purposes only and does not constitute A gency policy. 4-38 DRAFT--DO NOT CITE OR QUOTE 1 pack-years) did not differ by exposure group. Mean cumulative fiber estimates were 0.45, 1.13, 2 and 6.16 fibers/cc-years in Groups I, II, and III, respectively. An increased risk of costophrenic 3 angle blunting (n = 11), pleural, and parenchymal abnormalities (// = 11), or any of these 4 outcomes (n = 22) was observed in Group III compared with Group 1; the prevalence of any 5 radiographic change was 2.8% in Group I, 3.9% in Group II, and 5.8% in Group III. Using the 6 cumulative fiber metric, the prevalence of any radiographic change was 2.4% in the 7 <1 fiber/cc-year, 5.0% in 1-10 fibers/cc-year, and 12.5% in the >10 fibers/cc-year groups. 8 A follow-up study of this cohort was conducted in 2002-2005 (Rohs et aU 2008) (see 9 Table 4-10). This study included 298 workers, of which 280 completed the study interview and 10 chest X-ray. Details of the reasons for nonparticipation rates are described in Table 4-10. The 11 evaluation of each worker included an interview to determine work and health history, 12 spirometry, pulmonary examination, and chest X-ray. The study interview included information 13 about smoking history and asbestos exposure at the Marysville, Ohio plant and other worksites. 14 Exposure was estimated using the procedure previously described using the data on fiber levels 15 (Lockev et ah. 1984). Exposure was assumed to occur from 1963 to 1980 in this study, 16 assuming an 8 -hour workday and 365 days of exposure per year (J. Lockey, University of 17 Cincinnati, personal communication to R. Benson, U.S. EPA, July, 2007). Each worker supplied 18 a detailed work history (start and end date for each area within the facility). The exposure 19 reconstruction resulted in a cumulative exposure estimate for each individual. The estimated 20 cumulative exposure for this follow-up study ranged from 0.01 to 19.03 fibers/cc-years 21 (mean = 2.48). The time from first exposure ranged from 23 to 47 years. Twenty-eight workers 22 reported previous occupational exposure to asbestos. Exposure outside of work was assumed to 23 be zero. 24 Three board-certified radiologists independently classified the radiographs using the ILO 25 classification system (ILO. 2002). Radiologists were blinded to all identifiers. Pleural 26 thickening (all sites) was reported as either localized pleural thickening or diffuse pleural 27 thickening. Diffuse pleural thickening of the chest wall may be reported as in-profile or face-on, 28 and is recorded on the lateral chest wall "only in the presence of and in continuity with, an 29 obliterated costophrenic angle" (ILO. 2002). Localized pleural thickening may also be viewed 30 in-profile or face-on and was described by Rohs et ah (2008) as m...(pleural) thickening with or 31 without calcification, excluding solitary costophrenic angle blunting" consistent with current This docum ent is a draftfo r review purposes only and does not Constitute A gency policy. 4-39 DRAFT--DO NOT CITE OR QUOTE 1 ILO classification. Interstitial abnormalities were considered present if the reader identified 2 irregular opacities of profusion 1/0 or greater (ILO. 20021. For the analysis, a chest X-ray was 3 defined as positive for pleural abnormality and/or interstitial abnormality when the median 4 classification from the three readings was consistent with such effects. Radiographs classified as 5 unreadable were not used. Radiographic abnormalities found in the study population are 6 summarized in Tables 4-11 and 4-12. 7 8 9 Table 4-11. Prevalence of pleural radiographic abnormalities according to 10 quartiles of cumulative fiber exposure in 280 participants 111234567890 Exposure quartile Exposure, fiber-yr/cc, and (mean) Number of workers Number of workers with pleural thickening (%)b Crude OR (95% Cl) Age-adjusted OR (95% Cl) RMIadjusted OR (95% Cl) Number of workers with small opacities (%) First 0.01-0.28 70 (0.12) 5(7.1) 1.0 1.0 1.0 0(0) (referent) (referent) (referent) Second 0.29-0.85 IT (0.56) 17 (24.6) 4.0 3.2 4.9 0(0) (1.4-11.6) (1.0-9.7) (1.3-18.2) Third 0.86-2.20 68a 20G(29.4) 5.4 4.0 7.6 1(1.5) (1.33) (1.9-15.5) (1.3-12.8) (2.1-27.5) Fourth 2.21-19.03 70 (7.93) 38 (54.3) 15.4 (5.6-43) 10.0 (3.1-32) 17.0 (4.8-60.4) 7(10) Total (2.48) 280 80 (28.6) 8 (2.9) 12 13 aTwo observations in the second quartile and two in the third quartile had exact exposure values at the 50thpercentile 14 cutoff point. Rounding put these four observations in the second quartile. 15 bSignificant trend, p < 0.001. 16 cTypographical error in publication corrected. 17 ~ ' 18 The 80 workers with pleural thickening include 68 with localized pleural thickening (85%) and 12 with diffuse 19 pleural thickening (15%). 20 ' 21 Source: Rohs et ah (2008). Table 3 and Figure 2; mean exposure levels and number o f workers with parenchymal 22 abnormalities by quartile obtained from J. Lockey, University o f Cincinnati (personal communication to Robert 23 Benson, U.S. EPA). 24 25 This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-40 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-12. Prevalence of pleural thickening in 280 participants according to 2 various cofactors 31023456789 Variable Number of workers Number with pleural thickening (% ) Crude OR 95% Cl p-Value H ired on or before 1973 186 H ired after 1973 94 70 (37.6) 10(10.6) 5.07 Reference 2.47-10.41 <0.001 Body M ass Index. ' kg/rn2 <24.9 28 2 5 -2 9 .9 101 >30 110 8 (28.6) 31 (30.7) 27 (24.5) Reference 1.11 0 .4 4 -2 .7 9 0.52 0.81 0 .3 2 -2 .0 6 0.43 Ever smokedb Yes No 184 55 (29.9) 96 25 (26.04) 1.21 0.70-2.11 0.50 Reference Age at time o f interview 40-49 55 50-59 116 >60 109 5 (9 .1 ) 28 (24.1) 47 (43.1) Reference 3.18 7.58 1.16-8.76 2 .8 0 -2 0 .4 9 0.03 <0.001 F em ale M ale 16 1 (6.3) 264 79 (29.9) R eference 6.40 0 .8 3 -4 9 .3 2 0.07 4 5 7? = 239 for Body Mass Index due to 38 persons undergoing phone interview and 3 persons with onsite interviews 6 who were not measured for height and weight. 7 ' Smoking history as recorded in 2004 questionnaire. Of these 280 participants, 20 persons reported never smoking 8 in the 1980 questionnaire but subsequently reported a history o f smoking in the 2004 questionnaire (either current 9 or ex-smoker). 10 11 Source: Rolls et al. (2008) 12 13 14 Pleural thickening was observed in 80 workers (28.7%), and small opacities (>l/0) were 15 observed in 8 (2.9%). Six of the 8 participants with small opacities also had pleural thickening 16 (4 as LPT, 2 as DPT). The prevalence of pleural thickening increased across exposure quartiles 17 from 7.1% in the first quartile to 24.6%, 29.4%, and 54.3% in the second, third, and 18 fourth quartiles, respectively (see Table 4-11). The range of exposures was estimated as 19 0.01-0.28, 0.29-0.85, 0.86-2.20, and 2.21-19.03 ftber/cc-years in the first, second, third, and 20 fourth quartiles, respectively (Rohs et al., 2008). 21 Pleural thickening was associated with hire on or before 1973 and age at time of 22 interview but was not associated with body mass index (BMI) or smoking history (ever smoked) 23 (see Table 4-12). Body mass index is a potentially important confounder because fat pads can This document is a draftfor review purposes only and does not constitute Agency policy. 4-41 DRAFT--DO NOT CITE OR QUOTE 1 sometime be misclassified as localized pleural thickening. A hire date of on or before 1973 and 2 ages at time of interview are each highly correlated with cumulative exposure to fibers. The 3 small number of females (n= 16) in the cohort limits the analysis of the association with sex. 4 Modeling of odds ratios with cumulative fiber exposure and including various cofactors (age, 5 hired before 1973, or BMI) with the first exposure quartile as the reference was also conducted. 6 Each model demonstrated the same trend: increased prevalence of pleural thickening with 7 increasing cumulative exposure to fibers. Adjusting for age, date of hire, and body mass index 8 resulted in odds ratios of 2.7, 3.5, and 6.9 for the second, third, and fourth quartiles, respectively. 9 Age-adjusted and BMI-adjusted results were included in Table 4-11. There was no evidence of 10 significant interactions using this modeling. 11 There was potential coexposure to a number of herbicides, pesticides, and other 12 chemicals in the facility (personal communication to Robert Benson, EPA Region 8, from Ivan 13 Smith, The Scotts Company, June 7, 2007). The herbicides and pesticides used during the time 14 when Libby ore was used included atrazine, benomyl, bensulide, chloroneb, chlorothalonyl, 15 chlorpyrifos, 2,4-D, dacthal, diazinon, dicamba, dephenamid, disodium methanearsonate, dyrene, 16 ethoprop, linuron, MCPP, monuron, neburon, oxadiazon, terrachlor, pentachlorophenol, 17 phenylmercuric acetate, siduron, terrazole, thiophannate-methyl, thiram. Other chemicals used 18 included ammonium hydroxide, brilliant green crystals, caustic soda, corncobs, ferrous 19 ammonium sulfate, ferrous sulfate, florex RVM, frit-504, frit-505, hi sil, lime, magnesium 20 sulfate, mon-a-mon, potash, potassium sulfate, Sudan orange, Sudan red, sulfur, sulfuric acid, 21 UFC, urea, and Victoria green liquid dye. No quantitative information on exposure to these 22 chemicals is available. However, the addition of the other chemicals to the vermiculite carrier 23 occurred in a different part of the facility after expansion of the vermiculite ore. Industrial 24 hygiene monitoring in these areas showed very low levels of fibers in the air. In addition, none 25 of these other chemicals is volatile. Thus, it is unlikely that workers would be coexposed by 26 inhalation to these other chemicals. EPA has no information indicating that exposure to any of 27 these individual chemicals causes pleural thickening or evidence of small opacities typical of 28 those found in workers employed in the Marysville facility. The spectrum of radiographic 29 abnormalities observed in the lung and pleura are the same in the Marysville workers, the Libby 30 workers (see Section 4.1.1.4.2, Table 4-7), and the Libby community survey (including workers) 31 (see Section 4.1.2.2, Table 4-9). This document is a draftfor review purposes only and does not constitute Agency policy. 4-42 DRAFT--DO NOT CITE OR QUOTE 1 This study demonstrates that exposure to Libby Amphibole asbestos can cause 2 radiographic evidence of pleural thickening and parenchymal abnormalities (small opacities) in 3 exposed workers. The prevalences of radiographic abnormalities involving the pleura were 4 28.7% in 2004 (80/280), compared to a 2% prevalence observed in 1984 (10/501). This apparent 5 increase in prevalence is most likely due to the additional time between the two studies giving 6 additional time for the abnormalities to become apparent in conventional X-rays. The follow-up 7 study also shows an increasing prevalence of pleural thickening with increasing cumulative 8 exposure to Libby Amphibole asbestos. 9 The influence of some potential sources of selection bias in Rohs et al. (2008) is difficult 10 to qualitatively or quantitatively assess. One type of selection is the loss due to the death of 11 84 of the 513 (16%) workers in the first study; this group may represent less healthy or more 12 susceptible population. Exclusion of the very sick or susceptible may imply that the population 13 of eligible participants was somewhat healthier that the whole populati on of workers; this 14 exclusion may result in an underestimation of risk. Another type of selection is the loss due to 15 nonparticipation among the 431 individuals identified as alive in 2004 (n = 135 refusals and 16 nonresponders; 31%). Participation rates in epidemiologic studies can be associated with better 17 health status, and participation is often higher among nonsmokers compared with smokers. This 18 type of selection of a relatively healthier group (among the living) could also result in an 19 underascertainment of the risk of observed abnormalities within the whole exposed population. 20 However, if parti cipati on was related differenti ally based on exposure and outcome (i.e., if 21 workers experiencing pulmonary effects and who were more highly exposed were more likely to 22 participate than the highly exposed workers who were not experiencing pulmonary effects), the 23 result would be to overestimate the exposure response. This latter scenario is less likely to occur 24 for asymptomatic effects (i.e., abnormalities detected by chest X-ray), such as those that are the 25 focus of this study than for symptoms such as shortness of breath or chest pain. 26 Some information is available on differences by participation status in the Rohs et al. 27 (2008) study. Although current age was similar (mean: 59.1 and 59.4 years, respectively, in 28 participants and living nonparticipant groups, p = 0.53), participants were more likely to have 29 been hired before or during 1973 (66.4 and 49.7%, respectively,/) = 0.001), and had higher mean 30 exposure levels (mean cumulative exposure: 2.48 and 1.76 fiber/cc-years, respectively,/? = 0.06). 31 Participants were also somewhat less likely to be ever smokers (58.6%) compared with the living This docum ent is a draftfo r review purposes only and does not Constitute A gency policy. 4-43 DRAFT--DO NOT CITE OR QUOTE 1 nonparticipants (66.2%). Using a conservative assumption that all living nonparticipants would 2 have had normal X-rays, resulted in estimated prevalences of pleural abnormalities of 3.7, 13.9, 3 18.5, and 38.3%, respectively, in the lowest-to-highest exposure quartile, with corresponding 4 odds ratios of 1.0 (referent), 4.19 (95% Cl: 1.34, 13.08), 5.91 (95% Cl: 1.95, 17.93), and 16.15 5 (95% Cl: 5.53, 47.17). This pattern is similar to that observed in the analysis that excludes the 6 living nonparticipants, indicating the observed trend with exposure was not an artifact of a bias 7 introduced by differences in participation rates among the workers. 8 9 4.1.3.1. Summary o f Marysville, OH Vermiculite Processing Plant Worker Studies 10 The studies conducted in the 1980s (Lockev et al.. 19841 and the follow-up of the cohort 11 (Rohs et al .,,2008) indicate that pleural thickening can be seen among workers in this plant, with 12 increasing prevalence with increasing cumulative exposure. Radiographic evidence of small 13 opacities (interstitial changes in the lung) increased from 0.2% in the original study to 2.9% and 14 radi ographic evidence of pleural thickening increased from 2 to 28.6% of participants in the 15 follow-up study. No effects on lung function were found in the original study (Lockev et ah. 16 19841. Lung function was not reported for the cohort follow-up, despite greater prevalence of 17 radiographic abnormalities (Rohs et af. 20081. 18 19 4.1.4. Community Studies from Other Vermiculite Processing Plants 20 ATSDR has completed community evaluations of 28 sites, in addition to Libby, 21 surrounding exfoliation plants that require further evaluation by EPA because of current 22 contamination or evidence (based on a database of invoices) that the plant processed more than 23 100,000 tons of vermiculite from the Libby, MT mine (see Figure 4-1). Nine of these 24 evaluations included analyses conducted in conjunction with state health departments using 25 death certificate data (see Table 4-13). These community-level evaluations do not address 26 individual exposures or residential histories; therefore, the evidence in these evaluations 27 pertaining to disease risk is somewhat limited. 28 This document is a draftfor re\'iew purposes only and does not Constitute Agency policy. 4-44 DRAFT--DO NOT CITE OR QUOTE 1 Figure 4-1. Location of 28 sites included in the Phase 1 community 2 evaluations conducted by ATSDR. 3 Source: ATSDR (2008a) http://www.atsdr.cdc.gov/asbestos/sites/national_map/. 4 5 Table 4-13. Description of study areas in ATSDR health consultations 6 evaluating cancer incidence and mortality8 7 Site, exposure period Study area (w from 1990 census) Year of report Los Angeles, CA, 1950-1977 Incidence: census tract ( = 21.945) Mortality: zip code ( = 57,615) 2007 Newark, CA, 1967-1992 Incidence: census tract (n = 7,785) Mortality: zip code (n = 37,861) 2005 Santa Ana, CA. 1972-1993 Census tract (35,000) 2003 West Chicago, IL, 1974-1996 Mortality: zip code (n = 14,796) 2003 Dearborn, MI, early 1950s-1989 City limits (/; = 89,015) 2005 St. Louis, Missouri, 1956-1988 Census tracts (// = 20,112) 2006 Trenton, NJ, 1920s-1990 Census tracts and areas (n = 26.762) 2005 Edgewater, NJ. not reported Not reported 2005 Marysville, OH, 1963-19809 City limits (n = 9,656) 2005 8 9 aAll incidence studies used Surveillance, Epidemiology, and End Results (SEER) data as comparison group except 10 New Jersey, which used New Jersey state rates. All mortality studies used U.S. rates from the National Center for 11 Health Statistics. 12 bThe ATSDR health consultation data presented incidence data from 1979-2000, but the 1986-1995 incidence data 13 and the mortality data were obtained from the report o f the New Jersey Department o f Health and Social Services 14 http://www.atsdr.cdc.gov/asbestos/sites/liealth consultations/index.html. 15 cThe start date for the use o f the Libby, MT vermiculite was given as variously described as 1963 or 1967 in the 16 ATSDR health consultation report (ATSDR, 2008b): the studies by Lockey et al. (1984) and Rohs et al. (2008) 17 used 1957 and 1963, respectively, as the start date. This document is a draftfor review purposes only and does not constituteAgency policy. 4-45 DRAFT--DO NOT CITE OR QUOTE 1 The lung cancer standardized incidence ratios for these evaluations range from 2 0.74-1.07, and the SMRs range from 0.74-1.1, indicating little evidence of an increased risk of 3 lung cancer among these studies (see Table 4-14). As expected from the small number of 4 observations, the standardized incidence ratios for mesothelioma or the category of cancer of the 5 peritoneum, retroperitoneum, and pleura (excluding mesothelioma, but which could reflect some 6 misdiagnoses) are more variable, ranging from approximately 0.5-2.5. Breast and prostate 7 cancer were selected as negative controls (i.e., cancers that have not previously been associated 8 with asbestos exposure) in these evaluations. For breast cancer, the standard incidence ratios 9 (SIRs) ranged from 0.73 to 1.25, and for prostate cancer, the SIRs ranged from 0.58 to 1.11, 10 similar to the variability seen among the estimates for lung cancer. In summary, these studies do 11 not provide evidence of an increased risk of lung cancer in the communities surrounding plants 12 that processed vermiculite contaminated with Libby Amphibole asbestos; the small numbers of 13 mesothelioma cases and potential contribution of other asbestos-related sites in some areas make 14 it very difficult to interpret these data. A major limitation of these studies is the lack of 15 information on exposure. Selection of the study population is based on geographic area, with no 16 site-specific or individual-level assessment of relevant exposure pathways. Thus, the extent to 17 which community members were exposed around these facilities is unknown. The use of this 18 type of broad exposure characterization would be expected to result in considerable exposure 19 misclassification. As a result, more refined study designs are needed to evaluate risk to 20 individuals potentially exposed to Libby Amphibole asbestos in their community due to 21 operations at the expansion plants. 22 4.1.4.1. Summary o f Community Studiesfrom Other Vermiculite Processing Plants 23 The community-based mortality studies around the 28 exfoliation plants that processed 24 vermiculite contaminated with Libby Amphibole asbestos provide little evidence of an increased 25 risk of asbestos-related cancers in the surrounding communities. These studies are quite limited, 26 however, by the broad exposure classification and the inability to limit the analysis to individuals 27 who had resided in the specific areas during the relevant exposure periods. Additional studies 28 would be needed to more fully examine the potential risks associated with residential exposures 29 from these sources. 30 31 This document is a draftfor review purposes only and does not constitute Agency policy. 4-46 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-14. Incidence and mortality results for potential asbestos-related 2 cancers (by cancer site) in communities in the vicinity of 3 vermiculite-processing facilities (with ATSDR health consultations 4 evaluating potential pathways of exposure) 5 Incidence3 Mortality*1 Study areac Observed Expected' SIR (95% Cl) Observed Expected' SM R (95% Cl) Lung and bronchus Los Angeles, CAd 100 Newark, CAd 29 Santa Ana, CAd 79 West Chicago, IL - Dearborn, MI 757 St. Louis, MO - Trenton, NJ 496 117.4 27.2 95.4 764.4 671.0 0.85 (0.69, 1.04) 210 1.07 (0.71, 1.53) 125 0.83 (0.66, 1.03) - - 95 0.99 (0.92, 1.06) 1,133 - 319 0.74 (0.68, 0.81) 976 285.0 0.74 124.3 1.01 - - 98.6 0.96 1,261.3 0.90 286.6 1.1 1,100.3 0.89 (0.64, 0.84) (0.84, 1.2) (0.78, 1.18) (0.85, 0.95) (1.0, 1.2) (0.83, 0.94) Edgewater, NJ 35 30.7 1.14 (0.80, 1.59) 51 50 1.02 (0.76, 1.34) Marysville, OH - - - 106 98.1 1.1 (0.9, 1.3) M esothelioma Los Angeles, CAd 1 1.9 0.53 (0.01, 2.96) - - - - Newark, CAd 1 0.4 2.49 (0.03, 13.9) - - - - Santa Ana, CAd 4 1.5 2.68 (0.72, 6.87) - - - - West Chicago, IL - - - - - - - - Dearborn, MI 8 12.3 0.65 (0.28, 1.28) - - - - St. Louis, MO - - - - - - - - Trenton, NJ 6 10.6 0.57 (0.21, 1.24) - - - - Edgewater, NJ 1 0.5 2.11 (0.03, 11.7) - - - - Marysville, OH - - - - - - - - Peritoneum, retroperitoneum, and pleura Excluding mesothelioma Including mesothelioma Los Angeles, CAd 1 3.1 0.32 (0.00, 1.78) 0 2.1 0.0 - Newark, CAd 3 0.7 4.06 (0.82, 11.9) 0 0.9 0.0 (0,4.10) Santa Ana, CAd 6 2.7 2.24 (0.82, 4.87) - - - - West Chicago, IL - - - - 1 0.8 1.28 (0.02,7.12) Dearborn, MI 16 19.1 0.84 (0.48, 1.36) 9 9.6 0.93 (0.43, 1.77) St. Louis, MO - - - - 3 2.3 1.3 (0.3, 3.8) Trenton, NJ 10 16.7 0.60 (0.29, 1.10) 18 8.3 2.17 (1.29, 3.43) Edgewater, NJ 1 0.8 1.28 (0.02,7.13) 0 0.2 0.0 - Marysville, OH - - - - 0 0.8 0.0 - This document is a draftfor review purposes only and does not constitute Agency policy. 4-47 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-14. Incidence and mortality results for potential asbestos related 2 cancers (by cancer site) in communities in the vicinity of vermiculite 3 processing facilities (with ATSDR health consultations evaluating potential 4 pathways of exposure) (continued) 5 6 7 aAll incidence studies used Surveillance. Epidemiology, and End Results (SEER) data as the comparison group 8 except New Jersey, which used New Jersey state rates; incidence period in all analyses was 1986-1995. An 9 additional analysis compared the Hamilton, NJ mesothelioma rates to SEER rates: standard incidence ratio (SIR) 10 was reported to be "increased slightly but remained under 1.0," Incidence data, ICD-10 (International 11 Classification o f Diseases) codes: lung and bronchus, C340:C349; mesothelioma, M-9050:9053; peritoneum, 12 retroperitoneum. and pleura, C480:C488, C384; respiratory system and intrathoracic organs, C320:C399-excluding 13 mesothelioma; selective digestive organs, C150:C218, C260-C269-excluding mesothelioma. 14 'All mortality studies used U.S. rates from the National Center for Health Statistics. Mortality period was 15 1989-1998 in the Los Angeles and Newark, CA analyses and was 1979-1998 in all analyses. Mortality data, 16 ICD-9 codes: lung and bronchus, 162.2-162.9; peritoneum, retroperitoneum. and pleura, 158, 163; respiratory 17 system and intrathoracic organs, 161-165; selective digestive organs, 150-154, 159. 18 cExpected values have been rounded. 19 dSimilar results were observed in the CA analyses using alternative methods to calculate standardized risk ratios for 20 incidence and mortality. 21 22 Cl = confidence interval. 23 24 Source: Site-specific health consultations at 25 http://www.atsdr.cdc.gov/asbestos/sites/liealth consultations/index.html. 26 27 28 4.1.5. Case Reports 29 Progressive disease from exposure to Libby Amphibole was noted in a case report of fatal 30 asbestosis in an individual who died 50 years after working at a vermiculite processing plant for 31 a few months at about age 17 (Wright et ah, 2002). In another case report, exposures that 32 stemmed from playing for a few years as a child in contaminated vermiculite waste materials 33 around a former Libby vermiculite processing facility was reportedly associated with the 34 development of asbestosis and fatal lung cancer (Srebro and Roggli, 1994). 35 36 4.2. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS IN 37 ANIMALS--ORAL, INHALATION AND OTHER ROUTES OF EXPOSURE 38 Laboratory animal studies with exposure to Libby Amphibole or tremolite asbestos show 39 effects similar to those observed in occupationally exposed human populations including pleural 40 pathology, mesothelioma, and lung cancer. Tremolite is an amphibole asbestos fiber that is a 41 component of Libby Amphibole asbestos (~6%). Also, in early studi es Libby Amphibole 42 asbestos was defined as tremolite. Therefore, laboratory animal studies examining the effect of 43 tremolite exposure have been reviewed and are summarized below to potentially increase This docum ent is a draftfo r review purposes only and does not constitute A gency policy. 4-48 DRAFT--DO NOT CITE OR QUOTE 1 understanding of the effects and mechanisms of Libby Amphibole asbestos. Detailed study 2 summaries can be found in Appendix D and summarized in Tables 4-15 and 4-16. No inhalation 3 studies have been performed for Libby Amphibole asbestos, but chronic intrapleural injection 4 studies in hamsters demonstrate carcinogenicity following exposure. The chronic inhalation and 5 intrapleural injection laboratory animal studies with tremolite asbestos demonstrated pleural 6 pathology and carcinogenicity in rats. These studies support the epidemiology studies of Libby 7 Amphibole asbestos exposure (see Section 4.1), and aid in informing the mechanisms of Libby 8 Amphibole asbestos-induced disease. 9 10 4.2.1. Oral 11 No studies in laboratory animals with oral exposure to Libby Amphibole were found in 12 the literature. However, one chronic cancer bioassay was performed following oral exposure to 13 tremolite. McConnell et al. (1983b) describe part of a National Toxicology Program study 14 (NTP, 1990b) performed to evaluate the toxi city and carcinogenicity of ingesti on of several 15 minerals, including tremolite. The tremolite (Governeur Talc Co, Govemeur, New York) used 16 was not fibrous. No significant tumor induction was observed in the animals with oral exposure 17 to tremolite animals. Although nonneoplastic lesions were observed in many of the aging rats, 18 these were mostly in the stomach and occurred in both controls and exposed animals. The 19 observed lesions included chronic inflammation, ulcerati on, and necrosis of the stomach 20 (McConnell et al.. 1983b). McConnell et al. (1983b) suggested that nonfibrous tremolite could 21 account for the lack of toxicity following exposure in this group of animals. Also, oral studies of 22 asbestos, in general, show decreased toxicity and carcinogenicity as compared to inhalation and 23 implantation/injection studies (Condie. 1983). 24 25 4.2.2. Inhalation 26 There are no laboratory animal studies following inhalation exposure to Libby 27 Amphibole asbestos; however two studies have examined the effect of inhalation exposure to 28 tremolite in Wistar rats (Bernstein et al.. 2005; Bernstein et al, 2003; Davis et aL 1985). Davis 29 et al. (1985) performed a chronic inhalation study examining response in male Wistar rats 30 exposed in a chamber to 10 mg/m (-1,600 fibers/mL, >5 pm) of commercially mined tremolite 31 over a 12-month period. Bernstein et al. (2005; 2003) exposed Wistar rats to tremolite This docum ent is a draftfo r review purposes only and does not Constitute A gency policy. 4-49 DRAFT--DO NOT CITE OR QUOTE Table 4-15. In vivo data following exposure to Libby Amphibole asbestos This docum ent is a draftfo r re\'iew purposes only and does not constitute A gency policy. 4-50 DRAFT--DO NOT CITE OR QUOTE Species (sex) LVG:LAK Hamsters (M) (n ~ 60/group) Exposure route Intraperitoneal injection (once) 25 mg/0.5 mL 0.9% NaCl solution Fiber type Tremolite (Sample 60) and tremolite + vermiculite (Sample 63) C57B1/6 mice (M, F) (n = 7/group) intratracheal instillation Libby Amphibole asbestos (once) (Six Mix) and crocidolite 1 wk. 1 mo, 3 mo C57B1/6 mice P C F) (n = 7/group) 100 pg o f sample in 30 |iL saline Intratracheal instillation Libby Amphibole asbestos (once) (Six Mix) and crocidolite 1 \vk. 1 mo, 3 1110 Wistar Kyoto rats (M) (a = 12/group) 100 pg o f sample in 30 pL saline Intratracheal instillation Libby Amphibole asbestos (once) (Six Mix) 1 d, 1 wk, 1 mo Spontaneously Hypertensive (SH) (a = 6/group) 0.25 or 1.0 mg/rat SH Heart Failure (SHHF) (77 = 6/group) Effects3 Pleural adhesions (fibrosis): examined 10 animals/group at ~3 mo post exposure: Sample 60: 10/10; Sample 63: 10/10; Control: 0/10 Reference Smith (1978) (W.R. Grace study) Mesothelioma: Sample 60: 5/66; Sample 63: 5/64; Control: 0/60 Altered gene expression in mice exposed to both samples; increase in collagen in exposed animals Putnam et aJ. (2008) Collagen gene expression and protein levels increased following exposure to both forms of asbestos (~1 mo post exposure). Smartt et al. (2010) Strain-related differences observed in bio markers o f inflammation following exposure to Libby Amphibole asbestos. Shannahan et al. (2011a) No differences were observed in histopathology. Table 4-15. In vivo data following exposure to Libby Amphibole asbestos (continued) This docum ent is a draftfo r re\'iew purposes only and does not constitute A gency policy. Species (sex) Spontaneously Hypertensive (SH) (M) (// = 8/group) Fisher 344 rats (M) (;? = 8/group) Exposure route Fiber type Intratracheal instillation Libby Amphibole asbestos (once) (Six Mix) 4 h, 1 d 1.0 mg deferoxamine (DEF); 21 pgFeC l3; 0.5 mg LA, 0.5 mg FeLA; 0.5 mg LA + 1 mg DEF in 300 pL saline Intratracheal instillation (once) 1 d, .3 d, 7 d, 2 wk, 3 mo Libby Amphibole asbestos (Six Mix) Amosite 0.65 or 6.5 mg/rat LA; 0.65 mg amo site in 250 pL saline Effects3 Statistically significant increases in neutrophils was observed in B ALF in animals exposed to LA, FeLA and LA + DEF with the greatest increase observ ed in the LA+DEF animals. Reference Shannahan et al. (2011b) Statistically significant increases in inflammatoiy markers were observed following exposure to LA and amosite, including increased neutrophils and inflammatory gene expression, with the greatest increase in amosite-exposed rats. Padilla-Carlin et al. (2011) aWhen available, results are shown as number of animals with tumors/total number o f animals examined. 4-51 DRAFT--DO NOT CITE OR QUOTE Table 4-16. In vivo data following exposure to trem olite asbestos This docum ent is a draftfo r re\'iew purposes only and does not constitute A gency policy. 4-52 DRAFT--DO NOT CITE OR QUOTE Species (sex) F344 rats (M, F) (77 = 100 to 250/group) Wistar rats (M) (77 = 48) AF/Han rats (77 = 33-36/group) Exposure route Fiber type Oral 1% bw in feed pellets; lifetime exposure starting in dam Tremolite-nonfibrous (Govemeur Talc Co., Govemeur, NY) Inhalation 10 mg/m3 (7 h each day. 5 days per week, total o f 224 days) South Korean tremolite and brucite Intraperitoneal injection Tremolite (Six samples) 10 mg/2 mL PBS; single exposure Hamsters (77 < 35/group) Intrapleural injection 10 or 25 mg Four types o f tremolite (Sample FD-14; 275; 31; 72) Sprague-Dawley and Wistar rats (77 - 32 Wistar rats (Sample A); 48 Sprague-Dawley rats [Samples B and C]) Intrapleural injection 20 mg/rat Tremolite (Three samples) Effects3 Offspring from exposed mothers were smaller at weaning and throughout life; No toxicity or increase in neoplasia in tremolite rats as compared to controls. Reference McConnell et at. (1983b) Increased fibrosis (19/39) and carcinogenesis (18/39). Davis et al. (1985) All six fibers could induce mesothelioma: California: 36/36b Swansea: 35/36b Korea: 32/36b Italy: 24/36 Carr Brae: 4/33 Shininess: 2/36 Sample FD-14: 0/35 Sample 275: 0/34 (10 mg); 0/31 (25 mg) Samples 31: 3/41 (10 mg); 12/28 (25 mg) ` Sample 72: 4/13 (10 mg); 13/20 (25 mg) No tumors following exposure to Samples A and B; Sample C: 14/47 Davis et al. (1991) Smith et al. (1979) Wagner et al. (1982) Table 4-16. In vivo data following exposure to trem olite asbestos (continued) This docum ent is a draftfo r re\'iew purposes only and does not constitute A gency policy. Species (sex) Osbome-Mendel rats (n = 28/group) Wistar rats (F) (n = 40/group) Wistar rats (M) (;? = 56) Exposure route Hardened gelatin technique Fiber type Tremolite (Two samples) 40 mg Intraperitoneal injection Tremolite 1 l 3.3 and 1 x 15 mg, lifetime observation Inhalation (flow-past nose only) Tremolite C57B1/6 mice (F) (n = 10/group) 100 fibers/cm3 longer than 20 pm, 5 days, follow-up 1 year later Intratracheal instillation Tremolite and wollastonite Two doses of 60 pg each given 1 week apart in the first and second week o f a 7-month experiment Effects3 Sample 1: 21/28 pleural sarcomas Sample 2: 22/28 pleural sarcomas Reference Stanton et al. (1981) Limited details in text. Increase in mesothelioma following exposure to tremolite: 3.3 mg sample: 9/29; 15 mg sample: 30/37 Roller etal. (1997. 1996) Tremolite had a pronounced inflammatoiy Bernstein et al. (2005; response with rapid granuloma development (1 day 2003) post exposure); Slight interstitial fibrosis observed at 90 and 180 days postexposure. Tremolite-exposed mice demonstrated increased IgG immune complex deposition in the kidneys, increased size o f local lymph nodes, and increased total cell count. Plan et al. (2008) aWhen available, results are shown as number of animals with tumors/total number o f animals examined. hAsbestifonn types led to mesothelioma in most if not all exposed animals in this study. 4-53 DRAFT--DO NOT CITE OR QUOTE 1 (100 fibers/cm3 ) and chrysotile for 13 consecuti*ve weeks (6 hours per day, 5 days per week) with 2 1-year follow-up. The results of these inhalation studies produced pronounced inflammation and 3 very high levels of pulmonary fibrosis. Davis et al (1985) also demonstrated an increase in 4 carcinomas and mesotheliomas following exposure to tremolite, with no pulmonary tumors 5 observed in the controls. These results show that Wistar rats exposed to tremolite exhibited 6 increased numbers of pulmonary lesions and possibly tumors. 7 8 4.2.3. Intratracheal Instillation Studies 9 Intratracheal instillation has been used to examine the effect of exposure to Libby 10 Amphibole (Padilla-Carlin et al.. 2011; Sliannatian et al.. 201.1 a; Shannahan et al... 2011b; Smaitt 11 et al.. 2010; Putnam et al.. 2008) and tremolite asbestos (Blake et al.. 2008; Pfau et al, 2008; 12 Sahu et al.. 1975). These studies exposed C57B1/6 mice (100 pg/mouse), Wistar Kyoto (WKY) 13 rats (0.25 or 1 mg/rat) or Fisher 344 rats (0.65 or 6.5 mg/rat) once to Libby Amphibole asbestos 14 and analyzed the results up to 3 month postexposure. Putnam et al. (2008) observed nonsta 15 tistically significant increases in collagen following exposure to Libby Amphibole asbestos, as 16 well as gene expression alterations related to membrane transport, signal transduction, epidermal 17 growth factor signaling, and calcium regulation. Smartt et al. (2010) followed up this study by 18 analyzing specific genes by quantitative RT-PCR for genes involved in collagen accumulation 19 and scar formation (Coll Al, Coll A2, Col3Al). Libby Amphibole asbestos exposure led to 20 increased gene expression of Col lA2 at 1 week postinstillation and Col3Al at 1 month post 21 exposure. Both studies observed increased inflammation, however, Libby Amphibole asbestos 22 exposure demonstrated minimal inflammation that did not progress in the time points examined. 23 These studies demonstrate that exposure to Libby Amphibole asbestos may lead to inflammation 24 and fibrosis. Shannahan et al. (2011a) exposed two rat models of human cardiovascular disease 25 to Libby Amphibole asbestos to determine if the preexisting cardiovascular disease in these 26 models would impact the lung injury and inflammation following exposure. Healthy WKY rats 27 were compared to spontaneously hypertensive (SH) and spontaneously hypertensive heart failure 28 rats following exposure. All rats (male only) were exposed to 0, 0.25, or 1.0 mg/rat via 29 intratracheal instillation and were examined at 1 day, 1 week and 1 month postexposure. No 30 changes were observed histopathologically, however, changes were observed in markers of 31 homeostasis, inflammation and oxidative stress. While inflammation and cell injury were This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-54 DRAFT--DO NOT CITE OR QUOTE 1 observed in all strains, no strain-related differences were observed following exposure to Libby 2 Amphibole asbestos (Shannahan et aL 2011a). In a follow-up study to further examine the role 3 of iron in the inflammatory response to Libby Amphibole asbestos exposure, Shannahan et al. 4 (20Llb) exposed SH rats to Libby Amphibole asbestos alone and with bound Fe as well as with 5 an iron chelator (deferoxamine, DEF). Exposure to Libby Amphibole asbestos led to statistically 6 significant increases in inflammatory markers (e.g., neutrophils, IL-8) with the greatest increase 7 occurring in the presence of DEF. Iron bound to Libby Amphibole asbestos was not released 8 following instillation except in the presence of DEF as supported by the lack of increase in 9 BALF iron. These results suggest that chelation of iron bound to Libby Amphibole asbestos as 10 well as endogenous proteins increases the toxicity of Libby Amphibole asbestos in vivo 11 Padilla-Carlin et al. (2011) exposed Fisher 344 rats (male only) to Libby Amphibole 12 asbestos (0.65 or 6.5 mg/rat) or amosite (0.65 mg/rat; positive control) by intratracheal 13 instillation to examine inflammatory response for 3 months post-exposure. Libby Amphibole 14 asbestos exposure led to statistically significant increases of neutrophils in BALF as early 1 day 15 post-exposure, with other inflammatory markers (e.g., protein, LDFI, GGT) increased statistically 16 significantly at different timepoints during the 3 month period post-exposure. However, on a 17 mass basis, amosite produced a greater inflammatory response as measured by inflammatory 18 markers (e.g., neutrophil influx, gene expression changes) and histopathological analysis 19 demonstrating interstitial fibrosis. These studies demonstrate a statistically significant increase 20 in inflammatory response to Libby Amphibole asbestos in mice and rats as measured in BALF 21 by cytology, histopathology and gene expression analysis. Follow-up studies are needed to 22 inform the chronic effects of exposure to Libby Amphibole asbestos. 23 Laboratoiy animal studies of tremolite intratracheal instillation exposure have been 24 performed in mice in doses ranging from 60 pg to 5 mg. Male Swiss albino mice exposed to 25 tremolite (5 mg) via intratracheal instillation demonstrated histological changes (Sahu et ah. 26 1975). Microscopic results following exposure to tremolite showed acute inflammation of the 27 lungs at 7 days post exposure, including macrophage proliferation and phagocytosis similar to 28 that observed with amosite and anthophyllite. Limited progression of fibrotic response was 29 observed at 60 and 90 days post exposure, with no further progression of fibrotic response 30 Blake et al. (2008) and Pfau et al. (2008) examined the role of asbestos in autoimmunity. Blake 31 et al. (2008) performed in vitro assays with Libby Amphibole asbestos (see Section 4.4), and This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-55 DRAFT--DO NOT CITE OR QUOTE 1 both studies performed the in vivo assays with tremolite. C57BL/6 mice were instilled 2 intratracheally for a total of two doses each of 60-ug saline and wollastonite or Korean tremolite 3 sonicated in sterile phosphate buffer saline (PBS,) given 1 week apart in the first 2 weeks of a 4 7-month experiment. Sera from mice exposed to tremolite showed antibody binding colocalized 5 with SSA/Ro52 on the surface of apoptotic blebs (Blake et al.. 2008). In Pfau et al. (2008), by 6 26 weeks, the tremolite-exposed animals had a significantly higher frequency of positive 7 antinuclear antibody tests compared to wollastinate and saline. Most of the tests were positive 8 for dsDNA and SSA/Ro52. Serum isotyping showed no major changes in immunoglobulin 9 subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed mice decreased overall. 10 Further, IgG immune complex deposition in the kidneys increased, with abnormalities suggestive 11 of glomerulonephritis. No increased proteinuria was observed during the course of the study. 12 Local immunologic response was further studied on the cervical lymph nodes. Although total 13 cell numbers and lymph-node size were significantly increased following exposure to tremolite, 14 percentages of T- and B-cells did not significantly change. 15 16 4.2.4. Injection/Implantation Studies 17 There are no laboratory animal studies examining intraperitoneal injection or 18 implantation of Libby Amphibole asbestos. Biological effects following exposure to tremolite 19 have been examined in five intraperitoneal injection studies (Roller et al.. 1997. 1996; Davis et 20 al.. 1991; Wagner et al., 1982; Smith et al... 1979; Smith. 1978) and one implantation study 21 (Stanton et al,, 1981). 22 Studies by Smith and colleagues (Smith et al.. 1979; 1978). Wagner et al. (1982). Davis 23 et al. (1991) and Roller et al. (1997. 1996) demonstrated that intrapleural injections of tremolite 24 asbestos19 is associated with an increase in pleural fibrosis and mesothelioma in hamsters and 25 rats compared to controls or animals injected with less fibrous materials. Doses ranged from 26 10-25 mg/animal for each study, and although carcinogenesis was observed in these studies 27 there was a variable level of response to the different tremolite forms examined. Although these 28 studies clearly show the carcinogenic potential of Libby Amphibole or tremolite asbestos fibers, 29 intrapleural injections bypass the clearance and dissolution of fibers from the lung after 19 Smith (1978) used tremolite from Libby, MT; Smith et al. ( 1979) may also have used tremolite from Libby, MT (i.e., Libby Amphibole asbestos). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-56 DRAFT--DO NOT CITE OR QUOTE 1 inhalation exposures. Further, limited information was provided confirming the presence or 2 absence of particles or fibers less than 5 pm in length in these studies, limiting the interpretation 3 of results. 4 There is one laboratory animal study that examined the effect of tremolite exposure 5 following implantation of fibers in the pleural cavity. Stanton et al. (1981) also examined 6 tremolite and describe a series of studies on various forms of asbestos. Fibers, embedded in 7 hardened gelatin, were placed against the lung pleura. As an intrapleural exposure, results might 8 not be comparable to inhalation exposures, as the dynamics of fiber deposition and pulmonary 9 clearance mechanisms are not accounted for in the study design. Studies using two tremolite 10 asbestos samples from the same lot were described as being in the optimal size range for 11 carcinogenesis; the fibers were distinctly smaller in diameter than the tremolite fibers Smith et al. 12 (1979) used. These samples both had a high number of fibers in the size range (>8-pm long and 13 <0.25-pm diameter; i.e., "Stanton fibers"). Exposure to both tremolite samples led to 14 mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et al. (1981) study also used talc 15 that did not lead to mesothelioma production. 16 There are no studies currently available in laboratory animals exposed to Libby 17 Amphibole asbestos by inhalation. However, the chronic intraperitoneal injection study in 18 hamsters (Smith et al.. 1979; Smith. 1978) demonstrated tumor formation following exposure to 19 tremolite obtained from the Libby, MT mine. No other chronic studies of Libby Amphibole 20 asbestos are avai lable. A recent study in rats examining the impact of preexisting cardiovascular 21 disease on pulmonary inflammation demonstrated an increase in inflammatory markers 22 following exposure to Libby Amphibole asbestos via intratracheal instillation in SH rats as 23 compared to normal healthy controls exposed to the same dose (Shannahan et al.. 2011b). More 24 recent studies examined gene expression changes (Hillegass et al,, 2010; Putnam et al.. 2008) 25 and early protein markers of fibrosi s (Smartt et al.. 2010) in mice exposed to Libby Amphibole 26 asbestos via intraperitoneal injection. These studies demonstrated an increase in gene and 27 protein expression related to fibrosis following exposure to Libby Amphibole asbestos. 28 Tremolite fibers, although obtained from different locations throughout the world, consistently 29 led to pulmonary lesions and/or tumor formation with various routes of exposure (inhalation, 30 injection, instillation) and in multiple species (rats, hamsters, and mice) (Bernstein et al.. 2005; 31 Bernstein et al.. 2003; Roller et al... 1997. 1996; Davis et al.. 1985; Wagner et al.. 1982; Stanton This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-57 DRAFT--DO NOT CITE OR QUOTE 1 et al., 1981). Although comparing potency of the various forms of tremolite is difficult given the 2 limited information on fiber characteristics and study limitations (e.g., length of follow-up 3 postexposure), these results show potential increased risk for cancer (lung and mesothelioma) 4 following exposure to tremolite asbestos. 5 The results of the studies described above show the fibrogenic and carcinogenic potential 6 of Libby Amphibole and tremolite asbestos. Further, the more recent studies by Blake et al. 7 (2008) and Pfau et al. (2008) support human studies demonstrating potential autoimmune effects 8 of asbestos exposure (see Section 4.3.1). 9 10 4.2.5. Summary of Animal Studies for Libby Amphibole and Tremolite Asbestos 11 Tables 4-15 and 4-16 summarize the studies described in this section, with full study 12 details available in Appendix D. Limited in vivo studies have been performed exposing 13 laboratory animals to Libby Amphibole asbestos. One intrapleural injection study using 14 tremolite from the Libby, MT area is included in this section under Libby Amphibole asbestos 15 since earlier terminology for Libby Amphibole asbestos was often tremolite (Smith. 1978). 16 Hamsters in this study exposed to Libby Amphibole asbestos developed fibrosis and 17 mesothelioma following exposure. Subchronic studies in mice (Smartt et al.. 2010; Putnam el 18 al.. 2008) demonstrated gene and protein expression changes related to fibrosis production 19 following exposure to Libby Amphibole asbestos. Finally, short-term studies in rats 20 demonstrated an increase in inflammatory markers following exposure to Libby Amphibole 21 asbestos (Padilla-Carlin et al.. 2011; Shasmahan. et al.. 201 la; Shannahan et al.. 2011b). 22 Because tremolite is part of Libby Amphibole asbestos, results from tremolite studies 23 were also described. In general, fibrous tremolite has been shown to cause pulmonary 24 inflammation, fibrosis and/or mesothelioma or lung cancer in rats (Bernstein et al.. 2005; 25 Bernstein et al. 2003; Davis et al., 1991; Davis et al... 1985; Wagner et al.. 1982) and hamsters 26 (Smith et al.. 1979). The single short-term study on mice showed limited response to tremolite 27 (Sahu et al.. 1975). The one chronic-duration oral study (McConnell et al.. 1983b) did not show 28 increased toxicity or carcinogenicity; this study, however, used only nonfibrous tremolite, which 29 later studies showed to be less toxic and carcinogenic than fibrous tremolite (Davis et al.. 1991). 30 Chronic inflammation is hypothesized to lead to a carcinogenic response through the 31 production of reactive oxygen species and increased cellular proliferation (Hanahan and This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-58 DRAFT--DO NOT CITE OR QUOTE 1 Weinberg, 2011). Although limited, the data described in Section 4.2 suggest an increase in 2 inflammatory response following exposure to Libby Amphibole asbestos and tremolite asbestos 3 similar to that observed for other durable mineral fibers [reviewed in Mossman et al. (2007)1. 4 Whether this inflammatory response then leads to cancer is unknown. Studies examining other 5 types of asbestos (e.g., crocidolite, chrysotile, and amosite) have demonstrated an increase in 6 chronic inflammation as well as respiratory cancer related to exposure [reviewed in Kamp and 7 Weitzman (1999)1. Chronic inflammation has also been linked to genotoxicity and mutagenicity 8 following exposure to some particles and fibers (Driscoll et al., 1997; 1996; 1995). The evidence 9 described above suggests chronic inflammation is observed following Libby Amphibole asbestos 10 and tremolite asbestos exposure; however, the role of inflammation and whether it leads to lung 11 cancer or mesothelioma following exposure to Libby Amphibole asbestos is unknown. 12 ROS production has been measured in response to both Libby Amphibole asbestos and 13 tremolite asbestos exposure. Blake et al. (2007) demonstrated an increase in the production of 14 superoxide anion following exposure to Libby Amphibole asbestos. Blake et al. (2007) also 15 demonstrated that total superoxide dismutase was inhibited, along with a decrease in intracellular 16 glutathione, both of which are associated with increased levels of ROS. These results are 17 supported by a recent study in human mesothelial cells (Hillegass et al,, 2010) (described in 18 Section 4.4 and Appendix D). Increased ROS production was also observed in human airway 19 epithelial cells following exposure to Libby Amphibole asbestos (Duncan et al,, 2010) (described 20 in Section 4.4 and Appendix D). This increase in ROS and decrease in glutathione are common 21 effects following exposure to asbestos fibers and particulate matter. Although ROS production is 22 relevant to humans, based on similar human responses as compared to animals, information on 23 the specifics of ROS production following exposure to Libby Amphibole asbestos is limited to 24 the available data described here. Therefore, the role of ROS production in lung cancer and 25 mesothelioma following exposure to Libby Amphibole asbestos is unknown. 26 27 4.3. OTHER DURATION OR ENDPOINT-SPECIFIC STUDIES 28 4.3.1. Immunological 29 Two epidemiology studies have examined the potential role of Libby Amphibole asbestos 30 and autoimmunity. Noonan et al. (2006) used the data from the community health screening to 31 examine self-reported history of autoimmune diseases (rheumatoid arthritis, scleroderma, or This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-59 DRAFT--DO NOT CITE OR QUOTE 1 lupus) in relation to the asbestos exposure pathways described above (see Table 4-17). To 2 provide more specificity in the self-reported history of these diseases, a follow-up questionnaire 3 was mailed to participants to confirm the initial report and obtain clarifying information 4 regarding the type of disease, whether the condition had been diagnosed by a physician, and 5 whether the participant was currently taking medication for the disease. Responses were 6 obtained from 208 (42%) of the 494 individuals who had reported these conditions. Of these 7 208 responses, 129 repeated the initial report of the diagnosis of rheumatoid arthritis, and 8 161 repeated the initial report of the diagnosis of one of the three diseases (rheumatoid arthritis, 9 scleroderma, or lupus). Among people aged 65 and over (n = 34 rheumatoid arthritis cases, 10 determined using responses from the follow-up questionnaire), a two- to threefold increase in 11 risk was observed in association with several measures reflecting potential exposure to asbestos 12 (e.g., asbestos exposure in the military) or specifically to Libby Amphibole asbestos (e.g., past 13 work in mining and milling operations, use of vermiculite in gardening, and frequent playing on 14 vermiculite piles when young). Restricted forced vital capacity, presence of parenchymal 15 abnormalities, playing on vermiculite piles, and other dust or vermiculite exposures were also 16 associated with rheumatoid arthritis in the group younger than 65 {n = 95 cases). Restricted 17 forced vital capacity was defined as FVC <80% predicted and a ratio of FEV1 to 18 FVC >70% predicted. For all participants, an increased risk of rheumatoid arthritis was observed 19 with increasing number of exposure pathways. RRs of 1.0, 1.02, 1.79, 2.51, and 3.98 were 20 observed for 0 (referent), 1, 2-3, 4-5, and 6 or more pathways, respectively (trend p < 0.001, 21 adjusting for restrictive spirometry, parenchymal abnormalities, and smoking hi story). Although 22 the information gathered in the follow-up questionnaire and repeated reports of certain diagnoses 23 decreased the false-positive reports of disease, considerable misclassification (over-reporting and 24 under-reporting) is likely, given the relatively low confirmation rate of self-reports of 25 physician-diagnosed rheumatoid arthritis (and other autoimmune diseases) seen in other studies 26 (Karlson et a!.. 2003; Rasch et aL 2003; Ling et aL 2000). 27 Another study examined serological measures of autoantibodies in 50 residents of Libby, 28 MT, and a comparison group of residents of Missoula, Montana (Pfau et aL 2005); (see 29 Table 4-17). The Libby residents were recruited for a study of genetic susceptibility to 30 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-60 DRAFT--DO NOT CITE OR QUOTE 1 Table 4-17. A utoim m une-related studies in the Libby, MT community 2 Reference(s) Inclusion criteria and design details Results Noonan et al. (2006) Nested case-control study among 7,307 participants in 2000-2001 community health screening. Conducted interviews, gathered self-reported history o f rheumatoid arthritis, sclerodenna, or lupus. Follow-up questionnaire mailed to participants concerning self-report o f "physician-diagnosis" o f these diseases and medication use. Association with work in Libby mining/milling operations (ages 65 and older): Rheumatoid arthritis OR: 3.2 (95% Cl: 1.3,8.0) Rheumatoid arthritis, lupus, sclerodenna OR: 2.1 (95% Cl: 0.90,4.1) Risk increased with increasing number o f asbestos exposure pathways. Pfau et al. (2005) Libby residents (;? = 50) recruited for study o f genetic susceptibility to asbestos-related lung disease. Missoula, MT comparison group ( = 50), recruited for study o f itmnune function; age and sex-matched to Libby participants. Serum samples obtained; IgA levels, prevalence of antinuclear, anti-dsDNA antibodies, anti-RF antibodies, and anti-Sm, RNP, SS-A, SS-B, and Scl-70 antibodies determined. Increased prevalence o f high titer (>1:320) antinuclear antibodies in Libby sample (22%) compared to Missoula sample (6%). Similar increases for rheumatoid factor, anti-RNP, anti-Scl-60, anti-Sm, anti-Ro (SSA), and anti-La (SSB) antibodies observed in Libby sample. 3 4 5 asbestos-related lung disease, and the Missoula residents were participants in a study of immune 6 function The Libby sample exhibited an increased prevalence (22%) of high-titer (>1:320) 7 antinuclear antibodies when compared to the Missoula sample (6%), and similar increases were 8 seen in the Libby sample for rheumatoid factor, anti-RNP, anti-Scl-60, anti-Sm, anti-Ro (SSA), 9 and anti-La (SSB) antibodies. Although neither sample was randomly selected from the 10 community residents, an individual's interest in participating in a gene and lung disease study 11 likely would not be influenced by the presence of autoimmune disease or autoantibodies in that 12 individual 13 Hamilton et al. (2004), Blake et al. (2008), and Pfau et al. (2008) examined the role of 14 asbestos in autoimmunity in laboratory animal or in vitro studies. Blake et al. (2008) performed 15 in vitro assays with Libby Amphibole asbestos (see Section 4.4), and both studies performed the 16 in vivo assays with tremolite. C57BL/6 mice were instilled intratracheally for a total of two 17 doses each of 60-pg saline and wollastonite or Korean tremolite sonicated in sterile PBS, given 18 1 week apart in the first 2 weeks of a 7-month experiment. Sera from mice exposed to tremolite 19 showed antibody binding colocalized with SSA/Ro52 on the surface of apoptotic blebs (Blake et 20 al.. 2008). In Pfau et al. (2008), by 26 weeks, the tremolite-exposed animals had a significantly This document is a draftfor review purposes only and does not constitute Agency policy. 4-61 DRAFT--DO NOT CITE OR QUOTE 1 higher frequency of positive antinuclear antibody tests compared to wollastinate and saline. 2 Most of the tests were positive for dsDNA and SSA/Ro52. Serum isotyping showed no major 3 changes in immunoglobulin subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed 4 mice decreased overall. Further, IgG immune complex deposition in the kidneys increased, with 5 abnormalities suggestive of glomerulonephritis. No increased proteinuria was observed during 6 the course of the study. Local immunologic response was further studied on the cervical lymph 7 nodes. Although total cell numbers and lymph-node sizes were significantly increased following 8 exposure to tremolite, percentages of T- and B-cells did not significantly change. Hamilton et al. 9 (2004) investigated the ability of Libby Amphibole, crocidolite, and PM2.5 (collected over a 6 10 month period in Flouston, TX, from EPA site 48-201-1035) to alter the antigen-presenting cell 11 (APC) function was altered in cultured human alveolar macrophages. Asbestos exposure 12 (regardless of type) and PM2.5 up-regulated a THi lymphocyte derived cytokine, interferon 13 gamma (IFNy), and the Th2 lymphocyte-derived cytokines interleukin-4 (IL-4) and 14 interleukin-13 (IL-13). There was, however, extreme variation among subjects in the amount of 15 response. In addition, there was no correlation between an individual's cells' response to 16 asbestos versus PM, suggesting that more than one possible mechanism exists for a 17 particle-induced APC effect and individual differential sensitivities to inhaled bioactive particles. 18 Although limited number of studies, these results suggest a possible effect on 19 autoimmunity following exposure to Libby Amphibole asbestos. Further studies are needed to 20 increase understanding of this potential effect. 21 22 4.4. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF 23 ACTION 24 In vitro analysis of fibers depends on the characteristics of the fibers and cell types used 25 for the studies. Therefore, in reviewing the literature it is important to pay attention to cell types 26 used, particularly related to the ability to internalize fibers and produce an oxidative stress 27 response. Results from in vitro studies have demonstrated potential biological mechanisms of 28 oxidative stress and inflammation in response to exposure to Libby Amphibole and tremolite 29 asbestos. These studies are summarized below and in Tables 4-18 and 4-19, with detailed study 30 descriptions available in Appendix D. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-62 DRAFT--DO NOT CITE OR QUOTE 1 Limited in vitro studies have been conducted with Libby Amphibole asbestos from the 2 Zonolite Mountain mine. These studies demonstrated an effect of Libby Amphibole asbestos on 3 inflammation and immune function (Duncan et ah. 2010; Blake et a.L. 2008; Blake et. al... 2007; 4 Hamilton et ah, 2004). oxidative stress (Hillegass et al.. 2010). and genotoxicity (Pietruska et al.. 5 2010). Similar endpoints have been examined in vitro following exposure to tremolite asbestos 6 (Okavasu et al.. 1999; Wylie et al., 1997; Suzuki and Hei, 1996; Athanaslou et al... 1992; Wagner 7 et al.. 1982). 8 9 4.4.1. Inflammation and Immune Function 10 Hamilton et al. (2004) showed an increase in TH1 and TH2 cytokines following exposure 11 to both asbestos and particulate matter, suggesting a similar effect of exposure to both materials 12 on immune function. Analysis of these results is limited, as the use of primary cells in culture 13 that led to an extremely variable response. Two studies by Blake et al. (2008; 2007) further 14 examined the effect of Libby Amphibole asbestos on immune response in murine macrophages. 15 These studies demonstrated that Libby Amphibole asbestos was internalized, and this 16 internalization resulted in an increase in reactive oxygen species (ROS). These studies also 17 showed a variable cytotoxic response, as Libby Amphibole asbestos exposure did not result in a 18 statistically significant increase in cytotoxicity, while crocidolite did. DNA damage also was 19 increased in crocidoiite-exposed cells--but not in Libby Amphibole asbestos exposed-cells. An 20 increase (relative to controls) in autoantibody formation following exposure to Libby Amphibole 21 asbestos also was observed. Studies that examined cellular response to tremolite also found that 22 fiber characteristics (length and width) play a role in determining ROS production, toxicity, and 23 mutagenicity (Okavasu et al.. 1999; Wagner et at., 1982). 24 Mechani sms of oxi dative stress following exposure to Libby Amphibole asbestos were 25 also studied in human mesothelial cells (Hillegass et al.. 2010). Gene expression changes 26 following exposure to 15 x 106 p nr/cnr Libby Amphibole asbestos211as compared to the 27 nonpathogenic control (75 * 106 iim2/cnr glass beads) in the human mesothelial cell line 28 LP9/TERT-1 for 8 and 24 hours. Gene ontology of these results demonstrated alteration s in 29 genes related to signal transduction, immune response, apoptosis, cellular proliferation, 2ULibby Amphibole asbestos samples were characterized for this study with analysis o f chemical composition and mean surface area (Meeker et al.. 2003). Doses were measured in surface area and described based on viability assays as either the -nontoxic (15 h 106 p n f/cn r) or the toxic dose (75 m 106 fuir/cnr). This document is a draftfor review purposes only and does not constitute Agency policy. 4-63 DRAFT--DO NOT CITE OR QUOTE This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy, 1 Table 4-18. In vitro data following exposure to Libby Amphibole asbestos Test system Primary human alveolar macrophages and lymphocytes Murine macrophages (primary and R A W 264.7)3 Fiber type Libby Amphibole asbestos or crocidolite Libby Amphibole asbestos and crocidolite Murine macrophages (primary and RAW264.7) Human lung epithelial cells (wild-type and XRCC1-deficient) Libby Amphibole asbestos or crocidolite Libby Amphibole asbestos or crocidolite Human mesothelial cells (LP9/TERT-1 and HKNM-2) Primary human airway epithelial cells (HAECs) Libby Amphibole asbestos or crocidolite Libby Amphibole asbestos (fractionated and unfractionated), amosite (fractionated and unfractionated), crocidolite Dose/exposure duration 0, 25, 50 ug/mL 24 h Internalization: 0, 5,62.5 pg/cm2 3-24 h Oxidative stress: 0, 6.25, 32.5,62.5 pg/cm2 3, 7, 12, and 24 h Cell viability: 0. 6.25, 32.5, 62.5 pg/cm2 3, 7, 12, and 24 h DN A damage: 0, 6.25, 32.5, 62.5 pg/cm2 3, 7, 12, and 24 h0, 62.5 pg/cm2 0-72 h 5 pg/cm2 24 h 0. 15 x IQ6 pnr/crn2 (nontoxic) and 75 * 106 pm /cm2 (toxic) for 8 or 24 h 0,2.64, 13.2 or 26.4 pg/cm2 2, 4 or 24 h Effects Upregulated TH1 and TH2 cytokines (IFNy, IL-4, IL-13) ' Reference Hamilton et al. (2004) Internalized Libby Amphibole asbestos fibers were mostly less than 2 pm in length Blake et al. (2007) Increased ROS over control (wollastonite) and crocidolite Decreased GSH No effect was observed on cell viability No increase in DNA damage and adduct formation Time-course dose response for apoptosis; Redistribution o f autoantigen on cell surface Blake et al. (2008) Dose-dependent increase in micronuclei in both cell types, but increased in the XRCC1-deficient cells as compared to wild-type Pietruska et al. (2010) Alterations in genes related to oxidative stress, Hillegass et al. particularly SOD2 (2010) Increases in pro-inflammatory gene expression Duncan et al. and ROS production (2010) 4-64 DRAFT--DO NOT CITE OR QUOTE aAll results for RAW264.7. Data not shown for primary cells though authors state similar response to RAW264.7. PBS = phosphate buffer saline, ROS = reactive oxygen species, GSH = glutathione, DNA = deoxyribonucleic acid, LDH = lactic dehydrogenase, BGL = p-glucuronidase, SHE = Syrian hamster ovary, HTE = hamster tracheal epithelial, RPM = rat pleural mesothelial, NIEHS = National Institute of Environmental Health Sciences, HPRT = hvpoxa lithinc-gua nine phosphoribosyltransferase. Table 4-19. In vitro data following exposure to trem olite asbestos This document is a draftfo r re\'iew purposes only and does not constitute Agency policy, 4-65 Test system/species Fiber type Dose/exposure duration Effects Reference Primary murine macrophages Sample A (flake-like from California talc deposits); Sample B (medium-sized fibrous from Greenland); Sample C (fine-fiber material from S. Korea); Positive Control (crocidolite) 0, 50,100, and 150 pg/mL 18 h LDH and BGL levels increased following exposure to Sample C (longer, thinner fibers) and crocidolite (positive control). Sample C led to the greatest increases in giant cell formation and cytotoxicity o f samples tested. Sample B also led to some increased cytotoxicity. Wagner et al. (1982) TA98, TA100, TA102 S. Metsovo tremolite typhimurium TA98, TA100, and T A 102: 0-500 pg/pcr plate 2 days No significant revertants were observed in any o f the three Salmonella strains tested. Athanasiou et al. (1992) V79 and BPNi cells V79 and BPNi: 0-4 pg/cnr 6, 24, and 48 h No affect was observed on gap-junctional intercellular communication. BPNi cells BPNi: 0-2 pg/cnr 24 h Tremolite led to a dose-dependent increase in micronuclei induction. SHE cells SHE: 0-3 pg/cnr 24 h Tremolite exposure led to increased chromosomal aberrations but not in a dose-dependent fashion. A[L] cells (hamster hybrid cells containing human cliroino so me 11) UICC chrysotile, crocidolite, Metsovo tremolite, erionite 0, 2.5-40 pg/mL 24 h Relative increase in heme oxygenase as compared to control. Suzuki and Hei (1996) HTE and RPM cell lines NIEHS chrysotile. NIEHS crocidolite,'F D 14, S157. CPS 183 (talc fibers containing tremolite) Varied (based on weight, fiber Fibrous talc exposure led to limited proliferation Wylie et al. length, and surface area). of cells. (1997) A[L] cells (hamster hybrid cells containing human chromosome 11) Tremolite, erionite, RCF-1 0-400 pg/mL 24 h No significant increase in HPRT mutations for these three fibers; Dose-dependent induction of mutations in CD59 did occur for erionite and tremolite. Okayasu et al. (1999) PBS = phosphate buffer saline, ROS = reactive oxygen species, GSH = glutathione, DNA = deoxyribonucleic acid, LDH = lactic dehydrogenase, BGL = (3-glucuronidase, SHE = Syrian hamster ovary, HTE = hamster tracheal epithelial, RPM = rat pleural mesothelial, NIEHS = National Institute of Environmental Health Sciences, HPRT = hypoxantliine-guanine phosphoribosyltransferase. DRAFT--DO NOT CITE OR QUOTE 1 extracellular matrix, cell adhesion and motility, and only in one gene related to reactive oxygen 2 species processing. Oxidative stress was observed as both dose- and time-dependent in cells 3 exposed to Libby Amphibole asbestos but was increased following exposure to the higher dose 4 of Libby Amphibole asbestos (statistical analysis not possible). Glutathione (GSH) levels were 5 transiently depleted following 2-8 hours exposure to the higher dose of Libby Amphibole 6 asbestos, with a gradual recovery up to 48 hours in LP9/TERT-1 cells (HKNM-2 not analyzed). 7 These studies demonstrate that Libby Amphibole asbestos exposure leads to increases in 8 oxidative stress as measured by ROS production, gene expression, protein and functional 9 changes in oxidative stress proteins (SOD), and GSH level alterations in human mesothelial 10 cells. 11 Gene expression alterations of interleukin-8 (IL-8), cyclooxygenase-2 (COX-2), heme 12 oxygenase (HO)-l as well as other stress-responsive genes as compared to amosite (Research 13 Triangle Institute) was observed in primary human airway epithelial cells (HAEC) following 14 exposure to Libby Amphibole asbestos. Comparisons were made with both fractionated 15 (aerodynamic diameter <2.5 pm) and unfractionated fiber samples (Duncan et aL 2010). 16 Crocidolite fibers (UICC) were also included in some portions of this study for comparison. 17 Primary ITAECs were exposed to 0, 2.64, 13.2, and 26.4 pg/cm of crocidolite, amosite (AM), 18 amosite 2.5 (fractionated), Libby Amphibole asbestos, or Libby Amphibole asbestos 19 2.5 (fractionated) for 2 or 24 hours in cell culture. Cytotoxicity was determined by measurement 20 of lactate dehydrogenase (LDH) from the maximum dose (26.4 pg/cm ) of both amosite and 21 Libby Amphibole asbestos samples, with less than 10% LDH present following exposure to all 22 four samples. Minimal increases in gene expression of IL-8, COX-2, or HO-1 were observed at 23 2 hours postexposure to all five fiber types; at 24 hour postexposure, however, a dose response 24 was observed following exposure to all fiber types with the results showing a pro-inflammatory 25 gene expression response (Duncan et al.. 2010). These results support a limited cytotoxicity of 26 both amosite and Libby Amphibole asbestos under these concentrations and time frames. 27 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-66 DRAFT--DO NOT CITE OR QUOTE 1 4.4.2. Genotoxicity 2 Genotoxicity and, more specifically, mutagenicity, are associated with tumor formation 3 through alterations in genetic material.21 Mutagenicity refers to a permanent effect on the 4 structure and/or amount of genetic material that can lead to heritable changes in function, while 5 genotoxicity is a broader term including all adverse effects on the genetic information (Eastmond 6 et ah, 2009). Results of standard mutation assays like the Ames test, which analyze for point 7 mutations, have found asbestos and other mineral fibers to be negative or only marginally 8 positive (Walker et al,, 1992). Several other studies, however, have shown that asbestos 9 exposure can result in a variety of chromosomal alterations, which are briefly discussed below. 10 Genotoxicity following exposure to asbestos fibers has been described as the result of 11 two distinct mechanisms, either ROS production leading to direct DNA damage, or physical 12 interference of mitosis by the fibers. For both DNA damage and mitotic interference, the fibers 13 must first enter the cell. Some studies have shown that a direct interaction between fibers and 14 cellular receptors might also lead to increased ROS production. ROS production is likely to be a 15 key event in fiber-induced direct DNA damage, as observed following exposure to other forms 16 of asbestos, while the indirect DNA damage requires fiber interaction with cellular components 17 (e.g., mitotic spindle, chromosomes). 18 ROS production and genotoxicity (micronuclei induction) following exposure to Libby 19 Amphibole asbestos has been demonstrated in XRCC1-deficient human lung epithelial 20 H460 cells (Pietruska et al.. 2010). XRCC1 is involved in the repair mechanisms for oxidative 21 DNA damage, particularly single strand breaks. Micronuclei induction was measured following 22 treatment of cells by controls (positive, hydrogen peroxide; negative, paclitaxel) and by 23 5 pg/cm fibers or Ti02 particles for 24 hours. Spontaneous micronuclei induction was increased 24 in XRCC1 -deficient cells in a dose-dependent manner following exposure to crocidolite and 25 Libby Amphibole asbestos as compared to control. These results support a potential genotoxic 26 effect of exposure to both crocidolite and Libby Amphibole asbestos. 21 G enotoxicity. a broad term and refers to potentially harmful effects on genetic material which may be mediated directly or indirectly, and which are not necessarily associated with mutagenicity. Thus, tests for genotoxicity include tests which provide an indication o f induced damage to DNA (but not direct evidence o f mutation) via effects such as unscheduled DNA synthesis, sister chromatid exchange, or mitotic recombination, as well as tests for mutagenicity; M utagenicity: refers to the induction of permanent transmissible changes in the amount or structure of the genetic material o f cells or organisms. These changes, "mutations," may involve a single gene or gene segment, a block o f genes, or whole chromosomes. Effects on whole chromosomes may be structural and/or numerical (as defined in the European Union Technical Guidance on Risk Assessment (CEC. 1996). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-67 DRAFT--DO NOT CITE OR QUOTE 1 Athanasiou et al. (1992) performed a series of experiments to measure genotoxicity 2 following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction, 3 chromosomal aberrations, and gap-junction intercellular communication. Although a useful test 4 system for mutagenicity screening for many agents, the Ames assay is not the most effective test 5 to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through 6 generation of ROS or direct disruption of the spindle apparatus during chromatid segregation. 7 Fibers do not induce ROS in the Ames system, however, and the Salmonella typhimurium strains 8 do not endocytose the fibers. Only one study was found in the published literature that used the 9 Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown 10 to be the causative agent of endemic pleural calcification and an increased level of malignant 11 pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite, 12 S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 pg/plate of asbestos 13 (Athanasiou et al.. 1992). Metsovo tremolite did not yield a statistically significant increase in 14 revertants in the Ames assay, including in the TA102 Salmonella strain, which is generally 15 sensitive to oxidative damage. This study demonstrated clastogenic effects of tremolite, 16 including chromosomal aberrations and micronuclei induction. Tremolite exposure in Syrian 17 hamster embryo (SHE) cells did lead to a dose-dependent increase in chromosome aberrations 18 that was statistically significant at the highest doses tested (1.0--3.0 pg/cm ) (p < 0.01) 19 (Athanasiou et al.. 1992). A statistically significant dose-dependent increase in levels of 20 micronuclei was demonstrated following tremolite exposure at concentrations as low as 21 0.5 pg/cm (p < 0.01) in BPNi cells after 24-hour exposure. Literatures searches did not find 22 tremolite tested for clastogenicity in other cell types, but the results of this study suggest 23 interference with the spindle apparatus by these fibers. No analysis was performed to determine 24 if fiber interference of the spindle apparatus could be observed, which would have supported 25 these results. No effect on the gap-junctional intercellular communication following tremolite 26 exposure was observed in both Chinese hamster lung fibroblasts (V79) and Syrian hamster 27 embryo BPNi cells, which are sensitive to transformation (Athanasiou et al.. 1992). 28 Okayasu et ah (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the 29 man-made ceramic (RCF-1) fiber. ITuman-hamster hybrid A(L) cells contain a full set of 30 hamster chromosomes and a single copy of human chromosom e 11. Mutagenesis of the CD59 31 locus on this chromosome is quantifiable by antibody complement-mediated cytotoxicity assay. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-68 DRAFT--DO NOT CITE OR QUOTE 1 The authors state that this is a highly sensitive mutagenicity assay, and previous studies have 2 demonstrated mutagenicity of both crocidolite and chrysotile (Hei et ah, 1992). The cytotoxicity 3 analysis for mutagenicity was performed by exposing 1 x 10: A(L) cells to a range of 4 concentrations of fibers as measured by weight (0-400 pg/mL or 0-80 pg/'crrL) for 24 hours at 5 37C. CD59 mutant induction showed a dose-dependent increase in mutation induction for 6 erionite and tremolite, but RCF-1 did not. 7 In summary, one in vitro study examined genotoxicity of Libby Amphibole asbestos by 8 measuring DNA adduct formation following exposure via murine macrophages (primary and 9 immortalized) (Blake et ah. 2007). The data showed no increase in adduct formation as 10 compared to unexposed controls. A second study observed increases in micronuclei induction in 11 both normal human lung epithelial cells and XRCC1-deficient cells for both Libby Amphibole 12 and crocidolite asbestos (Pietruska et al.,, 2010). Two studies of tremolite examined 13 genotoxicity. The first found no significant increase in revertants in the Ames assay (Athanasiou 14 et al.. 1992). which is similar to results obtained for other forms of asbestos. This study did find, 15 however, that tremolite exposure led to a dose-dependent increase in chromosome number and 16 micronuclei formation, which has also been described for other asbestos fibers [as reviewed in 17 Hei et al. (2006) and Jaurand and Levy (1997)1. Hei and colleagues (Okavasu et al.. 1999) 18 performed mutation analysis with tremolite and found a dose-dependent increase in mutations in 19 CD59 in hamster hybrid cells. Genotoxicity analysis in humans, following exposure to Libby 20 Amphibole asbestos or tremolite, has not been measured, although other types of asbestos fibers 21 have led to increases in genotoxicity in primary cultures and lymphocytes (Popp et al.. 2005; 22 Poser et aL 2004). In general, these studies have examined genotoxicity with a focus on ROS 23 production as a key event. Although Libby Amphibole asbestos- and tremolite-specific data are 24 limited to in vitro studies, given the similarities in response to other forms of asbestos, there is 25 some evidence to suggest genotoxicity following exposure to Libby Amphibole and tremolite 26 asbestos. However, the potential role of this genotoxicity in lung cancer or mesothelioma 27 following exposure to Libby Amphibole asbestos is unknown. 28 29 4.4.3. Cytotoxicity and Cellular Proliferation 30 The initial stages of tumorigenicity may be an increased cellular proliferation at the site 31 of fiber deposition, which can increase the chance of cancer by increasing the population of This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-69 DRAFT--DO NOT CITE OR QUOTE 1 spontaneous mutations, thereby affording genotoxic effects an opportunity to multiply. 2 Increased cell proliferative regeneration is also a hallmark of tumor clonal expansion and 3 generally occurs in response to increased apoptosis. 4 Wagner et al. (19821 examined the in vitro cytotoxicity of three forms of tremolite used 5 in their in vivo studies. LDH and P-glucuronidase were measured in the medium following 6 incubation of unactivated primary murine macrophages to 50, 100, and 150 pg/mL of each 7 sample for 18 hours. The Korean tremolite (Sample C) produced results similar to the positive 8 control: increased toxicity of primary murine macrophages, increased cytoxicity of Chinese 9 hamster ovary (CHO) cells, and increased formation of giant cells from the A549 cell line. The 10 tremolite sample from Greenland (Sample B) did result in increased toxicity over controls; 11 although to a lesser degree (statistics are not given). Although differential toxicity of these 12 samples was noted on a mass basis, data were not normalized for fiber content or size. The 13 inference is that differential results may be due, at least in part, to differential fiber counts. 14 Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and 15 proliferative effects of asbestos in hamster tracheal epithelial (ITTE) and rat pleural mesothelial 16 (RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise 17 to tracheobronchi al carcinoma, while RPM cells give rise to mesotheliomas. The results of the 18 analysis with fiber exposure by mass (pg/cm ) show elevated colonies in HTE cells following 19 exposures to both asbestos fibers (p < 0.05) at the lowest concentrations, while significant 20 decreases were observed for both asbestos fibers at the higher concentrations (0.5 pg/cm , 21 p < 0.05) (Wylie et al.. 1997). No proliferation was observed for either chrysotile or crocidolite 22 asbestos fibers in RPM cells, but cytotoxicity was observed at concentrations greater than 23 0.05 pg/cm {p < 0.05). All talc samples were less cytotoxic in both cell types. Analyzing the 24 data for cytotoxicity and proliferation based on the exposure measurement demonstrated 25 differences in response depending solely on how the fibers were measured: by mass, number, or 26 surface area. These results show variability in interpreting the results of the same assay based on 27 the defined unit of exposure. Most early studies used mass as the measurement for exposure, 28 which can impact how the results are interpreted. When possible, further analysis of fiber 29 number and surface area would help elucidate the role of these metrics, particularly for in vivo 30 studies. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-70 DRAFT--DO NOT CITE OR QUOTE 1 Tremolite and Libby Amphibole asbestos exposure led to increases in both fibrosis and 2 tumorigenicity in all but one animal study, supporting a possible role for proliferation in 3 response to these fibers. However, there are limited data to demonstrate that increased 4 cytotoxicity and cellular proliferation following exposure to Libby Amphibole asbestos leads to 5 lung cancer or mesothelioma. 6 Summary. The review of these studies clearly highlights the need for more controlled 7 studies examining Libby Amphibole asbestos in comparison with other forms of asbestos and for 8 examining multiple endpoints--including ROS production, DNA damage, and pro-inflammatory 9 gene expression alterations--to improve understanding of mechanisms involved in cancer and 10 other health effects. Data gaps still remain to determine specific mechanisms involved in Libby 11 Amphibole asbestos-induced disease. Studies that examined cellular response to tremolite also 12 found that tremolite exposure may lead to increased ROS production, toxicity, and genotoxicity 13 (Okavasu et al.. 1999; Wagner et ah, 1982). As with the in vivo studies, the definition of fibers 14 and how the exposures were measured varies among studies. 15 16 4.5. SYNTHESIS OF MAJOR NONCANCER EFFECTS 17 The predominant noncancer health effects observed following inhalation exposure to 18 Libby Amphibole asbestos are effects on the lungs and pleural lining surrounding the lungs. 19 Recent studies have also examined noncancer health effects following exposure to Libby 20 Amphibole asbestos in other systems, including autoimmune effects and cardiovascular disease. 21 These effects have been observed primarily in studies of exposed workers and community 22 members and are supported by laboratory animal studies. 23 24 4.5.1. Pulmonary Effects 25 4.5.1.1. Pulmonary Fibrosis (Asbestosis) 26 Asbestosis is the interstitial pneumonitis and fibrosis caused by inhalation of asbestos 27 fibers and is characterized by a diffuse increase of collagen in the alveolar walls (fibrosis) and 28 the presence of asbestos fibers, either free or coated with a proteinaceous material and iron 29 (asbestos bodies). Fibrosis results from a sequence of events following lung injury, which 30 includes inflammatory cell migration, edema, cellular proliferation, and accumulation of 31 collagen. Asbestosis is associated with dyspnea, bibasilar rales, and changes in pulmonary This document is a draftfor review purposes only and does not constitute Agency policy. 4-71 DRAFT--DO NOT CITE OR QUOTE 1 function: a restrictive pattern, mixed restrictive-obstructive pattern, and/or decreased diffusing 2 capacity (ATS. 2004). Radiographic evidence of small opacities in the lung is direct evidence of 3 scarring of the lung tissue and as the fibrotic scarring of lung tissue consistent with mineral dust 4 and mineral fiber toxicity. The scarring of the parenchymal tissue of the lung contributes to 5 measured changes in pulmonary function, including obstructive pulmonary deficits from 6 narrowing airways, restrictive pulmonary deficits from impacting the elasticity of the lung as 7 well as decrements in gas exchange. 8 Workers exposed to Libby Amphibole asbestos from vermiculite mining and processing 9 facilities in Libby, MT, as well as plant workers in Marysville, OH, where vermiculite ore was 10 exfoliated and processed, have an increased prevalence of small opacities on chest X-rays, which 11 is indicative of fibrotic damage to the parenchymal tissue of the lung (Rohs et al.. 2008: 12 Amandus et al.. 1987b: McDonald et al.. 1986b: Lockev et al.. 1984). These findings are 13 consistent with a diagnosis of asbestosis, and the studies are described in detail in 14 Section 4.1.1.4.2. Significant increases in asbestosis as the primary cause-of-death have been 15 documented in studies of the Libby worker cohort report (see Table 4.6 for details) (Larson et al.. 16 2010b: Sull ivan. 2007: Amandus and Wheeler. 1987: McDonald et al.. 1986a). For both 17 asbestosis mortality and radiographic signs of asbestos (small opacities), positive exposure- 18 response relationships are described where these effects are greater with greater cumulative 19 exposure to Libby Amphibole asbestos. 20 Deficits in pulmonary function consi stent with pulmonary fibrosi s have been reported in 21 indivi duals exposed to Libby Amphibole asbestos. The initial study of the Marysville, OH 22 cohort measured but reported no change in pulmonary function (Lockev et al.. 1984). 23 Pulmonary function was not reported for the cohort follow-up, although prevalence of pleural 24 and parenchymal abnormalities was increased (Rohs et al.. 2008). Although studies of the 25 occupational Libby worker cohort do not include assessment of pulmonary function (Amandus et 26 al.. 1987b: McDonald et al.. 1986b) data from the ATSDR community screening, which included 27 workers, provide support for functional effects from parenchymal changes. The original report 28 of the health screening data indicated moderate-to-severe pulmonary restriction in 2.2% of men 29 (Peipins et al.. 2003: ATSDR. 2001b). A recent reanalysis of these data show that for study 30 participants with small opacities viewed on the radiographs (grade 1/0 or greater), and DPT the 31 mean FVC is reduced to 78.76 (3.64), 82.16 (3.34), respectively of the expected value (Weill This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-72 DRAFT--DO NOT CITE OR QUOTE 1 et al.. 2011), A mean FVC of 95.63 (0.76) was reported for those with other pleural 2 abnormalities versus 103.15 (0.25) in participants with no radiographic abnormalities. The 3 strongest effects of diffuse pleural thickening and/or costophrenic angle obliteration on FVC 4 were seen among men who had never smoked (-23.77,/) < 0.05), with smaller effects seen 5 among men who had smoked (-9.77,/) < 0.05) and women who had smoked (-6.73,/) < 0.05). 6 Laboratory animal and mechanistic studies of Libby Amphibole asbestos are consistent with the 7 noncancer health effects observed in both Libby workers and community members. Pleural 8 fibrosis was increased in hamsters after intrapleural injections of Libby Amphibole asbestos 9 (Smith. 1978). More recent studies have demonstrated increased collagen deposition consistent 10 with fibrosis following intratracheal instillation of Libby Amphibole asbestos fibers in mice 11 (Padilla-Carlin et al.. 2011; Sliannahan et al... 20i la; Shannahan et al... 201 lb; Smartt et al. 2010; 12 Putnam et al.. 2008). Pulmonary fibrosis, inflammation, and granulomas were observed after 13 tremolite inhalation exposure in Wistar rats (Bernstein et al.. 2005; Bernstein et al.. 2003) and 14 intratracheal instillation in albino Swiss mice (Sahu et al.. 1975). Davis et al. (1985) also 15 reported pulmonary effects after inhalation exposure in Wistar rats including increases in 16 peribronchiolar fibrosis, alveolar wall thickening, and interstitial fibrosis. 17 18 4.5.1.2. Other Nonmalignant Respiratory Diseases 19 Mortality studies of the Libby workers indicate that there is increased mortality, not only 20 from asbestosis, but other respiratory diseases. Deaths attributed to chronic obstructive 21 respiratory disease and deaths attributed to "other" nonmalignant respiratory disease were 22 elevated more than twofold (see Table 4-6) (Larson et al.. 2010b; Sullivan. 2007). These 23 diseases are consistent with asbestos toxi city, and the evidence of a positive exposure-response 24 relationship for mortality from all nonmalignant respiratory diseases, supports this association. 25 26 4.5.2. Pleural Effects 27 Pleural thickening that is caused by mineral fiber exposure includes two distinct 28 biological lesions: discrete pleural plaques in the parietal pleura and diffuse pleural thickening of 29 the visceral pleura. Both forms of pleural thickening can be viewed on standard radiographs. 30 Flowever, the two are not always clearly distinguishable on X-rays, and smaller lesions may not 31 be detected. Fligh resolution computed tomography is a method that can di stinguish between the This document is a draftfor review purposes only and does not constitute Agency policy. 4-73 DRAFT--DO NOT CITE OR QUOTE 1 lesions, as well as detect smaller lesions than are visible on X-rays. Pleural thickening may 2 restrict lung function, increase breathlessness with exercise, and contribute to chronic chest pain. 3 The potential for health effects and severity of health effects are increased with the extent and 4 thickness of the pleural lesions. 5 Data from the ATSDR community health screening study indicate that the prevalence of 6 pleural abnormalities, identified by radiographic examination, increases substantially with 7 increasing number of exposure pathways (Peipins et aU 2003). A reanalysis of these data also 8 considered age, smoking history, and types of exposures. Increased pleural thickening is 9 reported for Libby workers, those with other vermiculite work and those in "dusty trades." 10 Increased LPT is reported in both those exposed only as househole contacts or through 11 environmental exposure pathways, with greater incidence by age (38.3 and 12.7%, respectively, 12 in the 61-90 age group) (Weill et al.. 2011). DPT is reported at lower rates with 5.9 and 2.2%, 13 respectively, in these exposure groups in the highest age bracket evaluated (age 61-90). 14 Increased pleural thickening is reported for both of the studied worker cohorts, with 15 evidence of positive exposure response relationships (Larson et al., 2010a; Rohs et al.,, 2008; 16 Amandus et al., 1987b; McDonald et al... 1986b; Lockev et al.. 1984). Both McDonald et al. 17 (1986b) and Amandus et al. (1987b) indicate age is also a predictor of pleural thickening in 18 exposed individuals, which may reflect the effects of time from first exposure. Smoking data 19 were limited on the Libby workers and analyses do not indicate clear relationships between 20 smoking and pleural thickening (Amandus et al.. 1987b; McDonald et al., 1986b). Pleural 21 thickening in workers at the Scott Plant (Marysville, OH) was associated with hire on or before 22 1973 and age at time of interview but was not associated with BMI or smoking history (ever 23 smoked) (Rohs et al.. 2008). 24 25 4.5.3, Other Noncancer Health Effects (Cardiovascular Toxicity, Autoimmune Effects) 26 There is limited research available on noncancer health effects occurring outside the 27 respiratory system. Larson et al. (2010b) examined cardiovascular disease-related mortality in 28 the cohort of exposed workers from Libby (see Section 4.1.1.4.3). Mechanistic studies have 29 examined the potential role of iron and the associated inflammation for both the respiratory and 30 cardiovascular disease (Shannahan et aL 2011b). Two studies examined the association between 31 asbestos exposure and autoimmune disease (Noonan et al.. 2006) or autoantiboides and other This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-74 DRAFT--DO NOT CITE OR QUOTE 1 immune markers (Pfau et al., 2005) (see Table 4-17). Limitations in the number, scope, and 2 design of these studies make it difficult to reach conclusions as to the role of asbestos exposure 3 in either cardiovascular disease or autoimmune disease. 4 5 4.5,4. Libby Amphibole Asbestos Summary of Noncancer Health Effects 6 The studies in humans summarized in Section 4.1 have documented an increase in 7 mortality from nonmalignant respiratory disease, including asbestosis, in workers exposed to 8 Libby Amphibole asbestos (Larson et al.. 2010b; Sullivan. 2007; McDonald et al.. 2004; 9 Amandus and Wheeler. 1987). Radiographic evidence of pleural thickening and interstitial 10 damage (small opacities) are also well documented among employees of the Libby vermiculite 11 mining operations (Larson et al.. 2010a; Amandus et a l 1987b; McDonald et al.. 1986b). 12 Additional studies have documented an increase in radiographic changes in the pleura and 13 parenchyma among employees of a manufacturing facility in Marysville, OH that used Libby 14 vermiculite ore contaminated with Libby Amphibole asbestos (Rohs et al.. 2008; Lockev et al.. 15 1984). Positive exposure-response relationships for these health effects for both occupational 16 cohorts studied, as well as the observed latency, support an association between exposure to 17 Libby Amphibole asbestos and these pieuro-pulmonary effects. Studies of community members 18 exposed to Libby Amphibole asbestos have documented similar pleural abnormalities and 19 pulmonary deficits consistent with parenchymal damage (Weill et al.. 2011; Whitehouse, 2004; 20 Peipins et al.. 2003). Although limited, animal studies support the toxicity of Libby Amphibole 21 asbestos to pleural and pulmonary tissues. Developing research supports a role of inflammatory 22 processes in the toxic action of Libby Amphibole asbestos, consistent with the observed health 23 effects (Duncan et al.. 2010; Hamilton et al.. 2004). Taken together, the strong evidence in 24 human studies, defined exposure response relationships, and supportive animal studies provide 25 compelling evidence that exposure to Libby Amphibole asbestos causes nonmalignant 26 respiratory disease, including asbestosis, pleural thickening, and deficits in pulmonary function 27 associated with mineral fiber exposures. Existing data regarding cardiovascular effects and the 28 potential for autoimmune di sease are limi ted. 29 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-75 DRAFT--DO NOT CITE OR QUOTE 1 4.5.5. Mode-of-Action Information (Noncancer) 2 The precise mechanisms causing toxic injury from inhalation exposure to Libby 3 Amphibole asbestos have not been established. However, nearly all-durable mineral fibers with 4 dimensional characteristics that allow penetration to the terminal bronchioles and alveoli of the 5 lung have the capacity to induce pathologic response in the lung and pleural cavity (ATSDR. 6 2001a; Witschi and Last 1996). The physical-chemical attributes of mineral fibers are important 7 in determining the type of toxicity observed. Fiber dimension (width and length), density, and 8 other characteristics such as chemical composition, surface area, solubility in physiological 9 fluids, and durability all play important roles in both the type of toxicity observed and the 10 biologically significant dose. Fibrosis results from a sequence of events following lung injury, 11 which includes inflammatory cell migration, edema, cellular proliferation, and accumulation of 12 collagen. Fibers do migrate to the pleural space, and it has been hypothesized that a similar 13 cascade of inflammatory events may contribute to fibrotic lesions in the visceral pleura. 14 Thickening of the visceral pleura is more often locali zed to lobes of the lung with pronounced 15 parenchymal changes, and it has also been hypothesized that the inflammatory and fibiogenic 16 processes within the lung parenchyma in response to asbestos fibers may influence the fibrogenic 17 process in the visceral pleura. The etiology of parietal plaques is largely unknown with respect 18 to mineral fiber exposure. 19 There is currently insufficient evidence to establish the noncancer mode of action for 20 Libby Amphibole asbestos. Limited in vitro studies have demonstrated oxidative stress 21 following Libby Amphibole asbestos exposures in various cell types (Duncan et al.. 2010; 22 Hillegass et al.. 2010; Pietruska et al.. 2010; Blake et al,, 2007). Libby Amphibole asbestos 23 fibers increased intracellular ROS in both murine macrophages and human epithelial cells 24 (Duncan et al.. 2010; Blake et al.. 2007). Surface iron, inflammatory marker gene expression 25 was increased following exposure to Libby Amphibole asbestos in human epithelial cells 26 (Shannahan et al.. 201 lb ; Duncan et al. 2010; Pietruska et al.. 2010) (see Table 4-18). 27 Tremolite studies demonstrate cytotoxicity in various cell culture systems (see Table 4-19). 28 The initial stages of any fibrotic response involve cellular proliferation, which may be 29 compensatory for cell death due to cytotoxicity. Analysis of cellular proliferation has 30 demonstrated both increases and decreases following exposure to asbestos fibers in vitro and in 31 vivo depending on the specific fiber or cell type (Mossman et al.. 1985; Topping and Nettesheim. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-76 DRAFT--DO NOT CITE OR QUOTE 1 1980). Other studies have focused on the activation of cell-signaling pathways that lead to 2 cellular proliferation following exposure to asbestos (Scapoli et al.. 2004; Shukla et al... 2003; 3 Ding et al., 1999; Zanella et al., 1996). 4 Although slightly increased compared to controls, cytotoxicity in murine macrophage 5 cells exposed to Libby Amphibole asbestos was decreased compared to other fiber types (Blake 6 et al.. 2008). Cytotoxicity was slightly, but statistically significantly, increased compared to an 7 unexposed control at 24 hours post exposure to Libby Amphibole asbestos, while crocidolite 8 exposure resulted in even higher levels of cytotoxicity. No other in vitro study examined 9 cytotoxicity following exposure to Libby Amphibole asbestos, although an increase in apoptosis 10 was demonstrated in this same cell system (Blake et al.. 2008). Recent studies in mice exposed 11 to Libby Amphibole asbestos demonstrated increased collagen deposition and collagen gene 12 expression, markers of fibrosis (Smartt et al.. 2010; Putnam et al.. 2008). Short-term studies in 13 rats also demonstrated an increased inflammatory response (Padilla-Carlin et al.. 2011; 14 Shannahaa et al.. 2011a; Shannahan et aL 2011b). Tremolite and Libby Amphibole asbestos 15 exposure led to increases in both fibrosis in all but one animal study, supporting a role for 16 proliferation in response to these fibers. Taken together with studies on other asbestos fibers, 17 these data suggest that a cytotoxicity and cell proliferation may play a role in the noncancer 18 health effects following exposure to Libby Amphibole asbestos. 19 Although continued research demonstrates that the Libby Amphibole asbestos has 20 biologic activity consistent with the inflammatory action and cytotoxic effects seen with other 21 forms of asbestos, the data are not sufficient to establi sh a mode of action for the 22 pleura-pulmonary effects of exposure to Libby Amphibole asbestos. 23 24 4.6. EVALUATION OF CARCINOGENICITY 25 4.6.1. Summary of Overall Weight of Evidence 26 Under the EPA Guidelinesfor Carcinogen Risk Assessment (U.S. EPA. 2005a). Libby 27 Amphibole asbestos is carcinogenic to humans following inhalation exposure based on 28 epidemiologic evidence that shows a convincing association between exposure to Libby 29 Amphibole asbestos fibers and increased lung cancer and mesothelioma mortality (Larson et al.. 30 2010b: Moolgavkar et al.. 2010: Sullivan, 2007: McDonald et al, 2004: Amandus and Wheeler. 31 1987: McDonald et al.. 1986a). These results are further supported by animal studies that This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-77 DRAFT--DO NOT CITE OR QUOTE 1 demonstrate the carcinogenic potential of Libby Amphibole asbestos fibers and tremolite fibers 2 in rodent bioassays. As a durable mineral fiber of respirable size, this conclusion is consistent 3 with the extensive published literature that documents the carcinogenicity of amphibole fibers. 4 US. EPA's Guidelinesfor Carcinogenic Risk Assessment (US. EPA. 2005a) indicate 5 that for tumors occurring at a site other than the initial point of contact, the weight of evidence 6 for carcinogenic potential may apply to all routes of exposure that have not been adequately 7 tested at sufficient doses. An exception occurs when there is convincing information (e.g,, 8 toxicokinetic data) that absorption does not occur by other routes. Information on the 9 carcinogenic effects of Libby Amphibole asbestos via the oral and dermal routes in humans or 10 animals is absent. The increased risk of lung cancer and mesothelioma following inhalation 11 exposure to Libby Amphibole asbestos has been established by studies in humans, but these 12 studies do not provide a basis for determining the risk from other routes of exposure. 13 Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not considered a 14 portal-of-entry effect. However, the role of indirect or direct interaction of asbestos fibers in 15 disease at these extrapulmonary sites is still unknown. There is no infoimation on the 16 translocation of Libby Amphibole asbestos to extrapulmonary tissues following either oral or 17 dermal exposure, and limited studies have examined the role of these routes of exposure in 18 cancer. Therefore, Libby Amphibole asbestos is considered carcinogenic to humans by the 19 inhalation route of exposure. 20 21 4.6.1.1. Synthesis o f Human, Animal, and Other Supporting Evidence 22 Libby, MT workers have been the subject of multiple mortality studies demonstrating an 23 increased cancer mortality in relation to estimated fiber exposure. Occupational studies 24 conducted in the 1980s (Amandus and Wheeler. 1987) (McDonald et af. 1986a) as well as the 25 extended follow-up studies published in more recent years (Larson et aL 2010b; Sullivan. 2007; 26 McDonald et al.. 2004) and additional analyses of the extended follow-up (Moolgavkar et al.. 27 2010) provide evidence of an increased risk of lung-cancer mortality and of mesotheli oma 28 mortality among the workers exposed to Libby Amphibole asbestos in the Libby vermiculite 29 mining and processing operations. This pattern is seen in the lung cancer analyses using an 30 internal referent group in the larger follow-up studies (Larson et al.. 2010b; Sullivan. 2007; 31 McDonald et al.. 2004). with cumulative exposure analyzed using quartiles or as a continuous This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-78 DRAFT--DO NOT CITE OR QUOTE 1 measure, and in the studies reporting analyses using an external referent group [i.e., standardized 2 mortality ratios (Sullivan. 2007; Amandus and Wheeler. 1987; McDonald et ah. 1986a). 3 McDonald et al. (2004) also reported increasing risk of mesothelioma across categories of 4 exposure; the more limited number of cases available in earlier studies precluded this type of 5 exposure-response analysis. This association is also supported by the case series of 6 11 mesothelioma patients among residents in or around Libby, MT, and among family members 7 of workers in the mining operations (Whitehouse et al., 2008). 8 Although experimental data in animals and data on toxicity mechanisms are limited for 9 Libby Amphibole asbestos, tumors were observed in tissues similar to those in humans (e.g., 10 mesotheliomas, lung cancer) indicating the existing data are consistent with the cancer effects 11 observed in humans exposed to Libby Amphibole asbestos. Smith (1978) reported increased 12 incidence of mesotheliomas in hamsters after intrapleural injections of Libby Amphibole 13 asbestos. Additionally, studies in laboratory animals (rats and hamsters) exposed to tremolite via 14 inhalation (Bernstein et al.. 2005; Bernstein et al.. 2003; Davis et al. 1985). intrapleural injection 15 (Roller et al. 1997. 1996; Davis et al.. 1991; Wagner et al.. 1982; Smith et al.. 1979) or 16 implantation (Stanton et al.. 1981) have shown increases in mesotheliomas and lung cancers. 17 Tremolite from various sources was used and varied in fiber content and in potency (see 18 Section 4.2, Appendix D). Although McConnell et al. (1983b) observed no increase in 19 carcinogenicity following oral exposure to nonfibrous tremolite, the ability of this study to 20 inform the carcinogenic potenti al of fibrous tremolite through inhalation is unclear, and these 21 study results contribute little weight to the evaluation of the carcinogenicity of fibrous Libby 22 Amphibole asbestos. 23 The available mechanistic information suggests Libby Amphibole asbestos induces 24 effects that may play a role in carcinogenicity (see Section 4.3.4, Appendix D). Several in vitro 25 studies have demonstrated oxidative stress and genotoxicity following Libby Amphibole 26 asbestos exposures in various cell types (Duncan et al. 2010; Hillegass et al.. 2010; Pietruska et 27 al. 2010; Blake et al. 2007). Libby Amphibole asbestos increased intracellular ROS in both 28 murine macrophages and human epithelial cells (Duncan et al. 2010; Blake et al.. 2007). 29 Additionally, surface iron, inflammatory marker gene expression and aneugenic micronuclei 30 were increased following exposure to Libby Amphibole asbestos in human epithelial cells 31 (Duncan et al, 2010; Pietruska et al. 2010). Tremolite studies demonstrate cytotoxic and This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-79 DRAFT--DO NOT CITE OR QUOTE 1 clastogenic effects (e.g., micronucleus induction and chromosomal aberrations) of the fibers in 2 various cell culture systems. 3 In summary, the epidemiologic data demonstrate an association between exposure to 4 Libby Amphibole asbestos and increased cancer risk. Supporting evidence of carcinogenic 5 potential was observed in the limited number of laboratory animal studies exposed to Libby 6 Amphibole asbestos or tremolite (see Tables 4-15 and 4-16 summarizing in vivo studies). 7 Overall, the available evidence supports the conclusion that Libby Amphibole asbestos is 8 carcinogenic to humans. 9 10 4.6.2. Mode-of-Action Information 11 4.6.2.1. Description o f the Mode-of-Action Information 12 EPA guidance provides a framework for analyzing the potential mode(s) of action by 13 which physical, chemical, and biological information is evaluated to identify key events in an 14 agent's carcinogenicity (II.S. EPA. 2005a). Agents can work through more than one mode of 15 action (MOA), and MOA can differ for various endpoints (e.g., lung cancer versus 16 mesothelioma). Reasonably, the analysis of a MOA would start with some knowledge of an 17 agent's biological activity that leads to cellular transformation resulting in carcinogenicity. 18 Although early steps in the process often can be identified, carcinogenicity is a complex process 19 resulting from multiple changes in cell function. Due to the limited data available specific to 20 Libby Amphibole asbestos, the mode of action of Libby Amphibole asbestos for lung cancer and 21 mesothelioma following inhalation exposure cannot be established. 22 Research on vari ous types of mineral fibers supports the role of multiple biologic 23 responses following exposure to asbestos in general (i.e., chronic inflammation, generation of 24 ROS, direct genotoxicity, and cytotoxicity and cellular proliferation) in the carcinogenic 25 response to mineral fibers. However, the complexiti es of fiber toxicity make it diffi cult to define 26 modes of action for asbestos, in general [as reviewed in Aust et al. (2011); Mossman et al. 27 (2011); Huang et al. (2011); Bunderson-Schelvan et al. (2011); Broaddus et al. (2011)1. Further, 28 limitations in early study design and presentation of the results hinder understanding of mode 29 and mechanism of action for specific fiber types. Most studies lack information on the 30 characterizati on of fibers and cell types used, hindering understanding of the mode(s) of acti on. 31 Particularly of importance is the route of exposure utilized in the in vivo studies, as results This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-80 DRAFT--DO NOT CITE OR QUOTE 1 obtained from nonphysiologically relevant routes of exposure (i.e., intraperitoneal injection, 2 gelatin implant) may not accurately reflect the response in occupational inhalation exposures. 3 Occupational studies demonstrate human health effects (e.g., lung cancer, mesothelioma) 4 following exposure to Libby Amphibole asbestos. Although the limited mechanistic data 5 demonstrate biological effects similar to those of other mineral fibers following exposure to 6 Libby Amphibole asbestos, the existing literature are insufficient to establish a mode of action 7 for Libby Amphibole asbestos for lung cancer or mesothelioma. These biological effects 8 following exposure to Libby Amphibole asbestos and/or tremolite are demonstrated in a limited 9 number of laboratory animal and in vitro studies. Multiple key events for one particular MOA 10 have not been identified; therefore, the mode of action for Libby Amphibole asbestos 11 carcinogenicity cannot be established. 12 13 4.6.2,2. Application o f the Age-Dependent Adjustment Factors 14 As described above, the mode of action for Libby Amphibole asbestos is unknown . The 15 weight of evidence does not support a mutagenic mode of action for Libby Amphibole asbestos 16 carcinogenicity. Therefore, according to EPA's Supplemental Guidancefor Assessing 17 Susceptibility from Early-Life Exposure to Carcinogens (TJ.S. EPA. 2005b). the application of 18 the Age-Dependent Adjustment Factors is not recommended. 19 20 4.7. SUSCEPTIBLE POPULATIONS 21 Certain populations may be more susceptible to adverse health effects from exposure to 22 Libby Amphibole asbestos. Because the adverse health effects resulting from exposure to Libby 23 Amphibole asbestos have been, for the most part, studied in occupational cohorts of adult white 24 men (see Sections 4.1.1 and 4.1.3), there is limited information on the effects to a broader 25 population. A few studies, however, have examined health effects resulting from 26 nonoccupational exposure in other age groups, in other genders (i.e., females), and in different 27 race or ethnicity groups. The data from these studies could inform whether any differential risk 28 exists for these groups (see Sections 4.1.2 and 4.1.4). Flowever, it should be noted that the 29 ability to distinguish true differences from chance variation in effect estimates is related to the 30 sample size and statistical power, which, in most cases, is quite limited in these studies. In 31 addition, genetic polymorphisms, preexisting health conditions, and differences in nutritional This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-81 DRAFT--DO NOT CITE OR QUOTE 1 status may alter an individual's response to Libby Amphibole asbestos. Finally, coexposures to 2 other substances (e.g., tobacco smoke or particulate matter) may increase an individual's risk of 3 adverse health effects from exposure to Libby Amphibole asbestos. Where data are available, 4 each of these factors is discussed below with respect to increased susceptibility to noncancer 5 effects and cancer from exposure to Libby Amphibole asbestos, and where information specific 6 to Libby Amphibole asbestos is not available, the general literature on the toxicity of mineral 7 fibers is briefly referenced. 8 There are also factors that may influence one's exposure potential to asbestos based on 9 lifestage or other defined population. For example, children spend more hours outside and may 10 engage in activities which impact exposure level compared to adults (U.S. EPA. 2006b; NRC. 11 1993). Because lifestage and activity patterns can increase the potential for health effects from 12 exposure, these factors define those who may be more susceptible to health effects due to greater 13 exposure. Section 2.3 discusses this exposure potential, including how children workers, 14 household contacts and residents may be exposed to Libby Amphibole asbestos. 15 16 4.7.1. Influence of Different Lifestages on Susceptibility 17 Individuals at different lifestages differ from one another physiologically, anatomically, 18 and biochemically. Individuals in early and later lifestages differ markedly from adulthood in 19 terms of body composition, organ function, and many other physiological parameters, which can 20 influence the toxicokinetics and toxicodynamics of chemicals and their metabolites in the body 21 (Guzelian et aL 1992). This also holds true for mineral fibers, including asbestos fibers (see 22 Section 3). This section presents and evaluates the literature on how individuals in early or later 23 lifestages might respond differently and thus potentially be more susceptible to adverse health 24 effects of Libby Amphibole asbestos exposure. 25 26 4.7.1.1. Lifestage Susceptibility 27 Humans in early lifestages (i.e., conception through adolescence) can have unique 28 susceptibilities compared to those in later lifestages because they undergo rapid physiological 29 changes during critical periods of development (Selevan et aL 2000). Furthermore, they are 30 often exposed to xenobiotics via unique exposure pathways (i.e., transplacental transfer and 31 breast milk ingestion) (U.S. EPA. 2006b; NRC. 1993). Although no data exist for Libby This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-82 DRAFT--DO NOT CITE OR QUOTE 1 Amphibole asbestos, limited observations in stillborn infants indicate occurrence of 2 transplacental transfer of tremolite (Hague et al.. 1998; 1996) and other asbestos and nonasbestos 3 fibers (Hague et al.. 1998; Hague et al.. 1996; Hague et al.. 1992; Hague et al.. 1991). Haque et 4 al. (1992) hypothesized that maternal health conditions might influence the translocation of 5 fibers, as some of the mothers had preexisting health conditions. Transplacental transfer of 6 asbestos also has been demonstrated in animals following maternal exposure by gavage (Haque 7 et al.. 2001) or injection (Hague and Vrazeh 1998; Cunningham and Pontefract 1974) (see 8 Section 3). These studies did not evaluate sources or levels of exposure, and injection studies are 9 a less relevant route of exposure than inhalation. Based on these studies, Libby Amphibole 10 asbestos fibers may be transferred through the placenta, resulting in prenatal exposure at any 11 stage of fetal development. 12 Increased lung deposition of fibers in children compared with adults has been observed 13 (Bennett et al.. 2008; Isaacs and Martonen. 2005; Asgharian et al.. 2004; Phalen and Oldham., 14 2001; Oldham et al.. 1997; Schiller-Scotland et al.. 1994; Phalen et al.. 1985). Nasal deposition 15 of particles was shown to be lower in children compared to adults--particularly during exercise 16 (Becguemin et al.. 1991). The lung and nasal deposit!onal differences are due in part to 17 structural differences across lifestages, which can change the depositional pattern of different 18 fiber sizes and possibly alter the site of action and result in differential clearance and subsequent 19 health effects. It is unclear, however, whether the lung surface, body weight, inhalation volume, 20 or exposure patterns are most determinative of dose. One study reported that the ratio of lung 21 surface area to body weight does not differ considerably for a 10-month old, a 9-year old, and an 22 adult (Short. 1952). Another study suggested that deposition of fine particles (2-pm mass 23 median aerodynamic diameter, which is in the size range of those for Libby Amphibole asbestos 24 reported in Table 2-2) in the lung is increased for overweight (>95tn percentile BMI) children 25 who breathe more at rest compared to underweight children (<25fhpercentile BMI) (Bennett and 26 Zeman. 2004). 27 There are few studies analyzing noncancer outcomes in children exposed to Libby 28 Amphibole. A Libby medical screening program collected data on 7,307 participants, including 29 600 children aged 10-17 years old, representing 8.2% of the cohort (Peipins et al.. 2003). 30 Pulmonary function tests showed that none of these children had moderate or severely restricted 31 lung function (ATSDR. 2002. 2001b). This study also studied chest radiographs for those This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-83 DRAFT--DO NOT CITE OR QUOTE 1 18 years old or older (Noonan et al., 2006; Peipins et al., 2003; ATSDR, 2001b), but X-rays were 2 not conducted on children. In addition, the prevalence of some self-reported respiratory 3 symptoms among 10-29-year-old adolescents and young adults was associated with certain 4 exposure pathways. These participants were < age 18 in 1990 when the mining/milling 5 operations closed (Vinikoor et al., 20101. Understanding of the community health effects and the 6 examination of the potential progression of adverse health effect in this community would 7 benefit from additional research to establish the clinical significance of these findings. No other 8 studies of noncancer outcomes in early lifestages of humans or experimental animals exposed to 9 Libby Amphibole asbestos have been reported. 10 For exposure to other types of asbestos, studies have reported noncancer outcomes in 11 early lifestages. Those in the very young include reports of stillbirth (Hague et al., 1998; 19961 12 and death among infants (age 1-27 months) due to sudden infant death syndrome and 13 bronchopulmonary dysplasia (Hague and Kanz, 1988). These studies found higher levels of 14 asbestos in the lungs of those who died compared to controls. In the infant study, the authors 15 speculate that either there was a preexisting abnormal lung physiology in these children that may 16 contribute to a reduced ability to clear fibers from the lung, or that the children could have an 17 increased exposure to asbestos (Hague and Kanz, 1988). Those in older children include reports 18 of pleural and diaphragmatic calcifications (Epleretal., 1980) and altered immune and 19 respiratory conditions (Shtof et al.. 2000). 20 In experimental animals, offspring of rats exposed to tremolite had decreased body 21 weight gain at weaning and 8-weeks-old compared to controls (NTP. 1990b; McConnell et al.. 22 1983b). This was also observed in some similar studies of other forms of asbestos (NTP, 1990a, 23 1988, 1985; McConnell et al., 1983b) but not in others (McConnell et al., 1983a; NTP, 1983). 24 Embryonic toxicity was observed in a few experimental animal studies. Crocidolite injected into 25 pregnant mice resulted in altered limb differentiation in cultured embryos [Krowke et al. (1983). 26 abstract], and chrysotile in drinking water given to pregnant mice resulted in decreased 27 postimplantation survival in cultured embryos (Schneider and Maurer. 1977); however, pregnant 28 mice exposed to chrysotile in drinking water did not affect in vivo embryonic survival 29 (Schneider and Maurer. 1977). 30 It is possible that early lifestage exposure may increase the risk of noncancer outcomes in 31 adulthood compared to adult exposure. After tremolite exposure during childhood, one study This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-84 DRAFT--DO NOT CITE OR QUOTE 1 reported altered immunity in adulthood (Zerva et al., 1989), and one study described a case 2 report of asbestosis in adulthood (Voisin et al.. 1994). Another study also reported an increased 3 risk of asbestosis after childhood exposure to asbestos from parental occupational exposure to 4 asbestos (Kilburn et al.. 1985). To address the potential for increased susceptibility to cancer 5 from early lifetime exposures, one needs to consider if there is evidence of differential health 6 effects such as increased potency from early lifetime exposure, decreased latency based on the 7 age of exposure, or cancers observed with early lifetime exposures not seen with adult exposures. 8 There are no published reports that can directly answer these questions for exposure to Libby 9 Amphibole asbestos. 10 While cancers in adults have been documented following exposure to Libby Amphibole 11 asbestos, similar reports describing childhood cancers resulting from this exposure have not been 12 identified. Few cancers occurring in children have been documented in children exposed to any 13 form of asbestos. Examples of cases include a 17-year old exposed to chrysotile and tremolite 14 (Andrion et ah, 1994) and a 3-year old exposed to chrysotile (Lieben and Pistawka. 1967). both 15 of whom developed mesothelioma. ITowever, childhood mesothelioma, in particular, may have 16 an etiology that is different from that of the di sease that is seen in adults (Cooper et aL 1989). 17 No cancer bioassays have been performed in j uvenile animals exposed to Libby Amphibole 18 asbestos. 19 Of the 11 Libby Amphibole asbestos-related mesothelioma cases described by 20 Whitehouse et al. (2008), 2 reported potential exposure scenarios that were limited to childhood, 21 and both of these were diagnosed at a relatively young age at diagnosis (48, compared with 52 to 22 82 years of age for the other nine cases). Although these case studies support the link between 23 exposure to Libby Amphibole asbestos and mesothelioma, it is unclear if children are more 24 susceptible than adults. 25 Case reports of exposure to tremolite during childhood, and subsequent diagnosi s of 26 mesothelioma in adulthood (Senyigit et al.. 2000; Schneider et al. 1998; Sakellariou et al.. 1996; 27 Rev et al.. 1993; Magee et al.. 1986). support the limited data summarized above for Libby 28 Amphibole asbestos. Additional case studies of mesothelioma after childhood exposure to other 29 types of asbestos are available (Yano et al.. 2009; Ascoli et al.. 2003; Magnani et aL 2001; Rom 30 et al. 2001; Schneider et al... 1996a; Schneider et al., i996b; Schneider et al.. 1995; Roguin et al.. 31 1994; Cazzadori et at... 1992; Inase et al. 1991; K aneetal. 1990; 1,1 et al... 1989; Martensson et This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-85 DRAFT--DO NOT CITE OR QUOTE 1 al., 1984; Wassermann et al.., 1980; Li et al., 1978; Anderson et a!., 1976; Wagner et al., 1960). 2 These studies, however, do not clarify whether exposure during childhood yields different 3 adverse health effects compared with exposure during adulthood. 4 In experimental studies, the offspring of rats orally exposed to nonfibrous tremolite did 5 not demonstrate an increase in tumors compared to controls (NTP. 1990b; McConnell et al.. 6 1983b). Similar studies of other forms of asbestos did report an increase of various neoplasms in 7 the offspring (NTP, 1990a. 1988. 1985; McConnell et al,. 1983a; McConnell et al.. 1983b). but 8 another study reported none (NTP. 1983). 9 Studies of exposure to other types of asbestos have attempted to determine if exposure to 10 asbestos in early life results in an increased risk of developing cancer. An early study in the 11 United Kingdom described occupational exposure to chrysotile, crocidolite, and amosite for a 12 group of 900 women. First exposure from ages 15-24 years led to a higher relative mortality 13 risk for lung and pleural cancer compared with women who were first exposed at older ages 14 (SMR 30 based on 12 observed and 0.4 expected, SMR 8 based on 4 observed and 0.5 expected, 15 and SMR 6.7 based on 6 observed and 0.9 expected in the first exposure at ages 15-24, 25-34, 16 and >35 years, respectively) (Newhouse et al.. 1972). A study by Hansen et al. (1998) in 17 Wittenoom, Western Australia examined 27 individuals diagnosed with mesothelioma who had 18 been environmentally exposed to crocidolite (i.e., residents of the town but not directly employed 19 in the area's crocidolite mining and milling industry); 11 of these subjects were children 20 <15 years old at the time of exposure. One-third of all the subjects were less than 40 years old 21 when diagnosed, but the authors found no increase in mesothelioma mortality rates when 22 analyzed by age at first exposure. However, risk was significantly increased based on time from 23 first exposure, duration of exposure, and cumulative exposure (Hansen et al.. 1998). Additional 24 studies of this cohort found that the mesothelioma mortality rate was lower for those first 25 exposed (based on age residence in the area began) to crocidoli te at ages <15 years (n = 24; 26 mesothelioma mortality rate 47 per 100,000 person-years) compared with those first exposed at 27 ages >15 years (n = 43; mesothelioma mortality rate 112 per 100,000 person-years) (Reid et al.. 28 2007). The hazard ratio for age at first residential exposure of >15 years compared with 29 <15 years was 3.83 (95% Cl: 2.19, 6.71), adjusting for cumulative exposure, gender, and an 30 interaction term for gender and cumulative exposure. This document is a draftfor review purposes only and does not constitute Agency policy. 4-86 DRAFT--DO NOT CITE OR QUOTE 1 Based on these very limited and inconclusive studies on other forms of asbestos, no 2 conclusions can be drawn about differential risk of adverse health effects after early lifestage 3 exposure to Libby Amphibole asbestos compared to exposure during adulthood. It is unknown 4 whether early lifestage exposure compared to adult exposure increases susceptibility for adult 5 cancers, as measured by increased incidence, severity, or disease progression, or by decreased 6 latency. 7 Later lifestage is generally defined as >65 years old. Because pulmonary function 8 (volume and rate of breathing) decreases with age (Weiss. 2010). increased deposition of fibers 9 in the lung from exposures in later lifestages is unlikely. Clearance of fibers from the lung might 10 be reduced, however, as older adults have a less effective cough reflex and strength and the cilia 11 are less able to move mucus up and out of the airway (U.S. EPA. 2006a). Older adults could be 12 more susceptible to the effects of Libby Amphibole asbestos due to the gradual age-related 13 decline in physiological processes. Additionally, decreased immune function, increased genetic 14 damage, and decreased DNA repair capacity can result in increased susceptibility with age (U.S. 15 EPA. 2006a). These age-associated alterations could decrease fiber-induced DNA damage repair 16 but might also reduce the incidence of fiber-induced DNA damage due to decreased 17 phagocytosis or inflammation. Specific data pertaining to age-varying effects of Libby 18 Amphibole asbestos on these processes are not available. 19 Because the risk of many types of noncancer effects increases with age, an increasing rate 20 of specific diseases with increasing age can be expected among individuals exposed at some 21 point in their lives to Libby Amphibole asbestos. Radiographic tests among those exposed to 22 Libby Amphibole show that older age, which may be highly correlated with time since first 23 exposure in some occupational settings, is one of the factors most associated with pleural or 24 interstitial abnormalities (Rohs et al., 2008; Horton et aL 2006; Muravov et ah. 2005; Pei pins et 25 al., 2003; ATSDR. 2001b; Amandus et al... 1987b; McDonald et al... 1986b; Lockev et al.. 1984). 26 Abnormal radiographs also increase with age in general population studies (Pinsky et al.. 2006). 27 In the community health screening study, an increased risk of rheumatoid arthritis among 28 individuals ages >65 years was observed in relation to several measures reflecting exposure to 29 Libby Amphibole asbestos (e.g., worked for W.R. Grace, used vermiculite for gardening) 30 (Noonan. 2006). However, the available studies do not provide a basis for evaluating the timing 31 of the exposure in relation to these outcomes. No conclusions can be drawn about differential This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-87 DRAFT--DO NOT CITE OR QUOTE 1 risk of noncancer after later lifestage exposure to Libby Amphibole compared to exposure earlier 2 in life. 3 No studies assessing the carcinogenic effect of exposures occurring in older age groups 4 are available for Libby Amphibole asbestos. It should be noted that observed health effects 5 among individuals exposed to Libby Amphibole asbestos are likely to increase with increasing 6 age due to the long latency period for the exposure response for asbestos and lung cancer and 7 other chronic diseases. However this type of observation would not directly address the questi on 8 of whether exposures at older ages have a stronger or weaker effect compared with exposures at 9 younger ages. 10 11 4.7.2. Influence of Gender on Susceptibility 12 A discussion of gender-related differences in risk from asbestos exposure raises several 13 important issues, such as gender-related differences in exposure patterns, physiology, and 14 dose-response (Smith. 20021. For example, nasal breathing fdters out particles, and men tend to 15 breathe less through their nose during exercise than women do (Bennett et ah. 20031. Bennett 16 et al. (19961 showed a gender difference in fractional deposition (defined as the ratio of particles 17 not exhaled to total particles inhaled) of particles 2 pm in mass median aerodynamic diameter. 18 This particle diameter is within the range of Libby Amphibole asbestos particles reported in 19 Table 2-2. This study found that, in general, women had a greater retenti on of particles 20 compared to men because men had higher ventilation rates compared to women; however, the 21 overall deposition rate was higher in the men (Bennett et al.. 1996). 22 Most occupational studies for Libby Amphibole asbestos have examined the effects of 23 exposure only in men (Moolgavkar et al.. 2010; Sullivan. 2007; McDonald et al... 2004; 24 Amandus et al.. 1988; Amandus et al., 1987a; Amandus and Wheeler. 1987; McDonald et al.. 25 1986a; McDonald et al.. 1986b). There is limited information specifically on women exposed to 26 Libby Amphibole asbestos. In the Libby, MT community studies, no gender-related trends in 27 mortality due to lung or digestive cancer were observed (ATSDR. 2000). These limited data do 28 not provide a basis for drawing conclusions regarding gender-related differences in adverse 29 health effects from Libby Amphibole asbestos. 30 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-88 DRAFT--DO NOT CITE OR QUOTE 1 4.7.3. Influence of Race or Ethnicity on Susceptibility 2 Race and ethnicity often are used in medical and epidemiological studies to define 3 various groups of the population. These categories could be surrogates for differences in 4 exposure (e.g., occupation, socioeconomics, behavior) or biology (e.g., physiology, genetics), in 5 which case these factors may play a role in susceptibility as well. Nasal structure and lung 6 architecture can influence the depositional patterns for both particles and fibers. One study of 7 18 Caucasians (ages 8 to 30 years) and 14 African Americans (ages 8 to 25 years) reported 8 increased ventilation rates during exercise in the African Americans (matched on sex, age, 9 height, and weight) (Cerny, 1987). Another study (11 Caucasians and 11 African Americans, 10 ages 18 to 31 years) reported decreased nasal deposition efficiency (for particle sizes of 1-2 pm, 11 which is in the range of those for Libby Amphibole asbestos reported in Table 2-2) in African 12 Americans compared to Caucasians (Bennett and Zeman, 2005). Furthermore, nasal breathing 13 during exercise occurred less in Caucasians compared to African Americans in this study 14 (Bennett et al., 2003). 15 Of the occupational and residential studies for Libby Amphibole asbestos, the vast 16 majority of subjects with known race were white, precluding the ability to conduct an analysis of 17 racial and ethnicity-related differences in the mortality risks within the Libby worker cohort. In 18 a study of occupati onal exposure to chrysotile asbestos in a textile factor, lung-cancer mortality 19 risk in relation to exposure was lower in nonwhite males (0.84, 95% CI: 0.52-1.27) compared to 20 white males (2.34, 95% CL 1.94-2.79), although a statistically significant increase in SMR was 21 observed for nonwhite males at high exposure levels (>120 fiber-years/mL) (Hein et al.. 2007). 22 This observed difference could be due to a lower prevalence of smoking among nonwhite 23 compared with white males (Hein et al.. 2007). 24 25 4.7.4. Influence of Genetic Polymorphisms on Susceptibility 26 XRCC1 is a DNA damage repair gene. A recent study demonstrated that 27 XRCC1-deficient cells exposed to Libby Amphibole or crocidolite asbestos demonstrated 28 increased levels of micronuclei induction (Pietruska et al., 2010). Two other studies examined 29 XRCC1 polymorphisms in relation to disease risk with other types of asbestos exposure. Zhao 30 et al. (2006) found no association between XRCC1 polymorphisms and asbestosis in 31 asbestos-exposed workers. A study by Dianzani et al. (2006). however, did find an association This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-89 DRAFT--DO NOT CITE OR QUOTE 1 between XRCC1 and asbestos-induced lung disease in a population exposed to asbestos 2 pollution. Further work is necessary, with clear definitions of patient populations and their 3 exposure levels, so that these studies and others can be compared to determine if XRCC1 4 polymorphisms increase susceptibility to adverse health effects following exposure to Libby 5 Amphibole asbestos. 6 SODs are free radical scavengers that dismutate superoxide anion to oxygen and 7 hydrogen peroxide. SODs are expressed in most cell types exposed to oxygen. Several common 8 forms of SODs occur and are named by the protein cofactor: copper/zinc, manganese, iron, or 9 nickel. A recent study observed no significant alterations in levels of intracellular SOD 10 following a 3 hour exposure to Libby Amphibole asbestos in mice (Blake et al,, 2007). Other 11 studies in humans and mice have examined SOD expression in relation to other types of asbestos 12 exposure. Manganese superoxide dismutase activity was elevated in biopsies of human 13 asbestos-associated malignant mesothelioma, although no genotypic differences were found to 14 be related to this change in activity (Hirvonen et al., 2002). Other studies have focused on the 15 role of extracellular superoxide dismutase (EcSOD) and asbestos-induced pulmonary disease 16 (Kliment et al., 2009: Gao et al, 2008: Fattman et al.. 2006: Tan et al.. 2004). These studies 17 have suggested a protective effect of EcSOD, with mice that lack this form of SOD having 18 increased sensitivity to asbestos-induced lung injury (Fattman et al,, 2006). Familial studies 19 showing unusually high incidence of mesothelioma suggest that genetic factors might play a role 20 in the etiology of mesothelioma (Ugolini et al,, 2008: Huncharek, 2002: Roushdv-Hammadv et 21 eil., 2001). although whether a genetic factor or a common environmental element leads to the 22 similar responses in these families is difficult to determine. Increased interest in the role of 23 genetic factors in asbestos-related health outcomes has led to several analytical studies on 24 specific genetic polymorphisms. A review of 24 published reports (19 studies) discusses the 25 current state of knowledge regarding geneti c susceptibility associated with asbestos-related 26 diseases (in particular, malignant pleural mesothelioma). Results from several studies 27 demonstrated an association between asbestosis-related diseases and GSTM 1-null 28 polymorphism, whereas results for other polymorphisms were conflicting (Neri et al., 2008). 29 Some polymorphisms discussed in Neri et al. (2008) are in genes for A-acetyl-transferase 2; 30 glutathione-s-transferases (GSTs); SOD; CYP1 Al, CYP2D6; neurofibromatous 2 (Nf2); p53; This document is a draftfor review purposes only and does not constitute Agency policy. 4-90 DRAFT--DO NOT CITE OR QUOTE 1 and XRCC1. Although occupational asbestos exposure was assessed, the type of asbestos is 2 generally unknown in these studies. 3 Limited animal studies have examined the role of genetic variations related to asbestos 4 exposure, including specific signaling pathways (Shukla et aL 2007). DNA damage repair (Lin 5 et aL 2000; Ni et al.. 2000k and tumor suppressor genes (Vaslet et aL 2002; Kleymenova et aL, 6 1997; Marsella et al.. 19971. Genetic alterations of particular interest for mesothelioma include 7 those involved in tumor suppression (p53, Nf2) and oxidative stress (SOD, GSTs). Nf2 and p53 8 are frequently altered in mesotheliomas, but no consistent mutations have been found (Cheng et 9 al.. 1999; Mavall et al.. 1999; Bianchi et al.. 19951. Alterations in expression of antioxidant 10 enzymes like SOD and GST in mesothelioma can yield cells more resistant to oxidative stress as 11 compared to normal cells due to increased antioxidant activity (Ramos-Nino et al., 2002; 12 Rahman and MacNee. 19991. No studies that examine the role of cell-cycle control genes were 13 found following exposure to Libby Amphibole asbestos. Additionally, no information on other 14 genetic polymorphisms in relation to disease risk among those exposed to Libby Amphibole 15 asbestos was identified in the available literature. 16 17 4.7.5. Influence of Health Status on Susceptibility 18 Preexi sting health conditions could potentially alter the biological response to asbestos 19 exposure. Mesothelioma risk has been hypothesized to be related to immune impairment 20 (Bianchi and Bianchi. 2008) and simian virus 40 exposure in humans (Carbone et aL 2007; 21 Kroczynska et al.. 2006; Cristaudo et al... 2005; Foddis et al.. 2002; Bocchetta et al.. 2000; 22 Mavall et al.. 1999). Coexposure to asbestos and SV40 has been associated with p53-related 23 effects in vitro (Foddis et al.. 2002; Bocchetta et al... 2000; Mavall et al.. 1999), and cell signaling 24 aberrations in vivo (Kroczynska et al.. 2006; Cristaudo et al.. 2005). However, the influence on 25 cancer risk is unknown, as these lines of research are not fully developed and have not been 26 applied specifically to Libby Amphibole asbestos. 27 Obesity can compromise inhalation exposure, as increased particle deposition In the lungs 28 of overweight children (Bennett and Zeman. 2004) and adults (Graham et al.. 1990) has been 29 observed. Individuals with respiratory diseases could have compromised lung function that 30 alters inhalation exposure to Libby Amphibole asbestos. For example, individuals with chronic 31 obstructive pulmonary disease have increased inhalation volume (Phalen et al.. 2006) and This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 4-91 DRAFT--DO NOT CITE OR QUOTE 1 increased fine particle deposition (Phalen et ah, 2006; Bennett et a).., 1997; Kim and Kang, 1997) 2 and retention (Regnis et al.. 2000). Similarly, studies have reported an increase in coarse particle 3 (aerodynamic diameter >5 pm) deposition in individuals with cystic fibrosis (Brown and 4 Bennett. 2004; Brown et al., 2001). For people exposed to Libby Amphibole asbestos, an 5 increased risk for interstitial lung abnormalities was observed for those with a history of 6 pneumonia (Peipins et al.. 2003). In another study, bronchial asthma was examined as a 7 potential confounding variable for asbestos-related effects on pulmonary function, although no 8 confounding was observed (Whitehouse. 2004). 9 10 4.7.6. Influence of Lifestyle Factors on Susceptibility 11 No studies were identified that examined lifestyle factors specifically with respect to 12 Libby Amphibole asbestos. Lifestyle factors such as exercise, nutritional status, and smoking 13 habits could affect the biological effects of asbestos exposure through various mechanisms. For 14 example, those with more physically demanding jobs or those who regularly engage in vigorous 15 exercise might experience increased lung deposition from fine particles or fibers compared to 16 those with a more sedentary lifestyle (Phalen et al.. 2006; Becquemin et al.. 1991). Randomized 17 controlled trials of vitamin supplementation (beta-carotene and retinol) have been conducted for 18 asbestos-related lung cancer, but results do not support a protective effect (Cullen et al., 2005) 19 For lung cancer, a synergistic relationship between cigarette smoking and asbestos 20 exposure has been demonstrated (Wraith and Mengersen. 2007; Hammond et al... 1979; Selikoff 21 and Hammond. 1979). Research has suggested that asbestos fibers might also enhance the 22 delivery of multiple carcinogens in cigarette smoke, and that cigarette smoking decreases the 23 clearance mechanisms in the lungs and could, therefore, lead to an increase in fiber presence in 24 the lungs (Nelson and Kelsey, 2002). Smoking likely causes genetic alterations associated with 25 lung cancer (Landi et al.. 2008) that might increase the carcinogenic risk from exposure to 26 asbestos. Benzo(a)pyrene, a component of tobacco, also has been observed to enhance the 27 carcinogenic effects of asbestos (Loll et al.. 2004; Kimizuka et al.. 1987; Mossman et al.. 1984; 28 DiPaolo et al.. 1983; Mossman et al.. 1983; Reiss et al.. 1983). 29 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 4-92 DRAFT--DO NOT CITE OR QUOTE 1 4.7.7. Susceptible Populations Summary 2 A very limited amount of information is available on exposure to Libby Amphibole 3 asbestos early in life that could lead to increased risk of asbestos-induced disease later in life. 4 Due to the long latency period of some diseases in relation to asbestos exposure in general, 5 adverse effects may be more likely to be observed with an increase in age. This assumption 6 requires further investigation. The number of women who have been occupationally exposed to 7 Libby Amphibole asbestos is very small, and health risks have not been evaluated specifically 8 for this group. Differences between men and women in residential sources and types of exposure 9 (e.g., types of activities done in the household) also preclude the possibility of drawing 10 conclusions regarding the relative susceptibility of women compared with men to health effects 11 of exposure to Libby Amphibole asbestos. Similarly, sufficient data are not available to draw 12 conclusions regarding racial or ethnic variation in susceptibility to diseases caused by exposure 13 to Libby Amphibole asbestos. In addition, the potential modifying effects of genetic 14 polymorphisms, preexisting health conditions, nutritional status, and other lifestyle factors have 15 not been studied, specifically as related to exposure of Libby Amphibole asbestos and health 16 outcomes. This document is a draftfor review purposes only and does not constitute Agency policy. 4-93 DRAFT--DO NOT CITE OR QUOTE 1 5. EXPOSURE-RESPONSE ASSESSMENT 2 5.1. ORAL REFERENCE DOSE (RfD) 3 Data are unavailable to characterize the toxic effects of Libby Amphibole asbestos22 4 following oral exposure. Thus, an oral reference dose is not derived. 5 6 5.2. INHALATION REFERENCE CONCENTRATION (RfC) 7 5.2.1. Choice of Principal Study and Critical Effect 8 Studies in humans have shown radiographic evidence of health effects on the lung and 9 pleura (a thin tissue surrounding the lung and lining the chest cavity) such as pleural thickening 10 and fibrosis of the lung and pleura in exposed workers (Larson et al.. 2010a; Robs et al.,, 2008; 11 Amandus et al., 1987b; McDonald et al., 1986b; Lockev et al., 1984) as well as community 12 studies (Weill et al.. 2011; Muravov et al.. 2005; Pei pins et al.. 2004b; Whitehouse, 2004; 13 Peipins et al.. 2003) (see Sections 4.1.1.4 and 4.1.2). Five cohort mortality studies of workers 14 who mined, milled, and processed Libby vermiculite (henceforth described as the Libby 15 workers) identified increased risk of mortality from noncancer causes including nonmalignant 16 respiratory disease--especially asbestosis, chronic obstructive pulmonary disease, and silicosis 17 (Larson et al.. 2010b; Sullivan.. 2007; McDonald et al... 2004; Amandus and Wheeler. 1987; 18 McDonald et al.. 1986a) as well as cardiovascular disease (Larson et al.. 2010b). Additionally, 19 there is a potential for autoimmune effects following inhalation exposure to Libby Amphibole 20 asbestos (Noonan et al.. 2006; Pfau et al.. 2005) (see Section 4.3). The overall noncancer hazard 21 identification for exposure to Libby Amphibole asbestos is summarized in Section 4.5. A 22 reference concentration (RfC) is intended to define an exposure level that is likely to be without 23 an appreciable risk of adverse health effects; studies that relate these health effects to exposure 24 levels are necessary for RfC derivation23. Quantitatively, study characteristics preferred for RfC 25 derivation include adequate exposure-response information, ideally with analyses based on ~The tern "Libby Amphibole asbestos" is used in this document to identity the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby. MT. It is further described in Section 2.2. 23An RfC is defined as "An estimate (with uncertainty spanning perhaps an order o f magnitude) o f a continuous inhalation exposure to the human population (including sensitive subgroups) that is likely to be without an appreciable risk o f deleterious effects during a lifetime." This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-1 DRAFT--DO NOT CITE OR QUOTE 1 estimates including assignment of quantitative exposure estimates to distinguish exposure levels 2 in the study subjects. 3 Of the available human studies, only the worker mortality and morbidity studies provide 4 exposure estimates suitable for quantitative analysis to derive benchmark concentration estimates 5 or NOAELs/LOAELs and, thus, would allow for consideration for use in RfC derivation (Larson 6 et al.. 2010b; Rohs et al.. 2008; Sullivan. 2007; McDonald et ai... 2004; Amandus and Wheeler. 7 1987; Amandus et al.. 1987b; McDonald et al... 1986a; McDonald et al.. 1986b; Lockev et ah, 8 1984). Although there are data that define exposures from some activities in the community (see 9 Section 2.3), these data do not address all potential exposures nor are data available on activity 10 patterns, which would be needed to provide individual exposure measurements. There are no 11 studies in laboratory animals on the inhalation route of exposure suitable for derivation of an 12 RfC because available animal studies lack adequate exposure-response information and are of a 13 short-term duration. Therefore, only the worker studies that include adequate exposure 14 assessment and identify health effects are considered for RfC derivation. 15 Five cohort mortality studies of Libby workers identified increased risk of mortality from 16 noncancer causes (Larson et al . 2010b; Sullivan. 2007; McDonald et al.. 2009; Amandus and 17 Wheeler. 1987; McDonald et al.. 1986a). These studies were not considered as candidates for 18 RfC derivation because the radiographic parenchymal and pleural abnormalities are more 19 sensitive than the corresponding mortality causes. An RfC is intended to be a level at which no 20 category of adverse health outcome would occur. 21 Although one study (i.e.. Larson et al.. 2010b) has reported an increase in mortality from 22 various cardiovascular diseases, no studies have been conducted in a population exposed to 23 Libby Amphibole asbestos on cardiovascular endpoints other than mortality. The reported 24 excess mortality specific to vascular effects is unique, and further substantiation of this finding is 25 needed. Thus, the mortality represents a more severe health effect from related pulmonary and 26 pleural endpoints. The less severe indicator of the first radi ographic changes is the preferred 27 endpoint for RfC derivation. 28 Several morbidity studies examined the quantitative association between exposure to 29 Libby Amphibole asbestos and lesions in the lung or surrounding pleura in exposed human 30 populations; two are studies in Libby workers (Amandus et al.. 1987b; McDonald et al.. 1986b). 31 and two are studies in workers from the Marysville, OFI facility (Rohs et al.. 2008; Lockev et al.. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-2 DRAFT--DO NOT CITE OR QUOTE 1 1984). Rohs et al. (2008) was a follow-up study to Lockey et al. (1984) on a subset of the same 2 cohort and reported a higher prevalence of adverse effects following the longer time from first 3 exposure. These four studies, all of which demonstrate an association between Libby Amphibole 4 asbestos exposure and increased risk of effects on the lung and pleura, were considered for 5 selection as the principal study to serve as the basis for the derivation of the RfC. 6 All four candidate principal studies (Rohs et al.. 2008; Amandus et al., 1987b; McDonald 7 etal- 1986b; Lockey et al.. 1984) have adequate reporting of the studied populations, methods of 8 analysis, statistical analyses, and results. Each of the four candidate studies reports radiographic 9 signs of nonmalignant respiratory effects, which may be considered as endpoints for an RfC 10 derivation, specifically pleural thickening (localized and/or diffuse) and small opacities 11 (indicative of parenchymal damage) (ILQ. 2002. 1980. 1971). Table 5-1 summarizes the four 12 candidate principal studies. See Sections 4.1.1.4 and 4.1.3 for detailed study information. 13 14 5 .2 .1 .1 . Evaluation o f Candidate Studies and Selection o f Critical Study 15 The candidate studies were evaluated in terms of quality attributes that would support 16 their use as a principal study in the derivation of an RfC. When selecting among candidate 17 principal studies, there were several factors, summarized in Table 5-2, that were generally 18 considered. 19 20 5 .2 .1 .2 . Evaluation o f Exposure Paradigm in Candidate Studies 21 Each of the studies provided estimates of cumulative Libby Amphibole asbestos exposure 22 (in fibers/cc-year), rather than mean or peak exposure. Elowever, there were differences in 23 exposure intensity. In contrast to vermiculite facility workers in Libby, MT, the workers at the 24 O.M. Scott Plant in Marysville, OH, were generally exposed at lower levels (see Table 5-1), and 25 were primarily exposed in the workplace. Because of showering and changing into civili an 26 clothes at the end of the work shift for most employees, nonoccupational exposure in the 27 Marysville workers was minimal. Despite the uncertainty in the magnitude of pre-1972 28 exposures (discussed below), the available data indicate worker exposures in the Marysville 29 plant did not generally include the high intensity exposures observed for the Libby worker 30 cohort, with Rohs et al. (2008) reporting a mean exposure of 2.48 fibers/cc-year. The lower This document is a draftfor review purposes only and does not constitute Agency policy. 5-3 DRAFT--DO NOT CITE OR QUOTE Table 5-1. Sum m ary of candidate principal studies on Libby Am phibole asbestos for reference concentration (RfC) derivation This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy. Cohort and reference Study population Outcome assessment Radiographic endpoints evaluated Exposure assessment Exposure characteristics Libby W orker C ohort McDonald et al. (1986b) 244 employees, comprising 164 "current" workers (as o f July 1, 1983) and 80 "past" w orkers Age at exam (years): "current" "nast" <39 80 1 40-59 69 30 >60 15 49 No job tenure information; f10.7 years as reported by Armstrong et al. (1988)1 Radiographs taken at time o f cohort assembly (1983) Films independently read by three experienced readers using 1980 ILO standards Film quality: Good: 56% Fair: 36% Poor: 7% Unreadable: 0.4% 1) Parenchymal changes (small opacities >1/0) 2) Pleural changes (pleural thickening on chest wall, pleural calcification) Individual work histories and exposure levels for specific work locations were used to estimate cumulative exposures for cohort members. 1935-1967: Exposure estimated based on professional judgment. For mill locations only (1950-1967), exposure estimated using dust-to-fiber conversion and interviews with plant employees. Mean cumulative exposure "current" 40.1 fibers/cc-yr "past" 118.9 flbers/cc-yr Exposure categories: <10 fibers/cc-yr (n = 92) 10--<20 fibers/cc-yr ( = 64) 20-<100 fibers/cc-yr (n = 53) 100-<200 fibers/cc-yr (77 = 16) > = 200 fibers/cc-yr (77 = 19) 1968-1982: Air samples analyzed for fibers by PCM analysis. 5-4 DRAFT--DO NOT CITE OR QUOTE Table 5-1. Sum m ary of candidate principal studies on Libby Am phibole asbestos for reference concentration (RfC) derivation (continued) This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy. Cohort and reference Amandus et al. (1987b) Study population 184 men employed 1975-1982, with at least 5 years job tenure Mean (SD), years: Age at exam: 44 (12) Job tenure: 14 (8) Outcome assessment Company radiographs Source year: 1981-1982 (72.8%) 1976-1980 (26.6%) <1975 (1 worker) Films independently read by three readers using 1980 ILO standards Film quality (by reader): Excellent: 22.8, 24.4, 47.9% Acceptable: 60.9, 60.9, 29.3% Poor: 16.3, 14.7, 22.8% Unreadable: None Radiographic endpoints evaluated 1) Parenchymal changes (small opacities >1/0) 2) Pleural changes ("any pleural change"'1, pleural calcification, pleural thickening on chest wall only) Exposure assessment Exposure characteristics Individual work histories and Exposure categories: exposure levels for specific 0 -1 5 fibers/cc-year work locations were used to (h = 63) estimate cumulative exposures 16-30 fibers/cc-year for cohort members. (n = 29) 31-85 fibers/cc-year 1935-1967: Exposure (n = 44) estimated based on >86 fibers/cc-year ( - 48) professional judgment. For mill locations only (1950-1967), exposure estimated using dust-to-fiber conversion and interviews with plant employees. 1968-1982: Air samples analyzed for fibers by PCM analysis. 5-5 DRAFT--DO NOT CITE OR QUOTE Table 5-1. Sum m ary of candidate principal studies on Libby Am phibole asbestos for reference concentration (RfC) derivation (continued) This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy. Cohort and reference Study population O.M. Scott Plant Cohort, Marysville, OHb Lockey et al. (1984) 512 plant employees Mean (range), years: Age at exam: 37.5 (19-66) Mean (SE), years: Job tenure by exposure group and smoking status (N S=nonsmoker, EX=former smoker, CS=current smoker) Low, NS: 6.6 (1.1) Low, EX: 11.3 (1.6) Low, CS: 10.5 (1.2) Medium, NS: 8.4 (1.0) Medium, EX: 13.3 (1.3) Medium. CS: 8.9 (0.7) High. NS: 12.2 (0.9) High, EX: 13.0 (1.1) High, CS: 10.7 (0.9) Outcome assessment Radiographic endpoints evaluated Exposure assessment Exposure characteristics Posterior-anterior chest radiographs taken in 1980 Films independently read by 2 board-certified radiologists (B-readers) using modification of 1971ILO standards. A third B-reader was used to resolve any difference in diagnosis. 1) Parenchymal changes (only one small opacity recorded [grade 1/1], unclear if opacities graded 1/0 or 0/1 would have been reported) 2) Pleural changes (pleural plaque, pleural thickening, pleural calcification) 3) Costophrenic angle blunting only Self-reported individual work histories and exposure levels for specific work locations were used to estimate cumulative exposures for cohort members. 1957-1971: Exposure estimated based on interviews with plant employees and post-1972 air measurements. Some workplace exposure control measures were taken prior to 1972. Exposure categories: <1 fibers/cc-year (n = 253) 1-10 fibers/cc-year (ti = 200) >10 fibers/cc-year (;? = 48) 1972-1980: Air samples analyzed for fibers by PCM analysis. The exposure reconstruction in the original study was based on limited data, and air sampling data from 1972 on were not available for all jobs. Where data were not available, the earliest available sampling data informed early exposures (Lockev. 1985). 5-6 DRAFT--DO NOT CITE OR QUOTE Table 5-1. Sum m ary of candidate principal studies on Libby Am phibole asbestos for reference concentration (RfC) derivation (continued) This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy. Cohort and reference Rohs et al. (2008) Study population Outcome assessment Radiographic endpoints evaluated 280 plant employees [follow-up of cohort described in Lockey et al. (1984)] Mean (SD), range (years): Age: 59.1 (10.5), 4 4 -8 7 Mean (SD), median (years): Years since first exposure No pleural changes (/? = 200): 32.1 (5.5), 31.0 Pleural changes present (/? =. 80): 36.8 (4.9), 37.9 Posterior-anterior chest radiographs taken 2002-2005 Films independently read by three board-certified radiologists (B-readers) using 2000 ILO standards Seven employees had unreadable films and are not included in the cohort o f 280 participants 1) Parenchymal changes (small opacities, profusion score >1/0) 2) Pleural changes (localized pleural thickening [any pleural thickening excluding costoplirenic angle blunting], diffuse pleural thickening [any pleural thickening with costoplirenic angle blunting], pleural calcification) Exposure assessment Exposure assessment from Lockev et al. (1984) with change in start date to 1963. Exposure characteristics Exposure categories: 0.01-0.28 fibers/cc-year (n = 70) 0.29-0.85 fibers/cc-year (n = 72) ` 0.86-2.20 fibers/cc-year (n = 68) ` 2.21-19.03 fibers/cc-year (n = 70) aAmandus et al. (1987b p. 28) define "any pleural change" as "... any unilateral or bilateral pleural change, which included pleural plaque, diffuse pleural thickening o f the chest wall, diaphragm or other site, but excluded costoplirenic angle obliteration... ." bln addition to the exposure infonnation used by Lockey et al. (1984) and Rolls et al. (2008), the University o f Cincinnati augmented and refined these exposure estimates using additional exposure data, which included industrial hygiene measurements not previously available and measurements using industrial hygiene data from the facility to determine estimates o f exposure after 1980. 5-7 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-2. Summary of rationale for identifying candidate principal studies 2 on Libby Amphibole asbestos for RfC development 3 A ttrib u te Preferred characteristics for candidate principal studies for the Libby Am phibole Asbestos RfC Relevance o f exposure paradigm Studies o f subchronic or chronic duration are preferred over studies o f acute exposure duration because most relevant environmental exposure scenarios are expected to address chronic exposure scenarios (potentially including both continuous exposure from ambient conditions and episodic activity-related exposures). Measures o f cumulative exposure are a widely used metric to address asbestos risk. It is consistent with the expectation that toxic responses will reflect an accumulative effect o f asbestos inhaled and deposited in tissues over time. Additionally mean exposure, exposure duration, and time from first exposure (TSFE) have all been reported as predictors o f health effects from asbestos exposure. Cumulative exposure has the advantage that it reflects both duration and intensity (e.g., mean level) o f asbestos exposure. Relatively lower exposure intensities that may represent conditions more similar to environmental exposures are preferred as there may be less uncertainty in extrapolation of the results to lower exposure levels. Results from studies with high exposure intensity or cumulative exposure are, other things being comparable, judged less relevant for environmental risk assessment compared to studies defining effects at lower levels o f exposure. Some biological processes (e.g., potential decrease in effectiveness o f particle clearance processes) may more strongly influence responses at very high levels o f exposure and be less relevant at lower levels. Thus, exposure conditions with lower level exposures may remove some o f the uncertainty in estimating health effects from environmental exposures. Study design characteristics Sufficient follow-up time for outcomes to develop (which can depend on the health outcome being addressed). Study size and participation rates that are adequate to detect and quantify health outcomes being studied are preferred, with no indications o f bias in study population selection. Use o f a study design or analytic approach, which adequately addresses the relevant sources o f potential confounding, including age, sex, smoking, and exposure to other risk factors (such as non-Libby asbestos). This document is a draftfor review purposes only and does not constitute Agency policy. 5-8 DRAFT--DO NOT CITE OR QUOTE Table 5-2. Sum m ary of rationale for identifying candidate principal studies on Libby Amphibole asbestos for RfC development (continued) Measurement of exposure Emphasis is placed on the specificity of exposure assessment in time and place with a preference for greater detail where possible. Exposure measurements that are siteand task-specific provide appropriate exposure information, and individual, rather than area samples are preferred where available. Measurement techniques that are more specific to the agent of concern are preferred over less specific analytical methods. Better characterization of fibers is preferred. For asbestos fibers, TEM analysis, which can identity the mineral fibers present, provides the most specific information; PCM identifies fibers as defined by that method (NIOSH 7400) and, thus, is useful but do not confirm the mineral nature of the counted fibers. Total dust measurements are the least informative of those available. Stronger studies will often be based upon knowledge of individual work histories (job titles/tasks with consideration of changes over time); however, appropriate group-based exposure estimates may also be relevant. Measurement of effect(s) Exposure reconstruction and estimating exposures based on air sampling from other time periods and/or operations are less preferred methods of exposure estimation. Emphasis is placed on the more sensitive health outcome endpoints that are available. For parenchymal and pleural effects considered here, the radiographic abnormalities are more sensitive than the corresponding mortality causes. An RfC is intended to be a level at which no category of adverse health outcome wo uld occur. Pleural and parenchymal abnormalities assessed using good quality radiographs or high-resolution computed tomography (HRCT) and independently evaluated multiple qualified readers according to ILO standards. Evaluation of radiographs should not be influenced by knowledge of exposure status. 1 2 3 intensity exposures for the Marysville cohort and corresponding lower cumulative exposures are 4 advantages of this study, considering there are uncertainties inherent in exposure-response data 5 and extrapolating from the high intensity occupation exposures to lower level exposures often 6 seen in community and environmental exposures. 7 8 5.2.1.2.1. Evaluation o f study design in candidate studies 9 The candidate principal studies differed in the study populati ons, in terms of follow-up 10 time, study size and participation, and available information (see Table 5-1). The study sizes are 11 similar for the two Libby worker studies (n = 184 and n= 244, respectively) (Amandus et al.. 12 1987b; McDonald et al., 1986b) and the Marysvill e update (n = 280) (Rohs et aL 2008). 13 Adequate follow-up time allows for the health effect to manifest prior to sampling. In the 14 case of pleural abnormalities, there is some variability with latency based on intensity of This document is a draftfo r review purposes only and does not constitute Agency policy. 5-9 DRAFT--DO NOT CITE OR QUOTE 1 exposure as well as the nature of the pleural lesion where discrete pleural plaques have a shorter 2 latency than diffuse thickening of the visceral pleura. Larson et al. (2010a) studied the latency 3 for individuals in the Libby worker cohort, reporting a median latency of 8.6 years for localized 4 pleural thickening versus 27 years for diffuse pleural thickening and 19 years for minimal signs 5 of small opacities (parenchymal changes).24 Lockey et al. (1984) report the mean employment 6 duration for their exposure groups from 6.6 to 13.3 years at the time of their study (but do not 7 assess time since first exposure (TSFE); thus, it is unclear whether in the first examination these 8 workers had sufficient follow-up to assess the radiographic changes, especially diffuse pleural 9 thickening and small opacities. The Rohs et al. (2008) report includes 24 more years of 10 follow-up time and is preferred over the early Lockey et al. (1984) study on this basis. 11 Both studies of the Libby workers report duration of employment and average age of the 12 participants, but not TSFE. The McDonald et al. (1986b) study included both current and former 13 workers--these former workers likely have longer time from first exposure compared with 14 current workers. The study included all current plant employees (164 men, 9 women). 15 Flowever, there was a lower participation rate in former employees (80 of 110 eligible former 16 employees agreed to provide chest radiographs). Additionally, X-rays for all study participants 17 were taken in the same year, providing similar quality X-rays between past and current 18 employees. In contrast, Amandus et al. (1987b) only considered workers employed during 1975 19 to 1982 and relied on available radiographs regardless of year (radiographs were available for 20 93% of employees). Because workers terminated prior to 1975 were excluded from the study, 21 older individuals, and individuals with longer TSFE were less likely to be included than in the 22 study by McDonald et al. (1986b). which included former workers. Both Libby worker studies 23 do report radiographic abnormalities, so the follow-up is adequate for some effects to be 24 documented; however, compared with the Rohs et al. (2008) study, the Libby worker studies 25 have shorter follow-up times. 24Individual latency for visible LPT in Libby exposed workers was evaluated in 84 workers with radiographic evidence o f pleural and/or parenchymal changes (Larson et al., 2010a). By examining historical radiographs, researchers were able to identify the first appearance o f the lesions, although it is recognized that retrospective design o f this study likely identified lesions at earlier time points, as the readers were aware o f the later X-rays (Larson et al., 2010a). It is acknowledged that some of the workers at Libby may have been exposed through the community prior to working, and in fact, one individual had the first pleural change noted at 9 years o f age. prior to occupational exposure (Larson et al.. 2010a). Where data on prior exposures were available, workers with no prior exposure had an average latency o f 9.4 years versus 5.1 years for workers with potential exposures prior to hire (jV= 63 and 31, respectively). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-10 DRAFT--DO NOT CITE OR QUOTE 1 Among Marysville workers, there were very few employees who declined to participate 2 in the earlier study by Lockey et al. (1 9 8 4 ), where 5 1 2 out of 5 3 0 employees were included, but 3 there is potential for selection bias in the follow-up by Rohs et al. (2 0 0 8 ), where only 4 2 8 0 employees out of the original cohort were evaluated. Rohs et al. (2 0 0 8 ) state that employees 5 hired in 1973 or earlier (when exposure estimates were more uncertain) were more likely to 6 participate compared to employees hired after 1973, and while the range of cumulative Libby 7 Amphibole asbestos exposure was similar between participants and nonparticipants, participants 8 did have higher mean cumulative exposure estimates. While it is accurate that exposure levels 9 were uncertain before sampling began at Marysville in 1972, it is also accurate that exposures 10 were much lower beginning in 1974, when additional industrial hygiene controls were 11 implemented. Thus, persons hired < 1 9 7 3 had higher exposure (if less perfectly measured), while 12 those hired > 1 9 7 4 had lower exposure, and likely less disease (under an assumption of an 13 exposure-response effect). Thus, we might assume that the prevalence rates in nonparticipants 14 are likely lower than in participants. The self-selection to participate in the study is dependent 15 on the exposure, thus leading to dependent censoring and potential selection bias (see 16 Section 4.1 .3 for a discussion of this potential selection bias). However, Rohs et al. (2 0 0 8 ) 17 conducted a sensitivity analysis assuming that all living nonparticipants had no pleural changes 18 and report a similar significant trend of increased pleural changes by exposure quartile. In 19 contrast, participation rates for the Libby worker studies were much higher (see above), and there 20 is no indi cati on of potential bias in selection of these study parti cipants (Amandus et al., 1 9 8 7 b ; 21 McDonald et al., 1 9 8 6 b ). 22 Both studies of Libby workers also evaluated age and smoking as potential confounders 23 of the association between Libby Amphibole asbestos exposure and radiographic abnormalities. 24 McDonald et al. (1 9 8 6 b ) report that both age and cumulative exposure are significant predictors 25 of small opacities and pleural abnormalities in the study of current and former workers, 26 providing regression coefficients for cumulative exposure, age, and smoking status. Amandus et 27 al. (1 9 8 7 b ) report that although cumulative exposure and age are both significant predictors for 28 small opacities, cumulative exposure was not significantly related to pleural abnormalities when 29 age is included in the model, thus limiting the usefulness of these data for RfC derivation based 30 on pleural abnormalities. Neither study of Libby workers addressed gender, body mass index This document is a draftfor review purposes only and does not constitute Agency policy. 5-11 DRAFT--DO NOT CITE OR QUOTE 1 (BMI), or time from first exposure, although both studies excluded workers with other 2 asbestos/dusty trade occupations. 3 With respect to the Marysville, OH worker cohort, Lockey et al. (1984) only matched on 4 age in their analysis. The follow-up examination by Rohs et al. (2008) included information on 5 several important covariates, including age, gender, hire date, prior exposure to asbestos, BMI, 6 and smoking history. Hire date and age were significantly associated with the prevalence of 7 pleural abnormalities, and results are presented considering these covariates. 8 9 5.2.I.3. Evaluation o f Exposure Assessment in Candidate Studies 10 For both the O.M. Scott facility in Marysville, OH and the Libby, MT facilities, exposure 11 estimates rely primarily on fiber counts using phase contrast microscopy (PCM) and 12 reconstruction of earlier exposures from company records, employee interviews, and the 13 professional judgment of the researchers estimating historical exposures (Amandus et al.. 1987a; 14 McDonald et al. 1986a; Lockey et al., 1984). Work histories for the Libby worker cohort were 15 extracted from company employment records, while work histories for the Marysville cohort 16 were self-reported. 17 The two studies of workers in Libby, MT used similar exposure estimation, based on the 18 same fiber measurements and work records (Amandus et ah, 1987b; McDonald et al.. 1986a). 19 As discussed in Section 4.1.1.2, exposures prior to 1968 are not based on fiber measurements by 20 PCM and, thus, are more uncertain that later exposure estimates,25 The study population of 21 McDonald et al. (1986b) included current and former workers, with 26% of participants over 60 22 and 40% of participants between 40-59 years of age at the time of their X-ray in 1983. 23 Although tenure and dates of employment are not reported, exposure estimates for this study 24 group would include the less-certain exposure estimates prior to 1968 (McDonald et al.. 1986a). 25 However, Amandus et al. (1987b) studied workers still employed during 1975-1982 (i.e., 26 excluding those terminated prior to 1975) who had at least 5 years of employment. The average 27 tenure of the study parti cipants was 14 years. Although both studi es have the limitation of 28 less-certain exposure estimates prior to 1968, based on study design, the Amandus et al. (1987b) 25Exposures in the dry mill at Libby, MT. prior to 1967 were estimated from total dust measurements based on site-- specific conversion ratios. Exposures for all other location operations prior to 1968 were estimated because no air sampling data were available (Amandus et al.. 1987a: McDonald et al.. 1986b). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-12 DRAFT--DO NOT CITE OR QUOTE 1 study group includes a greater proportion of more recent workers. However, neither researcher 2 assessed these uncertainties nor the impact of early exposure estimates on the apparent 3 exposure-response relationship. 4 Another source of uncertainty in exposure estimates for this cohort is possible 5 community/nonoccupational exposures. Members of the Libby worker cohort may have lived in 6 Libby prior to/after employment and resided in Libby and surrounding areas during employment. 7 In both cases, there may have been community exposures to Libby Amphibole asbestos that are 8 not captured in occupational-based cumulative exposure metrics. This unmeasured 9 nonoccupational exposure may be low relative to the estimated occupational exposures, but is, 10 nevertheless, a source of uncertainty in estimating the exposure-response relationship. 11 The quality of the exposure assessment also changed over time in the Marysville cohort 12 (Rohs et aL 2008; hockey, 19851. Industrial hygiene measurements based on PCM analysis are 13 available for the O.M. Scott facility beginning in 1972, although personal breathing zone 14 samples were not available until 1976 (Rohs et aL 2008). Thus, exposure levels for all job tasks 15 prior to 1972 are estimates from later sampling events. Additionally, air sampling data were not 16 available for several job tasks until the late 1970s. For example, air-sampling data were only 17 available for two of seven job tasks in the trionizing department beginning in 1973 (expander 18 and dryer). All others have dates of 1976 or later [see Table 10, Lockey (1985)1. The 19 installation of exposure control equipment in 1974 adds to the uncertainty in early exposures 20 estimated from sampling in later years. There is uncertainty when the Libby ore was first used in 21 the facility. Company records indicated that the date was between 1957 and 1960, and the 22 University of Cincinnati used the best-available information from focus group interviews to 23 assign the first usage of Libby ore in 1959 (see Appendix F). 24 EPA has collaborated with the University of Cincinnati research team to better evaluate 25 historical exposures at the O.M. Scott facility in Marysville, OH (see Appendix F). Although no 26 air-sampling results were found prior to 1972, additional information on plant processes from 27 other records and employee interviews has resulted in updated exposure estimates (see 28 Section 5.2.3.1). These refined estimates of the historical exposure improve exposure 29 characterization for the Marysville worker cohort over previous publications. 30 This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-13 DRAFT--DO NOT CITE OR QUOTE 1 5.2.1.3.1. Evaluation o f outcome assessment in candidate studies 2 In all four candidate studies, outcomes were assessed using chest radiographs 3 independently evaluated by multiple readers. However, there were differences in the standards 4 used for evaluation of radiographic changes, as well as timing and quality of the radiographs. 5 The two studies in Libby workers (Amandus et ah. 1987b; McDonald et al., 1986b) used similar 6 outcome-assessment procedures, with radiographs evaluated by three readers according to 1980 7 ILO standards. Two different sets of standards were used to evaluate radiographs in the 8 Marysville cohort. The first study used modified 1971 ILO standards (modifications not 9 stipulated) (Lockev et al.. 1984). while the follow-up study used the updated 2000 ILO standards 10 (Rohs et aL 2008). 11 Radiograph quality may also impact outcome assessment. In McDonald et al. (1986b). 12 which used radiographs taken in 1983 specifically for the study, 7% of films were classed as 13 "poor quality" (some technical defect impairing the pneumoconiosis classification) and 0.4% as 14 "unreadable." Amandus et al. (1987b). which used available radiographs taken over a wide time 15 peri od (1975 to 1982), report that the proportion of films rated as "poor quality" ranged from 16 14.7% to 22.8% depending on the reader. In the Marysville cohort, Lockey et al. (1984) state 17 that ",. .radiographs that could not be interpreted because of poor quality were repeated" (p. 953). 18 Rohs et al. (2008) do not report the percentage of films rated as "poor quality" but do note that 19 7 out of 298 (2.3%) radiographs taken were considered unreadable. 20 21 5.2.1.3.2. Selection o fprincipal cohort 22 Based on the criteria set out in Table 5-2 and the above evaluation, the update of the 23 Marysville, OH worker cohort (Rohs et al.. 2008) is the preferred cohort. The main advantages 24 of the Marysville, OH worker cohort over the two studies of pleural and lung abnormalities in 25 the workers in Libby, MT are: 26 27 28 1) Adequate follow-up time and the availability of time from first exposure data for 29 evaluation, 30 2) Minimal exposure to Libby Amphibole asbestos outside of the workplace, This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-14 DRAFT--DO NOT CITE OR QUOTE 1 3) Better quality radiographs, and use of the most recent ILO reading guidelines in the 2 cohort update, 3 4) Data are more appropriate for low-dose extrapolation--a lower range of cumulative 4 exposures for the study participants (n = 280), compared to Libby workers, 5 5) The data allow consideration of more covariates and potential confounders (e.g., 6 BMI, smoking status, age), 7 6) The presence of a demonstrated exposure-response relationship for Libby amphibole 8 asbestos exposure and radiographic abnormalities--in contrast to the study by 9 Amandus et al. (1987b). which does not support an exposure-response relationship 10 for pleural abnormalities based on the cumulative exposure metric (when age is 11 included as a covariate). 12 13 14 The disadvantages of the Marysville, OH cohort compared to the two studies of pleural 15 and lung abnormalities in the workers in Libby, MT are: 16 17 18 1) Approximately 70% of the Marysville, OH cohort were hired before 1972 when there 19 were no measured exposure data [Rohs et al. (2008). and hockey et al. (1984) study], 20 2) Participants in Rohs et al. (2008) were self-selected, with greater participation among 21 older employees and those who began work prior to 1973 when exposures were 22 relatively higher. This is a potential source of bias in study population selection 23 analyzed by Rohs et al. (see Section 4.1.3). 24 3) Exposure estimates are based on self-reported work histories. In this case, there is 25 some uncertainty in the employment history, and some individuals had extensive 26 overtime work. Employment history was self-reported during interviews with each 27 individual for the original study (i.e.. Lockev et ah. 1984). and errors in this process 28 could affect assigned Libby Amphibole asbestos exposure estimates for this cohort . 29 30 31 5.2.1.4. Selection o f Critical Effect 32 There are several endpoints that are suitable for consideration for the derivation of an 33 RfC for Libby Amphibole asbestos where health effects data and exposure information are 34 available in the principal study (Rohs et ah. 2008: Lockev et ah. 1984): (1) parenchymal changes 35 viewed as small opacities in the lung; (2) blunti ng of the costophrenic angle (measured between 36 the rib cage and the diaphragm); or (3) pleural thickening (both localized and diffuse). Each of 37 these effects is an irreversible pathological lesion (ATS. 2004). As the available epidemiologic This document is a draftfor review purposes only and does not constitute Agency policy. 5-15 DRAFT--DO NOT CITE OR QUOTE 1 studies describe these endpoints as viewed on standard X-rays (see Text Box 5-1), it is important 2 to understand the distinction between what is viewed on the radiograph versus the underlying 3 biologic lesion. The following discussion reviews the health effects associated with each of 4 these radiographic abnormalities observed in workers exposed to Libby Amphibole asbestos. 5 6 Text Box 5-1. Radiographic Abnormalities of the Lung and Pleura Parenchymal changes in the lung (small opacities): The small opacities viewed within the lung (interstitial changes) are indicative of pneumoconiosis and are associated with exposure to not only mineral fibers, but also mineral dust and silica. The radiographic signs of pneumoconiosis begin as small localized areas of scarring in the lung tissue and can progress to significant scarring and lung function deficits. The ILO standards provide a scheme for grading the severity of the small opacities; the size, shape, and profusion of the small opacities are recorded, as well as the affected zone of the lung (ILO, 2002) . Obliteration of the costophrenic angle: The costophrenic angle (CPA) is measured as the angle between the ribcage and the diaphragm on a posterior anterior-viewed radiograph (the costophrenic recess). When CPA blunting or obliteration is noted on a radiograph, it is recorded as present or absent (ILO, 2002). Obliteration of the CPA may occur in the absence of other radiographic signs. Pleural thickening: The pleural lining around the lungs (visceral pleura) and along the chest wall and diaphragm (parietal pleura) may thicken due to fibrosis and collagen deposits. Pleural thickening (all sites) is reported as either localized pleural thickening (LPT) or diffuse pleural thickening (DPT). DPT of the chest wall may be reported as in-profile or face on, and is recorded on the lateral chest wall "only in the presence o f and in continuity with, an obliterated costophrenic angle" (ILO, 2002). Localized pleural thickening may also be viewed in-profile or face-on and is generally a pleural plaque (parietal). Calcification is noted where present (ILO, 2002). 7 8 9 5.2.2. Evaluation of Radiographic Lesions as Potential Critical Effects 10 5.2.2.1. Health Effects o f Parenchymal Changes as Small Opacities Viewed on Standard 11 Radiographs 12 Radiographic evidence of small opacities in the lung is evidence of tlbi otic scatring of 13 lung tissue consistent with mineral dust and mineral fiber toxicity. The scarring of the 14 parenchymal tissue of the lung contributes to measured changes in pulmonary function, 15 including obstructive pulmonary deficits from narrowing airways, restrictive pulmonary deficits 16 from impacting the elasticity of the lung as well as decrements in gas exchange. However, 17 although data across the mineral fiber literature strongly support a finding of functional deficits 18 where small opacities are visible on radiographs, the data also indicate that deficits in pulmonary 19 function (consistent with interstitial fibrosis) are seen before these changes are detected by This document is a draftfo r review purposes only and does not constitute Agency policy. 5-16 DRAFT--DO NOT CITE OR QUOTE 1 radiographic examination. Thus, changes in lung function may occur before the fibrotic lesions 2 can be detected on standard radiographs (ATS. 2004; Broderick et al.. 1992). For example, 3 decreased Carbon monoxide (CO) diffusion is a sign of reduced gas exchange in the pulmonary 4 region of the lung and is observed in workers exposed to other types of asbestos even when small 5 opacities are absent on radiographs. Similarly, obstructive deficits in lung function may be 6 observed without radiographic signs for fibrotic lesions of small opacities. As decreased 7 diffusion and obstructive deficits are mechanistically linked to changes in the parenchymal tissue 8 these data suggest radiographs may not be sensitive enough to detect and protect against small 9 localized lesions in parenchymal tissue of the lung. Radiographic evidence of small opacities 10 indicates interstitial damage of the lung paremchyma, is associated with decreased pulmonary 11 function and considered evidence of an adverse health effect. Thus, small opacities are an 12 appropriate endpoint for RfC derivation. However, as there is evidence of functional changes in 13 lung function from lesions not detectable on conventional radiographs, more sensitive endpoints 14 should be considered. 15 16 S.2.2.2. Health Effects o f Diffuse Pleural Thickening (DPT) Viewed on Standard 17 Radiographs 18 DPT is a fibrotic lesion (often described as a basket weave of collagen) in the visceral 19 pleura that encases each lobe of the lungs. The fibrotic lesion restricts the ability of the lung to 20 expand mechanically, as well as by reducing the available volume (where thickening has 21 progressed) (Jones et al.. 1988) and DPT is strongly associated with reduced lung function (ATS. 22 2004). There are consistent reports of impaired lung function associated with DPT in 23 asbestos-exposed populations (Broderick et al.. 1992; Kilburn and Warshaw. 1991; Bourbeau et 24 al,, 1990). A cross-sectional study of men (n = 1,298) exposed to asbestos through various 25 trades (e.g., boiler makers, welders, plumbers/pipefitters) included chest radiographs and 26 spirometry (Kilburn and Warshaw. 1991). When considering the effect of DPT (with 27 costophrenic angle [CPA] blunting) on radiographic function, FVC, FEV1, and FEF25-7526were 28 all significantly reduced (85, 79, and 66% of predicted values, respectively) as compared with 29 individuals with calcification or plaques only in men with no signs of small opacities (ILO 26 Forced Vital Capacity (FVC); Forced Expiratory Volume in 1 second (FEV1) and Percent FVC (FEV%) = [(100 x FEV1) - FVC, FEF25-75. is the expiratory flow between 25% and 75% o f the FEV.] This document is a draftfor review purposes only and does not constitute Agency policy. 5-17 DRAFT--DO NOT CITE OR QUOTE 1 profusion score of 0/0 or 0/1) (p < 0.0001). The relationship between pleural fibrosis and FVC 2 was studied in asbestos-exposed sheet metal workers (N = 1,211) where not only the type of 3 thickening (discrete versus diffuse) (ILO. 1980) but also CPA involvement and the location of 4 the thickening were taken into consideration (Broderick et ah. 1992). Univariate analysis 5 indicated FVC was decreased by both DPT (with CPA blunting) and circumscribed thickening, 6 diaphragm involvement, CPA involvement, and the extent of the thickening (Broderick et ah. 7 1992). Multivariate linear regression, allowing for control of potential confounders, found 8 decreased FVC was significantly related to DPT, plaques, CPA involvement, and extent of the 9 thickening, but not diaphragmatic involvement (Broderick et al.. 1992). 10 The mechanisms for reduced lung volume in individuals with asbestos-related DPT have 11 been examined by measuring lung function and changes in diaphragm length, rib-cage 12 dimensions, and subphrenic volume in 26 patients during breathing (Singh et aU 1999). DPT 13 reduced both total lung capacity and FVC with corresponding decreases in rib-cage expansion 14 and movement of the diaphragm, consistent with the restrictive nature of these lesions, which 15 may encase part of the lung (Singh et aU 1999). These direct measurements of the effect of DPT 16 chest wall and diaphragmatic motion illustrate the role of DPT in reducing lung volume, 17 contributing to restrictive deficits in pulmonary function. Taken together, the epidemiologic 18 evidence and the mechanistic information that support a restrictive effect of fibrotic lesion in the 19 visceral pleura, substantiate the associations between DPT and decreased pulmonary function. 20 As such, the observation of DPT on standard radi ographs is representative of pathological 21 changes directly related to reduced lung function and is, therefore, an indication of adversity, 22 and, can serve as an appropriate health endpoint for consideration in RfC derivation. 23 24 5.2.2.3. Health Effects o fLocalized Pleural Thickening (LPT) Viewed on Standard 25 Radiographs 26 Localized pleural thickening (LPT) viewed on a standard radiograph may include both 27 pleural plaques and pleural thickening that does not involve blunting of the costophrenic angle 28 (ILO. 2002). Thus, both parietal plaques and localized thickening of the visceral pleura may be 29 designated as LPT. Thickening of the parietal pleura is due to an acellular collagen plaque 30 (basket weave of collagen fibers) between the parietal pleura and the ribcage (or along the 31 diaphragm) often described as discrete or circumscribed pleural plaques (ATS. 2004; Jones. This document is a draftfor review purposes only and does not constitute Agency policy. 5-18 DRAFT--DO NOT CITE OR QUOTE 1 2002). Thickening of the visceral pleural is a fibrosis with diffuse borders and may extend into 2 the lung parenchyma (ATS. 2004; Jones. 2002). The pathology and health effects of the 3 different lesions are evaluated here in the characterization of the health significance of LPT. 4 Costal parietal plaques occur between the thoracic cage and parietal pleura, which is 5 normally adherent to the thoracic cage (ATS. 2004; Jones. 2002). Costal parietal plaques have 6 been described as collagen deposits with ragged irregular edges and up to 1 cm in depth and may 7 be calcified. These parietal plaques have been associated with constricting pain in the thoracic 8 cavity (Mukheriee et al.. 2000). The parietal pleura is well innervated by the intercostal and 9 phrenic nerves and is considered very sensitive to painful stimuli (Jones. 2002). With respect to 10 parietal plaques, pain during exertion or exercise could result in restrained chest wall motion 11 during exertion or exercise. Thus, Bourbeau et al. (1990) hypothesized that the dyspnea and 12 changes in pulmonary function noted in individuals with pleural plaques may be due to physical 13 irritation and perhaps a constricting action where parietal plaques are well progressed or 14 numerous and impact a large proportion of the parietal surface. 15 Kouris et al. (1991) examined the presence of dyspnea, and measures of pulmonary 16 function (i.e., FVC, FEV1, and FEV%27) in asbestos-exposed workers (n = 913) in relation to 17 radiographic signs of lung and pleural anomalies. Radiographs were contemporary to the study 18 and read in accordance with ILO (1980) guidelines. Pleural plaques were associated with 19 reduced FVC and FEV1.0 (87.6% and 84.1% of predicted, respectively,/? < 0.0005), although 20 deficits associated with diffuse thickening were greater (76.4% and 73.9%, p < 0.0005) (Kouris 21 et al. 1991). Correspondingly odds ratios for decreased FVC and FEV1.0 (80% decrement) 22 were increased by the presence of both plaques and diffuse thickening (1.5 for plaques and 23 4.2 and 4.7 for diffuse thickening, respectively). Interestingly, when history of lung disease was 24 considered, pleural plaques had a greater effect in individuals without previous lung disease 25 (OR of 2.1 for FVC and 1.7 for FEV1.0). 26 Pleural thickening in general is associated with decreased pulmonary function (Petrovic 27 et al... 2004; Wang et al.. 2001; Miller et al.. 1994)and this association is strengthened as the 28 severity of the pleural thickening increases (Lilis et al.. 1991). Few available studies have 29 examined the relationship between pleural plaques identified on standard radiographs (ILO. 'Forced Vital Capacity (FVC); Forced Expiratory Volume in 1 second (FEV1) and Percent FVC (FEV%) = [(100 x FEV1) *FVC]. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-19 DRAFT--DO NOT CITE OR QUOTE 1 1980) and pulmonary function without including DPT in the analysis and adequately controlling 2 for the presence of small opacities (indicative of parenchymal damagejf8. 3 Lilis et al. (1991) examined pulmonary function in long-term asbestos insulation workers, 4 and found that one measure (FVC) decreased significantly as the severity of pleural fibrosis (all 5 types, as indicated by a pleural index) increased. This decrease was more dramatic when 6 including parenchymal changes (small opacities) or if DPT was viewed separately. A second 7 analysis focusing on participants with pleural plaques found an inverse relationship between 8 severity of the pleural plaques and FVC (j? < 0.0001), when adjusting for the independent effects 9 of duration, smoking and presence of small opacities (Lilis et al.. 1991). This finding supports a 10 view that pleural plaques, when extensive, may contribute to restrictive lung deficits, but the 11 analysis included individuals with known small opacities (e.g., lung fibrosis). The authors do not 12 address the potential that the pleural index may also correspond to increased severity of 13 parenchymal changes, potentially confounding the analysis where accounting for small opacities 14 (profusion scores of 1/0 or greater) may not adequately control for asbestos-related parenchymal 15 damage 16 Oliver et al. (1988) studied the relationship between pulmonary function and pleural 17 plaques in asbestos-exposed railway workers (n = 383). Case selection included exclusion of 18 workers with DPT (1LQ. 1980) and exclusion of any indication of small opacities (only 19 profusion scores of 0/0 were included). Standard spirometry was conducted to evaluate 20 restrictive and obstructive pulmonary deficits. Additionally, single-breath diffusing capacity 21 (DLCO) was measured which would indicate parenchymal defects. The DLCO was similar in 22 subjects with and without circumscribed plaques, suggesting little or no subradiographic 23 parenchymal damage, which corresponded to the presence of pleural plaques. Pleural plaques 24 were associated with both decreased FVC and pulmonary restriction (p = 0.03 and 0.04, 25 respectively) where the diagnostic certainty for the plaques was considered `definite', and there 26 was an association between level of diagnostic certainty and these pulmonary deficits (p = 0.02) 27 (Oliver et al.. 1988). Quantitative pleural score, based on the number and extent of plaques, was28 28It is difficult to control for effects subradiographic parenchymal fibrosis on lung function, where it may not have progressed to visible small opacities, and it has been suggested that reduced lung fund ion. which has been associated with circumscribed plaques in some studies, may be reflecting the effects o f subradiographic parenchymal changes, rather than a direct effect o f DPP (ATS. 2004: Erdinc et al.. 2003 : Miller and Zurio, 1996: Broderick et al.. 1992). This document is a draftfor review purposes only and does not constitute Agency policy. 5-20 DRAFT--DO NOT CITE OR QUOTE 1 also associated with decreased FVC and pulmonary restriction (p = 0.0135 and 0.0126, 2 respectively) (Oliver et aE 1988). Of the available studies that assess pleural thickening with 3 standard radiographs, this study best controls for the possibility of subradiographic parenchymal 4 damage and is, therefore, strong evidence that circumscribed pleural plaques independently 5 impact pulmonary' function. The observed restrictive pulmonary deficit is consistent with the 6 potential for pleural plaques to restrict chest wall motion or the elasticity of the diaphragm. 7 Three high-resolution computed tomography (HRCT) studies were conducted specifically 8 to assess the potential for parietal plaques to impact lung function. Staples et al. (1989) report no 9 difference in lung function or diffusing capacity between participants (n = 76) with and without 10 pleural plaques. Soulat et al (1999) found no difference in FEV1 or FVC between 11 asbestos-exposed insulators with (n = 84) and without (n = 51) pleural plaques in the absence of 12 any parenchymal changes. As severity of pleural thickening has been shown to be positively 13 associated with decrease measures of pulmonary functi on, Van Cleemput et al. (2001) not only 14 examined the effect of HRCT defined pleural plaques on pulmonary function, but also assessed 15 the extent of the pleural plaques. Neither the presence nor extent of pleural plaques were 16 associated with lung function parameters (diffusing capacity or normalized spirometric values) 17 (van Cleemput et al.. 2001). Where pleural plaques and diffuse thickening (visceral pleura) were 18 both identified by F1RCT and correlated to pulmonary function, diffuse visceral thickening--but 19 not plaques--were associated with decreased lung volume and FVC (Copley et al.. 2001). 20 Although CPA involvement was not independently assessed, several scoring systems for severity 21 were compared which included CPA involvement, and as in other studies, increased severity 22 correlated to greater decrements. 23 The mechanisms for reduced lung volume in individuals with asbestos-related pleural 24 plaques and DPT have been examined by measuring lung function and changes in diaphragm 25 length, rib-cage dimensions and subphrenic volume in 26 patients during breathing (Singh et al.. 26 1999). Pleural plaques alone did not reduce any of the measures of lung function in this study, 27 but there were indications of reduced diaphragm movement (Singh et al.. 1999). This may be an 28 indication that diaphragmatic plaques in the parietal pleura have the potential to attenuate the 29 movement of the diaphragm during breathing. Because this study is relatively small (N= 26) 30 and a distinction was not made between costal and diaphragmatic plaques by the study authors, This document is a draftfor review purposes only and does not constitute Agency policy. 5-21 DRAFT--DO NOT CITE OR QUOTE 1 additional work is needed to better understand the direct effects of pleural plaques on lung 2 function. 3 Although some researchers have questioned that pleural plaques alone directly impact 4 pulmonary function, a critical review of the literature from 1965-1999 concludes: "1) 5 Individuals with asbestos-induced pleural plaques may have alterations in pulmonary function 6 and /or clinical symptoms that are independent of smoking and radiographic parenchymal 7 fibrosis and, 2) the respiratory changes dues to asbestos-induced pleural plaques are generally 8 less severe than those caused by pleural thickening" (Rockoff et aL 20021. Therefore, although 9 the evidence is mixed, pleural plaques may be independently associated with reduced pulmonary 10 function. 11 No studies correlating pulmonary function to radiographic signs of localized pleural 12 thickening (LPT) using the 1LO (ILO. 2002) guidelines could be located. However, several 13 researchers employed similar classification schemes, modifying earlier ILO classification 14 systems, such that DPT was diagnosed only in conjunction with blunting of the CPA. This 15 modification potentially includes cases of diffuse pleural thickening (without CPA blunting) in 16 their analysi s of pleural plaques, making their findings somewhat applicable to the current 17 classificati on of LPT (Garcia-Closas and Christianh 1995: Broderick et al.. 1992). Pleural 18 thickening (without CPA blunting) was associated with mixed respiratory impairment in a study 19 of asbestos-exposed construction carpenters (n = 631) (OR of 3.7 [95% Confidence Interval (Cl): 20 1.4-12.3]) but was only weakly associated when the outcome was restrictive deficit specifically 21 (1.3 [95% CL 0.4-3.9]) (Garcia-Closas and Christianh 1995). Broderick et al. (1992) found 22 decreased FVC was not only significantly associated with "diffuse thickening" (with CPA 23 blunting) but also with "pleural plaques" (which included all pleural thickening without CPA 24 blunting). The severity of pleural thickening (both as width or percentage of lateral wall) and 25 calcification was associated with reduced FVC as well (Broderick et al.., 1992). Kilburn and 26 Warshaw (1991) assessed pulmonary function in individuals with "plaques only," "diffuse 27 thickening only," and "diffuse thickening with CPA blunting," showing progressive deficits 28 across these categories in FVC, FEV1, and mid-expiratory flow (e.g., FEV1: 90.5, 86.2, and 29 49.4% \p < 0.05], respectively). Again, there is a trend that diffuse thickening has a greater 30 impact on lung functi on parameters, although an independent effect of plaques cannot be ruled 31 out by these data. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-22 DRAFT--DO NOT CITE OR QUOTE 1 In summary, the radiographic classification of localized pleural thickening (LPT) under 2 current ILO guidelines may include both parietal plaques (in the pleura lining the interior of the 3 ribcage) and diffuse visceral thickening (without CPA obliteration) (ILO, 2002). The two 4 lesions (parietal plaques and localized visceral thickening) are distinct and may contribute 5 independently to observed health effects. Parietal plaques are known to induce chronic 6 constricting chest pain that increases in severity as the extent of the plaques increases. Pleural 7 thickening in general is associated with reduced lung function parameters with increased effect 8 correlating with increased severity of the pleural thickening (Petrovic et ah, 2004; Wang et ah, 9 2001; Miller et al.. 1994; Lilis et ah, 1991). There is clear evidence from HRCT studies that the 10 presence and extent of visceral thickening does impair lung function, although, when evaluated 11 independently, parietal plaques were not statistically correlated with decreased pulmonary 12 function (Copley et al,, 2001; Schwartz et al,, 1993). Specifically considering the designation of 13 LPT, lung function impairment has been demonstrated in several studies where pleural 14 thickening without CPA involvement has been studied (Garcia-Closas and Christiani, 1995; 15 Broderick et al., 1992; Kilburn and Warshaw, 1991). Thus, the radiographic classification of 16 localized pleural thickening (LPT) (ILO, 2002) includes pleural lesions associated with chronic 17 chest pain, decreased lung volume, and decreased measures of lung function. Therefore, EPA 18 considers LPT an adverse effect and an appropriate endpoint for RfC derivation. 19 20 5.2.3. Methods of Analysis 21 5.2.3.1. Exposure Data and Choice o f Exposure Metric 22 EPA collaborated with a research team at the University of Cincinnati to update the 23 exposure reconstruction for use in the job-exposure matrix (JEM) for all workers in the 24 Marysville, OH cohort, taking into account additional industrial hygiene data that were not 25 available for previous studies conducted in this cohort. As discussed in detail in Appendix F, 26 exposure estimates for each worker in the O.M. Scott Marysville, OH plant were developed 27 based on available industrial hygiene data from the plant. Figure 5-1 shows the average 28 exposure concentrations of fibers in air (PCM fibers/cc)29of each department from 1957 to 2000, 29 29PCM. where fibers are viewed and counted by light microscopy, does not identity the composition o f the fiber. Thus, the mineralogy o f fibers identified under PCM cannot be determined. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-23 DRAFT--DO NOT CITE OR QUOTE 9.0 8.0 Estimated Measured 1 2 3 Figure 5-1. Estimated and measured exposure concentrations in Marysville, 4 OH facility3 5 6 aTrionizing is a term used in the Maiysville, OH facility and includes unloading o f rail cars 7 containing venniculite ore (track), using conveyers to move the vermiculite ore into the expander 8 furnaces, separation o f the expanded vermiculite from sand, blending in of lawn care chemicals, 9 and drying and packaging o f the final product. As no unexpanded ore was used in pilot plant, 10 research, polyfonn, office, packaging, or warehouse, jobs in these categories were assigned as 11 background. Workers assigned to plant maintenance activities spent 50% o f their time in 12 trionizing areas and 50% o f their time in areas assigned as plant background. Workers assigned to 13 central maintenance spend 10% o f their time in trionizing areas and 90% o f their time in areas 14 assigned as plant background. Central maintenance jobs were eliminated in 1982 and contracted 15 out (see Appendix F). 16 17 18 This document is a draftfor review'purposes only and does not constituteAgency policy. 5-24 DRAFT--DO NOT CITE OR QUOTE 1 indicating the time periods when fiber measurements were not available (`Estimated') and were 2 available (`Measured'). 3 In brief, the starting point for the JEM was the measured or estimated concentration of 4 fibers in air (fibers/cc) of each department from 1957-2000. The distribution of exposure by 5 department is summarized in Figure 5-1. Using available data on the year of hire and the 6 departments in which each person worked, the cumulative exposure (fibers/cc-year) for each 7 worker for each year since the date of hire was estimated. Each worker's cumulative exposure 8 was then adjusted to a cumulative human equivalent exposure for continuous exposure (CHEEC; 9 fibers/cc-year) to represent exposure 24 hours/day and 365 days/year (assuming that any 10 exposure off site was zero) for the full duration of employment. Adjustments for different 11 inhalation rates in working versus nonworking time periods were incorporated in this analysis. 12 The calculated value is similar to what EPA usually refers to as continuous human equivalent 13 exposure (US. EPA, 1994b). These calculations are somewhat more complex than the usual 14 conversions to equivalent continuous exposure concentrations that EPA makes in the analysis of 15 occupational studies. Conversions for noncancer effects are usually made using an adjustment 16 factor of 240 days 365 days * 10 n r 20 m3(ELS. EPA, 1994b). However, the adjustment 17 factor in this current assessment takes into account the extensive seasonal overtime for some job 18 codes at the Marysville facility, as well as other annual periods when work hours were reduced 19 (see Appendix F). The estimated CHEEC was used to represent Libby Amphibole asbestos 20 exposure in all subsequent analyses because it combines aspects of both intensity of exposure 21 and duration of exposure.3" For Libby Amphibole asbestos, the exposure metric is calculated as 22 cumulative exposure (fibers/cc-year). Cumulative exposure is a commonly evaluated exposure 23 metric in occupational studies, especially for mineral fibers, where fiber retention may be 24 relevant to toxicity. It should be noted that discrete parietal plaques have often been associated 25 with other exposure metrics (e.g., mean exposure, TSFE) (i.e., Paris et al, 2008; Jakobsson et aU 26 1995; Ehrlich et al., 1992; Copes et al.. 1985). Paris et al. (2008) show significant 27 exposure-response relationships for both mean and cumulative exposure metrics for pleural 28 plaques (identified by HRCT) among workers with mixed fiber exposures, when accounting for 29 age, smoking, and TSFE. Mean exposure provided a better overall fit (Paris et al.. 2009). Thus, 30 EPA has conducted an uncertainty assessment for the RfC derivation from the sub-cohort by also 3uThe University o f Cincinnati used the term CHEEC in its report (see Appendix F). This document is a draftfor review purposes only and does not constitute Agency policy. 5-25 DRAFT--DO NOT CITE OR QUOTE 1 exploring alternative methods to weight the BMCLio in units of cumulative exposure, to 2 represent the average exposure needed for RfC derivation (see Section 5.3.7). 3 Because localized pleural thickening does not generally occur immediately after exposure 4 and requires some time to develop to the state that it can be detected on a conventional chest 5 X-ray, exposures that occur close to the time of X-ray may not contribute to the occurrence of 6 observable disease and may obscure the exposure-response relationship. Accordingly, a lagged 7 exposure (i.e., cumulative exposure discounting the most recent time period) may be the most 8 appropriate measure to use. Therefore, exposure estimates with various lags were investigated 9 (lags of 0, 5, 10, 15, and 20 years). For example, a CHEEC value based on a lag of 5 years 10 excludes all exposures that occurred within 5 years of the date of X-ray. Looking at the 11 occurrence of the outcome for various categories of time elapsed since first exposure, the first 12 localized pleural thickening was detected ~10 years after the first exposure. 13 14 S.2.3.2. Data Sets for Modeling A nalyses 15 The individual health outcome data for all workers who participated in the Lockey et al. 16 (1984) study and the follow-up study by Rohs et al. (2008) were used for exposure-response 17 modeling. To avoid any bias from previous occupational exposure to asbestos, only the data 18 from those who did not report any previous occupational exposure to asbestos were used. The 19 data from Lockey et al. (1984) and Rohs et al. (2008) were combined for the full cohort to 20 provide a greater range in time from first exposure (described below). Outcome assessments, 21 i.e., chest X-rays, were performed at two different time points, 1980 and 2002-2005. While the 22 evaluation approaches were generally similar (independent readings by three certified 23 B-readers), it is important to note that X-ray readings were performed by different individuals, 24 under a different reading protocol in 1980 (modified 1971 ILO standards) compared to 2000s 25 [ILO (2002) standards], leading to some uncertainty in stati stical analyses that combine these 26 data sets. An additional consideration is human body composition--in some cases, difficulty in 27 distinguishing fat pads from true pleural thickening may lead to misclassification of the outcome. 28 BMI measurements are available for the latter study but not for the 1980 evaluation; the effect of 29 BMI was investigated and is discussed below. 30 Radiographs were evaluated by two B-readers with a consensus evaluati on by a third 31 reader in the case of disagreement in the original study by Lockey et al. (1984). In the follow-up This document is a draftfor review purposes only and does not constitute Agency policy. 5-26 DRAFT--DO NOT CITE OR QUOTE 1 by Rohs et al. (2008), a radiographic reading was considered positive "when the median 2 classification from the three independent B readings was consistent with pleural and/or 3 interstitial changes" (p. 631). Because the ILO criteria were updated in 2000, the reader forms 4 from Lockey et al. (1984) showing pleural changes were evaluated for consistency with the ILO 5 2000 criteria. This reevaluation did not result in any change in the diagnosis for any individual 6 from the 1980 reading.31 In addition, no difference in reported X-ray quality was noted between 7 the Lockey et al. (1984) data and the follow-up by Rohs et al. (2008). 8 The full data set of the exposure-response relationship for localized pleural thickening 9 was as follows. The radiographic data from Lockey et al. (1984) (n = 513) and Rohs et al. 10 (2008) (n = 280), were combined for a total of 793 X-ray evaluations (this includes repeated 11 X-rays on the same individual). X-rays obtained from workers who reported exposure to 12 asbestos at other locations were excluded from consideration (n = 793 - 105 = 688 X-ray 13 evaluations). 14 For workers who were X-rayed in both Lockey et al. (1984) and Rohs et al. (2008). one 15 of the observations was excluded so that there were no repeat observations for individual 16 workers in the data set used for modeling. For workers who were negative for localized pleural 17 thickening in Lockey et al., the (1984) study data were excluded, and the Rohs et al. (2008) data 18 were retained. For workers who were positive for localized pleural thickening in Lockey et al. 19 (1984) and also in Rohs et al. (2008). the 1984 study data were retained. One worker was 20 positive in 1984 and negative in 2008 (removing this worker from the analysi s did not change 21 results). The 2008 study data were retained for this worker. This procedure resulted in n = 688 22 X-rays - 252 duplicates = 436 X-rays, representing 436 individual workers. 23 Two workers from Lockey et al. (1984) were excluded because the start day and the 24 X-ray date were the same (n = 436 - 2 = 434). For each worker, the estimated cumulative 25 exposure corresponded to the date of the X-ray retained for analysis--if the 1980 X-ray was 26 used, the individual's cumulative exposure estimate covered the period from start of work 27 through the X-ray date in 1980. If the 2002-2005 X-ray was used, cumulative exposure covered 28 the period from start of work through the date ofjob stop or 2000, whichever occurred earlier. 31Personal communication (e-mail) from Dr. James Lockey, University o f Cincinnati, to Dr. Robert Benson in March 2011 reports that a review o f the 1980 B-reader forms using the ILO 2000 guidelines would not result in changes in individual diagnosis for study participants. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-27 DRAFT--DO NOT CITE OR QUOTE 1 The Marysville cohort data comprise 434 workers who were not previously exposed to 2 asbestos and had at least one X-ray observation. Because the concentration of Libby Amphibole 3 asbestos in workplace air was estimated rather than measured for all years prior to 1972, this data 4 set was stratified into two subsets: (1) workers hired in 1972 or after (for whom all exposure 5 values are measured), and (2) workers hired before 1972 (for whom some of the exposure values 6 are estimated). Distributions of cases and TSFE (7) at each outcome assessment are shown in 7 Table 5-3. 8 9 10 Table 5-3. Distribution of cases and time from first exposure (T) for cohort 11 of Marysville workers 12234567 All participants3 First exposed before 1972 First exposed 1972 or later Cases/Total Range of T Cases/Total Range of T Cases/Total Range of T Examined 1980 (Lockev et al., 1984) 5/434 0.42-23.43 4/236 8.75-23.43 1/198 0.42-8.42 Examined 2002-2005 (Rohs et al.. 2008) 57/252 23.14-47.34 45/133 31.07-47.34 12/119 23.14-32.63 Marysville cohort (n = 434, examination in either 1980 or 2002-2005) 61/434 0.42-47.34 48/236 8.75-47.34 13/198 0.42-32.63 13 14 aThe 252 individuals examined in 2002-2005 were also examined in 1980. Note that there were originally 15 513 individuals in the Lockey et al. (1984) cohort; o f these, 77 had previous asbestos exposure and were excluded 16 (;? = 436). Two individuals were excluded because their X-ray date was the same as their employment start date 17 (;? = 434). These exclusions are also reflected in the Rohs et al. (2008) cohort. 18 19 Source: Rohs et al. (2008) and Lockey et al. (1984). 20 21 22 The more accurate exposure data are considered to be those from 1972 and later, as these 23 data were based on analytical measurements. Due to the longer follow-up time and additional 24 covariate information, the most informative outcome data come from the 2002-2005 25 examinati on. Based on these consi derations, a sub-cohort of the Marysville workers, which 26 includes data from workers in the 2002-2005 examination, and who began work in 1972 or later 27 This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-28 DRAFT--DO NOT CITE OR QUOTE 1 (12 cases of localized pleural thickening and 106 unaffected individuals32) (Rohs et al., 2008), 2 was chosen as the preferred analysis to develop a point of departure (POD) for localized pleural 3 thickening to serve as the basis for the RfC. Additionally, sample POD estimates based on 4 statistical analyses of results from the full cohort [Lockey et al. (19841 and Rohs et al. (2008) 5 combined, as described above] were included for comparison. 6 7 5.2.3.3. Statistical Modeling o f the Sub-cohort 8 EPA performed analyses of study results for the sub-cohort whose exposures began on or 9 after 1/1/1972 when workplace PCM measurements were available, reducing uncertainties 10 associated with exposure assessment. Localized pleural thickening (LPT), as diagnosed from a 11 standard radiograph (ILO, 2002). was selected as the critical effect based on the health effects 12 associated with pleural thickening specific to this diagnosis (see Section 5.2.2.3). Alternative 13 critical effects were not considered for the sub-cohort analysis given the limited number of cases 14 (one case of DPT and no cases of small opacities). Epidemiologic methods were used to analyze 15 the exposure-response data, and benchmark concentration (BMC) methodology was used to 16 estimate PODs. In this approach, the available data are fit to a set of mathematical 17 exposure-response models to determine an appropriate empirical representation of the data. 18 General model fit is evaluated to determine whether the model form appropriately represents the 19 data; here, this was done using the Hosmer-Lemeshow test (a form of the Pearson %9 20 goodness-of-fit statistic). Among models with adequate general fit, a recommended model form 21 is then determined; commonly, this is the model with the best fit as measured by Akaike's 22 Information Criterion (AIC) value among these model forms judged to provide an appropriate 23 and statistically adequate representation of the data. For inhalation data, the BMC is defined as 24 the exposure level, calculated from the best-fit model, which results in a specified benchmark 25 response (BMR). The RfC is derived from the lower 95% confidence limit of the BMC, referred 26 to as the BMCL, which accounts for statistical uncertainty in the model fit to the data. All 32There was one individual whose radiographic examination indicated diffuse pleural thickening, who was excluded from further analyses o f the preferred sub-cohort. Diffuse pleural thickening represents a more severe outcome than the selected critical effect o f LPT-- including this individual as a case would not be appropriate given that the critical effect is selected to represent a most sensitive endpoint, and the subsequent selection o f a benclnnark response in modeling efforts. Diffuse pleural thickening is considered separately as an endpoint (with appropriate benchmark response) in sensitivity analyses o f alternative outcomes in the larger group o f workers examined in 2002-2005 (see Section 5.3.8). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-29 DRAFT--DO NOT CITE OR QUOTE 1 analyses were performed using SAS statistical software v. 9.1. BMCLs were obtained by the 2 profile likelihood method as recommended by Crump and Howe (1985) using the NLMIXED 3 (nonlinear mixed modeling) procedure in SAS (Wheeler. 2005) (see Appendix E for details). 4 For models where a background parameter is included, a 1% risk of localized pleural 5 thickening was assumed. Establishing a background rate for LPT prevalence is problematic for 6 several reasons. Little data exist to define background rates for LPT, as this designation is more 7 recent, and the majority of the published data use earlier ILO guidelines, which define discrete 8 pleural plaques (DPP). Secondly, it is difficult to define a population without exposure to 9 asbestos in any setting. As environmental and community exposures can increase pleural 10 thickening (Weill et al.. 2011; Luo et al. 2003; Hiraoka et al. 1998; Zitting et al.. 1996) the 11 question arises, Is there a true background rate? Also, in general, pleural thickening increases 12 with both age and TSFE in a population. There is a study that reports the LPT in Libby 13 community members with no reported pathways of exposure (Weill et al.. 2011). LPT 14 prevalence is reported at 0.4% in participants age 25-40, and 1.4% in participants age 41-50 15 (based on X-rays taken in 2000). Older study participants (61-90) had a LPT prevalence of 16 12.7%, likely influenced by high historical exposures, as well as the increased TSFE. In two 17 studies of persons not known to be previously exposed to asbestos, Anderson et al. (1979) and 18 Castellan et al. (1985) report DPP estimated prevalence of 1.2% (4/326) and 0.2% (3/1,422), 19 respectively. In cross-sectional studies, which may include persons with occupational exposure 20 to asbestos, Rogan reported DPP prevalence estimates of 1.2% in the National Health and 21 Nutrition Examination (NHANES) I study (1971-1975) (Rogan et al.. 1987) and 3.9% in the 22 NHANES II study (Rogan et al.. 2000). Among military populations, two studies have reported 23 an estimated DPP prevalence of 2.3% (Muller et al.. 2005; Miller and Zurlo. 1996). Based on 24 these reports, the 1% background rate was chosen as representing the prevalence among persons 25 without occupational exposure to asbestos in the age range of the Rohs et al. (2008) study 26 population. As there is some uncertainty regarding the true background rate for LPT, a 27 sensitivity analysis was performed where the model includes the background rate as an estimated 28 parameter rather than using the set value of 1%. There was little change in the resulting model 29 fits or BMCLs (see Section 5.3.4). 30 In the absence of agent-specific information to assist in identifying a BMR, a 10% extra 31 risk was judged to be a minimally biologically significant level of change, and is also This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-30 DRAFT--DO NOT CITE OR QUOTE 1 recommended for standard reporting purposes (U.S. EPA, 2000a). LPT is an irreversible 2 pathological change and associated with health effects including chronic pain, dyspnea, and 3 deficits in pulmonary function (see Section 5.2.2.3). The likelihood and severity of these health 4 effects increases with increased extent and severity of the pleural thickening. However, as the 5 data from the critical study do not provide information on the severity of the lesions, we cannot 6 assess the relative likelihood of any of these health effects. Thus, the observed LPT prevalence 7 may include a range of lesions from minimally adverse to severe. The biology of more severe 8 lesions (i.e., DPT and small opacities) could justify lower BMRs; however, there are not enough 9 cases to model these endpoints in this sub-cohort. A sensitivity analysis was conducted using the 10 data set included in Rohs et al. (2008) to examine the impact of choice of BMR and critical 11 effect on the POD (see Section 5.3.8). 12 13 5.2.3.3.I. Statistical model evaluation and selection 14 Dichotomous statistical models describing the probability of individual response as a 15 function of cumulative exposure (represented by CHEEC in units of fibers/cc-year) were used. 16 In order to investigate the key explanatory variables for analysis, a forward-selection process was 17 used to evaluate the association of each of the potential covariates with the risk of localized 18 pleural thickening, controlling for Libby Amphibole asbestos exposure. Covariates considered 19 for inclusion in the model were TSFE (7), age at X-ray, gender, smoking history, and BMI. This 20 initial modeling was done using a standard logistic regression model, as is commonly applied in 21 analysis of epidemiological data. The base model was a logistic regression model with 22 cumulative Libby Amphibole asbestos exposure (natural log transformed) as the independent 23 variable. This model provided an adequate fit to the data (Hosmer-Lemeshow/i-value of 0.64), 24 and the exposure variable was statistically significantly associated with the outcome 25 (beta = 0.5676, standard error, [SE] = 0.2420 increase in log odds for every unit increase in 26 CHEEC, /i-value = 0.02). Covariates were evaluated according to whether inclusion of the 27 covariate improved model fit as assessed by the AIC, and stati stical significance of the covariate. 28 When controlli ng for Libby Amphibole asbestos exposure, none of these covariates were 29 associated with odds of localized pleural thickening: T: /i-value = 0.89; age at X-ray: 30 /i-value = 0.77; gender: /i-value = 0.78; smoking history: /i-value = 0.17; BMI: /i-value = 0.41. 31 The inclusion of each of the covariates with the exception of smoking increased the AIC for the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-31 DRAFT--DO NOT CITE OR QUOTE 1 model, and the improvement in model fit with the addition of smoking was marginal (decrease of 2 0.1 AIC units). Therefore, only cumulative Libby Amphibole asbestos exposure (CHEEC) was 3 included in further analyses, although sensitivity analyses were performed to investigate the 4 potential impact of smoking (see Section 5.3.6 and Appendix E). 5 The candidate models (see Table 5-4 for model forms) were logistic (with CHEEC 6 considered as continuous, and continuous with a natural logarithm transformation), probit (with 7 CHEEC considered as continuous, and continuous with a natural logarithm transformation), 8 3-parameter log-logistic, dichotomous Hill, and dichotomous Michaelis-Menten models (with 9 only CHEEC for the latter three models). These are statistical models used to evaluate 10 dichotomous data that were considered appropriate here given the supralinear nature of the 11 observed relationship between Libby Amphibole asbestos exposure and prevalence of localized 12 pleural thickening. For each of the candidate models, exposure lags of 0, 5, 10, 15, and 20 years 13 were investigated. Although zero lag exposures are not likely to be biologically relevant (i.e., 14 some lag is expected for development of LPT), these models were included for completeness and 15 for comparison of relative model fits. Similarly, although we explored models with exposure 16 lagged by 20 years, there were cases of localized pleural thickening in the full cohort with fewer 17 than 20 years since first exposure; therefore, using such a long lag (which necessitates the 18 assumption that these are background cases) was not judged to be appropriate, and the results are 19 not further considered; these models are indicated by gray shading in Table 5-4. Further details 20 of these analyses are included in Appendix E. 21 All of the candidate models had adequate fit. Models were compared using the AIC-- 22 values were quite similar among the candidate models, ranging from 74.0 to 77.8 (see 23 Table 5-4). The model with the lowest AIC was the Michaelis-Menten model with 10-year 24 lagged exposure (AIC = 74.0). For this model form, the AIC values did not vary much for lags 25 of 5 to 15 years, but the 10-year lagged exposure provided the lowest AIC and was selected as 26 the preferred exposure metric. There were several models that had similar model fits (within 27 2 AIC units, a proximity that can be considered to be a range that cannot clearly differentiate 28 between models) (Burnham and Anderson. 2002) as the best-fitting model, including the logistic 29 and probit models with the natural log of CITEEC as the exposure metric (lags of 5, 10, and 30 15 years), the 3-parameter log-logistic model (lags of 5, 10, and 15 years), the Dichotomous Hill 31 model (lag of 10 years), and the Michaelis-Menten model with exposure lagged by 5 or 15 years. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-32 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-4. Candidate models for association between cumulative Libby 2 Amphibole asbestos exposure in the Marysville sub-cohort and localized 3 pleural thickening 4 _ M odel Exposure m etric FornT H osm er-L em e show GOF AIC /7-value BMC BMCL Logistic CHEEC P(LPT) = 1/[1 + exp(-fl -/CCHEEOI I I P 0.7423 IBS IBS CHEEC, lag 5 77.5 0.6914 1.5245 0.8836 CHEEC, lag 10 77.4 0.6751 1.4734 0.8540 CHEEC, lag 15 77.6 0.6474 1.4510 0.8242 CHEEC, lag 20 77.8 0.8800 IBS IBS Logistic CHEEC, lag 5 ln(CHEEC) P(LPT) = 1/[1 + exp{-a-b *ln(CHEEC))] 75.5 0.6537 IBS: IBS: 75.2 0.5454 0.2281 0.0601 CHEEC, lag 10 CHEEC, lag 15 74.6 0.5708 0.2028 0.0591 74.7 0.6620 0.1686 0.0463 CHEEC, lag 20 Probit model CHEEC, lag 5 CHEEC, lag 10 CHEEC, lag 15 CHEEC, lag 20 Probit model CHEEC, lag 5 CHEEC, lag 10 CHEEC, lag 15 CHEEC, lag 20 3 -parameter log-logistic CHEEC, lag 5 CHEEC, lag 10 CHEEC, lag 15 CHEEC, lag 20 5 75.4 CHEEC P(LPT)=<S>(a + /CCHEEC) 77.2 77.0 77.0 77.2 77.4 ln(CHEEC) P{LPT) = <D(fl + />*ln(CHEEC)) 76.0 75.7 75.2 75.0 75.7 ln(CHEEC) P(LPT) = bkg + (1 - bkg)/[l + exp(-o 74.9 -6*ln(CHEEC))] 74.6 74.1 74.3 75.2 0.8152 0.7698 0.7146 0.6864 0.6645 0.8884 0.6041 0.4967 0.5385 0.6166 0.7945 0.7030 0.4894 0.5853 0.7238 0.8277 1.3773 1.3336 1.3148 - 0.8481 0.8048 0.7776 - 0.2066 0.1843 0.1544 - 0.0502 0.0496 0.0441 - 0.3096 0.2696 0.2193 - 0.0979 0.0888 0.0693 - This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-33 DRAFT--DO NOT CITE OR QUOTE Table 5-4. Candidate models for association between cumulative Libby Amphibole asbestos exposure in the Marysville sub-cohort and localized pleural thickening (continued) Model Exposure Metric Form* Hosmer-Leme show GOF AIC /7-value BMC BMCL Dichotomous Hill1' ln( CHEEC) P(LPT) = bkg + (Plateau - 76.9 bkg)*CHEECb/[exp(-o) + CHEECb] 0.6040 CHEEC, lag 5 76.5 0.3598 0.3083 0.1015 CHEEC, lag 10 76.0 0.4244 0.2640 0.0923 CHEEC, lag 15 76.2 0.6659 0.2112 0.0724 CHEEC, lag 20 77.2 0.8277 - - Michaelis-Menten0 ln(CHEEC) P(LPT) = bkg + (Plateau - 74.9 bkg)*CHEEC/[exp(-) + CHEEC] 0.5243 -- -- CHEEC, lag 5 74.5 0.3351 0.3096 0.1352 CHEEC, lag 10d 74.0 0.4163 0.2642 0.1177 CHEEC, lag 15 74.3 0.5664 0.2097 0.0898 CHEEC, lag 20 76.0 0.5610 - - 1 abkg indicates background rate, fixed at 1%. 2 bFor statistical modeling, the equivalent model form was used: P(PT) = bkg + (Plateau - bkg)/[l + exp(-o - 3 b*ln(CHEEC))]. ................. 4 T or statistical modeling, the equivalent model form was used: P(PT) = bkg + (Plateau - bkg)/[l + c\p(- - 5 ln(CHEEC))]. 6 dParameter estimates for the best-fitting models are as follows: 7 intercept = -0.1801 (SE = 1.0178), plateau = 0.5577 (SE = 0.3568,/rvalue = 0.1207). 8 9 10 The range was relatively narrow among these similarly fitting models (BMCLs ranging from 11 0.0441 to 0.1352), with the lowest BMCL ~2.7 times lower than the BMCL for the 12 Michaelis-Menten model, with exposure lagged by 10 years. 13 The potential confounding effect of covariates was reexamined in the best-fitting model. 14 As in the initial assessment, after controlling for the effect of exposure (CHEEC, lagged by 15 10 years), there was no association between risk of LPT and TSFE (p-value = 0.997), age at 16 X-ray (p-value = 0.87), gender (p-value = 0.55) or BMI (p-value = 0.38), and inclusion of each 17 of these covariates increased the AIC (with the exception of BMI, due to missing information for 18 some individuals). The variable representing smoking history did not meet the alpha = 0.05 19 criterion for statistical significance (p-value = 0.08), although inclusion of this variable decreased 20 the AIC from 74.0 in the best-fitting model, to 72.3. Smoking was not considered further in the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-34 DRAFT--DO NOT CITE OR QUOTE 1 derivation of the RfC due to the lack of statistical significance at the alpha = 0.05 level. 2 However, because inclusion of the smoking variable did improve model fit, it is investigated 3 further as a sensitivity analysis (see Section 5.3.6 and Appendix E). 4 The Michaelis-Menten model using the 10-year lagged exposure had a /i-val ue for fit of 5 0.42, an AIC value of 74.0, and an estimated intercept = -0.1801 (SE = 1.0178) and plateau of 6 0.5577 (SE = 0.3568) (see Figure 5-2). This model yielded a BMCio of 0.2642 fibers/cc-year, 7 and corresponding BMCLio of 0.1177 fibers/cc-year for a 10% increase in prevalence of 8 localized pleural thickening. This BMCLio of 0.1177 fibers/cc-year is the preferred POD 9 estimate to support development of an RfC for Libby Amphibole asbestos. 10 11 sso0.6 a '3 0.5 0.4 aO3l. 0.3 .a 0.2 0.1 f 0 0 > * Estimated prevalence by cumulative exposure lagged by 10 years Observed prevalence by quartiles of unlagged cumulative exposure among cases 2 4 6 8 10 CHEEC (fibers/cc-yr) 12 13 Figure 5-2. Graph of observed and estimated prevalence of localized pleural 14 thickening calculated using the Michaelis-Menten model with 10-year lagged 15 exposure. 16 17 18 5.2.4. RfC Derivation--Including Application of Uncertainty Factors (UFs) 19 Among the available studies that could provide exposure-response data for the 20 relationship between Libby Amphibole asbestos exposure and risk of localized pleural 21 thickening (LPT), consideration of study attributes led to the selection of a study of the 22 Marysville, OLI worker cohort as the primary data set for RfC derivation (Rohs et a L 2 0 0 8 ) (see 23 Section 5.2.1). An updated job-exposure matrix is available for this follow-up of the original This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-35 DRAFT--DO NOT CITE OR QUOTE 1 cohort described by Lockey et al. (1984). The updated job-exposure matrix provides a more 2 refined understanding of exposure to Libby Amphibole asbestos throughout plant operation (see 3 Section 5.2.3.1 and Appendix F). Flowever, due to remaining uncertainties in exposures prior to 4 1972, EPA elected to model a sub-cohort of plant employees that consisted of individuals who 5 began their employment in 1972 or later (see Section 5.2.3.2). It is acknowledged that although 6 this provides a sub-cohort with less potential for exposure misclassification, there is reduced 7 power due to fewer individuals and fewer observed cases. Therefore, EPA provides a supporting 8 analysis using the combined results for the Marysville plant workers as reported in both the 9 original study and in the update (Rohs et al.. 2008; Lockey et al.. 1984)1 (Section 5.2.5). 10 LPT is an irreversible pathological change associated with constricting chest pain, 11 dyspnea, and decreased pulmonary function and, therefore, it is selected as the critical effect in 12 the sub-cohort. The Michaelis-Menten model, with a 10-year lag for exposure, provided the best 13 model fit for the sub-cohort data (AIC = 74.0, see Table 5-4). Using a 10% BMR for LPT, a 14 BMC of 0.2642, and a BMCLio of 0.1177 (fibers/cc)-years were calculated (see Table 5-4). As 15 this POD is in units of cumulative exposure, and the RfC is given in continuous lifetime 16 exposure, the POD was adjusted to 70 years of exposure, lagged by 10 years (nonoccupational, 17 lifetime exposure). Thus the adjusted lifetime BMCLio is 1.96 * 10' fibers/cc (as derived 18 below), and is the POD for RfC derivation. 19 20 21 Lifetime-BMCLio = BMCLio : (lifetime exposure duration) 22 = [0.1177 (fibers/cc) k year] [70 - 10 years] 23 = 1.96 ^ 10"1fibers/cc 24 25 26 Following EPA practices and guidance (U.S. EPA. 2002. 1994b). application of the 27 following uncertainty factors was evaluated resulting in a composite UF of 100. 28 29 30 An interspecies uncertainty factor, UFa, of 1 is applied for extrapolation from animals 31 to humans because the critical effect used as the basis for the RfC was observed in 32 humans. This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-36 DRAFT--DO NOT CITE OR QUOTE 1 An intraspecies uncertainty factor, UFr, of 10 was applied to account for human 2 variability and potentially susceptible individuals in the absence of quantitative 3 information to assess the toxicokinetics and toxicodynamics of Libby Amphibole 4 asbestos in humans. Only adults sufficiently healthy for full-time employment were 5 included in the principal study and the study population was primarily male. 6 A LOAEL to NOAEL uncertainty factor, UFL. of 1 was applied because the current 7 approach is to address this factor as one of the considerations in selecting a BMR for 8 BMC modeling. In this case, a BAIR of 10% extra risk was considered to be 9 minimally biologically significant. 10 A subchronic-to-chronic uncertainty factor, UF's. of 1 was applied because the 11 selected POD is from a study population including chronic exposure (Rohs et al., 12 2008). The average employment duration for the sub-cohort corresponding for the 13 RfC derivation is 18.7 years (SD = 8 6; range = 0.3-29.0). 14 A database uncertainty factor, UFD. of 10 was applied to account for database 15 deficiencies in the available literature for the health effects of Libby Amphibole 16 asbestos. Although there is a large database for asbestos in general, only three study 17 populations exist for Libby Amphibole asbestos specifically: the Marysville, OE1 18 worker cohort, the Libby worker cohort and the ATSDR community screening (which 19 includes some Libby worker cohort participants). Limitations of these studies are 20 described below. 21 1. Evidence exists for an association between exposure to Libby Amphibole asbestos 22 and other noncancer health effects with no exposure-response information. 23 Without additi onal data, it is unknown if a lower POD or RfC would be derived 24 for these effects. 25 a. Two studies have found a possibl e increased preval ence of autoimmune 26 disease and biological markers for autoimmune disease in Libby residents 27 (Noonan et al., 2006; Pfau et al., 2005), although these studies do not 28 indicate whether the autoimmune effects would be observed at exposures 29 lower than that observed for localized pleural thickening. Subsequent 30 animal studies have indicated that exposure to Libby Amphibole asbestos 31 does induce auto-antibodies in mice (Blake et al., 2008).33 32 b. A mortality analysis for the Libby worker cohort also found associations 33 between occupational exposures to Libby Amphibole asbestos and 34 mortality due to cardiovascular disease (Larson et al,, 2010b). 33It is unknown if autoimmune effects are secondary to the chronic inflammatory response expected from exposure to mineral fibers. However, one study o f individuals in a community exposed to tremolite found changes in iimnune parameters in exposed individuals without localized pleural thickening, and that additional iimnune markers, including autoantibodies, increased in individuals with localized pleural thickening (Zerva et al., 1989). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-37 DRAFT--DO NOT CITE OR QUOTE 1 c. Deficits in pulmonary function have been documented in those exposed to 2 Libby Amphibole asbestos occupationally or in the community. However, 3 exposure data are lacking to define an exposure response relationship on 4 this sensitive endpoint (Weill et al.. 2011; Whitehouse. 2004). 5 2. There are no data in laboratory animals or humans on general systemic effects for 6 Libby Amphibole asbestos. However, it is known that inhaled asbestos fibers 7 migrate out of the lung and into other tissues (see Section 3.1), lending 8 uncertainty to any assumptions that other effects would not be expected. 9 3. Although data do exist to define an exposure-response relationship for 10 radiographic abnormalities in the Marysville, OH worker cohort, these data are 11 limited by the dates of the available radiographs. The data for the sub-cohort of 12 workers exposed post-1972 allowed for assessing prevalence of LPT up to 13 approximately 30 years after first exposure (Mean = 28.2 years, 14 range = 23.2-32.7 years). However, there is evidence to indicate that the 15 prevalence of pleural plaques and pleural thickening in general is likely to 16 continue to increase more than 30 years after first exposure (Paris et al.. 2009; 17 Paris et a.L 2008; Jakobsson et aL 1995; Hlllerdal. 1994; Ehrlich et al.. 1992; 18 Jarvholm. 1992; Lilts et aL 1991; Merchant. 1990; McDonald et al.. 1986b). As 19 the RfC is intended for a lifetime of exposure, and pleural thickening is known to 20 progress across the lifetime (even with less-than-lifetime exposures), the lack of 21 health data assessed at end of lifetime is a data gap. 22 23 24 The derivation of the RfC from the morbidity studies of the Marysville, OH worker 25 cohort [i.e., Rohs et al. (2008)1was calculated from a POD, lifetime-BMCLio of 1.96 M10"' 26 fibers/cc for localized pleural thickening, (adjusted to 70 years of exposure, lagged by 10 years 27 (nonoccupational, lifetime exposure), and dividing by a composite uncertainty factor (UF) of 28 100. 29 As derived below, the chronic RfC is 2 >< 10 fibers/cc for Libby Amphibole asbestos 30 and was calculated by dividing the lifetime-POD by a total UF of 100: 31 32 33 Chronic RfC = Lifetime-BMCLio ^ UF 34 = 1.96 x JO'3fibers/cc x- 100 35 = 1.96 x IQ-5 fibers/cc, rounded to 2 x 10"3fibers/cc 36 37 This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-38 DRAFT--DO NOT CITE OR QUOTE 1 5.2.5. Alternative Analyses of the Full Marysville Cohort 2 Modeling of the full cohort was also conducted utilizing the full data set for localized 3 pleural thickening from the Marysville cohort. Since the full cohort includes data combined 4 from Lockey et al. (1984) and Rohs et al. (2008), there were individuals who had more than one 5 observation. As described in Section 5.2.3.2, for those workers X-rayed in both 1980 (Lockey et 6 al.. 1984) and 2004-2005 (Rohs et al.. 2008). one of the observ ations was excluded so that there 7 are no repeat observations for individual workers in the data used for the modeling. 8 Time from first exposure to X-ray (the variable T, in this model) is an important variable 9 in understanding the full Marysville data set, as can be seen by the much higher prevalence of 10 localized pleural thickening in the 2000s compared to the 1980 assessment, an increase which 11 cannot be fully explained by the increases in cumulative exposure occurring with continued 12 exposure. Consequently, in looking at the full cohort, Tis a strong predictor of localized pleural 13 thickening. Study '/'-values are measures of the time from first exposure to the event that an 14 X-ray was taken that detected an abnormality. As such, these values in themselves are not 15 measures of biological latency--an abnormality may be present for some time before the event 16 that an X-ray is taken. Given the occurrence of higher exposures in earlier years in this study, 17 higher /-values correspond to individuals who likely experienced the early higher intensity 18 exposures. This may lead to some uncertainty in the estimated models because uncertainty in the 19 estimated exposures can influence the apparent relationship between T and lesion prevalence. A 20 similar approach as described in Section 5.2.3.3.1 was used to evaluate candidate models for the 21 full cohort. Details are provided in Appendix E. However, as time from first exposure (7) was 22 an important covariate for these analyses, further efforts were needed to develop a model 23 incorporating T along with cumulative exposure. The logistic and probit models including 24 CHEEC as a continuous exposure had inadequate model fit as evaluated using the 25 Hosmer-Lemeshow test (/;-values of 0.003 for both) and so were not considered for further 26 analysis. The remaining candidate models (logistic and probit with the natural logarithm of 27 CHEEC, 3-parameter log-logistic, dichotomous Hill, and dichotomous Michaelis-Menten) had 28 adequate fit. Among these models, the AIC values ranged from 327.9 (Michaelis-Menten) 29 to 346.8 (logistic with the natural logarithm of CHEEC) (see Appendix E). Based on these 30 results, the Michaelis-Menten model was selected for further evaluation, and different This document is a draftfor review purposes only and does not constitute Agency policy. 5-39 DRAFT--DO NOT CITE OR QUOTE 1 approaches were investigated to represent T along with cumulative exposure to Libby Amphibole 2 asbestos using this model form. 3 The approach taken to incorporate T was through modification of the plateau term in the 4 Michaelis-Menten model to allow the plateau for the exposure-response relationship to change 5 for different values of T. After investigating various forms for the plateau (described in 6 Appendix E), the plateau term used took the form: Plateau = Background + (1-background) * 7 0(7]m,s), where <P(7]m,s) represents the cumulative normal probability distribution function. 8 Different exposure lags were then investigated for this model--as seen for the sub-cohort, the 9 AIC values were quite similar for lags of 0-15 years (AICs ranging from 277.72 to 278.04). 10 However, the 20-year lagged exposure had an increased AIC of 280.60 and was not judged an 11 appropriate choice. In order to estimate a BMCio and corresponding BMCLio for this model 12 form, a fixed value of T must be specified. 13 To facilitate comparison of the results of the two models, the Cumulative Normal 14 Michaelis-Menten model was run with the variables consistent with the sub-cohort hired in 1972 15 or later (see Section 5.2.3.3.1). A value of T= 30 years and a lag time of 10 years were used. 16 For the sub-cohort, the mean time from first exposure was 28 years. For the Cumulative Normal 17 Michaelis-Menten model, the BMCio was 0.1477 fibers/cc-year, and the BMCLio was 18 0.0580 fibers/cc-year. These values are generally similar to the results from the sub-cohort for 19 those hired in 1972 or later using the Michaelis-Menten model (BMCio and BMCLio of 0.2642 20 and 0.1177 fibers/cc-year, respectively). 21 One alternative analysis using the full cohort model, with a TSFE value of T= 40 years 22 was conducted. A BMCLio of 0.0136 fibers/cc-year was calculated with the Cumulative Normal 23 Michaelis-Menten model. The BMCLio with T= 40 years is used because it is near the upper 24 end of the range of T values available in the data set (Tmax= 47.375 years). This POD combined 25 with a lag time of 5 years [used because Larson et al. (2010a) showed that discrete pleural 26 thickening could be observed much earlier than previously thought] and a total UF of 100 was 27 used to derive an alternative RfC of 3.8 x 10 6fibers/cc, or rounding to one significant digit, 28 4 x 10 6 fibers/cc. See Appendix E for details. This alternative RfC is a factor of 5 lower than 29 the RfC derived from the sub-cohort. This alternative RfC is an order of magnitude lower 30 compared to both the preferred sub-cohort analysis and the full cohort analysis, with a fixed T of 31 30 years. This document is a draftfor review purposes only and does not constitute Agency policy. 5-40 DRAFT--DO NOT CITE OR QUOTE 1 Another alternative analysis is based on projection of risks using the full cohort model for 2 a "lifetime" time from first exposure of 70 years. Note that none of the workers had a 3 T> 50 years; therefore, this modeling represents a mathematical extrapolation beyond available 4 data. A BMCLio of 0.0042 fibers/cc-year was calculated using the Cumulative Normal 5 Michaelis-Menten model. This POD combined with a lag time of 5 years and a total UF 6 of 30 was used to derive an alternative RfC of 2.1 %10 *' fibers/cc, or rounding to one significant 7 digit, 2 x 10 " fibers/cc. See Appendix E for details. 8 Each of the candidate PODs (analyses from both the sub-cohort and full cohort) has 9 strengths and weaknesses. A major strength of the preferred analysis (Marysville sub-cohort) is 10 that by limiting the data set to those individuals hired in 1972 or later, the exposure 11 reconstruction relies only on data supported by industrial hygiene measurements in the facility. 12 The exposures were also lower after 1972 as compared to previous years. However, this 13 approach reduces the number of individuals in the data set from 434 to 119 and reduces the 14 number of cases from 61 to 12. In addition, this approach narrows the range in the time from 15 first exposure to 23.15-32.65 years (see Table 5-3). The analyses of the full cohort have the 16 strength of using all of the data available on the Marysville cohort and of using a model that 17 incorporates both cumulative exposure and time from first exposure as relevant explanatory 18 variables. One weakness of the full cohort analyses is that the exposure reconstruction relies on 19 estimates of the exposure conditions in the Marysville facility before industrial hygiene data 20 were available in 1972. 21 22 5.2.6. Previous Reference Concentration (RfC) Derivation 23 There is no previous RfC derivation for Libby Amphibole asbestos. 24 25 5.3. UNCERTAINTIES IN THE INHALATION REFERENCE CONCENTRATION 26 (RfC) 27 5.3.1. Uncertainty in the Exposure Reconstruction 28 As in all epidemiologic studies, there are uncertainties in the exposure reconstruction. In 29 this case, there is some uncertainty in the employment history, and some individuals had 30 extensive overtime work. Employment history was self-reported during interviews with each 31 individual for the original study (Lockev et al.. 1984). and errors in this process could affect This document is a draftfor review purposes only and does not constitute Agency policy. 5-41 DRAFT--DO NOT CITE OR QUOTE 1 assigned Libby Amphibole asbestos exposure estimates. As stated previously, fiber 2 measurements started in the Marysville plant in 1972; exposures prior to this time were estimated 3 by University of Cincinnati scientists, based on focus group interviews with 15 long-term former 4 workers and the times when engineering changes were made to control dust in the facility (see 5 Appendix F). Exposure estimates for the period prior to 1972, can, thus, be considered as 6 semiquantitative rather than directly based on industrial hygiene data. The University of 7 Cincinnati analysis assumed that early exposure levels in the plant are twice those measured in 8 1972 (see Appendix F). The greater uncertainty of the pre-1972 exposure estimates led to EPA's 9 decision to focus the analysis on the post-1972 group of workers rather than the full cohort. 10 Although it is generally true that the use of more data is an advantage for statistical analyses 11 because it allows for the computation of more statistically precise effect estimates, this increased 12 precision may be offset by a negative impact on the accuracy of the effect estimate if an increase 13 in sample size is accompanied by greater exposure misclassification or other biases. 14 While the uncertainties related to a lack of quantitative measurements are not relevant to 15 the sub-cohort analysis, it is important to recognize that exposure assessment post-1972 also has 16 some limitations. The main sources of uncertainty are incomplete exposure measurements for 17 some of the occupations/tasks before industrial hygiene improvements that started about 1973 or 18 1974 and continued throughout the 1970s (see Appendix F, Figure F-l). 19 There is uncertainty when the Libby ore was first used in the facility. Company records 20 indicated that the date was between 1957 and 1960, and the University of Cincinnati used the 21 best-available information from focus group interviews to assign the first usage of Libby ore in 22 1959 (see Appendix F). There is also uncertainty in the data regarding asbestos content in other 23 ore sources before and after Libby ore use. In 1957 and 1958, only ore from South Carolina was 24 used. From 1959 to 1971, ores from Libby and South Carolina were used. From 1972 to 1980, 25 ores from Libby, South Carolina, South Africa, and Virginia were used with Libby being the 26 major source. Libby ore was not used in the facility after 1980. However, industrial hygiene 27 measurements collected after 1980 showed low levels of fibers in the facility. PCM analysis 28 does not determine the mineral/chemical make-up of the fiber, and, thus, cannot distinguish 29 between different kinds of asbestos. 30 As reported in Appendix C, the EPA analysis of bulk ores from Virginia and South 31 Africa showed the presence of only a few or no Amphibole asbestos fibers; EPA could not obtain This document is a draftfor review purposes only and does not constitute Agency policy. 5-42 DRAFT--DO NOT CITE OR QUOTE 1 a sample of ore from South Carolina. However, the South Carolina ore is known to contain 2 fibers [see Appendix F; U.S. EPA (2000b); McDonald et al. (1988)1. Using the industrial 3 hygiene data, the University of Cincinnati estimated that the fiber content of the South Carolina 4 ore was about 10% of that of the Libby ore (see Appendix F). This result is consistent with data 5 comparing South Carolina and Libby ores from samples tested in 1982 (U.S. EPA. 2000b). EPA 6 believes that the overwhelming exposure to fibers in the Marysville facility is from the Libby 7 ore. Therefore, EPA has attributed all of the adverse health effects to exposure to fibers from 8 Libby ore from 1957 to 1980 and from the post-1980 exposure. However, because the 9 concentration of fibers in the workplace was near background after 1980, the post-1980 time 10 period makes only a small contribution to an individual's cumulative exposure. 11 There was potential coexposure to other chemicals in the Marysville facility (see 12 Section 4.1.3). These other chemicals were used after expansion of vermiculite ore in another 13 area of the facility. Industri al hygiene data showed very low levels of fibers in the areas where 14 the additional chemicals were added to the expanded vermiculite. In addition, none of these 15 chemicals are volatile. The most likely route of exposure to these chemicals is through dermal 16 contact. It is unlikely that any coexposure to these particular chemicals would alter the 17 exposure-response relati onship of Libby Amphibole asbestos in the respiratory system (see 18 Sections 4.1.3 and 5.3.1). 19 The University of Cincinnati Research Team assumed that there was no exposure to 20 Libby Amphibole asbestos outsi de of the workpl ace. The interviews with the Marysville 21 workers revealed that about 10% of the workers reported bringing raw vermiculite home. These 22 interviews also revealed that changing to street clothes from work-supplied coveralls was 23 standard practice at the end of the shift, and approximately 64% of the workers showered before 24 leaving the workplace. For these workers, it is likely that additional exposure outside the 25 workplace was minimal. However, for the remainder of the workers, it is reasonable to assume 26 that additional exposure could have occurred at home. Additional data collected by the 27 University of Cincinnati Research Team document that no increased prevalence of pleural or 28 parenchymal change consistent with asbestos exposure has been observed in household contacts 29 of the workers from the Marysville facility (J. Lockey, University of Cincinnati, personal 30 communication to Robert Benson, U.S. EPA, 2011). 31 This document is a draftfor review purposes only and does not constitute Agency policy. 5-43 DRAFT--DO NOT CITE OR QUOTE 1 5.3.2. Uncertainty in the Radiographic Assessment of Localized Pleural Thickening 2 The use of conventional radiographs to diagnose pleural thickening has several 3 limitations. The localized thickening must be of sufficient size and thickness to be viewed on the 4 X-ray; small lesions may exist but not be reported. More severe and larger lesions are more 5 reliably detected on radiographs. There are also potential interferences. Fat pads may be 6 mistaken as pleural plaques as they generally occur against the ribcage in a similar location 7 (Gilmartin. 1979); this is one source of uncertainty between readers. Although generally related 8 to mineral fiber exposure, pleural plaques may also be a result of trauma to the chest, and pleural 9 thickening may appear after an active TB infection. Often signs of trauma (e.g., fractured ribs) 10 and radiographic signs of past TB infection can be seen and are noted by the reader. In these 11 cases, LPT would not be diagnosed. There is a certain amount of subjectivity when viewing the 12 X-rays determining which features are representative of pleural thickening and if signs of 13 alternative etiology can be noted; thus, several certified readers are generally consulted, and a 14 consensus of opinions determines the diagnosis. Regardless, there is still potential for outcome 15 misclassificati on. For example, one of the workers in the Marysville cohort had a positive X-ray 16 in the 1980 evaluation but a negative X-ray at the 2002-2005 evaluation (excluding this worker 17 from the analysis did not change results). However, uncertainty in the presence or absence of 18 localized pleural thickening in each individual is considered minimal due to the use of three 19 highly qualified chest radiologists evaluating the radiographic films and the use of the majority 20 vote of the readers for the diagnosis. 21 BMI was investigated as a potential explanatory variable because fat pads can sometimes 22 be misdiagnosed as pleural thickening. BMI was not measured in the 1980 examination but was 23 available for most participants of the 2000s examination. To address whether fat deposits may 24 affect outcome classification, EPA considered the effect of adding BMI as a covariate in the 25 model. However, BMI did not display an association with odds of localized pleural thickening 26 in this population (see Appendix E). While these covariates were not associated with the risk of 27 localized pleural thickening in the sub-cohort after adjusting for exposure, it was not possible to 28 evaluate this relationship in the full cohort. In the general U.S. population, BMIs have increased 29 between 1980 and the 2000s, so one cannot necessarily assume the relationships will be the same 30 for the two examination periods. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 5-44 DRAFT--DO NOT CITE OR QUOTE 1 5.3.3. Uncertainty Due to Time From First Exposure 2 There is some uncertainty associated with the length of follow-up of the Marysville 3 cohort. The observed range of TSFE to X-ray in the full cohort is 0.4 47 years, and 23.2-32.7 4 years in the preferred sub-cohort (see Table 5-3). It is anticipated that the prevalence of 5 localized pleural thickening in the study population--and in the post 1972 exposure cohort-- 6 may continue to show some increase with passage of time. In this case, the modeling approach 7 may not accurately reflect the exposure-response relationship that would be seen with a longer 8 follow-up time. However, a recent study by Larson et al. (2010a) examined serial radiographs 9 conducted on a group of Libby vermiculite workers with pleural or parenchymal changes. They 10 found that among those workers with localized pleural thickening, all cases were identified 11 within 30 years, and that the median time from hire to the first detection of localized pleural 12 thickening was 8.6 years. Albeit the retrospective evaluation of radiographs is a different and 13 more sensitive procedure, these findings indicate that the range of follow-up time in the 14 Marysville sub-cohort is likely sufficient to support the exposure-response modeling developed 15 in this current assessment. Note that the likelihood that prevalence of localized pleural 16 thickening may further increase beyond 30 years after first exposure is a principal rationale cited 17 for the selection of a database UF of 10 in this current assessment. 18 19 5.3.4. Uncertainty in Background Rate of Localized Pleural Thickening 20 In the derivation of the RfC, a background rate of 1% for localized pleural thickening was 21 used. As discussed in Section 5.2.3.3, there is uncertainty in estimating the value of this 22 parameter. However, in statistical modeling of the Marysville sub-cohort, potential uncertainty 23 in the background rate of localized pleural thickening has little impact on the estimated POD. 24 The best-fitting model (Michaelis-Menten with 10-year lagged exposure) was rerun, allowing the 25 background rate to be estimated as a parameter rather than fixed, with a resulting estimated 26 background rate of 3.12% (SE = 2.84%). Both the fixed and estimated values are in the range of 27 estimates from previous studies described above, and the difference in the POD when the 28 background rate is fixed at 1% versus when it is estimated is ~15% (0.1177 compared with 29 0.1349 fibers/cc-year, and it does not affect the proposed RfC (after rounding to one significant 30 digit). 31 This document is a draftfor re\'iew purposes only and does not constituteAgency policy. 5-45 DRAFT--DO NOT CITE OR QUOTE 1 5.3.5. Uncertainty in Model Functional Form and Lagged Exposure 2 A number of model forms were explored in the initial stages of analysis (see Appendix E) 3 before selecting the Michaelis-Menten model. In this application, the ratio of the BMCio to the 4 BMCLio (0.2642 HHGS1177 ==2.2) was reasonable given the size of the available data set, 5 indicating acceptable statistical precision in the BMC estimate. In addition, BMCs and BMCLs 6 estimated from other candidate models for the post-1972 exposure sub-cohort were in a similar 7 range to the selected model. Finally, the complementary analysis with the full cohort (utilizing a 8 time from first exposure of 30 years, which was selected to be consistent with time since first 9 exposure values within the sub-cohort) provided similar results to the sub-cohort analysis. A 10 second model-based uncertainty is the choice of lag for cumulative exposure. The RfC 11 derivation is based on the exposure lagged by 10 years, since this lag yielded the lowest AIC. 12 However, if other lags (with similar AICs) are used, the difference in POD may fluctuate to be 13 approximately 20% higher or approximately 55% lower. Thus, the choice of lag does not affect 14 the proposed RfC (after rounding to one significant digit). 15 16 5.3.6. Uncertainty Due to Effect of Smoking 17 Smoking is an important variable to consider when evaluating respiratory health 18 outcomes. Although data are mixed, a few studies suggest smoking may affect risk of 19 developing pleural thickening or timing of pleural thickening development among persons 20 exposed to asbestos. However, no studies were identified that assessed the relationship between 21 LPT specifically and any measure of smoking status. Di screte pleural plaques as defined in 22 earlier ILO classification systems have not been associated with smoking in asbestos-exposed 23 workers (Mastrangelo et al.. 2009; Paris et al... 2009; Koskinen et al.. 1998). but there is evidence 24 that small opacities (asbestosis) and diffuse pleural thickening may be associated with smoking 25 in asbestos-exposed individuals.34 As the current classificati on of LPT includes cases of diffuse 34Studies among populations exposed to general asbestos have reported mixed effects on the impact o f smoking on risk o f radiographic abnormalities; two studies reported a significant association between risk o f all pleural thickening, including both pleural plaques and diffuse pleural thickening (McMillan et al.. 19801. or any pleural abnormality (Welch et al.. 2007) and smoking after controlling for some measure o f asbestos exposure. A larger number o f studies reported borderline-- or possible-- associations when examining risk o f pleural changes (Paris et al.. 2008: Dement et al- 2003: Zitting et al.. 1996: Yano et a l. 1993: Lilis et al.. 1991; Baker et al.. 1985) or no association with smoking (Soulat et al.. 1999: Neri et al,. 1996: Ehrlich et a l. 1992: Delclos et al., 1990: Rosenstock et al., 1988). Possible reasons for the different findings include varying quality o f smoking information (some used This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-46 DRAFT--DO NOT CITE OR QUOTE 1 pleural thickening where the CPA is not involved, investigation of the potential for smoking to 2 modify the effect of asbestos exposure on the prevalence of LPT is warranted. 3 Each of the four candidate studies considered for RfC derivation considered smoking in 4 their analytic approach. In the Libby workers cohort, McDonald et al. (1986b1assessed pleural 5 thickening of the chest wall (both discrete and diffuse regardless of CPA involvement) and found 6 smoking status (current, former, or never smoker) was of borderline statistical significance 7 (p = 0.10) in a regression model, controlling for Libby Amphibole asbestos exposure and age. 8 This is consistent with the broader asbestos literature, addressing all pleural thickening or all 9 pleural abnormalities. Amandus et al. (1987b) evaluated radiographic abnormalities consistent 10 with the current LPT designation; the authors took a different analytic approach to assess 11 smoking effects, constructing separate models for the full cohort and restricting to current and 12 former smokers. The parameter estimates were not significant for the two models, although the 13 coefficients corresponding to Libby Amphibole asbestos exposure were slightly higher for the 14 full cohort model. 15 In the Marysville workers cohort, smoking was characterized using pack-years in the 16 original study (Lockev et al.. 1984) and as ever/never smoking in the follow-up (Rohs et al.. 17 2008). Lockey et al. (1984) reported that the pack-years variable was significantly associated 18 with risk of all radiographic changes using discriminate analysis (any pleural thickening, small 19 opacities, and blunting of the CPA) but did not present results for effect of smoking controlling 20 for Libby Amphibole asbestos exposure. Rohs et al. (2008) did not find a difference in smoking 21 prevalence among those with and without any radiographic changes but also did not report 22 results controlling for Libby Amphibole asbestos exposure, or for LPT specifically. 23 Therefore, EPA explored the effect of smoking on the critical endpoint. In analyses for 24 RfC derivation, the variable representing smoking history (ever smoker vs. never smoker) was of 25 borderline significance in the best-fitting model (p = 0.08) and improved model fit (see 26 Appendix E). The limited sample size (only three cases were never smokers) and limited nature 27 of the smoking information precluded use of the smoking variable for RfC derivation. However, 28 the model including smoking was examined as a sensitivity analysis. In this analysis, BMCs and 29 BMCLs estimated separately for smokers and nonsmokers differed by approximately sixfold, categories o f ever/never or fonner/current/never, while others used pack-years) and differences in the specific outcome studied. This document is a draftfor review purposes only and does not constitute Agency policy. 5-47 DRAFT--DO NOT CITE OR QUOTE 1 suggesting that smokers may be at a higher risk for LPT from exposure to Libby Amphibole 2 asbestos than nonsmokers. Thus, an estimated BMCL for smokers would be lower than the POD 3 used for RfC derivation (0.04 fibers/cc-year for smokers versus 0.12 fibers/cc-year for the entire 4 sub-cohort). Conversely, a BMCL for nonsmokers would be slightly higher 5 (0.25 fibers/cc-year). These sensitivity analyses indicate a need for further research on the effect 6 of smoking in relation to LPT risk among asbestos-exposed populations. 7 8 5.3.7. Sensitivity Analysis: Derivation of a POD for Lifetime Exposure From the 9 Cumulative Exposure Metric 10 Exposure-response modeling for LPT in the Marysville sub-cohort used the cumulative 11 exposure (CE) metric (represented as CHEEC, described in Section 5.2.2.1) providing a POD in 12 fibers/cc-years. In order to derive an RfC in the units of continuous air concentration for a 13 lifetime (Li,, fibers/cc), the POD from the CE metric was weighted across a lifetime exposure. 14 Thus, the lifetime BMCLio is 1.96 ><10" (0.1177 fibers/cc-years : 60 years35). This procedure is 15 one way to account for the duration of exposure in the occupational study being less than 16 lifetime. There is some uncertainty as to whether and how to take account for less-than-lifetime 17 exposure in the occupational cohort. The cohort participants had a wide range of exposure 18 durations, all of which are less than lifetime36. As there are other reasonable alternatives to 19 derive a lifetime RfC, a sensitivity analysis was conducted to examine if RfC derivation was 20 greatly impacted by the method chosen to convert the POD in units of cumulative exposure, to 21 an air concentration for lifetime exposure. 22 Use of the CE metric adjusted based on ventilation rates and work schedule to a 23 continuous air concentration is consistent with EPA guidance (represented as CHEEC in this 24 assessment) (U.S. EPA. 2002. 1994b). Guidelines also recommend that if the human study is a 25 less-than-lifetime study, additional adjustment may be needed, depending on the nature of the 26 observed health effect for an RfC applicable to lifetime exposure (U.S. EPA. 1994b). Although 27 cumulative exposure is often associated with asbestosis (small opacities) and DPT, many other 28 studies have found pleural plaques are better predicted by other exposure metrics (e.g., average 35Because the best-fitting model had a 10-year lag, the lag is applied to the weighting across a lifetime as well. Sixty years represent lifetime exposure o f 70 years; 70 years - 10 years for the lag in exposure. 36This is especially true for the RfC derived from the sub-cohort hired after 1972, which had a more limited range o f employment duration (mean=18.7 years [SD =8.6f range=0.3-29.0). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-48 DRAFT--DO NOT CITE OR QUOTE 1 intensity, mean exposure, duration). The use of a measure of average exposure (averaged over 2 the period of exposure) is consistent with previous studies (asbestos in general) that report 3 associations of the prevalence of pleural plaques with mean or average asbestos exposure (Paris 4 et af. 2008; Jakobsson et at., 1995; Ehrlich et af. 1992). The first alternative method was to 5 weight the POD across duration of exposure in the sub-cohort, rather than a full lifetime. The 6 second alternative is model the exposure-response relationship for LPT against average exposure 7 (a measure of the cumulative exposure for each worker averaged over the individual worker's 8 duration of work exposure). 9 The first sensitivity analysis is calculated by dividing the modeled POD for the 10 sub-cohort (0.1177 fibers/cc-years [30-year BMCLio]) by the average employment duration for 11 the sub-cohort of 18.7 years. Therefore, the POD expressed as the equivalent concentration for 12 the mean worker exposure duration for the sub-cohort i*s 6.3 3x 10" (fibers/cc, continuous air 13 concentration) ([0.1177 fibers/cc-years] x- 18.7 years). 14 For the second analysis, the average exposure was calculated for each participant 15 (AvgExp = CITEEIT for each worker x duration of exposure for each worker). The 16 exposure-response relationship was defined using the best-fitting model for the sub-cohort 17 (Michaelis-Menten). The average exposure metric also provided an adequate fit to the data for 18 the preferred sub-cohort (Ilosmer-Eemeshow GOF37; P = 0.72) and provided a slightly better-- 19 but similar--fit to the CE metric [AIC = 72.2 versus 74.0). The Michaelis-Menten model 20 provided a BMC of 1.8 K 10"2fibers/cc, and a BMCLio of 8.5 M10"3fibers/cc for the average 21 work-duration exposure metric. This BMCLio is about 4-fold higher than the lifetime-BMCLio 22 above from the primary analysis (1.96 x 10" fibers/cc). 23 The three methods provide PODs that vary by a factor of up to 4 (2.0 x 10"', 6.1 x 10'3, or 24 8.5 x 10" fibers/cc) when expressed as a continuous air concentration. The primary analysis 25 assumes duration contributes to risk and thus calculates a concentration across a lifetime that 26 would yield the POD CE. The second analysis is consistent with assuming duration contributes 27 to risk but estimating the concentration only for the mean duration in the observed database. The 3 General model fit was evaluated with the I-Iosmer-Lemesliow (2000) test (a form o f the Pearson x2 goodness-of-fit [GOF] statistic). This is a goodness-of-fit test that compares observed and expected events. Observations are sorted in increasing order o f estimated probability o f the event occurring and then divided into ~10 groups; the test statistic is calculated as the Pearson j j statistic of observed and expected frequencies in these groups. This document is a draftfo r review purposes only and does not constitute Agency policy. 5-49 DRAFT--DO NOT CITE OR QUOTE 1 third analysis assumes duration does not contribute to risk and models the average work duration 2 continuous exposure equivalent for each worker. 3 The difference comes in whether the critical study is considered of adequate duration to 4 inform health effects from a lifetime exposure, or if further adjustment is needed across time. 5 The primary analysis provides this adjustment to a full lifetime. This sensitivity analysis 6 indicates that the approach taken to average the POD based on the CE metric (CHEEC) across a 7 lifetime was a reasonable approach, as similar results are obtained using different approaches 8 (i.e., within 4-fold). 9 10 5.3.8. Sensitivity Analysis for Choice of Critical Effect and Selection of Benchmark 11 Response (BMR) 12 The critical effect selected for RfC derivation is localized pleural thickening. Alternative 13 endpoints were not modeled using the preferred sub-cohort due to small numbers--there were 14 five cases of bilateral LPT, only one case of diffuse pleural thickening, and no individuals with 15 interstitial changes. As a sensitivity analysis, these three alternative endpoints (along with all 16 LPT) were modeled among all Marysville workers not previously exposed to other forms of 17 asbestos, with X-rays performed in 2002-2005 (n = 250). These analyses were performed using 18 the Michaelis-Menten model with a background rate of 1% and unlagged CHEEC as the 19 exposure metric. BMRs of 1, 5, and 10% were investigated (see Table 5-5). Else of the 10% 20 BMR for these alternative endpoints allows for comparison with a POD based on the selected 21 critical effect of LPT. In this larger cohort, the POD for a 10% increase in LPT was 22 0.06 fibers/cc-years (in comparison with the POD derived from the sub-cohort and used in RfC 23 derivation of 0.118 fibers/cc-years). Results for all pleural thickening (LPT and DPT) did not 24 differ from results for LPT. Bilateral localized pleural thickening was included as a rough 25 indication of increased severity within LPT, and as expected results in higher PODs at each 26 BMR than LPT. The resulting BMCLs for DPT and small opacities (1.17 and 2.89 27 fibers/cc-years, respectively, 10% BMR) are higher than the POD for LPT (0.06 fibers/cc-years). 28 Thus, use of an alternative endpoint at the same BMR would provide a higher POD, and 29 corresponding higher RfC. 30 However, a 10% BMR is not appropriate for more severe endpoints and BMCLs are 31 calculated at 1 and 5% BMRs as well. If DPT is used as a critical effect, PODs of 0.081 and This document is a draftfor review purposes only and does not constitute Agency policy. 5-50 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-5. Modeling of alternative endpoints in the Marysville worker 2 cohort members examined in 2002-2005 3 BMR P aram eter Bilateral localized pleural thickening (n = 33) vs. no abnorm alities (n = 181) Diffuse pleural thickening (n = 10) vs. no abnorm alities ( = 181) Interstitial changes (n = 7) vs. no abnorm alities (n = 181) AIC 164.6 64.1 45.9 Alpha (SE) 0.2670 (0.5420) -2.8434 (1.6617) -4.0674 (0.5014) Plateau (SE) 0.4120 (0.0962) 0.6166 (0.6307) 1.0000 (- ) BMR = 1% BMC 0.0193 0.2849 0.5899 BMCL 0.0097 0.0814 0.2425 BMR = 5% BMC 0.1075 1.5259 3.0739 BMCL 0.0552 0.4728 1.3158 BMR = 10% BMC 0.2501 3.3494 6.4894 BMCL 0.1337 1.1715 2.8923 4 5 6 0.473 fibers/cc-years would be calculated for a 1% and 5% BMR, respectively. If small 7 opacities are used as a critical effect, the PODs are higher at both a 1% and a 5% BMR 8 (0.243 and 1.32, respectively). In summary, the use of more severe alternative endpoints (with 9 appropriate BMRs) results in PODs higher than that estimated using the critical effect of LPT 10 (0.06 fibers/cc-year, BMR 10%), and all are higher than the POD used in RfC derviation, with 11 the exception of DPT at a 1% BMR (0.0814 fibers/cc-year). BMCLs for these more severe 12 endpoints using a 1% BMR were within -twofold of the preferred POD (0.0814 and 13 0.2425 fibers/cc-year for diffuse pleural thickening and interstitial changes, respectively). There 14 is uncertainty associated with these estimates due to the inclusion of individuals hired before 15 1972, when no quantitative exposure measurements were available. Thus, a choice of alternative 16 critical effects (even with lower BMRs) would not result in an RfC appreciably lower than the 17 proposed RfC based on LPT and a 10% BMR. 18 19 5.4. CANCER EXPOSURE-RESPONSE ASSESSMENT 20 5.4.1. Overview of Methodological Approach 21 The objective of this human health assessment is to derive a cancer estimate for 22 inhalation exposure to Libby Amphibole asbestos. The inhalation unit risk (IUR) is defined as 23 an upper-bound excess lifetime cancer risk estimated to result from continuous exposure to an 24 agent at a concentration of 1 pg/L in water, or 1 pg/m3 in air. However, current health standards This document is a draftfor review purposes only and does not constitute Agency policy. 5-51 DRAFT--DO NOT CITE OR QUOTE 1 for asbestos are given in fibers/cc of air as counted by PCM, since they are based on health 2 effects observed in occupational cohorts and this is the standard for measuring fiber exposures in 3 an occupational environment (OSHA. 1994; U.S. EPA. 1988a). Similarly, when examining the 4 available health effects data for Libby Amphibole asbestos, the best available exposure metric at 5 this time is fibers/cc counted by PCM (see Section 4.1.1.2). Therefore, for Libby Amphibole 6 asbestos, the IUR represents the lifetime risk of mortality from either mesothelioma or lung 7 cancer in the general U.S. population from chronic inhalation exposure to Libby Amphibole 8 asbestos at a concentration of 1 fiber/cc of air. 9 lURs are based on human data when appropriate epidemiologic studies are available. 10 The general approach to developing an 1UR from human epidemiologic data is to quantitatively 11 evaluate the exposure-response relationship (slope) for that agent to derive a specific estimate of 12 its cancer potency in the studied population. For this current assessment, the first step was to 13 identify the most appropriate data set available, which in this case can be used to quantitatively 14 estimate the effects of Libby Amphibole asbestos exposure on cancer mortality. Once the 15 relevant data describing a well-defined group of individuals along with their exposures and 16 health outcomes were selected, an appropriate statistical model was selected that adequately fit 17 the data, and then individual-level exposures were modeled using a variety of possible exposure 18 metrics (see Section 5.4.2). The available epidemiologic data allowed for modeling of the 19 effects of estimated ambient occupational exposures to Libby Amphibole asbestos on the 20 observed cancer mortality risk in workers. Exposure-response modeling was conducted for each 21 cancer mortality endpoint individually, and in some cases, the statistical model and the specific 22 metric of exposure used for each cancer endpoint may have been different. For example, the 23 exposure metric that best describes the exposure-response relationship for mortality from 24 mesothelioma attributable to occupational exposure to Libby Amphibole asbestos was found to 25 be different from the exposure metric that best describes mortality from lung cancer (see 26 Section 5.4.3). Potential covariates that may also be important predictors of cancer mortality are 27 included in the statistical models. These models were then statistically evaluated to determine 28 which exposure metric representing estimated ambient occupational exposures provided the best 29 statistical fit to the epidemiologic data 30 This cancer potency (slope) estimate derived from the epidemiologic data is then applied 31 to the general U.S. population to determine the exposure level that would be expected to result This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-52 DRAFT--DO NOT CITE OR QUOTE 1 in 1% extra cancer mortality risk over a lifetime of continuous exposure. For epidemiologic 2 studies, which may be based on larger numbers of individual observations, smaller response 3 levels that are closer to the background response levels are considered appropriate. Extra risk is 4 defined as equaling (R. - Ro) - (1 - R,,), where R, is the lifetime cancer mortality risk in the 5 exposed population, and R0is the lifetime cancer mortality risk in an unexposed population (i.e., 6 the background risk). For example, if the expected lifetime risk of lung-cancer mortality in the 7 unexposed general U.S. population is 5%, then this human health assessment seeks to estimate 8 the level of exposure to Libby Amphibole asbestos that would be expected to result in a lifetime 9 risk of lung-cancer mortality of 5.95%; this lifetime risk of mortality is equivalent to a 1% extra 10 risk: (0.0595-0.05) + (1-0.05) = 0.01. For mesothelioma mortality, an absolute risk of 1% was 11 considered, rather than extra risk, for two reasons. First, because mesothelioma is very rare in 12 the general population (Hillerdal 1983). and second, because mesothelioma is almost 13 exclusively caused by exposure to asbestos, including Libby Amphibole asbestos. 14 A life-table analysis (see Appendix G for details) was used to compute the 95% lower 15 bound on the lifetime exposure to Libby Amphibole asbestos that corresponds to a 1% extra risk 16 of cancer mortality in the general U.S. population using age-specific mortality statistics and the 17 exposure-response relationships for each cancer endpoint as estimated in the studied population. 18 This lower bound on the level of exposure serves as the POD for extrapolation to lower 19 exposures and for deriving the unit risk. Details of this analysis are presented in Section 5.4.5. 20 A cancer-specific unit risk was obtained by dividing the extra risk (1%) by the POD. The 21 cancer-specific unit risk estimates for mortality from either mesothelioma or lung cancer were 22 then statistically combined to derive the final KIR (see Section 5.4.5.3). Uncertainties in this 23 cancer assessment are described in detail in Section 5.4.6. 24 25 5.4.2. Choice of Study/Data--with Rationale and Justification 26 This human health assessment is specific to Libby Amphibole asbestos. This current 27 assessment does not seek to evaluate quantitative exposure-response data on cancer risks from 28 studies of asbestos that did not originate in Libby, MT. 29 The available sources of data included the cohort of workers employed at the vermiculite 30 mining and milling operation in and around Libby, MT. This cohort has been the subject of 31 several epidemiologic analyses (Larson et aU 2010b; Moolgavkar et aL 2010; Sullivan. 2007; This document is a draftfor review purposes only and does not constitute Agency policy. 5-53 DRAFT--DO NOT CITE OR QUOTE 1 Amandus and Wheeler, 1987; McDonald et al., 1986a) (and described in detail in Section 4.1). 2 There have also been published reports on cases of mesothelioma in the Libby, MT area 3 (Whitehouse et al.. 2008) and mortality data published by the Agency for Toxic Substances and 4 Disease Registry (ATSDR, 2000). However, published mortality data on Libby, MT residents 5 (Whitehouse et a f, 2008; ATSDR, 2000) could not be used in exposure-response modeling due 6 to lack of quantitative exposure data. 7 The other available cohort of workers exposed to Libby Amphibole asbestos was from an 8 Ohio vermiculite processing plant (see Section 4.1.3) (Rohs et al.. 2008; Lockev et al.. 1984). 9 Pleural changes were evaluated; however, no data were available pertaining to cancer incidence 10 or mortality in the Ohio cohort. No other worker cohorts exposed to Libby Amphibole asbestos 11 with cancer incidence or mortality data were available. 12 The most appropriate available data set with quantitative exposure data for deriving 13 quantitative cancer mortality risk estimates based on Libby Amphibole asbestos exposure in 14 humans is the cohort of workers employed at the vermiculite mining and milling operation in and 15 around Libby, MT (hereafter referred to as the Libby worker cohort). These data are considered 16 the most appropriate to inform this human health assessment for several reasons: (1) these 17 workers were directly exposed to Libby Amphibole asbestos, (2) detailed work histories and 18 job-specific exposure estimates are available to reconstruct estimates of each individual's 19 occupational exposure experience, (3) the cohort is sufficiently large and has been followed for a 20 sufficiently long period of time for cancer to develop (i.e., cancer incidence) and result in 21 mortality, and (4) the broad range of exposure experiences in this cohort provided an 22 information-rich data set, which allowed evaluation of several different metrics of exposure. 23 Uncertainties in these data are discussed in Section 5.4.6. 24 25 5.4.2.I. Description o f the Libby Worker Cohort 26 The Libby worker cohort has been extensively studied. McDonald et al. published three 27 studies on a subset of the cohort (2004, 2002; 1986a). Sci enti sts from NIOSH conducted two 28 epidemiologic investigations, resulting in several published reports on different subsets of the 29 cohort (Sullivan. 2007; Amandus et al... 1988; Amandus and Wheeler, 1987). Larson et al. 30 (2010b) analyzed an ATSDR reconstruction of the Libby worker cohort from company records 31 with exposure estimates obtained from NIOSH with mortality follow-up through 2006. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-54 DRAFT--DO NOT CITE OR QUOTE 1 Moolgavkar et al. (2010) reanalyzed the Sullivan (2007) data with mortality follow-up through 2 2001 using a different statistical approach. 3 According to Sullivan (2007), nearly all of these study subjects were workers at the 4 Libby, MT vermiculite mine, mill, and processing plant. Although the mine was several miles 5 from Libby, MT, some of the study subjects worked in the town (see Section 4.1.1.1). Workers 6 may have also been assigned jobs as truck drivers, or jobs working in the screening plant, 7 railroad loading dock, expansion plants, or an office. Individuals' demographic and work history 8 data were abstracted from company personnel and pay records. A database created by NIOSH in 9 the 1980s contained demographic data and work history starting from September 1935, and vital 10 status at the end of 1981 for 1,881 workers. NIOSH compared these data with company records 11 on microfilm, and work history data were reabstracted to ensure data quality. One person was 12 removed from the cohort because company records stated that he was hired but never worked 13 (Sullivan. 2007). Nine workers with Social Security numbers listed in company records were 14 excluded because demographic and work history data were not available, leaving 1,871 workers 15 in the cohort available for epidemiologic analysis. Table 5-6 shows the demographic and 16 exposure characteri stics of this cohort. 17 18 Table 5-6. Demographic and exposure characteristics of the Libby worker 19 cohort 20 Characteristic All workers Number o f workers 1,871 Number o f deaths from all causes 1,009 Number o f deaths from mesothelioma 18 Number o f deaths from lung cancer 111 Mean year o f birth 1929 Mean year o f hire 1959 Mean age at hire (years) 30.2 Mean person-years o f follow-up (no lag) 35.9 Total person-years o f follow-up (no lag) 67,101 Mean employment duration (years) 3.7 Mean cumulative exposure (fiber/cc-year) 96.0 Median cumulative exposure (fiber/cc-year) 9.8 Range o f cumulative exposures (no lag) (fiber/cc-year)a 0-1722 21 22 A ccording to the work histories and JEM, there were 26 workers who had zero exposure. These 23 individuals (7 men and 19 women) all worked at the office downtown. This document is a draftfor review purposes only and does not constituteAgency policy. 5-55 DRAFT--DO NOT CITE OR QUOTE 1 For the purposes of this current assessment, vital status follow-up was completed by 2 NIOSH through 2006 using the National Death Index [NDI-Plus; Bilgrad (1999)1. Workers 3 known to be alive on or after January 1, 1979 (the date NDI began tracking deaths nationwide), 4 but not found in the NDI search, were assumed to have been alive on December 31, 2006 5 (Sullivan. 2007). Nearly 54% of workers in the cohort (n = 1,009) had died by 6 December 31, 2006. NIOSH researchers obtained death certificates from across the United 7 States (while exposure occurred in and around Libby, deaths could have occurred elsewhere) for 8 deaths prior to 1979 and coded to the International Classification of Diseases (ICD) revision in 9 effect at the time of death by a single National Center for Health Statistics-trained nosologist. 10 After 1979, ICD codes were obtained from the NDI-Plus. For workers known to be deceased, 11 the underlying cause of death was determined from death certificates and coded to the ICD codes 12 using the rubrics of the ICD revision in effect at the time of death [ICD-5 (WHO. 1938). ICD-6 13 (WHO. 1948). ICD-7 (WHO. 1957). ICD-8 (WHO. 1967). ICD-9 (WHO. 1977). or ICD-10 14 (WHO. 1992)1 15 Basic demographic information on the occupational cohort members was largely 16 complete. However, when data were missing, they were imputed by NIOSH based on the 17 following assumptions regarding gender, race, and date of birth. Seven workers with unknown 18 gender were assumed to be male because 96% of the workforce was male, and NIOSH review of 19 names did not challenge that assumption (Sullivan. 2007). Workers of unknown race (n = 935) 20 were assumed to be white because workers at this facility were known to be primarily white, and 21 U.S. Census Bureau data indi cate that 90-95% of the local population identify themselves as 22 white (Sullivan. 2007). For four workers with unknown birth dates, date of birth was estimated 23 by subtracting the mean age at hire for the cohort from the worker's hire date. The potential 24 impact of this imputation procedure on the analytic results is discussed in Section 5.4.6. 25 26 S.4.2.2. Description o f Cancer Endpoints 27 This human health assessment of Libby Amphibole asbestos focuses on two cancer 28 endpoints: mesothelioma and lung cancer. The endpoint for both mesothelioma and lung cancer 29 was mortality, not incidence. Incidence data are not available for the Libby worker cohort. 30 However, there is evidence that other cancer endpoints may also be associated with exposure to 31 asbestos. The International Agency for Research on Cancer (IARC) concluded that there was This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-56 DRAFT--DO NOT CITE OR QUOTE 1 sufficient evidence in humans that other types of asbestos (chrysotile, crocidolite, amosite, 2 tremolite, actinolite, and anthophyllite) were causally associated with mesothelioma and lung 3 cancer, as well as cancer of the larynx and the ovary (Straif et al.. 2009). Among the entire 4 Libby worker cohort, only two deaths were found to be due to laryngeal cancer, and there were 5 no deaths from ovarian cancer among the 84 female workers. The EPA did not evaluate these 6 other outcomes as part of this current assessment. The limited number of female workers in this 7 cohort is discussed later as a source of uncertainty in the derived estimates (see Section 5.4.6). 8 Mesothelioma did not have a distinct ICD code prior to introduction of the 10'1' revision 9 PCD-10), which was not implemented until 1999. Therefore, for deaths in the Libby worker 10 cohort occurring from 1979 to 1998, death certificates were obtained if the NDI identified the 11 death as being from one of the possible mesothelioma codes identified by Marsh et al. (2001), or 12 from respiratory cancer, nonmalignant respiratory disease, digestive cancer, or unspecified 13 cancer. Death certificates (1940-1998) were reviewed by the NIOSH principal investigator 14 (Sullivan, 2007) to identify any mention of mesothelioma on the death certificate, as is the 15 standard procedure for assessing mesothelioma mortality and as has been used in other analyses 16 of Libby worker cohort mesothelioma mortality (Larson et al., 2010b; McDonald et al,, 2004). 17 In total, 18 mesothelioma deaths occurring from 1979 to 2006 were identified by NIOSH using 18 these methods, which serve as the basis for this current assessment; 19 mesothelioma deaths 19 were identified by Larson et al. (2010b) for the same cohort from death certificates for all causes 20 of death rather than the more targeted set of causes identified by Marsh et al. (2001) or Sullivan 21 (2007). 22 Whitehouse et al. (2008) identified four mesothelioma cases among workers that were 23 not included in Sullivan (2007) with mortality follow-up through 2001; no other information was 24 provided. Most likely, three mesothelioma cases from these four were accounted for during the 25 update of the NIOSH cohort to 2006, which serves as the basis for this current assessment. 26 Whitehouse et al. (2008) also provided detailed information on 11 residential cases, but this 27 information could not be used in exposure-response analyses for this current assessment because 28 there is no quantitative exposure information for these cases and no information defining or 29 enumerating the population from which these cases arose. 30 Mortality records (and death certificates) may not always reflect the true cause of death 31 for various reasons (e.g., misdiagnosis, improper recording on the death certificate, or miscoding This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-57 DRAFT--DO NOT CITE OR QUOTE 1 of the cause of death). For mesothelioma, the undercounting of cases (underascertainment) is a 2 particular concern given the limitations of the ICD classification systems used prior to 1999 3 [detection rates varied from 12% from ICD-9 codes alone to 83% from manual inspection of 4 death certificates (Davis et ah. 1992)1; recent studies demonstrated that ICD-10 coding has 5 detection rates similar to the latter rate above (Camidge et al,, 2006; Pinheiro et al.,, 2004). The 6 appropriate procedure for pre-ICD-10 codes is not to use ICD codes alone but to manually 7 inspect death certificates, as was done by Sullivan (2007). There is also evidence that the 8 detection rate of peritoneal mesothelioma is much lower than pleural mesothelioma (Selikoff and 9 Seidman, 1992). This current assessment has accounted for the impact of this 10 underascertainment on the final IUR (see Section 5.4.5.1.1). 11 Lung-cancer mortality was based on the underlying cause of death identified by the ICD 12 code on death certificates according to the ICD version in use at the time of death. Based on 13 these different ICD codes, lung-cancer mortality included malignant neoplasms of the trachea, 14 bronchus, and lung, and was identified by the following codes: ICD-5 code `047' (excluding 15 `47c, Cancer of unspecified respiratory organs'), ICD-6 codes `162' or `163,' ICD-7 codes `162' 16 or `163' (excluding `162.2, Cancer of the pleura'), 1CD-8 and ICD-9 code `162', and ICD-10 17 codes `C33' or `C34'. In all, there were 111 deaths, with an underlying cause attributed to lung 18 cancer. All deaths after 1960 were coded as bronchus or lung because the ICD versions in use as 19 that time distinguished malignant neoplasms of the trachea as distinct from bronchus and lung. 20 Other investigators of this cohort have used different definiti ons of lung cancer or used different 21 follow-up periods, as described in Section 4.1.1.2.2 (Description of Cohorts). 22 23 5.4.2.3. Description o f Libby Amphibole Asbestos Exposures 24 The mining, milling, and processing operations at the mine and in and around Libby, 25 conditions of exposure, and job-specific estimates of exposure intensity have been thoroughly 26 described in Section 4.1 (Sullivan. 2007: Amandus et al.., 1987a: McDonald et al,, 1986a). 27 Briefly, miners extracted vermicuiite ore from an open-pit mine that operated on Zonolite 28 Mountain outside the town of Libby, MT. The ore was processed in a dry mill (1935-1974) 29 and/or two wet mills (1950-1974 and 1974-1990). The resulting concentrate was transported by 30 railroad to processing plants around the United States where the vermicuiite was expanded for 31 use in loose-fill attic insulation, gardening, and other products (see Section 2.1). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-58 DRAFT--DO NOT CITE OR QUOTE Exposure Intensity 1 2 Figure 5-3. Plot of the NIOSH job-exposure matrix for different job 3 categories over time. The height of each bar represents the intensity of exposure 4 as an 8-hour TWA (fibers/cc) for a job in a particular year. Each row for 5 "Selected Jobs" represents a specific job category. 6 7 8 EPA adopted the JEM developed and used by Sullivan (2007) (see Figure 5-3), which 9 was, in turn, based on that used in the earlier NIOSH study for jobs through 1982 (Amandus et 10 al., 1987a; Amandus and Wheeler. 1987). As discussed in more detail in Section 4.1, Amandus 11 et al. (1987a) defined 25 location operations to which they assigned exposure intensity based on 12 available information (see Table 5-7). Ajob category may have involved more than one location 13 operation, and the 8-hour time-weighted average exposure (8-hour TWA) for each job category 14 in the JEM was calculated from the exposure intensity and time spent at each location operation 15 (Amandus et al., 1987a). 16 For the later data in Table 5-7 from 1967 through 1982, over 4,000 air samples analyzed 17 for fibers by PCM analysis were available to inform the exposure intensity for the 25 location 18 operations (see Table 5-7). Therefore, the JEM for 1968-1982 is based on direct analytic This document is a draftfor review purposes only and does not constitute Agency policy. 5-59 DRAFT--DO NOT CITE OR QUOTE DRAFT--DO NOT CITE OR QUOTE 5-60 This docum ent is a draftfo r re\'iew purposes only and does n ot constitute A gency policy. Table 5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of operations through 1982 [Amandus et al. (1987a): Table VII] Location operation Downtown office building Bus ride Mine office Mine raise. Mine-- nondrilling Transfer point Quality control lab Service area by mill Dry mill Dry mill sweeping Old and new wet mill-- millwright Old wet mill-- nonmillwright New wet mill-- nomnillwright Skip area Concentrate hauling River station binside River conveyor tunnel River office binside Verxite plant Bagging plant Tails belt <50 0 1.2 1.0 1.6 2.6 2.2 13.1 1.9 168.4 182.1 88.3 5.5 21.2 112.5 10.6 22.6 12.9 7.3 50-59 0 1.2 1.0 1.6 2.6 2.2 13.1 1.9 168.4 182.1 7.0 3.7 88.3 5.5 21.2 112.5 10.6 22.6 12.9 7.3 60-63 0 1.2 1.0 1.6 2.6 2.2 13.1 1.9 168.4 182.1 7.0 3.7 88.3 5.5 21.2 112.5 10.6 2.8 12.9 7.3 64-67 0 1.2 1.0 1.6 2.6 2.2 2.6 3.8 33.2 35.9 7.0 3.7 17.4 5.5 21.2 112.5 10.6 2.8 12.9 7.3 Year 68-70 71 0 0 1.2 1.2 1.0 1.0 1.6 1.6 2.6 2.6 2.2 2.2 2.6 2.6 1.9 1.9 33.2 33.2 35.9 35.9 7.0 7.0 3.7 3.7 - - 17.4 17.4 5.5 5.5 21.2 21.2 112.5 112.5 10.6 10.6 2.8 - 12.9 12.9 7.3 7.3 72-74 0 1.2 1.0 1.6 2.6 2.2 2.6 1.9 16.6 19 7.0 3.7 3.2 4.8 5.5 21.2 112.5 10.6 4.3 7.3 75-76 0 0 0.5 0.8 0.6 0.6 0.6 0.2 0.6 2.0 0.6 0.4 0.7 0.3 0.2 1.2 0.7 77-79 0 0 0.5 0.8 0.6 0.6 0.6 0.2 0.6 0.8 0.6 0.4 0.7 0.3 0.2 1.2 0.7 80-82 0 0 0.5 0.8 0.6 0.6 0.6 0.2 0.6 0.8 0.6 0.4 0.7 0.3 0.2 1.2 0.7 Table 5-7. Exposure intensity (fibers/cc) for each location operation from the beginning of operations through 1982 (continued) This document is a draftfor review purposes only and does not constitute Agency policy. Location operation Screen plant Drilling High Low Ore loading High Low River dock High Low Bagging plant High Low <50 23 6.7 82.5 24 116.9 38 12.9 4.6 50-59 23 6.7 27.7 15 42.5 19 12.9 4.6 60-63 23 6.7 10.7 9 17 6.4 12.9 4.6 64-67 23 6.7 10.7 9 17 6.4 12.9 4.6 Year 68-70 71 - - 9.2 9.2 6.7 9.2 3.2 3.2 3.2 3.2 17 5.1 5.1 5.1 12.9 12.9 4.6 4.6 72-74 9.2 9.2 3.2 3.2 5.1 5.1 4.3 4.3 75-76 0.5 0.6 0.6 0.2 0.2 0.5 0.5 1.2 1.2 77-79 0.5 0.6 0.6 0.2 0.2 0.5 0.5 1.2 1.2 80-82 0.5 0.6 0.6 0.2 0.2 0.5 0.5 1.2 1.2 5-61 DRAFT--DO NOT CITE OR QUOTE 1 measurements in air for each location operation (Amandus et al., 1987a). With the exception of 2 the dry mill, no air samples were available for other location operations at the mine and 3 processing facilities prior to 1967. In order to estimate exposures that occurred before that time, 4 the NIOSH researchers interviewed plant employees and based estimates of exposure intensities 5 on known changes in operations over the years and professional judgments regarding the relative 6 intensity of exposure; exposure intensity for 23 of the pre-1967 location operations was 7 extrapolated from post-1967 measurements based on reasoned assumptions for each location 8 operation (Amandus et al.. 1987a). 9 However, the amount and quality of measurement data in the facility in earlier years were 10 much more limited (Amandus et al.. 1987a). A total of 40 dust samples were taken, exclusively 11 in the dry mill, over the years 1950-1964. Using these measurements, much higher exposures 12 were inferred to occur prior to 1964 than those measured in later years. Although air sampling 13 for fibers by PCM was available beginning in 1967, average fiber concentrations (dry mill) 14 differed rather widely between limited data sets from different investigators up through the early 15 1970s: 1967-1968, NIOSH data, 65 fibers/cc (n = 14); 1970, company data, 11 fibers/cc 16 (n = 15); 1971, Mine Safety and Health Administration (MSHA) data, 31 fibers/cc (n = 52); 17 1972, MSHA and company data, 15 fibers/cc (n = 45). Thus, estimated exposure levels continue 18 to be uncertain during the period when fiber concentration measurements by PCM became 19 available in 1967. 20 Air samples collected by the State of Montana were available for the dry mill 21 from 1956-1969, but these were analyzed for total dust, not asbestos fibers. Total dust samples 22 (collected by a midget impinger) were examined by light microscopy, but no distinction was 23 made between mineral dusts, debris, and asbestos fibers. All objects were counted and reported 24 in the units of million particles per cubic foot of air (mppcf). Amandus et al. (1987a) developed 25 a relationship between total dust and asbestos fiber counts based on the comparison of 26 contemporaneous air sampling in the dry mill (see Section 4.1.1.2). The conversion ratio of 27 4.0 fibers/cc per mppcf was used to estimate exposure intensity for two location operations in the 28 dry mill for the years prior to 1967. 29 The exposure intensity (fibers/cc) for each of the location operations (see Table 5-7) was 30 used to calculate an estimate of daily occupational exposure for each job category in the JEM 31 (see Figure 5-3). For each job, the time spent at each location operation and the exposure This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-62 DRAFT--DO NOT CITE OR QUOTE 1 intensity for each location operation were averaged to derive an estimate of the 8-hour TWA. 2 The resulting JEM available for this current assessment and previous epidemiologic studies of 3 the Libby worker cohort is based on the air concentration of fibers as enumerated by PCM, 4 which measures fibers longer than 5 pm with an aspect ratio >3:1 [i.e., the fiber size regulated 5 under the OSHA standard (OSHA. 2006)1 Additionally, only fibers that are wide enough to be 6 viewed on PCM can be detected with this method. Amandus et al. (1987a) considered fibers 7 >0.44 pm in diameter to be visible by PCM in the historical filter analysis. More recent 8 techniques have refined the PCM method, and fibers greater than 0.25 pm in diameter are now 9 considered PCM fibers (IPCS. 1986). 10 There was one important limitation of the NIOSIT work history data In the earlier study 11 (Amandus and Wheeler. 1987). workers with "common laborer" job assignments and some 12 workers with unknown job assignments hired between 1935 and 1959 were assigned the 13 relatively low exposure levels estimated for the mill yard (Sullivan. 2007). Of the 991 workers 14 hired before 1960, 811 workers had at least one job with an unknown job assignment, with 15 706 having all department and job assignments prior to 1960 listed as unknown. In the more 16 recent study by Sullivan (2007), these workers were assigned the same relatively high time 17 weighted average estimated exposure intensity (absolute majority of these workers were assigned 18 66.5 fibers/cc) for all jobs during that time period. The lack of information on specific job 19 assignments for such a large portion of these early workers when exposures were higher resulted 20 in the mifclassification of the exposure and effectively yielded exposure metrics that were 21 differentiated only by the duration of each worker's employment. Because of the lack of more 22 specific measured fiber exposure data during this early period, the EPA experienced difficulties 23 in identifying an adequate exposure-response model fit for the complete cohort. These 24 difficulties are described in detail in Section 5.4.3.5. 25 As a result, the IUR analyses were based on the subset of workers hired after 1959 (i.e., 26 on or after January 1, 1960) and consisted of 880 workers. Of these 880 workers hired after 27 1959, 28 workers had at least one job with an unknown job assignment with 9 having all job and 28 department assignments between 1960-63 listed as unknown. These workers were assigned a 29 time-weighted average estimated exposure intensity of 66.3 fibers/cc. In addition, reabstracting 30 work histories for the more recent study (Sullivan. 2007) identified several job assignments not 31 mentioned in the earlier publications. Sullivan (2007) estimated exposure for the additional job This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-63 DRAFT--DO NOT CITE OR QUOTE 1 and calendar time period-specific combinations based on professional experience and review of 2 exposure records from earlier studies of the Libby worker cohort (Amandus et a!.. 1987a; 3 Amandus and Wheeler. 1987; McDonald et al.. 1986a). Uncertainties in the exposure 4 assessment for this sub-cohort are described in Section 5.4.6.1.2.4. While the Sullivan (2007) 5 study was limited to the white male workers, EPA's analysis includes all workers regardless of 6 race or gender. Table 5-8 shows the demographic and exposure characteristics of the sub-cohort 7 hired after 1959. Figure 5-3 shows a three-dimensional representation of the job-exposure 8 matrix used by Sullivan (2007) and in this current assessment. Not all jobs were included; thus, 9 the figure is not comprehensive but rather illustrative. The three axes show the intensity of fiber 10 exposure as an 8-hour TWA (fibers/cc, vertical axis) for selected job categories over time 11 (horizontal axes). For several jobs, the estimated 8-hour TWA was greater than 100 fibers/cc for 12 the decades prior to 1963. Figure 5-3 shows the variability in exposures across jobs and over 13 time. From 1967-1982, all exposure measurements that inform the JEM are based on 14 location-specific air samples analyzed for fibers by PCM. As stated above, pre-1968 exposures 15 in the dry mill were based on the measurement of dust levels from 1956-1967 that were 16 converted to PCM by Amandus et al. (1987a) and extrapolated backwards in time. Pre-1968 17 exposures for all other locations within the JEM were extrapolated from post-1967 fiber levels 18 based on reasoned assumptions (Amandus et al.. 1987a). 19 Amandus et al. (1987a) recognized the uncertainty in the pre-1968 exposures assigned to 20 the cohort. Although there is some uncertainty in the dust-to-fiber conversion, this conversion 21 (4.0 fibers/cc per mppcf) was based on contemporaneously collected dust and fiber data collected 22 in the dry mill and only applied to the dry mill environment. Amandus et al. (1987a) considered 23 a range of possible conversion factors (1.2-11.5 fibers/cc per mppcf). Greater uncertainty may 24 lie with the reasoned assumptions used to extrapolate exposures to the early decades for all 25 which Amandus et al. estimated a range of possible exposure intensities: drilling, ore loading, the 26 river dock, and the bagging plant, where intensity of exposure may vary as much as threefol d 27 between the low and high estimates (see Table 5-8). Finally, some workers were employed after 28 1982 through 1993 when demolition of the facilities was completed (Larson et al.. 2010b). 29 These exposures were not evaluated by Sullivan (2007) and were not included in the NIOSH 30 JEM. However, only 148 sub-cohort workers were still employed on May 31, 1982, according 31 to the NIOSH records. Because exposure concentrations in 1982 (see Table 5-7) were generally This document is a draftfor review purposes only and does not constitute Agency policy. 5-64 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-8. Demographic and exposure characteristics of the subset of the 2 Libby worker sub-cohort hired after 1959 3 C h aracteristic Sub-cohort hired after 1959 Number o f workers 880 Number o f deaths from all causes 230 Number o f deaths from mesothelioma 7 Number o f deaths from lung cancer 32 Mean year o f birth 1942 Mean year o f hire 1971 Mean age at hire (years) 28.6 Mean person-years o f follow-up (no lag) 32.2 Total person-years o f follow-up (no lag) 28,354 Mean employment duration (years) 3.3 Mean cumulative exposure (fiber/cc-year) 19.2 Median cumulative exposure (fiber /cc-year) 3.4 Range o f cumulative exposures (no lag) (fiber/cc-year)a 0-462 4 5 According to the work histories and JEM, there were 21 sub-cohort workers who had zero cumulative 6 exposure. These 21 individuals all worked at the office downtown. 7 8 9 location operations considered. For example, there were four location operations for below 1 10 ftber/cc, with only two locations having concentrations of 1.2 ftbers/cc, it is unlikely that these 11 workers' exposures were significantly underestimated. Uncertainties in all aspects of JEM are 12 described in Section 5.4.6.1.2. 13 14 5.4.2.4. Description o f Libby Worker Cohort Work Histories 15 NIOSH staff abstracted demographic data and work history data from company personnel 16 and payroll records, including W-4 federal tax forms. An individual's work history was 17 determined from job change slips, which recorded new job assignment, date of change, and 18 change in hourly pay rate (which differed by the job assignment). Work history records span the 19 time period from September 1935 to May 1982. Dates of termination were unknown for 58 of 20 640 workers (9%), who left employment before September 1953. EPA adopted the assumption 21 used by NIOSH (Sullivan. 2007) that these people worked for 384 days, based on the mean 22 duration of employment among all workers with known termination dates before September 23 1953. The majority of workers in this cohort as a whole and among those hired on or after 24 January 1, 1960, worked at multiple jobs; many of the workers switched jobs repeatedly or had This document is a draftfor review purposes only and does not constitute Agency policy. 5-65 DRAFT--DO NOT CITE OR QUOTE 1 the estimated exposure for a job change from one year to the next. Of the 880 workers hired in 2 1960 or afterwards in the sub-cohort, the mean number of times a worker's exposure level 3 changed according to the JEM was 5, the median was 2, and the maximum number of changes 4 was 32 (see Figure 5-4; see also Figure 5-3 for a depiction of job-exposure intensities for 5 different j obs over time). 6 7 Num ber of Different Jobs Heicl A m ong Workers Flired After 1959 3or' Number of Jobs Held by Worker 8 9 Figure 5-4. Histogram showing the number of workers who experienced 10 each incremental number of different jobs among the 880 workers hired 11 after 1959. 12 13 14 5.4.2.5. Estimated Exposures Based on Job-Exposure Matrix (JEM) and Work Histories 15 Exposure-response modeling of epidemiologic data is based on several considerations as 16 summarized by Finkelstein (1985): 17 18 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-66 DRAFT--DO NOT CITE OR QUOTE 1 After identification of an occupational hazard one of the goals of occupational 2 epidemiology is to quantify the risks by determining the dose-response relations 3 for the toxic agent. In many circumstances little is known about the dose received 4 by target tissues; the data available usually pertain only to exposure to various 5 concentrations of the toxic material in the workplace. The calculation of dose 6 requires additional physiological and chemical information relating to absorption, 7 distribution, biochemical reactions, retention, and clearance. 8 9 10 In asbestos epidemiology the usual measure of exposure is the product of the 11 concentration of asbestos dust in the air (fibers or particles per ml) and the 12 duration of exposure to each concentration summed over the entire duration of 13 exposure (years); this measure is the cumulative exposure.... 14 15 16 Cumulative exposure has been the traditional method of measuring exposure in 17 epidemiologic analyses of many different occupational and environmental exposures and was the 18 exposure metric applied to the risk of lung-cancer mortality in the Integrated Risk Information 19 System (IRIS) assessment for general asbestos (U.S. EPA. 1988a). Two alternative approaches 20 to developing exposure metrics to describe the effects of concentrations of asbestos dust in the 21 air on the risks of mortality have also been proposed. The first alternative was proposed by 22 Jahr (1974). who studied silica-induced pneumoconiosis and suggested that exposures to 23 occupational dusts could be weighted by the time since exposure. This yields an exposure metric 24 that gives greater weight to earlier exposures. Berry et al. (1979) subsequently suggested the 25 application of exposure metri cs that allowed for the clearance of dust or fibers by using a decay 26 term on exposures. For the evaluation of mortality risk from mesothelioma, U.S. EPA (1988a) 27 used a different exposure metric than was used for lung-cancer mortality, which factored in the 28 time since first exposure. As observed in U.S. EPA (1988a). it is important to note that different 29 characterizations of estimated ambient exposures may be reasonably expected to be associated 30 with different endpoints. 31 Most studies of asbestos-related mortality have evaluated either cumulative exposure, 32 exposure concentration, or the duration of employment as exposure metrics. Many studies have 33 been limited in the availability of detailed exposure data--especially at the individual level. In 34 the Libby worker cohort data developed by NIOSIT and used in this current assessment, detailed 35 work histories, together with job-specific exposure estimates, allowed for the reconstruction of 36 each individuaTs estimated occupational exposure over time to define multiple exposure metrics. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-67 DRAFT--DO NOT CITE OR QUOTE 1 From this information-rich, individual-level data set from NIOSH, EPA constructed a 2 suite of the different metrics of occupational exposure, which had been proposed in the asbestos 3 literature or used in the IRIS asbestos assessment (U.S. EPA, 1988a). This suite of models was 4 defined a priori to encompass a reasonable set of proposed exposure metrics to allow sufficient 5 flexibility in model fit to these data. These exposure metrics were evaluated in 6 analytic-regression models to test which exposure metrics were the best empirical predictors of 7 observed cancer mortality, and the better fitting models were advanced for consideration as the 8 basis of the exposure-response relationship for the 1UR. The types of exposure metrics evaluated 9 were intended to allow for variations of the classic metric of cumulative exposure, allowing for 10 more or less weight to be placed on earlier or later exposures. These simulated exposure metrics 11 were derived mathematically to approximate underlying processes that are not well understood 12 (see Section 5.4.6). Thus, the fit of exposure metrics is evaluated on the basis of maximizing the 13 likelihood for the Libby worker cohort, and the estimated parameters do not necessarily have 14 biological interpretations. 15 The first exposure metric--cumulative exposure (CE)--is a simple addition of each day 16 of exposure across time (see Eq. 5-1). CE has been widely used in modeling risk of cancer in 17 occupational epidemiology and has been used for modeling lung cancer (Larson et al., 2010b; 18 Moolgavkar et al., 2010; Sullivan, 2007; McDonald et al, 2004) and mesothelioma (McDonald 19 et al., 2004) in the Libby worker cohort. When using this exposure metric in the risk model, all 20 exposures have equal weight regardless of when they occurred and lead to the same estimated 21 cancer ri sk whether exposure happened early or later in life. 22 EPA calculated each individual's occupational CE to Libby Amphibole asbestos over 23 time from their date of hire until the date they ceased to be employed in the Libby operations or 24 until the date NIOSIT collected the work history data, if still employed in May 1982. Workers 25 were assumed to remain at their final occupati onal CE level until death or the end of the 26 follow-up period on December 31, 2006. Each worker's CE at any time point (daily increment) 27 since their date of hire was computed as the sum of their exposure intensity (fibers/cc) on each 28 specific occupational day (xt) from day 1 through day k. Mathematically, this was defined as 29 30 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-68 DRAFT--DO NOT CITE OR QUOTE 1 2 3 4 Where CE at time %= k ^ xtt 7=1 ` (Eq. 5-1) 5 6 xt. =the estimated job-specific exposure intensity for the day tp and 7 4 = the day on which the exposure is estimated. 8 9 10 A second exposure metric--residence time-weighted (RTW) exposure--gives additional 11 weight to early exposures. By doing so, the RTW exposure metric allows the possibility that 12 early exposures are more influential on cancer mortality predictions in the model. Unlike many 13 chemicals that are rapidly metabolized in the body and excreted, asbestos fibers are durable, and 14 some may remain in the body for years. Fibers that remain in the lung may continue to damage 15 lung cells and tissue until they are removed or cleared (see Section 3.2). Similarly, fibers that 16 translocate to the pleura may damage cells as long as they remain in this tissue. Therefore, a 17 fiber exposure may not only damage tissue during the exposure, but fibers may remain in these 18 tissues, with cellular and tissue damage accumulating over time. 19 The RTW exposure metric in this current assessment is sometimes called the cumulative 20 burden, or the area under the curve. A type of RTW metric was proposed for modeling of 21 mesothelioma mortality by Newhouse and Berry (1976) based on a general understanding of the 22 relationship between tumor incidence rate and time to cancer (Cook et aU 1969) as well as 23 animal models of mesothelioma (Berry and Wagner. 1969). Simil ar types of RTW metrics were 24 applied to the insulators asbestos cohort by Peto et al. (1982). discussed by Finkelstein (1985). 25 and applied in the derivation of the IUR in the IRIS assessment for asbestos (U.S. EPA. 1988a). 26 McDonald et al. (2004) and Moolgavkar et al. (2010) used RTW-type metrics for modeling 27 mesothelioma in the Libby worker cohort, and McDonald et al. (2004) applied an RTW metric 28 for modeling lung-cancer mortality in the Libby worker cohort. 29 In calculating RTW, each day's exposure is multiplied by the time since the exposure 30 occurred (see Eq. 5-2). RTW CE was calculated as a cumulative function of each time-interval's 31 CE such that earlier exposures contribute greater weight. 32 This document is a draftfor review purposes only and does not constitute Agency policy. 5-69 DRAFT--DO NOT CITE OR QUOTE 1 k J 2 RTW CE at time g- = Xtj (Eq. 5-2) ii i 3 4 Where 5 6 xt. = the estimated job-specific exposure intensity for the day th and 7 4- = the day on which the exposure is estimated. 8 9 10 The CE and RTW exposure metrics result in increasing or sustained metrics of exposure 11 across time. Elowever, it is known that some cellular and genetic damage may be repaired over 12 time, which could decrease cancer risk from exposure over time. Additionally, asbestos fibers 13 are cleared (removed) from the lung through natural processes and translocated to other tissues 14 (see Section 3.2.1.1). Therefore, when considering lung cancer, it is possible that removal of 15 asbestos fibers from the lung would reduce lung cancer risk over time. Although less is known 16 about removal of asbestos from the pleura, there may be clearance mechanisms operative in that 17 tissue as well (see Section 3.2.1.2). As noted earlier, Berry et al. (1979) proposed the use of 18 exposure metrics based on occupational exposures, which addressed the issue of clearance 19 through a mathematical decay term that modified measured ambient exposures. For 20 mesothelioma, modeling a decay term on exposure has been proposed by Berry (1999). Based 21 on this proposal, several recent papers applied a decay term to modeling mesothelioma mortality 22 (Berry et al., 2009; Reid et al.., 2009; Barone-Adesl et al... 2008; Gasparrini et al,, 2008; Clements 23 et al.. 2007; Ftodgson et al., 2005; Berry et al., 2004). Similarly, recent publications indicate that 24 the relative risk of lung cancer due to asbestos exposure declines 15-20 years after the cessation 25 of exposure to asbestos (Magnani et al., 2008; Elauptmann et al, 2002). 26 Mathematically allowing for the magnitude of earlier exposures to diminish with 27 advancing time was considered to be a method of giving less weight in the analyses to earlier 28 exposures compared to the previous two exposure metrics. Therefore, two additional exposure 29 metrics were considered, where a decay rate was applied to the CE and RTW exposure metrics 30 (see Eq. 5-3 and 5-4). This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-70 DRAFT--DO NOT CITE OR QUOTE 1 For each exposure metric, the application of a half-life was calculated by depreciating 2 each time-interval's (/ -i:/) exposure according to a model of exponential decay with various 3 half-lives {Tm) of 5, 10, 15, and 20 years. Note that the particular kinetics of Libby Amphibole 4 asbestos fibers are not fully understood, and the relevance of these particular half-lives was 5 determined from the statistical fit of these exposure metrics to the risk of cancer mortality, rather 6 than the biological half-life of the fibers. For a very large half-life, decay is very slow, and these 7 metrics would be very similar to the CE and RTW exposure metrics. 8 9 10 (Eq. 5-3) 11 12 Where 13 14 Xnj = the estimated job-specific exposure intensity for the day tj, and 15 tk = the day on which the exposure is estimated. 16 I Ik j xt *exp ln(0-5)*(t/c -(') 17 RTW with half-life at time i | = % (Eq. 5-4) /=! N 18 19 20 In addition to the exposure metrics used in the lung-cancer mortality analysis, modeling 21 of mesothelioma mortality (see Section 5.4.3.1) included the exposure model used in the IRIS 22 assessment for asbestos (U.S. EPA. 1988a). originally proposed in Peto et al. (1982): 23 24 25 Im = CO-KM (Eq. 5-5) 26 27 Where 28 29 Im = the observed deaths from mesothelioma/person-years, 30 C. = the average concentrati on of asbestos in the air, This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-71 DRAFT--DO NOT CITE OR QUOTE 1 KM = an estimated slope describing the relationship between Libby Amphibole asbestos 2 exposure and mesothelioma mortality, and 3 0 = the function of the time since first exposure (/) and the duration of employment 4 ' w 5 For t < 10, O = 0 6 For 1 0 < ^ < d + 10, 0 = ( t - 10)3 7 For t> d + 10,Q = ( t - \ 0 f 8 9 10 The asbestos IUR (U.S. EPA. 1988a) metric (see Eq. 5-5) was originally fit to aggregate 11 cohort data and was based on a function of average cohort exposure, time since first exposure, 12 and duration of employment. The analysis here of individual data for Libby Amphibole asbestos 13 is, therefore, a different application of this exposure metric, and its fit to the mesothelioma 14 mortality of the Libby worker cohort is evaluated in this current assessment. 15 In addition to the use of these methods of describi ng exposure metrics representing 16 estimated ambient exposure to Libby Amphibole asbestos dust for use in predicting the risk of 17 mortality, there is the important issue of potentially modifying the exposure metrics to account 18 for cancer latency. Without knowledge of the specific timing of etiologically relevant exposure 19 that may initiate and promote cancers ultimately resulting in mortality, any exposure metric may 20 include exposures during some time period that do not have bearing on the risk of mortality. In 21 the absence of such information on the specific cancer latency associated with a specific 22 exposure, Rothman (1981) suggested that the most relevant exposure period could be identified 23 by comparing the fit of exposure metrics across multiple lag periods to allow for the 24 identification of the optimal latency period as an expression of a lag time between exposure and 25 mortality. This has since become a standard practice in occupational and environmental 26 epidemiology. Accordingly, exposure estimates for all exposure metrics were adjusted to 27 account for the time period between the onset of cancer and mortality. The lag period defines an 28 interval before death, or end of follow-up, during which, any exposure is excluded from the 29 calculation of the exposure metric. Cohort members who died or were lost withi n the initial 30 years of follow-up were assigned lagged exposure values of zero if they had not been followed 31 for longer than the lag time. The various exposure metrics were lagged at 10, 15, and 20 years to 32 account for different potential cancer latencies within the limitations of the available data. This docum ent is a draftfo r review purposes only and does not constitute A gency policy. 5-72 DRAFT--DO NOT CITE OR QUOTE 1 Metrics without a lag were fit for comparison purposes but were not considered to be 2 biologically reasonable, given that the outcome under analysis is cancer mortality (specifically, 3 mesothelioma and lung cancer), for which latency periods of 10 years or more have been 4 established for asbestos (U.S. EPA, 1988a). Consequently, metrics that were not adjusted by 5 lagging exposure in the final years before mortality (or the end of follow-up) were not 6 considered further in the development of an IUR for Libby Amphibole asbestos. 7 8 5.4.3. Exposure-Response Modeling 9 Sufficient biological information to select models for the epidemiology data on the basis 10 of biological mechanism (see Section 3) is not available. In this situation, EPA's practice is to 11 investigate a range of model forms to determine how to best empirically model the 12 exposure-response relationship in the range of the observed data. For Libby Amphibole 13 asbestos, possible exposure metrics were explored for model fit to the chosen models. The 14 exposure metric options were selected to provide a range of shapes that was sufficiently flexible 15 to allow for a vari ety of ways that time and duration might relate to cancer risk in the data being 16 modeled. EPA then evaluated how well the models and exposure metric combinations fit the 17 data being modeled. Metrics that did not fit the data well were rejected. For purposes of 18 calculating a reasonable upper bound on the risk per exposure, two different types of uncertainty 19 were accounted for. The first uncertainty is in the estimated slope for each exposure metric, and 20 this was accounted for by using the upper bound estimated using the statistical variance of the 21 estimated slope. EPA accounted for the second uncertainty that stemmed from the choice of 22 exposure metrics among the set that fit the data by using the exposure metric (among those few 23 with a reasonable fit) that estimated the highest risk (because formal estimation of an upper 24 bound was not possible). This is explained in more detail below and in Section 5.4.5. 25 The risk estimates are based on epidemiological analysis of the primary NIOSFI data 26 (Libby worker cohort). The rationale for selection of the Libby worker cohort is presented in the 27 previous section (see Section 5.4.2). Analysis of this primary epidemiologic database allows the 28 comparison of multiple metrics of exposure to quantify the exposure-response relati onship. This 29 approach is intended to support the empirical representation of the exposure-response 30 relationship of estimated ambient occupational exposure to Libby Amphibole asbestos with 31 observed cancer mortality risk. The exposure-response modeling may be influenced by This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-73 DRAFT--DO NOT CITE OR QUOTE 1 uncertainties in the magnitude and time course of the exposure estimates and, therefore, may not 2 necessarily reflect the biologic disposition of inhaled fibers (see Section 5.4.6). 3 The following sections provide information about modeling of the full cohort first, the 4 difficulties in identifying adequately fitting models to these data, and the decision to base the 5 analysis on a sub-cohort of workers that did allow for identifying adequately fitting models. 6 7 5.4.3.1. Modeling o f Mesothelioma Exposure Response in the Libby Worker Cohort 8 The background incidence of mesothelioma is extremely rare (Hillerdah 1983). Since 9 there is a very low background risk, the exposure-response model applied here examines the 10 relationship of the absolute risk of mesothelioma mortality that is attributable to Libby 11 Amphibole asbestos exposure because there is not a true background risk of mesothelioma 12 mortality among people who were truly unexposed to Libby Amphibole asbestos (as opposed to 13 the relative risk model, which is used for lung-cancer mortality; see Section 5.4.3.3). Poisson 14 regression models are employed here for estimating the absolute risk of mesotheli oma, as the 15 Poisson distribution is an appropriate model for use with data that are counts of a relatively rare 16 outcome, such as observed mesothelioma deaths in the Libby worker cohort. Other analyses of 17 mesothelioma mortality in the Libby worker cohort have also used the Poisson regression model 18 (Moolgavkar et aL 2010; McDonald et aL 2004). In the Poisson regression model, probability 19 of k events is specified as 20 21 22 23 (Eq. 5-6) 24 25 26 where %is parameterized with the exposure metric (defined in Section 5.4.2.5). Then, life-table 27 analysis is used to estimate risks in the general U.S. population for the derivation of the unit risk 28 of mesothelioma mortality (see Section 5.4.5.1). 29 Estimati on of the exposure-response relationship for mesotheli oma mortality using the 30 Poisson regression model was performed using a Monte Carlo Markov Chain (MCMC) Bayesian 31 approach with an uninformative or diffuse prior [WinBUGS Version 1.4 (Spiegelhalter et ah. 32 2003)1. Use of diffuse priors is a standard procedure in Bayesian analysis, in situations like this This document is a draftfor review purposes only and does not constitute Agency policy. 5-74 DRAFT--DO NOT CITE OR QUOTE 1 one, when there is no prior knowledge about the toxicity of Libby Amphibole asbestos under a 2 particular model. Since this analysis focuses only on the Libby worker cohort and does not try to 3 factor in data from other sources in estimating potency, use of a diffuse prior is considered 4 appropriate for this analysis. 5 The benefit of using the WinBUGS software is its computational ease and that it provides 6 a posterior distribution of P (the mesothelioma slope factor) rather than just a point estimate. A 7 diffuse (high variance) Gaussian distribution, truncated to exclude negative parameter values, is 8 used as a diffuse prior. With such a prior, results of MCMC analysis are expected to be similar 9 to maximum likelihood estimation in a non-Bayesian analysis. Standard practices of MCMC 10 analysis were followed for verifying convergence and sensitivity to the choice of initial values. 11 The posterior distribution is based on three chains with a burn-in of 10,000 (i.e., the first 12 10.000 simulations are dropped so that remaining samples are drawn from a distribution close 13 enough to the true stationary distribution to be usable for estimation and inference) and thinning 14 rate of 10 (i.e., only each 10th simulation is used--thus reducing autocorrelation) such that 15 3.000 total simulations constitute the posterior distribution of p. The mean of the posterior 16 distribution served as a central estimate, and the 90% credible interval3defined the 5th percentile 17 and the 95thpercentile of the distribution, which served as bounds for the 95th lower and upper 18 one-sided confidence intervals, respectively. 19 Multiple metrics of exposure (see Section 5.4.2.5) as well as exposure intensity, duration 20 of employment, age at death or loss to follow-up, and time since first exposure were compared 21 using the Deviance Information Criterion (D1C). The DIC (Spieselhalter et aL 2002) is used in 22 Bayesian analysis and is an analogue of the AIC, with smaller values indicating a better 23 statistical fit to the data. Use of the DIC and AIC is standard practice in comparing the fit of 24 nonnested models to the same data set with the same dependent outcome variable but different 25 independent covariates. According to Burnham and Anderson (2002), "These methods allow the 26 data-based selection of a "best" fitting model and a ranking and weighting of the remaining 27 models in a predefined set." Because of the small number of deaths from mesotheliomas in 28 absolute terms, only uni- and bi-variate models (with age or time since first exposure as the 29 second covariate) were considered. Sex and race were not used as covariates since all 30 mesotheliomas were observed in men assumed to be white (Sullivan. 2007). Each exposure A credible interval is the Bayesian analogue o f a confidence interval. This document is a draftfor review purposes only and does not constitute Agency policy. 5-75 DRAFT--DO NOT CITE OR QUOTE 1 metric was lagged by 0, 10, 15, or 20 years. The use of a lag period aims to account for the 2 latency period between the onset of mesothelioma (which occurs some time before clinical 3 diagnosis) and mesothelioma mortality. 4 5 5.4.3.2. Mesothelioma Mortality Analysis in the Libby Worker Cohort 6 For the full Libby worker cohort (n = 1,871), the duration of employment provided a 7 considerably better univariate model fit than the other possible exposure metrics, indicating that 8 this exposure metric was the best single predictor of mesothelioma mortality in the full Libby 9 worker cohort. The bivariate model, which included duration of employment and age at death or 10 censoring, provided the overall best fit (DIC = 196). The inclusion of information on the 11 concentration of exposure beyond the duration of employment resulted in a degradation in model 12 fit (see Table 5-9). The metric used in the ILJR for asbestos (U.S. EPA. 1988a) (see Eq. 5-5) had 13 a much higher DIC of 233.7 in the analysis here. It is likely that the poorer fit seen when using 14 information on exposure concentration is the result of the fact that duration of employment is 15 measured with comparatively little error, while derivation of specific exposure concentrations 16 may be subject to a sizable measurement error. Moreover, as described in Section 5.4.2.3, for 17 706 of 991 (71%) workers hired from 1935 to 1959, only the duration of employment was 18 known, but not the job category or department code, and, thus, the same time-weighted average 19 estimated exposure intensity for that time peri od had been assigned to 653 of these workers39 20 (Sullivan. 2007). It is likely that because of the potential for particularly large exposure 21 measurement error among more than two thirds of the workers hired prior to 1960 who were 22 assigned the same exposure intensity, this resulted in the duration of employment being the best 23 predictor of mesothelioma mortality. Additionally, estimates of exposure intensity prior to 1968 24 have greater uncertainty associated with them than more recent exposure measurements, which 25 are based on fiber counts in air samples analyzed by PCM. For the majority ofjob locations 26 (23 of 25), no exposure measurements were available prior to 1968, and exposures were 27 estimated based on employee interviews (in 1982) and what was known about major changes in 28 operations between 1935 and 1967. For two exposure locations, the dust-to-fiber conversion 29 ratio is based on measurements taken in the late 1960s, so extrapolations from the mid-1960s to 39Note that Sullivan (2007) analyzed the population of 1,672 white male workers rather than all 1,871 workers so the numbers o f workers with missing job category- and department information were different. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-76 DRAFT--DO NOT CITE OR QUOTE 1 the early 1960s is likely to be more certain than extrapolation further back in time. The fact that 2 the metric using only duration of employment fit best and the additional incorporation of 3 exposure intensity information worsened the fit indicates that it is unlikely that IUR estimates 4 can be developed using the full cohort data because exposure values were not predictive of 5 mesothelioma mortality. 6 7 8 Table 5-9. Comparison of univariate model fit of various exposure metrics 9 for mesothelioma mortality in the full Libby worker cohort (n = l,871)a,b 101234567890 Variable DIC Duration o f employment 202.9 Age at death or censoring 209.2 CE lagged 15 yr 209.5 CE lagged 10 yr 209.9 RTW lagged 10 yr with 5-yr !/2life 210.4 CE lagged 10 yr with 20-yr !/2life 210.6 RTW with 5-yr Vi life 210.7 RTW with 10-yr Vi life 211.0 CE 211.4 Time since first exposure 211.4 11 12 "Since one o f the mesothelioma deaths occurred less than 20 years from start o f the exposure, lag 20 metrics 13 assigned no exposure to this case, which resulted in the very poor fit o f exposure metrics lagged 20 years. 14 bLower DIC values represent better fits. Models with DIC within 10 units o f the DIC o f the model with the lowest 15 DIC are shown. 16 17 DIC = Deviance Information Criterion. 18 19 20 The DIC values for models that included lag and/or half-life adjustments to the exposure 21 metrics were not penalized in the regression analyses for including these extra parameters 22 because those factors were not represented as covariates but rather were embedded in the 23 exposure metrics. While these results were obtained using each instance with lag and/or half-life 24 as a separate model fit, it may be appropriate to penalize the DIC values from these results for 25 inclusion of these parameters. Note that if the DIC values from the lag and/or half-life models 26 were penalized, this would serve to improve the relative fit of the model using only duration as a 27 parameter in comparison with the lag and/or half-life models because the DICs for the penalized 28 models would increase while the DIC for the unpenalized models would be unchanged. This document is a draftfor review purposes only and does not constitute Agency policy. 5-77 DRAFT--DO NOT CITE OR QUOTE 1 5.4.3.3. Modeling o f Lung Cancer Exposure Response in the Libby Worker Cohort 2 To develop an exposure-response relationship for lung cancer, the lung-cancer mortality 3 data were modeled as a function of the historical exposure data for the Libby worker cohort. The 4 mesothelioma mortality data were modeled to estimate the absolute risk because it is very rare in 5 the general population (Hillerdah 1983). Lung-cancer mortality does have a known background 6 risk, and, thus, modeling of lung-cancer mortality is based on the relative risk rather than the 7 absolute risk. As such, there are different analytic methods available that can use information on 8 time-varying exposures. The NIOSIT-developed individual-level exposure data for the Libby 9 worker cohort are very detailed, with start and stop dates for each of the workers'jobs and 10 estimated fiber exposures for 25 specific location-operations (Amandus et aL 1987a). It is, 11 therefore, important to find a model that makes efficient and effective use of these 12 time-dependent data. 13 The Cox proportional hazards model (Cox. 19721 is one of the most commonly used 14 statistical models for the epidemiologic analysis of survival and mortality in cohort studies with 15 extensive follow-up (Larson et al.. 2010b; Moolgavkar et ah. 2010). In the Cox proportional 16 hazards model, the conditional hazard function, given the covariate Z, is assumed to have the 17 form 18 19 20 A(t | Z ) = A0(t)Q xp(/Ji Z ) (Eq. 5-7) 21 22 23 where P is the vector of regression coeffici ents, Xo(t) denotes the baseline hazard function, and T 24 denotes transposition of the vector. One of the strengths of this model is that knowledge of the 25 baseline risk function is not necessary, and no particular shape is assumed for the baseline 26 hazard; rather, it is estimated nonparametrically. The contributions of covariates to the hazard 27 are multiplicative. When Z represents exposure and ft Z is small, the Cox proportional hazards 28 model is consistent with linearity of dose response for low doses. 29 When the proportional hazards assumption holds, it is possible to estimate the hazard 30 ratio of exposure (relative risk) without estimating the hazard function in the unexposed (or in 31 the lowest exposures seen within the study group) since this baseline hazard function drops out 32 of the calcul ations. The Cox proportional hazards model assumes that a function of covariates This document is a draftfor review purposes only and does not constitute Agency policy. 5-78 DRAFT--DO NOT CITE OR QUOTE 1 (i.e., exposures) result in risks that are a constant multiple of the baseline hazard in unexposed 2 individuals over some timescale, typically calendar time or age. This proportionality is assumed 3 to be constant across the range of observed exposures, given the set of modeled covariates, and 4 can be evaluated across time. 5 The Cox proportional hazards model was chosen to represent the lung-cancer mortality 6 data for several reasons. Of primary importance is that it takes statistical advantage of the 7 extensive exposure data collected for the cohort on time-varying exposures to Libby Amphibole 8 asbestos. There are no other standard model formulations that allow for the analysis of 9 time-varying exposures in the manner achieved by the Cox proportional hazards model. The 10 exposure-response relationship (proportional hazards ratio) determined in this model intrinsically 11 takes into account the effects of other causes of mortality that are unrelated to exposure (i.e., 12 independent censoring). Further, all comparisons are made within the cohort by comparing the 13 mortality experience of people with different exposures within the same cohort population. The 14 issue of competing risks that are dependent on exposure (e.g., asbestosis or nonmalignant 15 respiratory disease) is an acknowledged uncertainty for this type of analysis (see Section 5.4.6). 16 Other methods common to occupational epidemiology, such as the use of standardized 17 mortality ratios typically rely upon comparisons of the mortality experience in an exposed 18 population group compared to that in the general population. However, the comparison 19 population may not always be appropriate due to differences in general health status (e.g., the 20 healthy worker effect) and differences in exposure to other risk factors for a specific disease 21 (e.g., smoking history). The lack of comparability between the study population and the 22 comparison population can lead to confounding by other measured or unmeasured 23 characteristics, which may be statistically associated with both the exposure of interest and the 24 endpoint. The Cox proportional hazards model controls for such potentially confounding 25 characteristics by using a comparison group from within the study population (i.e., internal 26 controls). Internal controls are a statistically appropriate comparison group because they are 27 expected to be more similar in potentially confounding characteristics to the remainder of the 28 cohort, thereby controlling for both measured and unmeasured confounding and helping ensure 29 that comparisons are more statistically valid. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 5-79 DRAFT--DO NOT CITE OR QUOTE 1 5.4.3.4. Lung-Cancer Mortality Analysis in the Libby Worker Cohort 2 As described in the previous section, quantitative exposure-response relationships for 3 lung-cancer mortality were evaluated using the Cox proportional hazards model. Cox 4 proportional hazards models of this type require the specification of a timescale. Age is typically 5 the time-related variable with the strongest relationship to cancer mortality and was used as the 6 timescale in these analyses. Use of age as the timescale in a time-varying Cox proportional 7 hazards model controls for age as a risk factor by design rather than by parametric modeling and 8 effectively rules out age as a potential confounder. Individual covariates available to EPA in the 9 complete analytic data set compiled from the NIOSH data were evaluated for their ability to 10 explain the lung-cancer mortality. These included sex, race, birth year, age at hire, and various 11 exposure-related variables including TWA workplace intensity of exposure in fibers/cc, job type, 12 and the start and stop date of each different job. These data allowed for the computation of 13 cumulative exposure, cumulative exposure with application of a half-life, and RTW cumulative 14 exposure, with and without application of a half-life (see Section 5.4.2.5). Each exposure metric 15 was also lagged by 0, 10, 15, or 20 years. The use of a lag period aims to account for the latency 16 period between the onset of lung cancer (which occurs some time before clinical diagnosis) and 17 lung-cancer mortality. 18 All lung-cancer mortality analyses were conducted using SAS software version 9.1 (SAS, 19 Cary, NC). EPA fit the extended Cox proportional hazards model (Tableman and Kim. 2004; 20 Kleinbaum and Klein, 1996). which included both time-independent factors such as sex, race, 21 and date of birth, as well as time-dependent measures of Libby Amphibole asbestos exposure 22 over the entire time course of each individuals' lifetime from their date of hire until death or loss 23 to follow-up. This method allows for control of potential confounding by age by design rather 24 than through multivariate covariate modeling. The inclusion of date of birth in these analyses 25 controls for any potential birth cohort effect. 26 EPA's analyses of time-dependent exposure data included goodness-of-fit testing of the 27 proportionality assumption for the Libby worker cohort. Because Cox proportional hazard 28 models rely on the assumption that the hazard rate among the exposed is proportional to the 29 hazard rate among the unexposed, it is important to evaluate the model against this assumption. 30 Therefore, analyses of extended Cox proportional hazards models tested this assumption using a 31 Wald test on the model interaction term between the Libby Amphibole asbestos exposure metric This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-80 DRAFT--DO NOT CITE OR QUOTE 1 and the timescale (i.e., age). As a general rule, a nonzero slope that is either increasing or 2 decreasing indicates a violation of the proportional hazards assumption. Wald tests for the 3 complete cohort consistently showed that the interaction term was a statistically significant 4 predictor of lung-cancer mortality (p < 0.05) and was interpreted as evidence that the hazards did 5 not remain proportional over time. The cause of the lack of proportionality is unknown, but 6 several likely explanations are discussed in Section 5.4.3.5 below and in the discussion of 7 uncertainties in Section 5.4.6.1. 8 9 5.4.3.5. Summary o fMesothelioma and Lung Cancer Analysis o f Libby Worker Cohort 10 Several possible explanations exist for the finding that duration of employment was the 11 best fitting exposure metric for mesothelioma mortality, as well as the finding of the lack of 12 proportionality of hazards in the lung-cancer mortality modeling. 13 14 15 1) Duration of employment, but neither department code nor job category, was known for 16 706 of 991 (71%) workers hired from 1935 to 1959. Without knowledge of the job 17 category, the same exposure concentration had been assigned to almost all of these 18 workers, likely resulting in a particularly large measurement error for exposure in 19 approximately one third of the total cohort of 1,871 workers. This is a very likely 20 explanation for the superior fit for duration of employment in modeling of mesothelioma 21 mortality relative to the other exposure metrics based on measured exposures. Assigning 22 the same exposure concentration to so many of the workers hired before 1960, regardless 23 ofjob, likely resulted in significant exposure misclassification. Random error in 24 exposure measurements generally attenuates the strength of epidemiologic associations 25 between exposure and observed effect, weakening the predictive ability of any of the 26 exposure-based metrics compared to duration of employment, which was more accurately 27 determined for all workers in the cohort. 28 2) Even where the job category was identified, few exposure data exist prior to 1968. For 29 the majority ofjob locations (23 of 25), no exposure measurements were available prior 30 to 1967, and so exposures were estimated based on employee interviews (conducted in 31 1982) to determine what was known about major changes in operations between 1935 32 and 1967. For two job locations, dust-to-PCM extrapolations are based on measurements 33 taken in the late 1960s, so extrapolating from the mid-1960s to the early 1960s is likely to 34 be more certain than extrapolating further back in time. Random error in these exposure 35 measurements would also generally attenuate the strength of association between 36 exposure and observed effect during the earlier years of mine operation and, thus, a 37 greater degree of measurement error in the earlier years could have resulted in the lack in 38 proportionality of the hazard ratios for lung cancer over time. A greater degree of 39 measurement error in the earlier years could also provide an explanation for the worse fit 40 of the mesothelioma models that incorporated these exposure measures. This document is a draftfor review purposes only and does not constitute Agency policy. 5-81 DRAFT--DO NOT CITE OR QUOTE 1 3) Another explanation for the lack of proportional hazards in modeling lung-cancer 2 mortality may be that this cohort has an anomalous age structure due to the hiring of 3 much older individuals during the time of the Second World War. Among those workers 4 in the cohort hired prior to 1960, 9% were older than 50 years at the time of hire, and 5 22% were older than 40 years. Among those workers hired in 1960 or afterwards, only 6 4% were older than 50 years, and 14% were older than 40 years. Older workers differ 7 from younger workers in several potentially important ways that could alter their 8 response to exposures. Older workers were born in a different era, with different 9 nutritional and public health standards which may influence mortality patterns. 10 4) The lack of proportional hazards in modeling lung-cancer mortality may also be a 11 reflection of confounding or effect modification, which can change in magnitude over 12 time. The most likely candidate for confounding or effect modification is smoking. 13 NIOSH records show that of the 1,871 workers in the full Libby workers cohort, 14 1,121 workers (60%) were missing smoking status data, while 750 (40%) had data with 15 values "S" (Smoker), "Q" (Former Smoker), or "N" (Nonsmoker). Given this high 16 percentage of missing values, EPA did not consider these smoking data to be adequate 17 for use in the evaluation of confounding or effect modification. 18 5) Smoking rates, over time, among the sub-cohort of workers hired after 1959 are likely to 19 have been more similar since smoking rates change more slowly over shorter periods of 20 time than over longer ones. This restriction in time period of hiring would also result in 21 less variation by birth year cohort, which is strongly related to smoking patterns as people 22 of different generations developed different smoking rates. Thus, this restriction in the 23 time period of hiring may make the cohort members more similar to each other, thereby 24 reducing the potential impact of any smoking-related confounding. Further discussion of 25 the relevance of smoking can be found in the section on uncertainties (see Section 5.4.6). 26 27 28 When the assumption of proportionality is not met, the potential influence of 29 confounding factors in the full-cohort analysis is of concern. Additionally, the lack ofjob 30 category information for 69% of the workers hired prior to 1960 and greater measurement error 31 in early exposures may result in significant random exposure measurement error, which may bias 32 the observed exposure-response relationships towards the null. 33 Although duration of employment was the best exposure metric for modeling 34 mesothelioma mortality in the full cohort, it made quantitatively estimating an exposure-response 35 relationship difficult. In addition, violation of the underlying statistical assumptions adversely 36 impacted modeling of lung-cancer mortality in the full cohort. Therefore, EPA chose to 37 undertake a sub-cohort analysis. This document is a draftfor review purposes only and does not constitute Agency policy. 5-82 DRAFT--DO NOT CITE OR QUOTE 1 In particular, because uncertainty in retrospective assessment of workplace exposures is 2 reduced in the later years, EPA decided to analyze a sub-cohort of all the workers with as late a 3 starting employment date as possible, while still maintaining a sufficient number of lung cancer 4 and, especially, mesothelioma mortalities. Nearly all of the workers with completely missing 5 data on job category or department code and only duration of employment available were hired 6 before 1960, and so EPA developed a sub-cohort analysis by dividing the total cohort into those 7 hired prior to 1960 (n = 991) and those hired after 12/31/1959 (n = 880). This cut point roughly 8 divided the cohort in half. For the sub-cohort of those workers hired after 1959, there were 9 sufficient numbers of both mesothelioma and lung cancer mortalities to apply the Poisson and 10 Cox proportional hazards model, correspondingly. EPA initially examined the fit of these 11 models using several exposure metrics to predict mortality from mesothelioma and found that in 12 this sub-cohort, the exposure metrics that included information on exposure concentration 13 provided superior statistical fits to the exposure metrics based only on employment duration. In 14 this same sub-cohort, the assumptions of the Cox proportional hazards model were also satisfied 15 for the modeling of time-varying exposure. 16 While it is generally true that the use of more data is an advantage in statistical analyses 17 because it allows for the computation of more statistically precise effect estimates, this advantage 18 could not be utilized, because of the difficulty in deriving risks from the full cohort analysis (see 19 also Section 5.4.6 on uncertainties remaining in the sub-cohort). 20 21 5.4.3.6. Analysis of Sub-Cohort of Employees Hired After 1959 22 The reasons stated in Section 5.4.2 for choice of Libby worker cohort data are still valid 23 for the sub-cohort. In particular, (1) these workers were directly exposed to Libby Amphibole 24 asbestos, (2) detailed work histories and job-specific exposure estimates are available to 25 reconstruct estimates of each individual's occupational exposure experience with only 9 workers 26 completely missing job and department codes during the period when relatively high average 27 time-weighted estimated exposure intensity was assigned, (3) the sub-cohort is still sufficiently 28 large and has been followed for a sufficiently long period of time for cancer to develop (i.e., 29 cancer incidence) and result in mortality, and (4) the broad range of exposure experiences in the 30 sub-cohort provided an information-rich data set. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 5-83 DRAFT--DO NOT CITE OR QUOTE 1 5.4.3.6.I. Results o f analysis o f mesothelioma mortality in the sub-cohort 2 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The 3 number of mesothelioma deaths in the sub-cohort is 7 (2 deaths coded in ICD-10 and 5 deaths 4 coded in ICD-9), and the mesothelioma death rate of 24.7 per 100,000 person-years for the 5 sub-cohort is similar to the mesothelioma death rate of 26.8 per 100,000 person-years for the full 6 cohort (18 mesothelioma deaths), with a difference of less than 10%. 7 Table 5-10 shows the relative fit of various exposure metrics for mesothelioma mortality 8 in the sub-cohort hired after 1959, including only those exposure metrics whose information 9 weight was greater than 0.01. Information weights are computed from the DICs (Burnham and 10 Anderson. 2002). As discussed below, metrics with higher DICs and lower information weights 11 are unlikely to provide a good fit and are, thus, not included in Table 5-10. Information weights 12 are commonly used in Bayesian analyses. Information weights can be computed by first 13 assessing the differences between the best DIC and each of the others (A DIC.). 14 15 16 DIC w, =exp| - | A DIC.,yXexp( ] A l),C> (EcI 5~8) 17 18 19 The other exposure metrics that were fit included those metrics used in the full cohort 20 analysis [duration of employment, time since first exposure, age at death or censoring, RTW 21 metrics, CE with lag metrics, and IRIS IUR (1988a) metric], but all of them fit worse than any of 22 the metrics in Table 5-10, irrespective of possible penalization for extra parameters as discussed 23 in the analysis of the full cohort. The two metrics with cumulative exposure lagged 15 and 24 10 years, both with 5-year half life, provided the two best fits as indicated by their lower DIC 25 values and higher information weights (see Table 5-10). Cumulative exposures lagged 10 or 26 15 years, both with 10-year half life, provided the next two best fits accordi ng to DIC values, but 27 models including each of these metrics exhibited noticeably lower information weights than the 28 best metric. All metrics in Table 5-10 contain a decay term and have the same number of 29 parameters in their correspondi ng model, allowing for a direct comparison of the DIC values 30 (DICs are similar to AICs in what is considered an important difference) and information 31 weights. It is important to note that the suite of exposure metrics that were applied in this current 32 assessment to modeling mesothelioma mortality encompass the range of choices described in the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-84 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-10. Comparison of model fit of exposure metrics for mesothelioma 2 mortality in the sub-cohort hired after 1959a,b. Only the model fits with 3 information weights greater than 0.010 are shown 4 Exposure metric Lag(yr) DIC Information Weight CE with 5-year % life 15 70.6 0.428 CE with 5-year f | life 10 72.8 0.143 CE with 10-year J/ 2 life 10 73.9 0.082 CE with 10-year ks life 15 74.0 0.078 CE with 10-year Yi life 0 74.5 0.061 CE with 5-year f | life 0 75.0 0.047 CE with 15-year J/ 2 life 10 75.7 0.033 CE with 15-year ks life 0 76.0 0.029 CE with 15-year '/? life 15 76.1 0.028 CE with 20-year H life 10 76.7 0.020 CE with 20-year J/ 2 life 0 77.0 0.017 CE with 20-year life 15 77.2 0.016 5 6 aLower DIC values represent better fits. 7 bSince one o f mesothelioma deaths occurred in less than 20 years from start o f the exposure, lag 20 metrics assigned 8 no exposure to this case, and the very poor fit o f lag 20 metrics is a result. 9 10 DIC = Deviance Infonnation Criterion. 11 12 13 asbestos literature including CE, RTW, and decay metrics as well as the IRIS IUR (US. 14 EPA, 1988a) metric.In the sub-cohort hired after 1959, the DIC value for mesothelioma using the 15 IRIS IUR (U.S. EPA, 1988a) metric (see Eq. 5 -5 ) is substantially higher (DIC = 9 8 .4 ) than for 16 any of the metrics in Table 5-10. This indicates that the IRIS IUR (U.S. EPA, 1988a) metric 17 does not provide as good a fit for the Libby Amphibole asbestos worker cohort, using the 18 estimated historical exposure levels, as the other metrics in Table 5-10. Setting the exponents in 19 the IRIS IUR (U.S. EPA, 1988a) metric to the values of 2 and 4, as suggested by EPA (1986a). 20 did not improve the fit of the metric to the Libby Amphibole asbestos worker cohort data (results 21 not shown). A substantial difference of this analysis from the IRIS IUR (U.S. EPA, 1988a) 22 modeling is that this analysis is based on individual-level data, whereas the IRIS IUR (U.S. EPA, 23 1988a) application was to aggregate data. Also, cohorts used in the IRIS IUR (U.S. EPA, 1988a) 24 did not include cohorts exposed to Libby Amphibole asbestos. Alternately, the relative fit of this 25 model may have been affected by uncertainties in the estimated exposure described in detail in 26 Section 5.4.6. This document is a draftfor review purposes only and does not constitute Agency policy. 5-85 DRAFT--DO NOT CITE OR QUOTE 1 Next, EPA considered which covariates should be added to the model with the exposure 2 metric that provided the best fit. The addition of covariates" age at death or censoring" and 3 "time since first exposure" did not improve the fit, as measured by DIC (results not shown). 4 As described in Section 5.4.2.5, only metrics with nonzero lag were retained for 5 derivation of unit risks. Table 5-11 shows slopes and credible intervals for all retained metrics 6 from Table 5-10. The units of the slopes are fiber/cc-year. These slopes and credible intervals 7 represent calendar year continuous environmental exposure as described above and define the 8 9 10 Table 5-11. Mesothelioma mortality exposure metrics fits, slopes, and 11 credible intervals 12 Exposure metric Lag years DIC Slope x 10"5 90% C l fo r slope x 10"5 CE - 5-yr Vi life 15 70.6 20.6 (10.2, 34.3) CE - 5-yr Vi life 10 72.8 31.1 (15.2, 50.8) CE - 10-yr '/> life 10 73.9 9.93 (5.00, 16.3) CE - 10-yr '/> life 15 74.0 7.78 (3.72, 12.9) CE - 15-yr '/> life 10 75.7 6.17 (3.04, 10.1) CE - 15-yr '/> life 15 76.1 5.30 (2.63, 8.69) CE - 20-yr '/>life 10 76.7 4.71 (2.34, 7.71) CE - 20-yr '/>life 15 77.2 4.27 (2.12,6.98) 13 14 Cl = credible interval. 15 16 17 "Exposed Hazard Rate" in the life-table procedure when multiplied by the exposure level (see 18 Appendix G for details). 19 Based on the results from the exposure metric with the lowest DIC (cumulative exposure 20 with a 5-year half life for decay and a 15-year lag for cancer mortality latency), the slope was 21 2.06 x ICE4 per fiber/cc-year based on a 365-day calendar year, and the 95% upper bound on the 22 slope was 3.43 x 10 4per fiber /cc-year. This point estimate and 95% upper bound represent the 23 relative risk (including statistical uncertainty within the exposure metric) of mesothelioma 24 mortality observed from exposure to Libby Amphibole asbestos fibers in the worker cohort for 25 this exposure metric. Issues related to uncertainty in the choice of exposure metric are described 26 further in the section on the derivation of the combined IUR of mesothelioma and lung cancer 27 (see Section 5.4.5.3). This document is a draftfor review purposes only and does not constitute Agency policy. 5-86 DRAFT--DO NOT CITE OR QUOTE 1 2 5.4.3.6.2. Results of the analysis of the lung-cancer mortality in the sub-cohort 3 EPA based its final analyses for lung-cancer mortality on the subset of workers hired 4 after 1959. Thus, this analysis is based on 32 deaths from lung cancer40(ICD-8: two deaths with 5 the code 162.1; ICD-9: one death with the code 162.2, 20 deaths with the code 162.9; ICD-10: 6 nine deaths with the code C349) out of 230 deaths that occurred in the sub-cohort of 880 7 workers. 8 All multivariate Cox proportional hazards models with time-varying exposures were 9 initially fit, using one exposure metric at a time, to the sub-cohort hired after 1959 with 10 covariates for sex, race, and date of birth. Lung-cancer mortality was modeled using CE and 11 RTW exposure, where each metric was potentially modified by four different half-lives (5, 10, 12 15, or 20 years). Each of these exposure metrics was also evaluated with four different lag 13 periods to allow for cancer latencies of 0, 10, 15, or 20 years. The lag period is defined as 14 immediately prior to observed cancer death, where exposure is not considered to be causally 15 related to mortality. In all, 40 exposure response multivariate models were evaluated for the 16 adequacy of the exposure metric to fit the epidemiologic data. Each exposure metric and the 17 comparative model fit statistics are presented in Table 5-12. 18 The assumptions of the Cox proportional hazards model were reevaluated for the 19 sub-cohort. Restricting the cohort addressed each of the previously listed potential explanations 20 for the lack of hazard proportionality (see Section 5.4.3.3). First, measurement error for 21 exposures is likely to have been smaller after 1959 for several reasons. One reason is that the 22 706 workers for whom job category and department code information was missing during all of 23 their employment prior to 1960 were removed from the analysis. Also, beginning in 1968, fiber 24 concentrations by PCM analysis of site-specific air samples were available for all location 25 40Note that in the full cohort, it was unclear whether there were cases o f tracheal cancer included in the definition of lung cancer as many o f the recorded ICD codes on death certificates did not provide sufficient detail to distinguish tracheal cancer cases from lung cancer cases. However, among the sub-cohort o f workers hired after 1959, all the deaths from the broader category o f cancers o f the lung, bronchus, and trachea did provide sufficient detail to show that there were no deaths from tracheal cancer. This document is a draftfor review purposes only and does not constitute Agency policy. 5-87 DRAFT--DO NOT CITE OR QUOTE 1 2 Table 5-12. Model fit comparison for different exposure metrics and 3 lung-cancer mortality associated with Libby Amphibole asbestos, controlling 4 for age, gender, race, and date of birth 5 Ordered by exposure metric Ordered by model fit Lag Exposure metric (yr) Lag AIC Exposure metric (y r) Multivariate Exposure AIC model /7-value /7-value CE 0 361.610 CE 10-yr Vi life 10 358.400 0.0071 0.0009 CE 10 361.073 CE 5-yr Vi life 10 358.502 0.0075 0.0010 CE 15 363.124 CE 15-yr Vi life 10 358.777 0.0084 0.0015 CE 20 364.964 CE 20-yr Vi life 10 359.122 0.0098 0.0022 CE 20-yr '/>life 0 361.123 CE 5-yr Vi life 15 359.910 0.0138 0.0032 CE 20-yr '/>life 10 359.122 CE 10-yr Vi life 15 360.543 0.0181 0.0079 CE 20-yr '/>life 15 361.533 CE 10 361.073 0.0227 0.0188 CE 20-yr '/>life 20 364.703 CE 20-yr Vi life 0 361.123 0.0232 0.0155 CE 15-yr 'A life 0 361.382 CE 15-yr Vi life 15 361.129 0.0232 0.0162 CE 15-yr Vi life 10 358.777 CE 15-yr Vi life 0 361.382 0.0258 0.0184 CE 15-yr Vi life 15 361.129 CE 20-yr Vi life 15 361.533 0.0276 0.0254 CE 15-yr Vi life 20 364.588 RTW 5-yr Vi life 0 361.593 0.0283 0.0309 CE 10-yr '/> life 0 362.169 CE 0 361.610 0.0285 0.0307 CE 10-yr '/> life 10 358.400 CE 10-yr Vi life 0 362.169 0.0360 0.0358 CE 10-yr '/> life 15 360.543 RTW 10-yr Vi life 0 362.283 0.0378 0.0588 CE 10-yr '/> life 20 364.342 RTW 15-yr Vi life 0 362.714 0.0452 0.0863 CE 5-yr Vi life 0 364.225 RTW 20-yr Vi life 0 362.973 0.0503 0.1084 CE 5-yr Vi life 10 358.502 CE 15 363.124 0.0535 0.1215 CE 5-yr Vi life 15 359.910 RTW 5-yr Vi life 10 363.224 0.0558 0.1343 CE 5-yr Vi life 20 363.644 CE 5-yr Vi life 20 363.644 0.0662 0.1751 RTW 0 363.869 RTW 0 363.869 0.0726 0.2397 RTW 10 364.835 RTW 10-yr Vi life 10 364.041 0.0778 0.2810 RTW 15 364.990 CE 5-yr Vi life 0 364.225 0.0838 0.2908 RTW 20 364.502 RTW 15-yr Vi life 10 364.336 0.0876 0.3733 RTW 20-yr Vi life 0 362.973 CE 10-yr Vi life 20 364.342 0.0878 0.3661 RTW 20-yr Vi life 10 364.477 RTW 20-yr Vi life 10 364.477 0.0927 0.4314 RTW 20-yr Vi life 15 365.011 RTW 20 364.502 0.0936 0.5307 RTW 20-yr Vi life 20 364.628 CE 15-yr Vi life 20 364.588 0.0969 0.4815 RTW 15-yr Vi life 0 362.714 RTW 20-yr Vi life 20 364.628 0.0985 0.5763 RTW 15-yr Vi life 10 364.336 RTW 15-yr Vi life 20 364.662 0.0998 0.5909 RTW 15-yr Vi life 15 365.001 CE 20-yr Vi life 20 364.703 0.1014 0.5530 RTW 15-yr Vi life 20 364.662 RTW 10-yr Vi life 20 364.719 0.1021 0.6188 RTW 10-yr Vi life 0 362.283 RTW 5-yr Vi life 15 364.768 0.1041 0.6021 RTW 10-yr Vi life 10 6 364.041 RTW 5-yr Vi life 20 364.831 0.1067 0.6884 This document is a draftfor review purposes only and does not constitute Agency policy. 5-88 DRAFT--DO NOT CITE OR QUOTE Table 5-12. Model fit comparison for different exposure metrics and lung-cancer mortality associated with Libby Ainphibole asbestos, controlling for age, gender, race, and date of birth (continued) Ordered by exposure metric Ordered by model fit Lag Exposure metric (yr) Lag AIC Exposure metric (yr) Multivariate Exposure AIC model /7-value /7-value RTW 10-yr % life 15 364.962 RTW 10 364.835 0.1069 0.6586 RTW 10-yr A life 20 364.719 RTW 10-yr & life 15 364.962 0.1124 0.8173 RTW 5-yr U life 0 361.593 CE 20 364.964 0.1125 0.8204 RTW 5-yr U. life 10 363.224 RTW 15 364.990 0.1136 0.8809 RTW 5-yr M life 15 364.768 RTW 15-yr hk life 15 365.001 0.1141 0.9100 RTW 5-yr M life 20 364.831 RTW 20-yr i life 15 365.011 0.1146 1 2 CE: Cumulative exposure with or without exponential decay modeled with different half-lives. 3 RTW: Residence-time weighted exposure with or without exponential decay with different half-lives. 4 AIC: Akaike Information Criterion. 0.9599 5 6 7 operations to inform the JEM. Prior to 1968, the exposure intensity for 23 of 25 location 8 operations was estimated based on reasoned assumptions informed by employee interviews in 9 the early 1980s. It is likely the uncertainty of these reasoned assumpti ons increased the farther 10 back in time that exposures were estimated, making the earliest exposure estimates (1940s and 11 1950) less certain than those only a few years before fiber count data were available. Finally, 12 between 1956 and 1967, dust-to-PCM extrapolation data were used to estimate exposures in the 13 dry mill based on measurements taken in the late 1960s. Although there is some uncertainty in 14 the conversion ratio selected by Amandus et al. (1987a). dust-to-fiber conversions are likely to 15 be less uncertain than extrapolations further backwards in time to the 1950s and 1940s, where 16 only one air sample for dust was available in 1944. Thus, the potential attenuation effect of 17 nondifferential measurement error is likely to be reduced by examining the post-1959 cohort 18 alone compared to the entire cohort. 19 In addition, by focusing on the more homogeneous age distribution of workers hired after 20 1959, concerns about differential cancer mortality latency were diminished. Third, smoking 21 rates among this more narrowly defined sub-cohort are likely to have been more homogeneous, 22 and, thus, restricting analysis to this sub-cohort would help to limit any potential confounding 23 due to smoking. Finally, EPA conducted goodness-of-fit testing of the extended Cox 24 proportional hazards model as applied to the sub-cohort hired post-1959. There was no evidence This document is a draftfor review purposes only and does not constitute Agency policy. 5-89 DRAFT--DO NOT CITE OR QUOTE 1 to reject the hypothesis of proportionality, and the exposure models demonstrated adequate fits to 2 the data, with statistically significant effect estimates. In each of the Cox proportional hazards 3 model analyses with time-varying exposures--across all the exposure metrics and across all the 4 lag lengths--no violations of the assumption of proportionality of hazards were found. 5 As the exposure-response models cannot strictly be considered to be nested, a standard 6 measure of fit called the Akaike Information Criterion [AIC; Burnham and Anderson (2002)] 7 was used for comparison of goodness of fit across models based on the same data set. In their 8 text on model selection, Claeskens and Hjort (2008) state that ".. .for selecting a model among a 9 list of candidates, Akaike's information criterion (AIC) is among the most popular and versatile 10 strategies." Claeskens and Hjort (2008) also state that the model yielding the smallest AIC is 11 judged the best one and it is a common practice in environmental epidemiology to simply select 12 the single model with the best statistical fit (i.e., the lowest AIC) among the models that were 13 evaluated. Smaller AIC values generally indicate a better fitting model relative to larger AIC 14 values. While large differences in AIC values can reveal important differences in model fit, 15 small differences are less conclusive. For example, models differing in AIC by 2 or less units 16 can be considered to have a substantial level of empirical support [Burnham and Anderson 17 (2002); p. 70], 18 Table 5-12 shows the models and exposure metrics ordered by fit. Of interest is whether 19 there are models with distinct exposure metrics that adequately fit these data (as measured by 20 statistical significance of the model p-value) and then, a measure of relative fit among these 21 adequately fitti ng models. Of the 40 exposure-response metri cs, 14 demonstrated an adequate fit 22 to the data as measured by the overall model fit, with the likelihood ratio test being statistically 23 significant (p < 0.05), as well as having statistically significant exposure metrics (p < 0.05). 24 However, note that only the nine models that demonstrated adequate model and exposure metric 25 fit and incorporated a lag period to account for lung-cancer mortality latency were advanced for 26 potential use in developing a unit risk. While metrics that did not include an adjustment for lag 27 on the exposure metric to account for cancer mortality latency were fit to these data for the sake 28 of completeness, they were dropped from further consideration because they implicitly assume 29 no passage of time between the initiation of cancer, subsequent promotion of that cancer, and 30 mortality. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-90 DRAFT--DO NOT CITE OR QUOTE 1 Several general patterns were discemable with respect to which exposure metrics best 2 predicted lung-cancer mortality when comparing AICs for relative model fit. The data show that 3 lagging exposure by 10 years best predicts lung-cancer mortality compared to other lags. This 4 trend is seen across both the cumulative exposure without decay and the various half-life 5 cumulative exposure metrics where a 10-year lag of exposure best predicts lung-cancer mortality 6 for all cumulative exposure metrics compared to other lags; metrics with 15-year lags were 7 generally the next best in terms of fit. Another conclusion is that the models that included RTW 8 exposure metrics, regardless of half-life or lag, were less suitable than the models that employed 9 cumulative exposure and its variants. 10 Among the 40 exposure metric models that were evaluated, the exposure model with the 11 lowest AIC value was for cumulative exposure with a 10-year half life for decay and a 10-year 12 lag for cancer mortality latency and had a model /i-value of 0.0071 (see Table 5-12). This 13 multivariate model controlled for age, gender, race, and date of birth. This model estimated a 14 slope (beta) of 1.26 x 10 2 per fiber/cc-year based on a 365-day calendar year,41and the 15 95thpercentile upper bound on this parameter was 1.88 x 10 2 per fiber/cc-year. The /7-value for 16 the Libby Ampbibole asbestos regression coefficient (slope) was <0.001, indicating that this 17 parameter was statistically significantly greater than zero. Table 5-13 shows the slopes and 18 confidence intervals for all retained metrics from Table 5-12. 19 According to the model results presented in Table 5-12, there were other exposure 20 metrics that predicted lung-cancer mortality and exhibited statistically significant effect 21 estimates. Several other metrics were considered to fit nearly as well as the model with the 22 smallest AIC since their AIC values were within two units of the exposure model with the lowest 23 AIC, a proximity that can be considered to be a range that cannot clearly differentiate between 24 models (Burnham and Anderson. 2002). As each of the other exposure metrics was based on a 25 different reorganization of the same exposure data, the different slopes are not directly 26 comparable, but all adequately fitting lagged models also produce statistically significant slopes 27 for the exposure-response relationship (p < 0.05). Of particular note are the results of the 28 cumulative exposure model, with a 10-year lag for latency, but without a decay function, since it 29 showed the lowest AIC among nondecay models. 41The two-sided 90% confidence interval is (6.00 H 10"3, 1.88 I 10'"): the two-sided 95% confidence interval is (5.12 x Ilf. 2.00 10 A. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-91 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-13. Lung-cancer mortality exposure metrics fits, slopes, and 2 confidence intervals for all retained metrics from Table 5-12. Subset of lung 3 cancer models with lagged exposures that yielded statistically significant 4 model fit (p < 0.05) and exposure metric fit (p < 0.05) to the epidemiologic 5 data 6 Lag Exposure metric years AIC Slope (Beta) Exposure SE /7-value 90% C l for the slope CE 10-yr `/ 2 life 10 358.400 0.0126 0.0038 0.0009 (0.0063, 0.0188) CE 5-yr 'A life 10 358.502 0.0179 0.0055 0.0010 (0.0089, 0.0269) CE 15-yr Vi life 10 358.777 0.0106 0.0033 0.0015 (0.0052, 0.0160) CE 20-yr '/> life 10 359.122 0.0095 0.0031 0.0022 (0.0044, 0.0146) CE 5-yr Vi life 15 359.910 0.0155 0.0052 0.0032 (0.0069,0.0241) CE 10-yr '/> life 15 360.543 0.0115 0.0043 0.0079 (0.0044, 0.0186) CE 10 361.073 0.0058 0.0025 0.0188 (0.0017,0.0099) CE 15-yr Vi life 15 361.129 0.0097 0.0040 0.0162 (0.0031,0.0163) CE 20-yr '/> life 15 361.533 0.0087 0.0039 0.0254 (0.0023,0.0151) 7 8 Cl = confidence interval 9 10 11 The AIC values for models that included lag and/or half-life adjustments to the exposure 12 metrics were not penalized in the regression analyses for using these extra parameters because 13 these factors were not represented as covariates but rather were embedded in the computation. 14 While these results were obtained using each instance of lag and/or half-life terms in separate 15 model fit, it may be appropriate to mathematically penalize the AICs for inclusion of these 16 additional parameters. AIC values, as typically computed by regression software, include the 17 addition of a penalty for model complexity as measured by the number of parameters that are fit 18 in the regression model (thereby increasing the AIC). In the AIC calculations presented in 19 Table 5-12, the models are treated as having the same number of parameters since each model 20 represents the same exposures in a different way but with a single exposure parameter in the 21 regression models and are, therefore, equally penalized in the software's AIC calculation. 22 Because an argument can be made that exposure metrics that do not include a decay function 23 with their half-life term are implicitly more parsimonious (simpler), a comparison of the AICs is 24 not straightforward. If the decay model fits were penalized for the inclusion of the decay 25 function in the computation of the exposure metric, then with such an adjustment, the relative fit This document is a draftfor review purposes only and does not constitute Agency policy. 5-92 DRAFT--DO NOT CITE OR QUOTE 1 of the CE models would be somewhat improved in terms of their comparison with the values in 2 Table 5-12 (AICs are generally penalized 2 units for each additional parameter). 3 Table 5-13 displays the lagged exposure-response models and metrics with adequate 4 model fit (p < 0.05) to the epidemiologic data that were further considered. The units of the 5 slopes are fiber/cc-year. These slopes and confidence intervals represent calendar year 6 continuous environmental exposure as described above and define the "Exposed Hazard Rate" in 7 the life-table procedure when multiplied by the exposure level (see Appendix G for details). 8 9 5.4.3.6.3. Summary of results of the analysis of the lung-cancer mortality in the sub-cohort 10 As presented in Table 5-13, the CE model with 10-year half life and lag provided an 11 adequate fit to the data (p < 0.05) and had the lowest AIC value. The cumulative exposure 12 model with a 10-year lag also yielded a statistically adequate fit to these data (p < 0.05), as did 13 several decay models with a 15-year lag. These results demonstrate reasonable uncertainty in the 14 metric of exposure such that no single exposure model can be definitively selected based on 15 goodness of fit alone, because IUR is based on the plausible upper bound of the effect estimate. 16 Based on the results from the lowest AIC multivariate model (i.e., cumulative exposure with a 17 10-year half life for decay and a 10-year lag for cancer mortality latency), the slope was 18 1.26 x 10_2 per fiber/cc-year based on a 365-day calendar year, and the 95% upper bound on the 19 slope was 1.88 x 10_2 per fiber/cc-year. This point estimate and 95% upper bound represent the 20 relative risk (including statistical uncertainty within exposure metric) of lung-cancer mortality 21 observed from exposure to Libby Amphibole asbestos fibers in the worker cohort for this 22 exposure metric. Issues related to uncertainty in the choice of exposure metric are described 23 further in the section on the derivation of the combined IUR of mesothelioma and lung cancer 24 (see Section 5.4.5.3). 25 26 5.4.3.6.4. Sensitivity analysis of the influence o f high exposures in early 1960s on the model 27 fit in the sub-cohort 28 As discussed in Section 5.4.2.5, the comparison of model fit between various exposure 29 metrics is an empirical process and does not necessarily reflect either a specific biological or 30 other factor as an underlying cause for model fit. Although data do not exist to evaluate 31 biological bases for model fit, other potential factors can be explored where data allow. For This document is a draftfor review purposes only and does not constitute Agency policy. 5-93 DRAFT--DO NOT CITE OR QUOTE 1 example, because of concerns that very high (>100 fibers/cc) 8-hour TWA exposures during 2 1960-1963 (see Table 5-7) could have influenced the relative fit of the various exposure metrics, 3 EPA conducted a sensitivity analysis of the impact on the relative model fit of reducing all 4 estimated exposure intensities for 1960-1963 by 50%. 5 For modeling mesothelioma mortality on this revised data set, there was one change in 6 the relative fit of 3ld and 4th best fit decay models, but the observation that exposure metrics 7 including decay fit better than exposure metrics without decay was unchanged (see Table 5-14). 8 Flowever, the fit of all the metrics decreased slightly, with each DIC increased between 0.3 and 9 1.1. The metrics without decay and RTW metrics had DIC values higher than those in 10 Table 5-14. The revised data set DIC for the model used in IRIS IUR (U.S. EPA. 1988a) was 11 97.9. 12 13 14 Table 5-14. Sensitivity analysis of model fit comparison for different 15 exposure metrics and lung-cancer mortality associated with Libby 16 Amphibole asbestos. Estimated exposure intensities for all jobs during 17 1960-1963 were reduced by 50%. 18245 Lag Exposure Metric (yr) All workers hired after 1959 (n = 880) Based on seven mesothelioma deaths (as shown in Table 5-11) DIC All workers hired after 1959 (n = 880) Based on seven mesothelioma deaths Exposures during 1960-1963 at 50% DIC CE 5-yr Vi life 15 70.6 71.2 CE 5-yr %life 10 72.8 73.9 CE 10-yr > life 10 73.9 74.9 CE 10-yr &S life 15 74 74.6 CE 15-yr 'A life 10 75.7 76.4 CE 15-yr > life 15 76.1 76.7 CE 20-yr > life 10 76.7 77.3 CE 20-yr &S life 15 77.2 77.7 19 20 CE = Cumulative Exposure with exponential decay modeled with different half-lives; DIC = Deviance Information Criterion. 21 22 23 For modeling lung-cancer mortality on this revised data set, there was no difference in 24 the order of the relative fit between the same exposure models that fit the sub-cohort of workers 25 hired after 1959 and included the exposures as estimated by Amandus et al. (1987a) during This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-94 DRAFT--DO NOT CITE OR QUOTE 1 1960-1963 (see Table 5-15). The models based on the revised data set fit marginally better 2 based on AIC. 3 4 5 Table 5-15. Sensitivity analysis of model fit comparison for different 6 exposure metrics and lung-cancer mortality associated with Libby 7 Amphibole asbestos, controlling for age, gender, race, and date of birth. 8 Estimated exposure intensities for all jobs during 1960-1963 were reduced by 9 50%. Lung cancer models presented include those with statistically significant 10 multivariate model p-value and nonzero lag in exposure. 111234567890 A ll w orkers hired after 1959 (n = 880) based on 32 deaths from lung cancer (as shown in Table 5-13) A ll w orkers hired after 1959 (n = 880) based on 32 deaths from lung cancer exposures during 1960-1963 at 50% Exposure m etric CE 10-yr `A life CE 5-yr A life CE 15-yr A life CE 20-yr A life Lag (yr) AIC 10 358.400 10 358.502 10 358.777 10 359.122 M ultivariate model /j-value 0.0071 0.0075 0.0084 0.0098 Exposure /7-value 0.0009 0.0010 0.0015 0.0022 AIC 357.644 357.781 357.966 358.283 M ultivariate model /7-value 0.0051 0.0054 0.0059 0.0068 Exposure /7-value 0.0004 0.0005 0.0006 0.0009 CE 5-yr A life 15 359.910 0.0138 0.0032 359.456 0.0113 0.0025 CE 10-yr A life CE 15 360.543 10 361.073 0.0181 0.0227 0.0079 0.0188 360.167 360.238 0.0154 0.0159 0.0067 0.0086 CE 15-yr A life 15 361.129 0.0232 0.0162 360.810 0.0203 0.0138 CE 20-yr A life 15 361.533 0.0276 0.0254 361.245 0.0244 0.0217 12 13 CE = Cumulative Exposure with or without exponential decay modeled with different half-lives. 14 AIC = Akaike Information Criterion. 15 16 17 This sensitivity analysis reduces some of the potential uncertainty in the results that may 18 have been attributed to exposure measurement error specific to the 1960-1963 time period when 19 some of the estimated exposures were particularly high. 20 21 5.4.3.6.5. Additional analysis of the potentialfor confounding of lung cancer results by 22 smoking in the sub-cohort of workers hired after 1959 23 In the full cohort analysis, the proportional hazard assumption was not found to hold, and 24 it was possible that one of the reasons for this failure was the presence of confounding by 25 smoking, which altered the proportionality of the hazard rate in the exposed workers compared 26 to the baseline hazard rate over time. By restricting the dates of hire in the sub-cohort, those This document is a draftfor review purposes only and does not constitute Agency policy. 5-95 DRAFT--DO NOT CITE OR QUOTE 1 workers in the sub-cohort may be made more similar to each other in ways that would reduce the 2 potential for confounding by smoking and, in this sub-cohort, the proportional hazards 3 assumption was found to hold, thus statistically eliminating concern regarding confounding by 4 smoking (because smoking, in general, is known as a very strong confounder). 5 6 As an additional check on the potential for confounding, a new method was evaluated to 7 test for confounding by smoking in occupational cohorts that do not have data on smoking. 8 Confounding, which can bias observed results when there is an uncontrolled variable, which is 9 correlated with both the explanatory variable and the outcome vari able, is a distinct concept from 10 effect-measure modification (i.e., synergy), which might reflect different observed effects of 11 exposure to Libby Amphibole asbestos among smokers as compared to nonsmokers. The extent 12 of effect-measure modification cannot be assessed without adequate data on smoking; however, 13 the issue is discussed in Section 5.4.6. 14 A method has been described by Richardson (2 0 1 0 ) to determine if an identified 15 exposure relationship with lung cancer is confounded by unmeasured smoking in an occupational 16 cohort study. Richardson (2 0 1 0 ) demonstrated that an exposure of interest (i.e., Libby 17 Amphibole asbestos) can be used to predict an outcome other than lung cancer such as chronic 18 obstructive pulmonary disease (COPD), which is known to be caused by smoking, but not 19 thought to be related to the exposure of concern.42 If a positive relati onship is identified where 20 no causal association is suspected, this would suggest that smoking and the exposure metric 21 (Libby Amphibole asbestos) were positively correlated and that the identified exposure-response 22 relationship was, in fact, confounded by smoking. EPA implemented this methodology to model 23 the potential effects of Libby Amphibole asbestos on the risk of COPD mortality on the 24 sub-cohort of workers hired after 1959. Using the exposure metric defined as cumulative 25 exposure with a 10-year lag, the extended Cox proportional hazards model with time-varying 26 exposures estimated a slope (beta) for COPD of -0.056 per fiber/cc-year based on a 365-day 27 calendar year. Thep-value for the coefficient (slope) was 0.102, indicating that this parameter 28 was not statistically significantly different from zero. Using the exposure metric defined as 29 cumulative exposure with a 10-year half life for decay and a 10-year lag for cancer latency, the 1Ricliardson (2010) cited articles by Rushton (2007a. b) with possible associations between asbestos and COPD w hich, if true, would have explained a positive association among the Libby workers cohort but should not detract from the use o f the Richardson method as applied to these Libby workers, where a negative association is found. This document is a draftfor review purposes only and does not constitute Agency policy. 5-96 DRAFT--DO NOT CITE OR QUOTE 1 extended Cox proportional hazards model with time-varying exposures estimated a slope (beta) 2 of -0.135 per fiber/cc-year based on a 365-day calendar year. Thep-value for the coefficient 3 (slope) was 0.116, indicating that this parameter was not statistically significantly different from 4 zero. 5 Summarizing these findings, EPA used the method described by Richardson (2010) to 6 evaluate whether exposures to Libby Amphibole asbestos predicted mortality from COPD as an 7 indication of potential confounding by smoking and found a nonsignificant negative relationship, 8 which was inconsistent with confounding by smoking in the sub-cohort of workers hired after 9 1959. 10 11 5.4.4. Exposure Adjustments and Extrapolation Methods 12 The estimated exposures based on JEM and work histories are discussed in 13 Section 5.4.2.5. Note that all slopes presented with units of fiber/cc-year are for calendar year 14 and not for occupational year. 15 16 5.4.5. Inhalation Unit Risk (IUR) of Cancer Mortality 17 The derivation of the unit risk estimates, defined as the lifetime risk of mortality from 18 either mesothelioma or lung cancer from chronic inhalation of Libby Amphibole asbestos at a 19 concentration of 1 fiber/cc of air, is presented in the following subsections. Note that all slopes 20 are presented as per fiber/cc-year for a 365-day calendar year rather than for an occupational 21 year. Also, note that while the slopes are not adjusted for differences in breathing rates and the 22 number of hours of exposure in an occupational (8-hour) day as compared to a whole (24-hour) 23 day, the central risk and unit risk estimates do incorporate this adjustment. 24 25 5.4.5.1. Unit Risk Estimatesfo r Mesothelioma Mortality 26 Computational details of the methodology and tables for deriving the unit risk for 27 mesothelioma mortality are presented in Appendix G. The modeling analysis presented above 28 showed that metrics including lag and half-life parameters provided the best empirical fit to the 29 Libby worker sub-cohort data. Although there is uncertainty in applying these models for 30 occupational mortality to estimation of risks for different exposure levels and time patterns (see 31 Section 5.4.6), following the recommendations of the Guidelinesfor Carcinogen Risk This document is a draftfor review purposes only and does not constitute Agency policy. 5-97 DRAFT--DO NOT CITE OR QUOTE 1 Assessment (U.S. EPA, 2005a), a linear low-dose extrapolation below the POD was used because 2 the mode of action for Libby Amphibole asbestos for mesothelioma is largely unknown. Using 3 the results of the cumulative exposure model with best-fitting lag and decay parameters, the 4 LECoi for the adult-only-exposures was determined to be 0.245 fibers/cc, which yielded an 5 adult-based unit risk of mesothelioma mortality of 0.041 (POD of 1% divided by the LECoi), 6 which when scaled by 70/54 to encompass the whole lifespan, yielded a lifetime unit risk of 7 0.053 per fibers/cc. The value of the risk corresponding to the measure of central tendency 8 involves ECoi rather than LECoi. The ECoi for the adult-only-exposures was determined to be 9 0.406 per fibers/cc, which when divided into a POD of 1%, yielded an adult-based central 10 estimate for mesothelioma mortality of 0.025, which when scaled by 70/54 to encompass the 11 whole lifespan, yielded a lifetime central estimate of 0.032 per fibers/cc. 12 The mesothelioma unit risks for model results presented in Table 5-11 and discussed in 13 Section 5.4.3.6.1 are presented in Table 5-16. All of the metrics in Table 5-16 are CE metrics 14 lagged 10-15 years (the fit of 20-year lag models was much worse since one of seven 15 mesothelioma deaths occurred before 20 years; lags longer than 15 years are possible, and this is 16 an uncertainty described in Section 5.4.6). Issues related to uncertainty in the choice of exposure 17 metric are described further in the section on the derivation of the combined IUR of 18 mesothelioma and lung cancer (see Section 5.4.5.3). 19 20 21 Table 5-16. Mesothelioma mortality exposure metrics unit risks 22 Exposure metric Lag years DIC Information weight Central risk estimate Unit risk CE - 5-yr II. life 15 70.6 0.428 0.032 0.053 CE - 5-yr S life 10 72.8 0.143 0.054 0.088 CE - 10-yr H life 10 73.9 0.082 0.028 0.047 CE - 10-yr '/2 life 15 74.0 0.078 0.020 0.032 CE - 15-yr 'A life 10 75.7 0.033 0.022 0.036 CE - 15-yr !/> life 15 76.1 0.028 0.017 0.027 CE - 20-yr H life 10 76.7 0.020 0.020 0.032 CE - 20-yr '/2 life 15 77.2 0.016 0.015 0.025 23 24 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-98 DRAFT--DO NOT CITE OR QUOTE 1 5.4.5.1.1. Adjustmentfo r mesothelioma underascertainment 2 For mesothelioma, the undercounting of cases (underascertainment) is a particular 3 concern given the limitations of the ICD classification systems used prior to 1999. In practical 4 terms, this means that some true occurrences of mortality due to mesothelioma are missed on 5 death certificates and in almost all administrative databases such as the National Death Index. 6 Even after the introduction of a special ICD code for mesothelioma with the introduction of 7 ICD-10 in 1999, detection rates are still imperfect (Camidge et al., 2006; Pinheiro et al.. 2004). 8 and the reported numbers of cases typically reflect an undercount of the true number. Kopylev et 9 al. (2011) reviewed the literature on this underascertainment and developed general methodology 10 to account for the likely numbers of undocumented mesothelioma deaths using the Libby worker 11 cohort as an example. Because the analysis of mesothelioma mortality was based on absolute 12 risk, it was possible to compensate for mesothelioma underascertainment in the Libby worker 13 sub-cohort. As the number of peritoneal mesotheliomas is partially known in the Libby worker 14 sub-cohort, the appropriate adjustment factor for the sub-cohort is 1.39 [Kopylev et al. (2011). 15 Table 3], 16 The adjusted mesothelioma central risk (based on the ECoi), corresponding to the best-fit 17 metric, was 0.044 (0.032 -x 1.39) per fibers/cc, and adjusted mesothelioma mortality unit risk was 18 0.074 (0.053 x 1.39) per fibers/cc. Mesothelioma mortality-adjusted unit risks are listed in 19 Table 5-17 along with their information weights. 20 21 22 Table 5-17. Adjusted for underascertainment unit risks for the sub-cohort 23 hired after 1959 corresponding to the different metrics 24 Exposure metric Lag years Information weight Adjusted central risk estimate Adjusted unit risk CE - 5-yr 14 life 15 0.428 0.044 0.074 CE - 5-yr 14 life 10 0.143 0.075 0.122 CE - 10-yr M life 10 0.082 0.039 0.065 CE - 10-yr 14 life 15 0.078 0.028 0.044 CE - 15-yr 14 life 10 0.033 0.031 0.050 CE - 15-yr 14 life 15 0.028 0.024 0.038 CE - 20-yr 14 life 10 0.020 0.028 0.044 CE - 20-yr 14 life 15 0.016 0.022 0.035 25 26 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-99 DRAFT--DO NOT CITE OR QUOTE 1 5.4.5.2. Unit Risk Estimatesfo r Lung-Cancer mortality 2 Computational details of the methodology and tables for deriving the unit risk for 3 lung-cancer mortality are presented in Appendix G. Although there is uncertainty in applying 4 these models for occupational mortality to the estimation of risks for different exposure levels 5 and time patterns (see Section 5.4.6), following the recommendations of the Guidelinesfor 6 Carcinogen Risk Assessment (U.S. EPA. 2005a). a linear low-dose extrapolation below the POD 7 was used because the mode of action for Libby Amphibole asbestos for lung cancer is 8 undetermined. The nine exposure-response models retained from Table 5-12 (shown in 9 Table 5-13) all had reasonably similar goodness of fits. No single model stands out as clearly 10 statistically superior; however, there is a range of quality of fit within the set that could be 11 considered adequate. The lung-cancer mortality unit risks are shown in Table 5-18. 12 Using the results of the exposure model with the lowest AIC value (i.e., cumulative exposure 13 with a 10-year half life for decay and a 10-year lag for cancer latency) alone, the LECoi for the 14 adult-only-exposures was determined to be 0.333 fibers/cc, which yielded an adult-based unit 15 risk of lung-cancer mortality of 0.0300 (POD of 1% divided by the LECoi), which when scal ed 16 by 70/54 to encompass the whole lifespan, yielded a lifetime unit risk of 0.0389 per fibers/cc. 17 The value of the risk that would correspond to the measure of central tendency involves ECoi 18 rather than LECoi. The ECoi for the adult-only exposures was determined to be 0.499 per 19 fibers/cc, which when divided into a POD of 1%, yielded an adult-based central estimate for 20 lung-cancer mortality of 0.0200, which when scaled by 70/54 to encompass the whole lifespan, 21 yielded a lifetime central estimate of 0.0260 per fibers/cc. 22 Using the results of the exposure model based on cumulative exposure with a 10-year lag 23 for cancer latency, the LECoi for the adult-only-exposures was determined to be 0 .191 fibers/cc, 24 which yielded an adult-based unit risk of lung-cancer mortality of 0.0524 (POD of 1% divided 25 by the LECoi), which when scaled by 70/54 to encompass the whole lifespan, yielded a lifetime 26 unit risk of 0.0679 per fibers/cc. The ECoi for the adult-only exposures was determined to be 27 0.325 per fibers/cc, which when divided into a POD of 1%, yielded an adult-based central 28 estimate for lung-cancer mortality of 0.0308, which when scaled by 70/54 to encompass the 29 whole lifespan, yielded a lifetime central estimate of 0.0399 per fibers/cc. 30 31 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-100 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-18. Unit risks for subset of lung cancer models with lagged 2 exposures that yielded statistically significant model fit (p < 0.05) and 3 exposure metric fit (p < 0.05) to the epidemiologic data 4 Exposure metric Lag AIC Exposure /7-value Central risk estimate (based on EC0i) Unit risk (based on LEC0i) CE 10-yr '/>life 10 358.400 0.0009 0.0260 0.0389 CE 5-yr Vi life 10 358.502 0.0010 0.0195 0.0293 CE 15-yr Vi life 10 358.777 0.0015 0.0300 0.0455 CE 20-yr '/>life 10 359.122 0.0022 0.0326 0.0501 CE 5-yr Vi life 15 359.910 0.0032 0.0167 0.0260 CE 10-yr '/>life 15 360.543 0.0079 0.0231 0.0375 CE 10 361.073 0.0188 0.0399 0.0679 CE 15-yr Vi life 15 361.129 0.0162 0.0258 0.0434 CE 20-yr '/>life 15 361.533 0.0254 0.0280 0.0486 5 6 7 The resulting unit risks in Table 5-18 ranged from 0.0260 to 0.0679 fibers/cc. This 8 shows that the unit risk (i.e., 0.0389 per fibers/cc) based on the exposure metric with the lowest 9 AIC value (i.e., cumulative exposure with a 10-year half life for decay and a 10-year lag for 10 cancer latency) is in the center of this range and is, thus, statistically robust. However, because 11 this estimate is in the middle of the range, it does not capture the uncertainty across metrics with 12 similar goodness of fit. As noted (see Section 5.4.3.6.2), an argument can be made that the CE 13 metric with a 10-year lag and no half-life is implicitly more parsimonious (simpler) because it 14 was not explicitly adjusted to include decay, although this metric is mathematically equivalent to 15 CE metric with a 10-year lag and an infinitely long decay half-life. Conceptually, the AIC 16 values are penalized for increased model complexity (thereby increasing the AIC). The AIC for 17 the CE models may reasonably be thought to be somewhat lower than through the standard 18 calculation of AIC. The CE metric with a 10-year lag does fit these data, is a simpler and more 19 straightforward metric, and has an extensive tradition of use in the epidemiologic literature and 20 in the practice of risk assessment. 21 Issues related to uncertainty in the choice of exposure metric are described in the section 22 on the derivation of the combined IUR of mesothelioma and lung cancer below. 23 This document is a draftfor review purposes only and does not constitute Agency policy. 5-101 DRAFT--DO NOT CITE OR QUOTE 1 5.4.5.3. IUR Derivationfor Combined Mesothelioma and Lung-Cancer Mortality 2 Before risks can be combined, it is important to understand several concepts that are 3 pertinent to the evaluation and comparison of the cancer-specific mortality unit risks that will be 4 combined. First, there is statistical uncertainty in the potency estimate within the 5 exposure-response model defined by each exposure metric. This within-metric uncertainty is 6 accounted for by the Bayesian credible interval around the potency estimates (slopes) for 7 mesothelioma mortality (see Table 5-11) and by the confidence interval around the potency 8 estimates (slopes) for lung-cancer mortality (see Table 5-13). Next, there is uncertainty in the 9 choice of metrics for developing an IUR (called cross-metric uncertainty, described below). 10 Finally, when unit risks corresponding to metrics are chosen accounting for uncertainty, these are 11 statistically combined into the IUR. Details are provided below. 12 For this current assessment, EPA obtained the best available demographic, exposure, and 13 vital status data from NIOSH. Subsequently, the best-fitting statistical models were identified, 14 which were then applied to derive central estimates of the lifetime combined mesothelioma and 15 lung-cancer mortality risk in the general population exposed to a continuous concentration of 16 1 fiber/cc of Libby Amphibole asbestos. Then, the individual exposure metric-specific risks 17 were calculated as the statistical (95%) upper confidence bounds on these central estimates. Use 18 of the upper confidence bound accounts for uncertainty in the effect estimate for each metric-- 19 otherwise referred to as the within-metric uncertainty. 20 Another source of uncertainty is the choice of the appropriate exposure metric among a 21 set of results that appear to fit the data similarly well. This uncertainty is referred to as the 22 between-metric or cross-metric uncertainty. For the Libby worker cohort data, the best-fit 23 (lowest information criterion values) metrics lead to estimates of risks that are more like 24 mid-range estimates among the other metrics (see Tables 5-17 and 5-18) with sufficiently close 25 information criterion values, rather than upper bound estimates. While the lung cancer unit risk 26 computed from the model with the lowest AIC appears to be robust, Table 5-18 shows that there 27 is a range of possible unit risk values from the set of models with adequate fit (as measured by a 28 statistically significant p-value for the exposure metric term) and similar goodness of fit. 29 Likewise, for mesothelioma mortality, among the models with adequate fit shown in Table 5-17, 30 there is a range of possible unit risk values. This document is a draftfor review purposes only and does not constitute Agency policy. 5-102 DRAFT--DO NOT CITE OR QUOTE 1 The IUR should be a reasonable upper bound on the extra risk. As is clear from 2 Tables 5-17 and 5-18 in the preceding sections, the unit risks based on the metrics with the 3 lowest information criterion values provide a lower estimate of cancer mortality risk than some 4 other similarly fitting metrics. While the models with the lowest information criterion values 5 have the greatest statistical support, other models that yield higher unit risks are also statistically 6 plausible. This current assessment selected the upper bound unit risk among the plausible 7 exposure metrics (regardless of the small residual differen ces in quality of fit) to account for 8 cross-metric uncertainty. Because there were few metrics with unit risks higher than the best 9 fitting metric's unit risk for each cancer mortality endpoint, this method effectively selects the 10 highest unit risk among those considered for each cancer mortality endpoint. 11 Once the cancer-specific mortality unit risks are selected, the two are then combined. 12 Because each of the unit risks is itself an upper bound estimate, summing such upper bound 13 estimates across mesothelioma and lung-cancer mortality is likely to overstate the overall risk. 14 Therefore, following the recommendations of the Guidelinesfor Carcinogen Risk Assessment 15 (U.S. EPA. 2005a). a statistically appropriate upper bound on combined risk was derived in order 16 to gain an understanding of the overall risk of mortality resulting from mesotheli oma and from 17 lung cancers. It is important to note that this estimate of overall potency describes the risk of 18 mortality from cancer at either of the considered sites and is not just the risk of both cancers 19 simultaneously. 20 Because the estimated risk for both mesothelioma and lung-cancer mortality was derived 21 using Poisson and Cox proportional hazards models, correspondingly, it follows from statistical 22 theory that each of these estimates of risk is approximately normally distributed. For 23 independent normal random variables, a standard deviation for a sum is easily derived from 24 individual standard deviations, which are estimated from confidence intervals: standard 25 deviation = (unit risk - central risk) + Z0.95, where Z0.95is a standard normal quantile equal 26 to 1.645. For normal random variables, the standard deviation of a sum is the square root of the 27 sum of the squares of individual standard deviations. 28 The upper bound among the mesothelioma mortality unit risks was 0.122 per fibers/cc. 29 The upper bound among the computed lung-cancer mortality unit risks was 0.0680 per fibers/cc 30 The central estimate of risk was 0.075 for mesothelioma mortality per fibers/cc and 0.0399 per 31 fibers/cc for lung-cancer mortality (see Tables 5-17 and 5-18, respectively). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-103 DRAFT--DO NOT CITE OR QUOTE 1 In order to combine the unit risks, one first obtains an estimate of standard deviation of 2 the sum of the individual unit risks as 3 4 5 V[ [[(0.122 - 0.075) + 1,645]2 + (0.068 - 0.0399) - 1.645 ]2 ] = 0.033 per fibers/cc (Eq. 5-9) 6 7 8 Then, the combined central estimate of risk of mortality from either mesothelioma or 9 lung cancer isO.0399 + 0.075 = 0.115 per fibers/cc, and the combined IUR is 10 0.115 +0.033 x 1.645 = 0.169 per fibers/cc. 11 Selecting the upper bound unit risk estimates for use in combining unit risks accounts for 12 many potential uncertainties. It accounts for uncertainty in the effect estimate (i.e., the 13 within-metric uncertainty) and the uncertainty attributable to the choice of exposure metric (i.e., 14 the cross-metric uncertainty). The combined IUR from the best fitting mesothelioma and 15 lung-cancer mortality models (using two different model selection criteria) can be computed for 16 comparison with Tables 5-17 and 5-18, respectively, by the same steps as above, and the results 17 are shown in Table 5-19. 18 19 20 Table 5-19. Reasonable upper bound and lowest information criteria 21 estimates of central risks and unit risks, per fibers/cc, for mesothelioma 22 mortality, lung-cancer mortality, and the IUR for the combined mortality 23 risk from mesothelioma and lung cancer 24 M esothelioma Lung cancer Combined mesothelioma and lung cancer M odel Central estimate Unit risk C en tral estim ate Unit risk C en tral estim ate IUR Reasonable upper bound3 0.075 0.122 0.040 0.068 0.115 0.169 Lowest information criteria13 0.044 0.074 0.026 0.040 0.070 0.103 25 26 T or mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay 27 half-life o f 5 years and a 15-year lag. For lung cancer, the selected model parameterized exposure as cumulative 28 exposure without decay and a 10-year lag. 29 bFor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay 30 half-life o f 5 years and a 10 -year lag. For lung cancer, the selected model parameterized exposure as cumulative 31 exposure with exponential decay half-life o f 10 years and a 10-year lag. 32 33 This document is a draftfor review purposes only and does not constitute Agency policy. 5-104 DRAFT--DO NOT CITE OR QUOTE 1 Compared to the combined IUR from the best fitting exposure models, the EPA's 2 selected combined 1UR of mesothelioma and lung-cancer mortality accounts for both the 3 demonstrated cross-metric uncertainty as well as several additional potential uncertainties, which 4 could have resulted in underestimates of the mesothelioma and lung-cancer mortality risks from 5 the epidemiologic data. These additional uncertainties are discussed in Section 5.4.6. The 11JR. 6 value of 0.169 per fibers/cc accounts for important quantitative uncertainties in the selection of 7 the specific exposure metric that may have remained in an IUR that might otherwise have been 8 based on the best fitting exposure models alone. 9 10 5.4.5.3.I. Comparison with other published studies o fLibby workers cohort 11 For lung cancer, two alternative analytic approaches to the use of EPA's extended Cox 12 proportional hazards models could have been used for the calculation of a unit risk of 13 lung-cancer mortality. All of the choices are based on different analyses of the Libby worker 14 cohort; however, inclusion criteria differ among the analyses as does the length of mortality 15 follow-up. Each of the two approaches has two options to estimate the slope of the 16 exposure-response relationship in place of the regression slope estimated from the Cox 17 proportional hazards model and follow through with the same life-table procedure to calculate 18 the unit risk of lung-cancer mortality. 19 The first approach would be to use the published categorical results based on Sullivan 20 (2007). The first option in this approach was for EPA to estimate a slope to those categorical 21 data. The second option was to use the slope estimated in a published reanalysis of categorical 22 data of the Sullivan (2007) cohort by Berman and Crump (2008). The second approach would 23 be to use the published regression results of other researchers who modeled the underlying 24 continuous data. The first option in this approach was to use the slope estimated by Larson et al. 25 (2010b). The second option was to use the slope estimated by Moolgavkar et al. (2010). 26 For comparison purposes, the lung cancer unit risk from these alternatives is computed, 27 however, as all analyses are based upon different subsets of the Libby workers cohort and used 28 different analytic methods, the results are not necessarily interchangeable. Table 5-20 29 summarizes lung cancer risks derived from these studies. 30 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-105 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-20. Lung cancer regression results from different analyses of 2 cumulative exposure in the cohort of workers in Libby, MT. All analyses 3 used NIOSH-collected exposure data but used different cohort definitions, lengths 4 of follow-up, and lengths of exposure lags to account for cancer latency 5 ' " " ' Lung cancer analysis Cohort definition Follow-up Lung cancer c a s e s /A 1' Slope per fiber/cc-year x l()"3 (calendar year) Risk based on Upper Confidence Limit UCL on the slope (per fibers/cc) This current Hired post-1959 2006 32/880 5.8 assessment Exposures 1960-1982 0.068 Sullivan (2007) Still alive post-1959 2001 99/1.672 4.2 White males Exposures 1960-1982 0.037 Moolgavkar et al. Still alive post-1959 2001 95/1,662 1.69 (2010)b Wliite males Exposures 1960-1982 0.011 Berman and Still alive post-1959 2001 93/1,672 3.96 Cramp (2008)3 Wliite males Exposures 1960-1982 0.079 Larson et al. Full cohort 2006 98/1,862 1.61 (2010b) Exposures 1935-1993 0.010 6 7 aSullivan (2007) and reanalysis o f Sullivan (2007) state slightly different number o f lung cancers. It is impossible to 8 reconcile these numbers from published information. 9 bReanalysis o f Sullivan (2007). 10 11 12 The first alternative analytic approach to estimating the extra risk from a linear regression 13 of individual mortality data was to use a standard technique used in EPA cancer risk assessments 14 (U.S. EPA. 2005a) when individual-level data are not available. This approach used a weighted 15 linear regression of standardized rate ratio (SRR) estimators for lung-cancer mortality in white 16 males, as calculated in the NIOSH cohort analysis (Sullivan. 2007), with categorical cumulative 17 exposure and a 15-year lag. The Sullivan (2007) analysis was based only on those who have not 18 died or been lost to follow-up before January 1, 1960 (in contrast to employment beginning after 19 January 1, 1960), because the NIOSH software program (Life Table Analysis System) used for 20 this analysis only has statistics on external comparison rates for asbestosis [one of the primary 21 outcomes of interest in the Sullivan (2007) analysis] beginning in 1960. The SRR analysis 22 involves internal comparisons of lung-cancer mortality rates in the higher exposure categories This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-106 DRAFT--DO NOT CITE OR QUOTE 1 to the lung-cancer mortality rates in the lowest exposure category. The weights used for the 2 SRRs were the inverses of the variances. Midpoints of the exposure intervals were used, and for 3 the unbounded interval, the midpoint was assumed to be twice the starting point of that interval. # . i # _o 4 Using this approach, a regression coefficient of 4.2 e 10 ' per fiber/cc-year 5 ([SE] = 7.7 x 10 4 per fiber/cc-year, p = 0.03) was obtained from the weighted linear regression 6 of the categorical SRR results. Because the data from Sullivan (2007) were already adjusted for 7 the length of an occupational year (240 days) to the length of a calendar year (365 days), only the 8 standard adjustment for inhaled air volume was performed. The concentration estimate obtained 9 using this regression modeling and the life-table analysis procedure was LECoi = 0.272 fibers/cc, 10 resulting in the lung cancer unit risk of 0.0368 per fibers/cc. 11 The Berman and Crump (2008) reanalysis was based on the Sullivan (2007) summary 12 results except they used a lag of 10 years (Sullivan, 2008, personal communication to Berman 13 and Crump). They fit the IRIS IUR (U.S. EPA. 1988a) lung cancer model to aggregate data 14 using an extra multiplicative parameter a (in this model, the relative risk at zero exposure is 15 estimated a rather than 1). In this model, the relative risk at zero exposure is a rather than 1 16 (unity). With a = 1, their model did not fit, and with a estimated, the fit was satisfactory, 17 Berman and Crump (2008) chose the central estimate of the slope from the fit with a estimated, 18 but constructed an "informal" 90% confidence interval by the union of two confidence intervals 19 (this upper bound is shown in see Table 5-20). This was done to address uncertainty in the 20 estimated parameter a, similar to what is done in this current assessment with estimated lag and 21 decay. Note also, that Berman and Crump (2008) also provide an UF to adjust for several 22 sources of uncertainty in exposures, resulting in an upper bound risk of 0.3162. 23 The second alternative analytic approach to estimating the extra risk of lung cancer from 24 a Cox regression with time-dependent covariates of individual mortality data was to use the 25 results published by Larson et ai. (2010b), with cumulative exposure and a 20-year lag. This 26 analysis of lung-cancer mortality was based on the full cohort of 1,862 workers updated until 27 2006 and using the same model form as the current EPA analysis (the extended Cox proportional # . _o 28 hazards model). Larson et al. (2010b) reported a regression coefficient of 1.06 1 10 per This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-107 DRAFT--DO NOT CITE OR QUOTE 1 fiber/cc-year (SE = 3.1 K 10 per fiber/cc-year, p = 0.0006)/' EPA assumed that the cumulative 2 exposures reported by Larson et al. (2010b) were based on years of occupational exposure 3 (240 days per year) during a 365-day calendar year. In order to account for exposure on every 4 day of the year for a calculation of unit risk, an adjustment for exposures during the length of an 5 occupational year (240 days) to the length of an calendar year (365 days) and an adjustment for 6 the volume of inhaled air were performed to match EPA's analyses. The concentration estimate 7 obtained using the Larson et al. (2010b) regression modeling and the life-table analysis 8 procedure was LECoi = 1.26 fibers/cc, resulting in a lung cancer unit risk of 0.0103 per fibers/cc. 9 Moolgavkar et al. (2010) also used the Cox proportional hazards model with 10 time-dependent covariates for analysis of the Sullivan (2007) cohort with a 15-year lag. The 11 parameter in this study estimates 1.11 x 10 per fiber/cc-year (SE = 2.5 * id 4 per 12 fiber/cc-year), which is very close to Larson et al. (2010b), and, therefore, the lung cancer unit 13 risk based on their analysis would be very close to Larson et al. (2010b). Comparison with 14 McDonald et al. (2004) is difficult, since their outcome is defined as respiratory cancer (ICD-9 15 160-165), which is more expansive than other researchers' definitions of the outcome as lung 16 cancer, and their sub-cohort of 406 white men employed before 1963--a time period when 17 exposure assessment was less reliable and more likely to include exposure-measurement error; 18 nonetheless, the parameter estimate resulting from the Poisson analysis by McDonald et al. 19 (2004) was 3.6 x |() 3per fiber/cc-year. 20 EPA based their analyses on the exposures that occurred after 1959, while the Sullivan 21 (2007), Larson et al. (2010b), and Moolgavkar et al. (2010) analyses were based on the cohort 22 including those hired before 1960, and McDonald et al. (2004) included only workers hired 23 before 1964. As explained in detail in the discussion (see Section 5.4.6) on uncertainty in the 24 exposure assessment, there were only several measurements from the 1950s and one from 1942, 25 and most of the exposure estimation for the early years of the cohort's experience were based on 26 estimates of the ratio of dust to fibers estimated in the late 1960s and extrapolated backwards in 27 time for several decades. Moreover, 706 of the workers hired before 1960 (not necessarily 28 short-term) did not have an exposure measurement assigned to them at all, leading to much 29 larger measurement error. These limitations in the underlying exposure assessment for the years43 43Note that EPA results based on the sub-cohort hired after 1959 were from the same model form but based on the cumulative exposure with a 10-year lag and had a slope o f 5.81 x 10"3 per fibers/cc-year (SE = 2.48 x 10"3 per fiber-cc/year, p = 0.018). This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-108 DRAFT--DO NOT CITE OR QUOTE 1 prior to 1968 likely resulted in exposure measurement error that could have attenuated the 2 analytic regression results, thereby yielding a smaller effect estimate for the whole cohort 3 compared to the sub-cohort hired after 1959. It appears the differences in results are mostly 4 attributable to the time periods of analysis and corresponding to the time period measurement 5 errors rather than the analytic approach. The small discrepancy between observed lung cancer 6 deaths between this current assessment and Larson et al. (2010b), described in Section 4.1.1.1, is 7 unlikely to play a role in the difference between risk estimates. Moreover, for the sub-cohort 8 hired after 1959, all deaths are included in the Larson et al. (2010b) lung cancer-counting rules. 9 None of the approaches used by McDonald et al. (2004), Sullivan (2007), nor Larson et 10 al. (2010b) could have been appropriately used for the unit risk of mesothelioma as they are not 11 based on absolute risk metrics of association, and the current assessment considered the relevant 12 metric of association to be the absolute risk. Berman and Crump (2008) did not evaluate risk of 13 mesothelioma. Moolgavkar et al. (2010) used an absolute risk model for mesothelioma. These 14 results are summarized in Table 5-21. The upper bound results for the full cohort presented by 15 Moolgavkar et al. (2010) are about 80% of the IRIS IUR (U.S. EPA. 1988a) estimate of 16 mesothelioma slope factor in a similar RTW-type metric, leading to an approximately 80% 17 estimate of the mesothelioma unit risk, as dependence is linear in the mesothelioma slope factor 18 (see Eq. 5-5). This is very close to this current assessment's estimate based on the sub-cohort, 19 which is also about 80% of the IRIS IUR (U.S. EPA. 1988a) estimate of mesothelioma risk. 20 Durati on of employment is the best metric for the full cohort, and it does not support 21 exposure-response estimation. 22 23 5.4.5.4. Applications o f the Combined Mesothelioma and Lung-cancer mortality IUR to 24 Partial Lifetime Environmental Exposure Scenarios 25 In the application of the IUR, scenarios other than lifetime environmental exposure are often of 26 interest to risk assessors. The life-table analysis in the (general) IRIS IUR for asbestos (U.S. 27 EPA. 1988a) predicts risk increases as the age of the first exposure decreases. The authors of 28 that analysis recommended the life-tables in that analysis be consulted when assessing partial 29 lifetime exposures (U.S. EPA. 1986a). In 2008, EPA (Office of Solid Waste and Emergency 30 Response) provided guidance for calculating risk estimates for less-than-lifetime exposures 31 based on the source life-table analysis (U.S. EPA. 20081. The age-at-onset of exposure and This document is a draftfor review purposes only and does not constitute Agency policy. 5-109 DRAFT--DO NOT CITE OR QUOTE 1 Table 5-21. Mesothelioma regression results from different analyses of 2 cumulative exposure in the cohort of workers in Libby, MT. All analyses 3 used NIOSH-collected exposure data but different cohort definitions, lengths of 4 follow-up, and lengths of exposure lags to account for cancer latency 5 Mesothelioma analysis Cohort definition Mesothelioma Follow-up cascs/TV Mesothelioma risk (absolute risk model) (per fibers/cc) This current assessment Hired post-1959 Exposures 1960-1982 2006 7/880 Upper Bound = 0.12 Central = 0.08 Sullivan (2007) Still employed post-1959 White males Exposures 1960-1982 2001 15/1,672 No estimates o f absolute risk Moolgavkar et al. ( 2 0 10)a Still employed post-1959 White males Exposures 1960-1982 2001 15/1,662 Upper Bound ~ 0.13 Central ~ 0.08 Larson et al. (2010b) Full cohort Exposures 1935-1993 2006 19/1,862 No estimates o f absolute risk Berman and Crump Still employed post-1959 (2008)a White males Exposures 1960-1982 6 7 aReanalysis o f Sullivan (2007). 8 9 2001 15/1,672 No estimates provided 10 duration-dependent unit risks reflect the influence of the time-cubed function in the 11 mesothelioma model (see Eq. 5-5) (U.S. EPA, 2008. 1986a) used in the 1986 assessment. 12 Because the time-cubed mesothelioma model, or parameterization of exposure metrics, did not 13 fit the data for mesothelioma mortality from exposure to the Libby Amphibole asbestos, the 14 approach to estimating risk of partial life exposure recommended by EPA when applying the 15 general IRIS IUR for asbestos (US. EPA. 2008) is not appropriate when applying the Libby 16 Amphibole asbestos-specific IUR. 17 Thus, this current assessment recommends that estimates of the risks of less-than-lifetime 18 exposures be computed by simple calculations of average lifetime exposure concentration 19 multiplied by the IUR. This recommendation is consistent with standard Superfund guidance 20 where exposures are estimated and averaged across a lifetime exposure, and the IUR is simply 21 applied to calculate excess cancer risk (U.S. EPA. 2008. 2001b). 22 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-110 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6. Uncertainties in the Cancer Risk Values 2 It is important to consider uncertainties in the derivation of the mesothelioma and 3 lung-cancer mortality risks in this assessment in the context of uncertainties in animal-based 4 health assessments. This assessment does not involve extrapolation from high doses in animals 5 to low doses in humans. This assessment is based on a well-documented and well-studied cohort 6 of workers with adequate years of follow-up to evaluate mesothelioma and lung-cancer mortality 7 risks with PODs within the range of the data. The discussions below explore uncertainty in the 8 derivation of the IUR in order to provide a comprehensive and transparent context for the 9 resulting cancer mortality risk estimates. 10 11 5.4.6.I. Sources of Uncertainty 12 Sources of uncertainty in this assessment include 13 14 15 1) Uncertainty in low-dose extrapolation, 16 2) Uncertainty in exposure assessment, including analytical measurements 17 uncertainty, 18 3) Uncertainty in modelform, 19 4) Uncertainty in selection o f exposure metric, 20 5) Uncertainty in assessing mortality corresponding to the cancer endpoints, 21 6) Uncertainty in control o fpotential confounding in modeling lung-cancer 22 mortality, 23 7) Uncertainty due to potential effect modification, 24 8) Uncertainty due to length o ffollow-up, 25 9) Uncertainty in use o f life-tables to calculate cancer mortality unit risks, 26 10) Uncertainty in combining o f mortality risks to derive a composite cancer 27 mortality IUR, 28 11) Uncertainty due to extrapolation o ffindings in adults to children. 29 30 This document is a draftfor review purposes only and does not constitute Agency policy. 5-111 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6.I.I. Uncertainty in low-dose extrapolation 2 A common source of uncertainty in quantitative cancer risk assessments generally derives 3 from extrapolating from high doses in animals to low doses in humans. Compared to 4 assessments based on animal data, the uncertainty from low-dose extrapolation in this 5 assessment employing occupational epidemiology data is considered to be somewhat reduced for 6 the following reasons. The NIOSH worker cohort developed by Sullivan (2007) includes 7 410 workers employed less than 1 year among the 880 workers hired on or after January 1, 1960. 8 Although short-term workers, on average, experience a mean exposure intensity per day worked 9 greater than workers employed more than a year (Sullivan. 2007). the cohort nevertheless 10 includes many short-term workers with relatively low cumulative occupational exposures. 11 Further, inclusion of salaried workers in the NIOSF1 cohort (Sullivan. 2007) adds many workers 12 with lower workplace exposure. Thus, while occupational exposure concentrations may be 13 generally higher than typical ongoing environmental concentrations, the low-dose exposures in 14 this occupational database may be representative of nonoccupational exposures. 15 While many occupational epidemiology studies are based on relatively high exposure 16 levels that are beyond the range of common environmental exposures, many in the Libby 17 workers cohort experienced exposures that were near or below the PODs derived from the 18 life-table analysi s. The POD for the selected lung-cancer mortality exposure metric was 19 0.191 fibers/cc. The POD for the selected mesothelioma mortality exposure metric was 20 0.106 fibers/cc. Among the workers hired after 1959 who had at least 1 year of occupational 21 exposure (n = 470; 20 lung cancer deaths), there were 19 (4%) with average occupational 22 exposure concentrations of less than 0.3 fibers/cc, includi ng 1 lung cancer death (5%). 23 Although data might have been modeled down to a very low cumulative exposure level, 24 the Guidelinesfor Carcinogen Risk Assessment (U.S. EPA. 2005a) recommend defining a POD 25 for low-dose extrapolation in order to increase the stability of the 1TJR estimate at lower 26 exposures, where fewer cancers might be expected. Thus, the uncertainty associated with 27 low-dose extrapolation is somewhat mitigated since the linear extrapolations from the dose 28 associated with the POD from the life-table analyses of each cancer endpoint were encompassed 29 within the observed data range. Nonetheless, some uncertainty remains in the extrapolation from 30 occupational exposures to lower environmental exposures when using a POD. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 5-112 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6.I.2. Uncertainty in exposure assessment 2 Accurate exposure assessment is generally considered to be a major challenge for 3 occupational epidemiologic studies and is a challenge that is well recognized by the NIOSH 4 investigators (Amandus et al.. 1987a). As stated previously in Section 5.4.3.3, while it is 5 generally true that the use of more data is an advantage in statistical analyses because it allows 6 for the computation of more statistically precise effect estimates, this advantage in precision may 7 be offset by a negative impact on the accuracy of the effect estimate if an increase in sample size 8 is accompanied by greater exposure misclassification or other biases. Therefore, EPA decided to 9 base this Libby Amphibole asbestos-specific human health risk assessment upon the mortality 10 experience of workers hired on or after January 1, 1960. EPA's use of the sub-cohort analysis is 11 based on the belief that it is important to accurately estimate the true underlying 12 exposure-response relationships by relying on the most accurate exposure data. The use of this 13 sub-cohort greatly reduces the uncertainty in exposure error compared to evaluations based on 14 the entire cohort. More specifically, 15 16 17 a) Job category and department codes were compl etely unknown for 706 of the 18 991 workers'jobs from 1935 to 1959 (71% of the cohort for this time period). These 19 workers were assigned the same exposure concentration (66.5 fibers/cc) for all years 20 without this information. Examination of the post-1959 cohort removes this 21 significant source of exposure misclassification (only 9 of 880 sub-cohort workers did 22 not have department code and job category information). 23 b) Using the more recently hired cohort minimizes the uncertainty in estimated worker 24 exposures based on the JEM, which was informed by air sampling data available in 25 1956 and later years. Although there are still uncertainties in the task-specific 26 exposure estimates from 1960-1967, uncertainty in the assessment of earlier 27 exposure levels is considerably greater. 28 c) Exposure measurements were collected from the area samples and represented 29 exposures for all the workers with the same job code. Statistically, this causes 30 Berkson measurement error effect, which is described later in this section. 31 32 As the EPA exposure-response modeling for mesothelioma and lung-cancer mortality is 33 based on the post-1959 sub-cohort, the remaining discussion of uncertainty in exposure 34 measurement will address these data. 35 This document is a draftfor review purposes only and does not constitute Agency policy. 5-113 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6.I.2.I. Sources o f uncertainty injob history information 2 Worker exposures for the EPA exposure-response modeling were calculated based on job 3 histories and the JEM from 1960 through 1982 (see Figure 5-3). Overall, there is little 4 uncertainty in the job history information. Regarding exposure estimation for the occupational 5 cohort, the NIOSH investigators (Amandus et al,, 1987a) conducted a detailed retrospective 6 exposure assessment to estimate the individual worker exposures. NIOSH used extensive 7 occupational exposure data to construct the time-specific JEM, spanning decades (Amandus et 8 ah, 1987a). These data were reabstracted from the workers' employment records for quality 9 assurance (Sullivan, 2007). NIOSH records on work histories and job-specific exposure 10 extended from the 1930s through May 1982. But, the vermiculite mining and milling operation 11 continued on for several years, and some workers were retained through 1993 for plant close-out 12 activities. Only 148 members of the post-1959 cohort (n = 880) were employed as of the May 13 1982 employment records when the cohort was enumerated by NIOSH (Sullivan. 2007). 14 Because exposure concentrations in 1982 (see Table 5-7) were generally below 1 fiber/cc with 15 only two locations having concentrations of 1.2 fibers/cc, it is unlikely that these workers' 16 exposures were significantly underestimated. 17 18 Sources o f uncertainty in exposure intensityfor the identified location operations 19 The available exposure data that inform the JEM include over 4,000 air samples, the 20 majority of which were collected after 1967 (see Table 4-1). All of the job location exposure 21 estimates (see Table 5-7) from 1968-1982 were directly informed from air samples collected on 22 membrane filters and analyzed for fibers by PCM. The availability of site- and task-specific air 23 samples for these years provides a good basis for the exposure estimates. However, there are 24 some uncertainties in estimating asbestos exposures using air samples analyzed by PCM. 25 26 27 1) PCM analysis does not determine the mineral or chemical make-up of the fiber: The 28 PCM method defines and counts fibers based on the size (aspect ratio and length) of the 29 particle without regard for the material that makes up the fiber being viewed. The PCM 30 method was developed for use in occupational environments where asbestos was present, 31 and the nature of the fibers should be further evaluated to confirm the fibers viewed 32 under PCM are asbestos. McGill University researchers evaluated the fibers collected on 33 membrane filters in the early 1980s and confirmed the presence of asbestos fibers in the 34 tremolite-actinolite solution series consistent with the Libby Amphibole asbestos This document is a draftfor review purposes only and does not constitute Agency policy. 5-114 DRAFT--DO NOT CITE OR QUOTE 1 (McDonald et al.. 1986a). NIOSH researchers confirmed the presence of tremolite 2 asbestos in bulk dust samples but not in air samples from the facility (Amandus et ah. 3 1987a). Although less specific to fibers, 60-80% of the airborne dust in the mills in 1968 4 was tremolite, further supporting the presence of asbestos in the air (based on State of 5 Montana air sampling, and X-ray diffraction analysis by the Public Health Service [PHS 6 correspondence, October 17, 1968]). However, although the presence of mineral fibers in 7 the actinolite-tremolite series was confirmed in the work environment, it is possible that 8 there were also fibers counted by PCM from other materials (such as textiles from clothes 9 and packaging materials). Therefore, it is unknown from these data what proportion of 10 the counted PCM fibers was mineralogically asbestos, or other materials present in the 11 workplace. 12 2) PCM defines fibers as particles with an aspect ratio greater than 3:1: There is an ongoing 13 debate in the literature on asbestos toxicity regarding the influence of aspect ratio on 14 relative toxicity. Specifically, in mining environments, it has been speculated that a 15 larger proportion of low aspect ratio fibers from mineral dusts may significantly impact 16 the apparent cancer potency of the measured PCM fibers in those environments (Berman. 17 2010; U.S. EPA. 1988a). There are few data available to understand fiber morphology 18 and fiber aspect ratios in the Libby cohort working environment. Considering the 19 post-1959 cohort, PCM fiber size distribution and aspect ratio data only exist for a set of 20 eight air samples (599 fibers) collected from the wet mill and screening operations and 21 analyzed by the NIOSH researchers (Amandus et ah. 1987a). For these air samples, over 22 96% of the fibers viewed by PCM had an aspect ratio greater than 10:1 (Table 4-2) 23 (Amandus et ah. 1987a)44' However, because these samples were provided by the 24 company in the early 1980s, they do not represent conditions in the old wet mill or dry 25 mill operations, which were significantly dustier environments (Amandus et ah. 1987a). 26 It is possible that prior to industrial hygiene (IH) modifications in 1974, the dry and old 27 wet mills generated proportionally more mineral dusts than screening and new wet mill 28 operations after IH modifications. No data are available for the mining environment, 29 which would also be expected to generate a range of mineral dusts. Therefore, there is a 30 significant uncertainty about the size and aspect ratio of fibers included in PCM fiber 31 counts for the majority of the post-1960 workers cohort. 32 3) The resolution of visible PCM fibers: Current analytical instruments used for PCM 33 analysis have resulted in a standardization of minimum fiber width considered visible by 34 PCM between 0.2 and 0.25 pm. Historical PCM analysis (1960s and early 1970s) 35 generally had less resolution, and fibers with minimum widths of 0.4 or 0.44 pm were 36 considered visible by PCM (Amandus et ah. 1987a; Rendall and Skikne. 1980). 37 McDonald et al. (1986a) compared fibers viewed by PCM and TEM and estimated that 38 approximately 1/3 of the total fibers could be viewed by the optical microscope. Because 39 38% of the fibers were <5 pm in length, this implies approximately 30% were not 40 viewable by optical microscopy for other reasons, such as width. However, it is 41 unknown what proportion of that 30% would be viewed with the minimum width 44Although Amandus et al. (1987a) report the sizing o f PCM fibers, the details o f the methodology are not given regarding how these fibers were identified. No method is cited, and it is unclear if the sizing was done by PCM or TEM for fibers in the reported size categories. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-115 DRAFT--DO NOT CITE OR QUOTE 1 resolution of 0.25 urn for later optical microscopy. It is likely that early PCM counts 2 were underestimated relative to the later data for the cohort but by less than a factor of 2. 3 4 5 Prior to 1968, no air sampling data were available for 23 of the 25 job location operations 6 (see Table 4-2), and the exposure estimates were extrapolated from later air sampling data. 7 Amandus et al. (1987a) recognized there is significant uncertainty in the extrapolation of 8 available air sampling data to previous time periods. The researchers took into account major 9 changes in operations and interviewed employees in the early 1980s regarding previous years of 10 operati on. The assumption s used to make these extrapolations are clearly stated for each of the 11 plant operations. For four operations, high and low estimates of pre-1968 exposures were 12 provided based on different sets of exposure assumptions (see Table 5-7). For ore loading, there 13 were negligible differences in the exposure estimates for the period from 1960-1967 (10 7 14 versus 9 fibers/cc). For drilling, the river dock, and the bagging plant, there were 3.4-, 2.6-, and 15 2.8-fold differences, respectively, between the high and low estimates of exposure between 1960 16 and 1968. 17 Dry mill exposures between 1960 and 1968 were informed by air sampling for total dust 18 collected in the dry mill facility from 1956-1969 (where total dust was collected by midget 19 impingers). Amandus et al. (1987a) derived a conversion factor of 4.0 fibers/cc per mppcf to 20 apply to the two location operations in the dry mill during these years. There was a range of 21 conversion factors considered for the dry mill depending on how the dust and fiber air samples 22 (PCM) were grouped and averaged (1.2 to 11.5 fibers/cc per mppcf). A subset of dust and fiber 23 samples available over the same time period (1967-1968) resulted in a ratio of 8.0 fibers/cc per 24 mppcf. In contrast, a ratio of 1.9 fibers/cc resulted when total dust samples from 1969 were 25 compared with fiber samples from 1970. However, both of these subsets had limited numbers of 26 samples available. Therefore, the conversion factor of 4.0 fibers/cc per mppcf was selected 27 based on using the maximum samples available over a time period when the dry mill exposures 28 were considered similar: dust samples (1965-1969) and fiber samples (1967-1971). 29 30 5.4.6.1.2.2. Sources o f uncertainty in the calculation o f thejob-exposure matrix (JEM) 31 The exposures in the JEM (see Figure 5-3) were calculated from the exposure intensities 32 of the various task-specific exposure intensities shown by job location operation (see Table 5-7). This document is a draftfor review purposes only and does not constitute Agency policy. 5-116 DRAFT--DO NOT CITE OR QUOTE 1 The uncertainties in the exposure intensity for the job location operations will impact the JEM 2 Additionally, for each of the job categories in the JEM, NIOSFI researchers defined which tasks 3 (job location operations) were conducted and for what proportion of the work day. A TWA 4 exposure for each job category across time was calculated based upon these assumptions and the 5 task-specific exposure estimates. There is a measure of uncertainty in these assumptions for 6 each job category. Additionally, there is inter-individual variation within the job categories. 7 These uncertainties are common to exposure reconstruction for epidemiological cohorts. 8 9 5.4.6.1.2.3. Uncertainty in the exposure metric 10 The PCM measurement is the available exposure metric for analysis of Libby worker 11 cohort at this time. Currently, there is no optimal choice of the best dose metric for asbestos, in 12 general, and, in particular, for Libby Amphibole asbestos, even if a TEM-based dose-response 13 JEM was available. Uncertainties related to PCM analytical method are discussed in Section 2. 14 Briefly, PCM cannot di stingui sh between asbestos and nonasbestos material or differentiate 15 between specific types of asbestos. Further, due to limitations of this methodology, PCM does 16 not take into account fibers smaller than 5 pm in length. 17 18 5.4.6.1.2.4. Evaluation o f the effects o f uncertainties in exposure measurement 19 An understanding of the effects of exposure measurement error on the risks estimated 20 from epidemiologic analyses is important to place these possible exposure measurement errors in 21 context. The effect of exposure measurement error on estimates of the risk of mesothelioma or 22 lung-cancer mortality attributable to exposure depends upon the degree to which that error may 23 be related to the likelihood of mesothelioma or lung-cancer mortality. Exposure measurement 24 error that is similar in pattern among workers who died of lung cancer to exposure measurement 25 error in people who did not die of lung cancer is a nondifferenti al exposure measurement error. 26 Differential exposure measurement error that is associated with the outcome can cause bias in an 27 effect estimate towards or away from the null, while nondifferenti al exposure error typically 28 results in bias towards the null (Rothman and Greenland. 1998). From the above evaluation of 29 uncertainties, there is no indication that the uncertainties in job history information, exposure 30 estimates for specific tasks, or calculation of the JEM would be differential based on the cancer 31 health outcome data. Therefore, these uncertainties are considered nondifferential, and the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-117 DRAFT--DO NOT CITE OR QUOTE 1 general result is likely to be an attenuation in risk estimates towards the null (that is, the addition 2 of random noise to a clear signal tends to reduce the clarity of the observed signal and the 3 avoidance of random noise--here from poor quality exposure measurements--results in a 4 stronger observed signal). 5 Generally speaking, if the exposure concentrations estimated by NIOSH were 6 systematically too high, then the associated risks of exposure estimated in the regression analysis 7 would be low since the same actual risk would be spread across a larger magnitude of exposure. 8 Similarly, if the exposure concentrations estimated by NIOSH were systematically too low, then 9 the associated risks of exposure estimated in the regression analysis would be too high. From the 10 above evaluation, the majority of the sources of uncertainty are not systematic. There are a few 11 areas of uncertainty that may be classified as biased : 12 13 14 1) High- and low-exposure estimates for four job location operations were provided 15 between 1960 and 1967. Amandus et al. (1987a) chose the high estimates of 16 exposure for these job location operations when calculating the JEM. Therefore, 17 there will be a bias towards the high end for the job categories informed by these 18 data. There was a 1.1- to 3.4-fold difference between the high and low estimates. 19 This difference will be less pronounced where these exposure concentrations are 20 averaged with other job location operations in the JEM and across multiple jobs for 21 the majority of the workers (see Figure 5-3). 22 2) Current PCM analysis would count more fibers relative to early PCM methods based 23 on minimum fiber width resolution. For example, Amandus et al. (1987a) used a 24 minimum width cutoff of 0.44 in their review of PCM fibers in the 1980s, which may 25 have resulted in as much as a twofold underestimate compared to current PCM 26 methods with a width resolution of 0.25 pm. Additionally, as PCM methodology has 27 developed over time, it is unknown when PCM results from company records would 28 be considered relatively standard to a minimum width resolution between 0.2 and 29 0.25 pm. Also, prior to standardization of PCM to 0.25-pm minimum width, there 30 was inter-laboratory variability as well. Therefore, the size distribution of PCM 31 fibers (e.g., minimum width) reported in the JEM may have changed over time. 32 Although theoretically a systematic bias, given the years for which PCM data are 33 available, this is likely an insignificant effect. 34 3) Asbestos was a contaminant of vermiculite that was the primary object of production. 35 Mine, old dry mill, and wet mill ambient air may have contained material other than 36 asbestos that could have contributed to PCM fiber count. The exposures in the old 37 dry and wet mills and mine location may have included a greater proportion of dust to 38 fibers than tasks using the ore and refined vermiculite after the new wet mill became 39 operational. It is possible there is a systematic over-count of fibers in the dusty This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-118 DRAFT--DO NOT CITE OR QUOTE 1 environment due to interference from mineral fragments. This likely impacts the 2 exposure intensity for 23 of 25 job location operations within the mine and old dry 3 mill. Estimated exposures from job categories that include these operations may be 4 biased upwards. 5 6 7 Nondifferential measurement error in a continuous exposure can be of the classical or 8 Berkson type and typically arises in environmental and occupational settings as a mixture of the 9 two forms (Zeger et al.. 2000). Classical measurement error occurs when true exposures are 10 measured with additive error (Carroll et al,, 2006) and the average of many replicate 11 measurements, conditional on the true value, equals the true exposure (Armstrong. 1998). This 12 error is statistically independent of the true exposure that is being measured and attenuates true 13 linear effects of exposure, resulting in effect estimates in epidemiologic studies that are biased 14 towards the null (Heid et ah. 2004; Zeger et al.. 2000; Armstrong. 1998). Such errors occur 15 when the mean values of multipl e local air samples are used. 16 Berkson measurement error is independent of the surrogate measure of exposure (Heid et 17 ah. 2004; Berkson. 1950) and is present when the average of individuals' true exposures, 18 conditional on the assigned measurement, equals the assigned measurement. Berkson 19 measurement error can arise from the use of local area mean sampled exposures to represent the 20 individual exposures of people in that area--even when the estimated area mean is equal to the 21 true underlying mean (i.e., no classical measurement error). Examples of random variability in 22 personal behavior that may produce Berkson measurement error in personal exposure estimates 23 include the volume of air breathed per day among the workers and the effectiveness of an 24 individual's nasal filtration at removing contaminants. In general, Berkson measurement error is 25 not thought to bias effect estimates but rather increases the standard errors of effect estimates 26 (Zeger et al.. 2000). However, some epidemiologic studies have suggested that Berkson 27 measurement error can produce a quantitatively small bias towards the null in some analyses 28 (Bateson and Wright. 2010; Kim et al., 2006; Reeves et al.. 1998; Burr. 1988).Uncertainties in 29 the levels and time course of asbestos exposure for the Libby workers also adds uncertainty to 30 the evaluation of the relative fit of different exposure metrics. 31 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-119 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6.I.2.5. Exposure to other kinds o f asbestos and residential exposure 2 Another source of uncertainty in the estimation of exposures in the Libby workers cohort 3 is the potential contribution of nonoccupational or residential exposures as well as exposures to 4 other kinds of asbestos in employment before or after working in Libby. 5 Many of the workers resided in Libby, MT, before and/or after their employment at the 6 mining and milling facilities ended. The vermiculite from the mine had been used at numerous 7 sites around the town, including baseball fields around the expansion plant and as filler in 8 gardens (TJ.S. EPA. 2010a. 2001a). Exposure to asbestos could have occurred among individuals 9 outside of the workplace, particularly through activities with the potential of stirring up of dirt or 10 other materials that had been mixed with the vermiculite (Weis. 2001a). The results of 11 community sampling indicated that even 10 years after mill operations ceased during some 12 activities, asbestos fiber concentrations in the air could exceed OSHA standards established for 13 the protection of workers (Weis. 2001a). 14 Therefore, the workers' actual personal exposures as the sum of occupational and 15 nonoccupational exposures are likely to have been underestimated by the use of estimated 16 Libby-related occupational exposure alone. The difficulty stems from the lack of data on 17 residential exposures and lack of information on pre- and postemployment residence of the 18 Libby workers. Nonoccupational exposures were likely to have been smaller in magnitude than 19 the occupational exposures, but workers may have lived in and around Libby, MT, for many 20 more years than they were exposed occupationally. The impact of residential exposure could be 21 more prominent for workers with lower occupational exposure who resided in Libby for a long 22 time. Whitehouse et al. (2008) has reported several cases of mesothelioma among residents of 23 the Libby, MT region who were not occupationally exposed. However, since the report by 24 Whitehouse et al. (2008) details only the cases and does not define or enumerate the population 25 from which those cases were derived, computed relative risks from nonoccupational exposures 26 were not available. ATSDR (2000) reported higher relative risks of mesothelioma among the 27 population of Libby, MT, including former workers residing in Libby, but did not provide 28 relative risk for nonoccupational exposure. Instead, the ATSDR report on mortality (2000) 29 grouped cases among the former workers with nonoccupationally exposed cases. Therefore, it is 30 not clear what the magnitude of the contribution of workers' nonoccupational exposures was to 31 their overall risk. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-120 DRAFT--DO NOT CITE OR QUOTE 1 Some of the occupational workers with lower exposures, such as short-term workers, may 2 have either been high school or college students working during the summer or may have been 3 transient workers who may not have stayed for a long time in Libby. Sullivan (2007) analyzed 4 differences between short- and long-term workers and reported little difference between the 5 groups except for age at hire. As the short-term workers were younger on average, this 6 supported the suggestions that some of the short-term workers may have been college students 7 working during the summer. This population of short-term workers is not well defined; 8 however, it is possible that short-term transient workers could potentially have been exposed to 9 other kinds of asbestos or other lung carcinogens in their non-Libby occupational career, which 10 might have affected their pre- and post-Libby risk profile for asbestos exposure. While their 11 occupational histories other than working in Libby are unknown, it is very unlikely that they 12 include exposures of the magnitude that were encountered in the Libby mine and mill. The 13 impact of these uncertainties on regression slopes is difficult to evaluate. However the slope 14 may be somewhat underestimated as an observed increase in risk would be attributed to a larger 15 exposure differential than might have been present due to the addition of nonoccupational 16 exposures. There will also be a downward bias from random exposure measurement error with 17 lower occupati onal exposure affected disproportionately; however, the magnitude of this bias 18 would be expected to be small. 19 20 5.4.6.I.2.6. Conclusion regarding uncertainty in exposure assessment 21 Overall, there are likely to be multiple sources of uncertainty attributable to exposure 22 measurement error. It is possible that systematic error may have been introduced into the 23 exposure intensities assigned to several of the job location operations discussed above. In each 24 case, these errors in estimating exposures were overestimates. The magnitude of the potential 25 overestimates of drilling and dry and old wet mill exposures is uncertain. The dust-to-fiber 26 conversion ratio applied to the dry mill during 1960-1967 could be an over or underestimate by 27 as much as twofold. Random error in the measurement of dust or fibers would likely have 28 produced an underestimation of ri sk. There is no known bias in the assumpti ons to extrapolate 29 exposure to pre-1968 location operations outside of the dry mill, and random bias would also 30 likely have produced an underestimation of risk. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 5-121 DRAFT--DO NOT CITE OR QUOTE 1 5.4.6.I.3. Uncertainty in modelform 2 For mesothelioma mortality, the Poisson regression model is commonly used for rare 3 outcomes and has been applied by McDonald et al. (2004) and Moolgavkar et al. (2010) to 4 model mesothelioma risk in the Libby worker cohort. For lung-cancer mortality, the Cox 5 proportional hazards model is a well-established method that is commonly used in cohort studies, 6 including by Larson et al. (2010b) and Moolgavkar et al. (2010) for the Libby worker cohort, 7 because this type of survival analysis takes into account differences in follow-up time among the 8 cohort. Larson et al. (2010b) conducted Poisson regression analyses and reported that their lung 9 cancer results using this different model form were similar to those from their extended Cox 10 proportional hazards models, but those results were not shown. 11 Both of these model forms allow for the evaluation and control of important potential 12 confounding factors such as age, sex, and race, and for the modeling of exposure as a continuous 13 variable. Both model forms yielded exposure-response results with good fit to the occupational 14 exposure data. The default assumpti on of the extended Cox proporti onal hazards model as well 15 as the Poisson regression model is that all censoring (due to death or loss to follow-up) is 16 assumed to be independent of exposure to the Libby Amphibole asbestos (e.g., death in an 17 automobile accident or moved to Canada). However, exposure to Libby Amphibole asbestos 18 may be causing deaths from other causes such as asbestosis or nonmalignant respiratory disease 19 (Larson et al.. 2010b), which is referred to as dependent censoring. The concern is that the 20 observation of lung-cancer mortality may be precluded by mortality from other causes. 21 In the cohort of 880 workers hired after 1959, 32 died of lung cancer, while 10 died of 22 asbestosis, and 21 died of nonmalignant respiratory disease. The mean length of follow-up from 23 the date of hire until death for the workers who died of lung cancer was 24.9 years. However, 24 the mean length of follow-up for the workers who died of asbestosis or nonmalignant respiratory 25 disease was 30.4 years, so it does not appear that early deaths from other causes associated with 26 exposure to the Libby Amphibole asbestos (Larson et al.. 2010b) would have precluded many 27 cases of lung cancer. This implies that any potential bias in the lung cancer risk estimates due to 28 dependent competing risks is small. 29 With respect to mesothelioma mortality, it should be noted that the exposure-response 30 modeling is limited by the number of deaths. However, dependent censoring, as described 31 above, is not accounted for in the Poisson regression model and likely causes a downward bias in This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-122 DRAFT--DO NOT CITE OR QUOTE 1 the estimation of risk. The mean length of follow-up for the workers who died of mesothelioma 2 was 30.1 years, and there is some evidence that early deaths from other exposure-related causes 3 precluded an individual's risk of death from mesothelioma; only lung cancer exhibited a shorter 4 average follow-up time compared to mesothelioma, and in 419 cases of mesothelioma, 5 mesothelioma and lung cancer were never coidentified (Roggli and Volliner. 20081. 6 7 5.4.6.I.4. Uncertainty in selection o f exposure metric 8 There is uncertainty about what metric should be used for modeling exposure to Libby 9 Amphibole asbestos. The previous IRIS 1UR assessment for asbestos (U.S. EPA. 1988a) found 10 that cumulative exposure with a 10-year lag was the best metric for lung-cancer mortality, and a 11 more complicated model (see Eq. 5-5) based on average cohort exposure intensity, average 12 cohort time since first exposure, and average duration of employment was the best metric for 13 mesothelioma mortality. This current assessment evaluated these models, but also models that 14 include unlagged and lagged cumulative exposure with and without a half-life of various lengths, 15 and RTW exposure with and without a half-life. In the analysis of comparative model fit, lagged 16 cumulative exposure with a half-life provided the best fits for both mesothelioma and 17 lung-cancer mortality associated with Libby Amphibole asbestos. However, evaluation of 18 20-year lag and longer lag times for mesothelioma was not possible, as the earliest mesothelioma 19 death happened less than 20 years from the start of the exposure, and, hence, exposure was 20 zeroed out, and the fit of any model with 20-year lag was very poor. Latency time for 21 mesothelioma may be as long as 60-70 years [e.g., Bianchi and Bianchi (2009)1. so the precise 22 lag time is uncertain. 23 In evaluating the data on lung fiber burden, Berry et al. (2009) estimated the range of the 24 half-life for crocidolite to be between 5 and 10 years. That range is consistent with the finding of 25 a 5 tolO-year half life with 10-15 years lag that provided the best fit to the Libby workers cohort 26 mesothelioma mortality data. Similarly, recent publications indicate that the relative risk of lung 27 cancer due to asbestos exposure declines 15-20 years after the cessation of exposure to asbestos 28 (Magnani et al.. 2008; Hauptmann et al.. 2002). The marginally best fit for the Libby workers 29 cohort lung-cancer mortality data was for CE models with a 5 to 20-year half life and 10-year 30 lag. However, the precise lag and half-life times are somewhat uncertain. Sensitivity analysis 31 that excluded people with high exposure during 1960-1963 (see Section 5.4.3.6.4) provides This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-123 DRAFT--DO NOT CITE OR QUOTE 1 further evidence that distinguishing between various lags and decays may be difficult with these 2 data. A limitation of this sensitivity analysis is the decrease in the number of cases, especially 3 for mesothelioma. Resolving this uncertainty would require longer follow-up time, which would 4 allow for a sub-cohort analysis of workers hired in 1967 or afterwards (when exposure estimates 5 began to be based on PCM measurements) until a sufficient number of cases would be available 6 for additional analysis. 7 These simulated decay models were derived mathematically to approximate underlying 8 biological processes that are not well understood, and their better fit is based on maximizing the 9 likelihood for the workers cohort and may not necessarily apply to the environmental exposure 10 patterns. Nonetheless, while the mode of action for carcinogenicity is unknown, the models 11 incorporating a half-life in the exposure metric were clearly preferable for mesothelioma 12 mortality, and the goal of the regression modeling effort was to identify the best fitting exposure 13 model for the Libby worker cohort. 14 The selection of the exposure metric is a source of cross-metric variability discussed in 15 Section 5.4.5.3, and the IUR incorporates this variability. The difference between this value and 16 the value derived from the best fitting exposure model describes the quantitative uncertainty, 17 which is less than twofold. 18 19 5.4.6.I.5. Uncertainty in assessing o f mortality corresponding to the cancer-specific endpoints 20 As well established in the literature, mortality rates calculated from death certificates are 21 lower than the true rate of death due to both lung cancer, and to a larger degree, mesothelioma 22 [lung cancer sensitivity: ranging from 86% in an asbestos cohort (Selikoff and Seidman. 1992) to 23 95% in general (Percy et al., 1981): mesothelioma sensitivity: ranging from 40% for ICD-9 24 (Selikoff and Seidman. 1992) to about 80% for ICD-10 (Camidge et aL 2006: Pinhei.ro et al... 25 2004). This underestimation of the true rate will result in a lower estimated risk compared with 26 that which would be estimated based on the true rate. The underestimation of risk is much more 27 pronounced for the absolute risk model (mesothelioma) than for the relative risk model (lung 28 cancer). Misdiagnosis rates would need to be quite disparate in the cohort and the comparison 29 population to impact relative risks, and this is unlikely for internal controls that were used in the 30 lung cancer analysis using the Cox proportional hazards model. Therefore, EPA considered use 31 of a procedure to adjust risks for mesothelioma--but not for lung cancer--underascertainment This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-124 DRAFT--DO NOT CITE OR QUOTE 1 (see Section 5.4.5.1.1). This procedure makes certain assumptions, in particular, that an 2 adjustment factor derived for the full cohort applies to the sub-cohort hired after 1959, and that 3 the rate of misdiagnosis of peritoneal mesotheliomas has not improved recently, and that the 4 proportion of peritoneal mesotheliomas in the cohort is estimated from the available information 5 on the type of mesothelioma in one-third of mesothelioma cases. However, overall uncertainty 6 in this adjustment is low, and the application of the adjustment reduces the bias associated with 7 the diagnostic underascertainment. 8 The endpoint for both mesothelioma and lung cancer was mortality, not incidence. The 9 latter is generally desirable, but median survival with lung cancer and, especially, mesothelioma 10 is not very long, so uncertainty related to the endpoint being death and not incidence is low. 11 There is evidence that other cancer endpoints may also be associated with exposure to the 12 commercial forms of asbestos. IARC concluded that there was sufficient evidence in humans 13 that commercial asbestos (chrysotile, crocidolite, amosite, tremolite, actinolite, and 14 anthophyllite) was causally associated with lung cancer and mesothelioma, as well as cancer of 15 the larynx and the ovary (Straif et aL 2009). Among the entire Libby workers cohort, only 16 2 deaths were found to be due to laryngeal cancer, and there were no deaths from ovarian cancer 17 among the 24 deaths of 84 female workers. The lack of sufficient number of workers to estimate 18 risk of ovarian cancer is an uncertainty in an overall cancer health assessment. 19 The remaining uncertainties attributed to assessing mortality corresponding to the cancer 20 endpoints are considered to be low. 21 22 5.4.6.1.6. Uncertainty in control o fpotential confounding in modeling lung cancer 23 It is well known that smoking is a strong independent risk factor for lung cancer and may 24 have a synergistic effect with asbestos exposure (Wraith and Mengersen. 2007). In contrast, 25 smoking is not considered a risk factor for mesothelioma (Selikoff and Lee. 1978; Anderson et 26 al,, 1976). 27 As an important potenti al confounder of the lung-cancer mortality analysis, the possibl e 28 effect of smoking on the estimated risk of lung-cancer mortality associ ated with exposure to 29 Libby Amphibole asbestos needs to be evaluated to the fullest extent possible. This 30 consideration was discussed in Amandus and Wheeler (1987) and in Section 4.1.1.3. This document is a draftfor review purposes only and does not constitute Agency policy. 5-125 DRAFT--DO NOT CITE OR QUOTE 1 Additionally, W.R. Grace and Co. instituted a smoking ban on the property in 1979 2 (Peacock. 2003). Information is not available as to the effect of this smoking ban at work on 3 smoking patterns outside of the work environment. About 30% of the sub-cohort was still 4 employed in 1979 and all of the post-1959 cohort had been terminated by May 1982, so the 5 impact of a workplace smoking ban on cohort smoking history may explain the higher proportion 6 of former smokers in the Amandus and Wheeler (19871 data. Lung cancer risks in ex-smokers 7 decrease over time compared to lung cancer risks in continued smokers. A reduction of smoking 8 in the Libby worker population may lead to fewer observations of lung cancer deaths in later 9 years of the cohort study than would have occurred in the absence of the smoking restrictions. 10 Changes in smoking behavior during the course of the epidemiological observation period would 11 lead to changes in the observed time course of lung cancer death rates. This issue is related to 12 potential effect modification of lung-cancer mortality described in Section 5.4.6.1.7. 13 Without high-quality individual-level data on smoking that could be used to control for 14 potential confounding, it is still possible to comment upon the likelihood and potential magnitude 15 of confounding and the impact any confounding would be expected to have on the lung-cancer 16 mortality risk estimates. Confounding can be controlled for in a number of ways including by 17 modeling and by restriction. Restriction of the study popul ation can reduce any potential 18 confounding by making the resulting population more similar. For instance, there can be no 19 confounding by gender when a study population is restricted to only men. This assessment 20 restricted the study population to those workers hired after 1959. Smoking habits have changed 21 over time, and it can reasonably be assumed that the range of smoking habits among those hired 22 after 1959 is less variable than that among the whole cohort, parti cularly because of the narrower 23 range of birth cohorts represented in this sub-cohort. This should have the effect of reducing 24 some of the potential for confounding. Analytic examinations of potential confounding are 25 discussed below. 26 Additionally, the extended Cox proportional hazards models controlled for date of birth, 27 which effectively controls for any secular trends in confounders over time (Tableman and Kim. 28 2004). Amandus and Wheeler (1987) cite data from the U.S. Public Health Service (HEW. 29 1979) showing a steady decrease in the prevalence of current smoking from 52.9% in 1964 when 30 the U.S. Surgeon General's report on smoking was released to 42.3% in 1970 and 37.5% in 1978 31 (HHS. 1990). If current smoking were a meaningful confounder, such a reduction in smoking This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-126 DRAFT--DO NOT CITE OR QUOTE 1 rates over time should have produced a noticeable distortion in the proportionality of the hazards 2 as the magnitude of confounding by smoking changes with smoking prevalence. No violation of 3 the proportional hazards assumption was observed in the context of the Cox proportional hazards 4 model; hence, there is no evidence of confounding by smoking in the analyses of workers hired 5 after 1959. 6 Lastly, Richardson (2010) describes a method to determine if an identified exposure 7 relationship with lung cancer is confounded by unmeasured smoking in an occupational cohort 8 study. EPA implemented this methodology to model the potential effects of Libby Amphibole 9 asbestos on the risk of COPD mortality on the sub-cohort of workers hired after 1959 (see 10 Section 5.4.3.6.5). Summarizing these findings, EPA used the method described by Richardson 11 (2010) to evaluate whether exposures to Libby Amphibole asbestos predicted mortality from 12 COPD as an indication of potential confounding by smoking and found a nonsignificant negative 13 relationship, which was inconsistent with confounding by smoking. 14 15 5.4.6.1.7. Uncertainty due to potential effect modification 16 Among the 32 deaths from lung cancer in workers hired after 1959 that were used to 17 estimate the unit risk of lung-cancer mortality (see Section 5.4.5.2), data on smoking listed 16 as 18 smokers, 4 as former smokers, and 12 of the 32 had missing data. Thus, data to support an 19 estimate of the risk of Libby Amphibole asbestos among known nonsmokers were not avail able. 20 It is theoretically possible that the risk of lung-cancer mortality estimated in this current 21 assessment is a reflection of a positive synergy between smoking and asbestos, and that the 22 adverse effect of Libby Amphibole asbestos among the potentially nonsmoking workers has been 23 overestimated. The unit risk of the lung cancer estimate herein and the combi ned mesotheliom a 24 and lung-cancer mortality IUR would then be health protective for any population that had a 25 lower prevalence of smoking than that of the Libby worker cohort. However, if the smoking ban 26 did diminish the effect of smoking, then any overestimation would be somewhat mitigated. 27 28 5.4.6.1.8. Uncertainty due to length o ffollow-up 29 There is some potential uncertainty regarding the length of follow-up for cancer 30 mortality, even more so with the restri ction of the cohort to those workers hired after 1959. The 31 hire dates among this subset of the cohort ranged from January 1960 to November 1981 (the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-127 DRAFT--DO NOT CITE OR QUOTE 1 mean date of hire was May 1971). Follow-up continued until the date of death or 2 December 31, 2006, whichever occurred first. Therefore, the range of follow-up was from 25 to 3 46 years, with a mean of more than 35 years. 4 Flowever, for mesothelioma mortality, the length of the latency period is considerably 5 longer. Suzuki (2001) reviewed 1,517 mesothelioma cases from 1975 through 2000 and was 6 able to estimate the latency for 800. Suzuki (2001) reported 17% of cases had a latency of less 7 than 30 years with 52% of cases with a latency of less than 40 years. Bianchi and Bianchi (2009) 8 estimated the mesothelioma latency in 552 cases and reported mean latency periods of 35 years 9 among insulators, 46 years among various industries, and 49 years among shipyard workers. 10 The effect of insufficient length of follow-up for mesothelioma mortality would be to 11 underestimate the risk of exposure since there would be workers who may eventually die of 12 mesothelioma that are not counted in this current assessment. Because the risk of mesothelioma 13 mortality is evaluated as an absolute risk, the unit risk of mesothelioma mortality may reasonably 14 be expected to rise with time moderated by the increase in person-years of follow-up. According 15 to the results of Suzuki (2001) and of Bianchi and Bianchi (2009), a mean length of follow-up of 16 35 years may only have captured half of all eventual mesothelioma mortality cases among the 17 Libby workers hired after 1959. If this were so, then the unit risk of mesothelioma mortality 18 could be larger than was estimated from existing data, depending on the relationship between the 19 number of additional deaths and increase in person-years. 20 21 5.4.6.I.9. Uncertainty in use o f life-tables to calculate cancer mortality IUR 22 The life-table procedure computes the extra risk of death from birth up to 85 years of age, 23 in part, because this is how national cancer incidence and mortality rate data that are one basis of 24 the life-tables are made available (see 2003-2007 SEER Table 15.10, age-specific U.S. death 25 rates). Because the prevalence of cancer mortality is a function of increasing age, this cut-off at 26 age 85 ignores a small additional risk oflung-cancer mortality among a small percentage of 27 people who have the higher background risk. This has the effect of slightly underestimating the 28 IUR that would be derived if the life-table were extended for an additional period of time, 29 accounting for longer life spans. Extension of the life-table analysis to people over the age of 30 85 requires an additional assumption. Assuming that having attained the age of 85 years, the 31 additional life expectancy is 5 years, then the lung-cancer mortality unit risk based on the LECoi This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 5-128 DRAFT--DO NOT CITE OR QUOTE 1 would be somewhat larger--on the order of 5-10%--slightly more than the additional 2 mesothelioma mortality risk if the life-tables were extended. 3 4 5.4.6.1.10. Uncertainty in combining of riskfor composite cancer IUR 5 For the purpose of combining risks, it is assumed that the unit risks of mesothelioma and 6 lung-cancer mortality are normally distributed. Since risks were derived from a large 7 epidemiological cohort, this is a reasonable assumption supported by the statistical theory, and 8 uncertainty related to it is low. 9 10 5.4.6.1.11. Uncertainty in extrapolation o ffindings in adults to children 11 The analysis of lung-cancer mortality specifically tested and confirmed the assumption 12 that the relative risk of exposure is independent of age within the age range of the occupational 13 sub-cohort hired after 1959. However, no comparable data are available to estimate the lifetime 14 risk from early life exposures. The life-table procedure is conducted so as to initiate exposure at 15 age 16 to represent adult exposures. Then, the adult-only-exposure IUR estimates derived from 16 the life-table analysis need to be rescaled to a 70-year lifespan in order to yield the standard 17 lifetime IUR, allowing risk estimate calculations involving less-than-lifetime exposure scenarios, 18 in the standard manner. After rescaling, the resulting "adult-based" IUR estimate (in contrast to 19 the unsealed "adult-only-exposure" IUR estimate obtained from the life-table calculations) can 20 be employed seamlessly by the end-user in the same manner as for an adult-based IUR estimate 21 derived from a rodent bioassay. Lack of published information on risks associated with Libby 22 Amphibole asbestos-specific exposure during childhood is the uncertainty associated with the 23 proposed extrapolation. If such information is subsequently published, the extrapolation 24 procedure can be updated. 25 26 5.4.6.2. Summary 27 In the discussion of the overall uncertainty in the IUR, it is important to distinguish 28 between uncertainty that encompasses both the direction and the magnitude from uncertainty 29 with known directional effects on the IUR but of unknown magnitude. In this summary, only the 30 latter uncertainties, which may result in underestimated or overestimated risk, are listed below. This document is a draftfor review purposes only and does not constitute Agency policy. 5-129 DRAFT--DO NOT CITE OR QUOTE 1 Uncertainties that are not thought to alter the estimated magnitude of the risk in a systematic 2 direction are not included in this summary. 3 The sources of uncertainty that could lead to a likely underestimation of the cancer risk 4 value include the following: 5 6 7 Use o f historical PCM exposure measurements. Because asbestos was a 8 component of vermiculite that was the primary object of production, mine and dry 9 and old wet mill ambient air may have contained material other than asbestos that 10 could have contributed to fibers counted by PCM. Therefore, it is possible that 11 exposure estimates for some, or possibly a large portion of the cohort, are 12 overestimated, and, therefore, the resulting IUR may be underestimated. 13 Measurement error in exposure assessment and assignment. This current 14 assessment showed that unit risk results from analysis of the lung-cancer 15 mortality in the full cohort (see Table 5-21) compared to the sub-cohort hired 16 after 1959 may have been attenuated as much as 2-6 times (see Section 17 5.4.6.1.2.4). By excluding those cohort members hired before 1960 for whom 18 there was insufficient work history information to estimate their exposures, the 19 unit risk for lung cancer was less attenuated due to exposure measurement error. 20 However, exposure measurements from the 1960s are also imperfect and include 21 a lesser degree of exposure measurement error, which could have led to 22 underestimated risk even in the sub-cohort hired after 1959. 23 Limited length o ffollow-up. Absolute risk is used for mesothelioma; therefore, 24 the unit risk of mesothelioma mortality could be larger than was estimated from 25 existing data, depending on the relationship between the number of additional 26 deaths and an increase in person-years. 27 Use o f life-tables to calculate the IUR based on cancer mortality. The 28 lung-cancer mortality unit risk based on the LECoi would be somewhat larger, 29 about 5-10%, and the mesothelioma unit risk would be slightly less (about 3%) 30 than that if the life-tables were extended from 85 to 90 years to account for longer 31 life spans. 32 Small number o f women and ovarian cancer. While asbestos is causally 33 associated with increased risks of ovarian cancer (Straif et aU 2009). there were 34 only 84 women in the whole cohort, and there were no deaths from ovarian cancer 35 among 24 total deaths. To the extent that there was an increased risk of ovarian 36 cancer in the Libby workers cohort due to inhalation exposures that was 37 unobserved, then the IUR would be somewhat underestimated. However, it was 38 not possible to estimate the magnitude of this underestimation on the total cancer 39 risk. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-130 DRAFT--DO NOT CITE OR QUOTE 1 Dependent competing risks. Competing risk of mortality from other diseases 2 related to exposure may have resulted in underestimates of the risk of mortality 3 from either mesothelioma or lung cancer. The mean length of follow-up for the 4 Libby workers who died of mesothelioma was to 30.1 years, and evidence exists 5 (Bianchi and Bianchi. 2009; Suzuki and Yuen. 2001) that early deaths from other 6 exposure-related causes could have precluded an individual's risk of death from 7 mesothelioma. However, it was not possible to estimate the magnitude of this 8 effect on the total cancer risk. 9 10 11 The sources of uncertainty that could lead to a likely overestimation of the cancer risk 12 value include the following: 13 14 15 Potential residual confounding and effect modification. The unit risk of 16 lung-cancer mortality estimated herein, and the combined mesothelioma and 17 lung-cancer mortality IUR, would over-estimate the risk in any population that 18 had a lower prevalence of smoking than that of the Libby worker cohort. Since 19 the Libby worker cohort had a large prevalence of smokers and ex-smokers and 20 no known nonsmokers developed lung cancer, it is also possible that estimated 21 risk for lung cancer is actually ri sk for an interaction of lung cancer and smoking, 22 and effects of smoking and asbestos are known to be between additive and 23 multiplicative (see Section 4). 24 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 5-131 DRAFT--DO NOT CITE OR QUOTE 1 6. MAJOR CONCLUSIONS IN THE CHARACTERIZATION OF HAZARD AND 2 EXPOSURE RESPONSE 3 Libby Amphibole asbestos,45present in vermiculite from the mine near Libby, MT, is a 4 complex mixture of amphibole fibers--both mineralogically and morphologically (see 5 Section 2.2). The mixture primarily includes tremolite, winchite, and richterite amphibole 6 minerals which exhibit a range of fiber morphologies (e.g., asbestiform, acicular, prismatic) 7 (Meeker et al.. 2003). Given the exposure potential to Libby Amphibole asbestos--and its 8 characteristic mineral composition--a hazard characterization and cancer exposure-response 9 assessment are presented. 10 As discussed in Section 1, there is currently no reference concentration (RfC) for 11 asbestos, and the U.S. Environmental Protection Agency (EPA) Integrated Risk Information 12 System (IRIS) inhalation unit risk (IUR) for asbestos is based on a synthesis of 14 epidemiologic 13 studies that included occupational exposure to chrysotile, amosite, or mixed mineral fibers 14 (chrysotile, amosite, and crocidolite) (U.S. EPA. 1988a). There is uncertainty in applying the 15 resulting IUR to environments and minerals that are not included in the studies considered for 16 the asbestos IUR derivation (U.S. EPA, 1988a). Published mortality studies on the Libby, MT 17 worker cohort have become available since the derivation of the IRIS asbestos IUR [i.e., 18 McDonald et al. (2004; 1986a); Amandus and Wheeler (1987); Sullivan (2007); Larson et al. 19 (2010b)]. This assessment documents noncancer and cancer health effects from inhalation 20 exposure to Libby Amphibole asbestos. Data are not available to support derivation of either a 21 reference dose (RfD) or a cancer oral slope factor (OSF) following oral exposures to Libby 22 Amphibole asbestos. 23 24 6.1. HUMAN HAZARD POTENTIAL 25 6.1.1. Exposure 26 Vermiculite ore mined near Libby, MT, contained Libby Amphibole asbestos, which 27 remained in the vermiculite concentrate (VC) and exfoliated product shipped from the facilities 28 (see Section 2). Vermiculite from the Libby, MT mine was used commercially from the 1920s to 45 The term "Libby Amphibole asbestos" is used in this document to identity the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-1 DRAFT--DO NOT CITE OR QUOTE 1 1990, and a review of company records available from (1964-1990) indicates approximately 2 6,109,000 tons of VC was shipped to over 200 facilities (ATSDR. 2008b). Vermiculite was 3 most notably used as attic insulation, a soil amender for gardening, and in the manufacturing of 4 gypsum wallboard. The exposure potential to Libby Amphibole asbestos includes historical 5 exposures (both occupational and community), as well as the potential for ongoing exposures to 6 waste materials, contaminated soils and vegetation, and consumer products (e.g., vermiculite 7 attic insulation; see Section 2.3) (ATSDR. 2008b. 2001b). 8 There are many ways in which workers and residents in Libby, MT, and the surrounding 9 communities may have been exposed while the mining and milling operations were active. 10 Historical routes of exposure include (1) occupational exposure; (2) take-home exposure for 11 household contacts of the workers (including children); (3) dust/fiber emissions to the 12 community from the milling and exfoliating facilities; (4) distribution of waste material into the 13 community as fill (including yards and recreational areas); (5) use of vermiculite attic insulation 14 in homes; (6) use of vermiculite in gardeni ng/hoiliculture; and (7) chil dren playing in the waste 15 stoner rock piles (Peipins et al,, 2003). Other than documentation of dust and fiber exposure 16 levels for mine and mill workers, there are few data to inform the levels of exposure to 17 household contacts and community members during mine and mill operations. Although no 18 historical exposure measurements are available from the homes of the workers, the EPA has 19 conducted sampling to determine exposure levels from vermiculite and waste materials that 20 remain in the community (U.S. EPA. 2006c; Weis. 2001a. b) (see Appendix B). These data 21 provide information useful to understand what historical exposures might have been for similar 22 activities. More recently, EPA has characterized exposures for various exposure pathways in the 23 community and continues to evaluate exposure potential in the ongoing efforts for cleanup (U.S. 24 EPA. 201 Oaf 25 Outside of Libby, MT, vermi culite concentrate and exfoli ated product was shipped to 26 271 domestic sites that served as processing facilities (U.S. GAO. 2007). These sites included 27 exfoliation plants (e.g., for the production of vermiculite insulation) as well as nonexfoliation 28 facilities (e.g., production of gypsum wallboard). The vermiculite concentrate was exfoliated by 29 heat-induced expansion resulting in vermiculite produced for commercial purposes. Both the 30 commercial vermiculite and the waste stoner rock (i.e., residual waste stoner rock from 31 exfoliation) contained Libby Amphibole asbestos fibers. Potential exposure routes in these This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-2 DRAFT--DO NOT CITE OR QUOTE 1 communities located around the country parallel the exposures in Libby, MT, including 2 occupational exposures, take-home exposures from workers, and children playing in the piles of 3 waste stoner rock near the facility (ATSDR.. 2008b. 2005b, 2003a). Waste materials (expanded 4 vermiculite and waste stoner rock) from some of these facilities were also used for fill in local 5 communities, potentially creating additional exposure pathways based on an Agency for Toxic 6 Substances and Disease Registry (ATSDR) review of 28 facilities, and a survey of the Western 7 Minerals Plant, MN (ATSDR. 2008b. 2003a). Few historical samples are available from these 8 facilities that could be used to quantify the exposure potential for workers or for the surrounding 9 communities (ATSDR. 2008b. 2005a, 2003a). Air modeling conducted for one exfoliating 10 facility in Minnesota does provide support for the potential of dust/fiber emissions from 11 exfoliating plants to impact ambient air quality in the vicinity of the plant (ATSDR. 2003a). 12 While the mine was active, there was potential exposure to commercial products 13 containing vermiculite from Libby, MT, especially in gardening soils and vermiculite attic 14 insulation. No studies have evaluated the potential for consumer exposure when vermiculite 15 from Libby, MT, was employed as a soil amender, but air sampling at one facility where this was 16 produced (O.M. Scott facility in Marysville, OH) demonstrated that workers handling this 17 material during manufacture were exposed to Libby Amphibole asbestos fibers (see 18 Section 5.2.3.1). There is potential for exposure in homes that contain vermiculite attic 19 insulation from Libby, MT, where residents and workers might enter attics for various uses, 20 repairs, and renovations (see Section 2.3.3). 21 22 6.1.2. Fiber Toxicokinetics 23 There is no specific informati on available on the fiber toxicokinetics of Libby Amphibole 24 asbestos. However, as a mineral fiber, the characteristics that define the deposition, clearance, 25 and translocation of other amphibole fibers might apply to Libby Amphibole asbestos. As 26 discussed in Section 3, the specific fiber dimensions and density of Libby Amphibole asbestos 27 will determine the probable pattern of deposition in the respiratory tract and other tissues (e.g., 28 pleura, peritoneum). Based on the fiber-size profile of airborne Libby Amphibole asbestos 29 fibers, deposition is expected throughout the respiratory tract including the alveolar regions. 30 Less is known about mineral fiber translocation to other target tissues in general, and, to date, no This document is a draftfor review purposes only and does not constitute Agency policy. 6-3 DRAFT--DO NOT CITE OR QUOTE 1 studies have specifically examined translocation following exposure to Libby Amphibole 2 asbestos. 3 As with other mineral fibers, clearance is likely to occur via the mucociliary apparatus in 4 the upper respiratory tract and the mucociliary escalator for those fibers deposited in the trachea 5 and bronchioles. This clearance is enhanced by macrophage action, which may transport some 6 of the fibers from the alveolar sac to the mucociliary system. Fibers may also be dissolved in 7 lung fluids or through the more aggressive action of alveolar macrophages. In general, 8 amphibole asbestos is considered more persistent and less likely to dissolve than other natural 9 mineral fibers, including serpentine asbestos (i.e., chrysotile) fibers. However, no data are 10 available for Libby Amphibole asbestos specifically, and it is unknown if Libby Amphibole 11 asbestos fibers would split or break in the pulmonary compartment as has been shown with some 12 amphibole fibers (e.g., ferroactinolite) (Coffin et al.. 19831. 13 Any fibers deposited in the respiratory tract and not cleared via the mucociliary system, 14 or not dissolved, can remain in the lung or can be transported to other tissues. Although data 15 specific to Libby Amphibole asbestos are not yet available, other asbestos fiber types can 16 translocate from the lung via macrophage action and transport through the lymph system, or 17 direct migration may occur through tissues from the mechanical action of the lung. Pleural and 18 peritoneal effects documented in Libby Amphibole asbestos-exposed individuals support the 19 potential for translocation of Libby Amphibole asbestos into the pleura. 20 21 6.1.3. Noncancer Health Effects in Humans and Laboratory Animals 22 Noncancer health effects identified in humans following inhalation exposure to Libby 23 Amphibole asbestos include pleural abnormalities, asbestosis, and reduced lung function as well 24 as increased mortality from noncancer causes. Two cohorts of workers exposed to Libby 25 Amphibole asbestos have been studied: workers at the mine and related operations in Libby, MT 26 and employees in the G.M. Scott plant in Marysville, OH, where the vermiculite product was 27 exfoliated and used as an inert carrier in lawn care products. Radiographic assessments of study 28 participants in both cohorts indicate radiographic abnormalities consistent with asbestos-related 29 disease, specifically pleural thickening (localized [LPT] and diffuse [DPT]) and small opacities 30 (indicative of interstitial fibrosis) (Rohs et aL 2008: Amandus et al... 1987b: McDonald et ah. 31 1986b: Lockev et ah. 1984). These studies provided quantitative exposure estimates and were This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-4 DRAFT--DO NOT CITE OR QUOTE 1 considered suitable for exposure-response analysis to support an RfC derivation. Additionally, 2 five cohort mortality studies of Libby, MT workers identified increased risk of mortality from 3 noncancer causes, including nonmalignant respiratory disease (e.g., asbestosis) (Larson et al.. 4 2010b; Sullivan, 2007; McDonald et al.. 2004; Aroandus and Wheeler, 1987; McDonald et al., 5 1986a) and cardiovascular disease (Larson et al.. 2010b). 6 ATSDR conducted health screening of community members in and around Libby, MT 7 (including past workers), and identified an increase in radiographic abnormalities with an 8 increased number of exposure pathways (Peipins et al., 2004a; Peipins et al., 2003; ATSDR. 9 2001b). Other researchers have also used these data to identify the increased prevalence of 10 respiratory symptoms in children (Vinikoor et al.. 2010) and to evaluate the prevalence of 11 radiographic abnormalities and reduced lung function in nonworker participants (Weill et al.. 12 2011). Radiographic abnormalities were more prevalent in mine/mill workers versus other 13 exposure categories (i.e., household contacts, dusty trades, and community-only exposures) 14 (Weill et al.. 2011). Pleural thickening (LPT or DPT) increased with age, within each exposure 15 group. Decreased pulmonary function (as percent of the predicted forced vital capacity) are 16 reported for participants with radiographic abnormalities (small opacities, DPT, and LPT) with 17 greater effects seen in participants with small opacities and DPT (Weill et al.. 2011). A nested 18 case-control study based on this study group also identified a potential for increased prevalence 19 of autoimmune disease (Pfau et al.. 2006). although additional research is needed to explore this 20 potential health outcome. 21 Although laboratory animal data and experimental data on toxicity mechanisms are 22 limited for Libby Amphibole asbestos, the existing data are consistent with the health effects 23 observed in both workers and community members exposed to Libby Amphibole asbestos. 24 Experimental animal studies have demonstrated increased collagen deposition consistent with 25 fibrosis following intratracheal instillati on of Libby Amphibole asbestos fibers in C57B16 mice 26 (Smartt et al.. 2010; Putnam et al.. 2008) and Fisher 344 rats (Padilla-Carlin et al.. 2011) as well 27 as increased markers of pulmonary inflammation in a rat model for human cardiovascular 28 disease (Shannahan et al.. 201 la; Shannahan et al.. 2011b). Pulmonary fibrosis, inflammation, 29 and granulomas were observed after tremolite, which comprises approximately 6% of the fiber 30 mixture in Libby Amphibole asbestos, inhalation exposure in specific-pathogen-free (SPF) male 31 Wistar rats (Bernstein et al.. 2005; Bernstein et al.. 2003). and intratracheal instillation in male This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-5 DRAFT--DO NOT CITE OR QUOTE 1 albino Swiss mice (Sahu et al., 1975). Davis et al. (1985) also reported pulmonary effects after 2 inhalation exposure to tremolite in SPF male Wistar rats including increases in peribronchiolar 3 fibrosis, alveolar wall thickening, and interstitial fibrosis. 4 5 6.1.4. Carcinogenicity in Humans and Laboratory Animals 6 There is convincing evidence of a causal association between exposure to Libby 7 Amphibole asbestos mesothelioma and lung cancer in workers from the Libby, MT vermiculite 8 mining and milling operations (Larson et al.. 2010b; Sullivan. 2007; McDonald et al... 2004; 9 Amandus et al . 1988; Amandus and Wheeler, J987: McDonald et al.. 1986a). No other 10 occupational cohort with exposures to Libby Amphibole asbestos has been studied with respect 11 to mortality risks. Whitehouse et al. (2008) documented 11 mesothelioma cases in nonworkers 12 exposed to Libby Amphibole asbestos in Libby, MT. Increased lung cancer and mesothelioma 13 deaths are also reported for worker cohorts exposed to other forms of amphibole fibers (amosite 14 and crocidolite) (de Klerk et al.. 1989; Seidman et al... 1986; Henderson and Enterline. 1979). 15 These findings are consistent with the increased cancers reported for communities exposed to 16 various rocks and soils containing tremolite fibers (Flasanoglu et al.., 2006; Slchletidis et al.. 17 1992; Baris et al.. 1987; Langer et al.. 1987; Baris et al. 1979; Yazicioglu. 1976). Although 18 potency, fiber dimension, and mineralogy differ between amphiboles, these studies are 19 supportive of the hazard identification of Libby Amphibole asbestos fibers described in this 20 assessment. 21 Although there is a limited laboratory animal database, the studies that are available 22 support the determi nation of carcinogenicity of Libby Amphibole asbestos fibers. Smith (1978) 23 demonstrated mesotheliomas in hamsters given a single intrapleural injection of Libby 24 Amphibole asbestos material (see Table 4-15). Tremolite is also carcinogenic in studies in rats, 25 hamsters, and mice, resulting in pleural mesothelioma, peritoneal mesothelioma, and lung cancer 26 depending on the route of exposure (see Table 4-16) (Bernstein et al.. 2005; Bernstein et al.. 27 2003; Roller et al... 1996; Davis et a.L 1991; Davis et al.. 1985; Stanton et al.. 1981). Although 28 comparing the potency of the tremolite used in these studies is difficult given the lack of 29 information on fiber characteri stics and other study limitations, these results demonstrate an 30 increased risk for lung cancer and mesothelioma following exposure to tremolite asbestos. 31 This document is a draftfor review purposes only and does not constitute Agency policy. 6-6 DRAFT--DO NOT CITE OR QUOTE 1 6.1.5. Susceptible Populations 2 Certain populations could be more susceptible than the general population to adverse 3 health effects from exposure to Libby Amphibole asbestos. In general, factors that may 4 contribute to increased susceptibility from environmental exposures include lifestage, gender, 5 race/ethnicity, genetic polymorphisms, health status, and lifestyle. However, little data exist to 6 address the potential of increased susceptibility to cancer or noncancer effects from exposure to 7 the Libby Amphibole asbestos. 8 Most occupational studies of workers exposed to Libby Amphibole asbestos have 9 examined the effects only in men because this group represents the vast majority of workers in 10 these settings (Moolgavkar et ah, 2010; Sullivan. 2007. McDonald et al., 2004; Amandus et al.. 11 1988; Amandus et al... 1987a; Amandus and Wheeler. 1987; Amandus et al.. 1987b; McDonald 12 et al.. 1986a; McDonald et al.. 1986b). The analysis presented here includes all workers, 13 however, there were few women in the cohort, and therefore no determination can be made 14 regarding increased susceptibility to lung cancer or mesothelioma by gender. Gender-related 15 differences in exposure patterns, physiology, and dose-response are some of the factors that may 16 contribute to gender-related differences in risk from asbestos exposure (Smith. 2002). The 17 limited data available from community-based studies (ATSDR. 2000) do not provide a basis for 18 drawing conclusions regarding gender-related differences in carcinogenic effects from Libby 19 Amphibole asbestos. Racial diversity among workers exposed to Libby Amphibole asbestos is 20 also limited, and data on ethnic groups are absent, precluding the ability to examine racial and 21 ethnicity-related differences in the mortality risks within the Libby, MT worker cohort. Finally, 22 the potential modifying effects of genetic polymorphisms, pre-existing health conditions, 23 nutritional status, and other lifestyle factors have not been studied sufficiently to determine their 24 potential contribution to variation in risk in the population. 25 26 6.1.6. Mode-of-Action Information 27 Due to the limited data that are available specific to Libby Amphibole asbestos, the mode 28 of action (MOA) of Libby Amphibole asbestos for lung cancer and mesothelioma following 29 inhalation exposure cannot be established. Laboratory animal studies of mice (Smartt et al.. 30 2010; Putnam et al.. 2008). hamsters (Smith. 1978) or rats (Padilla-Carlin et al.. 2011; 31 Shannahan et al... 2011a; Shannahan et al.. 201 lb) exposed to Libby Amphibole asbestos suggest This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-7 DRAFT--DO NOT CITE OR QUOTE 1 a similar type of inflammatory response to that observed with other mineral fibers; however, no 2 inhalation studies were available in the published literature. In vivo studies in rats, hamsters, or 3 mice exposed to tremolite (Roller et al.. 1997. 1996: Davis et al., 1991; McConnell et ai... 1983b; 4 Wagner et af. 1982; Stanton et al... 1.981; Smith et al.. 19791 show results similar to other 5 amphibole asbestos fibers including lung cancer and mesothelioma, with limited inhalation 6 studies (Bernstein et al., 2005; Bernstein et al., 2003; Davis et al., 19851. In vitro studies 7 demonstrate that the uptake of Libby Amphibole asbestos fibers by macrophage, mesothelial, 8 and lung epithelial cell lines may lead to an increase in oxidative stress as measured by reactive 9 oxygen species production, gene expression changes or genotoxicity (Hillegass et al.. 2010; 10 Pietrnska et al.. 2010; Blake et al., 20071. Thus, the available data indicate that Libby 11 Amphibole asbestos induces biological responses similar to other forms of asbestos such as 12 oxidative stress, chronic inflammation, genotoxicity, and increased cell proliferation. These 13 biological effects following exposure to Libby Amphibole asbestos and/or tremolite are 14 demonstrated in a limited number of laboratory animal and in vitro studies. Multiple key events 15 for one particular toxicity pathway or MOA have not been identified and adequately supported; 16 therefore, the MOA for Libby Amphibole asbestos carcinogenicity cannot be established. 17 18 6.1.7. Weight-of-Evidence Descriptor for Cancer Hazard 19 Under the EPA Guidelinesfor Carcinogen Risk Assessment (U.S. EPA, 2005a), Libby 20 Amphibole asbestos is carcinogenic to humans following inhalation exposure based on 21 epidemiologic evidence that shows convincing evidence of a causal association between 22 exposure to Libby Amphibole asbestos fibers and increased lung cancer and mesothelioma 23 mortality (Larson et al.. 2010a; Moolgavkar et al.. 2010; Sullivan, 2007; McDonald et al., 20Q4; 24 Amandus and Wheeler. 1987; McDonald et al., 1986a). These results are further supported by 25 animal studies that demonstrate the carcinogenic potential of Libby Amphibole asbestos fibers 26 and tremolite fibers in rodent bioassays. As a durable mineral fiber of respirable size, this 27 conclusion is consistent with the extensive published literature that documents the 28 carcinogenicity of amphibole fibers. 29 U.S. EPA's Guidelinesfor Carcinogenic Risk Assessment (U.S. EPA, 2005a) indicate 30 that for tumors occurring at a site other than the initial point of contact, the weight of evidence 31 for carcinogenic potential may apply to all routes of exposure that have not been adequately This document is a draftfor review purposes only and does not constitute Agency policy. 6-8 DRAFT--DO NOT CITE OR QUOTE 1 tested at sufficient doses. An exception occurs when there is convincing information (e.g., 2 toxicokinetic data) that absorption does not occur by other routes. Information on the 3 carcinogenic effects of Libby Amphibole asbestos via the oral and dermal routes in humans or 4 animals is absent. The increased risk of lung cancer and mesothelioma following inhalation 5 exposure to Libby Amphibole asbestos has been established by studies in humans, but these 6 studies do not provide a basis for determining the risk from other routes of exposure. 7 Mesothelioma occurs in the pleural and peritoneal cavities and, therefore, is not consi dered a 8 portal-of-entry effect. However, the role of indirect or direct interaction of asbestos fibers with 9 tissues at extrapulmonary sites is still unknown. There is no information on the translocation of 10 Libby Amphibole asbestos to extrapulmonary tissues following either oral or dermal exposure, 11 and limited studies have examined the role of these routes of exposure in cancer. Therefore, 12 Libby Amphibole asbestos is considered carcinogenic to humans by the inhalation route of 13 exposure. 14 15 6.2. EXPOSURE RESPONSE 16 This assessment contains a derivation of an RfC for noncancer effects and an IUR for 17 cancer based on epidemiologic data. It does not contain an RID or OSF. 18 19 6.2.1. Noncancer/Inhalation 20 Of the observed noncancer health effects from exposure to Libby Amphibole asbestos, 21 data that provide exposure-response information are only available for increased pleural 22 thickening (localized and diffuse) and signs of interstitial fibrosis (i.e., small opacities) in the two 23 worker cohorts (i.e., Libby worker cohort and Marysville worker cohort). Both cohorts provide 24 individual exposure estimates, and document increased hazard of pleural and parenchymal 25 effects. As detailed in Section 5.2.1, each of the available studies has strengths and weaknesses. 26 The cohort of Marysville, OH workers [Lockey et al. (1984) and the follow-up by Rohs et al. 27 (2008)1was selected as the principal cohort over the Libby worker cohort for five reasons: 28 (1) lack of confounding by residential and community exposure; (2) information on important 29 covariates (e.g., BMI); (3) exposure-response relationship defined for lower cumulative exposure 30 levels (in the post-1972 sub-cohort); (4) adequate length of follow-up; and (5) use of more recent 31 criteria for evaluating radiographs (ILCX 2002) (see Section 5.2.1 for details). Of the observed This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-9 DRAFT--DO NOT CITE OR QUOTE 1 radiographic abnormalities in exposed workers, localized pleural thickening (LPT) was selected 2 as the critical effect due to its higher prevalence relative to the other outcomes, minimal 3 adversity (compared with other effects), and specificity for durable mineral fiber exposure. LPT 4 is an irreversible pathological change associated with constricting chest pain, dyspnea, and 5 decreased pulmonary function and considered adverse (see Section 5.2.1.4). For an RfC 6 derivation, analyses focused on the cohort of Marysville, OH workers described by Rohs et al. 7 (2008). Specifically, the RfC was derived from the sub-cohort of the Marysville, OH workers 8 who started employment after 1972, due to the greater certainty in exposure assessment in this 9 group. 10 Benchmark dose (BMC) modeling, with a benchmark response of 10% extra risk, was 11 used to derive the point of departure (POD). A Michaelis-Menten regression model was the 12 best-fitting model for the sub-cohort and used to estimate the exposure-response relationship for 13 Libby Amphibole asbestos and LPT. Cumulative exposure with a lag of 10 years was selected as 14 the exposure metric, based on evidence for biological latency and model fit considerations. A 15 background rate of LPT of 1% was assumed based on a limited number of published estimates. 16 The resulting BMCio under these modeling assumptions was 0.2642 fibers/cc-year; the 17 corresponding lower 95% confidence limit of the BMCio (BMCLio) is 0.1177 fibers/cc-year as a 18 cumulative lifetime exposure. The RfC is for continuous exposure (i.e., 24 hours/day, 19 365 days/year, with exposure beginning at birth and continuing for 70 years). Thus, the modeled 20 BMCLio as CE was adjusted to 70 years of exposure, lagged by 10 years (non-occupalionai. 21 lifetime exposure) resulting in a value of 60 years (see Section 5.2.4). 22 23 24 POD = BMCLio + (lifetime exposure duration) 25 = [0 .1 1 7 7 (fibers/cc) year] = [70 - 10 years] 26 = 1 .9 6 * 10 5 fibers/cc 27 28 29 The RfC is obtained by applications of uncertainty factors as needed. Two uncertainty 30 factors (UF) have been applied for a composite UF of 100 (intraspecies uncertainty factor, 31 UFa = 10; database uncertainty factor, UFD. = 10) (see Section 5.2.4). As shown below, the This document is a draftfor review purposes only and does not constitute Agency policy. 6-10 DRAFT--DO NOT CITE OR QUOTE 1 chronic RfC is 2 x 10 5fibers/cc for Libby Amphibole asbestos; it was calculated by dividing the 2 lifetime-POD by a composite UF of 100: 3 4 5 Chronic RfC = POD + UF 6 = 1.96 * 10 ' fibers/cc - 100 7 = 1.96 k 10 Nfibers/cc, rounded to 2 * 10~ fibers/cc 8 9 10 Modeling was also conducted in the full cohort of workers described in Lockey et al. 11 (1984) and Rohs et al. (2008). These analyses used a different modeling approach, due to the 12 wider range of exposures and time from first exposure. A modified Michaelis-Menten model 13 provided the best fit to the full cohort data, which incorporated time from first exposure via the 14 plateau term for the model. For a time from first exposure of 30 years and exposure lag of 15 10 years, the BMC and BMCL corresponding to a 10% extra risk of LPT were 0.1477 and 16 0.0580 fibers/cc-year, respectively. This BMC and BMCL are quite similar to the values 17 obtained in the analysis for the RfC and provide important support for the selected modeling 18 approach. When time from first exposure is set at 40 years, the calculated RfC is 19 4 x 10 6 fibers/cc. 20 Confidence in the principal study is considered medium. The data used are human, 21 epidemiological data which are preferred to animal bioassays, and the principal study is 22 conducted in a population of occupationally exposed workers with long-term, relatively low 23 intensity exposures. Flowever, use of the sub-cohort resulted in a smaller data set, and fewer 24 cases to model. Additionally there are weaknesses in the primary study. Exposure estimates are 25 based on self-reported job histories. The study used a cross-sectional design and may be 26 negatively biased as individuals with more severe disease could have left employment or may 27 have died and not been included in the follow-up study, resulting in an underestimation of 28 overall toxicity. Flowever, for a less severe effect, such as LPT, this bias should be minimal. As 29 discussed in Sections 4.1.3 and 5.2.1.3.2, there may have been potential for selection bias due to 30 exposure-dependent censoring in this population, based on information provided by Rohs et al. 31 (2008) regarding the higher average exposure in participants compared to nonparticipants. In 32 terms of sensitivity of the study to detect a health effect, it is known that high-resolution This document is a draftfor review purposes only and does not constitute Agency policy. 6-11 DRAFT--DO NOT CITE OR QUOTE 1 computed tomography can identify mineral fiber-related lesions in the respiratory tract, which 2 cannot be identified by standard radiographs (Muravov et al.. 2005; ATS, 2004; Staples et ah, 3 1989). Thus, the technology employed for determining the prevalence of radiographic changes 4 in the Marysville cohort may underestimate the actual prevalence of localized pleural thickening. 5 Confidence in the database is low-to-medium. The database contains long-term mortality 6 and morbidity studies in humans exposed via inhalation to Libby Amphibole asbestos. The 7 morbidity studies do provide appropriate data for RfC derivation for pleural and lung 8 abnormalities. However, although decreased pulmonary function, a potential for autoimmune 9 effects, and cardiovascular disease are noted in exposed individuals, data do not provide an 10 exposure-response relationship. It is known that inhaled asbestos fibers migrate out of the lung 11 and into other tissues (see Section 3.1), lending uncertainty to any assumptions that other effects 12 would not be expected. There are no data in laboratory animals or humans on general systemic 13 effects. Therefore, overall confidence in the RfC is low-to-medium, reflecting medium 14 confidence in the principal study and low-to-medium confidence in the database. 15 Uncertainty and Sensitivity Analysesfo r RfC Derivation: It is important to consider the 16 sources of uncertainties in the derivation of the RfC for Libby Amphibole asbestos. These 17 include the following: 18 Measurement error in exposure assessment and assignment. The estimated exposure for 19 each individual relied on self-reported employment history, which may be subject to recall error. 20 Only data from 1972 and later were used for an RfC derivation, based on lack of fiber 21 measurements prior to this date; although better there remains some uncertainty in exposures 22 prior to installation of IH controls (1974). There is also uncertainty in the post-1972 data 23 regarding asbestos content in other ore sources (Virginia, South Carolina, and South Africa). 24 Although Libby Amphibole asbestos was not used in the facility after 1980, industrial hygiene 25 measurements collected after 1980 showed low levels of fibers. However, because the 26 concentration of fibers in the workplace was near background after 1980, this exposure makes 27 only a small contribution to an individual's cumulative exposure estimate. Similarly, any 28 exposure to Libby Amphibole asbestos outside of the workplace is not likely to contribute 29 significantly to cumulative exposure---- 10% of workers reported bringing raw veimiculite 30 home, and the majority showered and changed clothes before leaving the workplace. This document is a draftfor review purposes only and does not constitute Agency policy. 6-12 DRAFT--DO NOT CITE OR QUOTE 1 Radiographic assessment o f localizedpleural thickening. Conventional radiographs-- 2 rather than the more sensitive high-resolution computed tomography--were used to determine 3 the health outcome. Localized pleural thickening may be difficult to detect on these radiographs, 4 leading to the potential for outcome misclassification. However, uncertainty in the detection of 5 LPT in each individual is considered minimal due to the use of a team of highly qualified chest 6 radiologists evaluating the radiographic films and the use of consensus diagnosis. 7 Length offollow-up. Time from first exposure to X-ray was 23.2-32.7 years in the 8 preferred sub-cohort (mean of 28.2 years). The literature suggests that the prevalence of LPT 9 may increase with time, beyond this observed range of time from first exposure. The lack of 10 observed data beyond ~30 years after first exposure (on average) is a source of uncertainty when 11 characterizing the exposure-response relationship for a full lifetime of exposure (e.g., 70 years). 12 This likelihood that the prevalence of localized pleural thickening may increase further with time 13 beyond 30 years after first exposure, and lack of data to support characterization of the 14 exposure-response curve outside this range, is a principal rationale cited for the selection of a 15 database UF of 10 for an RfC derivation. 16 Background rate o f localizedpleural thickening. In the derivation of the RfC, a 17 background rate of 1% for LPT was used. Previous studies have reported a range of prevalence 18 estimates (0.02 to 3.9%) in populations not known to be occupationally exposed to asbestos. 19 However, in statistical modeling of the Marysville, OH sub-cohort, uncertainty in the 20 background rate of localized pleural thickening is very low. The difference in the POD when the 21 background rate is fixed at 1% versus when it is estimated (estimated background rate of 3.12%) 22 is -15% (0.1177 compared to 0.1349 fibers-year/cc), and it does not affect the proposed RfC 23 (after rounding to one significant digit). 24 Model Form. A number of model forms were explored in the initial stages of analysis 25 (see Appendix E) before selecting the Michaelis-Menten model. The BMC and the BMCL 26 estimated from other candidate models for the sub-cohort, as well as those obtained in modeling 27 from the full cohort were in a similar range to the selected model. A second model-based 28 uncertainty is the choice of lag for cumulative exposure. The RfC derivation is based on the 29 exposure lagged by 10 years, since this lag yielded the lowest Akaike Information Criterion 30 (AIC) value, and indication of superior fit. However, if other lags (with similar AICs) are used, 31 the difference in POD may fluctuate to be approximately 20% higher or approximately 55% This document is a draftfor review purposes only and does not constitute Agency policy. 6-13 DRAFT--DO NOT CITE OR QUOTE 1 lower. However, the choice of lag does not affect the proposed RfC (after rounding to one 2 significant digit). 3 Effect o f smoking. Information on ever/never smoking was available for the preferred 4 sub-cohort. This individual variable did not meet statistical significance in the best-fitting 5 model, although inclusion did improve model fit (see Appendix E). When including smoking in 6 the best-fitting model, BMCs and BMCLs estimated separately for smokers and nonsmokers 7 differed by approximately sixfold. Smoking was not included in the model selected for RfC 8 derivation due to the lack of statistical significance, limited sample size (only three cases were 9 never smokers out of a total of 12 cases), and lack of detailed information on smoking history, 10 but these sensitivity analyses indicate a need for further research on the effect of smoking in 11 relation to LPT risk among asbestos-exposed populations. 12 Sensitivity analysisfor the derivation o f a PODfor lifetime exposurefrom the CE metric 13 o f the worker cohort. Exposure-response modeling for LPT in the Marysville sub-cohort used 14 the cumulative exposure (CE) metric (represented as CHEEC, described in Section 5.2.3.1) 15 providing a POD in fibers/cc-years. In order to derive an RfC in the units of continuous air 16 concentration for a lifetime (i.e., fibers/cc), the POD from the CE metric was weighted across a 17 lifetime exposure. Thus, the lifetime BMCLio is 1.96 x 1CT3 [0.1177 (fibers/cc)-years ^ 18 60years], This procedure is one way to account for the duration of exposure in the occupational 19 study being less than lifetime. There is some uncertainty as to whether--and how--to take 20 account for less-than-lifetime exposure in the occupational cohort. A sensitivity analysis was 21 done to consider other procedures for this averaging. The primary analysis assumes duration 22 contributes to risk and thus calculates a concentration across a lifetime that would yield the POD. 23 The second analysis is consistent with assuming duration contributes to risk but estimating the 24 concentration only for the mean duration in the observed database. The third analysis assumes 25 duration does not contribute to risk and models the average work duration continuous exposure 26 equivalent for each worker. This sensitivity analysis indicates that the approach taken to average 27 the POD based on the CE metric (CHEEC) across a lifetime was a reasonable approach, as 28 similar results are obtained using different approaches (i.e., within 4 fold). 29 Choice o f critical effect. The critical effect selected for RfC derivation is localized 30 pleural thickening. Alternative endpoints were not modeled using the preferred sub-cohort due 31 to small numbers--there were five cases of bilateral LPT, only one case of diffuse pleural This document is a draftfor review purposes only and does not constitute Agency policy. 6-14 DRAFT--DO NOT CITE OR QUOTE 1 thickening, and no individuals with interstitial changes. As a sensitivity analysis, these three 2 alternative endpoints (along with all LPT) were modeled among all Marysville workers not 3 previously exposed to other forms of asbestos, with X-rays performed in 2002-2005 (n = 250). 4 These analyses were performed using the Michaelis-Menten model with a background rate of 1% 5 and unlagged CHEEC as the exposure metric. BMRs of 1, 5, and 10% were investigated (see 6 Table 5-5). Use of the 10% BMR for these alternative endpoints allows for comparison to a 7 POD based on the selected critical effect of LPT. In this larger cohort, the POD for a 8 10% increase in LPT was 0.06 fibers/cc-years (in comparison with the POD derived from the 9 sub-cohort and used in RfC derivation of 0.1177 fibers/cc-years). Results for all pleural 10 thickening (LPT and DPT) did not differ from results for LPT. Bilateral localized pleural 11 thickening was included as a rough indication of increased severity within LPT, and as expected 12 results in higher PODs at each BMR than LPT. The resulting BMCLs for DPT and small 13 opacities (1.17 and 2.89 fibers/cc-years respectively, 10% BMR) are higher than the POD for 14 LPT (0.06 fibers/cc-years). Thus, use of an alternative endpoint at the same BMR would provide 15 a higher POD, and corresponding higher RfC. 16 However, a 10% BMR is not appropriate for more severe endpoints and BMCLs are 17 calculated at 1 and 5% BMRs as well. If DPT is used as a critical effect, PODs of 0.081 and 18 0.473 fibers/cc-years would be calculated for a 1% and 5% BMR respectively. If small opacities 19 are used as a critical effect, the PODs are higher at both a 1% and a 5% BMR (i.e., 0.243 and 20 1.32, respectively). In summary, the use of more severe alternative endpoints (with appropriate 21 BMRs) results in PODs higher than that estimated using the critical effect of LPT (i.e., 22 0.06 fibers/cc-year, BMR 10%), and all are higher than the POD used in RfC derviation, with the 23 exception of DPT at a 1% BMR (0.0814 fibers/cc-year). BMCLs for these more severe 24 endpoints using a 1% BMR were within ~2-fold of the preferred POD (0.0814 and 25 0.2425 fibers/cc-year for diffuse pleural thickening and interstitial changes, respectively). There 26 is uncertainty associated with these estimates due to the inclusion of individuals hired before 27 1972, when no quantitative exposure measurements were available. Thus, a choice of alternative 28 critical effects--even with lower BMRs--would not result in an RfC appreciably lower than the 29 proposed RfC based on LPT and a 10% BMR. 30 This document is a draftfor review purposes only and does not constitute Agency policy. 6-15 DRAFT--DO NOT CITE OR QUOTE 1 6.2.2. Cancer/lnhalation 2 6.2.2.1. Background and Methods 3 The most appropriate data set for deriving quantitative cancer risk estimates based on 4 Libby Amphibole asbestos exposure in humans is the cohort of workers employed at the 5 vermiculite mining and milling operation near Libby, MT (see Section 4.1). No data were 6 available pertaining to cancer incidence or mortality in the Marysville, OH cohort, and mortality 7 and exposure data for other populations exposed to Libby Amphibole asbestos are very limited. 8 Whitehouse et al. (2008) provided detailed information on 11 mesothelioma cases among 9 nonworkers, but this information could not be used in exposure-response analyses for this 10 assessment, because there is no quantitative exposure information for these cases and no 11 information on the population from which these cases arose. 12 The Libby, MT worker cohort has been the focus of two epidemiologic investigati ons by 13 National Institute for Occupational Safety and Health (NIOSH) scientists. A database created by 14 NIOSH in the 1980s contains demographic data, work history, and vital status at the end of May 15 of 1982 for 1,881 workers at the vermiculite mine, mill, and processing plant in Libby, MT (see 16 Section 4.1.1.1). Vital status follow-up was completed by NIOSH through 2006 using the 17 National Death Index (Bilgrad, 1997). Nearly 54% of workers in the cohort {n = 1,009) had died 18 by December 31, 2006. The data from this update (provided by NIOSH) is the basis of the EPA 19 exposure-response modeling. 20 EPA does not have sufficient biological information to select models for the 21 epi demiology data on the basis of bi ological mechanism (see Section 5). In this situation, EPA's 22 practice is to investigate a range of model forms to determine how to best empirically model the 23 exposure-response relationship in the range of the observed data. In this case, different exposure 24 metrics were explored for model fit in the analytic models. The exposure metric options were 25 selected to provide a range of shapes that was suffici ently flexible to allow for a variety of ways 26 that time and duration might relate to cancer risk in the data being modeled. EPA then evaluated 27 how well the models and exposure metric combinations fit the data being modeled. Metrics that 28 did not fit the data well were rejected. For purposes of calculating a reasonable upper-bound on 29 the risk per exposure EPA accounted for uncertainty in the choice of exposure metrics by using 30 the exposure metric (among those of reasonable fit) that estimated the highest risk. This is 31 explained in more detail below and in Sections 5.4.3-5.4.5. However, there are other This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-16 DRAFT--DO NOT CITE OR QUOTE 1 uncertainties in the modeling of the epidemiological data that may impact the IUR and these are 2 described in detail in Section 5.4.6. 3 Cumulative exposure has been the traditional method of measuring exposure in 4 epidemiologic analyses of many different occupational and environmental exposures and was the 5 exposure metric applied for to the risk of lung-cancer mortality in the EPA general asbestos 6 evaluation (U S. EPA. 1988a). Two alternative approaches to developing exposure metrics to 7 describe the effects of air concentrations of asbestos dust in the air on the risks of mortality have 8 also been proposed. The first alternative was proposed by Jahr (1974) who studied 9 silica-induced pneumoconiosis. Hie also suggested that exposures to occupational dusts could be 10 weighted by the time since exposure yielding an exposure metric which gives greater weight to 11 earlier exposures. Berry et al. (1979) subsequently suggested the application of exposure metrics 12 that allowed for the clearance of dust or fibers by using a decay term on exposures. For the 13 evaluation of mortality risk from mesothelioma, U.S. EPA (1988a) used a different exposure 14 metric than was used for lung-cancer mortality, which factored in the time since first exposure. 15 It is important to note that different characterizations of ambient exposures may be reasonably 16 expected to be associated with different endpoints (i.e., lung cancer or mesothelioma). 17 In the Libby, MT worker cohort data developed by NIOSH and used by the EPA in this 18 assessment, detailed work histories, together with job-specific exposure estimates, allowed for 19 the reconstruction of each individual's occupational exposure experience over time to define 20 multiple exposure metrics. From this information-rich individual-level data set from NIOSH, the 21 EPA constructed a suite of the different metri cs of occupational exposure which had been 22 proposed in the asbestos literature or used in the EPA health assessment on general asbestos 23 exposures (U.S. EPA, 1988a). This suite of models was defined a priori to encompass a 24 reasonable set of proposed exposure metrics to allow sufficient flexibility in model fit to these 25 data. These exposure metrics were evaluated in analytic-regression models to test which 26 exposure metrics were the best empirical predictors of observed cancer mortality and the better 27 fitting models were advanced for consideration as the basis of the exposure-response relationship 28 for the IUR. The types of exposure metrics evaluated were intended to allow for variations of 29 the classic metric of cumulative exposure, allowing for more or less weight to be placed on 30 earlier or later exposures. These simulated exposure metrics were derived mathematically to 31 approximate underlying processes that are not well understood, and their fit is evaluated on the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-17 DRAFT--DO NOT CITE OR QUOTE 1 basis of maximizing the likelihood for the workers cohort and estimated parameters does not 2 necessarily have biological interpretation (see Section 5.4.2.5 for details). 3 Exposure estimates for all exposure metrics were adjusted to account for the time period 4 between the onset of cancer and mortality. The lag period defines an interval before death, or 5 end of follow-up, during which, any exposure is excluded from the calculation of the exposure 6 metric. Modeling of mesothelioma mortality included two additional exposure metrics: duration 7 of exposure and the exposure metric including a cubic function of time (see Eq. 5-5), originally 8 proposed in Peto et al. (1982) and employed in derivation of the IUR for asbestos (U S. EPA. 9 1988a. 1986a). 10 Analyses of mesothelioma mortality were conducted using a Poisson model with a 11 Markov chain Monte Carlo (MCMC) Bayesian approach, whereas analyses of lung-cancer 12 mortality were conducted using the Cox proportional hazards model with time-varying 13 exposures. There was one important limitation of the NIQSH job exposure matrix (JEM). Of 14 the 991 workers hired before 1960, 706 workers with unknown department code and unknown 15 job assignments hired between 1935 and 1959 were assigned the same average estimated 16 exposure intensity. The lack of information on specific job assignments for such a large portion 17 of these early workers when exposures were higher resulted in the misclassification of the 18 exposure and effectively yielded exposure metrics that were differentiated only by the duration 19 of each worker's employment. For this reason and because there was little measured fiber 20 exposure data during the earlier period, identifying an adequate exposure-response model fit was 21 unsuccessful. The two biggest problems were that the duration of employment was the 22 best-fitting metric for modeling mesothelioma and that the Cox model assumptions were violated 23 in modeling lung-cancer mortality (see Section 5.4.3.5). As a result, this assessment developed a 24 sub-cohort analysis by dividing the whole cohort into two groups: those hired prior to 1960 and 25 those hired after 1959. This removed all but nine cohort members with missing department code 26 and job category information and lessened the effect of estimates of early exposures where no air 27 sampling data were available. For the sub-cohort of those hired after 1959, those two biggest 28 problems were resolved: the assumptions of the Cox model were satisfied, and a lagged 29 cumulative exposure with a decay (rather than duration of exposure, as for the full cohort) was 30 the best-fitting metric for mesothelioma. This document is a draftfor review purposes only and does not constitute Agency policy. 6-18 DRAFT--DO NOT CITE OR QUOTE 1 Of the 880 workers hired after 1959, 230 (26%) had died by December 31, 2006. The 2 number of mesothelioma deaths in the sub-cohort is relatively small (n = 1,2 deaths coded in 3 ICD-10 and 5 deaths coded in ICD-9), but the rate of mesothelioma mortality was very similar in 4 the subcohort (24.7 per 100,000 person-years vs. 26.8 per 100,000 person-years for the full 5 cohort [18 mesothelioma deaths], a difference of less than 10%). 6 7 6.2.3. Modeling of Mesothelioma Exposure Response 8 A Poisson model is employed for estimating the absolute risk of mesothelioma following 9 exposure to Libby Amphibole asbestos, as the Poisson distribution is an appropriate model for 10 use with data that are counts of a relatively rare outcome, such as observed mesothelioma deaths 11 in the Libby, MT worker cohort. Estimation of the exposure-response relationship for 12 mesothelioma using the Poisson model was performed in WinBUGS software by a MCMC 13 Bayesian approach with an uninformative (diffuse) prior. The model was run to fit the mortality 14 data to exposure data for various exposure metrics described above. To comparatively evaluate 15 how much better one model fits than another, the Deviance Information Criterion (DIC) was 16 used. DIC is used in Bayesian analysis and is an analogue of AIC (Burnham and Anderson. 17 2002). Use of the DIC and AIC is standard practice in comparing the fit of nonnested models to 18 the same data set with the same dependent outcome variable but different independent 19 covariates. 20 Two cumulative exposure metrics with decay provided the best model fits. Both metrics 21 had a common 5-year half life, with lag times of either 10 or 15 years. In the sub-cohort hired 22 after 1959, the DIC value for mesothelioma using the IRIS IUR (U.S. EPA. 1988a) metric (see 23 Eq. 5-5) is substantially higher (DIC = 98.4) than for any of the metrics in Table 5-10, where the 24 lowest DIC is 70.6. This difference of over 20 DIC units, is an indication that the model used for 25 mesothelioma in the U.S. EPA (1988a) IUR derivation (see Eq. 5-5), does not fit these data from 26 the Libby, MT work cohort, compared to other exposure metrics presented (see Table 5-10). It 27 should be noted that the data modeled here are very different from the data on which the IRIS 28 assessment for asbestos (U.S. EPA. 1988a) is based--and one does not necessarily expect the 29 same model to fit different data sets--this is why EPA goes through a process to determine the 30 best-fitting model in each case. One difference with the IRIS IUR (U.S. EPA. 1988a) modeling 31 is that the analysis in this assessment is based on individual-level data, whereas the IRIS IUR This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-19 DRAFT--DO NOT CITE OR QUOTE 1 (U.S. EPA, 1988a) application was to aggregate data. Also, cohorts used in the IRIS IUR (U.S. 2 EPA. 1988a) did not include cohorts exposed to Libby Amphibole asbestos and Libby 3 Amphibole asbestos may be different from other types of asbestos. Alternately, the relative fit of 4 this model may have been affected by uncertainties in the estimated exposure described in detail 5 in Section 5.4.6. 6 As it is less likely that exposure during the last few years before death were contributory 7 to the development of the cancer and cancer mortality, the zero lag metrics were dropped from 8 further consideration. All eight models retained for derivation of IUR include a decay half-life in 9 the exposure metric. For the sub-cohort hired after 1959, the best-fitting exposure metric was 10 cumulative exposure with a 5 year half-life and a 15 year lag time with a central estimate for the 11 P of 2.07 x 10 4with 95% upper confidence limit (UCL) of 3.42 x 10 '. 12 13 6.2.4. Unit Risk Estimates for Mesothelioma Mortality 14 The increased risk of mesothelioma mortality attributable to continuous fiber exposure 15 was estimated using a life-table procedure based on the general U.S. population. The life-table 16 procedure involved the application of the estimated Libby Amphibole asbestos toxicity to a 17 structured representation of the general U.S. population in such a manner as to yield age-specific 18 risk estimates for cancer mortality in the presence or absence of exposure to Libby Amphibole 19 asbestos (see Section 5.4.5; Appendix G). 20 A default linear low-dose extrapolation below the POD was used because the mode of 21 action by which Libby Amphibole asbestos causes mesothelioma cannot be established. The 22 lower limit on the effective concentration (LECoi) for adult-only exposures was determined to be 23 0.245 fibers/cc, which yielded an adult-based unit risk for mesothelioma mortality of 0.053 per 24 fiber/cc (POD of 1% divided by the LECoi). 25 The value of the effective concentration (E C ) that would correspond to the measure of 26 central tendency is the ECoi- This value is used in the derivation of a combined risk of 27 mesothelioma and of lung cancer. The ECoi was determined to be 0.406 per fiber/cc, which 28 when divided into a POD of 1% and scaled (by 70/54) to encompass the whole lifespan, gives a 29 lifetime central estimate value of 0.032 per fiber/cc. 30 For mesothelioma, the undercounting of cases (underascertainment) is a particular 31 concern given the limitations of the International Classification of Diseases (ICD) classification This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-20 DRAFT--DO NOT CITE OR QUOTE 1 systems used prior to 1999. In practical terms, this means that some true occurrences of 2 mortality due to mesothelioma are missed on death certificates and in almost all administrative 3 databases such as the National Death Index. Even after introduction of special ICD code for 4 mesothelioma with introduction of ICD-10 in 1999, detection rates are still imperfect (Camidye 5 et al.. 2006; Pinheiro et al.. 20041 and the reported numbers of cases typically reflect an 6 undercount of the true number. Kopylev et al. (2011) reviewed the literature on this 7 underascertainm ent and developed methods to account for the likely numbers of undocumented 8 mesothelioma deaths. 9 To compensate for mesothelioma underascertainment attributable to ICD coding, the 10 mesothelioma mortality unit risk was further adjusted following the analysis of Kopylev et al. 11 (2011). The adjusted mesothelioma central (i.e., maximum likelihood estimate) risk, 12 corresponding to the best-fit metric, was 0.044 per fiber/cc, and the adjusted mesothelioma 13 mortality unit risk was 0.074 per fiber/cc. The adjusted mesothelioma mortality unit risks from 14 all eight exposure parameterization models with adequate fit produced a range of unit risk values 15 (see Table 5-17) from 0.044 to 0.122 Thus, there is uncertainty in mesothelioma risks generated 16 from similar-fitting models from different exposure metrics (see details in Section 5.4.6.1.3). 17 18 6.2.5. Modeling of Lung Cancer Exposure Response 19 All multivariate extended Cox model s were fit to the sub-cohort hired after 1959 with 20 covariates for sex, race, and date of birth, and exposure. Exposure for each of the 40 exposure 21 parameterizations was calculated independently and fit of these exposure metrics was evaluated 22 one at a time. As the exposure-response models cannot strictly be considered to be nested, a 23 standard measure of fit, the AIC (Burnham and Anderson. 2002). was used for comparison of 24 model fit with smaller values of AIC, indicating better goodness of fit. Of the 25 40 exposure-response metrics, 14 demonstrated an adequate fit to the data as measured by the 26 overall model fit with the likelihood ratio test (p < 0.05) as well as having statistically significant 27 exposure metrics (p < 0.05). However, only the nine models that demonstrated adequate model 28 and exposure metric fit and incorporated a lag period to account for cancer latency were 29 considered further in the development of the IUR (see Table 5-18). 30 Lagging exposure by 10 years was a better predictor of lung-cancer mortality compared 31 to other lags. As it is less likely that exposure during the last few years before death were This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-21 DRAFT--DO NOT CITE OR QUOTE 1 contributory to the development of the cancer and cancer mortality, the zero lag metrics were 2 dropped from further consideration. The residence time-weighted cumulative exposure, both 3 with and without decay of the exposure metric, did not fit these lung-cancer mortality data well 4 compared to the other models (see Table 5-12); this form of exposure metric does not 5 demonstrate evidence of an empirical fit to these epidemiologic data. 6 The model with the smallest AIC was for cumulative exposure with a 10-year half-life for 7 decay and a 10-year lag for cancer latency. The extended Cox model estimated a slope (beta) of 8 1.26 x 1CT2per fiber/cc-year based on a 365-day year, and the 95th percentile upper bound was _2 9 1.88 x 10 per fiber/cc-year. The/?-value for the Libby Amphibole asbestos regression 10 coefficient (slope) was <0.001. The slopes and confidence interval for the other exposure 11 metrics, which had similar fits to these data are reported in Table 5-13. Uncertainty in the choice 12 of the exposure metric (cross-metric uncertainty) is considered in the derivation of the final unit 13 risk (see details in Section 5.4.5.3), representing the range of unit risks that are derived from 14 these similarly fitting metrics. The model results that were ultimately selected to reflect the 15 upper-bound among the range of results were based on the cumulative exposure with a 10-year 16 lag exposure metric (CE10). The extended Cox model estimated a slope (beta) of 17 5.28 x 10"3per fiber/cc-year based on a 365-day year, and the 95thpercentile upper bound was _2 18 1.00x10 per fiber/cc-year. 19 20 6.2.6. Unit Risk Estimates for Lung-Cancer Mortality 21 The increased risk of lung-cancer mortality attributable to continuous fiber exposure was 22 estimated using a life-table procedure based on the general U.S. population. The life-table 23 procedure involved the application of the estimated Libby Amphibole asbestos-specific toxicity 24 to a structured representation of the general U.S. population in such a manner as to yield 25 age-specific risk estimated for cancer mortality in the presence or absence of exposure to Libby 26 Amphibole asbestos (see Section 5.4.5; Appendix G). 27 The nine exposure-response models retained in Table 5-13 all had reasonably similar 28 goodness of fits. No single model stands out as clearly statistically superior; however, there is a 29 range of quality of fit within the set that could be considered to have adequate fit. The 30 lung-cancer mortality unit risks are shown in Table 5-18. This document is a draftfor review purposes only and does not constitute Agency policy. 6-22 DRAFT--DO NOT CITE OR QUOTE 1 Using the results of the exposure model with the lowest AIC value (i.e., cumulative 2 exposure with a 10-year half-life for decay and a 10-year lag for cancer latency) alone, the LECoi 3 for the adult-only exposures was determined to be 0.333 fibers/cc. This yields an adult-based 4 unit risk of lung-cancer mortality of 0.0300 (POD of 1% divided by the LECoi). This estimate 5 was then scaled by 70/54 to encompass the whole lifespan; it yielded a lifetime unit risk of 6 0.0389 per fiber/cc. The value of the concentration that would correspond to the measure of 7 central tendency was based on the ECoi rather than LECoi. The ECoi for the adult-only 8 exposures was determined to be 0.499 per fiber/cc, which, when divided into a POD of 1%, 9 yielded an adult-based central estimate for lung-cancer mortality of 0.0200. This estimate was 10 then scaled by 70/54 to encompass the whole lifespan to, yielded a lifetime central estimate of 11 0.0260 per fiber/cc. 12 Using the results of the exposure model based on cumulative exposure with a 10-year lag 13 for cancer latency, the LECoi for the adult-only exposures was determined to be 0.191 per 14 fibers/cc, yielding an adult-based unit risk of lung-cancer mortality of 0.0524 (POD of 1% 15 divided by the LECoi). When scaled by 70/54 to encompass the whole lifespan, it yielded a 16 lifetime unit risk of 0.0679 per fiber/cc. The value of the risk that would correspond to the 17 measure of central tendency involves the ECoi rather than the LECoi. The ECoi for the 18 adult-only exposures was determined to be 0.325 per fiber/cc, which, when divided into a POD 19 of 1%, yielded an adult-based central estimate for lung-cancer mortality of 0.0308. This estimate 20 was then scaled by 70/54 to encompass the whole lifespan to, yielded a lifetime central estimate 21 of 0.0399 per fiber/cc. 22 The resulting unit risks in Table 5-18 ranged from 0.0260 to 0.0679 per fibers/cc, for a 23 lifetime continuous exposure. This shows that the unit risk based on the exposure metric with 24 the lowest AIC value (i.e., cumulative exposure with a 10-year half-life for decay and a 10-year 25 lag for cancer latency) is in the center of this range (i.e., 0.0389 per fiber/cc). This estimate is in 26 the middle of the range of possible unit risks and does not capture the uncertainty across metrics 27 with similar goodness of fit (see details in Section 5.4.6.1.3). 28 The model results selected to represent the upper bound risk among the range of 29 reasonable results are based on CE10 metric with a 10-year lag. The model results selected to 30 reflect the upper-bound among the range of results are based on the CE10 exposure metric with a 31 10-year lag, providing an IUR of 0.0679 per fibers/cc. This document is a draftfor review purposes only and does not constitute Agency policy. 6-23 DRAFT--DO NOT CITE OR QUOTE 1 6.2.7. IUR Derivation Based on Combined Mesothelioma and Lung-Cancer Mortality from 2 Exposure to Libby Amphibole Asbestos 3 When risks are combined, it is important to understand several concepts that are pertinent 4 to the evaluation and comparison of the cancer-specific mortality unit risks that will be 5 combined. First, there is statistical uncertainty in the potency estimate within the exposure 6 response model defined by each exposure metric. This within metric uncertainty is accounted 7 for in the confidence interval on slope. Next, there is an uncertainty in the choice of metrics for 8 developing IUR (cross-metric uncertainty). Finally, when unit risks corresponding to metrics are 9 chosen accounting for uncertainty, these are statistically combined into IUR. Details are 10 provided in Section 5.4.5.3. 11 Table 6-1 shows cancer-specific unit risks as well as combined risk of mesothelioma and 12 lung cancer. The IUR value of 0.17 per fiber/cc, continuous lifetime exposure, accounts for 13 important quantitative uncertainties in the selection of the specific exposure metric that may have 14 remained in an IUR that might have been based on the best-fitting exposure models alone. 15 Additional uncertainties are discussed in detail in Section 5.4.6. 16 17 18 Table 6-1. Reasonable upper bound and lowest information criteria 19 estimates of central risks and unit risks, for mesothelioma mortality, 20 lung-cancer mortality, and the IUR for the combined mortality risk from 21 mesothelioma and lung cancer (IURs are presented in the units of excess 22 cancers per fibers/cc, continuous lifetime exposure) 23 M esothelioma Lung cancer Combined mesothelioma and lung cancer M odel C en tral estim ate Unit risk C en tral estim ate Unit risk Central estim ate IUR Reasonable upper bound3 0.075 0.122 0.040 0.068 0.115 0.169 Lowest information criteria13 0.044 0.074 0.026 0.040 0.070 0.103 24 25 T or mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay 26 half-life o f 5 years and a 15-year lag. For lung cancer, the selected model parameterized exposure as cumulative 27 exposure without decay and a 10-year lag. 28 bFor mesothelioma, the selected model parameterized exposure as cumulative exposure with exponential decay 29 half-life o f 5 years and a 10-year lag. For lung cancer, the selected model parameterized exposure as cumulative 30 exposure with exponential decay half-life o f 10 years and a 10-year lag. 31 32 This document is a draftfor review purposes only and does not constitute Agency policy. 6-24 DRAFT--DO NOT CITE OR QUOTE 1 6.2.7.1. Comparison with Other Published Studies o f Libby, M T Workers Cohort 2 Several published studies have previously evaluated risk of mesothelioma and lung 3 cancer (i.e., Larson et ah. 2010b; Moolsavkar et ah. 2010; Berman and Crump. 2008; Sullivan. 4 2007) in Libby, MT workers cohort. 5 For mesothelioma, only Moolgavkar et al. (2010) provided an exposure-response 6 relationship for absolute risk of mesothelioma mortality that would be comparable with this 7 current assessment. Based on the full cohort, with mortality data through 2001 and a 8 modification of the Peto/Nicholson exposure metric, life-table analysis would provide an upper 9 bound unit risk of approximately 0.13 per fibers/cc continuous lifetime exposure. Therefore, 10 utilization of the exposure response modeling of Moolgavkar et al. (2010), would provide an 11 IUR for excess mesothelioma mortality in close agreement with the 1UR derived in this 12 assessment (see Section 5.4.5.3.1 for more details). 13 For lung cancer, all of the studies provide exposure-response relationships in terms of 14 relative risk of lung-cancer mortality and, thus, may provi de risk estimates comparable number 15 to this assessment. However, inclusion criteria, length of mortality follow-up, and analytic 16 methods differ among the analyses--thus, the results are not necessarily interchangeable. For 17 comparison purposes, the lung cancer unit risk from these studies are computed from life-table 18 analyses (see Table 5-20). The lung cancer IlJRs calculated based on the published literature, 19 ranged from 0.010 to 0.079 per fiber/cc (based on the upper-confidence limit). This is in close 20 agreement with this current assessment where an upper-bound estimate of 0.068 per fiber/cc, 21 continuous lifetime exposure is derived (see Section 5.4.5.3.1 for more details). 22 23 6.2.8. Sensitivity Analysis 24 6.2.8.1. Sensitivity Analysis o f Influence o f High Exposures in Early 1960s on Model Fit 25 Although data do not exist to evaluate biological bases for model fit, other potential 26 factors can be explored where data allow. For example, because of concerns that very high 27 (>100 fibers/cc) early (1960-1963) 8-hour Libby Amphibole asbestos TWA exposures (see 28 Table 5-7) could have influenced the relative fit of the various exposure metrics, EPA conducted 29 a sensitivity analysis of the impact on the relative model fit of reducing all estimated exposure 30 intensities for 1960-1963 by 50%. This document is a draftfor review purposes only and does not constitute Agency policy. 6-25 DRAFT--DO NOT CITE OR QUOTE 1 For modeling mesothelioma and lung-cancer mortality on this revised data set, there was 2 very little difference in the order of the relative fits of the exposure models as was seen for the 3 subcohort of workers hired after 1959 and included the exposures as estimated by Amandus et al. 4 (1987a) during 1960-1963 (see Tables 5-14 and 5-15). The models based on the revised data set 5 fit approximately as well for mesothelioma and for lung cancer. 6 7 6.2.8.2. Analysis o f Potential Confounding o f Lung Cancer Results by Smoking in the 8 Sub-cohort 9 EPA used three approaches to address the confounding issue, including restriction of the 10 cohort and two analytic evaluations of the potential for confounding by smoking including the 11 method described by Richardson (2010). Richardson (2010) describes a method to determine if 12 an identified exposure relationship with lung cancer is confounded by unmeasured smoking in an 13 occupational cohort study. EPA implemented this methodology to model the potential effects of 14 Libby Amphibole asbestos on the risk of COPD mortality on the subcohort of workers hired after 15 1959 (see Section 5.4.3.6.5). Summarizing these findings, EPA used the method described by 16 Richardson (2010) to evaluate whether exposures to Libby Amphibole asbestos predicted 17 mortality from COPD as an indication of potential confounding by smoking and found a 18 nonsignificant negative relationship, which was inconsistent with confounding by smoking. 19 20 6.2.9, Uncertainty in the Cancer Risk Values 21 It is important to consider the uncertainties in the derivation of the mesothelioma and 22 lung-cancer mortality risks in this assessment in the context of uncertainties in animal-based 23 health assessments. This assessment does not involve extrapolation from high dose in animals to 24 low dose in humans. The current assessment is based on a well-documented and well-studied 25 cohort of workers with adequate years of follow-up to evaluate mesothelioma and lung-cancer 26 mortality risks with PODs within the range of the data. The discussions in Section 5.4.6 explore 27 uncertainty in the derivati on of the KIR in order to provide a comprehensive and transparent 28 context for the resulting cancer mortality risk estimates. 29 The summary below includes likely one-sided uncertainties (biases) associated with the 30 derivation of the IUR in order to provide a context for the resulting cancer risk estimates. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 6-26 DRAFT--DO NOT CITE OR QUOTE 1 The sources of uncertainty that could lead to a likely underestimation of the cancer risk 2 value include the following: 3 4 5 Use o f historical phase contrast microscopy (PCM) exposure measurements. As asbestos 6 was a contaminant of vermiculite that was the primary object of production, mine and dry 7 and old wet mill ambient air may have contained material other than asbestos that could 8 have contributed to fibers counted by PCM. Therefore, it is possible that exposure 9 estimates for some or possibly a large portion of the cohort are overestimated, and, 10 therefore, the resulting IUR may be underestimated. 11 Measurement error in exposure assessment and assignment. This current assessment 12 showed that unit risk results from analysis of the lung-cancer mortality in the full cohort 13 (see Table 5-21) compared with the sub-cohort hired after 1959 may have been 14 attenuated as much as 2-6 times (see Section 5.4.6.1.2.4). By excluding those cohort 15 members hired before 1960 for whom there was insufficient work history information to 16 estimate their exposures, the unit risk for lung cancer was less attenuated due to exposure 17 measurement error. However, exposure measurements from the 1960s are also imperfect 18 and include a lesser degree of exposure measurement error, which could have led to 19 underestimated risk--even in the sub-cohort hired after 1959. 20 Limited length o ffollow-up. The IUR for mesothelioma mortality could be larger than 21 was estimated from existing data, since latency of mesothelioma can be as long as 22 60 years. The maximum length of follow-up was 46 years in this cohort. The magnitude 23 of underestimation would depend on the relationship between the number of additional 24 deaths and the increase in person-years. 25 Use o f life-tables to calculate the IUR based on cancer mortality. The life-table 26 procedure computes the extra risk of death from birth up to 85 years of age. This cut-off 27 at age 85 ignores a small additional risk of lung-cancer mortality among a small 28 percentage of people who have a higher background risk because of the increase in lung 29 cancer risk that is seen with increasing age. The lung-cancer mortality unit risk based on 30 the LECoi would be somewhat larger, on the order of 5-10%. On the other hand, the 31 additional mesothelioma mortality risk, if the life-tables were extended to account for 32 longer life spans, would be about 3%. 33 Small number o f women and ovarian cancer. While asbestos is causally associated with 34 increased risks of ovarian cancer (Straif et aU 2009). there were only 84 women in the 35 whole cohort, and there were no deaths from ovarian cancer among 24 total deaths. The 36 lack of observed ovarian cancer in this cohort may be a function of the limited number of 37 female deaths in the cohort allowing for the possibility that exposure to Libby Amphibole 38 asbestos could result in increased risk of ovarian cancer. However, it was not possible to 39 estimate the magnitude of this underestimation on the total cancer risk. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 6-27 DRAFT--DO NOT CITE OR QUOTE 1 Dependent competing risks. Competing risk of mortality from other diseases related to 2 exposure may have resulted in underestimates of the risk of mortality from either 3 mesothelioma or lung cancer. The mean length of follow-up for the Libby, MT workers 4 who died of mesothelioma was to 30.1 years, and evidence exists (Bianchi and Bianchi, 5 2009; Suzuki and Yuen, 2001) that early deaths from other exposure-related causes could 6 have precluded an individual's risks of death from mesothelioma. However, it was not 7 possible to estimate the magnitude of this effect on the total cancer risk. 8 9 10 The source of uncertainty that could lead to a likely overestimation of the cancer risk 11 value; 12 13 14 Potential residual confounding and effect modification. The unit risk of lung-cancer 15 mortality estimated herein, and the combined mesothelioma and lung-cancer mortality 16 IUR, would over-estimate the risk in any population that had a lower prevalence of 17 smoking than that of the Libby worker cohort. Because the Libby worker cohort had a 18 large prevalence of smokers and ex-smokers and no known nonsmokers developed lung 19 cancer, it is also possible that estimated risk for lung cancer is actually risk for an 20 interaction of lung cancer and smoking, and effects of smoking and asbestos are known 21 to be between additive and multiplicative (see Section 4). However, the company 22 imposed smoking ban, and the observation that there were many ex-smokers in the 23 cohort, would tend to lessen risks that would have occurred if these individuals continued 24 smoking. 25 26 27 6.3. APPLICATION OF THE LIBBY AMPHIBOLE ASBESTOS RFC AND IUR 28 6.3.1. Sites and Materials 29 This Libby Amphibole asbestos specific assessment is based on the evaluation of worker 30 cohorts, exposed to asbestos from a single mine in Libby, MT, and it is intended to allow for 31 estimates of the risk due to exposure to the asbestos fibers from that mine, or exposures to 32 asbestos fibers that arise from the management or use of the vermiculite ore and exfoliated 33 vermiculite from this mine. Therefore, it is appropriate to apply the Libby Amphibole 34 asbestos-specific RfC and/or IUR to sites which are believed to have been contaminated by these 35 materials when assessing risk from the amphibole fibers present from this contamination. This 36 may include sites where the ore was shipped or handled, where the vermiculite was exfoliated 37 and further processed, facilities which in other ways shipped or handled the exfoliated 38 vermiculite, where products containing the raw or exfoliated vermiculite were present, the 39 consumer products themselves (e.g., vermiculite attic insulation) and any waste streams from the This docum ent is a draftfo r review purposes only and does n ot constitute A gency policy. 6-28 DRAFT--DO NOT CITE OR QUOTE 1 above processes which contain vermiculite and the related Libby Amphibole asbestos-fibers. 2 The assessment was derived from PCM measurements taken at the Libby, MT occupational sites 3 and the mixture of minerals found in those measurements. It does not estimate the risk 4 attributable to specific subsets of those fibers whether based on size, shape, or mineral 5 composition other than the limitations on size and shape reflected in the PCM methodology and 6 counting rules. As detailed in Section 2, the amphibole asbestos present in the mine, ore and 7 expanded vermiculite, does not fit cleanly into a single category of nomenclature for amphibole 8 minerals. Most Libby Amphibole fibers are classified as winchite (84%), with lesser amounts of 9 richterite (11%) and tremolite (6%), based on the nomenclature proposed by Leake et al. (1997). 10 There are also trace amounts of magnesioriebeckite, edenite, and magnesio-arfvedsonite present 11 in Libby Amphibole asbestos (Meeker et al.. 2003). Within the 30 samples taken from the mine 12 the proportion of these minerals differed between samples (Meeker et al.. 2003) and the relative 13 proportions of these species may have varied over time (as ore from different locations was 14 processed). This assessment estimates the risk of exposure to the varying range of mineral fiber 15 mixtures that result from material originating from the geological deposit, recognizing there is 16 variation and uncertainty as to variations in the exposure to the underlying cohort and complex 17 variation in settings to which these estimates will be applied. 18 19 6.3.2. Exposure Units for Libby Amphibole Asbestos 20 As with the IRIS assessment for asbestos (U.S. EPA, 1988a). the RfC and IUR specific to 21 Libby Amphibole asbestos are presented here as fibers/cc exposure continuous lifetime exposure, 22 where exposure measurements are based on analysis of air filters by PCM. Early PCM analytical 23 techniques did not have the same resolution as current analytical methods, and it is understood 24 that PCM data for the majority of the exposures characterized for the Libby, MT workers and 25 Marysville, OH workers would likely have a width resolution of 0.4-0.44 pm (Amandus et al.. 26 1987a; IPCS. 1986; Rendall and Skikne. 1980). Therefore, as with the IRIS assessment for 27 asbestos (U.S. EPA. 1988a). the dimensions of the PCM fibers for the Libby Amphibole asbestos 28 unit risk are defined as fibers >5 pm in length with an aspect ratio of 3:1 or greater and a width 29 >0.4 pm. 30 Environmental air sampling for asbestos is now often analyzed by transmission electron 31 microscopes (TEM) to confirm that the fibers viewed are asbestos, and often it is used to identify This document is a draftfor review purposes only and does not constitute Agency policy. 6-29 DRAFT--DO NOT CITE OR QUOTE 1 the mineralogy of the fiber. Although some historical data do exist providing TEM analysis of 2 airborne fibers from the Libby, MT mill operation (McDonald et al.. 1986a; Langer et al.. i974). 3 these data are not sufficient to provide an alternative set of exposure measurements in TEM units 4 for the Libby, MT worker cohort, or provide a PCM to TEM conversion across the various work 5 environments. 6 Different sampling environments and varied site conditions may pose the potential for 7 airborne fibers from various materials. Because of that, it is expected that for many 8 environmental risk assessments conducted now and in the near future, measures of exposure may 9 be done with methods such as TEM and then adjusted through fiber-counting rules to estimate 10 the number of PCM-countable asbestos fibers. Site-specific environmental conditions should be 11 considered in determining how to best identify PCM-countable asbestos fibers in relevant air 12 samples for exposure assessments used in conjunction with this health assessment to yield 13 estimates of risk. 14 15 6.3.3. Applications to Early Lifetime and Partial Lifetime Environmental Exposure 16 Scenarios for IUR 17 The Libby Amphibole asbestos-specific unit risk derived in this assessment is a combined 18 risk of lung cancer and mesothelioma, each with its own adjustment for uncertainty in metrics. 19 The life-table analyses for Libby Amphibole asbestos do not predict greater risk from early-life 20 exposures. Thus, this assessment recommends that estimates of the risks of less-than-lifetime 21 exposures be computed by simple calculations of average lifetime exposure concentration 22 multiplied by IUR. This recommendation is consistent with standard Superfund guidance, where 23 exposures are estimated, averaged across a lifetime exposure, and the IUR simply applied to 24 calculate excess cancer risk (U.S. EPA. 2008. 2001b). The weight of evidence does not support 25 a mutagenic mode of action for Libby Amphibole asbestos carcinogenicity. Therefore, 26 according to EPA's Supplemental Guidancefor Assessing Susceptibility from Early-Life 27 Exposure to Carcinogens (U.S. EPA. 2005b), the application of the age-dependent adjustment 28 factors are not recommended. 29 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 6-30 DRAFT--DO NOT CITE OR QUOTE 1 6.3.4. Applications to Lifetime and Partial Lifetime Environmental Exposure Scenarios for 2 RfC 3 The Libby Amphibole asbestos specific RfC should be used to derive estimates of hazard 4 from exposure to airborne materials containing Libby Amphibole asbestos as described above. 5 The Libby Amphibole asbestos RfC was derived from an evaluation of the O.M. Scott, 6 Marysville, OH worker cohort (Rohs et al.. 2008; Lockev et al.. 1984). Exposure-response 7 modeling of cumulative Libby Amphibole asbestosis exposure with the best-fitting model 8 (Michaelis-Menten with 10-year lagged exposure) resulted in a BMCLio of 0.1177 fibers/cc-year 9 yielding an RfC for a 70-year lifetime of 2 x 10 5fibers/cc by calculating the average 10 concentration over a 60-year averaging period (70 years minus 10-year lag). 11 The estimate of hazard should be calculated by dividing the average daily exposure 12 concentration using an averaging period of 60 years by the reference concentration outlined in 13 Superfund Guidance to yield a quotient representing hazard (U.S. EPA. 2001b). The use of the 14 reference concentration in risk assessment is further clarified in RAGs, Part F, Supplemental 15 Guidancefor Inhalation risk Assessment (U.S. EPA. 2009a). The guidance provides for 16 addressing hazard for children and adults by estimating time-dependent average daily exposures. 17 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 6-31 DRAFT--DO NOT CITE OR QUOTE 7. REFERENCES AGI. (American Geological Institute). (2005). Glossary of Geology. In (5th ed.). Alexandria, VA. Amandus. HE: Wheeler. R: Jankovic, J; Tucker. J. (1987a). The morbidity and mortality of venniculite miners and millers exposed to tremolite-actinolite: Part I. Exposure estimates. Am J Ind Med 11: 1-14. Amandus. HE: Wheeler. R. (1987). The morbidity and mortality of venniculite miners and millers exposed to tremolite-actinolite: Part II. Mortality. Am J Ind Med 11: 15-26. Amandus. HE: Althouse. R: Morgan. WKC: Sargent. EN: Jones. R. (1987b). The morbidity and mortality of venniculite miners and millers exposed to tremolite-actinolite: Part III. Radiographic findings. Am J Ind Med 11: 27-37. Amandus. PIE: Wheeler. R: Armstrong, BG: McDonald. AD: McDonald. JC: Sebastien. P, (1988). Mortality of venniculite miners exposed to tremolite. Ann Occup Hyg 32: 459-465. http://dx.doi.Qrg/10.1093/annhvg/32.inhaled particles VI.459. Andersoa HA: Lilis. R: Damn. SM: Fischbein. AS: Selikoff. IJ. (1976). Household-contact asbestos neoplastic risk. Ami N Y Acad Sci 271: 311 -323. http://dx.doi.org/10.1111/i,1749-6632,1976,tb23127.x. Andersoa HA: Lilis. R: Damn. SM: Selikoff. IJ, (1979). Asbestosis among household contacts of asbestos factory workers. A nnN Y Acad Sci 330: 387-399. http://dx.doi.Org/10.llll/i.1749-6632.1979.tbl8741.x. Andrioa A: Bosia. S: Paoletti. L: Fevles. E; Lanfranco. C: Beilis. D: Mollo. F, (1994). Malignant peritoneal mesothelioma in a 17-year-old boy with evidence of previous exposure to chrysotile and tremolite asbestos. Hum Pathol 25: 617-622. http://dx.doi.org/10.1016/0046-8177(94)90230-5. ' Annstrong. B, (1998). Effect of measurement enor on epidemiological studies of environmental and occupational exposures. Occup Environ Med 55: 651-656. http://dx.doi.org/10.1136/oem.55.10.651. Annstrong. BG: Mcdonald. JC: Sebastien. P: Althouse. R: Amandus. HE: Wheeler. R, (1988). Radiological changes in venniculite workers exposed to Tremolite. Arm Occup Hyg 32: 469-474. http://dx.doi.ong/10.1093/annlivg/32.inlialed particles VI.469. Ascoli- V: Camovale-Scalzo. C; Nardi. F: Efrati. C: Menegozzo. M, (2003). A one-generation cluster of malignant mesothelioma within a family reveals exposure to asbestos-contaminated jute bags in Naples, Italy. Eur J Epidemiol 18: 171-174. http7/dx.doi.org/10.1023/A:1023032214444. ' Asgharian. B: Menache. MG: Miller. FJ, (2004). Modeling age-related particle deposition in humans. J Aerosol Med PulinDrag Deliv 17: 213-224. http://dx.doi.org/10.1089/iam.2004.17.213. Athanasiou. K: Constantopoulos. SH: Rivedal. E: Fitzgerald. DJ: Yamasaki. H, (1992). Metsovo-tremolite asbestos fibres: In vitro effects on mutation, chromosome abenation. cell transfonnation and intercellular communication. Mutagenesis 7: 343-347. http://dx.doi.Org/10.1093/inutage/7.5,343. ATS. (American Thoracic Society). (2004). Diagnosis and initial management of nonmalignant diseases related to asbestos. Am I Respir Crit Care Med 170: 691-715. http://dx.doi.org/10.1164/rccm,200310-1436ST. ATSDR. (Agency for Toxic Substances and Disease Registry). (2000). Health consultation: Mortality from asbestosis in Libby. Montana. Atlanta. GA. http://www.atsdr.cdc.gov/hac/pha/pha.asp?doeid=1225&pg=0. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 7-1 DRAFT--DO NOT CITE OR QUOTE ATSDR. (Agency for Toxic Substances and Disease Registry). (2001a). Toxicological profile for asbestos [ATSDR Tox Profile]. Atlanta, GA: U.S. Department of Health and Human Services, Public Health Service. http://www.atsdr.cdc. gov/ToxProfiles/tp61.pdf. ATSDR. (Agency for Toxic Substances and Disease Registry). (2001b). Year 2000 medical testing of individuals potentially exposed to asbestiform minerals associated with venniculite in Libby, Montana: A report to the community. Atlanta, GA: Agency for Toxic Substances & Disease Registry. http://www.atsdr.ede.gov/asbestos/sites/libbv montana/medical testing.html. ATSDR (Agency for Toxic Substances and Disease Registry). (2002). Libby tests: Lung abnormality rates high. Atlanta, GA. ATSDR. (Agency for Toxic Substances and Disease Registry). (2003a). Health consultation: Exposure assessment: Western Mineral Products Site. Atlanta, GA. http://www.health.state.mn.us/divs/eh/hazardous/sites/hennepin/western/wstmhcl003.pdf. ATSDR. (Agency for Toxic Substances and Disease Registry). (2003b). Report on the expert panel on health effects of asbestos and synthetic vitreous fibers: The influence of fiber length. ATSDR. (Agency for Toxic Substances and Disease Registry). (2004). Toxicological profile for synthetic vitreous fibers [ATSDR Tox Profile]. Atlanta, GA: U.S. Department of Health and Human Sendees, Public Health Service, http://www.atsdr,ede.gov/ToxProfiles/tp 161, pdf. ATSDlA. (Agency for Toxic Substances and Disease Registry). (2005a). Final report of the Northeast Minneapolis Community Venniculite Investigation (NMCVI) and worker/household study: Cohort identification and characterization. Atlanta, GA. http://www.health.state.mn.us/divs/eh/hazardous/sites/hennepin/western/nmcvicohort.pdf. ATSDR. (Agency for Toxic Substances and Disease Registry). (2005b). Health Consultation: Former O.C. Scoh and Sons Company, 14111 Scottslawn Road. Marysville. Union County, Ohio. EPA Facility ID: OHD990834483. (PB2008-104070). Atlanta, GA. ATSDR. (Agency for Toxic Substances and Disease Registry). (2008a). Fact sheet on the Summary report: Exposure to asbestos-containing vermiculite from Libby, Montana, at 28 processing sites in the United States [Fact Sheet]. Atlanta, GA. http://www.atsdr.ede.gov/asbestos/sites/national map/Summarv Report Fact Sheet.pdf. ATSDR. (Agency for Toxic Substances and Disease Registry). (2008b). Summary report: Exposure to asbestoscontaining venniculite from Libby, Montana, at 28 processing sites in the United States. Atlanta, GA: Agency for Toxic Substances & Disease Registry. http://www.atsdr.cde.gov/asbestos/sites/national map/Summarv Report 102908.pdf. Aust. AE: Cook, PM: Dodson. RF, (2011). Morphological and chemical mechanisms of elongated mineral particle toxicities. J Toxicol Environ Health B Crit Rev 14: 40-75. http://dx.doi.org/10.1080/10937404.201L556046. Bailey, KF: Kelse, J: Wylie, AG: Lee. RJ. (2006). The asbestiform and prismatic mineral growth habit and their relationship to cancer studies: A pictorial presentation. Alexandria. VA: National Stone, Sand & Gravel Association. http://www.regulations.gov/#!documentDetail:D=EPA-HO-QRD-2003-0016-0Q76. Baker, EL: Dagg. T: Greene. RE. (1985). Respiratory illness in the construction trades. I. The significance of asbestos-associated pleural disease among sheet metal workers. J Occup Med 27: 483-489. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 7-2 DRAFT--DO NOT CITE OR QUOTE Baris. I: Simonato, L; Artvinli. M; Poolev. F; Saracci. R: Skidmore, J: Wagner. C. (1987). Epidemiological and environmental evidence of the health effects of exposure to erionite fibres: A four-year study in the Cappadocian region of Turkey. Ini J Cancer 39: 10-17. http://dx.doi.org/10.1002/iic.2910390104. Baris. YI: Artvinli. M; Sahin. AA, (1979). Environmental mesothelioma in Turkey. Ann N Y Acad Sci 330: 423 432. http://dx.doi.Org/10.llll/i.1749-6632.1979.tbl8744.x. Barone-Adesi. F: Ferrante. D: Bertolotti. M: Todesco. A: Mirabelli. D: Terracini. B: Magnani. C. (2008). Long-term mortality from pleural and peritoneal cancer after exposure to asbestos: Possible role of asbestos clearance. Int J Cancer 123: 912-916. http://dx.doi.org/10.1002/iic.23609. Bateson. T: Wright. J. (2010). Regression calibration for classical exposure measurement error in environmental epidemiology studies using multiple local surrogate exposures. Am J Epidemiol 172: 344-352. http://dx.doi.org/10,1093/aie/kwa 123. Bateson. TF: Schwartz. J. (2008). Children's response to air pollutants. J Toxicol Environ Health A 71: 238-243. http://dx.doi.org/10.1080/152873907Q1598234. Becauem ia MH: Swift. DL: Boucliikhi. A: Roy, M: Teillac. A, (1991). Particle deposition and resistance in the noses of adults and children. Eur Respir J 4: 694-702. Bennett. W: Zeman. K: Jarabek. A, (2003). Nasal contribution to breathing with exercise: effect of race and gender. J Appl Physiol 95: 497-503. http://dx.doi.org/10.1152/iapplphvsiol.00718.2002. Bennett. WD: Zeman. KL; Kim. C. (1996). Variability of fine particle deposition in healthy adults: effect of age and gender. Am J Respir Crit Care Med 153: 1641-1647. Bennett. WD: Zeman. KL: Kim. C: Mascarella. J. (1997). Enhanced deposition of fine particles in COPD patients spontaneously breathing at rest. Inlial Toxicol 9: 1-14. http://dx.doi.org/10.1080/Q89583797198376. Bennett. WD: Zeman. KL, (2004). Effect of body size on breathing pattern and fine-particle deposition in children. J Appl Physiol 97: 821-826. http://dx.doi.org/10.1152/iapplphvsiol.01403.20Q3. Bennett. WD: Zeman. KL, (2005). Effect of race on fine particle deposition for oral and nasal breathing. Inlial Toxicol 17: 641-648. http://dx.doi.org/10.1080/0895837050Q188984. Bennett. WD: Zeman. KL: Jarabek. AM. (2008). Nasal contribution to breathing and fine particle deposition in children versus adults. J Toxicol Environ Health A 71: 227-237. http://dx.doi.ora/10.1080/15287390701598200. Berksoa J. (1950). Are there two regressions? J Am Stat Assoc 45: 164-180. Berman. DW: Crump. KS. (2008). Update of potency factors for asbestos-related lung cancer and mesothelioma. Crit Rev Toxicol 38: 1-47. http://dx.doi.org/10.1080/104084408Q2276167. Berinaa DW. (2010). Comparing milled fiber, Quebec ore, and textile factory dust: has another piece of the asbestos puzzle fallen into place. Crit Rev Toxicol 40: 151-188. http://dx.doi.org/10.3109/104084409Q3349137. Bernstein, D: Rogers. R: Smith. P, (2005). The biopersistence of Canadian chrysotile asbestos following inhalation: final results through 1 year after cessation of exposure. Inlial Toxicol 17: 1-14. http://dx.doi.org/10.1080/08958370590885663. This document is a draftfor review purposes only and does not constitute Agency policy. 7-3 DRAFT--DO NOT CITE OR QUOTE Bernstein. DM: Mast. R; Andersoa R; Hesterberg. TW: Musselmaa R: Kamstrup. O: Hadley, J. (1994). An experimental approach to the evaluation of the biopersistence of respirable synthetic fibers and minerals. Environ Health Perspect 102 Suppl 5: 15-18. Bernstein. DM; Chevalier. J; Smith. P. (2003). Comparison of Calidria clirysotile asbestos to pure tremolite: Inhalation biopersistence and histopathology following short-term exposure. Inlial Toxicol 15: 1387-1419. http://dx.doi.org/10.1080/08958370390248888. Benv. G: Wagner. JC. (1969). The application of a mathematical model describing the times of occurrence of mesotheliomas in rats following inoculation with asbestos. Br J Cancer 23: 582-586. Berry, G: Gilson. JC: Holmes. S: Lewinsolm. ITC: Roach. SA, (1979). Asbestosis: A study of dose-response relationships in an asbestos textile factory. Occup Environ Med 36: 98-112. http://dx.doi.Org/10.1136/oem.36.2.98. Berry . G. (1999). Models for mesothelioma incidence following exposure to fibers in terms o f timing and duration of exposure and the biopersistence of the fibers. Inlial Toxicol 11: 111-130. http://dx.doi.org/10.1080/0895837991972Q3. Berry, G: de Klerk. N: Reid. A: Ambrosini. G: Fritsclii. L: Olsen. N: Merler. E; Musk. A. (2004). Malignant pleural and peritoneal mesotheliomas in former miners and millers of crocidolite at Wittenoom, Western Australia. Occup Environ Med 61: e l 4. Berry, G: Poolev. F: Gibbs. A: Harris. J: McDonald. J. (2009). Lung fiber burden in the Nottingham gas mask cohort. Inlial Toxicol 21: 168-172. http://dx.doi.org/10.1080/089583708022913Q4. BGS. (British Geological Survey). (2011). World mineral production. Keyworth. Nottingham, UK. Bianchi. AB: Mitsunaga. SI: Cheng. JO: Klein. WM: Jhanwar. SC: Seizinger. B: Kiev. N: Klein-Szanto. AJ: Testa. JR. (1995). High frequency of inactivating mutations in the neurofibromatosis type 2 gene (NF2) in primary malignant mesotheliomas. PNAS 92: 10854-10858. Bianchi. C: Bianchi. T, (2008). Susceptibility and resistance in the genesis of asbestos-related mesothelioma. Indian J Occup Environ Med 12: 57-60. http://dx.doi.org/10.4103/0019-5278.43261. Bianchi. C: Bianchi. T. (2009). Malignant pleural mesothelioma in Italy. Indian J Occup Environ Med 13: 80-83. http://dx.doi.org/10.4103/0019-5278.55124. ' Bignon. J: Monchaux. G: Sebastien. P: Hirsch. A: Lafuma. J. (1979). Human and experimental data on translocation of asbestos fibers through the respiratory system. A nnN Y Acad Sci 330: 745-750. Bignon. J: Sebastien. P: Gaudichet, A, (1978) Measurement of asbestos retention in the human respiratory system related to health effects. In CC Gravatt; PD LaFleur; KFJ Heinrich (Eds.), Proceedings of the workshop on asbestos: definitions and measurement methods (Vol. 506, pp. 95-119). Washington, DC: National Bureau of Standards. Bilgraa. R. (1997). National Death Index user's manual. Hyattsville, MD: National Center for Health Statistics. Bilgrad. R. (1999). National Death Index Plus: Coded causes of death: Supplement to the National Death Index user's manual. Hyattsville, MD: National Center for Health Statistics. Blake. D: B olia C: Cox, D: Cardozo-Pelaez. F: Pfau. J. (2007). Internalization of Libby amphibole asbestos and induction of oxidative stress in murine macrophages. Toxicol Sci 99: 277-288. http://dx.doi.org/10.1093/toxsci/kfml66. This document is a draftfor review purposes only and does not constitute Agency policy. 7-4 DRAFT--DO NOT CITE OR QUOTE Blake. D; Wetzel. S; Pfau. J. (2008). Autoantibodies from mice exposed to Libby amphibole asbestos bind SSA/Ro52-enriched apoptotic blebs o f murine macrophages. Toxicology 246: 172-179. http://dx.doi.Org/10.1016/i.tox.2008.01.008. Bocchetta. M: P i Resta. I: Powers. A: Fresco. R; Tosolini. A, Testa. J: Pass. H: Rizzo. P: Carbone. M. (2000). Human mesothelial cells are unusually susceptible to simian vims 40-mediated transformation and asbestos cocarcinogenicity. PNAS 97: 10214-10219. http://dx.doi.org/10.1073/pnas.170207Q97. Boettcher. AL. (1966). Vermiculite, hydrobiotite, and biotite in the Rainy Creek igneous complex near Libby, Montana. Clay Minerals 6: 283-296. Boettcher. C: Stark. LI: Vanheel, M, (1996). Stacked bilayer helices: A new structural organization o f amphiphilic molecules. Ultramicroscopy 62: 133-139. http://dx.doi.org/10.1016/0304-3991(95)00140-9. Bourbeau. J: Ernst. P: Chrome. J: Armstrong. B: Becklake. MR, (1990). The relationship between respiratory impairment and asbestos-related pleural abnormality in an active work force. Am Rev Respir Dis 142: 837 842. http://dx.doi.Org/10.1164/airccm/142.4.837. Broaddus. VC: Everitt. JI: Black. B: Kane. AB. (2011). Non-neoplastic and neoplastic pleural endpoints following fiber exposure. J Toxicol Environ Health B Crit Rev 14: 153-178. http://dx.doi.Org/10.1080/10937404.201L556049. Broderick. A: Fuortes. LJ: Merchant. JA: Galvin, JR: Schwartz. DA. (1992). Pleural detenninants o f restrictive lung function and respiratory symptoms in an asbestos-exposed population. Chest 101: 684-691. Brody. AR: Hill. LH: Adkins. B. Jr: O'Connor. RW. (1981). Clnysotile asbestos inhalation in rats: deposition pattern and reaction of alveolar epithelium and pulmonary macrophages. Am Rev Respir Dis 123: 670-679. Brown. J: Bennett. W. (2004). Deposition o f coarse particles in cystic fibrosis: Model predictions versus experimental results. J Aerosol Med 17: 239-248. http://dx.doi.org/10.1089/iam.2004.17.239. Brown. JS: Zeman. KL: Bennett WD. (2001). Regional deposition o f coarse particles and ventilation distribution in healthy subjects and patients with cystic fibrosis. J Aerosol Med Pulm Drag Deliv 14: 443-454. http://dx.doi.org/10.1089/Q894268dl52744659. Brown. JS: W ilsoa WE: Grant. LD, (2005). Dosimetric comparisons o f particle deposition and retention in rats and humans, lnlial Toxicol 17: 355-385. Bunderson-Schelvan. M: Pfau. JC: Crouch. R: Holian. A, (2011). Nonpulmonary outcomes o f asbestos exposure. J Toxicol Environ Health B Crit Rev 14: 122-152. http://dx.doi.org/10.ld80/10937404.201L556048. Burnham. KP: Anderson. DR, (2002). Model selection and multimodel inference: a practical information-theoretic approach. In. New York: Springer. Burr. D, (1988). On errors-in-variables in binary regression-Berkson case. J Am Stat Assoc 83: 739-743. Camidge. DR: Stockton. DL: Bain. M, (2006). Factors affecting the mesothelioma detection rate within national and international epidemiological studies: Insights from Scottish linked cancer registry-mortality data. Br J Cancer 95: 649-652. http://dx.doi.org/10.1038/si.bic.66Q3293. Carbone. M: Albelda. SM: Broaddus. VC: Flores. RM: Hillerdal. G: Jaurand, MC: Kiaerheim. K: Pass. HI: Robinsoa B: Tsao. A, (2007). Eighth International Mesothelioma Interest Group [Review]. Oncogene 26: 6959-6967. http://dx.doi.org/10.1038/si.onc.121Q515. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 7-5 DRAFT--DO NOT CITE OR QUOTE Carroll. RL Ruppert. D; Stefanski. LA. (2006). Measurement error in nonlinear models: A modem perspective. In. London. England: Chapman & Hall. Case. BW: McDonald. AD: McDonald. JC. (2000). Asbestos fiber type and length in lungs o f chrysotile textile and production workers: Fibers longer than 18 mum. Inlial Toxicol 12: 411-418. http://dx.doi.org/10.1080/0895837005Q165319. Case. BW: Abraham. JL: Meeker. G: Poolev. FD: Pinkerton. KE, (2011). Applying definitions o f "asbestos" to enviromnental and "low-dose" exposme levels and health effects, particularly malignant mesothelioma. J Toxicol Environ Health B CritRev 14: 3-39. http://dx.doi.org/10.1080/10937404.2011.556Q45. Castellan RM: Sanderson. WT: Petersen. MR, (1985). Prevalence o f radiographic appearance o f pneumoconiosis in an unexposed blue collar population. Am Rev Respir Dis 131: 684-686. Castranova. V: Pailes. W: Judy. D: Blake. T: Schwegler-Berrv. D: Jones. W. (1996). In vitro effects o f large and small glass fibers on rat alveolar macrophages. J Toxicol Environ Health 49: 357-369. Cazzadori. A: Malesani. F: Romeo. L. (1992). Malignant pleural mesothelioma caused by non-occupational childhood exposme to asbestos. Br J Ind Med 49: 599. http://dx.doi.Org/10.1136/oem.49.8.599. CPC. (Centers for Disease Control and Prevention). (2009). Malignant mesothelioma mortality: United States, 1999-- 2005. MMWR Morb Mortal Wkly Rep 58: 393-396. CEC. (Commission o f the Emopean Communities). (1996). Guidance on risk assessment at work. Lnxembourg: Office for Official Publications o f the Emopean Communities. http://osha.europa.eu/en/topics/riskassessment/guidance.pdf. Cernv. FJ, (1987). Breathing pattern during exercise in young Black and Caucasian subjects. J Appl Physiol 62: 2220-2223. " " " .... Cheng. JO: Lee. WC: K lein MA: Cheng. GZ: Jhanwar. SC: Testa. JR. (1999). Frequent mutations o f NF2 and allelic loss from chromosome band 22ql2 in malignant mesothelioma: Evidence for a two-hit mechanism o f NF2 inactivation. Genes Chromosomes Cancer 24: 238-242. http://dx.doi.org/10.1002/(SICI)10982264(199903)24:3<238::AID-GCC9>3,0.CQ:2-M. Churg. A, (1982). Fiber counting and analysis in the diagnosis o f asbestos-related disease. Hum Pathol 13: 381-392. http://dx.doi.org/10,1016/S0046-8177(82)80227-X. Churg. A. (1988). Chrysotile, tremolite, and malignant mesothelioma in man. Chest 93: 621-628. http://dx.doi.orgU0.1378/chest.93.3,621. Churg, A: Wright. J: Gilks, B: DePaoli. L, (1989). Rapid short-term clearance o f chrysotile compared with amosite asbestos in the guinea pig. Am Rev Respir Dis 139: 885-890. Churg. A, (1994). Deposition and clearance o f chrysotile asbestos. Ann Occup Hyg 38: 625-633. http://dx, doi.org/10.1093/annliv g/3 8,4 ,625. Churg. A: Vedal. S. (1994). Fiber burden and patterns o f asbestos-related disease in workers with heavy mixed amosite and chrysotile exposure. Am J Respir Crit Care Med 150: 663-669. Claeskens. G: Hiort. NL, (2008). Model selection arid model averaging. In. Cambridge. England: Cambridge University Press. Clements. M: Berry, G: Shi. J: Ware. S: Yates. D: Johnson A. (2007). Projected mesothelioma incidence in men in New South Wales. Occup Environ Med 64: 747-752. http://dx.doi.org/10.1136/oem.2006.031823. This document is a draftfo r review purposes only and does not constitute Agency policy. 7-6 DRAFT--DO NOT CITE OR QUOTE C offia DL; Palekar. LD: Cook. PM. (1983). Correlation of in vitro and in vivo methods by means o f mass dose and fiber distribution for amosite and fibrous ferroactinolite. Environ Health Perspect 51: 49-53. C oia PG: Roggli. VL; Brody. AR, (1992). Deposition, clearance, and translocation of chrysotile asbestos from peripheral and central regions of the rat lung. Environ Res 58: 97-116. http://dx.doi.org/10.1016/S00139351(05)80207-5. C oia PG: Roggli. VL: Brodv, AR, (1994). Persistence o f long, thin chrysotile asbestos fibers in the lungs o f rats. Environ Health Perspect 102: 197-199. Condie, LW. (1983). Review o f published studies o f orally administered asbestos. Environ Health Perspect 53: 3-9. Constantopoulos. SH: Goudevenos, JA: Saratzis, N: Langer, AM: Selikoff, IJ: Moutsopoulos. HM, (1985). Metsovo hmg: Pleural calcification and restrictive lung function in northwestern Greece: Environmental exposure to mineral fiber as etiology. Environ Res 38: 319-331. http://dx.doi.org/10.1016/0013-9351(85)90096-9. Cook, PJ; Doll, R; Fellingham, SA, (1969). A mathematical model for the age distribution o f cancer in man. Int J Cancer 4: 93-112. http://dx.doi.org/10.1002/iic.2910040113. Cook, PM: O lsoa GF, (1979). Ingested mineral fibers: Elimination in human urine. Science 204: 195-198. http://dx.doi.org/10.1126/science.219478. Cooper, SP: Fraire. AE: Buffler, PA: Greenberg. SD: Langstoa C. (1989). Epidemiologic aspects o f childhood mesothelioma. Pathol Immunopathol Res 8: 276-286. http://dx.doi.org/10.1159/000157156. Copes, R: Thomas, D: Becklake. MR. (1985). Temporal patterns o f exposure and nomnalignant pulmonary abnormality in Quebec chrysotile workers. Arch Environ Health 40: 80-87. Copley, SJ: Wells, AU: Rubens, MB: Chabat, F: Sheehaa RE: Musk, AW: Hansell, DM, (2001). Functional consequences o f pleural disease evaluated with chest radiography and CT. Radiology 220: 237-243. Cox. DR, (1972). Regression models and life-tables. J Roy Stat Soc B 34: 187-220. Cristaudo, A: Foddis, R: Vivaldi, A: Buselli, R: Gattini, V: Guglielmi, G: Cosentino, F: Ottenga, F: Ciancia, E: Libener, R; Filiberti, R: Neri, M: Betta, P: Tognoa M: Mutti, L: Puntoni, R, (2005). SV40 enhances the risk o f malignant mesothelioma among people exposed to asbestos: A molecular epidemiologic casecontrol study. Cancer Res 65: 3049-3052. http://dx.doi.org/10.1158/0008-5472.CAN-04-2219. Crump. KS: Howe. RB, (1985) A review o f methods for calculating statistical confidence limits in low dose extrapolation. InD B Clayson; D Krewski; I Munro (Eds.), Toxicological risk assessment: Vol 1 Biological and statistical criteria (pp. 187-203). Boca Raton, FL: CRC Press, Inc. Cullea MR: Barnett, MJ: Balmes, JR; Cartmel, B: Redlicli, CA: Brodkia CA: Bamliart, S: Rosenstock, L: G oodiata GE: Haimnar, SP: Thomquist, MD: Omenn. GS. (2005). Predictors o f lung cancer among asbestos-exposed men in the beta-carotene and retinol efficacy trial. Am J Epidemiol 161: 260-270. http://dx.doi.org/10.1093/aie/kwi034. Cunningham. HM: Pontefract. RD. (1974). Placental transfer o f asbestos [Letter], Nature 249: 177-178. http://dx.doi.org/10.1038/249177a0. Dai. YT: Yu, CP. (1998). Alveolar deposition o f fibers in rodents and humans. J Aerosol Med 11: 247 -258. http://dx.doi.org/10.1089/iaial998.ll.247. Dana. DW: Hurlbut. CS: Klein. C. (1977). Manual o f mineralogy. In (19th ed.). New York, NY: John Wiley and Sons. This document is a draftfor review purposes only and does not constitute Agency policy. 7-7 DRAFT--DO NOT CITE OR QUOTE Davis. JM. (1989). Mineral fibre carcinogenesis: Experimental data relating to the importance o f fibre type, size, deposition, dissolution and migration. In Non-occupational exposure to mineral fibres (Vol. 90, pp. 33-45). Lyon, France: International Agency for Research on Cancer. Davis, JM, (1994). The role of clearance and dissolution in determining the durability or biopersistence o f mineral fibers. Environ Health Perspect 102: 113-117. Davis. JMG: Addison. J: B oltoa RE: Donaldson. K: Jones. AD: Miller. BG. (1985). Inhalation studies on the effects o f tremolite and brucite dust in rats. Carcinogenesis 6: 667-674. http://dx.doi.Org/10.1093/carcin/6.5.667. Davis. JMG: Addison. J: McIntosh. C: Miller. BG: N ivea K, (1991). Variations in the carcinogenicity o f tremolite dust samples o f differing morphology. Ann N Y Acad Sci 643: 473-490. http://dx.doi.org/10,1111/i,17496632,1991,tb24497.x. Davis. L: Martin. T: Kligler. B, (1992). Use o f death certificates for mesothelioma surveillance. Public Health Rep 107: 481-483. de Klerk. NH: Armstrong. BK: Musk. AW: Hobbs. MST, (1989). Cancer mortality in relation to measures o f occupational exposure to crocidolite at Wittenoom Gorge in Western Australia. Br J Ind Med 46: 529-536. http://dx.doi.Org/10.1136/oem.46.8.529. Delclos. GL: W ilsoa RK: Bradley, BL, (1990). Influence o f smoking on radiographic profusion and pleural changes in asbestos-exposed subjects. J Occup Med 32: 577-581. Dement. JM: Harris. RL, (1979). Estimates o f pulmonary and gastrointestinal deposition for occupational fiber exposure. Washington, DC: National Institute o f Occupational Safety and Flealth. http://www.ntis.gov/search/product.aspx? A B B R = P B 8 0 1 4 9 6 4 4 . Dement. JM; Welch. L; Bingham. E; Cameron. B: Rice. C: Quinn. P; R ingea K, (2003). Surveillance o f respiratory diseases among coastmction and trade workers at Department o f Energy- nuclear sites. Am J Ind Med 43: 559-573. http://dx.doi.org/10.1002/aiim, 10226. Dianzani. I: Gibello. L: Biava. A: Giordano. M: Bertolotti. M: Betti. M: Ferrante. D: Gnarrera. S: Betta. GP: Mirabelli. D: Matullo. G: Magnani. C. (2006). Polymorphisms in DN A repair genes as risk factors for asbestos-related malignant mesothelioma in a general population study. Mutat Res-Fundam Mol Mech Mutagen 599: 124-134. http://dx.doi.Org/10.1016/i.mrfmimn.2006.02.005. Ding. M: Dong. Z: Chea F: Pack. D: Ma. W: Ye. J: Shi. X: Castranova. V: Vallvathan. V, (1999). Asbestos induces activator protein-1 transactivation in transgenic mice. Cancer Res 59: 1884-1889. DiPaolo. JA: DeMarinis. AJ: Doniger. J, (1983). Asbestos and benzo(a)pyrene synergism in the transformation of Syrian hamster embryo cells. Pharmacology 27: 65-73. http://dx,doi.org/10 ,1159/000137837. D odsoa R: Williams. M: Com. C: Brollo. A: Bianchi. C. (1990). Asbestos content o f lung tissue, lymph nodes, and pleural plaques from former shipyard workers. Am Rev Respir Dis 142: 843-847. D odsoa R: O'Sullivan. M: Huang. J: Holiday, D: Haimnar. S. (2000a). Asbestos in extrapulmonary sites: Omentum and mesentery. Chest 117: 486-493. http://dx.doi.org/10,1002/aiim. 10164. D odsoa R: O'Sullivaa M: Brooks. D: Bruce. J. (2001). Asbestos content o f omentum and mesentery in nonoccupationally exposed individuals. Toxicol Ind Health 17: 138-143. Dodson. R: O'Sullivaa M: Brooks. D: Haimnar. S. (2003). Quantitative analysis o f asbestos burden in women with mesothelioma. Am J Ind Med 43: 188-195. http://dx.doi.org/10 ,1002/aiim, 10 164. This document is a draftfor review purposes only and does not constitute Agency policy. 7-8 DRAFT--DO NOT CITE OR QUOTE Dodson. RF: O'Sullivan. M: Cora CJ. (1996). Relationsliips between ferraginous bodies and uncoated asbestos fibers in lung tissue. Arch Environ Health 51: 462-466. http://dx.doi.org/10.1080/00039896.1996.9936Q47. Dodson. RE: Huang, J; Bruce. JR. (2000b). Asbestos content in the lymph nodes o f nonoccupationally exposed individuals. Am J Ind Med 37: 169-174. http://dx.doi.org/10,1002/(SICI)10970274(200002)37:2<169::AID-AJIM2>3.0.CQ:2-V. Dodson. RF: Graef. R: Shepherd. S: O'Sullivan. M: Levin. J. (2005). Asbestos burden in cases o f mesothelioma from individuals from various regions o f the United States. Ultrastruct Pathol 29: 415-433. http://dx.doi.org/10.1080/01913129Q945682. Popp. E: Yadav. S: Ansari. FA: Bhattacharya. K: von Recklinghausen. U: Rauen, U: Rodelsperger. K: Shokouhi. B: Geh. S: Rahman. O, (2005). ROS-mediated genotoxicity o f asbestos-cement in mammalian lung cells in vitro. Part Fibre Toxicol 2: 9. http://dx.doi.org/10.1186/1743-8977-2-9. Driscoll. KE: Maurer. JK: Higgins. J: Povnter, J. (1995). Alveolar macrophage cytokine and growth factor production in a rat model o f crocidolite-induced pulmonary inflammation and fibrosis. J Toxicol Environ Health 46: 155-169. http://dx.doi.org/10.1080/15287399569532Q26. Driscoll. KE: Carter. JM: Howard. BW: Hassenbein, DG: Pepelko. W: Baggs. RB: Oberdorster. G. (1996). Pulmonary inflammatory, chemokine, and mutagenic responses in rats after subchronic inhalation of carbon black. Toxicol Appl Pharmacol 136: 372-380. http://dx.doi.org/10.1006/taap.1996.0Q45. Driscoll. KE: Devo. LC: Carter. JM: Howard. BW: Hassenbein. DG: Bertram. TA, (1997). Effects o f particle exposure and particle-elicited inflammatory cells on mutation in rat alveolar epithelial cells. Carcinogenesis 18:423-430. ' Dumortier. P; Coplu. L: De Maertelaer. V: Emri. S: Baris. I: De Vuvst. P. (1998). Assessment o f environmental asbestos exposure in Turkey by bronclioalveolar lavage. Am J Respir Crit Care Med 158: 1815-1824. Duncan. K: Gliio. A: Dailey. L: Bern. A: Gibbs-Floumov. E: Padilla-Carlin. D: Roggli. V: D evlia R. (2010). Effect o f size fractionation on the toxicity o f amosite and Libby amphibole asbestos. Toxicol Sci 118: 420-434. http://dx.doi.org/10.1093/toxsci/kfq281. Eastmond. DA: Hartwig. A: Anderson. D: Anwar. WA: Cimino. MC: Dobrev. I: Douglas. GR: Nolnni. T: Plhllips. DLL Vickers. C. (2009). Mutagenicity testing for chemical risk assessment: Update o f the WHO/1PCS harmonized scheme. Mutagenesis 24: 341-349. http://dx.doi.org/10.1093/inutage/gep014. Edwards. GH: Lynch, JR. (1968). The method used by the U.S. public health service for enumeration o f asbestos dust on membrane filters. Aim Occup Hyg 11: 1-6. http://dx.doi.org/10.1093/annhvg/l 1,1.1. Ehrlich. R: Lilis. R: Chan. E: Nicholson. WJ: Selikoff. IJ, (1992). Long term radiological effects o f short term exposure to amosite asbestos among factory workers. Br J Ind Med 49: 268-275. Epler, GR: Fitz Gerald. MX: Gaensler. EA: Carrington CB. (1980). Asbestos-related disease from household exposure. Respiration 39: 229-240. http://dx.doi.org/10.1159/000194221. Erdinc. M: Erdinc. E: Cok. G; Polatli. M, (2003). Respiratory impairment due to asbestos exposure in brake-lining workers. Environ Res 91: 151-156. http://dx.doi.org/10.1016/S0Q 13-9351(02)00063-4. Ewing. WM: Hays, SM: Hatfield. R: Longo. WE: Millette. JR. (2010). Zonolite attic insulation exposure studies. Int J Occup Environ Health 16: 279-290. This document is a draftfor revie\vpurposes only and does not constitute Agency policy. 7-9 DRAFT--DO NOT CITE OR QUOTE Fattman. CL: Tan. RJ: Tobolewski. JM; Oury. TD. (2006). Increased sensitivity to asbestos-induced lung injury in mice lacking extracellular superoxide dismutase. Free Radic Biol Med 40: 601-607. http://dx.doi.Org/10.1016/i.freeradbiomed.2005.09.030. Finkelsteia MM, (1985). A study of dose-response relationships for asbestos associated disease. Br J Ind Med 42: 319-325. Finkelsteia MM: Dufresne. A, (1999). Inferences on the kinetics o f asbestos deposition and clearance among chrysotile miners and millers. Am J Ind Med 35: 401-412. http://dx.doi.org/10.1002/(SICI) 10970274(199904)35:4<401::AID-AJIM12>3.0.CQ:2-4. Foddis, R: De Rienzo. A: Broccoli. D: Bocchetta. M: Stekala. E: Rizzo. P; Tosolini. A: Grobelnv. JV: Jhanwar. SC: Pass. HI: Testa. 1R: Carbone. M. (2002). SV40 infection induces telomerase activity in human mesothelial cells. Oncogene 21: 1434-1442. http://dx.doi.org/10,1038/si.one. 1205203. Gao. F: Koenitzer. JR: Tobolewski. JM: Jiang. D: Liang. J: Noble. PW: Oury. TD. (2008). Extracellular superoxide dismutase inhibits inflammation by preventing oxidative fragmentation of hvaluronan. J Biol Chem 283: 6058-6066. http://dx.doi.org/10.l674/ibc.M709273200. Careia-Closas, M: Christiani. DC. (1995). Asbestos-related diseases in construction carpenters. Am J Ind Med 27: 115-125. http://dx.doi.org/10.1002/aiim .4700270111. Gasparrini. A: Pizzo. AM: Gorini. G: Costantini, AS: Silvestri. S: Ciapini. C: Innocenti. A: Berry, G. (2008). Prediction o f mesothelioma and lung cancer in a cohort o f asbestos exposed workers. Eur J Epidemiol 23: 541-546. http://dx.doi.org/10.1007/slQ654-008-9257-z. Ghio. AJ: Kennedy, TP: Whorton, AR; Crumbliss. AL: Hatch. GE: Hoidal. 1R. (1992). Role o f surface complexed iron in oxidant generation and lung inflammation induced by silicates. Am J Physiol 263: L 511-L518. Gibbs. AR: Stephens. M: Griffiths. DM: Blight. BJ: Poolev. FD. (1991). Fibre distribution in the lungs and pleura of subjects with asbestos related diffuse pleural fibrosis. Br J Ind Med 48: 762-770. http://dx.doi.org/10.1136/oem .48.ll.762. Gilmartia D. (1979). The serratus anterior muscle on chest radiographs. Radiology 131: 629-635. http://dx.doi.org/10.1148/131.3.629. Graham. DR: Chamberlaia MJ: Hutton. L: King. M: Morgaa WK. (1990). Inhaled particle deposition and body habitus. Br J Ind Med 47: 38-43. Green GM, (1973). Alveolobronchiolar transport mechanisms. Arch Intern Med 131: 109-114. Griffis. LC: Pickrell. JA: Carpenter. RL: Wolff, RK: M cA llen SJ: Yerkes. KL. (1983). Deposition o f crocidolite asbestos and glass microfibers inhaled by the beagle dog. Am Ind Hyg Assoc J 44: 216-222. http://dx.doi.org/10.1080/152986683914Q4671. Gunter. ME: Sanchez. MS. (2009). Amphibole forensics: Using the composition o f ampliiboles to determine their source, the Libby, Montana, example. American Mineralogist 94: 837-840. h ttp://d x.d oi.org/10.2138/am .2009.3224. Guzelian. PS: Henry, CJ: Olin, SS. (Eds.). (1992). Similarities and differences between children and adults: Implications for risk assessment. In PS Guzelian; CJ Henry; SS Olin (Eds.). New York: ILSI Press. Hamilton. R: Holian. A: Morandi. M. (2004). A comparison o f asbestos and urban particulate matter in the in vitro modification o f human alveolar macrophage antigen-presenting cell function. Exp Lung Res 30: 147-162. This document is a draftfor review purposes only and does not constitute Agency policy. 7-10 DRAFT--DO NOT CITE OR QUOTE Hammond. EC; Selikoff. 1J: Seidman. H. (1979). Asbestos exposure, cigarette smoking and death rates. Aim N Y Acad Sci 330: 473-490. Hanahan. D: Weinberg. RA, (2011). Hallmarks o f cancer: The next generation. Cell 144: 646-674. http://dx.doi.Org/10.1016/i.cell.2011.02.013. Hansen. J: de Klerk. NH: Musk. AW: Hobbs. MST, (1998). Enviromnental exposure to crocidolite and mesothelioma: Exposure-response relationships. Am J Respir Crit Care Med 157: 69-75. Hague, A: Kanz. M: Mancuso. M: Williams, G: Dodson. R, (1991). Asbestos in the lungs o f children. Ann N Y Acad Sci 643: 419-429. Hague, AK: Kanz. ME. (1988). Asbestos bodies in children's lungs: An association with sudden infant death syndrome and bronchopulmonary dysplasia. Arch Pathol Lab Med 112: 514-518. Hague. AK: Mancuso. MG: Williams. MG: Dodson. RF, (1992). Asbestos in organs and placenta o f five stillborn infants suggests transplacental transfer. Environ Res 58: 163-175. Hague, AK: Vrazel. DM: Burau. KD: Cooper. SP: Downs. T, (1996). Is there transplacental transfer o f asbestos? A study o f 40 stillborn infants. Pediatr Pathol Lab Med 16: 877-892. Hague, AK: Vrazel. DM: Ucliida, T. (1998). Assessment o f asbestos burden in the placenta and tissue digests of stillborn infants in South Texas. Arch Environ Contain Toxicol 35: 532-538. http://dx.doi.org/10.1007/s00244990Q413. Hague, AK: Vrazel DM, (1998). Transplacental transfer of asbestos in pregnant mice. Bull Environ Contain Toxicol 60: 620-625. http://dx.doi.org/10.1007/sQ01289900671. Hague. AK: Ali. I: Vrazel. DM: Ucliida. T. (2001). Chrysotile asbestos fibers detected in the newborn pups following gavage feeding of pregnant mice. J Toxicol Environ Health A 62: 23-31. Harris, f.i Jr: Timbrel!. V. (1975). The influence o f fibre shape in lung deposition-mathematical estimates. Inhaled Particles 4: 75-89. Harris. RL,, Jr: Fraser. DA. (1976). A model for deposition o f fibers in the human respiratoiy system. Am Ind Hyg Assoc J 37:73-89. http://dx.doi.org/10.1080/00028897685Q7416. ' ' " Hasanoglu, HC: Yildirim, Z: Ennis, H: Kilic, T: Koksal. N, (2006). Lung cancer and mesothelioma in towns with enviromnental exposure to asbestos in Eastern Anatolia. Int Arch Occup Environ Health 79: 89-91. http://dx.doi.org/10.1007/s00420-005-001Q-6. Hauptmann. M: Polilabeln. H: Lubia JH: Jockel. KH: Ahrens. W: Briiske-Hohlfeld, I: Wichmann. HE, (2002). The exposure-time-response relationship between occupational asbestos exposure and lung cancer in two German case-control studies. Am J Ind Med 41: 89-97. Hawthorne. FC. (1981). Amphibole spectoscopy. Reviews in Mineralogy and Geochemistry 9: 103-139. Hei. TK: Piao. CO: He. ZY; Vannais. D: Waldrera CA, (1992). Chrysotile fiber is a strong mutagen in mammalian cells. Cancer Res 52: 6305-6309. Hei. TK: Xu. A: Huang. SX: Zhao., Y, (2006). Mechanism o f fiber carcinogenesis: From reactive radical species to silencing o f the beta igH3 gene. Inlial Toxicol 18: 985-990. http://dx.doi.org/10.1080/0895837060083531Q. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 7-11 DRAFT--DO NOT CITE OR QUOTE Heid. IM: Kticlienhoff H: Miles. J: Kreienbrock. L: Wjdunami. HE. (2004). Two dimensions o f measurement error: Classical and Berkson error in residential radon exposure assessment. J Expo Anal Environ Epidemiol 14: 365-377. http://dx.doi.org/10.1038/si.iea.750Q332. Hein. MJ: Stavner. LT: Lehman. E; Dement. JM. (2007). Follow-up study o f chrysotile textile workers: Cohort mortality and exposure-response. Occup Environ Med 64: 616-625. http://dx.doi.org/10.1136/oein.2006.0310Q5. Henderson. VL: Enterline. PE. (1979). Asbestos exposure: Factors associated with excess cancer and respiratory disease mortality. Ann N Y Acad Sci 330: 117-126. http://dx.doi.Org/10.llll/i.1749-6632.1979.tbl8712.x. HEW. (U.S. Department o f Health. Education and Welfare). (1979). Smoking and health: A report o f the Surgeon General. (PHS 79-50066). Wasliingtoa DC: Government Printing Office. HHS. (U.S. Department o f Health and Human Services). (1990). The health benefits o f smoking cessation: A report o f the surgeon general. Rockville. MD. http://profiles.nlm.nih.gov/ps/access/NNBBCV.pdf. Hillegass. JM: Shukla. A: MacPherson. MB: Lathrop. SA: Alexeeva. V: Perkins. TN: van der Vliet. A: Vacek. PM: Gunter. ME: M ossm aa BT, (2010). Mechanisms o f oxidative stress and alterations in gene expression by Libby six-mix in human mesothelial cells. Part Fibre Toxicol 7: 26. http://dx.doi.org/10.1186/1743-8977-726. ' Hillerdal. G. (1983). Malignant mesothelioma 1982: Review o f 4710 published cases. Br J Dis Chest 77: 321-343. http://dx.doi.org/10.1016/0007-0971(83)90068-2. Hillerdal. G. (1994). The human evidence: Parenchymal and pleural changes. Ami Occup Flyg 38: 561-567. http://dx, doi.org/'l0 , 1093/annlivg/38 .4 ,561. Hiraoka. T: Qhkura. M: Morinaga. K: Kolwama. N: Shimazu. K: Ando. M. (1998). Anthophyllite exposme and endemic pleural plaques in Kumamoto. Japan. Scand J Work Environ Health 24: 392-397. Hirvonen. A: Tuirnala. J: Ollikainen. T: Linnaimnaa. K: Kinnula. V. (2002). Manganese superoxide dismutase genotypes and asbestos-associated pulmonary disorders. Cancer Lett 178: 71-74. http://dx.doi.org/10.1016/S0304-3835(01)00819-9. Hodgson. J: McElvennv. D: D am toa A: Price. M: Peto. J. (2005). The expected burden o f mesothelioma mortality in Great Britain from 2002 to 2050. Br J Cancer 92: 587-593. http://dx.doi.org/10.1038/si.bic.6602307. Holt. P. (1982). Translocation of asbestos dust through the bronchiolar wall. Environ Res 27: 255-260. http://dx.doi.orgU0.1016/0013-9351(82)90080-9. Holt. PF, (1983). Translocation of inhaled dust to the pleura. Environ Res 31: 212-220. http://dx.doi.org/10,1016/0013-9351(83)90077-4. Horton. K: Kapil. V: Larsoa T; Muravov. O: Melnikova. N: Anderson. B. (2006). A review o f the federal government's health activities in response to asbestos-contaminated ore found in Libby, Montana. Inlial Toxicol 18: 925-940. http://dx.doi.org/10.1080/0895837060Q835161. Hosmer, DW. Jr: Lameshow. S. (2000). Applied logistic regression. In. New York, N Y : John Wiley & Sons. Huang. SX; Jairrand, MC: Kamp, DW: Whvsner. J: Hei. TIC (2011). Role o f mutagenicity in asbestos fiber-induced carcinogenicity and other diseases. J Toxicol Environ Health B Crit Rev 14: 179-245. http://dx.doi.org/10.1080/10937404.2011.556051. Huncharck. M. (2002). Non-asbestos related diffuse malignant mesothelioma. Tumori 88: 1-9. This document is a draftfor review purposes only and does not constitute Agency policy. 7-12 DRAFT--DO NOT CITE OR QUOTE ICRP. (International Commission on Radiological Protection). (1994). Human respiratory tract model for radiological protection: A report o f a task group of the International Commission on Radiological Protection. ICRP Publication 66. Ami ICRP 24: 1-482. ILO. (International Labour Organization). (1971). International classification of radiographs o f pneumoconioses (revised, 1968). In (Red. Ed. 1968 ed., Vol. 22). Geneva, Switzerland: International Labour Office. ILO, (International Labour Organization). (1980). Guidelines for the use o f the ILO international classification of radiographs o f pneumoconioses. In (Vol. 22). Geneva, Switzerland: International Labour Office. ILO, (International Labour Organization). (2002). International classification of radiographs o f pneumoconioses. In (Rev. Ed. 2000 ed., Vol. 22). Geneva, Switzerland: International Labour Office. Inase. N: Takavama. S: Nakavama, M: Miura. H: Kimula, Y, (1991). Pleural mesothelioma after neighborhood exposure to asbestos during cliildhood. Jpn J Med 30: 343-345. http://dx.doi.org/10.2169/intemalmedicinel962.30.343. IPCS. (International Programme on Chemical Safety). (1986). Asbestos and other natural mineral fibres (Vol. 53). Geneva, Switzerland: World Health Organization, http://www.inchem.org/documents/ehc/ehc/ehc53,htm. Isaacs, KK: Martonea TB, (2005). Particle deposition in children's lungs: Theory and experiment. J Aerosol Med 18: 337-353. http://dx.doi.org/10.1089/iam.2005.18.337. lain, J ,, (1974). Dose-response basis for setting a quartz threshold limit value. Arch Environ Health 29: 338-340. Jakobsson, K: Stromberg, U: Albin, M; Welinder, H: Hagmar, L, (1995). Radiological changes in asbestos cement workers. Occup Environ Med 52: 20-27. Jarabek, AM: Asgliarian. B: Miller. FJ. (2005). Dosimetric adjustments for interspecies extrapolation o f inhaled poorly soluble particles (PSP). Inlial Toxicol 17: 317-334. http://dx.doi.org/10.1080/0895837059Q929394. Jarvliolm B, (1992). Pleural plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21: 1180 1184. ' ' Janrand, MC. (1997). Mechanisms o f fiber-induced genotoxicity. Environ Health Perspect 105: 1073-1084. Jones. AD: McMillan. CH: Johnston. AM: McIntosh. C: Cowie, H: Bolton. RE: Borzucki, G: Vincent. JH. (1988). Pulmonary clearance o f UICC amosite fibres inhaled by rats during chronic exposure at low concentration. BrJInd Med 45: 300-304. Jones. JS. (2002). The pleura in health and disease. Lung 179: 397-413. http://dx.doi.org/10.1007/s004080000Q75. Kamp. DW: Weitzman. SA, (1999). The molecular basis o f asbestos induced lung injury [Review]. Thorax 54: 638 652. http://dx.doi.Org/10.1136/tlix.54.7.638. Kane. MJ: Chahinian. AP: Holland. JF. (1990). Malignant mesothelioma in young adults. Cancer 65: 1449-1455. http://dx.doi.org/10.1002/1097-0142( 19900315)65:6<1449::AID-CNCR2820650633>3.0.CQ:2-0. Karlsora EW: Mandl. LA: Aweh. GN: Grodsteira F, (2003). Coffee consumption and risk o f rheumatoid arthritis. Arthritis Rheum 48: 3055-3060. http://dx.doi.org/10,1002/art. 11306. Kelly, J: Pratt. G: Johnson. J: Messing. R, (2006). Community exposure to asbestos from a venniculite exfoliation plant inN E Minneapolis. Inlial Toxicol 18: 941-947. littp://dx.doi.org/10.1080/08958370600834883. This document is a draftfor review purposes only and does not constituteurgencypolicy. 7-13 DRAFT--DO NOT CITE OR QUOTE Kilbum. KH: Lilis. R: Anderson. HA; Bovlen. CT; Einsteia HE: Johnson. SJS: Warshaw. R. (1985). Asbestos disease in family contacts o f shipyard workers. Am J Public Health 75: 615-617. Kilburn. KH: Warshaw. RH, (1991). Difficulties o f attribution o f effect in workers exposed to fiberglass and asbestos. Am J Ind Med 20: 745-751. http://dx.doi.ora/10.1002/aiim.4700200605. Kim. CS; Kang. TC. (1997). Comparative measurement o f lung deposition of inhaled fine particles in normal subjects and patients with obstructive airway disease. Am J Respir Crit Care Med 155: 899-905. Kim. I-I: Yasui. Y : Burstva I. (2006). Attenuation in risk estimates in logistic and Cox proportional-hazards models due to group-based exposure assessment strategy. Ann Occup Hyg 50: 623-635. http://dx.d oi.ora/10.10 9 3 /a n n h y g /m e l0 2 1 . Kimizuka, G: Ohwada. H: Havashi. Y, (1987). Co-carcinogenic effect o f asbestos and benzo(a)pvrene in the lung of hamster. Acta Pathol Jpn 37: 465-474. http://dx.doi.org/10.I l l 1/i. 1440-1827,1987,tb00380.x. Kleinbaum. DG: Klein. M. (1996). Survival analysis: A self-learning text. In. New York: Springer-Verlag. Kleymenova EV: Bianchi. AA: Kiev. N: Pvlev. LN: Walker. CL. (1997). Characterization o f the rat neurofibromatosis 2 gene and its involvement in asbestos-induced mesothelioma. Mol Carcinog 18: 54-60. http://dx.doi.org/10.1002/(SICI) 1098-2744(199701) 18:1<54::AID-MC7>3.0.CQ:2-R. Kliment. CR: Englert, JM: Gochuico. BR: Guoving. Y: Kaminski. N; Rosas. I: Ourv. TD. (2009). Oxidative stress alters syndecan-1 distribution in lungs with pulmonary fibrosis. J Biol Chem 284: 3537-3545. http://dx.doi.Org/10.1074/ibc.M807001200. Kohvama. N: Suzuki. Y. (1991). Analysis o f asbestos fibers in lung parenchyma, pleural plaques, and mesothelioma tissues o f North American insulation workers. Ami N Y Acad Sci 643: 27-52. http://dx.doi.Org/10.llll/i.1749-6632.1991.tb24442.x. Kopylev, L: Sullivan. PA: Vinikoor. LC: Bateson. TF. (2011). Monte Carlo analysis o f impact of underascertaimnent o f Mesothelioma cases on underestimation o f risk. 4: 45-53. Koskinen. K: Zitting. A: Tossavainea A: Rinne. JP: Roto. P: Kivekas. J: Reiiula. K: Huuskonen. MS. (1998). Radiographic abnormalities among Finnish construction, shipyard and asbestos industry workers. Scand J Work Environ Health 24: 109-117. Komis. SP: Parker. DL: Bender, AP: Williams. AN. (1991). Effects o f asbestos-related pleural disease on pulmonary function. Scand J Work Environ Health 17: 179-183. Kroczvnska. B; Cutrone. R: Bocchetta, M: Yang. H: Elmishad, AG: Vacek. P: Ramos-Nlno. M: Mossman, BT: Pass. HI: Carbone. M, (2006). Crocidolite asbestos and SV40 are cocarcinogens in human mesothelial cells and in causing mesothelioma in hamsters. PNAS 103: 14128-14133. http://dx.doi.org/10.1073/pnas.06045441Q3. Kroschwitz. JI: Seidel. A. (2010). Kirk-Othmer concise encyclopedia o f chemical technology. In (5th ed.). New York: John Wiley and Sons. Krowke. R: Bluth. V: Merker. JH: Neubert. D. (1983). Placental transfer and possible teratogenic potential of asbestos in mice [Abstract]. Teratology 32: 26A-27A. Landi. M: Dracheva. T: Rotuimo, M: Figueroa. J: Liu. H: Dasgupta. A: Mann. F: Fukuoka. J: Hames. M: Bergea A: Murphy, S: Yang. P: Pesatori. A: Consonni. D: Bertazzi. P: Wacholder. S: Shjh. J: Caporaso. N: Jea J. (2008). Gene expression signature o f cigarette smoking and its role in lung adenocarcinoma development and survival. PLoS ONE 3: el651. http://dx.doi.org/10.137l/ioumal.pone.0001651. This document is a draftfo r review purposes only and does not constitute Agency policy. 7-14 DRAFT--DO NOT CITE OR QUOTE Langcr. AM: Mackler. AD; Poolev. FD. (1974). Electron microscopical investigation o f asbestos fibers. Environ Health Perspect 9: 63-80. Langer. AM; N olaa RP: Constantopoulos. SH; Moutsopoulos. HM. (1987). Association o f Metsovo lung and pleural mesothelioma with exposure to tremolite-containing whitewash. Lancet 329: 965-967. http://dx.doi.org/10.1016/S0140-6736(87)90305-9. Langseth. H: Johansen. BY: Nesland. JM; Kiaerheixm K, (2007). Asbestos fibers in ovarian tissue from Norwegian pulp and paper workers. Int J Gynecol Cancer 17: 44-49. http://dx.doi.org/10. I l l 1/i. 15251438.2006.00768.x. Larsen, ES. (1942). Alkalic rocks o f IronPIill, Gunnison County, Colorado. Washington, DC: U.S. Geological Survey. Larson. T: Mever, C: Kapil. V: Gurney, J: Tarver. R: Black, C: Lockev, J. (2010a). Workers with Libby amphibole exposure: retrospective identification and progression o f radiographic changes. Radiology 255: 924-933. http://dx. doi.org/10.1148/radiol. 10091447. Larson. TC: Antao, VC: Bove, FJ. (2010b). Venniculite worker mortality: Estimated effects o f occupational exposure to Libby amphibole. J Occup Environ Med 52: 555-560. http://dx.doi.org/10.1097/JQM.0b013e3181dc6d45. Leake, 6 , , (1978). Nomenclature of amphiboles. Mineral Mag 42: 533-563. Leake. B , , E,: Woolley, A , , R.: Arps. C .. E. S.: Birch, W .. D.: Gilbert, M ,, C.: Grice. J .. D.: Hawthorne, F , , C.: Kato, A ,K i s c h , H , , J,: Krivovichev. V . . G.: Linthout, K ,L a ir d . J , .: Mandarino, J ,.: Maresch. W .. V.: Nickel, E , , H,: Rock, N , , M, S.: Schumacher. J ,, C.: Smith. D . . C.: Shephensoa N , , C. N,: Ungarehi. L, .: Whittake. E . . J. W.: Youzhi, G ,. (1997). Nomenclature o f amphiboles: Report o f the Subcommittee on Amphiboles o f the International Mineralogical Association Commission on New Minerals and Mineral Names. Mineral Mag 61: 295-321. Lentz. TJ: Rice. CH: Succop, PA: Lockev. JE: Dement. JM: LeMasters. GK, (2003). Pulmonary deposition modeling with airborne fiber exposure data: a study o f workers manufacturing refractory ceramic fibers. Appl Occup Environ Hyg 18: 278-288. http://dx.doi.org/10.1080/10473220301404. Li. FP: Lokich. J: Lapev, J: Neptune, WB: Wilkins, EW. (1978). Familial mesothelioma after intense asbestos exposure at home. JAMA 240: 467. http://dx.doi.org/10.1001/iama. 1978,03290050057022. Li. FP: Dreyfus, MG; Antman. KH, (1989). Asbestos-contaminated nappies and familial mesothelioma. Lancet 333: 909-910. http://dx.doi.org/10.1016/SQ 140-6736(89)92916-4. Lieben. J: Pistawka. H, (1967). Mesothelioma and asbestos exposure. Arch Environ Health 14: 559-563. Lilis, R: Miller. A: Godbold. J: Chan. E: Benkert, S: Selikoff. IJ, (1991). The effect o f asbestos-induced pleural fibrosis on pulmonary function: quantitative evaluation. In The third wave o f asbestos disease: Exposure to asbestos in place (Vol. 643, pp. 162-168). New York, N Y : New York Academy o f Sciences. L ia F: Liu. Y: Keshava. N; Li. S. (2000). Crocidolite induces cell transformation and p53 gene mutation in BALB/c-3T3 cells. Teratog Carcinog Mutagen 20: 273-281. http://dx.doi.org/10.1002/15206866(2000)20:5<273::AID-TCM3>3.0.CQ:2-1. Ling. S: Fried. L: Garrett. E; ITirsch, R: Guralnik, J: Hochberg. M. (2000). The accuracy o f self-report of physician diagnosed rheumatoid arthritis in moderately to severely disabled older women. Women's Health and Aging Collaborative Research Group. J Rheumatol 27: 1390-1394. This document is a draftfor review purposes only and does not constitute Agency policy. 7-15 DRAFT--DO NOT CITE OR QUOTE Lippmann, M. (1990). Effects o f fiber characteristics on lung deposition, retention, and disease. Environ Health Perspect 88: 311-317. http://dx.doi.org/10,1289/ehp,9088311. Lockev. JE: Brooks, SM; Jarabek, AM: Khourv. PR: McKay, RT: Carson. A: Morrison. JA: Wiot, JF: Spitz, HB. (1984). Pulmonary clianges after exposure to vermiculite contaminated with fibrous tremolite. Am Rev RespirDis 129: 952-958. Lockev. JE. (1985) Pulmonary hazards associated with vermiculite exposure. Master's Thesis, University of Cincinnati, Cincinnati, OH. Loli, P: Topinka, J: Georgiadis, P: Dusinska, M: I-Iurbankova, M: Kovacikova. Z; Volkovova, K: Wolff, T: Oesterle, D: Kvrtopoulos. S. (2004). Benzo[a]pyrene-enhanced mutagenesis by asbestos in the lung o f lambda-lacl transgenic rats. Mutat Res 553: 79-90. http://dx.doi.Org/10.1016/i.mrfmimn.2004.06.025. Lund. LG: Williams, MG: Dodson. RF: Aust. AE. (1994). Iron associated with asbestos bodies is responsible for the formation o f single strand breaks in phi X174 RFIDNA. Occup Environ Med 51: 200-204. Luo, S: Liu, X: Mu, S: Tsai, SP: Wen, CP. (2003). Asbestos related diseases from environmental exposure to crocidolite in Da-yao, China. I. Review o f exposure and epidemiological data. Occup Environ Med 60: 35 41; discussion 41-32. http://dx.doi.org/10.1136/oem.60.1.35. Magee, F: Wright. XL; Chan. N: Lawson, L: Churg, A. (1986). Malignant mesothelioma caused by childhood exposure to long-fiber low aspect ratio tremolite. Am J End Med 9: 529-533. http://dx.doi.org/10.1002/aiiin.47000906Q4. Magnani, C: Dalmasso, P; Biggeri, A; Ivaldi, C; Mirabelli, D; Terracini, B. (2001). Increased risk o f malignant mesothelioma o f the pleura after residential or domestic exposure to asbestos: A case-control study in Casale Monferrato, Italy. Environ Health Perspect 109: 915-919. http://dx.doi.org/10.1289/ehp.011Q9915. Magnani. C; Ferrante. D; Barone-Adesi, F: Bertolotti. M; Todesco. A; Mirabelli. D; Terracini. B. (2008). Cancer risk after cessation o f asbestos exposure: A cohort study o f Italian asbestos cement workers. Occup Environ Med 65: 164-170. http://dx.doi.org/10.1136/oem.20Q7.b32847. Malomi, W; Iosi, F; Falclii, M; Donelli, G. (1990). On the mechanism o f cell internalization o f chrysotile fibers: An immunocytochemical and ultrastructural study. Environ Res 52: 164-177. http://dx.doi.org/10.1016/S0Q139351(05)80251-8. Marsella, J; Liu. B; Vaslet, C; Kane, A. (1997). Susceptibility of p53-deficient mice to induction o f mesothelioma by crocidolite asbestos fibers. Environ Health Perspect 105: 1069-1072. Marsh. GM; Gula. MJ; Youk, AO; Buchanich. JM; Churg. A; Colby, TV. (2001). Historical cohort study o f US man-made vitreous fiber production workers: II. Mortality from mesothelioma. J Occup Environ Med 43: 757-766. Martensson. G; Larsson, S: Zettergrea L, (1984). Malignant mesothelioma in two pairs o f siblings: Is there a hereditary predisposing factor. Eur J Respir Dis 65: 179-184. Mastrangelo, G; Ballarin, MN; Bellini. E; Bicciato. F; Zannol. F; Gioffre, F; Zedde, A; Tessadri, G: Fedeli, U; Valentini, F; Scoizzato, L; Marangi, G: Lange, JH. (2009). Asbestos exposure and benign asbestos diseases in 772 formerly exposed workers: dose-response relationships. Am J Irid Med 52: 596-602. http://dx.doi.org/10.1002/aiim.2Q713. Mavall, FG: Jacobson, G: Wilkins. R. (1999). Mutations o f p53 gene and SV40 sequences in asbestos associated and non-asbestos-associated mesotheliomas. J Clin Pathol 52: 291-293. This document is a draftfor review purposes only and does not constitute Agency policy. 7-16 DRAFT--DO NOT CITE OR QUOTE McConnell. EE: Shefner. AM; Rust. JH: Moore. JA. (1983a). Chronic effects o f dietary exposure to amosite and chrvsotile asbestos in Syrian golden hamsters. Environ Health Perspect 53: 11-25. McConnell. EE; Rutter. HA; Ulland. BM; Moore. JA, (1983b). Chronic effects o f dietary exposure to amosite asbestos and tremolite in F344 rats. Environ Health Perspect 53: 27-44. McDonald. J: Edwards. C: Gibbs. A: Llovd. H: Poolev. F: Ross. D: Rudd. R. (2001). Case-referent survey o f young adults with mesothelioma: II. Occupational analyses. Arm Occup Hyg 45: 519-523. http://dx.doi.Org/10.1093/annliyg/45.7.519. McDonald. JC: McDonald. AD: Armstrong. B: Sebastien, P, (1986a). Cohort study o f mortality o f venniculite miners exposed to tremolite. Occup Environ Med 43: 436-444. http://dx.doi.Org/10.1136/oern.43.7.436. McDonald. JC: Sebastiea P: Armstrong. B. (1986b). Radiological survey o f past and present venniculite miners exposed to tremolite. Br J Ind Med 43: 445-449. http://dx.doi.Org/10.1136/oern.43.7.445. McDonald. JC: McDonald. AD: Sebastiea P: Mov, K, (1988). Health o f venniculite miners exposed to trace amounts o f fibrous tremolite. Occup Environ Med 45: 630-634. http://dx.doi.Org/10.1136/oern.45.9.630. McDonald. JC: Harris. J: Annstrong. B, (2002). Cohort mortality study o f venniculite miners exposed to fibrous tremolite: An update. Ann Occup Hyg 46: 93-93. http://dx.doi.org/10.1093/annhyg/46.suppl 1.93. McDonald. JC: Harris. J: Annstrong. B, (2004). Mortality in a cohort o f venniculite miners exposed to fibrous amphibole in Libby, Montana. Occup Environ Med 61: 363-366. http://dx.doi.org/10.1136/oern.2003.008649. McMillan. GH; Pethvbridge. RJ: Sheers. G. (1980). Effect o f smoking on attack rates o f pulmonary and pleural lesions related to exposure to asbestos dust. Br J Ind Med 37: 268-272. http://dx.doi.Org/10.1136/oern.37.3.268. Meek. ME: Grasso. P. (1983). An investigation o f the penetration o f ingested asbestos into the nonnal and abnonnal intestinal mucosa o f the rat. Food Chern Toxicol 21: 193-200. http://dx.doi.org/10.1016/Q2786915(83)90236-3. Meeker. GP: Bern, AM: Brownfield. IK: Lowers. HA: Sutlev. SJ: H oefea TM: Vance. JS. (2003). The composition and morphology o f amphiboles from the Rainy Creek Complex, near Libby, Montana. American Mineralogist 88: 1955-1969. Merchant. JA. (1990). Human epidemiology: A review o f fiber type and characteristics in the development of malignant and nonmalignant disease. Environ Health Perspect 88: 287-293. Metintas. M: Metintas. S: Hillerdal. G: Ucgrtn, I: Erginel, S: Alalas. F: Yildirim. H, (2005). Nonmalignant pleural lesions due to enviromnental exposure to asbestos: a field-based, cross-sectional study. Eur Respir J 26: 875-880. http://dx.doi.org/10.1183/09031936.05.001364Q4. Miller. A: Libs, R: Godbold. J: Chan. E: Wu, X: Selikoff. IJ. (1994). Spirornetric impairments in long-term insulators. Relationships to duration o f exposure, smoking, and radiographic abnormalities. Chest 105: 175 182. Miller. JA: Zurlo. JV. (1996). Asbestos plaques in a typical Veteran's hospital population. Am J Ind Med 30: 726 729. http://dx.doi.org/10.1002/(SICI) 1097-0274(199612)30:6<726::AID-AJIM9>3.0.CO:2-Q. Miserocchl, G: Sancini. G: Mantegazza, Chiappino, G. (2008). Translocation pathways for inhaled asbestos fibers. Environ Health Global Access Sci Source 7: 4. http://dx.doi.org/10.1186/1476-069X-7-4. This document is a draftfor review purposes only and does not constitute Agency policy. 7-17 DRAFT--DO NOT CITE OR QUOTE Moatamed. F; Lockcv. JE: Parry. WT. (1986). Fiber contamination o f venniculites: A potential occupational and environmental health hazard. Environ Res 41: 207-218. http://dx.doi.org/10.1016/S0013-9351(86)80183-9. Moolgavkar. SH; Turim. J; Alexander. DP: Lau. EC: Cushing. CA, (2010). Potencv factors for risk assessment at Libby, Montana. Risk Anal 30: 1240-1248. http://dx.doi.Org/10.llll/i.1539-6924.2010.01411.x. Morgan, A: Talbot. RJ: Holmes. A, (1978). Significance o f fibre length in the clearance o f asbestos fibres from the lung. Br J Ind Med 35: 146-153. http://dx.doi.Org/10.1136/oem.35.2.146. M ossm aa BT: Eastman, A: Landesmaa JM: Bresnick, E, (1983). Effects of crocidolite and chrysotile asbestos on cellular uptake and metabolism of benzo(a)pvrene in hamster tracheal epithelial cells. Environ Health Perspect 51: 331-335. Mossman, B T : Eastman, A: Bresnick, E. (1984). Asbestos andbenzo[a]pyrene act synergistically to induce squamous metaplasia and incorporation o f [3H]thymidine in hamster tracheal epithelium. Carcinogenesis 5: 1401-1404. http://dx.doi.Org/10.1093/carcin/5.ll.1401. Mossman. BT; Cameroa GS; Yotti, LB. (1985). Cocarcinogenic and tumor promoting properties o f asbestos and other minerals in tracheobronchial epithelium. Carcinogenesis 8: 217-238. Mossman. BT: Bonn, PJ: Castranova, V: Costa. DL: Donaldson. K: Kleeberger, SR. (2007). Mechanisms o f action o f inhaled fibers, particles and nanoparticles in lung and cardiovascular diseases. Part Fibre Toxicol 4: 4. http://dx.doi.org/10.1186/1743-8977-4-4. M ossm aa BT: Lippmaim, M: Hesterberg, TW; Kelsey, KT: Barchowskv, A: Bonner. JC. (2011). Pulmonary endpoints (lung carcinomas and asbestosis) following inhalation exposure to asbestos. J Toxicol Environ Health B Crit Rev 14: 76-121. http://dx.doi.org/10.1080/10937404.201L556Q47. MSHA, Asbestos Exposure Limit. 30, 56, 57, and 71, (2008). Mukheriee. S: de Klerk. N: Palmer. LJ: O lsea NJ: Pang, SC: William Musk, A, (2000). Chest pain in asbestosexposed individuals with benign pleural and parenchymal disease. Am J Respir Crit Care Med 162: 1807 1811. Muller. JG: Bohnker. BK: Philippi AF: Litow. FK: Rudolph. G: Hernandez. JE, (2005). Trends in pleural radiographic findings in the Navy Asbestos Medical Surveillance Program (1990-1999). Mil Med 170: 375-380. ' Muravov. O: Kave. W: Lewin. M: Berkowitz, Z: Lvbarger. J: Campolucci, S: Parker. J, (2005). The usefulness of computed tomography in detecting asbestos-related pleural abnormalities in people who had indeterminate chest radiographs: The Libby, MT, experience. Int J Hyg Environ Health 208: 87-99. Mvoio, T: Takava. M, (2001). Estimation o f fibrous aerosol deposition in upper bronchi based on experimental data with model bifurcation. Ind Health 39: 141-149. NCI. (National Cancer Institute). (2011). Fast Stats: An interactive tool for access to SEER cancer statistics Retrieved August 3, 2011, from http://seer,cancer.gov/faststats Nelson. HH, Kelsey. KT, (2002). The molecular epidemiology o f asbestos and tobacco in lung cancer. Oncogene 21: 7284-7288. http://dx.doi.org/10.1038/si.onc.12058Q4. Neri. M: Ugolini, D: Dianzani. I: Gemignani, F: Landi. S: Cesario. A: Magnani. C: Mutti. L: Puntoni. R: Bonassi. S. (2008). Genetic susceptibility to malignant pleural mesothelioma arid other asbestos-associated diseases. Mutat Res 659: 126-136. http://dx.doi.Org/10.1016/i.mrrev.2008.02.002. This document is a draftfor review purposes only and does not constitute Agency policy. 7-18 DRAFT--DO NOT CITE OR QUOTE Neri. S; Boraschi. P: Anlonelli. A; Falasclii. F; Bascliieri. L. (1996). Pulmonary function, smoking habits, and high resolution computed tomography (HRCT) early abnormalities o f lung and pleural fibrosis in shipyard workers exposed to asbestos. Am J Ind Med 30: 588-595. http://dx.doi.org/10.1002/(SICI)10970274(199611)30:5<588::AID-AJIM6>3.0.CQ:2-P. Newliouse, ML: Berry, G: Wagner. JC: Turok, ME. (1972). A study o f the mortality o f female asbestos workers. Occup Environ Med 29: 134-141. Newhouse. ML: Berry, G. (1976). Predictions o f mortality from mesothelial tumours in asbestos factory workers. Br J Ind Med 33: 147-151. Ni, Z: Liu. Y: Keshava. N: Zhou. G: Whong, W: Ong, T, (2000). Analysis o f K-ras and p53 mutations in mesotheliomas from humans and rats exposed to asbestos. Mutat Res 468: 87-92. http://dx.doi.Org/10.1016/S 1383-5718(00)00043-7. NIQSH (National Institute for Occupational Safety and Health). (1994a). Method 7400: asbestos and other fibers by PCM. InNIOSH manual o f analytical methods (4 ed., pp. N/A). Cincinnati, OH: U.S.: Department of Health and Human Serv ices, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health (NIOSH). NIQSH. (National Institute for Occupational Safety and Health). (1994b). Method 7402: Asbestos by TEM. In NIOSH manual o f analytical methods (4 ed., pp. N/A). Cincinnati, OH: U.S.: Department o f Health and Human Services, Centers for Disease Control and Prevention, National Institute for Occupational Safety and Health (NIOSH). NIOSH. (National Institute for Occupational Safety and Health). (2011). Asbestos fibers and other elongate mineral particles: State o f the science and roadmap for research. (2011-159). http://www.cdc. gov/niosh/docs/2011- 159/. ` Noonan, C. (2006). Exposure matrix development for the Libby cohort. Inlial Toxicol 18: 963-967. http://dx.doi.org/10.1080/08958370600835Q21. Noonan C: Pfau, J: Larson, T; Spence, M, (2006). Nested case-control study o f autoimmune disease in an asbestosexposed population. Environ Health Perspect 114: 1243-1247. NRC. (National Research Council). (1983). Risk assessment in the federal government: Managing the process. In. Washington, DC: National Academies Press. NRC. (National Research Council). (1993). Pesticides in the diets of infants and children. In. Washington, DC: National Academy Press. N IP , (National Toxicology Program). (1983). NTP lifetime carcinogenesis studies o f amosite asbestos (cas no. 12172-73-5) in Syrian golden hamsters (feed studies) (Vol. 249). Research Triangle Park. NC. http://nip,niehs.nih. gov/ntp/htdocs/LT rpts/tr249,pdf. NTP, (National Toxicology Program). (1985). NTP toxicology and carcinogenesis studies o f chrysotile asbestos (cas no. 12001-29-5) inf344/nrats (feed studies) (Vol. 295). Research Triangle Park, NC. http://www.ntis.gov/search/product.aspx?ABBR=PB86167103. NTP, (National Toxicology Program). (1988). NTP technical report on the toxicology and carcinogenesis studies of Crocidolite Asbestos (CAS No. 12001-28-4) InF344/NRats (Feed Studies) (Vol. 280). Research Triangle Park, NC. http://ntp.nielis.mh.gov/ntp/htdocs/LT rpts/tr280,pdf. This document is a draftfor revie\vpurposes only and does not constitute Agency policy. 7-19 DRAFT--DO NOT CITE OR QUOTE NTP. (National Toxicology- Program). (1990a). NTP technical report on the toxicology and carcinogenesis studies o f amosite asbestos (CAS No. 12172-73-5) in F344/N Rats (Feed Studies) (Vol. 279). Research Triangle Park, NC. http://ntp.niehs.nih.gov/ntp/htdocs/LT rpts/tr279.pdf. NTP. (National Toxicology Program). (1990b). NTP technical report on the toxicology and carcinogenesis studies of tremolite (CAS no. 14567-73-8) in F344/N rats (feed studies) (Vol. 277). Research Triangle Park, NC. http://ntp.niehs.nih.go\7ntp/htdocs/LT rots/tr277,pdf. Oberdorster, G. (1991). Lung dosimetry and extrapolation o f results from animal inhalation studies to man. J Aerosol Med Pulm Drag Deliv 4: 335-347. Okavasu, R: Wu, L: Hei, TK, (1999). Biological effects o f naturally occurring and man-made fibres: In vitro cytotoxicity and mutagenesis in mammalian cells. Br J Cancer 79: 1319-1324. http://dx.doi.org/10.1038/si.bic.669Q213. Oldham MJ: Maimix, RC: Phalen, RF, (1997). Deposition o f monodisperse particles in hollow models representing adult and child-size tracheobronchial airways. Health Phys 72: 827-834. Oldham. P , , D. (1965). On estimating the arithmetic means o f lognormally-distributed populations. Biometrics 21: 235-239. Oliver. LC: E isea EA: Greene, R: Sprince, NL, (1988). Asbestos-related pleural plaques and lung function. Am J Ind Med 14: 649-656. http://dx.doi.org/10.1002/aiim.47001406Q4. QSHA, (Occupational Safety & Health Administration). (1994). Occupational Safety and Health Standards: Toxic and Flazardous Substances - Asbestos. Standard No. 1910.1001 [Standard]. Washington, DC: United States Department o f Labor. QSHA, Asbestos. 29, 1910,1001, (2006). Padilla-Carlin, DL Schladweiler, MCJ: Shannahan, JH: Kodavanti, UP: Nvska, A: Burgoon, LD: Gavett, SH, (2011). Pulmonary inflaimnatory andfibrotic responses in Fischer 344 rats after intratracheal instillation exposure to libby amphibole. J Toxicol Environ Health A 74: 1111-1132. http://dx.doi.org/10.1080/15287394.201L58694Q. Paris, C: Martin. A: Letoumeux, M: Wild, P. (2008). Modelling prevalence and incidence o f fibrosis and pleural plaques in asbestos-exposed populations for screening and follow-up: a cross-sectional study. Environ Health Global Access Sci Source 7:30. http://dx.doi.org/10,1186/1476-069X-7-30. Paris. C: Thierry, S: Brochard. P: Letoumeux, M: Schorle, E: Stoufflet. A: Ameille, J: Conso. F: Pairon, JC. (2009). Pleural plaques and asbestosis: dose- and time-response relationships based on HRCT data. Eur Respir J 34: 72-79. http://dx.doi.org/10.1183/09031936.000940Q8. Peacock. A, (2003). Libby, Montana: Asbestos and the deadly silence o f an American corporation. In. Neenah, WI: Big Earth Publishing. Peipins, LA: Lewin, M: Campolucci, S: Lvbarger, JA: Miller. A: Middleton. D; Weis, C: Spence, M: Black. B: Kapil, V. (2003). Radiographic abnormalities and exposure to asbestos-contaminated vermiculite in the community o f Libby, Montana, USA. Environ Health Perspect 111: 1753-1759. Peipins, LA: Lewin. M: Campolucci. S: Lvbarger. JA: Kapil, V: Middleton. D: Miller. A: Weis, C: Spence, M; Black. B, (2004a). Overstating the consequences: Peipins et al.'s response. Environ Health Perspect 112: A84-A85. http://dx.doi.org/10.1289/ehp, 112-a84b. This document is a draftfor revie\vpurposes only and does not constitute Agency policy. 7-20 DRAFT--DO NOT CITE OR QUOTE Peipins. LA; Lewin. M: Campolucci. S; Lvbarger. JA; Kapil. V: Middleton. D: Miller. A: Weis. C: Spence. M: Black. B. (2004b). Radiographic abnormalities: Response from Peipins et al. Environ Health Perspect 112: A83. http://dx.doi.org/10.1289/ehp. 112-a83. Percy, C: Stanek. E, III: Gloeckler. L, (1981). Accuracy o f cancer death certificates and its effect on cancer mortality statistics. Am J Public Health 71: 242-250. http://dx.doi.org/10.2105/AJPH.71.3,242. Peto. J: Seidmaa H: Selikoff. IJ, (1982). Mesothelioma mortality in asbestos workers: implications for models of carcinogenesis and risk assessment. Br J Cancer 45: 124-135. http://dx.doi.org/10,1038/bic. 1982,15. Petrovic, P: Ostoiic. L: Peric. I: Mise. K: Ostoiic. Z: Bradaric. A: Bota. B: Jankovic. S: Tocili. J. (2004). Lung function changes in pleural asbestosis. Coll Antropol 28: 711-715. Pfau. JC: Sentissi. JJ: Weller. G: Putnam. EA, (2005). Assessment o f autoimmune responses associated with asbestos exposure in Libby, Montana, USA. Environ Health Perspect 113: 25-30. http://dx.doi.org/10.1289/ehp,7431. Pfau, JC: Pershouse, M; Putnam. EA, (2006). Conference summary. Directions and needs in asbestos research: new insights. Inlial Toxicol 18: 919-923. .http://dx.doi.org/10.i080/0895837060Q834735. Pfau. JC: Sentissi. JJ: Li. S: Calderon-Garciduenas. L: B row a JM: Blake. DJ. (2008). Asbestos-induced autoimmunity in C57BI/6 mice. J Innnunotoxicol 5: 129-137. http://dx.doi.Org/10.1080/15476910802085756. Phalen. RF: Oldham. MJ: Beaucage. CB: Crocker. TT: Mortensea JD. (1985). Postnatal enlargement o f human tracheobronchial airways and implications for particle deposition. Anat Rec 212: 368-380. http://dx.doi.org/10.1062/ar.10921204Q8. Phalen. RF: Oldham. MJ, (2001). Methods for modeling particle deposition as a function o f age. Respir Physiol Neurobiol 128: 119-130. http://dx.doi.org/10.1016/S0034-5687(01)00270-5. Phalen. RF: Oldham. MJ: Nel. AE, (2006). Tracheobronchial particle dose considerations for in vitro toxicology studies. Toxicol Sci 92: 126-132. http://dx.doi.org/10.1093/toxsci/kfi 182. Pietmska. JR: Johnstoa T: Zhitkovich. A: Kane. AB, (2010). XRCC1 deficiency sensitizes human lung epithelial cells to genotoxicity by crocidolite asbestos and Libby amphibole. Environ Health Perspect 118: 1707 1713. http://dx.doi.org/10.1289/ehp.1002312. ' Pinheiro. GA: Antao. VC: Bang. KM: Attfield. MD, (2004). Malignant mesothelioma surv eillance: a comparison of ICD 10 mortality data with SE ER incidence data in nine areas o f the United States. Int J Occup Environ Health 10: 251-255. ' Pinskv. PF: Freedman. M; Kvale. P: Oken. M; Caporaso. N: Gohagan. J. (2006). Abnormalities on chest radiograph reported in subjects in a cancer screening trial. Chest 130: 688-693. http://dx.doi.Org/10.1378/chest.130.3.688. Poser. I: Ralnnaa O: Loliani. M: Yadav. S: Becker. HH: Weiss. DG: Schiffmann. D: Popp, E. (2004). Modulation o f genotoxic effects in asbestos-exposed primary human mesothelial cells by radical scavengers, metal chelators and a glutathione precursor. Mutat Res Genet Toxicol Environ Mutagen 559: 19-27. http://dx.doi.Org/10.1016/i.mrgentox.2003.12.006. Price. B. (2004). Radiographic abnormalities and asbestos exposure: Libby, Montana. Environ Health Perspect 112: A82-A83. http://dx.doi.org/10.1289/ehp.112-a82. This document is a draftfor revie\vpurposes only and does not constitute Agency policy. 7-21 DRAFT--DO NOT CITE OR QUOTE Putnam EA: Smartt. A: Groves. A; Scliwanke. C; Brezinski. M: Pershouse. MA. (2008). Gene expression changes after exposure to six-mix in a mouse model. J Immunotoxicol 5: 139-144. http://dx.doi.org/10.1080/15476910802Q85772. Rahman. I: MacNee. W. (1999). Lung glutathione and oxidative stress: Implications in cigarette smoke-induced airway disease. Am J Physiol 277: L1067-L1088. Ramos-Nino. ME: Haegens. A: Shukla. A: M ossmaa BT. (2002). Role o f mitogen-activated protein kinases (MAPK) in cell injury and proliferation by environmental particulates. Mol Cell Biochem 234-235: 111 118. http://dx.doi.Org/10.1023/A:1015924413043. Ra.scli. EK: Hirsch. R: Paulose-Ram. R: I-Iochberg. MC. (2003). Prevalence o f rheumatoid arthritis in persons 60 years o f age and older in the United States: Effect o f different methods o f case classification. Arthritis Rheum 48: 917-926. http://dx.doi.org/10.1002/art. 10897. Reeves. GK: Cox, DR: Darby, SC: Whitley, E, (1998). Some aspects o f measurement error in explanatory variables for continuous and binary regression models. Stat Med 17: 2157-2177. http://dx.doi.org/10.1002/fSICI) 1097-0258(19981015)17:19<2157: :AID-SIM916>3.0.CQ:2-F. Regnis, JA: Zeman, KL: Noone, PG: Knowles, MR: Bennett, WD. (2000). Prolonged airway retention o f insoluble particles in cystic fibrosis versus primary ciliary dyskinesia. Exp Lung Res 26: 149-162. http://dx.doi.org/10.1080/01902140Q269844. ' Reid, A: Berne G: de Klerk. N: Hansen. J: Hevworth, L Ambrosini, G: Fritsclii, L: O lsea N: Merler, E: Musk, A, (2007). Age and sex differences in malignant mesothelioma after residential exposure to blue asbestos (crocidolite). Chest 131: 376-382. http://dx.doi.org/10.1378/chest.06-1690. Reid. A: Berry, G: Hevwortli. J: de Klerk. NH: Musk, AW. (2009). Predicted mortality from malignant mesothelioma among women exposed to blue asbestos at Wittenoom Western Australia. Occup Environ Med 66: 169-174. http://dx.doi.org/10.1136/oem.2007.038315. Reiss, B: Tong, C: Telang, S: Williams, GM, (1983). Enhancement o f benzo[a]pyrene mutagenicity by chrysotile asbestos in rat liver epithelial cells. Environ Res 31: 100-104. http://dx.doi.org/10.1016/0Q139351(83)90065-8. Rendall, R: Du Toit, R. (1994). The retention and clearance o f glass fibre and different varieties o f asbestos by the lung. Ami Occup Hyg 38: 757-761. http://dx.doi.org/10.1093/annlivg/38.inlialed particles VII.757. Rendall, RE: Skikne, ML (1980). Submicroscopic fibres in industrial atmospheres. In Biological effects o f mineral fibres (Vol. 30, pp. 837-843). Lyon, France: International Agency on the Research o f Cancer. Rev. F: Boutin. C: Steinbauer, J: Viallat, JR: Alessandroni, P: Jutisz, P: P i Giambattista. D: Billon-Galland, MA: Hereng, P: Dumortier, P, (1993). Environmental pleural plaques in an asbestos exposed population of northeast Corsica. EurRespir J 6: 978-982. Richardson DB, (2010). Occupational exposures and lung cancer: Adjustment for unmeasured confounding by smoking. Epidemiology 21: 181-186. http://dx.doi.org/10.1097/EDE.0b013e3181c6f7d9. Rockoff, SD: Chu. J: Rubia LJ, (2002). Special Report: Asbestos-induced pleural plaques-A disease process associated with ventilatory impairment and respiratory symptoms. Clinical Pulmonary Medicine 9: 113-- 124. ....................... Ragan, WJ: Gladea BC: Ragan, NB: Anderson, HA, (1987). US prevalence o f occupational pleural thickening: A look at chest X-rays from the first National Health and Nutrition Examination Survey. Am J Epidemiol 126:893-900. ' This document is a draftfo r review purposes only and does not constitute A gency policy. 7-22 DRAFT--DO NOT CITE OR QUOTE R ogaa W.L Ragan. NB: Dinse. GE. (2000). X-ray evidence o f increased asbestos exposure in the US population from NHANES I and NHANES II, 1973-1978. Cancer Causes Control 11: 441-449. h ttp ://d x .d o i.O rg /1 0 .1 0 2 3 /A :1 0 0 8 9 5 2 4 2 6 0 6 0. Roggli, VL: Vollmer. RT: Butnor, KJ; Spom, TA, (2002). Tremolite and mesothelioma. Ami Occup Hyg 46: 447 453. http://dx.doi.org/10.1093/annlryg/mef056. Roggli, VL: Vollmer. RT, (2008). Twenty-five years o f fiber analysis: What have we learned? Hum Pathol 39: 307 315. http://dx.doi.Org/10.1016/i.hmnpath.2007.07.005. Roguin, A: Ben-Shahar, M: Ben-Dror, G: Cohen. I: Hazani, E. (1994). Malignant mesothelioma in families of asbestos workers. Harefuah 126: 702-704, 764. Rolls, A: Lockev. I: Dunning. K: Sliukla, R: Faa H: Hilbert, T; Borton, E: Wiot, J: Mever, C: Slriplev, R: Lemasters, G: Kapil, V. (2008). Low-level fiber-induced radiographic changes caused by Libby venniculite: A 25-year follow-up study. Am I Respir Crit Care Med 177: 630-637. http://dx.doi.org/10.1164/rccin.200706-841QC. Roller. M: Pott, F; Kamino, K: Althoff, GH: Bellmamr, B. (1996). Results o f current intraperitoneal carcinogenicity studies with mineral and vitreous fibres. Exp Toxicol Pathol 48: 3-12. Roller. M: Pott, F: Kamino. K: Althoff. GH: Bellmamr. B, (1997). Dose-response relationship o f fibrous dusts in intraperitoneal studies. Environ Health Perspect 105: 1253-1256. Rom, WN: Hamnrar. SP: Rusch. V: Dodson. R: Hoffman. S. (2001). Malignant mesothelioma from neighborhood exposure to antliophyllite asbestos. Am I Ind Med 40: 211-214. http://dx.doi.org/10.1002/aiim, 1089. Rosenstock. L; Bamlrart, S: Hever. NJ: Pierson. PI: Hudson. LD, (1988). The relation among pulmonary function, chest roentgenograpliic abnormalities, and smoking status in an asbestos-exposed cohort. Am Rev Respir Dis 138: 272-277. http://dx.doi.Org/10.1164/airccm/138.2.272. Ross. M: N olaa RP. Langer. AM: Cooper. WC. (1993) Health effects o f mineral dusts other than asbestos. In GD Guthrie, Ir.; BT Mossman (Eds.), Health effects o f mineral dusts (Vol. 28, pp. 361-407). Washington, DC: Mineralogical Society o f America. Rothman, KJ. (1981). Induction and latent periods. Am I Epidemiol 114: 253-259. Rothman. KJ. (1986). Significance questing. Ann Intern Med 105: 445-447. Rothmaa KJ: Greenland. S. (1998). Modem epidemiology. In (2nd ed.). Philadelphia, PA: Lippincott, Williams, & Wilkins. Roushdv-Hammadv. I: Siegel. J: Einri. S: Testa. J: Carbone. M, (2001). Genetic-susceptibility factor and malignant mesothelioma in the Cappadocian region of Turkey. Lancet 357: 444-445. http://dx.doi.org/10.1016/S014Q- 6736(00)04013-7. " Rushtoa L, (2007a). Chronic obstructive pulmonary disease and occupational exposure to silica. Rev Environ Health 22:255-272. ' ' Rushtoa L, (2007b). Occupational causes o f chronic obstructive pulmonary disease. Rev Environ Health 22: 195 212. Sahu. AP: Dogra. RK; Shanker. R: Zaidi. SIT, (1975). Fibrogenic response in murine lungs to asbestos. Experimentelle Pathologie 11: 21-24. This document is a draftfor review purposes only and does not constitute Agency policy. 7-23 DRAFT--DO NOT CITE OR QUOTE Sakellariou. K: Malamou-Mitsi, V: Haritou. A: Koumpaniou. C: Stacliouli. C: Dimoliatis. ID: Constantopoulos. SH (1996). Malignant pleural mesothelioma from nonoccupational asbestos exposure in Metsovo (north-west Greece): Slow end o f an epidemic? Em Respir J 9: 1206-1210. http://dx.doi.org/10.1183/09031936.96.090612Q6. Sanchez. MS: Gunter. ME: Dvar. MD, (2008). Characterization o f historical ampliibole samples from the former vermiculite mine near Libby, Montana, U.S. A. European Journal o f Mineralogy 20: 1043-1053. http://dx.doi.org/10.1127/0935-1221/2008/002Q-1866. Scapoli. L: Ramos-Nino. M: Martinelli. M: M ossmaa B, (2004). Src-dependent ERK5 and Src/EGFR-dependent ERK1/2 activation is required for cell proliferation by asbestos. Oncogene 23: 805-813. http://dx.doi.org/10.1038/si.onc.1207163. Schlller-Scotland. CF: I-Ilawa, R: Gcbiiari J. (1994). Experimental data for total deposition in the respiratory tract of children. Toxicol Lett 72: 137-144. http://dx.doi.org/10.1016/0378-4274(94)90020-5. Schneider. J: Grossgartea K: Woitowitz. HJ, (1995). [Fatal pleural mesothelioma diseases caused by familial household contacts with asbestos fiber dust], Pneumologie 49: 55-59. Schneider. J: Straif. K: Woitowitz. HJ, (1996a). Pleural mesothelioma and household asbestos exposure [Review]. Rev Environ Health 11: 65-70. http://dx.doi.org/10.1515/REVEH.1996.ll.l-2.65. Schneider. J: Rodelsperger. K: Pohlabeln. H: Woitowitz. HJ, (1996b). Umwelt-und Innenraum-Gefahrdung durch Asbestfaserstaub als Risiko-und Einflubfaktor des diffusen malignen Plemamesothelioms (DMM). Zentralblatt fuer Hygiene und Umweltmedizin 199: 1-23. Schneider. J: Rodelsperger. K: Briickel, B: Kavser. K: Woitowitz. HJ, (1998). Environmental exposure to tremolite asbestos: Pleural mesothelioma in two Turkish workers in Germany. Rev Environ Health 13: 213-220. http://dx.doi.Org/10.1515/REVEH.1998.13.4.213. Schneider. V: Maurer. RR. (1977). Asbestos and embryonic development. Teratology 15: 273-279. http://dx.doi.org/10.1002/tera.1420150309. Schwartz. DA: Galvin. JR: Yagla. SJ: Speakmaa SB: Merchant. JA: Hnnninghakc. GW. (1993). Restrictive lung function and asbestos-induced pleural fibrosis: A quantitative approach. J Clin Invest 91: 2685-2692. http://dx.doi.org/10.1172/JCI116507. Sebastien. P; Billion-Galland. MA; Dufour. G: Bignon, J, (1980). Measmement o f asbestos air pollution inside buildings sprayed with asbestos. (EPA-560/13-80-026). Washington, DC: U. S. Environmental Protection Agency, http://nepis.epa. gov/Exe/ZvPURL.cgi?Dockey=9100BHQI.txt. Sebastien. P: McDonald. JC: McDonald. AD: Case. B: Harlev. R. (1989). Respiratory cancer in chrysotile textile and mining industries: Exposure inferences from lung analysis. Br J Ind Med 46: 180-187. Seidmaa H: Selikoff IJ: Gelb. SK, (1986). Mortality experience o f amosite asbestos factory workers: Doseresponse relationships 5 to 40 years after onset o f short-tenn work exposure. Am J Ind Med 10: 479-514. Selevaa SG: Kimmel. CA: Mendola. P. (2000). Identifying critical windows o f exposure for children's health. Environ Health Perspect 1083: 451-455. Selikoff. IJ: Lee. DHK, (1978). Asbestos and disease. In. New York, NY: Academic Press. Selikoff IJ: Hammond. EC. (1979). Asbestos and smoking [Editorial]. JAMA 242: 458-459. http://dx.doi.org/10.1001/iama.1979.03300050048Q29. This document is a draftfor review purposes only and does not constitute Agency policy. 7-24 DRAFT--DO NOT CITE OR QUOTE Selikoff. 1J: Seidman. H. (1992). Use o f death certificates in epidemiological studies, including occupational hazards: Variations in discordance o f different asbestos-associated diseases on best evidence ascertainment. Am J Ind Med 22: 481-492. http://dx.doi.org/10.1002/aiim.4700220403. Senvigit, A: Babavigit, C: Gokirmak, M; Topcu, F; A saa E: Coskunsel, M: Isik. R: Ertem, M. (2000). Incidence of malignant pleural mesothelioma due to environmental asbestos fiber exposure in the southeast o f Turkey. Respiration 67: 610-614. http://dx.doi.org/10.1159/000056289. Shannahaa J: Schladweiler, M: Padilla-Carlin, D: Nvska. A: Ricliards, J: Gliio, A: Gavett, S: Kodavanti. U. (2011a). The role o f cardiovascular disease-associated iron overload in Libby amphibole-induced acute pulmonary injury and inflammation. Inhal Toxicol 23: 129-141. http://dx.doi.org/10.3109/08958378.2011.55185Q. Shannalian. JH: Gliio, AJ: Schladweiler. MC: McGee, JK: Ricliards, JTL Gavett, SH: Kodavanti, UP. (2011b). The role o f iron in Libby amphibole-induced acute lung injury and inflammation. Inhal Toxicol 23: 313-323. http://dx.doi.org/ld.3109/08958378.201L569587. Short, RHP, (1952). Aspects o f comparative lung growth. Proc Biol Sci 140: 432-441. http://dx.doi.org/10.1098/rspb.1952.0071. Shtol', AV: Plotko, EG: Seliankina, KP. (2000). [Children's health and environmental air pollution with dust containing asbestos], Med TrProm Ekol 11: 10-13. Slinkla, A: Gulumian, M: Hei, TK: Kamp, D: Ralimaa O: Mossman, BT. (2003). Multiple roles of oxidants in the pathogenesis o f asbestos-induced diseases [Review]. Free Radic Biol Med 34: 1117-1129. http://dx.doi.org/10.1016/SQ891-5849(03)00060-1. Shukla, A: Lounsbmv, KM: Barrett. TF: Gell, J: Rincoa M: Butnor, KJ: Taaties, DJ: Davis. GS: Vacek, P: Nakavama, KI: Nakavama, K: Steele, C: Mossmaa BT, (2007). Asbestos-induced peribronchiolar cell proliferation and cytokine production are attenuated in lungs o f protein kinase C-delta knockout mice. Am J Pathol 170: 140-151. http://dx.doi.org/10.2353/aipath.2007.060381. Sichletidis, L: Daskalopoulou, E: Tsarou, V: Pnevmatikos, I: C111pros. D: Vamvalis, C. (1992). Five cases o f pleural mesothelioma with endemic pleural calcifications in a mral area in Greece. Med Lav 83: 326-329. Singh. B: Eastwood, PR: Finucane, KE; Panizza, JA: Musk, AW. (1999). Effect o f asbestos-related pleural fibrosis on excursion o f the lower chest wall and diaphragm. Am J Respir Crit Care Med 160: 1507-1515. Smartt. AM: Brezinski, M: Trapkus, M; Gardner, D: Putnam. EA, (2010). Collagen accumulation over time in the murine lung after exposure to crocidolite asbestos or Libby amphibole. Environ Toxicol 25: 68-76. http://dx.doi.org/10.1002/tox.2Q472. Smith, DD. (2002). Women and mesothelioma [Editorial]. Chest 122: 1885-1886. http://dx.doi.org/10.1378/chest. 122,6 ,1885. Smith, WE, (1978). Final report on biologic tests o f samples 22260p5 and 22263p2. M adisoa NJ: Health Research Institute, Fairleigh Dickinson University. Smith. WE: Hubert. DD. (1974) The intrapleural route as a means for estimating carcinogenicity. In E Karbe; JF Park (Eds.), Experimental lung cancer: Carcinogenesis and bioassays: International symposium held at the Battelle Seattle Research Center, Seattle, WA, USA, June 23-26, 1974 (pp. 92-101). New York, NY: Springer-Verlag. Smith, WE: Hubert. DD: Sobel, HJ: Marquet. E, (1979) Biologic tests o f tremolite in hamsters. InR Lemen; JM Dement (Eds.), Dusts and disease (pp. 335-339). Park Forest South, IL: Pathotox Publisher. This document is a draftfo r review purposes only and does not constitute A gency policy. 7-25 DRAFT--DO NOT CITE OR QUOTE Smith. WE: Hubert, DP; Sobel. HJ. (1980) Dimensions o f fibres in relation to biological activity. In JC Wagner; W Davis (Eds.), Biological effects o f mineral fibres (Vol. 30, pp. 357-360). Lyon, France: World Health Organization. Sonlat. JM; Lauaue, D: Esanirol. Y; Depres, M: Giroa J: Claudel. R: Carles. P. (1999). High-resolution computed tomography abnormalities in ex-insulators annually exposed to asbestos dust. Am J Ind Med 36: 593-601. http://dx,doi.org/10.1002/(SICI)1097-0274(199912)36:6<593::AID-AJIMl>3.0.CO;2-Z, Sniegelhalter, D; Thomas. A: Best, N; Lunn, D, (2003). WinBugs version 1.4 user manual. In. Cambridge, UK: MRC Biostatistics Unit. Sniegelhalter, DJ: Best. NG: Carlia BP: Van der linde. A, (2002). Bayesian measures o f model complexity and fit. J R Stat Soc Series B Stat Methodol 64: 583-639. http://dx.doi.org/10.llll/1467-9868.00353. Srebro. SH: Roggli. VL, (1994). Asbestos-related disease associated with exposure to asbestiform tremolite. Am J Ind Med 26: 809-819. http://dx.doi.org/10.1002/aiiin.470026061Q. Stanton. MF: Lavard, M; Tegeris. A: Miller. E; May. M; Morgan. E; Smith. A. (1981). Relation o f particle dimension to carcinogenicity in ampliibole asbestoses and other fibrous minerals. J Natl Cancer Inst 67: 965-975. Staples, CA; Gamsu, G: Ray, CS; Webb, WR. (1989). High resolution computed tomography and lung function in asbestos-exposed workers with normal chest radiographs. Am Rev Respir Dis 139: 1502-1508. http://dx.doi.Org/10.1164/airccm/139.6.1502. Strait, K: Benbrahim-Tallaa. L: Baan. R; Grosse, Y; Secretaa B; El Gliissassi, F; Bouvard. V: Gulia. N; Freemaa C; Galichet, L: Cogliano. V, (2009). A review of human carcinogens: Part C: Metals, arsenic, dusts, and fibres. Lancet Oncol 10: 453-454. http://dx.doi.org/10.1016/S1470-2045(09)70134-2. Sturm, R . . (2009). A theoretical approach to the deposition of cancer-inducing asbestos fibers in the human respiratory tract. The Open Lung Cancer Journal 2: 1-11. Sullivaa PA. (2007). Venniculite, respiratory disease, and asbestos exposure in Libby, Montana: Update of a cohort mortality study. Environ Health Perspect 115: 579-585. http://dx.doi.org/10.1289/ehp.9481. Suzuki. K: Hei, TK, (1996). Induction o f heme oxygenase in mammalian cells by mineral fibers: Distinctive effect o f reactive oxygen species. Carcinogenesis 17: 661-667. http://dx.doi.org/10,1093/carcin/17,4.661. Suzuki. Y; Kolwama. N, (1991). Translocation o f inhaled asbestos fibers from the lung to other tissues. Am J Ind Med 19: 701-704. http://dx.doi.org/10.1002/aiiin.47001906Q3. Suzuki. Y; Yuen SR. (2001). Asbestos tissue burden study on human malignant mesothelioma. Ind Health 39: 150 160. http://dx.doi.org/10.2486/indhealth.39.150. Suzuki. Y; Y u ea SR: Ashley. R, (2005). Short, thin asbestos fibers contribute to the development o f human malignant mesothelioma: Pathological evidence. Int J Hyg Environ Health 208: 201-210. http://dx.doi.Org/10.1016/i.iiheh.2005.0L015. Tableman. M . K i m . J . , S. (2004). Survival analysis using S: Analysis o f time-to-event data. In. Virginia Beach. VA: Chapman and Hall/CRC Press. Tan. RJ; Fattman. CL: Watkins. SC: Ourv. TD. (2004). Redistribution o f pulmonary EC-SOD after exposure to asbestos. J Appl Physiol 97: 2006-2013. http://dx.doi.org/10,1152/iapplplivsiol.00480.2004. This document is a draftfor review purposes only and does not constituteurgencypolicy. 7-26 DRAFT--DO NOT CITE OR QUOTE Topping. DC; Nettesheim. P. (1980). Two-stage carcinogenesis studies with asbestos in Fischer 344 rats. J Natl Cancer Inst 65: 627-630. Tossavainen. A: Karialainen. A: Karhunea PJ. (1994). Retention o f asbestos fibers in the human body. Environ HealthPerspect 102: 253-255. Truhaut, R: Chouroulinkov. I, (1989). Effect o f long-tenn ingestion o f asbestos fibres in rats. In Non-occupational exposure to mineral fibres (Vol. 90). Lyon, France: International Agency for Research on Cancer. ITS. EPA. (U.S. Environmental Protection Agency). (1986a). Airborne asbestos health assessment update [EPA Report|. (EPA/60(M8-84/0()3F). Washington DC. http://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=20009EBT.txt. U.S. EPA. (U.S. Environmental Protection Agency). (1986b). Guidelines for mutagenicity risk assessment [EPA Report]. (EPA/630/R-98/003). Washington. DC. http://www.epa.gov/iris/backgrdhtml. U.S. EPA (U.S. Environmental Protection Agency). (1986c). Guidelines for the health risk assessment o f chemical mixtures [EPA Report], (EPA/630/R-98/002). Washington. DC. U.S. EPA. (U.S. Enviromnental Protection Agency). (1988a). IRIS summary for Asbestos (CASRN 1332-21-4). Washington. DC. http://www.epa.gov/iris/subst/037 1,htm. U.S. EPA. (U.S. Enviromnental Protection Agency). (1988b). Recommendations for and documentation of biological values for use in risk assessment. (EPA/600/6-87/008). Cincinnati. OH: U.S. Enviromnental Protection Agency. Enviromnental Criteria and Assessment Office. http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm7dekU34855. U.S. EPA (U.S. Enviromnental Protection Agency). (1991a). Guidelines for developmental toxicity risk assessment [EPA Report]. (EPA/600/FR-91/001). Washington, DC: U.S. Enviromnental Protection Agency. Risk Assessment Fomin, http://www.epa.gov/iris/backgrd.html. U.S. EPA. (U.S. Enviromnental Protection Agency). (1991b). Health assessment document for venniculite [EPA Report]. (EPA/600/8-91/037). Research Triangle Park, NC. http://nepis.epa. gov/Exe/ZvPURL.cgi?Dockev=3000 ILHV.txt. U.S. EPA. (U.S. Environmental Protection Agency). (1994a). Interim policy for particle size and limit concentration issues in inhalation toxicity studies [EPA Report]. Washington, DC: U.S. Enviromnental Protection Agency, Office of Pesticide Products, Health Effects Division. http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=186068. U.S. EPA. (U.S. Environmental Protection Agency). (1994b). Methods for derivation o f inhalation reference concentrations and application o f inhalation dosimetry. (EPA/600/8-90/066F). Research Triangle Park, NC: U.S. Enviromnental Protection Agency, Office o f Research and Development, Office o f Health and Environmental Assessment, Enviromnental Criteria and Assessment Office. http://cfbub.epa. gov/ncea/cfm/recordisplav.cfin?deid=71993. U.S. EPA. (U.S. Environmental Protection Agency). (1995). The use o f the benchmark dose approach in health risk assessment [EPA Report]. (EPA/630/R-94/007). Washington, DC. http://www.epa.gov/raf/publications/useof-bda-healtlirisk.htm. U.S. EPA, (U.S. Enviromnental Protection Agency). (1996). Guidelines for reproductive toxicity risk assessment [EPA Report]. (EPA/630/R-96/009). Washington. DC: U.S. Enviromnental Protection Agency, Risk Assessment Fomin http://www.epa.gov/raf/publications/pdfs/REPR051.PDF. This document is a draftfor re\'iew purposes onIv and does not constitute Agency policy. 7-27 DRAFT--DO NOT CITE OR QUOTE U.S. EPA. (U.S. Environmental Protection Agency). (1998). Guidelines for neurotoxicity risk assessment. (EPA/630/R-95/001F). Washington, DC: U.S. Environmental Protection Agency, Risk Assessment Forum. http://www.epa.gov/raf/publications/pdfs/NEUROTOX.PDF. U.S. EPA. (U.S. Environmental Protection Agency). (2000a). Benchmark dose technical guidance document [external review draft]. (EPA/630/R-00/001). Washington. DC: U.S. Environmental Protection Agency, Risk Assessment Forum, http://www.epa.gov/raf/publications/benclnnark-dose-doc-draft.htm. U.S. EPA (U.S. Enviromnental Protection Agency). (2000b). Sampling and analysis o f consumer garden products that contain venniculite. (epa 744-r-00-010). Washington, DC. http://nepis.ep8. gov/E.xe/ZYPURL.cgi?Dockev=P 10014IJ.txt. U.S. EPA. (U.S. Environmental Protection Agency). (2000c). Science policy council handbook: Risk characterization. (EPA 100-B-00-002). Washington, D C.: U.S. Enviromnental Protection Agency, Office o f Research and Development, Office o f Science Policy, http://www.epa.gov/osa/spc/pdfs/rcliandbk.pdf. U.S. EPA. (U.S. Environmental Protection Agency). (2000d). Supplementary guidance for conducting health risk assessment o f chemical mixtures. (EPA/630/R-00/002). Washington, DC: U.S. Enviromnental Protection Agency. Risk Assessment Forum, http://cfoub.epa.gov/ncea/cfm/recordisplav.cfm?deid=20533. U.S. EPA. (U.S. Enviromnental Protection Agency). (2001a). EPA's actions concerning asbestos-contaminated venniculite in Libby, Montana. (2001-S-7). Washington. DC. http://www.epa. gov/oi g/reports/2001/montana.pdf. U.S. EPA. (U.S. Environmental Protection Agency). (2001b). Risk assessment guidance for superfund: Volume III part A, process for conducting probabilistic risk assessment. (EPA 540-R-02-002). Washington, DC: U.S. Environmental Protection Agency, Office o f Solid Waste and Emergency Response. http://www.epa.gov/oswer/riskassessment/rags3adt/index.htm. U.S. EPA. (U.S. Enviromnental Protection Agency). (2002). A review o f the reference dose and reference concentration processes. (EPA/630/P-02/002F). Washington, DC. http://cfpub.epa. gov/ncea/cfm/recordisplav.efm?deid=51717. U.S. EPA (U.S. Enviromnental Protection Agency). (2004). Dosimetry o f inhaled particles in the respiratory tract. (EPA/600/P-99/002bF). Research Triangle Park, NC. U.S. EPA. (U.S. Enviromnental Protection Agency). (2005a). Guidelines for carcinogen risk assessment. (EPA/630/P-03/00IF). Washington. DC. http://www.epa.gov/cancerguidelines/. U.S. EPA. (U.S. Enviromnental Protection Agency). (2005b). Supplemental guidance for assessing susceptibility from early-life exposure to carcinogens. (EPA/630/R-03/003F). Washington, DC: U.S. Enviromnental Protection Agency, Risk Assessment Forum, http://www.epa.gov/cancerguidelines/guidelines-carcinogensupplement.htm. U.S. EPA. (U.S. Environmental Protection Agency). (2006a). Aging and toxic response: Issues relevant to risk assessment [EPA Report]. (EPA/600/P-03/004A). Washington. DC. http://cft)ub.epa.goy./ncea/cfm/recQrdisplay.cfm?deid=156648. U.S. EPA. (U.S. Environmental Protection Agency). (2006b). A framework for assessing health risk of enviromnental exposures to children. (EPA/600/R-05/093F). Washington, DC. http://clipub.epa. gov/ncea/cfm/recordisplav.cfm?deid=158363. U.S. EPA, (U.S. Enviromnental Protection Agency). (2006c). Phase 2 study data summary report for Libby, Montana, enviromnental monitoring for asbestos, evaluation o f exposure to airborne asbestos fibers during routine and special activities. Libby, MT. This document is a draftfo r review purposes only and does not constitute A gency policy. 7-28 DRAFT--DO NOT CITE OR QUOTE U.S. EPA. (U.S. Environmental Protection Agency). (2006d). Science policy council handbook: Peer review, 3rd edition. (EPA/100/B-06/002). Wasliington, DC: U.S. Environmental Protection Agency, Science Policy Council. http://www.epa.gov/OSA/sDc/2peerrev.htm. U.S. EPA. (U.S. Environmental Protection Agency). (2008). Framework for investigating asbestos-contaminated superfund sites. Washington, DC. http://www.epa. gov/smrerfund/health/contarninants/asbestos/pdfs/framework asbestos guidance.pdf. U.S. EPA. (U.S. Enviromnental Protection Agency). (2009a). Risk assessment guidance for superfund volume I: Human health evaluation manual (Part F, supplemental guidance for inhalation risk assessment): Final. (EPA/540/-R-070/002). Wasliington, DC: U.S. Enviromnental Protection Agency, Office o f Superfund Remediation and Technology Innovation, http://www.epa.gov/oswer/riskassessment/ragsf/index.htm. U.S. EPA. (U.S. Environmental Protection Agency). (2009b). Summary o f outdoor ambient air monitoring for asbestos at the Libby asbestos site, Libby, Montana (October 2006 to June 2008). Denver, CO. U.S. EPA. (U.S. Environmental Protection Agency). (2010a). Activity-based sampling summary report, operable unit 4, Libby, Montana, Superfund site. Denver, CO. http://www.epa.gov/region8/superfund/libbv/OU4 ABS ResultsSummarvReportJune2010.pdf. U.S. EPA, (U.S. Enviromnental Protection Agency). (2010b). Particle size distribution data for Libby Amphiboles structures observed in air at the Libby Asbestos Superfund Site. Denver, CO: U.S. Enviromnental Protection Agency, Superfund Records Center. U.S. EPA, (U.S. Environmental Protection Agency). (2010c). Public schools asbestos sampling report. Denver, CO. U.S. GAO. (U.S. Government Accountability Office). (2007). Flazardous materials: EPA may need to reassess sites receiving asbestos-contaminated ore from Libby, Montana, and should improve it's public notification process. (GAO-08-71). Wasliington, DC. http://www.gao.gov/products/GAQ-08-71. Ugolini, D: Neri, M: Ceppi. M: Cesario, A: Dianzani, I: Filiberti, R: Gemignani, F: Landi, S: Magnani, C: Mutti, L; Puntoni, R: Bonassi, S. (2008). Genetic susceptibility to malignant mesothelioma and exposure to asbestos: the influence o f the familial factor. Mutat Res 658: 162-171. http://dx.doi.Org/10.1016/i.mrrev.2007.08.001. van Cleemput, J: de Raeve, H; Verscliakelen, JA: Rombouts. J: Lacauet, LM: Nemerv. B. (2001). Surface of localized pleural plaques quantitated by computed tomography scanning: No relation with cumulative asbestos exposure and no effect on lung function. Am J Respir Crit Care Med 163: 705-710. Vaslet, CA: Messier. NJ: Kane, AB, (2002). Accelerated progression o f asbestos-induced mesotheliomas in heterozygous p53+/- mice. Toxicol Sci 68: 331-338. http://dx.doi.Org/10.1093/toxsci/68.2.331. Versar, (Versar Inc.). (2003). EPA's pilot study to estimate asbestos exposure from vermiculite attic insulation. Washington. DC: U.S. Enviromnental Protection Agency. http://www.epa.gov/asbestos/pubs/insulatiomeport.pdf. Vinikoor. LC: Larson. TC: Bateson TF: Bimbaum. L, (2010). Exposure to asbestos-containing vermiculite ore and respiratory symptoms among individuals who were children while the mine was active in Libby, Montana. Environ Health Perspect 118: 1033-1038. http://dx.doi.org/10.1289/ehp.090168Q. Voisin, C: Marin. I: Brochard. P: Pairon, J, (1994). Enviromnental airborne tremolite asbestos pollution and pleural plaques in Afghanistan. Chest 106: 974-976. http://dx.doi.Org/10.1378/chest.106.3.974. Wagner, J: Chamberlain, M, B row a R: Berry, G: Poolev. F: Davies. R: Griffiths. D. (1982). Biological effects of tremolite. Br J Cancer 45: 352-360. This document is a draftfor review purposes only and does not constitute Agency policy. 7-29 DRAFT--DO NOT CITE OR QUOTE Wagner. JC: Sieges, CA: Marchand. P. (1960). Diffuse pleural mesothelioma and asbestos exposure in the north western Cape Province. Occup Environ Med 17: 260-271. http://dx.doi.org/10.1136/oein. 17,4.260. Walker. C: Everitt, J: Barrett, JC. (1992). Possible cellular and molecular mechanisms for asbestos carcinogenicity. Am J Ind Med 21: 253-273. http://dx.doi.org/10.1002/aiiin.470021Q214. Wang, XR: Yano. E: Wang. M: Wang. Z: Christian!, DC. (2001). Pulmonary function in long-term asbestos workers inChina. J Occup Environ Med 43: 623-629. Wanner. A: Salathe, M: Q'Riordara TG. (1996). Mucociliary clearance in the airways. Am J Respir Crit Care Med 154: 1868-1902. Wassermann, M: Wassermami, D: Steinitz. R: Katz. L: Lemesch, C. (1980). Mesothelioma in children. In Biological effects o f mineral fibres (Vol. 30). Lyon, France: International Agency for Research on Cancer. Weill. D: D liilloa G: Frevder. L: Lefante. J: Glindmever. H, (2011). Lung fimction, radiologic changes and exposure: analysis o f ATSDR data from Libby, Montana. Eur Respir J 38: 376-383. http://dx.doi.org/10.1183/09031936.000502l6. Weis. C. (2001a). Amphibole mineral fibers in source materials in residential and coimnercial areas o f Libby pose an imminent and substantial endangennent to public health Weis. C. (2001b). Fibrous Amphibole Contamination in Soil and Dust at Multiple Locations in Libby Poses an Imminent and Substantial Endangennent to Public Health: an Addendum to my Memorandum o f May 10, 2000 Weiss. ST, (2010). Lung function and airway diseases. Nat Genet 42: 14-16. http://dx,doi.org/10,1038/ngO 110-14. W elch LS: Haile. E: Dement. J: Michaels. D. (2007). Change in prevalence o f asbestos-related disease among sheet metal workers 1986 to 2004. Chest 131: 863-869. http://dx.doi.org/10.1378/chest.06-1155. Wheeler. MW. (2005). Benchmark dose estimation using SAS. In Proceedings o f the Thirtieth Annual SAS Users Group International Conference. Cary, NC: SAS Institute, http://www2,sas.com/proceedings/sugi30/20130.pdf. Whitehouse. A, (2004). Asbestos-related pleural disease due to tremolite associated with progressive loss o f lung function: Serial observations in 123 miners, family members, and residents o f Libby, Montana. Am J Ind Med 46: 219-225. http://dx.doi.org/10.1002/aiim.20053. Whitehouse. A: Black. C: Heppe. M: Ruckdeschel. J: L evia S. (2008). Environmental exposure to Libby Asbestos and mesotheliomas. Am J Ind Med 51: 877-880. http://dx.doi.org/10.1002/aiim.2062Q. WHO. (World Health Organization). (1938). Manual o f the international statistical classification of diseases, injuries, and causes o f death. Geneva, Switzerland. WHO. (World Health Organization). (1948). Manual o f the international statistical classification of diseases, injuries, and causes o f death. Geneva, Switzerland. WHO, (World Health Organization). (1957). Manual o f the international statistical classification of diseases, injuries, and causes o f death. Geneva, Switzerland. WHO. (World Health Organization). (1967). Manual o f the international statistical classification o f diseases, injuries, and causes o f death. Geneva, Switzerland. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. 7-30 DRAFT--DO NOT CITE OR QUOTE WHO. (World Health Organization). (1977). Manual o f the international statistical classification of diseases, injuries, and causes o f death. Geneva, Switzerland. WHO. (World Health Organization). (1992). Manual o f the international statistical classification of diseases, injuries, and causes o f death. Geneva, Switzerland. Witschi, HR: Last, JA, (1996) Toxic responses o f the respiratory system. In CD Klaassen (Ed ), Casarett and Doull's toxicology (5th ed., pp. 443-462). New York, N Y : McGraw Hill. Witschi, HR: Last, JA, (2001) Toxic responses o f the respiratory system. In LJ Casarett; CD Klaassen; J Doull (Eds.), Casarett and Doull's toxicology: The basic science o f poisons (6 ed., pp. 515-534). Columbus, OH: McGraw-Hill Professional. Wraith. D: Mengersen, K, (2007). Assessing the combined effect o f asbestos exposure and smoking on lung cancer: a Bayesian approach. Stat Med 26: 1150-1169. http://dx.doi.org/10.1002/sim.2602. Wright, RS: Abraham. JL: Harber, P; Burnett, BR: Morris, P; W est P. (2002). Fatal asbestosis 50 years after brief high intensity exposure in a vermiculite expansion plant. Am J Respir Crit Care Med 165: 1145-1149. http://dx.doi.org/10.1164/rccm,2110034. Wylie, AG: Skinner. HCW: Marsh, J: Snyder, IT: Garzione, C: Hodkinsoa D; Winters, R: M ossmaa BT, (1997). Mineralogical features associated with cytotoxic and proliferative effects o f fibrous talc and asbestos on rodent tracheal epithelial and pleural mesothelial cells. Toxicol Appl Pharmacol 147: 143-150. http://dx.doi.org/10.10Q6/taap. 1997,8276. Wvlie, AG: Verkouterea JR. (2000). Amphibole asbestos from Libby, Montana: Aspects o f nomenclature. American Mineralogist 85: 1540-1542. Yano, E: Tanaka, K: Funaki, M; Maeda, K: Matsunaga. C; Yamaoka, K. (1993). Effect o f smoking on pleural thickening in asbestos workers. Br J Ind Med 50: 898-901. http://dx.doi.org/10.1136/oem.50.10,898. Yano, E; Wang, ZM: Wang, XR: Wang, MZ: Takata, A: Kohvama, N: Suzuki. Y, (2009). Mesothelioma in a worker who spun chrysotile asbestos at home dining childhood. Am J Ind Med 52: 282-287. http://dx.doi.org/10.1002/aiim.20675. Yazicioglu, S , . (1976). Pleural calcification associated with exposure to chrysotile asbestos in southeast Turkey. Chest 70: 43-47. http://dx.doi.org/10.1378/chest.70,1,43. Yu. CP: Asgharian B: Yen, BM, (1986). Impaction and sedimentation deposition o f fibers in airways. Am Ind Hyg Assoc J 47: 72-77. http://dx.doi.org/10.108Q/15298668691389388. Yu, CP: Ding, YJ: Zhang, L; Oberdorster, G; Mast. RW; Maxim. LD: Utell, M l, (1997). Retention modeling of refractory ceramic fibers (RCF) in humans. Regul Toxicol Pharmacol 25: 18-25. http://dx.doi.org/10.1006/rtph, 1996,1064. Zanella, C; Posada, J: Tritton, T; M ossm aa B, (1996). Asbestos causes stimulation o f the extracellular signalregulated kinase 1 mitogen-activated protein kinase cascade after phosphorylation o f the epidermal growth factor receptor. Cancer Res 56: 5334-5338. Zeger, SL: Thomas, D: Dominici, F; Samet, JM: Schwartz. J: Dockery, D: Cohen. A, (2000). Exposure measurement error in time-series studies o f air pollution: Concepts and consequences. Environ Health Perspect 108: 419-426. Zerea EH: Gumurdiihi. D: Roggli, VL: Zorludemir, S: Erkisi. M: Tuncer, I, (2000). Environmental malignant mesothelioma in southern Anatolia: A study o f fifty cases. Environ Health Perspect 108: 1047-1050. This document is a draftfor review purposes only and does not constitute Agency policy. 7-31 DRAFT--DO NOT CITE OR QUOTE Zerva. LV; Constantopoulos, SH: Moutsopoulos, HM. (1989). Humoral immunity alterations after environmental asbestos exposure. Respiration 55: 237-241. http://dx.doi.org/10.1159/000195740. Zhao. XH: Jia. G: Liu. YO: Liu. SW: Y a a L: Jia Y: Liu. N, (2006). Association between poh morphisms o f DNA repair gene XRCC1 and DNA damage in asbestos-exposed workers. Biomed Environ Sci 19: 232-238. Zhou. Y : Su, WC; Cheng. YS. (2007). Fiber deposition in the tracheobronchial region: Experimental measurements. Inhal Toxicol 19: 1071-1078. http://dx.doi.org/10.1080/089583707Q1626634. Zitting. AJ: Karialainea A: Impivaara. O: Kuusela. T: Maki. J: Tossavainea A: Jarvisalo. J. (1996). Radiographic small lung opacities and plemal abnormalities in relation to smoking, urbanization status, and occupational asbestos exposure in Finland. J Occup Environ Med 38: 602-609. This document is a draftfor re\'ie\v purposes only and does not constitute Agency policy. 7-32 DRAFT--DO NOT CITE OR QUOTE APPENDIX A. SUMMARY OF EXTERNAL PEER-REVIEW AND PUBLIC COMMENTS AND DISPOSITION A-l [This page intentionally left blank.] 1 2 3 4 5 PARTICLE SIZE DISTRIBUTION DATA FOR 6 LIBBY AMPHIBOLE STRUCTURES OBSERVED IN AIR 7 AT THE LIBBY ASBESTOS SUPERFUND SITE 8 9 10 11 12 13 July 14, 2010 14 15 16 17 18 19 20 Prepared by: 21 U.S. Environmental Protection Agency 22 Region 8 23 Denver, CO 24 25 26 27 28 29 30 31 32 33 34 With Technical Assistance from: 35 36 SRC, Inc. 37 Denver, CO 38 This document is a draftfor review purposes only and does not constitute Agency policy. B-l DRAFT--DO NOT CITE OR QUOTE APPROVAL PAGE T h is report, Particle Size Distribution Datafor Libby Amphibole Structures Observed in Air at the Libby Asbestos Superfund Site, is a p p r o v e d fo r d is tr ib u tio n . This document is a draftfor re\'iew purposes only and does not constituteAgency policy. B-2 DRAFT--DO NOT CITE OR QUOTE PARTICLE SIZE DISTRIBUTION DATA FOR LIBBY AMPHIBOLE STRUCTURES OBSERVED IN AIR AT THE LIBBY ASBESTOS SUPERFUND SITE 1.0 INTRODUCTION Libby is a community in northwestern Montana that is located near a large open-pit vermiculite mine. Vermiculite from this mine contains varying levels of a form of asbestos referred to as Libby Amphibole (LA). In 1999, EPA Region 8 initiated environmental investigations in the town of Libby and in February, 2002, EPA listed the Libby Asbestos Site (the Site) on the National Priorities List. The Site includes the former vermiculite mine and residential homes, commercial businesses, schools and parks that may have become contaminated with asbestos fibers as a result of vermiculite mining and processing conducted in and around Libby as well as other areas in the vicinity that may have been impacted by mining-related releases of asbestos. Historic mining, milling, and processing operations at the Site, as well as bulk transfer of mining-related materials, tailings, and waste to locations throughout Libby Valley, are known to have resulted in releases of vermiculite and LA to the environment. As part of the response actions taken pursuant to the Comprehensive Environmental Response, Compensation and Liability Act, EPA has performed a number of investigations to characterize the nature and extent of LA contamination of air, soil, dust and other media in and around the community of Libby. Because available information suggests that the toxicity of asbestos is at least partially influenced by the size of the inhaled asbestos particles, these investigations have included the measurement of the dimensions (length and width) of LA particles observed in samples collected from the Libby site. The purpose of this report is to summarize size distribution data for LA particles that have been observed in air samples collected at the site, and to utilize these data to make comparisons between various subsets of the data to determine if any important differences in particles size distributions can be recognized. 2.0 METHODS 2.1 Data Overview EPA has been collecting samples of air since 2001 at the Libby site. Table 1 provides an overview of the sampling programs that have generated these data. The raw data for the air samples included in this assessment are provided in Appendix A. Most of the samples that have been collected have been analyzed for asbestos by transmissi on electron microscopy (TEM) using either ISO 10312 (1 9 9 5 ) or AHERA (1 9 8 6 ) counting rules, as modified by site-specific modifications as described in modifications forms LB-000016 and LB00003 1 (provided in Appendix B). In all cases, the data that are recorded during the analysis of a This document is a draftfor review purposes only and does not constitute Agency policy. B-3 DRAFT--DO NOT CITE OR QUOTE sample include the length, width and aspect ratio (length/width) of all particles that meet the counting rules specified for the analysis. 2.2 Data Presentation One convenient method for comparing the size distributions of two different sets of LA particles is through a graph that plots the cumulative distribution function (CDF) for each particle set. This graphical format shows the fraction of all particles that have a dimension less than some specified value. This format is used in this document to present the distributions of length, width and aspect ratio. There are a number of statistical tests that can be used to compare two distributions in order to support a statistical statement about whether the distributions are "same" or "different". Such comparisons are complicated by the fact that the distributions may be similar over some intervals and dissimilar over other intervals. However, at present, data are not sufficient to know which parts of the distribution are most important from a toxicological perspective. Therefore, this document relies upon simple visual inspection to assess the degree of difference between various regions of differing distributions. This document is a draftfor review purposes only and does not constitute Agency policy. B-4 DRAFT--DO NOT CITE OR QUOTE 3.0 RESULTS 3.1 Data Validation The Libby2 database and Libby OU3 database have a number of built-in quality control checks to identify unexpected or unallowable data values during upload into the database. Any issues identified by these automatic upload checks were resolved by consultation with the analytical laboratory before entry of the data into the database. After entry of the data into the database, several additional data verification steps were taken to ensure the data were recorded and entered correctly. A total of 29,504 LA structures are included in Table 1. Of these structures, 25% have undergone data validation in accord with standard site-wide operating procedures (SRC, 2008) to ensure that data for length, width, particle type, and mineral class are correct. Of the structures that have undergone validation, only 39 of 7,464 (0.5%) structures had errors in length, width, or mineral class. These errors were corrected and the database updated as appropriate. 3.2 Consolidated Data Set Originally, most samples of air at Libby were analyzed using a counting rule based on a fiber aspect ratio of 5:1. More recently, most air samples are counted using an aspect ratio rule of 3:1. Because this rule has varied over time, Libby-specific laboratory modifications LB-000016 and LB-000031 (see Attachment 1) were created to document the historic modifications and instructions that laboratories have followed throughout the Libby program. Figure 3-1 presents the particle size distributions for 29,504 LA particles observed to date1in air samples collected at the Libby Asbestos Superfund site that have an aspect ratio of 5:1 or more, along with the distributions for 11,451 particles that were counted using an aspect ratio rule of 3:1. As seen, the di stributions are very similar. This is because the number LA particles that have an aspect ratio >3:1 and <5:1 is a relatively small fraction of the total (7%). For simplicity, all remaining analyses focus on the set of particles with an aspect ratio of 5:1 or more. 3.3 Frequency of Complex Structures Asbestos particles occur not only as fibers but also in more complex structures including bundles, clusters, and matrix complexes. The frequency of these structure types in air samples from Libby are summarized below: 'Based on a query o f the Libby2 database on 12/08/09 and the Libby GU3 database on 2/9/10. This document is a draftfor review purposes onlv and does not constitute Agency policy. B-5 DRAFT--DO NOT CITE OR QUOTE Type2 Fiber Bundle Matrix Cluster Total Number 23,933 2,366 3,150 54 29,504 Frequency 81% 8% 11% 0.2% 100% As shown, most (81%) of the enumerated structures are fibers, with less than 20 % complex structures. 3.4 Comparisons of Stratified Data Sets The data sets shown in Figure 3-1 are based on air samples that were collected at a number of different locations around the site, and which were analyzed by several different methods. In order to investigate whether there are any important differences in size distributions between operable units, sampling locations (indoor, outdoor), activity (e.g., active or passive), and /or analytical method, the consolidated data set was partitioned into a number of subsets, as follows: Figure 3-2 3-3 3-4 3-5 3-6 Comparison LA particles observed in air stratified by structure type LA particles observed in air stratified by Operable Unit LA particles observed in air stratified by sample type (ambient, indoor, outdoor ABS) LA particles observed in air stratified by preparation method (direct vs indirect) LA particles observed in air stratified by analysis method (ISO vs AHERA) Figure 3-2 is a comparison of different structure types (fiber, bundles, and matrices). Clusters were not included because there were too few for a distribution to be meaningful. As seen, the length distribution for matrix particles is somewhat left-shifted compared to fibers. This is perhaps expected because some portion of the fiber length in matrix fibers is obscured by the matrix particle. In contrast, the length and thickness distributions for bundles are right-shifted compared to fibers. This is expected because a bundle is several fibers lying in parallel. Figure 3-3 compares the size distributions of LA at different operable units (OUs) at the site. As seen, there appears to be little difference in structures from the different OUs. 2 In some cases, the structure type assignment provided by the laboratory was not a valid choice according to the recording rules for the specified analysis method. Table A -l in Appendix A presents the types of invalid structure types and the structure class assumption that was made in order to include the structure in this report. This document is a draftfor review purposes only and does not constitute Agency policy. B-6 DRAFT--DO NOT CITE OR QUOTE Figure 3-4 shows the distribution of structure sizes for different types of air samples. Samples have been placed into three groups: ambient air, indoor ABS, and outdoor ABS. As shown, the length and width distributions for indoor and outdoor ABS samples are relatively similar, while the length and width distribution for ambient air samples appear to be right shifted. However, this observation should be considered to be relatively uncertain because of the small number (136) of particles that constitute the ambient air data set. Figure 3-5 compares the size distributions for samples using direct and indirect preparation methods. As shown, there is little difference in the distributions or either length of width, suggesting that preparation method does not have a significant impact on particle size. Figure 3-6 compares the particle size distributions as a function of analytical counting rules. As shown, the length and width distributions for particles analyzed using AHERA rules tend to be somewhat right-shifted relative to the distributions for particles analyzed using ISO 10312 rules. This apparent difference might be related either to differences in counting rules between methods, or possibly to differences in the nature of samples analyzed by each method. In either event, the difference between methods appears to be relatively small. 4.0 SUMMARY Particle size data are available for nearly 30,000 LA structures that have been observed in air samples collected at the Libby Asbestos Superfund site. Most (about 80%) LA particles are fibers, with less than 20% complex structures (bundles, clusters, or matrices). LA particle lengths typically range from a little less than 1 pm up to 20-30 pm, and occasionally higher. The average length is about 7 pm. Thicknesses typically range from about 0.1 pm up to about 2 pm, with an average of about 0.5 pm. Although some variations occur, particle size distributions are generally similar between different locations and between different types of samples. This document is a draftfor review purposes only and does not constitute Agency policy. B-7 DRAFT--DO NOT CITE OR QUOTE APPENDIX A RAW DATA: LA STRUCTURE DATA FROM THE LIBBY 2 DATABASE AND THE LIBBY OU3 DATABASE Libby2DB based on a download date of 12/8/09 Libby OU3 DB based on a download date of 2/9/10 See attached compact disc. This document is a draftfor review purposes only and does not constitute Agency policy. B-8 DRAFT--DO NOT CITE OR QUOTE APPENDIX B LIBBY-SPECIFIC LABORATORY MODIFICATION FORMS LB-000016 LB-000031 Table L Air Sample Collection Programs P ro g ram Phase 1 Phase 1R Phase 2 Phase 2R CSS SQAPP A m bient A ir :(A A ) O U 4 Indoor/ O utdoor A BS Indoor S ch o o ls O utdoor S ch o o ls Phase 2 (O U3) Phase 3 (O U3) C lean-up E valuation O ther P ro g ram D escription Initial investigation sam pling to assess nature and extent o f potential contam ination. Includes source areas (e.g., screening plant, export plant), com m ercial buildings, and residential properties. M onitoring and confirm ation sam pling as p art o f clean-up activities. A ctivity-based sam pling (A B S) included four scenarios: 1) routine indoor activities, 2) active cleaning, 3) sim ulated rem odeling disturbances, 4 ) garden ro to tillin g . M onitoring and confirm ation sam pling as part o f Phase 2 C ontam inant Screening Study o f L ibby properties to determ ine need for rem ediation. Sam pling to address risk assessm ent data gaps. Included indoor A B S (routine activities) and outdoor A B S (raking, m ow ing, playing), as w ell as clean-up evaluation sam ples. A m bient air m onitoring program for 14 stations in O U 4, 2 stations in O U2, 2 stations in O U6. Sam ples represent long-term (continuous 5-day) collection periods. Sam pling to assess exposures during indoor A B S (passive & active activities) and outdoor A B S (raking, m ow ing, playing) in OU4. S tationary air sam ple collection from w ithin L ibby public schools O utdoor A B S sam pling from L ibby public schools sim ulating exposures to students and m aintenance staff A m bient air sam pling. Sam ples represent long-term (continuous 5-day) collection periods. A B S air sam pling o f A T V riding, hiking, cam p fire construction Sam pling to m onitor air and dust levels alter com pletion o f clean-up activities at 31 properties. Includes v arious site-specific sam pling investigations (e.g., Stim son L um ber, Flyw ay, B N SF) and sm aller-scale sam pling program s. P ro g ram D ate R ange D ec 1999 - present Jun 2000 - present M ar - N ov 2001 Apr 2008 - N ov 2009 Apr 2003 - O ct 2006 Jun 2005 - O ct 2006 O ct 2006 - Jun 2008 Jul 2 0 0 7 -Jun 2008 Dec 2008 Jul - Sept 2009 July - O ct 2008 Aug - N ov 2009 Nov 2003 - Feb 2004 Aug 2001 - present Sam pling and A nalysis P lan (s) U .S. E P A (2000) U .S. E P A (2000) U .S. E P A (2001) U .S. E P A (2002) U .S. E P A (2005) U .S. E P A (2006): (2007c) U .S. E P A (2007b): (2007a) U .S. E P A (2008a) U .S. E P A (2009a) U .S. E P A (2008b) U .S. E P A (2009b) U .S. E P A (2003) various N um ber of LA S t r u c t u r e s <a) 328 18,525 867 1,717 3 1,456 136 5,603 2 5 67 59 5 731 (a) Restricted to LA structures recorded in accordance with a 5:1 aspect ratio rule. LA structure counts are based on a download o f Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10. Program 1A BN CR DM El EP FC FL SL LA Structures 9 17 3 1 1 104 184 146 266 Other Description AIRS Site (418 Mineral Ave) BNSF Cumulative Risk Study Demolition Sampling from 2006 only BNSF Rail Yard Exclusion Zones Export Plant Flower Creek WR Grace (Flyway site) Stimson Lumber This document is a draftfor review purposesonly and does not constitute Agency policy. B-9 DRAFT--DO NOT CITE OR QUOTE Figure 3-1. Particle Size Distributions of LA Particles in Libby Air Samples 1.0 0.9 0.8 0.7 0.6 u. OO 0 .5 0.4 0.3 0.2 0.1 0.0 0.1 L ength / 1 10 Length (um) Ml Air Samples -- 100 1.0 0.9 0.8 0.7 0.6 8 '5 0.4 0.3 0.2 0.1 0.0 0.01 W idth f J r J 0.1 1 Width (um) W\ Air Samples 10 Aspect Ratio Data are based on a download of Libby 2DB performed on 12-8-09 and the Libby OU3 DB on 2-9-10. All Air S a m p le s Number of Structures (29,504) Type Number Frequency F 23,933 1 81% B 2,366 ` 8% M 3,150 , 11% C 54 i 0.2% This document is a draftfor review purposesonly and does not constitute Agency policy. B-10 DRAFT--DO NOT CITE OR QUOTE Figure 3-2. Particle Size Distributions of LA Particles in Libby Air Samples by Structure Type S tru c tu re Type F B M N S tru c tu re s 2 3 ,9 3 3 2 ,366 3 ,150 Clusters have not been included in this figure because N = 54 and this in not believed to be a sufficient number of structures. This document is a draftfor review purposesonly and does not constitute Agency policy. B-ll DRAFT--DO NOT CITE OR QUOTE Figure 3-3, Particle Size Distributions of LA Particles in Libby Air Sam ples by Operable Unit (OU) OU N S tru c tu re s 1 447 2 7,421 3 4,382 4 13,005 5 335 This document is a draftfor review purposesonly and does not constitute Agency policy. B-12 DRAFT--DO NOT CITE OR QUOTE Figure 3-4. Particle Size Distributions o f LA Particles in Libby Air Sam plesby Air Type Sam ples Source Am bient Air Indoor ABS Outdoor ABS N Structures 136 891 5,953 This document is a draftfor review purposesonly and does not constitute Agency policy. B-13 DRAFT--DO NOT CITE OR QUOTE Figure 3-5. Particle S ize D istributions o f LA Particles in L ib by A ir Sam ples by Preparation M ethod P re p a ra tio n Direct Indirect N Structures 17,578 11,926 This document is a draftfor review purposesonly and does not constitute Agency policy. B-14 DRAFT--DO NOT CITE OR QUOTE Figure 3-6. Particle S ize D istributions o f LA Particles in L ib b y A ir S am ples b y Analysis M ethod A nalysis M ethod N S tru ctu res ISO 12,657 AHERA 16,847 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. B-15 DRAFT--DO NOT CITE OR QUOTE 5.0 REFERENCES AHERA. 1986. Asbestos Hazardous Emergency Response A ct. Title 20, Chapter 52, Sec. 4011. Public Law 99-519. ISO. 1995. International Organization for Standardization (ISO). Ambient Air - Determination of Asbestos Fibres - Direct-Transfer Transmission Electron Microscopy Method. ISO 10312:1995(E). USEPA. 2000. Sampling and Quality Assurance Project Plan Revision 1 for Libby, Montana. Environmental Monitoring for Asbestos. Baseline Monitoring for Source Area and Residential Exposure to Tremolite-Actinolite Asbestos Fibers. Report prepared by U.S. Environmental Protection Agency Region. January 4, 2000. USEPA. 2001. Phase 2 Sampling and Quality Assurance Project Plan (Revision 0) for Libby, Montana. Environmental Monitoring for Asbestos. Evaluation of Exposure to Airborne Asbestos Fibers During Routine and Special Activities. Report prepared by U.S. Environmental Protection Agency Region 8. March 2001. USEPA. 2002. Final Sampling and Analysis Plan, Remedial Investigation, Contaminant Screening Study, Libby Asbestos Site, Operable Unit 4. Report prepared by U.S. Environmental Protection Agency Region 8, with technical support from CDM. April 30, 2002. USEPA. 2003. Final Sampling and Analysis Plan Addendum, Post Clean-up Evaluation Sampling, Contaminant Screening Study, Libby Asbestos Site, Operable Unit 4. Report prepared by U.S. Environmental Protection Agency Region 8, with technical support from CDM and Syracuse Research Corporation. December 1, 2003. USEPA. 2005. Supplemental Remedial Investigation Quality Assurance Project Plan for Libby, Montana. Revision 1. U.S. Environmental Protection Agency Region 8. August 5, 2005. USEPA. 2006. Sampling and Analysis Plan for Outdoor Ambient Air Monitoring at the Libby Asbestos Site. Revision 1. Report prepared by U.S. Environmental Protection Agency Region 8, with technical support from CDM and Syracuse Research Corporation. December 7, 2006. This document is a draftfor review purposes only and does not constitute Agency policy. B -16 DRAFT--DO NOT CITE OR QUOTE USEPA. 2007a. Sampling and Analysis Plan for Outdoor Ambient Air Monitoring - Operable Units 1, 2, 5, and 6. Final Addendum prepared by U.S. Environmental Protection Agency Region 8, with technical support from CDM and Syracuse Research Corporation. July 3, 2007 USEPA. 2007b. Sampling and Analysis Plan for Activity-Based Outdoor Air Exposures, Operable Unit 4, Libby, Montana, Superfund Site. Final. U.S. Environmental Protection Agency, Region 8. July 6, 2007. USEPA. 2007c. Sampling and Analysis Plan for Activity-Based Indoor Air Exposures, Operable Unit 4, Libby, Montana, Superfund Site. Final. U.S. Environmental Protection Agency, Region 8. July 6, 2007. USEPA 2008a. Request for Modification to Laboratory Activities LB-000031A. Requested by Lynn Woodbury of Syracuse Research Corporation. January 18, 2008. USEPA. 2008b. Standard Operating Procedure for TEM Data Review and Data Entry Verification. SOP No. EPA-LIBBY-09 (rev 1). Prepared by U.S. Environmental Protection Agency, Region 8, with technical support from SRC, Inc. March 5, 2008. USEPA. 2008c. Phase II Sampling and Analysis Plan for Operable Unit 3 Libby Asbestos Superfund Site. Part B: Ambient Air and Groundwater. Prepared by U.S. Environmental Protection Agency Region 8, with technical support from Syracuse Research Corporation and NewFields Boulder LLC. July 2, 2008. USEPA. 2008d. Final Sampling and Analysis Plan Libby Public Schools - Stationary Air Sample Collection Libby Asbestos Site Libby, Montana. Prepared by U.S. Dept, of Transportation and CDM Federal Programs Corp. with technical support from Syracuse Research Corporation. December 5, 2008. USEPA. 2009a. Remedial Investigation for Operable Unit 3 Libby Asbestos Superfund Site. Phase m Sampling and Analysis Plan. Prepared by U.S. Environmental Protection Agency Region 8, with technical support from Syracuse Research Corporation and NewFields Boulder LLC. May 26, 2009. USEPA. 2009b. Final Sampling and Analysis Plan for Activity-Based Outdoor Air Exposures at Libby Public Schools Libby Asbestos Site Libby, Montana. Prepared by U.S. Dept, of Transportation and CDM Federal Programs Corp. with technical support from SRC, Inc. July 17, 2009. This document is a draftfor review purposes only and does not constitute Agency policy. B -17 DRAFT--DO NOT CITE OR QUOTE 1 APPENDIX C. CHARACTERIZATION OF AMPHIBOLE FIBERS FROM ORE 2 ORIGINATING FROM LIBBY, MT; LOUISA COUNTY, VA; AND PALABORA, 3 REPUBLIC OF SOUTH AFRICA By David L. Berry, PhD 4 U S. EPA Region 8 5 Ecosystems Protection and Remediation 6 1595 Wynkoop Street [8EPR-PS] 7 Denver, CO 80202-1129 8 This document is a draftfor review purposes only and does not constitute Agency policy. C-l DRAFT--DO NOT CITE OR QUOTE 1 The O.M. Scott plant in Marysville, OH manufactured a number of products including 2 fertilizers, dyes, and pesticides that were bound to a vermiculite carrier as a delivery vehicle. 3 The plant received ore from Enoree, SC; Louisa County, VA; Libby, MT; and Palabora, 4 Republic of South Africa, which was processed in an exfoliation furnace to produce vermiculite 5 used in the manufacture of their commercial products. Only ore from South Carolina was used 6 in 1957 and 1958. From 1959 to 1971, ores from South Carolina and Libby, MT were used. 7 From 1972 to 1980, ores from Libby, MT, South Africa, and Virginia were used. No ore from 8 Libby, MT was used after 1980. Only ore from South Africa and Virginia was used after 1980 9 (see Appendix F). 10 The U.S. Environmental Protection Agency (EPA) Region 8 obtained samples of ore 11 from Libby, MT, South Africa, and Virginia from Dr. James Lockey, University of Cincinnati, 12 and analyzed the samples to determine mineralogy and particle size distribution (length, width, 13 and aspect ratio) using transmission electron microscopy (TEM) and energy dispersive 14 spectroscopy (EDS) to identify the nature of the amphibole fibers. Dr. Lockey obtained the 15 South African and Virginia ore samples from the Marysville, OFI facility in 1980 and the Libby, 16 MT ore (Libby #3 ore) from an expansion plant in Salt Lake City, UT, in 1981. Region 8 was 17 unable to obtain vermiculite or ore from the Enoree, SC mine complex. 18 The ore from the Rainey Creek complex (Vermiculite Mountain Mine, Libby, MT) 19 resides in large ultramafic intrusive bodies that are rich in biotite, pyroxenite, and biotitite, a rock 20 compri sed of almost pure biotite. The ultramafic intrusi ons are cut by deposits of syenite and 21 carbonatite, and much of the bi otite has been hydrothermally altered to hydrobiotite and 22 vermiculite (Meeker et al,, 2003; Frank and Edmund, 2001). The pyroxenite has been altered to 23 fibrous soda-rich amphiboles, and contacts with pyroxenite surrounding the biotitite contain the 24 vermiculite ore zone containing diopside, hydrobiotite, and apatite. Fibrous and nonfibrous 25 amphiboles are located in both veins and disseminated throughout the intrusive rock along 26 cleavage planes of pyroxene. Amphiboles from Vermiculite Mountain had been referred to as 27 soda tremolite, richterite, soda-rich tremolite, tremolite asbestos, and richterite asbestos by a 28 number of investigators. In 2000, Wylie and Verkouteren (2000) identified winchite as the 29 principal amphibole in the Vermiculite Mountain deposit based on chemical investigation 30 referencing the classificati on system of Leake et al. (1997) and optical properties. Meeker et al. 31 (2003) investigated amphibole types from the mine complex using electron probe microanalysis This document is a draftfo r review purposes only and does not constitute.A gency policy. C-2 DRAFT--DO NOT CITE OR QUOTE 1 and X-ray diffraction analysis and reported the presence of winchite, richterite, tremolite, and 2 magnesioriebeckite. Magnesio-arfvedsonite and edenite were detected in low abundance. The 3 amphibole composition of the Libby Amphiboles is roughly winchite, richterite, tremolite, 4 magnesio-riebeckite, magnesio-arfvedsonite, and edenite (84:11:6:<1:<1:<1). The O.M. Scott 5 facility received ore from the Vermiculite Mountain mine complex, Libby, MT from 1959 6 through 1980. 7 The Palabora Igneous Complex, located near Phalaborwa, Republic of South Africa, is 8 the location of the Palabora mine. The Palabora ore deposit shares many features with the 9 Vermiculite Mountain mine complex--including zoned deposits with ultramafic rocks 10 (pyroxenite) and intrusion by alkalic rock, primarily syenite. The primary mica at Palabora is 11 phlogopite rather than biotite, and the primary alteration product that forms vermiculite ore is 12 hydrophlogopite rather than hydrobiotite (Schoeman. 19891. 13 The Palabora ore is reported to contain little or no asbestiform fibers based on polarized 14 light microscopy by the Institute of Occupational Medicine in Edinburgh (TOM Consulting. 15 2008). Crude vermiculite from the Palabora complex was also reported to be free of asbestiform 16 fibers by polarized light microscopy (TOM Consulting. 2008). In both reports, the analysis by 17 polarized light microscopy was conducted with a detection limit of 1 ppm, and, since no 18 chrysotile or amphibole structures were detected, no further analysis by electron microscopy and 19 X-ray diffraction were conducted. 20 The ore from the Virginia Vermiculite mine in Louisa County, VA is described as mafic 21 rock intruded by a series of small pegmatites (Gooch. 1957). Meisinger (1979) classified the 22 deposits as Type 3, similar to the ores from Enoree, SC. The formations consist of potassic 23 ultramafic bodies, primarily biotite. The vermiculite ores are found primarily in hydrobiotite 24 portions of the biotite intrusions. The hydrobiotite deposits are preferentially mined because of 25 better commercial properties compared to vermiculite. 26 There is limited information on the asbestos content of the ores from the Louisa County 27 deposit. Rohl and Langer (1977) reported both chrysotile and amphibole fibers in six ore 28 samples from the Louisa County deposit. The chrysotile was reported as fibers and bundles 29 while the amphiboles fibers were classified as actinolite. Moatamed et al. (1986) analyzed a 30 Virginia ore sample collected at a processing plant in Salt Lake City, UT and reported traces of 31 fibrous amphibole asbestos identified as actionlite in the form of cleavage fragments having low This document is a draftfo r review purposes only and does not constitute.A gency policy. C-3 DRAFT--DO NOT CITE OR QUOTE 1 aspect ratios. Amphibole content for both unexfoliated and exfoliated ores ranged up to 1.3% 2 amphibole asbestos. 3 Ores from the Enoree, SC deposits are primarily hydrobiotite and biotite in origin. 4 Fluroapatite is a common mineral collocated with the hydrobiotite. Zircon is also widely 5 dispersed throughout the plutons along with minor accessory minerals including talc, chlorite, 6 chromite, rutile, titanite, corundum, anatase, and amphibole asbestos (Hunter, 1950). The 7 amphibole asbestos identified in the vermiculite deposit at Enoree, SC has been classified as 8 tremolite (Libby, 19751. 9 As previously noted, EPA Region 8 obtained samples of ore from Libby, MT, South 10 Africa, and Virginia from Dr. James Lockey, University of Cincinnati, and analyzed the samples 11 to determine the particle-size distribution (length, width, and aspect ratio), using TEM and EDS 12 to identify the mineral composition of the amphibole fibers. Region 8 was unable to acquire a 13 sample of ore from the South Carolina Enoree mine complex for analysis. Region 8 conducted 14 analysis of the ore and exfoliated materials to connect the exposures of workers to mineral fibers 15 in Marysville, OH, to the ore originating in Libby, MT. The connection is based on fiber 16 morphology, mineralogy, and fiber-size similarities. 17 In order to analyze the fibers from the ore and vermiculite bulk material, the fibers must 18 be loaded onto filters and prepared for analysis by TEM. Three potential methods were 19 considered for transferring the fibers from the bulk material to filters: water elutriation, 20 glove-box transfer, and the fluidized bed asbestos segregator (FBAS). Of these three methods, 21 only the glove-box and FBAS involved physical disturbance of the bulk material to elutriate 22 fibers into the air that might be similar to handling and processing of ore in the Marysville, OH 23 plant. Due to the limited quantity of test material available for analysis, Region 8 employed the 24 FBAS as an analytical instrument to load the mineral fibers onto filters for TEM analysis. 25 Bri efly, samples of ore and vermi culite were prepared following the procedure outlined 26 by Bern et al. (2002). Samples were dried, ground with a Wylie mill and mortar and pestle, and 27 sieved through a 230-pm (60 mesh) sieve. Samples (exactly 2.0 g) were mixed with 18 g of 28 analytical silica sand and placed in a FBAS vessel to load 25-mm mixed cellulose ester air 29 sampling filters (0.8-p pore size). The FBAS was run for 3 minutes to load the filter cassettes 30 with sufficient fibers for analysis by TEM . Five filters were loaded for each of the ore and This document is a draftfor review purposesonly and does not constitute Agency policy. C-4 DRAFT--DO NOT CITE OR QUOTE 1 vermiculite samples. After loading, the filters were prepared for TEM analysis by mounting on 2 copper grids, carbon coating, and subjected to TEM analysis (TEM-ISO 10312 method). 3 The laboratory followed fiber counting rules detailed in the Quality Assurance Project 4 Plan for the specific study using Libby-specific laboratory modifications. Total amphibole fibers 5 and Phase Contrast Microscopy equivalent (PCMe) fibers were counted for each of the 6 ore/vermiculite samples as described in Appendix B. A total of 1.0 mm area or a total of 7 200 asbestos structures were counted to achieve the desired analytical sensitivity (1/g; 1.5 x 104). 8 DS was performed on selected samples from each of the vermiculite/ore samples to provide 9 mineral characterization of individual fibers. Fiber counts were recorded on National Asbestos 10 Data Evaluation Sheet data sheets for further analysis. Only the Libby, MT vermiculite and 11 Libby, MT ore samples had sufficient fibers detected to construct a fiber-size distribution. 12 Fiber counts were determined by counting fiber numbers for a specific area of the filter 13 grid or a specific number of grid openings (whichever was achieved first) to determine total 14 fibers present. As shown in Table C-l, the number of fibers for the test materials varied greatly 15 depending on the source, and the grid area measurement was exceeded prior to the fiber count 16 metric (167 grid openings ~1.0 mm ). 17 18 Table C-l. Fiber detected in ore and expanded product Structures counted Concentration (s/g) Sample type Grid openings LA OA c LA OA c Virginia Ore 167 00 0 0 0 0 Virginia Expanded 167 10 0 13,008 0 0 South Africa Ore 167 20 2 26,403 0 26,403 South Africa Expanded 167 00 0 0 0 0 Libby #3 Ore 167 320 0 0 1,393,873 0 0 Libby Expanded 167 100 0 0 468,213 0 0 19 20 LA = Libby Amphibole, OA = Other amphibole, C = Chrysotile. Note: the designation of fibers as Libby Amphibole 21 in this instance reflects only a qualitative morphological comparison to amphiboles o f the Libby, MT series. 22 23 24 The Libby #3 ore and the Libby #3 expanded material contained the greatest number of 25 fibers both in fiber counts on the filters and in calculated structures per gram of bulk material. This document is a draftfor review purposes only and does not constitute Agency policy. C-5 DRAFT--DO NOT CITE OR QUOTE 1 Virginia expanded and South African ore contain amphibole structures represented by low fiber 2 counts. South African ore also contained chrysotile fibers as determined by morphology and 3 EDS analysis. The absence of fibers detected in the Virginia ore and the South African- 4 expanded materials probably represents actual low fiber content of the ore and is a function of 5 the detection limit for the structure analysis. The estimation of structures per gram of material 6 indicated that there were 13,000 to 26,000 fibers per gram of bulk material, which was 7 approximately 18 times lower than the Libby, MT ore samples. The decrease in fibers found in 8 the Marysville, OH facility after 1980 when only ore from Virginia, Palabora, and South 9 Carolina was used (see Appendix F) is consistent with the findings of low fiber counts for the 10 Virginia and Palabora materials. In addition, numerous nonasbestiform minerals were also 11 detected including biotite, micas, and pyroxenes in the bulk materials from Virginia and South 12 Africa. 13 Amphiboles are a complex group of minerals characterized by double chains of silicate 14 tetrahedrons and the generic chemical formula of A() ]B2Cs where A, B, C, and T 15 represent the various cations. The modem classification system of amphiboles is described in 16 Leake et al. (19971. To classify the mineral speci es of the amphibole, it is not sufficient to 17 determine its composition; the various cations must be assigned to the specific^, B, C, and T 18 sites. The cutoffs of the compositional ranges allowed for each amphibole mineral species are 19 based on the number of the cations in the various sites. The methodology to classify an 20 amphibole is to first determine its elemental compositions (e.g., as expressed as weight percent 21 oxide for each element or as atomic percent for each element). Then a normalized routine is 22 applied to the raw elemental measurements to calculate the number of each of the cations 23 contained in one formula unit. (This is a simple arithmetic calculation since the cation percents 24 have been measured, and the stoichiometry must balance the charges of the cations and anions.) 25 Generally, one formula unit is assumed to contain 23 oxygens. Next, the sites are filled up by 26 assigning cations to them subsequently, specifically: 27 28 This document is a draftfor review purposesonly and does not constitute Agency policy. C-6 DRAFT--DO NOT CITE OR QUOTE 1 T: Si4 , A!3 , and Ti4 , 2 C: Al3+and T i1 (only after the T sites are filled first) and then Mg' . IT , Fe34, and 3 then M if . 4 B: Any remaining Mg2+, Fe2+, and Mir (after the C sites are filled), all Ca2+, then 5 N a' if there is any room left. 6 A: Na+and K7 only. 7 8 9 Once the cations are assigned to their sites, it is a simple matter to classify the minerals 10 based on the cutoffs of the composition field allowed for each mineral. 11 The Libby Amphibole asbestos1group of minerals is a complex group of amphiboles 12 consisting of six minerals: 13 14 15 Winchite, CaNa[Mg, Fe' ]+[AI, Fe' ]Six022 [OHJ3 16 Richterite, NaCaNa [Mg, Fe2+, Mn, Fe3 ]}Six022[0 H]f 17 Tremelite, Ca2Mg5Si8022[0FI]2 18 Magnesio-riebeckite, Na2[Mg3, Fe3 2]Six0 22[0 B]2 19 Magnesio-arfvedsonite, NaNa2[Mg4,Fe' ]Six022[0H]2 20 Edenite, NaCa2Mg5Si7A1022[OFI]2 21 22 23 Libby Amphibole is characterized by a low amount of A1 in the T site--and a 24 correspondingly high Si content--so, according to Leake's (1997) classification, if the Si 25 (expressed as atoms per formula unit, apfu) is at least 7.5, and A1 content in the T site is <0.5, all 26 6 Libby Amphibole types can be plotted on a graph of Na content of the B site versus the 27 (Na + K) content in the A site. This approach was described by Meeker et al. (2003) for the 28 Rainy Creek complex. 29 EDS spectra (TEM/EDS) were collected from all amphibole fibers found in the South 30 Africa and Virginia samples, and six randomly selected Libby Amphibole asbestos fibers in each 'The tenn "Libby Amphibole asbestos" is used in this document to identity the mixture of amphibole mineral fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.), that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfo r review purposes only and does not constitute Agency policy. C-7 DRAFT--DO NOT CITE OR QUOTE 1 of the Libby, MX ore and Libby, MT expanded samples. Two bundles of asbestiform serpentine 2 (chrysotile) were found in the South African ore sample. EDS spectra were collected for one of 3 the bundles. The chemical formula of serpentine is Mg3Si20s[0H]4. The EDS software package 4 collected and summarized each spectrum to determine the atomic percent of each element of 5 interest. 6 Several assumptions were made in the treatment of the TEM/EDS data: 7 8 9 1. Numbers of cations per formula unit are calculated on the basis of 23 oxygens. This may 10 or may not be correct because an [OH] site in the amphibole crystal can be occupied by 11 either OLE, IT, CL. or O . The calculated cation numbers will be affected if a significant 12 quantity of O" is in the OH site. 13 2. A persistent problem with amphiboles is that they can contain both ferric [3+] and ferrous 14 [2+] iron in the same crystal. For the purposes of this report all Fe was assumed to be 15 Fe . A method for calculating the ratio of Ic to Fe ' is described in Leake et al. 16 Q997), but it is very complex, applies to polished sections, and was not attempted for this 17 report. 18 3. For the purposes of this report, the T sites were assumed to be filled completely full to 19 8 apfu, and the C sites were assumed to be completely full to 5 apfu. All Ca and any Mg, 20 Fe, and Mn remaining after the C site was full were then assigned to the B site. Next, Na 21 was assigned to the B site until it was full (2 apfu), then any remaining Na and all K were 22 assigned to the A site. 23 24 25 Applying these assumptions to the TEM/EDS data produces a useable graph of the Na 26 and K content of the amphibole fibers. As shown in Figure C-l, Libby #3 ore and Libby #3 27 Expanded amphiboles were characteristic of winchite and tremolite. Virginia Expanded and 28 South African ore both contained amphibole fibers characteristic of non-Libby (Na and K) in the 29 tremolite series. 30 31 This document is a draftfor review purposesonly and does not constitute Agency policy. C-8 DRAFT--DO NOT CITE OR QUOTE Sodium and Potassium Content of Individual Amphibole Fibers as Measured by TEM/EDS 1.0 i_ o 0Q}. W 3 ~ --(0' E3 0) C/> < E * + m E d e n ite R ic h te rite A 0.5 _____________________ _-Aj T re m o lite A c tin o lite " W in c h ite . A A A j 0.0 0.0 0.5 1.0 1.5 Na in B Site (atoms per formula unit) o Virginia Expanded <" South Africa Ore a Libby #3 Ore Libby #3 Expanded 1 2 Figure C-l. Cation values for Na in the B site and the Na + K in the A site 3 from individual amphibole fibers. 4 5 6 Following all assumptions described above and the approach of plotting Na in the B site 7 versus Na + K in the A site as described by Meeker et al. (2003), the mineral species of the 8 Marysville, OH fibers can be described as: 9 10 11 The single Virginia amphibole asbestos fiber is an actinolite 12 Both of the South African amphibole fibers are tremolite 13 8 of the Libby Amphibole asbestos fibers from Libby, MT are winchite 14 4 of the Libby Amphibole asbestos fibers from Libby, MT are tremolite This document is a draftfor review purposesonly and does not constitute Agency policy. C-9 DRAFT--DO NOT CITE OR QUOTE Length Length (urn) Aspect Ratio 1 2 Figure C-2. Fiber-size distribution of Libby Amphibole asbestos. This document is a draftfor review purposes only and does not constitute Agency policy. C-10 DRAFT--DO NOT CITE OR QUOTE 1 Fiber-size distributions for amphibole fibers from the Libby #3 ore and Libby #3 2 expanded sources were conducted on the fibers counted during the TEM analysi s of the filter 3 grids. Due to the low fiber count detected in the Virginia and South Africa sources, it was not 4 possible to develop a fiber-size distribution for these fibers. The Libby Amphibole asbestos 5 fiber-size data were plotted as a cumulative distribution frequency for fiber length, fiber width, 6 and aspect ratio. These data were compared to Libby Amphibole asbestos fibers collected in 7 Libby, MT as part of EPA's ongoing ambient air monitoring program and the Libby Asbestos 8 Superfund site (see Appendix B). The Libby, MT ore and expanded material showed an 9 increased frequency of longer and wider fibers than the fibers from the Libby, MT ambient 10 air-sampling program. Aspect ratios were nearly identical The differences between the length 11 and width frequency were not outside of the expected range for Libby Amphibole asbestos fibers 12 and were consistent with fiber-size distributions for soil activity-based-sampling data from 13 Libby, MT. 14 Based on the TEM'morphological analysis of filter grids, TEM/EDS analysis for the fiber 15 mineralogy, and the fiber-size distribution data, it can be concluded that the amphibole fibers 16 detected in the Libby # 3 ore samples from the Salt Lake Expansion facility are consistent with 17 data from authentic Libby Amphibole fibers (Meeker et al.. 2003) found in Libby, MT (see also 18 Appendix B). Further, ore samples from Virginia and South Africa contained amphibole and 19 chrysotile fibers but at a much lower frequency of detection than the Libby Amphibole ore as 20 reported in Appendix F. 21 22 This document is a draftfor review purposesonly and does not constitute Agency policy. C-ll DRAFT--DO NOT CITE OR QUOTE C.l. REFERENCES Bern, AG; Meeker. GP: Brownfield. I. (2002). Guide to analysis o f soil samples from Libby, Montana for asbestos content by scanning electron microscopy and energy dispersive spectroscopy. In USGS Open File Report. USGS. ' .......... ' " ' ' Frank. D; Edmund. L. (2001). Feasibility for identifying mineralogical and geochemical traces from venniculite ore deposits. (EPA 910-R-01-002). Seattle, WA: U.S. EPA Region 10. Gooch. EO, (19571. Venniculite. 3:1-5. Hunter. CE. (1950). Venniculite o f the southeastern states. In Symposium on mineral resources o f the southeastern United States. Knoxville, TN: University o f Teimessee Press. IOM Consulting. (2008). Sampling and analysis o f cmde venniculite samples for possible asbestifonn fibre and quartz content. (609-02386). Suney, England: Palabora Mining Co, Palabora Europe Ltd. Leake. BE: Woollev, AR: Arps. CES: Birch. WD: Gilbert. MC: Grice. JD: Hawthorne. FC: Kato. A: K iscli HJ: Krivovichev. VG: Linthout. K: Laird. J: Mandarino. J: Maresch. WV: Nickel. ELI: Rock. NMS: Schumacher. JC: Smith. DC: Shenhenson, NCN: Ungaretti. L: Whittakc. EJW: Youzlii. G. (1997). Nomenclatme o f amphiboles: Report o f the Subcoimnittee on Ampliiboles o f the International Mineralogical Association Commission on New Minerals and Mineral Names. Mineral Mag 61: 295-321. Libby, SC. (1975) The origin o f potassic ultramafic rocks in the Enoree "Venniculite" District, South Carolina Pennsylvania State University, University Park, PA. Meeker. GP: Bern. AM: Brownfield. IK: Lowers. HA: Sutlev. SJ: Hoefen. TM: Vance. JS. (2003). The composition and morphology o f ampliiboles from the Rainy Creek Complex, near Libby, Montana. American Mineralogist 88: 1955-1969. http://dx.doi.org/10.1093/annlivg/38.inlialed particles V1I.639. Meisinger. AC. (1979). Venniculite. In Minerals Yearbook 1978-1979, Metals and Minerals. Washington, DC: U.S. Bureau o f Mines. Moatamed, F: Lockev. JE: Parry. . (1986). Fiber contamination o f vermiculites: a potential occupational and environmental health hazard. Environ Res 41: 207-218. Rolil. AN: Langer. AM. (1977). Mineral analysis o f core samples from the Green Springs area. Virginia venniculite deposit: Unpublished letter report from Mt. Sinai School o f Medicine. Schoemaa JJ. (1989). Mica and venniculite in South Africa. J South African Institute o f Mining and Mineralogy 89: 1- 12. Wylie, AG: Verkouteren. JR. (2000). Amphibole asbestos from Libby, Montana: Aspects o f nomenclature. American Mineralogist 85: 1540-1542. This document is a draftfor review purposesonly and does not constitute Agency policy. C-12 DRAFT-- DO NOT CITE OR QUOTE 1 APPENDIX D. ANALYSIS OF SUBCHRONIC AND CHRONIC STUDIES AND 2 CANCER BIOASSAYS IN ANIMALS AND MECHANISTIC STUDIES 3 D.l. SUBCHRONIC AND CHRONIC STUDIES AND CANCER BIOASSAYS 4 D .l.l. Oral 5 McConnell et al. (1983) describe part of a National Toxicology Program study (NTP. 6 1990a, b, 1988, 1985) performed to evaluate the toxicity and carcinogenicity of ingestion of 7 several minerals. This study examined chrysotile and amosite in both hamsters and rats, and 8 crocidolite and tremolite only in rats. This chronic bioassay was designed to encompass the 9 lifeti me of the animal, includi ng exposure of the dams from which the test animals were derived, 10 Although the study examined chrysotile, amosite, crocidolite, and tremolite, for the purposes of 11 this assessment, the focus is on the results from exposure to tremolite. The tremolite (Govemeur 12 Talc Co., Govemeur, NY) used was not fibrous. Instead, the material was crystalline, as this 13 form was a common contaminant in talc at the time of these studies (McConnell et al., 1983) (see 14 Table D-l). Citing the Stanton et al. (1981) paper, McConnell et al. (1983) stated that crystalline 15 tremolite can become fibrous upon grinding. Tremolite was incorporated by 1% weight into 16 NIH-31 feed and given to 250 male and female F344 rats from birth until death (118 male and 17 female controls). 18 19 20 Table D-l. Fiber characteristics and distribution of fibers analyzed in feed 21 studies in F344 rats 22 Length interval3 Characteristic <3 pm >3 pm, <5 pm >5 pm, <10 pm >10 pm Mean width 0.77 1.78 2.87 5.22 Tremolite particles 120 61 17 49 % o f Tremolite particles 19.4 9.85 3 8 23 24 aAverage groups, more detailed in primary paper. 25 ' .............. 26 Source: McConnell et al. ( 1983). 27 28 29 No significant tumor induction was observed in the animals with oral exposure to 30 tremolite. Although non-neoplastic lesions were observed in many of the aging rats, these were This document is a draftfor review purposes only and does not constitute Agency policy. D-l DRAFT--DO NOT CITE OR QUOTE 1 mostly in the stomach and occurred in both controls and exposed animals. The lesions included 2 chronic inflammation, ulceration, and necrosis of the stomach (McConnell et al.. 1983). 3 McConnell et al. (1983) suggested that nonfibrous tremolite could account for the lack of 4 toxicity following exposure in this group of animals. Also, oral studies of asbestos, in general, 5 show decreased toxicity and carcinogenicity as compared to inhalation and 6 implantation/injection studies. 7 8 D.1.2. Inhalation 9 Davis et al. (1985) performed a chronic inhalation study examining response to tremolite 10 asbestos. Groups of 48 specific-pathogen-free (SPF) male Wistar rats were exposed in a 11 chamber to 10 mg/m (~1,600 fibers/mL, >5 pm) of commercially mined tremolite (South 12 Korea) for a total of 224 days (7 hours per day, 5 days per week) over a 12-month period. The 13 tremolite sample contained approximately 50% fibers 10-100-pm long, using a fiber definition 14 of length = >5 pm, diameter = <3 pm, and aspect ratio = >3:1. The results of the inhalation 15 study produced very high levels of pulmonary fibrosis, as well as 16 carcinomas and 16 2 mesotheliomas, among the 39 tremolite-exposed animals (see Tables D-2 and D-3). No 17 pulmonary tumors were observed in the controls. 18 19 20 Table D-2. Pulmonary fibrosis and irregular alveolar wall thickening 21 produced by tremolite exposure 22 Time after start of exposure (num ber of rats examined) 12 mo ( = 3) 18 mo (n = 4) 27-29 mo (n =12) Peribronchiolar fibrosis (SD)a 23.0 (21.4-24.2) 13.4 (9.7-18.9) - Irregular alveolar wall thickening (SD)b 35.2 (27.7-41.0) 27.7 (20.8-35.4) - Interstitial fibrosis (SD)b 0 23 24 aPercentage o f 100 squares counted in lung tissue area. 25 `'Percentage o f total lung tissue area. 26 27 SD = standard deviation. 28 29 Source: Adapted from Davis et al. (1985). 30 31 3.0 (0-5.6) 14.5 (3.8-26.9) This document is a draftfor review purposes only and does not constitute Agency policy. D-2 DRAFT--DO NOT CITE OR QUOTE 1 Table D-3. Tum ors (benign and malignant) produced by tremolite exposure 2 Tumor site Control (n = 36) Tremolite (n = 39) Pulmonary Adenomas 0 2 Adenocarcinomas 0 8 Squamous carcinomas 0 8 Mesotheliomas 0 2 Other organ systems Digestive/peritoneal 5 3 Urinogenital 3 1 Endocrine 3 5 Musculoskeletal, integumentary 5 5 Rcticulocndothelial/vascular 20 15 3 4 Source: Adapted from Davis et al. (1985). 5 6 7 Although Davis et al. (1985) did not describe the data, the difference between tremolite 8 and chrysotile was stated to be statistically significant, with tremolite exposure inducing more 9 fibrotic and carcinogenic lesions (see Table D-2). These results show that rats exposed to 10 tremolite exhibited increased numbers of pulmonary lesions and tumors. Tumors observed in 11 other organ systems are also listed in Table D-3 and appear to be unrelated to exposure. 12 Although a method for an injection study is described in Davis (1985). only the inhalation results 13 are presented. This same tremolite was used in later intraperitoneal injection experiments (Davis 14 et al., 1991) and might be what the authors are referring to in this article. 15 Wistar rats were exposed for 13 consecutive weeks (6 hours per day, 5 days per week) to 16 either Calidria chrysotile asbestos or tremolite asbestos in a flow-past, nose-only inhalation study 17 (Bernstein et al., 2003) (see Table D-4). The long-term effects from the same exposure were 18 described in Bernstein et al. (2005) (6 hours per day, 5 days per week). This study describes the 19 full results through 1 year after cessation of tremolite exposure in Wistar rats (n = 56). The 20 tremolite samples were chosen to have 100 fibers/mL of fibers longer than 20 pm present in the 21 exposure aerosol. Fibers were defined as any object with an aspect ratio >3:1, length >5 pm, and 22 diameter <3 pm, and all other objects were considered nonfibrous particles. Counting was 23 stopped when nonfibrous particle counts reached 30, and fiber counting was stopped at 500 with This document is a draftfor review purposes only and does not constitute Agency policy. D-3 DRAFT--DO NOT CITE OR QUOTE 1 Table D-4. C hrysotile and trem olite fiber characteristics of fibers used in 2 inhalation exposure studies in rats 3 Fiber type Mean no. fibers evaluated Mean no. total fibers/mL Mean % total fibers, >20 pm length Mean diameter (pm) SD Mean length Diameter Length (pm) SD range (pm) range (pm) Chrysotile 2,016 48,343.2 0.4 0.08 0 .0 7 3.61 7 .3 7 0.02-0.7 0.07-37.6 Tremolite 1,627 3,128.1 3.4 0.32 3 .5 2 5.49 : 13.97 0.1-3.7 0.9-75 4 5 Source: Bernstein et al. (2003). 6 7 8 length >5 urn, diameter <3 pm, or a total of 1,000 fibers and nonfibrous particles were recorded 9 (Bernstein et a!.., 20031. Lung tissue and associated lymph nodes were examined by 10 histopathology following tissue digestion. Associated lymph nodes showed erythrophagocytosis 11 (minimal severity) in one animal at all time points, compared to chrysotile and control, which 12 showed erythrophagocytosis (minimal severity) only at 180 days. 13 Table D-4 shows the comparison of number, concentration, and mean size distribution of 14 fibers used in this study. Note that the mean tremolite fiber diameter and length are much greater 15 than those of chrysotile, but the size ranges do overlap somewhat (Bernstein et al.. 2003). The 16 long tremolite fibers, once deposited in the lung, remain throughout the rat's lifetime. Even the 17 shorter fibers, following early clearance, remain with no dissolution or additional removal. At 18 365 days postexposure, the mean lung burden was 0.5 million tremolite fibers >20-pm long and 19 7 million fibers 5-20-pm long with a total mean lung burden of 19.6 million tremolite fibers. 20 The tremolite-exposed rats showed a pronounced inflammatory response in the lung as early as 21 1 day postexposure, with the rapid development of granulomas (1 day postexposure) followed by 22 the development of pulmonary fibrosis characterized by collagen deposition within the 23 granulomas. Increases in alveolar macrophages and granulomas were observed at all time points 24 (1, 2, 14, 90, and 180 days) measured except 365 days. Pulmonary fibrosis increased starting at 25 14 days and continued to be observed for up to 365 days. Slight interstitial fibrosis also was 26 observed, but only at 90 and 180 days postexposure. This study demonstrates that tremolite 27 exposure leads to pronounced inflammation and fibrosis (Bernstein et al.. 2006). Tumors were 28 not observed in this study, which is a consistent observation with the time frame observed in 29 other studies (i.e., 1-year postexposure) (Smith. 1978). This document is a draftfor review purposes only and does not constitute Agency policy. D-4 DRAFT--DO NOT CITE OR QUOTE 1 D.1.3. Intratracheal Instillation 2 A recent study by Putnam et al. (2008) was designed to explore gene-environment 3 interactions in the development of asbestos-related diseases. C57B1/6 mice were exposed once 4 to either Libby Amphibole asbestos 1 (Six Mix) (100 pg via intratracheal instillation); crocidolite 5 (100 pg via intratracheal instillation); or saline (30 pL via intratracheal instillation). 6 Characteristics of fibers are described in Table D-5. Animals were sacrificed, and the lungs were 7 harvested 6 months postinstillation. The left lung was used for ribonucleic acid (RNA) isolation, 8 and the right lung was used for histology (personal communication, e-mail from E. Putnam 9 [University of Montana] to M. Gwinn [U.S. EPA] 02/26/09). Histology on mouse lungs from 10 each treatment group demonstrated an increase in fibrosis, as viewed by Gomori's trichrome 11 staining, following exposure to crocidolite and, to a lesser extent, Libby Amphibole asbestos. 12 Histologic tissue was also exposed to Lucifer Yellow stain to further analyze variability in 13 collagen following exposure. Lucifer Yellow staining revealed an increase in collagen following 14 exposure to both crocidolite and Libby Amphibole asbestos, but only crocidolite exposure led to 15 a statistically significant increase (p < 0.05). RNA was isolated from homogenized lungs and 16 purified for use in microarray analysis. Pooled RNA samples from mice in each exposure group 17 were analyzed on a OK-element mouse oligonucleotide array (MWG Biotech), and expression 18 was compared to a mouse reference standard RNA. Gene-expression results were analyzed by 19 GO Miner, and genes exhibiting at least 1.25-fold up- or down-regulation in treated lungs were 20 described. These included genes involved in membrane transport, signal transduction, epidermal 21 growth factor signaling, and calcium regulation for both crocidolite and Libby Amphibole 22 asbestos exposures, which support the increase in collagen observed above. Some limitations to 23 this study are the use of a standard reference for gene-expression comparisons (as opposed to the 24 saline controls), the practice of describi ng genes only if a greater than twofold difference in 25 expression is observed, and the use of pooled samples of homogenized whole lung that in some 26 cases could dilute variability between different areas of exposed lung (different lobes, fibrotic 27 versus nonfibrotic). 28 'The tern "Libby Amphibole asbestos" is used in this document to identity the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfo r review purposes only and does not constitute A gency policy. D-5 DRAFT--DO NOT CITE OR QUOTE 1 Table D-5. Fiber characteristics for in tratrach eal instillation studies in mice 2 M aterial Diameter Length Aspect Ratio Libby Amphibole asbestos (Six Mix) 0.61 1.22 pm 7.21 7.01 pm 22.52 22.87 Crocidolite 0.16 0.09 pm 4.59 4.22 pm 34.05 43.29 3 4 Source: Blake et al. (2008; 2007): Putnam et al. (2008); Smartt et al. (2010). 5 6 7 A follow-up paper to Putnam et al. (2008), prepared by Smartt et al. (2010) examined the 8 increase of collagen in C57B1/6 mouse lung following exposure to crocidolite or Libby 9 Amphibole asbestos and also examined a few specific gene alterations by quantitative reverse 10 transcription polymerase chain reaction (RT-PCR). Animals (n = 3 to 6 mice per group) were 11 dosed with the same samples (see Table D-5) as described above (Putnam et aL 2008) but were 12 euthanized at 1 week, 1 month, and 3 months postinstillation. Treated mice were then divided 13 into two groups, with the left lung from the first group used for RNA isolation and the right lung 14 used for histology. The lungs from the second group were used for protein isolation and 15 hydroxyproline assay (personal communication, e-mail from E. Putnam [University of Montana] 16 to M. Gwinn [U.S. EPA] 02/26/09). Similar to results from Putnam et al. (2008), GomorEs 17 staining demonstrated increased collagen and inflammation at the airways in lungs of mice 18 exposed to either Libby Amphibole asbestos or crocidolite. These results were similar following 19 exposure to both amphiboles, with crocidolite effects appearing more severe at all time points 20 examined. No changes in the pleura of the lungs that were indicative of potential mesothelioma 21 were observed; such changes, however, would not be expected in such a short time-frame. This 22 study also examined severity of inflammation and found that, on average, crocidolite-exposed 23 animals demonstrated minimal inflammation at 1 week postinstillation, which then progressively 24 worsened at 1 and 3 months postinstillation. Although both asbestos exposures led to increased 25 inflammation, Libby Amphibole asbestos exposure demonstrated minimal inflammation that did 26 not progress in the time points examined. Gene-expression alterations were measured by 27 quantitative RT-PCR for genes involved in collagen accumulation and scar formation (Coll Al, 28 CollA2, Col3Al). Although exposure to both forms of asbestos at 1 week and 1 month 29 postinstillation led to increased Col gene expression, the levels and subtypes altered varied. 30 Libby Amphibole asbestos exposure led to increased gene expression of Coll A2 at 1 week This document is a draftfor review purposes only and does not constitute Agency policy. D-6 DRAFT--DO NOT CITE OR QUOTE 1 postinstillation and Col3Al at 1 month postexposure, while crocidolite led to no significant 2 alterations in the expression of these genes. Both crocidolite and Libby Amphibole asbestos 3 exposure led to increased Coll A1 gene expression as compared to saline control at 1 week and 4 1 month postexposure. Due to these differences in expression, the authors also examined the 5 collagen protein levels in the lungs to compare to the gene-expression changes. Total collagen 6 content was determined by measuring the hydroxyproline content in the caudal aspect of the left 7 lung. As compared to saline-exposed mice, a significant increase in hydroxyproline was 8 observed at 1 week and 1 month following exposure to both crocidolite and Libby Amphibole 9 asbestos; however, only lungs from crocidolite-exposed animals demonstrated a significant 10 increase at 3 months postexposure. These studies demonstrate that exposure to Libby 11 Amphibole asbestos lead to inflammation and fibrosis, although with differences in the time and 12 level of response from those of crocidolite. 13 Shannahan et al. (2011a) exposed two rat models of human cardiovascular disease to 14 Libby Amphibole asbestos to determine if the preexisting cardiovascular disease in these 15 models would impact lung injury and inflammation following exposure. Llealthy Wistar Kyoto 16 (WKY) rats were compared to spontaneously hypertensive (SH) and spontaneously hypertensive 17 heart failure (SHHF) rats following exposure. These rat models demonstrate pulmonary iron 18 homeostasis dysregulation (Shannahan et al.. 2010). All rats (male only) were exposed to 19 0, 0.25, or 1.0 mg/rat via intratracheal instillation and were examined at 1 day, 1 week and 20 1 month postexposure. No changes were observed histopathologically, however, changes were 21 observed in markers of homeostasis, inflammation, and oxidative stress. Bronchoalveolar lavage 22 fluid (BALF) protein was significantly increased in both the SIT and SHHF rat models as 23 compared to controls as early as 1 week postexposure, y-glutamyl transferase (GGT) activity was 24 increased in a concentration-dependent manner with exposure to Libby Amphibole asbestos at 25 the earliest time point measured (1 day), and was more pronounced in WKY rats as compared to 26 SH and SHHF rats. Lactate dehydrogenase (LDH) activity was also elevated in all strains but 27 was more pronounced in the SITFLF rat model. Neutrophil increases were observed following 28 exposure in all strains, peaking at 1 day postexposure in all strains and persisting in the SH and 29 SHHF rats until 1 month postexposure. Macrophages showed similar results but persisted only 2Median fiber dimensions as determined by TEM: length= 3.59 pm; width = 0.23 pm; aspect ratio >5:1. This document is a draftfo r review purposes only and does not constitute A gency policy. D-7 DRAFT--DO NOT CITE OR QUOTE 1 in the SH rat model until 1 month postexposure. In order to determine any impact of exposure 2 on iron homeostasis, BALF ferritin and transferrin levels were measured in the lung. Increases 3 in ferritin and transferrin were observed in both SH and SHHF rats as compared to WKY 4 controls. Nonheme iron was also observed to be increased in only the SH rats at 1 day and 5 1 week postexposure. Markers of inflammation (macrophage inflammatory protein [MIPj-2) and 6 oxidative stress (heme oxygenase-1 [HO-1]) were elevated in both SH and SHHF as compared to 7 WKY rats at baseline, but limited exposure-related differences were observed. Limited changes 8 were also observed in ascorbate and glutathione levels in BALF and lung tissue. While 9 inflammation and cell injury were observed in all strains, no strain-related differences were 10 observed following exposure to Libby Amphibole asbestos (Shannahan et al., 2011a). In 11 conclusion, this study showed the potential for population variability related to cardiac disease in 12 response to exposure to Libby Amphibole asbestos, including markers of cellular injury, iron 13 homeostasis, and inflammation. 14 Shannahan et al. (201 lb) tested the hypothesis that Libby Amphibole asbestos will bind 15 iron and increase the inflammogenic activity of fibers in vitro and acute lung injury and 16 infl ammation in vivo. The authors examined the ability of Libby Amphibole asbestos to bind 17 exogeneous iron in an acellular system and evaluated iron-related alterations in the production of 18 reactive oxygen species (ROS). The authors also investigated the role of iron in the acute 19 inflammogenic response in vitro, using human bronchiolar epithelial cells, and in vivo using SH 20 rats by modulating fiber-associated iron concentrations. In a cell-free medium, Libby 21 Amphibole asbestos bound about 16 pg of iron/mg of fiber and increased ROS generati on about 22 threefold. Generation of ROS was reduced by treatment with deferoxamine (DEF), an iron 23 chelator. To determine the role of iron in Libby Amphibole asbestos ROS generation and 24 inflammation, BEAS2B cells (bronchiolar epithelial cell line) were exposed to Libby Amphibole 25 asbestos (50 pg), iron-loaded Libby Amphibole asbestos, or Libby Amphibole asbestos treated 26 with DEF. No conditions altered HO-1 or ferritin mRNA expression. Libby Amphibole 27 asbestos by itself markedly increased EL-8 gene expression, which was significantly reduced by 28 iron loaded Libby Amphibole asbestos, but increased with Libby Amphibole asbestos treated 29 with DEF. To determine the role of iron in Libby Amphibole asbestos-induced lung injury in 30 vivo, spontaneously hypertensive rats were exposed intratracheally to either saline (300 pi), DEF 3Median fiber dimensions as determined by TEM: length = 3.59 pm; width = 0.23 pm; aspect ratio >5:1. This document is a draftfor review purposes only and does not constitute Agency policy. D-8 DRAFT--DO NOT CITE OR QUOTE 1 (1 mg), ferric chloride (21 jig), Libby Amphibole asbestos (0.5 mg), iron loaded Libby 2 Amphibole asbestos (0.5 mg), or Libby amphibole asbestos plus DEF (0.5 mg). Neither ferric 3 chloride nor DEF increased bronchoalveolar lavage fluid (BALF) neutrophils compared to saline 4 at 24 hours after treatment. Libby Amphibole asbestos exposure led to a statistically significant 5 increase in HALF neutrophils (p < 0.05). Loading of iron on Libby Amphibole asbestos, but not 6 chelation, slightly decreased inflammation (Libby Amphibole asbestos + DEF > Libby 7 Amphibole asbestos > iron loaded Libby Amphibole asbestos). At 4 hours after exposure, Libby 8 Amphibole asbestos-exposed lung mRNA expression of MIP-2 was significantly reduced in rats 9 exposed to iron loaded Libby Amphibole asbestos, but increased by DEF (Libby Amphibole 10 asbestos + DEF > Libby Amphibole asbestos > iron loaded Libby Amphibole asbestos). Ferritin 11 mRNA expression was elevated in rats exposed to iron loaded Libby Amphibole asbestos 12 compared to the Libby Amphibole asbestos control. HO-1 expression was unchanged following 13 treatment with Libby Amphibole asbestos. The authors concluded that the acute inflammatory 14 response following exposure to Libby Amphibole asbestos might be modified by the fiber's 15 ability to complex iron, rather than redox cycling of fiber associated iron. The authors further 16 concluded that iron overload conditions may influence susceptibility to Libby Amphibole 17 asbestos-induced pulmonary disease. 18 Padilla-Carlin et al. (2011) investigated pulmonary and histopathological changes in a 19 male Fisher 344 rats following exposure to Libby Amphibole asbestos1 The rats were 20 administered a single dose of either saline, amosite (0.65 mg/rat), or Libby Amphibole asbestos 21 (0.65 or 6.5 mg/rat) by intratracheal instillation. At time from 1 day to 3 months after exposure, 22 bronchoalveolar lavage (BAL) was performed and the right and left lung was removed for 23 Rt-PCR and histopathological analysis, respectively. The results showed that amosite exposure 24 (0.65 mg/rat) resulted in a higher degree of pulmonary injury, inflammation, and fibrotic events 25 than the same mass dose of Libby Amphibole. Both amosite and Libby Amphibole resulted in 26 higher levels of cellular permeability and inj ury, inflammatory enzymes, and iron-binding 27 protein in both BAL fluid and lung tissue compared to saline controls. In addition 28 histopathological examination showed notable thickening of interstitial areas surrounding the 29 alveolar and terminal bronchioles in response to amosite and Libby Amphibole. However, 30 mRNA levels for some growth factors (e.g., PDGF-A and TGF-ip), which contribute to fibrosis, 4Median fiber dimensions as determined by TEM: length = 3.59 (.un; width = 0.23 pm; aspect ratio >5. This document is a draftfor review purposes only and does not constitute Agency policy. D-9 DRAFT--DO NOT CITE OR QUOTE 1 were down regulated at several time points. The authors concluded that on a mass basis amosite 2 produced greater acute and persistent lung injury in this study. 3 In an early study, Sahu et al. (19751 described histological changes in the lungs of mice 4 exposed individually to amosite, anthophyllite, and tremolite. Fibers were described only as 5 <30-pm long. Groups of 20 male albino Swiss mice were exposed to amosite, anthophyllite, and 6 tremolite at a single dose of 5 mg, and two animals from each group were sacrificed at 1,2, 7, 7 15, 30, 60, 90, 120, and 150 days postexposure. Microscopic results following exposure to 8 tremolite showed acute inflammation of the lungs at 7 days postexposure, including macrophage 9 proliferation and phagocytosis similar to that observed with amosite and anthophyllite. Limited 10 progression of fibrotic response was observed at 60 and 90 days postexposure, with no further 11 progression of fibrotic response. 12 Blake et al. (2008) and Pfau et al. (2008) examined the role of asbestos in autoimmunity. 13 Blake et al. (2008) performed in vitro assays with Libby Amphibole asbestos, and both studies 14 performed the in vivo assays with tremolite. C57BL/6 mice were instilled intratracheally for a 15 total of two doses each of 60-ug saline and wollastonite or Korean tremolite sonicated in sterile 16 PBS, given 1 week apart in the first 2 weeks of a 7-month experiment. Detailed fiber 17 characteristics were described in Blake et al. (2007) for wollastonite and Libby Amphibole 18 asbestos, but not for Korean tremolite (see Table D-5; wollastonite and Korean tremolite not 19 shown). 20 Blake et al. (2008) described autoantibody production, monitored biweekly with blood 21 samples from saphenous vein bleeds and then by cardiac puncture following euthanization. 22 Specific autoantibodies were identified by immunoblotting with known nuclear antigens. These 23 autoantibodies were then incubated with murine macrophage cells previously exposed to Libby 24 Amphibole asbestos, wollastonite, or vehicle control (binding buffer containing 0.01 M Hepes, 25 0.14 M NaCl and 2.5 mM CaCL). Only sera from mice exposed to tremolite showed antibody 26 binding colocalized with SSA/Ro52 on the surface of apoptotic blebs (Blake et aL 2008). 27 In Pfau et al. (2008), collected serum samples, and urine were checked for protein 28 bi-weekly for 7 months. By 26 weeks, the tremolite-exposed animals had a significantly higher 29 frequency of positive antinucl ear antibody tests compared to wollastinate and saline. Most of the 30 tests were positive for dsDNA and SSA/Ro52. Serum isotyping showed no major changes in 31 immunoglobulin subclasses (IgG, IgA, IgM), but serum IgG in tremolite-exposed mice decreased This document is a draftfo r review purposes only and does not constitute A gency policy. D-10 DRAFT--DO NOT CITE OR QUOTE 1 overall. Further, IgG immune complex deposition in the kidneys increased, with abnormalities 2 suggestive of glomerulonephritis. No increased proteinuria was observed during the course of 3 the study. Local immunologic response was further studied on the cervical lymph nodes. 4 Although total cell numbers and lymph-node size were significantly increased following 5 exposure to tremolite, percentages of T- and B-cells did not significantly change. Because 6 tremolite is part of the makeup of Libby Amphibole asbestos (6%), using tremolite-exposed mice 7 might yield a similar response to Libby Amphibole asbestos-exposed mice. This same effect has 8 been demonstrated following exposure to ultraviolet radiation in skin cells, suggesting a similar 9 mechanism (Saegusa et aL 20021. 10 11 D.1.4. Injection/Im plantation 12 LVG:LAK hamsters were intrapleurally injected with tremolite obtained from the Libby, 13 MT mine in an unpublished study by Smith (19781 prepared for W.R. Grace and Company. 14 These samples were identified as tremolite (22260p5; Sample 60) and 50% tremolite + 50% 15 vermiculite (22263p2, Sample 63). Both fiber samples were measured by optical phase 16 microscopy, and fibers were described as amorphous, irregularly shaped particles of about 5--15 17 pm diameter, with Sample 60 (tremolite) also containing the occasional fiber up to 30 pm long. 18 Fiber size for Sample 60 (tremolite) also was measured by scanning electron microscopy (SEM) 19 and was determined to have a geometric mean length of 2.07 pm, a geometric mean diameter of 20 0.2 pm, and an average aspect ratio of 10.36:1. Twenty-five milligrams of each of the two 21 samples were individually injected intraperitoneally into the pleural cavity of LVG:LAK 22 hamsters. Pathology was examined at approximately 3 months postexposure in 10 animals from 23 each group, with the remaining animals observed until death, or 600 days postexposure, 24 depending on the health of the animal. Average survivorships were 410, 445, and 421 days in 25 groups exposed to Sample 60, Sample 63, and saline, respectively (see Table D-6). Pleural 26 fibrosis was observed 3 months postexposure, and mesothelioma was observed in both treatment 27 groups between 350 and 600 days postexposure, with no mesotheliomas in control groups. 28 This document is a draftfor review purposes only and does not constitute Agency policy. D-l 1 DRAFT--DO NOT CITE OR QUOTE 1 Table D-6. Pleural adhesions and tumors following intraperitoneal injection 2 exposure in LVG:LAK hamsters (25 mg) 3 Endpoint Control Sample 63 (tremolite Sample 60 (tremolite) and vermiculite) Average adhesion rating"1 0 (n = 10) 3.3 (;?= 10) 3.6 (;?= 10) Total tumors/animalsc 8/59 8/58 16/61 Benign 3/59 2/58 5/61 Malignant 5/59 6/58 9/61 Mesothelioma 0/59 5/58 5/61 4 5 aAs analyzed in first group sacrificed (between 41 and 92 days postexposure). 6 bRating for pleural adhesions: 0 = no adhesions; 1 = minimal adhesions; 4 = extensive adhesions. 7 cThese include adrenal adenoma, adrenal adenocarcinoma, lymphoma, pulmonary adenocarcinoma, adrenal 8 and salivary carcinoma, mesothelioma, rhabdomyosarcoma, hepatoma, thyroid carcinoma, subcutaneous 9 carcinoma, and malignant melanoma. 10 11 Source: Smith (1978). 12 13 14 The Smith et ah (1979) study was designed to determine whether mesothelioma is a 15 nonspecific result of mesothelial cells trapped in fibrous pleural adhesions, occurring regardless 16 of fiber type. Earlier studies by this group suggested that fibrosis and tumors resulting from liber 17 exposure (chrysolite or glass) were related to fiber dimensions (>20-um long, >0.75-pm 18 diameter) (Smith and Elubert, 1974). Injected fibrous talc (FD-14) was used as a negative 19 control in earlier studies and led to limited fibrosis and no tumor formation. The characteristics 20 of the FD-14 sample are described in the proceedings of Smith (1974). No further information 21 could be found on the characteristics of the samples used in this study/ Because the talc 22 contained 50% tremolite, 35% talc, 10% antigorite, and 5% chlorite, it was considered a 23 tremolite sample by Smith (1978). When the sample was later analyzed independently by Wylie 24 et al. (1993), only 64 (12.8%) of 500 tremolite particles measured met the National Institute for 25 Occupational Safety and Flealth definition of a fiber (>3:1 aspect ratio). Wylie et al. (1993) note, 26 however, that very long fibers of the mineral talc, with narrow widths and fibrillar structure, 27 occur in this sample. A second tremolite sample (Sample 275) used by Smith et al. (1979) was 28 described as similar to FD-14, although no details were given. The last two samples were5 5This fiber is also analyzed in Wylie et al. (1993) and Stanton et al. (1981). This document is a draftfo r review purposes only and does not constitute A gency policy. D-12 DRAFT--DO NOT CITE OR QUOTE 1 prepared from a deposit of tremolitic talc from the western United States (Sample 31) and from a 2 specimen of asbestiform tremolite (Sample 72),6 respectively. 3 Each of the four samples was examined microscopically, although the data were not 4 reported in the paper by Smith et al. (1979). The average fibers in Sample 72 were long, thin, 5 crystalline fibers (>20-pm long, 0.4-ptn diameter). Sample 31 appeared to have fewer long, thin 6 fibers than Sample 72, and many of the fibers in this sample were acicular. The characteristics 7 of the FD-14 sample were determined by phase microscopy (Smith and Fhibert, 1974). but no 8 characterization method was reported for the other three samples in this study. Other samples 9 used by this group have been analyzed by both optical and electron microscopy (Smith. 1978; 10 Smith and Hubert, 1974). The limited information on the fiber characteristics of the samples 11 used in these studies is provided in Table D-7. Note that no information was provided 12 confirming the presence or absence of particles or fibers less than 5 pm in length in any of the 13 three papers by Smith (1974) or Smith et al. (1979; 1978). These data deficiencies limit the 14 interpretation of results from this study. 15 16 17 Table D-7. Fiber characteristics and numbers of resulting tumors following 18 intrapleural injection of 10- or 25-mg fiber samples into Syrian hamsters 19 Tumors/survivors at 10 mgb Tumors/survivors at 25 mgb Average Average 350 500 600 350 500 600 Sample length3 (pm) diameter3 (pm) days days days days days days FD-14 5.7 1.6 ND N/D ND 0/35 0/26 0/20 275 N/D ND 0/34 0/14 0/6 0/31 0/15 0/3 31 >20 <0.4 1/41 1/19 1/11 2/28 4/9 6/5 72 >20 <0.4 0/13 1/6 3/2 3/20 5/6 5/1 20 21 aAlthough average length and diameter are reported, what range o f fibers was counted is unclear. Smith (1978) 22 (unpublished) states that only fibers greater than 5 pm long are included. No other information is provided for 23 these samples. 24 'Numerator = cumulative number of animals with tumors; denominator = number o f survivors. 25 26 N/D = not described. 27 28 Source: Smith et al. (1979): Smith (1978): Smith (1974). 6Although the source o f this material is not reported, these studies parallel those in the unpublished studies performed by Smith et al. (1979) for W.R. Grace that used material from Libby, MT. Whether Sample 72 is material from Libby. MT, or another location is unknown. This document is a draftfo r review purposes only and does not constitute A gency policy. D-13 DRAFT--DO NOT CITE OR QUOTE 1 Following analysis of Syrian hamsters intrapleurally injected with 10 or 25 mg of each of 2 the four samples of tremolite, Smith (19781 reported tumors at 350 days postexposure (25 mg) or 3 600 days postexposure (10 mg) for Samples 31 and 72 (see Table D-7). Although number of 4 animals was not provided by Smith et al. (1979). previous studies by these authors reported using 5 50 animals per exposure group (Smith. 1978; Smith and ITubert, 1974). The results in Table D-7 6 Present the cumulative number of tumors (numerator) at each time point analyzed over the 7 remaining survivors (denominator). The survival rate without tumor presentation was decreased 8 for animals exposed to Samples 72, 31, and 275. Smith et al. (1979) concluded that the FD-14 9 and 275 samples were noncarcinogenic, and Sample 31 was less carcinogenic than Sample 72. 10 Hamsters exposed to Sample 72 had extensive pleural fibrosis, which was observed to a lesser 11 degree in hamsters exposed to the other samples (Sample 72 > Sample 31 > Sample 12 275 = FD - 14). No statistical information was reported for these results, and because the 13 number of background tumors in control animals was not provided, no statistical analysis can be 14 performed. 15 ... . . H Both studies demonstrate that intrapleural injections of Libby Amphibole asbestos leads 16 to an increase in pleural fibrosis and mesothelioma in hamsters compared to controls or animals 17 injected with less fibrous materials. The use of doses of equal mass for both studies makes it 18 difficult to compare potency between samples, as each sample could have vastly different fiber 19 number and total surface area. Although these studies clearly show the carcinogenic potential of 20 Libby Amphibole asbestos fibers, intrapleural injections bypass the clearance and dissolution of 21 fibers from the lung after inhalation exposures. 22 Stanton et al. (1981) also examined tremolite and describe a series of studies on various 23 forms of asbestos. Fibers, embedded in hardened gelatin, were placed against the lung pleura. 24 As an intrapleural exposure, results might not be comparable to inhalation exposures, as the 25 dynami cs of fiber deposition and pulmonary clearance mechanisms are not accounted for in the 26 study design. Studies using two tremolite asbestos samples from the same lot were described as 27 being in the optimal size range for carcinogenesis; the fibers were distinctly smaller in diameter 28 than the tremolite fibers that Smith et al. (1979) used. These samples both had a high number of 29 fibers in the Stanton et al. (1981)-- size range (>8-pm long and <0.25-pm diameter). Exposure 30 to both tremolite samples led to mesotheliomas in 21 and 22 of 28 rats exposed. The Stanton et Assum ing Smith et al. (1979) used Libby Amphibole asbestos. This document is a draftfor review purposes only and does not constitute Agency policy. D-14 DRAFT--DO NOT CITE OR QUOTE 1 al. (1981) study also used talc that did not lead to mesothelioma production. This talc was found 2 to be the same as that used by Smith et al. (1979) and later by Wylie et al. (1993). Wylie et al. 3 (1993) stated that, although the two tremolites were consistent by size with commercial 4 amphibole asbestos, the talc used contained fibers that were much thinner and shorter, which is 5 not typical of prismatic tremolite fibers. 6 Wagner et al. (1982) examined three types of tremolite (California talc, Greenland, and 7 Korea) using SPF Sprague-Dawley (// = 48) and Wistar (n = 32) rats, then followed up with a 8 range of in vitro tests using the same fiber samples. Rats were injected intrapleurally 9 (20-mg tremolite) at 8-10 weeks of age and allowed to live out their lives. Median survival 10 times after injections were 644 days (California talc), 549 days (Greenland tremolite), and 11 557 days (Korean tremolite). Positive controls had a decreased survival time due to an infection, 12 which limits the interpretation of these data. Also, this study was performed separately using 13 different rat strains for the three tremolite samples. The authors state that, although the 14 decreased control survival time and use of different rat strains limit the usefulness of the study 15 for quantitative analysis, the results can be described qualitatively. Of the three tremolites, only 16 the Korean tremolite showed carcinogenic activity producing mesothelioma (14/47 rats, 30%). 17 Analysis of the fiber characteristics showed the Korean sample had fibers that were longer than 8 18 pm and a diameter of less than 1.5 pm. The California talc and Greenland tremolite had 19 little-to-no fibers in this size range (see Table D-8). Follow-up in vitro assays in the sample 20 publication (Wagner et al.. 1982) confirmed the in vivo results, with the exposure to Korean 21 tremolite resulting in increased LDH and [3-glucuronidase (BGL) release, cytotoxicity, and 22 giant-cell stimulation. 23 Davis et al. (1991) examined six tremolites with differing morphologies through 24 intraperitoneal injections with male SPF Wistar rats. Four of the tremolites were from 25 Jamestown, California; Korea; Wales; and Italy; and two were from Scotland. Of these, the three 26 from California, Korea, and Wales were asbestiform, and the other three were fiber bundl es or 27 prismatic (see Table D-9). Rats were exposed (n = 33 or 36) with one intraperitoneal injection 28 with samples that were 10 mg/2 mL-sterile PBS. Animals were allowed to live out their full life 29 spans or until signs of debility or tumor formati on developed. Although exposure was perform ed 30 based on sample weight, each sample was analyzed to determine the number of expected fibers 31 per milligram and, therefore, per exposure. These samples also were characterized further by This document is a draftfo r review purposes only and does not constitute A gency policy. D-15 DRAFT--DO NOT CITE OR QUOTE 1 Table D-8. Fiber characteristics of three tremolite samples analyzed by in 2 vivo and in vitro methods (TEM measurements) 3 Sample Location Fiber type Length Total no. of No. of fibers >8-pm No. of nonfibrous fibers long (xlO3) Diameter particles (xlO4) (x lO 4) <1.5-pm diameter A California Flake-like <6 pm <0.8 pm 6.9 5.1 1.7 material B Greenland Medium-sized <3 pm <1.2 pm 20.7 4.8 0 fibrous material C Korea Fine-fiber >8 pm <1.5 pm 3.3 15.5 56.1 material 4 5 TEM = transmission electron microscopy. 6 7 Source: Wagner et al. (1982). 8 9 10 Table D-9. Fiber characteristics in a 10-mg dose (as numbers of fibers) 11 Sample No. of animals No. of fibers in No. of 1 mg of injected mesotheliomas dust (x 10s) No. of fibers >8-pm long, <0.25-pm diam eter3 (x 10s) No. of particles in 1-mg injected dust ' (x 10s) Morphology California 36 36 13,430 121 18,375 Asbestiform W ales 36 35 2,104 8 4,292 Asbestiform Korea 33 32 7,791 48 13,435 Asbestiform Italy 36 24 1,293 1 20,137 Fiber bundles Carr Brae 33 4 899 0 9,490 Fiber bundles Shininess 36 2 383 0 5,901 Prismatic 12 13 aStanton fibers. 14 15 Source: Davis et al. (1991). 16 17 18 counting fibers versus particles. Data were collected for all fibers (aspect ratio >3:1) and 19 particles (aspect ratio <3:1) of total fibers. A fiber was defined as any component >8-pm long 20 and <0.25-pm diameter as measured by SEM (i.e., Stanton fibers). 21 The authors' overall conclusions were that all materials studied could cause 22 mesothelioma by this method of exposure, and the number of Stanton fibers was not sufficient to 23 explain the differences in response. Mesothelioma incidence was not correlated to Stanton 24 fibers, total particles, or mass of dust. The best predictor of mesothelioma incidence was total This document is a draftfor review purposes only and does not constitute Agency policy. D-16 DRAFT--DO NOT CITE OR QUOTE 1 fibers (see Table D-9). Although three samples were considered asbestiform (California, 2 Swansea, Korea), all samples had <1% of counted fibers defined as Stanton fibers. The highest 3 mesothelioma incidence was observed for the California sample, which contained the most 4 Stanton fibers (121 fibers per mg dust). The tremolite from Swansea, resulted in 97% 5 mesothelioma incidence yet contained only eight Stanton fibers per milligram (more than 90% 6 less than in the California sample). In contrast, the Italy tremolite, although containing only 7 0.08% Stanton fibers, resulted in 67% mesothelioma incidence. Little is known, however, about 8 the characteristics of particles or fibers <5-pm long. This study highlights two issues associated 9 with all fiber studies: the limits of analytical techniques and the variability in response based on 10 the metric used to measure exposure. This study also supports the premise that asbestos samples 11 containing fibers that are not long and thin can be carcinogenic. 12 The Roller et al. (1996) study was designed to provide data on the dose response of 13 various fiber types in relation to their fiber dimensions (as measured by SEM). Fibers were 14 defined in this study as having an aspect ratio of >5:1 for all lengths and widths. Female Wistar 15 rats (n = 40) were given either one intraperitoneal injection of 3.3 mg or 15 mg of tremolite 16 Rats were examined for tumors in the abdominal cavity following a lifetime (up to 30 months) of 17 observation. This paper described the fiber dimensions in depth (see Table D-10), while limited 18 discussion is focused on the exposure results. This table shows the characteristics of the fibers 19 sorted first by aspect ratio and diameter, and the fiber size distribution binned by the length and 20 diameter for those fibers with a length >5 pm. Results were described in this study in a table as 21 "positive rats" being those with histologically confirmed mesothelioma or macroscopically 22 supposed mesothelioma. No information was provided on how these determinations were made. 23 Exposure to 3.3-mg and 15-mg tremolite resulted in 9 mesotheliomas in 29 animals (64 weeks 24 postexposure) and 30 mesotheliomas in 37 animals (42 weeks postexposure), respectively. This 25 study demonstrates that intraperitoneal injection of tremolite led to mesothelioma in Wistar rats. 26 Analysis of other tissues was not described. 27 28 This document is a draftfor review purposes only and does not constitute Agency policy. D-17 DRAFT--DO NOT CITE OR QUOTE 1 Table D-10. Characteristics of tremolite fibers intraperitoneally injected into 2 Wistar rats 3 Fiber number per ng dust and mass fraction (%) Aspect Ratio (L/D) >5/1; D <2 pm (Roller study) Aspect Ratio (L/D) <3/1; D <3 pm (WHO, 1985) Length: >5 pm >10 pm >20 pm Diameter: >5 pm >10 pm >20 pm % % % No. Mass No. Mass No. Mass 17.4 32 6.9 27 1.9 18 % % % No. Mass No. Mass No. Mass 18.4 43 7.0 35 2.0 26 Fiber-size distribution for aspect ratio (L/D) >3/1 (all lengths, all diameters; SEM) % Total fibers L >5 pm 10% < Length (pm) 50% < 90% < 99% < 10% < Diameter (pm) 50% < 90% < 99% < 22% 0.8 2.4 9.2 29.4 0.14 0.27 0.67 1.49 4 5 SEM = scanning transmission microscopy. 6 7 Source: Roller et al. (1996). 8 9 10 D.2. MECHANISTIC DATA AND OTHER STUDIES IN SUPPORT OF THE MODE OF 11 ACTION 12 D.2.1. In Vitro Studies--Libby Amphibole Asbestos 13 Hamilton et al. (2004) examined the potential for fibers, including Libby Amphibole 14 asbestos, to modify the functi on of antigen-presenting cells (APC). Analysis was performed at 15 24 hours with two forms of asbestos (crocidolite [25 or 50 pg/mL] and Libby Amphibole 16 asbestos obtained from Site No. 30, Libby, MX [25 or 50 pg/mL]) and ultrafine particulate 17 matter (PM2.5 [particulate matter 2.5 microns diameter or less] [50 or 100 pg/mL]). Limited 18 information is provided by Hamilton et al. (2004) on fiber characteristics. Samples from Site 19 No. 30, however, are described as predominantly richterite and winchite by Meeker et al. (2003). 20 Primary human alveolar macrophages were incubated for 24 hours with Libby Amphibole 21 asbestos (25 or 50 pg/mL), crocidolite (25 or 50 pg/mL), or ultrafine particulate matter (50 or 22 100 pg/mL). Following incubation, cells were isolated from remaining particles and nonviable 23 cells, after which 0.25 x 10 macrophages were cocultured with autologous lymphocytes 24 (1 x 106 cells) in an 11-day APC assay. This assay analyzes the antigen-presenting function of 25 the pretreated macrophages by stimulating the lymphocytes using tetanus toxoid as the antigen. 26 The supernatant was assayed for cytokines on Day 11, and Hamilton et al. (2004) found that This document is a draftfor review purposes only and does not constitute Agency policy. D-18 DRAFT--DO NOT CITE OR QUOTE 1 p r e tr e a tm e n t w ith e ith e r a s b e s to s o r P M 2 .5 s ig n if ic a n tly u p r e g u la te d b o th T Hi a n d T H 2 c y to k in e s 2 (in te rf e ro n g a m m a [IF N y ]; in te rle u k in -4 [IL -4 ]; a n d in te r le u k in - 13 [ I L - 13]) (p < 0 .0 5 ). 3 T herefore, pre-exposure to either fibers or particles increased A PC function, as reflected in 4 increased cytokine release after tetanus challenge. N o significant differences, how ever, w ere 5 d iscem ab le b etw een asbestos and P M 2.5 pretreatm ent. T he authors speculated th at th e v ariability 6 in response betw een sam ples assayed-- presum ably due to the use o f prim ary cells-- obscures 7 statistical significance. A lthough this study su pports a ro le fo r fibers and P M 2.5 in potentiating 8 im m une response, the im plications o f these findings to hum an health are unclear because m any 9 agents can activate m acrophages prior to antigen challenge. 10 R ecen t studies (B lake et al., 2 0 0 8 ; B lake et al., 2 0 0 7 ) com pared the resp o n se o f m u rin e 11 m a c r o p h a g e s ( p r im a r y a n d c e ll lin e R A W 2 6 4 .7 ) to L ib b y A m p h ib o le a s b e s to s fib e r s a n d 12 c r o c id o lite a s b e s to s fib e r s . T h e L ib b y A m p h ib o le a s b e s to s fib e r s (7 .2 J 7 .0 1 -p m lo n g , 0 .6 1 13 1 .2 2 -p m d ia m e te r) u s e d in th e s e stu d ie s w e re o b ta in e d fro m th e U .S . G e o lo g ic a l S u rv e y an d 14 w ere ch em ically rep resen tativ e o f the L ibby, M T m in e (M eek er et al.. 2 0 0 3 ). T he crocidolite 15 fib e rs (4 .5 9 4 .2 2 p m -lo n g , 0 .1 6 0 .0 9 p m -d ia m e te r) u se d in th e s e stu d ie s w e re p ro v id e d b y 16 R esearch T riangle Institute, N C , and th e n o n cy to to x ic control fib er (w ollastonite, 17 4.4 6 5.53 p m -long, 0.75 1.02 p m -d iam eter) w as p ro v id ed b y N Y C O M in erals, N Y . C ells 18 w e re e x p o se d fo r 2 4 h o u rs to fib e r s a m p le s m e a s u re d b y re la tiv e m a s s (5 p g /c m ), a fte r w h ic h 19 th e ce lls w e re an aly zed b y tra n sm issio n elec tro n m ic ro sc o p y (T E M ) to m easu re in tern alizatio n . 20 T h e r e s u lts o f th e fir s t s tu d y ( B la k e e t al., 2 0 0 7 ) in d ic a te th a t L ib b y A m p h ib o le a s b e s to s fib e rs 21 c a n b o th a tta c h to th e p la s m a m e m b r a n e a n d b e in te r n a liz e d b y m a c r o p h a g e s , s im ila r to th e 22 c r o c id o lite fib e rs . T h e s e in te r n a liz e d fib e r s w e r e p r im a r ily le s s th a n 2 - p m lo n g a n d w e r e f o u n d 23 localized in the cytoplasm , in cytoplasm ic vacuoles, and near the nucleus follow ing 3-hour 24 exposure, 62.5 pg/cm . T his sam e co ncentration (62.5 pg/cm ) w as selected fo r the rem aining 25 studies b ecau se cell v iab ility w as n o t d ecreased at th is co n cen tratio n fo r th e L ib b y A m p h ib o le 26 asbestos (92% ); cell viability w as decreased fo r crocidolite (62% ), how ever, at this 27 concentration. A s a result, the rem aining assays w ould be expected to have decreased viability 28 follow ing exposure to crocidolite, w hich m ay im pact the levels o f various responses. F or 29 exam ple, th e reactiv e o x y g en sp ecies (R O S ) m easu rem en t w o u ld in crease w ith in creased cell 30 num ber; therefore, som e o f the quantitative results w ould be difficult to com pare betw een fiber 31 ty p es u n less n o rm a liz e d to cell num ber. This document is a draftfo r review purposes only and does not constitute A gency policy. D -19 D R A FT -- D O N O T C ITE O R Q U O TE 1 Oxidative stress was measured by the induction of ROS and the reduction in glutathione 2 (GSH) levels. These two measurements generally complement each other, as GSH is used in 3 cells to maintain intracellular redox balance in cells in response to increased ROS levels. Both 4 Libby Amphibole asbestos and crocidolite fiber internalization generated a significant increase 5 (p < 0.05) in intracellular ROS as quantified by the oxidation of 2,7-dichlorodihydrofluorescein 6 to dichlorofluorescein with hourly readings on a fluorescent plate reader. Libby Amphibole 7 asbestos exposure significantly increased ROS in a dose-dependent manner (6.25, 32.5, and 8 62.5 pg/cm ), as early as 1 hour postexposure at the highest dose (p < 0.05), as compared to a 9 no-treatment group. Only the highest concentration of crocidolite was tested. The lower 10 concentrations of Libby Amphibole asbestos were not compared to crocidolite and wollastonite, 11 but a comparison of the highest exposure concentrations (62.5 ug/cm ) of Libby Amphibole 12 asbestos, crocidolite, and wollastonite revealed greater ROS production following Libby 13 Amphibole asbestos exposure (1 hour,/; < 0.05). Blake et al. (2007) stated that similar results 14 were seen in the primary cell line but did not report the data. To differentiate the type of ROS 15 produced, dehydroergosterol fluorescence intensity levels were used, revealing that superoxide 16 anion was significantly increased following exposure to Libby Amphibole asbestos as compared 17 to controls. This observation was further confirmed with use of a free radical scavenger 18 (PEG-SOD [polyethylene glycol-superoxide dismutase]) specific to superoxide anion. This 19 coexposure of Libby Amphibole asbestos and PEG-SOD led to a significant decrease in ROS as 20 compared to cells exposed only to Libby Amphibole asbestos (p < 0.05). Total intracellular 21 superoxide dismutase (SOD) activity also was measured following exposure to Libby Amphibole 22 asbestos and showed a decrease in activity at 3 hours postexposure as compared to controls 23 (p < 0.05). Crocidolite appears to increase intracellular SOD activity at 24 hours postexposure. 24 These three assays demonstrate that Libby Amphibole asbestos exposure leads to increased 25 superoxi de anion in macrophages, most likely by suppressing activity of intracellul ar SOD. 26 GSH levels were found to be decreased in response to Libby Amphibole asbestos and 27 crocidolite exposure in the macrophage cell line as compared to unexposed cells (p < 0.05). The 28 decreased GSH levels were more prominent following crocidolite exposure as compared to 29 Libby Amphibole asbestos. Crocidolite exposure has been shown in other studies to lead to 30 increased hydrogen peroxide but not superoxide anion (Kamp and Weitzman. 1999; Kamp el al , 31 1992). The increased hydrogen peroxide from crocidolite exposure can then lead to increased This document is a draftfo r review purposes only and does not constitute A gency policy. D-20 DRAFT--DO NOT CITE OR QUOTE 1 hydroxyl radical production (through interactions with endogenous iron), and potentially, 2 deoxyribonucleic acid (DNA) adduct formation. DNA adduct formation 3 (8-hydroxy-2'deoxyguanosine, 8-OHdG), 8-oxoguanine-DNA-glycosylase 1 (Oggl) levels, and 4 DNA damage (comet assay) also were measured. A significant increase in DNA damage in 5 exposed macrophages, as measured by increases in both 8-OHdG formation and expression of 6 Oggl, a DNA repair enzyme that excises 8-OHdG from DNA following oxidative stress, was 7 observed following exposure to crocidolite but not Libby Amphibole asbestos. Increased 8 superoxide anion following Libby Amphibole asbestos exposure does not appear to yield 9 oxidative damage similar to crocidolite. These results suggest a chemical-specific response to 10 each type of amphibole that yields varied cellular responses. Therefore, the mechanism of action 11 following response to Libby Amphibole asbestos might be different than that of crocidolite, also 12 an amphibole fiber. 13 To determine if the ROS production was related to fiber number for both Libby 14 Amphibole asbestos and crocidolite, cell-fiber interactions and fiber internalization were 15 measured following exposure to equal concentrations of crocidolite, Libby Amphibole asbestos, 16 and wollastonite (62.5 pg/cm , 3 hours). With phase contrast light microscopy, the number of 17 cells interacting with one or more fibers were counted (100 cells counted for each treatment). 18 All murine macrophages bound or internalized at least one fiber from the Libby Amphibole 19 asbestos sample (mean SD, 4.38 1.06 internalized) or the crocidolite sample (3.28 20 1.58 internalized) but not the wollastonite sample (Blake et aL 2007). No significant differences 21 were observed in the responses to Libby Amphibole asbestos or crocidolite samples, suggesting 22 that the differences in measured ROS were not related to cell number. Fiber sizes varied 23 between the two samples, with the crocidolite sample containing a more homogeneous mixture 24 of long fibers (exact size not given), while the Libby Amphibole asbestos sample contained a 25 mixture of sizes and widths. These characteristics were not analyzed to determine what, if any, 26 role they might play in the varied response. 27 The second study by Blake et al. (2008) reports the effects of in vitro exposure to Libby 28 Amphibole asbestos on apoptosis by exploring autoimmune response following asbestos 29 exposure. Although Libby Amphibole asbestos was not directly used in the autoimmune studies, 30 the autoantibody (SSA/Ro52) is a known marker of apoptosis, and the in vitro studies included 31 treatment with Libby Amphibole asbestos. RAW264.7 cells exposed to Libby Amphibole This document is a draftfo r review purposes only and does not constitute A gency policy. D-21 DRAFT--DO NOT CITE OR QUOTE 1 asbestos induced apoptosis over 72 hours, as measured by induction of poly (ADP-ribose) 2 polymerase cleavage and increased Annexin V staining. Redistribution of SSA/Ro52 in 3 apoptotic blebs was demonstrated in Libby Amphibole asbestos-exposed RAW264.7 cells but 4 not in the unexposed controls and wollastonite-exposed RAW264.7 murine macrophages, further 5 confirming apoptosis. 6 The role of ROS in chromosomal damage from asbestos was examined in a recent study 7 of Libby Amphibole asbestos and Union for International Cancer Control (UICC) crocidolite in 8 XRCC1-deficient human lung epithelial H460 cells (Pietruska et al.. 20101. XRCC1 is involved 9 in the repair mechanisms for oxidative DNA damage, particularly single-strand breaks. This 10 study examined the effect of XRCC1 deficiency (induced in cells by shRNA knockdown) 11 following exposure to genotoxic (crocidolite and Libby Amphibole asbestos) and nongenotoxic 12 compounds (wollastonite, titanium dioxide) on micronucleus formation. Cells were exposed to 13 chemicals with known oxidants hydrogen peroxide (0-60 pM) or bleomycin (0-10 pg/ml) for 1 14 and 3 hours, or the nonoxidant paclitaxel (0-5 nM, 24 hours) to confirm the clonogenic survival 15 of the knockout cells, and as positive and negative controls. Fiber-size distribution for 16 crocidolite and Libby Amphibole asbestos is shown in Table D -lL Micronuclei induction was 17 measured following treatment of cells by controls as described above, and by 5-pg/cm fibers or 18 TiCL particles for 24 hours. Following treatment, cells were fixed, permeabilized, and blocked 19 before being exposed to anticentromere antibodies, and micronuclei were counted and scored as 20 centromere negative arising from DNA breaks (clastogenic) or centromere positive arising from 21 chromosomal loss (aneugenic). Spontaneous micronuclei induction was increased in 22 XRCC1-deficient cells as compared to control. Wollastonite and titanium dioxide did not induce 23 micronuclei in either cell type. Crocidolite and Libby Amphibole asbestos induced 24 dose-dependent increases in micronuclei formation in both cell types including an increase in the 25 proportion of micronuclei in XRCC1-deficient cells (see Table D-12). Libby Amphibole 26 asbestos exposure led to a decreased amount of micronuclei as compared to crocidolite. 27 Specifically in relation to clastogenic versus aneugenic micronuclei, crocidolite exposure led to 28 mainly clastogenic micronuclei while Libby Amphibole asbestos exposure led to a mixture of 29 aneugenic and clastogenic micronuclei . Nuclear bud formation was also observed but only with This document is a draftfor review purposes only and does not constitute Agency policy. D-22 DRAFT--DO NOT CITE OR QUOTE 1 Table D -ll. Size distribution of UICC crocidolite and Libby Amphibole 2 asbestos used in Pietruska et al. (2010)a 3 % fibers in size range Length (pm) Crocidolite Libby Amphibole Asbestos 0.1 1.0 46.4 12.6 1.1-5.0 44.8 38.5 5.1-8.0 3.8 23.1 8.1-10.0 0.9 10.4 10.1-20.0 2.4 11.6 >20.1 1.7 3.6 5 distribution by diameter also given in original manuscript. 6 Source: Adapted from Supplemental Material o f Pietruska et al. (2010). 7 ' ' 8 9 Table D-12. Percent clastogenic micronuclei following exposure to Libby 10 Amphibole asbestos or crocidolite 11 H460 cells X RCC1-deficient Libby Amphibole asbestos (5 Lig/cnf) 71.5 3.4% 86.0 1.2%* Crocidolite (5 pg/cnf) 57.2 2.2% 65.1 2.2% a 12 13 "p < 0.05 as compared to control cells. 14 15 Source: Pietruska et al. (2010). 16 17 18 exposure to crocidolite and bleomycin. Western blot analysis was performed to analyze protein 19 expression related to DNA damage repair (XRCC1) and cell cycle progression (p53, p21) (data 20 not shown in publication). The differences observed between crocidolite and Libby Amphibole 21 asbestos are most likely related to their physicochemical differences, particularly related to their 22 iron content. However, these results support a genotoxic effect of exposure to both crocidolite 23 and Libby Amphibole asbestos. 24 Mechanisms of oxidative stress following exposure to Libby Amphibole asbestos were 25 also studied in human mesothelial cells (Llillegass et aL 2010). Gene-expression changes were 26 measured with Affymetrix U133A microarrays (analysis with GeneSifter) following exposure to This document is a draftfor review purposes only and does not constitute Agency policy. D-23 DRAFT--DO NOT CITE OR QUOTE 1 15x106-pm2/cm2 Libby Amphibole asbestos8 as compared to the nonpathogenic control 2 (75 x IO -pm'/cm' glass beads) in the human mesothelial cell line LP9/TERT-1 for 8 and 3 24 hours. Gene expression of only one gene (manganese superoxide dismutase [MnSOD; 4 SOD2]) was altered following exposure to Libby Amphibole asbestos for 8 hours, while 5 111 genes had an altered gene expression following exposure to Libby Amphibole asbestos for 6 24 hours (altered by at least twofold as compared to control). 7 The gene for MnSOD; SOD2 was observed to be significantly upregulated at both time 8 points (p < 0.05) as compared to nonpathogenic controls. This gene was confirmed in normal 9 human pleural mesothelial cells (HKNM-2) by quantitative RT-PCR at 24 hours following 10 exposure to the nontoxic dose of Libby Amphibole asbestos. Upregulation of three genes from 11 this and previous studies by these authors was confirmed by quantitative RT-PCR (SOD2, A Tl\ 12 and IL8) in HKNM-2 cells exposed to both Libby Amphibole and crocidolite asbestos. Gene 13 ontology of these results demonstrated alterations related to signal transduction, immune 14 response, apoptosis, cellular proliferation, extracellular matrix, cell adhesion and motility, and in 15 only one gene related to ROS processing. Follow-up studies at both the nontoxic dose 16 (15 x lo6 pnr/cnr) and the toxic dose (75 x 106 um VcnL) exposure levels in LP9/TERT-1 cells 17 examined SOD protein and activity, ROS production, and glutathione (GSH) levels. At 18 24 hours, SOD2 protein levels were increased following exposure to the toxic dose of Libby 19 Amphibole asbestos (p < 0.05) but not at 8 hours. Cells exposed to all doses of Libby 20 Amphibole and crocidolite asbestos had increased copper-zinc superoxide di smutase 21 (Cu/ZnSOD; SOD1) protein at 24 hours (p < 0.05) but not at 8 hours. Although total SOD 22 activity remained unchanged, a dose-related SOD2 activity was observed following exposure to 23 both doses of Libby Amphibole asbestos for 24 hours, but this appeared to be minimal and was 24 not statistically significant (8 hours was not examined). Oxidative stress was measured by 25 dichiorodi hydrofluorescein diacetate fluorescence staining detected by flow cytometry and was 26 observed as both dose- and time-dependent in cells exposed to Libby Amphibole asbestos but 27 was increased following exposure to the toxic dose of Libby Amphibole asbestos (statistical 28 analysis not possible). Oxidative stress was further supported by analysis of gene expression of 8Libby Amphibole asbestos samples for this study were characterized by analysis o f chemical composition and mean surface area (Meeker et al.. 2003). Doses were measured in surface area and described based on viability assays as either nontoxic (15 106 p n f/c n f) or toxic (75 * 106 p n f/c n f). This document is a draftfo r review purposes only and does not constitute A gency policy. D-24 DRAFT--DO NOT CITE OR QUOTE 1 heme oxygenase 1 (HO-1) following exposure to Libby Amphibole asbestos in both 2 LP9/TERT-1 and HKNM-2 cells for 8 and 24 hours. HO-1 was significantly increased 3 following exposure to the toxic dose of Libby Amphibole asbestos in both cell lines (p-value not 4 given). GSH levels were transiently depleted following 2-8 hours exposure to 5 75 x 106-pm2/cm2-levels of Libby Amphibole asbestos, with a gradual recovery up to 48 hours in 6 LP9/TERT-1 cells (HKNM-2 not analyzed). Exposure to crocidolite asbestos at the toxic dose 7 led to a significant GSH decrease at all times points up to 24 hours (p < 0.05). These studies 8 demonstrate that Libby Amphibole asbestos exposure leads to increases in oxidative stress as 9 measured by ROS production, gene expression, protein and functional changes in oxidative 10 stress proteins (SOD), and GSEI-level alterations in human mesothelial cells. 11 The relative toxicity of Libby Amphibole asbestos was measured by gene-expression 12 changes of interleukin-8 (IL-8), cyclooxygenase-2 {COX-2), heme oxygenase {HO)-l as well as 13 other stress-responsive genes as compared to amosite (Research Triangle Institute, NC) in 14 primary human airway epithelial cells (HAEC) in vitro. Comparisons were made with both 15 fractionated (aerodynamic diameter <2.5 pm) and unfractionated fiber samples (Duncan et aL 16 2010). Crocidolite fibers (UICC) were also included in some portions of this study for 17 comparison. Fractionation was performed using the water elutriation method (Webber et al., 18 2008) and characterized as described in Lowers and Bern (2009). Primary HAECs were 19 exposed to 0, 2.64, 13.2, and 26.4 pg/em of crocidolite, amosite, AM2.5 (fractionated), Libby 20 Amphibole asbestos, or LA2.5 (fractionated) for 2 or 24 hours in cell culture. Confocal 21 microscopy was used to determine fiber content in cells exposed for 4 or 24 hours to 22 26.4-pg/cm AM2.5 or LA2.5 only. At 4 hours post exposure, fibers were mainly localized on 23 the periphery of the cell with some fibers internalized. By 24 hours post exposure, most fibers 24 appeared to be internalized and localized by the nucleus. Cytotoxicity was determined by 25 measurement of lactate dehydrogenase (LDH) from the maximum dose (26.4 pg/cm ) of both 26 amosite and Libby Amphibole asbestos samples, with less than 10% LDH present following 27 exposure to all four samples. Cytotoxicity was also determined for just the fractionated samples 28 of amosite and Libby Amphibole asbestos by measuring intracellular calcein fluorescence 29 emitted by live cells and showed 95% and 99% viability for AM2.5 and LA2.5, respectively. 30 These results support a limited cytotoxicity of both amosite and Libby Amphibole asbestos under 31 these concentrations and time frames. This document is a draftfo r review purposes only and does not constitute A gency policy. D-25 DRAFT--DO NOT CITE OR QUOTE 1 Gene-expression changes in specific inflammatory markers (IL-8, COX-2, HO-1) were 2 analyzed by quantitative RT-PCR for amosite, AM2.5, Libby Amphibole asbestos, LA2.5, and 3 CRO at both 2 and 24 hours post exposure (all doses). Minimal increases in gene expression of 4 IL-8, COX-2, or HO-1 were observed at 2 hours post exposure to all five fiber types; at 24 hours 5 post exposure, however, a dose response was observed following exposure to all fiber types. The 6 smaller size fractions resulted in differences in magnitude of gene-expression changes between 7 AM2.5 and LA2.5, with AM2.5 leading to greater induction of IL-8 and COX-2 as compared to 8 LA2.5. HO-1 levels were comparable between the two samples (see Table D-13). Gene 9 expression of transforming growth factor (TGF)-Bl was also quantified but only following 10 exposure to AM2.5 and LA2.5 (all doses; data not shown in publication). Levels of IL-8 protein 11 were also measured following 24 hours exposure to AM2.5 and LA2.5 (all doses) and were 12 statistically significant at the two highest exposures (13.2 and 26.4 ug/cm ). Gene-expression 13 changes were also examined for 84 genes involved in cellular stress and toxicity using a 96-well 14 RT-PCR array format following 24 hours exposure to 13,2-ug/cm amosite, Libby Amphibole 15 asbestos, AM2.5, or LA2.5 or to 26.4-pg/cm LA2.5 only. The results show a pro-inflammatory 16 gene-expression response. Gene-expression profiles were similar between amosite and Libby 17 Amphibole asbestos, but differences were observed between AM2.5 and LA2.5. 18 19 20 Table D-13. Gene-expression changes following exposure to 26.4-pg/cm2 21 amphibole asbestos for 24 hours51 22 Genes for specific inflam m atory markers Am osite (AM) A m osite, fractionated (AM 2.5) Libby A m p h ib ole Asbestos Libby A m p h ib ole Asbestos, fractionated (LA2.5) IL-8 50 7.5 120 25 46 8.3 37 7 .8 COX-2 5.4 0 .5 16 2 .8 9.0 1.7 1.6 0 .3 HO-1 2.9 0 .2 4.5 0 .3 2.5 0 .2 5.1 0 .6 23 24 aAll results in fold change as compared to untreated control cells. 25 26 Source: Duncan et al. (2010). 27 28 This document is a draftfor review purposes only and does not constitute Agency policy. D-26 DRAFT--DO NOT CITE OR QUOTE 1 To determine if surface iron on the fibers played a role in the inflammatory response, 2 Duncan et al. (2010) also examined surface iron concentrations by two methodologies: 3 inductively coupled plasma optical emission spectroscopy and citrate-bicarbonate-dithionite . 4 Both assays determined AM2.5 appeared to have the measured by thiobarbituric acid -reactive 5 product formation following exposure to amosite, AM2.5, Libby Amphibole asbestos, and 6 LA2.5. Both amosite samples were found to generate the greatest amount of hydroxyl radicals 7 compared to the two Libby Amphibole asbestos samples, with the fractionated AM2.5 and LA2.5 8 exhibiting small increases in ROS produced compared to the unfractionated samples. 9 10 D.2.2. In Vitro Studies--Tremolite 11 In general, all fibrous tremolite samples were shown to be carcinogenic, with those 12 containing more of the longer, thinner fibers (>10-pm-length, <l-pm-diameter) being more 13 potent carcinogens. Most studies described here used weight as the measurement of fibers for 14 exposure, with the doses ranging from 0 to 40 mg/animal. One set of studies did expose animals 15 with fibers measured by number (100 fibers/cm ) (Bernstein et aL 2006: Bernstein et aL 2005). 16 17 D.2.2.1. Cytotoxicity 18 Wagner et al. (1982) examined the in vitro cytotoxicity of three forms of tremolite (see 19 Table D-8) used in their in vivo studies. LDH and BGL were measured in the medium following 20 incubation of unactivated primary murine macrophages to 50, 100, and 150 pg/mL of each 21 sample for 18 hours. Cytotoxicity of Chinese ham ster lung fibroblasts V79-4 was measured by 22 methylene blue staining (fiber concentrations not given). Giant-cell formation in A549 human 23 basal alveolar epithelial cell cultures was measured, using 100 and 200 pg/mL of each sample for 24 5 days. Crocidolite fibers were used as the positive control. 25 In all three assay systems, the Korean tremolite produced results similar to the positive 26 control: increased toxicity of primary murine macrophages, increased cytoxicity of Chinese 27 hamster ovary (CHO) cells, and increased formation of giant cells from the A549 cell line. The 28 tremolite sample from Greenland (Sample B) did result in increased toxicity over controls, 29 although to a lesser degree (statistics are not given). The authors speculate that the iron content 30 in Sample B might have contributed to these results. Although differential toxicity of these This document is a draftfor review purposes only and does not constitute Agency policy. D-27 DRAFT--DO NOT CITE OR QUOTE 1 samples was noted on a mass basis, data were not normalized for fiber content or size. The 2 inference is that differential results are due, at least in part, to differential fiber counts. 3 In a study to further elucidate the role of ROS following exposure to asbestos, Suzuki and 4 Hei (19961 examined the role of heme oxygenase (HO) in response to asbestos. HO is induced 5 in response to oxidative stress and functions to degrade heme; it might, therefore, prevent 6 iron-mediated hydroxyl radical production. All fibers tested led to an increase in HO, though 7 chrysotile (UICC) and crocidolite (UICC) led to a greater increase than tremolite (Metsovo, 8 Greece) and erionite (Rome, Oregon). No statistics, however, are described for these results. 9 This study focused on responses to 20 and 40 pg/mL of chrysotile and then used doses that 10 yielded 0.5 and 0.3 relative survival fractions for all other fibers (crocidolite, 20 and 40 pg/mL; 11 tremolite, 150 and 300 pg/mL; erionite, 200 and 400 pg/mL). Fibers were not characterized in 12 this paper. When normalized by survival fraction, the inductions of HO above control were 13 3.89-, 3.86-, 2.75-, and 2.78-fold above background for chrysotile, crocidolite, tremolite, and 14 erionite, respectively. Limited information is provided on the results of tremolite exposures 15 beyond an increase in HO following an 8-hour exposure. This increased FIO following exposure 16 to tremolite demonstrates a response similar to that observed for crocidolite and chrysotile in this 17 study. Crocidolite is further analyzed, with exposures to the antioxidants, superoxide dismutase 18 and catalase, leading to a dose-dependent decrease in HO induction, which supports the role of 19 HO in oxidative stress. 20 Wylie et al. (1997) examined the mineralogical features associated with cytotoxic and 21 proliferative effects of asbestos in hamster tracheal epithelial (HTE) and rat pleural mesothelial 22 (RPM) cells with a colony-forming efficiency assay. HTE cells are used because they give rise 23 to tracheobronchial carcinoma, while RPM cells give rise to mesotheliomas. Cells were exposed 24 to fibers by weight, number, and surface area (see Table D-14). 25 Colony-forming efficiency assay results are expressed as the number of colonies in 26 exposed cultures divided by the control colonies multiplied by 100. Increases in colony numbers 27 indicate increased cell proliferation or survival in response to the exposure. Decreases in colony 28 numbers indicate toxicity or growth inhibition in response to the exposure. The results of the 29 analysis with fiber exposure by mass (pg/cm ) show elevated colonies in HTE cells following 30 exposures to both asbestos fibers (p < 0.05) at the lowest concentrations, while significant This document is a draftfor review purposes only and does not constitute Agency policy. D-28 DRAFT--DO NOT CITE OR QUOTE 1 decreases were observed for both asbestos fibers at the higher concentrations (0.5 ug/cirC2, 2 p < 0.05) (Wvlie et al.. 1997V 3 4 5 Table D-14. Fiber characteristics of five fibers examined in vitro for 6 cytotoxic (HTE cells) and proliferative effects (RPM cells) 7 ` Sample Description (% of sample) Surface area (mm2/g) Fibers/pg Fibers >5 pm/pg FD14 Talc (37), tremolite (35), serpentine (15), other (<2), unknown (12) 6.2 0 .2 2.5 x 103 0.8 x 103 SI57 CPS183 Talc (60), tremolite (12), unknown (21), other (4), anthophyllite (3), quartz (1) Talc (50), quartz (12), unknown (28), tremolite (4), other (4), anthophyllite (3) 4.9 0.2 4.9 0.4 1.1 x JO4 4.8 x 103 1.1 x 104 9.2 x 103 NIEHS crocidolite Riebeckite (100) 10.3 : 1.3 5.3 * 105 3.8 10" NIEHS chrysotile Chrysotile (100) 25.4 0.5 5.3 x 104 3.4 x 104 8 9 NIEHS = National Institute o f Environmental Health Sciences. 10 11 Source: Wylie et al. (1997). 12 13 14 No proliferation was observed for either chrysotile or crocidolite asbestos fibers in RPM 15 cells, but cytotoxicity was observed at concentrations greater than 0.05 pg/cm (p < 0.05). All 16 talc samples were less cytotoxic in both cell types. Comparing results of these samples when 17 exposure is measured by fiber number, the same number of crocidolite asbestos fibers >5-pm 18 long leads to proliferation in HTE cells, but proliferation did not occur for FD14 fibers. The 19 other two talc samples showed both insignificant cytotoxicity (SI57) and significant cytotoxicity 20 (CPS 183, p < 0.05). Therefore, when measured by fiber number, the results show differential 21 responses for the fibers analyzed, suggesting the mineralogy of the fibers is more important in 22 determining the biological response to fibers. In the RPM cells, however, similar responses were 23 seen for all fibers analyzed, except for the slight cytotoxicity of I D14 at 2.6 fibers/cm . This 24 suggests that fiber number does play a role in biological response in this cell type. 25 Data analysis by surface area of these samples is shown in Table D-14. The results of 26 these samples in both cell lines demonstrated that the cellular responses seemed unrelated to the 27 surface area, which demonstrates the impact of the dose metric on data. Analyzing the data for This document is a draftfor review purposes only and does not constitute Agency policy. D-29 DRAFT--DO NOT CITE OR QUOTE 1 cytotoxicity and proliferation based on the exposure measurement demonstrated differences in 2 response depending solely on how the fibers were measured (e.g., by mass, number, or surface 3 area). These results show variability in interpreting the same assay based on the defined unit of 4 exposure. Most early studies used mass as the measurement for exposure, which can impact how 5 the results are interpreted. When possible, further analysis of fiber number and surface area 6 might help elucidate the role of these metrics, particularly for in vivo studies. 7 8 D.2.2.2. Genotoxicity 9 (Athanasiou et al.,, 1992) performed a series of experiments to measure genotoxicity 10 following exposure to tremolite, including the Ames mutagenicity assay, micronuclei induction, 11 chromosomal aberrations, and gap-junction intercellular communication. Although a useful test 12 system for mutagenicity screening for many agents, the Ames assay is not the most effective test 13 to detect mutations induced by mineral fibers. Mineral fibers can cause mutation through 14 generation of ROS or direct disruption of the spindle apparatus during chromatid segregation. 15 Fibers do not induce ROS in the Ames system; however, and the Salmonella typhimurium strains 16 do not endocytose the fibers. Only one study was found in the published literature that used the 17 Ames assay to measure mutagenicity of tremolite. Metsovo tremolite asbestos has been shown 18 to be the causative agent of endemic pleural calcification and an increased level of mali gnant 19 pleural mesothelioma (see Section 4.1). To measure the mutagenicity of Metsovo tremolite, 20 S. typhimurium strains (TA98, TA100, and TA102) were exposed to 0-500 pg/plate of asbestos 21 (Athanasiou et al., 1992). This assay demonstrated that, like most asbestos fiber types tested in 22 earlier studies, Metsovo tremolite did not yield a significant increase in revertants in the Ames 23 assay, including in the TA102 Salmonella strain, which is generally sensitive to oxidative 24 damage. Although these strains can detect ROS mutations, they would not be able to produce 25 ROS from fibers alone or through necessary signaling pathways, and they do not endocytose 26 fibers. Thus, negative results in the Ames assay do not inform the cytotoxicity of Metsovo 27 tremolite. 28 Furthermore, this study demonstrated the clastogenic effects of tremolite, including 29 chromosomal aberrations and micronuclei induction. Tremolite exposure (0-3.0 gg/cm ) in 30 Syrian hamster embryo (SHE) cells resulted in a statistically significant increase in chromosomal 31 aberrations (p < 0.02) when all treatment groups were combined and then compared to controls; This document is a draftfo r review purposes only and does not constitute A gency policy. D-30 DRAFT--DO NOT CITE OR QUOTE 1 however, no clear dose-response relationship was evident (Athanasiou et al., 1992). Tremolite 2 exposure in SHE cells did lead to a dose-dependent increase in chromosome aberrations that was 3 statistically significant at the highest doses tested (1.0-3.0 pg/cm ) (p < 0.01) (see Table D-15). 4 5 6 Table D-15. Micronuclei induction (BPNi cells) and chromosomal 7 aberrations (SHE cells) following exposure to tremolite for 24 hours 8 Asbestos dose (pg/cm2) Micronuclei incidence/1,000 cells Chromosomal aberrations (including chromatid gaps, breaks, isochromatid breaks, and chromosome type) 0 17 3 0.5 31a 4 1.0 70b 12c 2.0 205b 9a 3.0 Not tested 13c 9 10 aSignificantly different from control (p < 0.05). 11 bSignilrcantly different front control (p < 0.01). 12 Significantly different from control (p < 0.02). 13 14 Source: Athanasiou et al. (1992). 15 16 17 Micronuclei induction was measured in BPNi cells after 24-hour exposure to 18 0-2.0-pg/cm tremolite. A statistically significant dose-dependent increase in levels of 19 micronuclei was demonstrated following tremolite exposure at concentrations as low as 20 0.5 pg/cm (p < 0.01). Literatures searches did not find tremolite tested for clastogenicity in 21 other cell types, but the results of this study suggest interference with the spindle apparatus by 22 these fibers. No analysis was performed to determine if fiber interference of the spindle 23 apparatus could be observed, which would have supported these results. 24 To determine if tremolite has some tumor promoter characteristics, Athanasiou et al. 25 (1992) further exami. ned i. ntercellular communi. catio. n fo.llowing exposure to 0-4.0-pg/cm9 26 tremolite in both Chinese hamster lung fibroblasts (V79) and SHE BPNi cells, which are 27 sensitive to transformation. Inhibition of gap-junctional intercellular communication has been 28 proposed to detect tumor-promoting activity of carcinogens (Trosko et al.. 1982). No effect on 29 gap-junction intercellular communication following tremolite exposure was observed. This document is a draftfor review purposes only and does not constitute Agency policy. D-31 DRAFT--DO NOT CITE OR QUOTE 1 Okayasu et al (1999) analyzed the mutagenicity of Metsovo tremolite, erionite, and the 2 man-made ceramic (RCF-1) fiber. Whether this tremolite is the same as that used in previous 3 studies from this group is unclear. Tremolite from Metsovo, Greece, used in this study was 4 characterized as 2.4 3.1-pm long and 0.175 0.13-pm diameter (arithmetic mean) with the 5 number of fibers per microgram of sample equal to 1.05 x 1O' . Human-hamster hybrid A(L) 6 cells contain a full set of hamster chromosomes and a single copy of human chromosome 11. 7 Mutagenesis of the CD59 locus on this chromosome is quantifiable by antibody 8 complement-mediated cytotoxicity assay. The authors state that this is a highly sensitive 9 mutagenicity assay, and previous studies have demonstrated mutagenicity of both crocidolite and 10 chrysotile (Hei et al,, 1992). The cytotoxicity analysis for mutagenicity was performed by 11 exposing I x 105A(L) cells to a range of concentrations of fibers as measured by weight 12 (0-400 Lig/mL or 0-80 pg/cm2) for 24 hours at 37C. CD59 mutant induction showed a 13 dose-dependent increase in mutation induction for erionite and tremolite, but RCF-1 did not. 14 15 D.3. SUMMARY 16 In vitro studies have been conducted with Libby Amphibole asbestos from the Zonolite 17 Mountain mine. These studies demonstrated an effect of Libby Amphibole asbestos on 18 inflammation and immune function (Duncan et al.. 2010; Blake et al., 2008; Blake et al., 2007; 19 Hamilton et al., 2004). oxidative stress (Hillegass et al.. 2010). and genotoxicity (Pietruska et al.., 20 2010). These results suggest that Libby Amphibole asbestos may act through similar 21 mechanisms as other forms of asbestos, but data gaps still remain to determine specific 22 mechanisms involved in Libby Amphibole asbestos-induced disease. 23 Studies that examined cellular response to tremolite also found that fiber characteristics 24 (length and width) play a role in determining ROS production, toxicity, and mutagenicity 25 (Okayasu et al.. 1999; Wagner et al.. 1982). As with the in vivo studies, the definition of fibers 26 and the methods of fiber measurement vary among studies. 27 This document is a draftfor review purposes only and does not constitute Agency policy. D-32 DRAFT--DO NOT CITE OR QUOTE D.4. REFERENCES Athanasiou. K: Constantopoulos. SH; Rivedal. E; Fitzgerald. DJ: Yamasaki. H. (1992). Metsovo-tremolite asbestos fibres: In vitro effects on mutation, chromosome aberration, cell transformation and intercellular communication. Mutagenesis 7: 343-347. http://dx.doi.Org/10.1093/mutage/7.5.343. Bernstein, D; Rogers, R: Smith, P. (2005). The biopersistence o f Canadian chrysotile asbestos following Inhalation: final results through 1 year after cessation o f exposure. Inlial Toxicol 17: 1-14. http://dx.doi.org/10.1080/0895837059Q885663. Bem steia D: Rogers, R: Smith. P: Chevalier. J, (2006). The toxicological response o f Brazilian chrysotile asbestos: a multidose subchronic 90-day inhalation toxicology study with 92-day recovery to assess cellular and pathological response. Inlial Toxicol 18: 313-332. http://dx.doi.org/10.1080/0895837050Q515871. Bernstein. DM: Chevalier. J: Smith. P, (2003). Comparison o f Calidria chrysotile asbestos to pure tremolite: Inhalation biopersistence and histopathology following short-term exposure. Inlial Toxicol 15: 1387-1419. http://dx.doi.org/10.1080/0895837039Q248888. Blake. D: Bolin, C: Cox, D; Cardozo-Pelaez, F; Pfau. J. (2007). Internalization o f Libby amphibole asbestos and induction o f oxidative stress in murine macrophages. Toxicol Sci 99: 277-288. http://dx.doi.org/10.1093/toxsci/kfml66. Blake, D: Wetzel, S: Pfau, J. (2008). Autoantibodies from mice exposed to Libby amphibole asbestos bind SSA/Ro52-enriched apoptotic blebs o f murine macrophages. Toxicology 246: 172-179. http://dx.doi.Org/10.1016/i.tox.2008.0L008. Davis. JMG: Addison. J: Bolton. RE: Donaldsoa K: Jones, AD: Miller. BG. (1985). Inhalation studies on the effects o f tremolite and brucite dust in rats. Carcinogenesis 6: 667-674. http://dx.doi.Org/10.1093/carcin/6.5.667. Davis, JMG: Addison. J: McIntosh. C: Miller, BG: Niven. K, (1991). Variations in the carcinogenicity o f tremolite dust samples o f differing morphology. Ami N Y Acad Sci 643: 473-490. http://dx.doi.0 r g /lO .llll/i.l 7496632,199Ltb24497.x. Duncaa K: Gliio. A: Dailey. L: Bern. A: Gibbs-Flournov. E: Padilla-Carlin. D: Roggli. V: Devlin. R, (2010). Effect o f size fractionation on the toxicity o f amosite and Libby amphibole asbestos. Toxicol Sci 118: 420-434. http://dx.doi.org/10.1093/toxsci/kfa281. Hamilton, R: Hoiian. A: Morandi. M. (2004). A comparison o f asbestos and urban particulate matter in the in vitro modification o f human alveolar macrophage antigen-presenting cell function. Exp Lung Res 30: 147-162. Hei. TK: Piao. CO: He. ZY; Vannais. D: Waldrea CA, (1992). Chrysotile fiber is a strong mutagen in mammalian cells. Cancer Res 52: 6305-6309. Hillegass. JM: Shukla. A: MacPhersoa MB: Lathrop. SA: Alexeeva. V: Perkins. TN: van der Vliet. A: Vacek. PM: Gunter. ME: Mossrnan. BT, (2010). Mechanisms o f oxidative stress and alterations in gene expression by Libby six-mix in human mesothelial cells. Part Fibre Toxicol 7: 26. http://dx.doi.org/10.1186/1743-8977-726. ' Kamp. DW: Graceffa. P: Pryor, WA: W eitzmaa SA, (1992). The role o f free radicals in asbestos-induced diseases. Free Radic Biol Med 12: 293-315. http://dx.doi.org/10.1016/0891-5849(92)90117-Y. Kamp. DW: Weitzmaa SA, (1999). The molecular basis o f asbestos induced lung injury [Review]. Thorax 54: 638 652. http://dx.doi.Org/10.1136/tlix.54.7.638. This document is a draftfor review purposes only and does not constitute Agency policy. D-33 DRAFT--DO NOT CITE OR QUOTE Lowers. HA: Bern, AM. (2009). Particle size characterization o f water-elutriated Libby amphibole 2000 and RTI international amosite. Reston, VA: U.S. Geological Survey, http://pubs.usgs.gov/of/2009/1242/. McConnell. EE: Rutter. HA: Ulland. BM: Moore. JA. (1983). Chronic effects o f dietary exposure to amosite asbestos and tremolite inF344 rats. Environ Health Perspect 53: 27-44. Meeker. GP: Bern. AM: Brownfield. IK: Lowers. HA: Sutlev. SJ: Hoefen. TM: Vance. JS. (2003). The composition and morphology o f amphiboles from the Rainy Creek Complex, near Libby, Montana. American Mineralogist 88: 1955-1969. NTP, (National Toxicology Program). (1985). NTP toxicology and carcinogenesis studies o f chrysotile asbestos (cas no. 12001-29-5) inf344/nrats (feed studies) (Vol. 295). Research Triangle Park, NC. http://www.ntis. gov/search/product.aspx?ABBR=PB86167103. NTP. (National Toxicology Program). (1988). NTP technical report on the toxicology' and carcinogenesis studies of Crocidolite Asbestos (CAS No. 12001-28-4) In F344/N Rats (Feed Studies) (Vol. 280). Research Triangle Park, NC. http://ntp.nielrs.nih.gov/ntp/lrtdocs/LT rpts/tr280,pdf. NTP. (National Toxicology7Program). (1990a). NTP technical report on the toxicology and carcinogenesis studies o f amosite asbestos (CAS No. 12172-73-5) inF344/N Rats (Feed Studies) (Vol. 279). Research Triangle Park, NC. http://ntp.iriehs.irilr.gov/ntpdrtdocs/LT rpts/tr279.pdf. NTP. (National Toxicology Program). (1990b). NTP technical report on the toxicology and carcinogenesis studies o f tremolite (CAS no. 14567-73-8) inF344/N rats (feed studies) (Vol. 277). Research Triangle Park, NC. http://ntp.niehs.nih.gov/iitp/htdocs/LT rptsZtr277.pdf. Okavasu. R: Wu, L: Hei. TK, (1999). Biological effects o f naturally occurring and man-made fibres: In vitro cytotoxicity and mutagenesis in mammalian cells. Br J Cancer 79: 1319-1324. http://dx.doi.org/10.1038/si.bic.669Q213. Padilla-Carlin. DL Sclrladweiler. MCJ: Shamralraa JH: Kodavanti. UP: Nvska. A: Burgoon. LD: Gavett, SH, (2011). Pulmonary7inflammatory and fibrotic responses in Fischer 344 rats after intratracheal instillation exposure to libby amphibole. J Toxicol Environ Health A 74: 1111-1132. http://dx.doi.org/10.1080/15287394.201L58694Q. Pfau. JC: Sentissi, JJ: Li. S: Calderon-Garciduenas, L: Brown. JM: Blake. DJ. (2008). Asbestos-induced autoimmunity in C57BI/6 mice. J Immunotoxicol 5: 129-137. http://dx.doi.org/10.1080/15476910802Q85756. Pietruska, JR: Johnston. T; Zhitkovich. A: Kane, AB, (2010). XRCC1 deficiency sensitizes human lung epithelial cells to genotoxicity by crocidolite asbestos and Libby amphibole. Environ Health Perspect 118: 1707 1713. http://dx.doi.org/10.1289/ehp.1002312. Putnam. EA: Smartt A: Groves, A: Schwanke, C: Brezinski. M: Pershouse, MA, (2008). Gene expression changes after exposure to six-mix in a mouse model. J Immunotoxicol 5: 139-144. http://dx.doi.org/10.1080/15476910802Q85772. Roller. M: Pott, F: Kamino. K: A lthoff GH: Bellmann. B, (1996). Results o f current intraperitoneal carcinogenicity shidies with mineral and vitreous fibres. Exp Toxicol Pathol 48: 3-12. Saegusa. J: Kawano. S: Koshiba. M: Havashi. N: Kosaka. H: Funasaka. Y: Kumagai. S. (2002). Oxidative stress mediates cell surface expression o f SS-A/Ro antigen on keratinocytes. Free Radic Biol Med 32: 1006-1016. http://dx.doi.org/10.1016/S0891-5849(02)00797-9. This document is a draftfor review purposes only and does not constitute Agency policy. D-34 DRAFT--DO NOT CITE OR QUOTE Sahu. AP: Dogra. RK; Shanker. R: Zaidi. SH. (1975). Fibrogenic response in murine lungs to asbestos. Experimentelle Pathologie 11: 21-24. Shannahan. J; Scliladweiler. M: Padilla-Carlin. D: Nvska. A: Richards. J: Gliio. A: Gavett. S: Kodavanti. U. (2011a). The role o f cardiovascular disease-associated iron overload in Libby amphibole-induced acute pulmonary injury and inflammation. Inlial Toxicol 23: 129-141. http://dx.doi.org/10.3109/08958378.2011.55185Q. Shannahan, JH: Scliladweiler. MC: Richards. JH: Ledbetter. AD: Gliio, AJ: Kodavanti. UP. (2010). Pulmonary oxidative stress, inflammation, and dvsregulated iron homeostasis in rat models o f cardiovascular disease. J Toxicol Environ Health A 73: 641-656. http://dx.doi.org/10.1080/152873909035782Q8. Shannahan. JH: Gliio. AJ: Scliladweiler. MC: McGee. JK: Richards. JH: Gavett. SH: Kodavanti. UP. (201 lb). The role o f iron in Libby amphibole-induced acute lung injury' and inflammation. Inlial Toxicol 23: 313-323. http://dx.doi.org/ld.3109/08958378.201L569587. ' Smartt AM: Brezinskl. M: Trapkus. M; Gardner. D: Putnam, EA, (2010). Collagen accumulation over time in the murine lung after exposure to crocidolite asbestos or Libby amphibole. Environ Toxicol 25: 68-76. http://dx.doi.org/10.1002/tox.20472. Smith. WE, (1978). Final report on biologic tests o f samples 22260p5 and 22263p2. Madison, NJ: Health Research Institute, Fairleigh Dickinson University. Smith. WE: Hubert. DP. (1974) The intrapleural route as a means for estimating carcinogenicity. In E Karbe; JF Park (Eds.), Experimental lung cancer: Carcinogenesis and bioassays: International symposium held at the Battelle Seattle Research Center, Seattle, WA, USA, June 23-26, 1974 (pp. 92-101). New York, NY: Springer-Verlag. Smith. WE: Hubert. DP: Sobel. HJ: Marquet. E, (1979) Biologic tests o f tremolite in hamsters. In R Lenten; JM Dement (Eds.), Dusts and disease (pp. 335-339). Park Forest South, IL: Patliotox Publisher. Stanton. MF; Lavard. M; Tegeris. A: Miller, E: Mav. M; Morgan. E; Smith. A, (1981). Relation o f particle dimension to carcinogenicity in amphibole asbestoses and other fibrous minerals. J Natl Cancer Inst 67: 965-975. ' Suzuki. K; Hei. TK, (1996). Induction o f heme oxygenase in mammalian cells by mineral fibers: Distinctive effect o f reactive oxygen species. Carcinogenesis 17: 661-667. http://dx.doi.Org/10.1093/carcin/17.4.661. Trosko. JE: Yotti. LP: Warrea ST: Tsushimoto. G: Chang. CC. (1982). Inhibition o f cell-cell coimnunication by tumor promoters. Carcinog Compr Surv 7: 565-585. Wagner. J; Chamberlain, M; Brown. R: Berry, G: Poolev. F; Davies. R: Griffiths. D. (1982). Biological effects of tremolite. Br J Cancer 45: 352-360. Webber. JS: Blake. DJ: Ward. TJ: Pfau. JC. (2008). Separation and characterization o f respirable amphibole fibers from Libby, Montana. Inlial Toxicol 20: 733-740. http://dx.doi.org/10.1080/089583708Q1932544. Wylie, AG: Bailey, KF: Kelse. JW: Lee. RJ. (1993). The importance o f width in asbestos fiber carcinogenicity and its implications for public policy. Am Ind Hyg Assoc J 54: 239-252. http://dx.doi.org/10.1080/15298669391354621. Wylie, AG: Skinner. HCW: Marsh. J: Snyder, H: Garzione. C: Hodkinson. D: Winters. R: M ossmaa BT, (1997). Mineralogical features associated with cytotoxic and proliferative effects o f fibrous talc and asbestos on rodent tracheal epithelial and pleural mesothelial cells. Toxicol Appl Pharmacol 147: 143-150. http://dx.doi.org/10.1006/taap. 1997.8276. This document is a draftfo r review purposes only and does not constitute A gency policy. D-35 DRAFT--DO NOT CITE OR QUOTE 1 APPENDIX E. EVALUATION OF EXPOSURE-RESPONSE DATA FOR 2 LOCALIZED PLEURAL THICKENING IN WORKERS FROM THE 3 MARYSVILLE, OH COHORT 4 E .l. STATISTICAL ANALYSIS OF THE 2004 DATA SET FOR WORKERS HIRED IN 5 1972 AND LATER 6 All analyses were performed using SAS statistical software v. 9.1. Benchmark dose 7 lower bound 95% confidence intervals (BMCLs) were obtained by the profile likelihood method 8 as recommended by Crump and Howe (19851 using the NLMIXED procedure in SAS (Wheeler. 9 2005). As described in Section 5.2.1.4, the critical endpoint for RfC derivation is radiographic 10 evidence of localized pleural thickening (LPT; n = 12 cases), compared with the referent group 11 with no radiographic evidence of pleural abnormality (n = 106). 12 13 E.1.1. Investigation of Explanatory Variables 14 Dichotomous statistical models describing the probability of individual response as a 15 function of cumulative exposure (CE) as measured by cumulative human equivalent exposure for 16 continuous exposure (CEIEEC) in units of fiber/cc-year were used for this analysis. In order to 17 investigate the key explanatory variables for analysis, a forward-selection process was used to 18 evaluate the association of each of the potential covariates with odds of locali zed pleural 19 thickening, controlling for CEIEEC. Covariates considered for inclusion in the model were time 20 since first exposure, age at X-ray, gender, smoking history, and body mass index (BMI). This 21 initial modeling was done using a standard logistic regression model as commonly applied in the 22 analysis of epidemiological data. The base model was a logistic regression model with CE 23 (natural log transformed) as the independent variable. This model provided an adequate fit to the 24 data (Hosmer-Lemeshow/i-value of 0.6357), and the exposure variable was statistically 25 significantly associated with the outcome (beta [standard error, SE] = 0.5676 [0.2420], 26 /i-val ue = 0.0190). Covariates were evaluated according to whether inclusion of the covariate 27 improved model fit as assessed by the Akaike Information Criterion (AIC), and statistical 28 significance of the covariate. When controlling for CE, inclusion of each of the covariates with 29 the exception of smoking increased the AIC for the model (with the exception of BMI, due to 30 mi ssing information for some individuals), and none were associated with odds of discrete 31 pleural thickening: time since first exposure--p = 0.8879; age at X-ray--p = 0.7735; This document is a draftfor w fe * purposes only and does not constitute Agency policy. E-l DRAFT--DO NOT CITE OR QUOTE 1 gender--p = 0.7660; smoking--p = 0.1669; BMI--p = 0.4095. Therefore, only exposure (i.e., 2 CHEEC) was included in further analyses (see Table E-l). 3 4 5 Table E-l. Evaluation of covariates for the 2004 data set for workers hired 6 in 1972 and later 7 Covariate Wald p-value for Wald p-value for beta coefficient beta coefficient corresponding to corresponding to covariate exposure AIC Base model (only ln[CHEEC]) -- 0.0190 75.5 Time since first exposure 0.8879 0.0310 77.5 Age at X-ray 0.7735 0.0186 77.4 Gender 0.7660 0.0195 77.4 Smoking history 0.1669 0.0231 75.4 BMIa 0.4095 0.0102 56.7 8 9 aNote that only 97 observations were used, due to missing values (AIC not comparable). 10 11 12 E.1.2. Investigation of Candidate Models 13 The candidate models were logistic (with CHEEC considered as continuous and 14 continuous with a natural logarithm transformation), probit (with CHEEC considered as 15 continuous and continuous with a natural logarithm transformation), 3-parameter log-logistic, 16 dichotomous Hill, and dichotomous Michaelis-Menten models. These are statistical models used 17 to evaluate dichotomous data and were considered appropriate given the supralinear nature of the 18 observed relationship between Libby Amphibole asbestos1exposure and the prevalence of 19 localized pleural thickening (LPT); model forms are provided in Table E-2. For each of the 20 candidate models, exposure lags of 0, 5, 10, 15, and 20 years were investigated. Although zero 21 lag exposures are not likely to be biologically relevant (i.e., some lag is expected for 22 development of LPT), these models were included for completeness and for comparison of 'The term "Libby Amphibole asbestos" is used in this document to identify the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby, MT. It is further described in Section 2.2. This document is a draftfor review purposes only and does not constitute Agency policy. E-2 DRAFT--DO NOT CITE OR QUOTE 1 Table E-2. Evaluation of different model forms for the 2004 data set for 2 w orkers hired in 1972 and later 3 Model Exposure Metric Form3 Hosmer-Lemeshow AIC GOF /J-valuc BMC BMCL Logistic CHEEC P(LPT) = 1 E [1 + exp(-a - b 77.7 CHEEC)] 0.7423 -- -- CHEEC, lag 5 77.5 0.6914 1.5245 0.8836 CHEEC, lag 10 77.4 0.6751 1.4734 0.8540 CHEEC, lag 15 CHEEC, lag 20 77.6 0.6474 1.4510 0.8242 77.8 0.8800 -- -- Logistic ln(CHEEC) P(LPT) = 1 + [1 + exp(-a - b x 75.5 ln( CHEEC))] 0.6537 -- -- CHEEC, lag 5 75.2 0.5454 0.2281 0.0601 CHEEC, lag 10 74.6 0.5708 0.2028 0.0591 CHEEC, lag 15 74.7 0.6620 0.1686 0.0463 CHEEC, lag 20 75.4 0.8152 -- -- Probit model CHEEC P(LPT) = <D(a + b x CHEEC) 77.2 0.7698 -- -- CHEEC, lag 5 77.0 0.7146 1.3773 0.8481 CHEEC, lag 10 77.0 0.6864 1.3336 0.8048 CHEEC, lag 15 CHEEC, lag 20 77.2 0.6645 1.3148 0.7776 77.4 0.8884 -- -- Probit model ln(CHEEC) P(LPT) = cl>(a + b x ln(CHEEC)) 76.0 0.6041 -- -- CHEEC, lag 5 75.7 0.4967 0.2066 0.0502 CHEEC, lag 10 75.2 0.5385 0.1843 0.0496 CHEEC, lag 15 75.0 0.6166 0.1544 0.0441 CHEEC, lag 20 75.7 0.7945 -- -- 3-parameter log-logistic ln(CHEEC) P(LPT) = bkg + (1 - bkg) : |1 ` 74.9 e x p ( - a - b x ln(CHEEC))] 0.7030 -- -- CHEEC, lag 5 74.6 0.4894 0.3096 0.0979 CHEEC, lag 10 74.1 0.5853 0.2696 0.0888 CHEEC, lag 15 74.3 0.7238 0.2193 0.0693 CHEEC, lag 20 75.2 0.8277 -- -- Dichotomous Hillb ln(C'HEEC) P(LPT) = bkg + (Plateau - bkg) * 76.9 CHEECb- [exp(-a) + C'HEECb] 0.6040 -- -- CHEEC, lag 5 76.5 0.3598 0.3083 0.1015 CHEEC, lag 10 76.0 0.4244 0.2640 0.0923 CHEEC, lag 15 CHEEC, lag 20 76.2 0.6659 0.2112 0.0724 77.2 0.8277 -- -- Michaelis-Menten0 ln(CHEEC) P(LPT) = bkg + (Plateau - bkg) x 74.9 CHEEC - [exp(-a) + CHEEC] 0.5243 -- -- CHEEC, lag 5 74.5 0.3351 0.3096 0.1352 This document is a draftfor w fe * purposes only and does not constitute Agency policy. E-3 DRAFT--DO NOT CITE OR QUOTE Table E-2. Evaluation of different model forms for the 2004 data set for w orkers hired in 1972 and later (continued) Model Exposure Metric Form* Hosmer-Lemeshow AIC GOF /j-value BMC BMCL CHEEC, lag 10d 74.0 0.4163 0.2642 0.1177 CHEEC, lag 15 74.3 0.5664 0.2097 0.0898 CHEEC, lag 20 76.0 0.5610 -- -- 1 2 abkg indicates background rate, fixed at 1%. 3 bFor statistical modeling, the equivalent model form was used: P(LPT) = bkg + (Plateau - bkg) ^ 11 + exp(-a - b x 4 ln(CHEEC))]. ..................... 5 T or statistical modeling, the equivalent model form was used: P(LPT) = bkg + (Plateau - bkg) - 11 + exp(-a - 6 In(CHEEC))]. 7 dParameter estimates for the best-fitting models are as follows: 8 intercept = -0.1801 (SE = 1.0178), plateau = 0.5577 (SE = 0.3568, p e = 0.1207). 9 -- = no data. 10 11 12 relative model fits. Similarly, although we explored models with exposure lagged by 20 years, 13 there were cases of LPT in the full cohort with fewer than 20 years since first exposure; 14 therefore, using such a long lag (which necessitates the assumption that these are background 15 cases) was not judged to be appropriate, and the results are not further considered. 16 The various model forms were compared using AIC, and general model fit was evaluated 17 with the Hosmer-Lemeshow (2000) test (a form of the Pearson % goodness-of-fit [GOF] 18 statistic). This is a goodness-of-fit test that compares observed and expected events. 19 Observations are sorted in increasing order of estimated probability of the event occurring and 20 then divided into ~10 groups; the test statistic is calculated as the Pearson % statistic of observed 21 and expected frequencies in these groups. The benchmark concentration (BMC) was estimated 22 for each candidate model using a Benchmark Response (BMR) of 10% and assuming a 23 background rate of 1% (see Section 5.2.3.3). BMCs and corresponding BMCLs were estimated 24 for each of the candidate models. 25 All of the candidate models had adequate fit as assessed by the FIosmer-Lemeshow test. 26 Models were compared using the AIC values, ranging from 74.0 to 77.8. The model with the 27 lowest AIC was the Michaelis-Menten model with 10-year lagged exposure (AIC = 74.0) (see 28 Table E-2). Note that models with exposure lagged by 0 or by 20 years, which are considered 29 not to be biologically relevant, are shaded grey and not included as candidate models. This document is a draftfor w fe * purposes only and does not constitute Agency policy. E-4 DRAFT--DO NOT CITE OR QUOTE 1 There were several models that had similar model fits (within two AIC units) as the 2 best-fitting model, including the logistic and probit models with the natural log of CHEEC as the 3 exposure metric (lags of 5, 10, and 15 years), the 3-parameter log-logistic model (lags of 5, 10, 4 and 15 years), the dichotomous Hill model (lag of 10 years), and the Michaelis-Menten model 5 with exposure lagged by 5 or 15 years. All but one of these models would yield a BMCL lower 6 than that for the best-fitting model. However, the range was relatively narrow among these 7 similarly fitting models (BMCLs ranging from 0.0441 to 0.1352), with the lowest BMCL 8 ~2.7 times lower than the BMCL for the Michaelis-Menten model with exposure lagged by 9 10 years. 10 The Michaelis-Menten model using the 10-year lagged exposure had a/t-value for fit 11 of 0.42, an AIC value of 74.0, and an estimated plateau of 0.5577 (SE = 0.3568). This model 12 yielded a BMC of 0.2642 fiber/cc-year, and corresponding BMCL of 0.1177 fiber/cc-year for a 13 10% increase in prevalence of LPT (see Table E-2 and Figure E-l). 14 15 to 0.6 '3a 0.5 0.4 ft .Oaal 0.3 0.2 0.1 f 0 0 Estimated prevalence by cumulative exposure lagged by 10 years Observed prevalence by quartiles of unlagged cumulative exposure among cases 2 4 6 8 10 CHEEC (fibers/cc-year) 16 17 Figure E-l. Observed prevalence of localized pleural thickening and 18 estimated probability of localized pleural thickening. 19 20 This document is a draftfor w fe * purposes only and does not constitute Agency policy. E-5 DRAFT--DO NOT CITE OR QUOTE 1 The potential confounding effect of covariates was re-examined in the best-fitting model 2 (see Table E-3). As in the initial assessment, after controlling for the effect of exposure (i.e., 3 CHEEC, lagged by 10 years) there was no association between risk of LPT and time since first 4 exposure (p = 0.9973), age at X-ray (p = 0.8734), gender (p = 0.5544), or BMI (p = 0.3806), and 5 inclusion of each of these covariates increased the AIC (with the exception of BMI, due to 6 missing information for some individuals). The variable representing smoking history did not 7 meet the alpha = 0.05 criteria for statistical significance (p = 0.0841), although inclusion of this 8 variable decreased the AIC from 74.0 in the base model, to 72.3. Smoking was not considered 9 further in the derivation of the RfC due to the lack of statistical significance at the alpha = 0.05 10 level. However, because inclusion of the smoking variable did improve model fit, it is 11 investigated further as a sensitivity analysis in Section E.2. 12 13 14 Table E-3. Evaluation of covariates for the 2004 data set for workers hired 15 in 1972 and later in the best-fitting model 16 W ald /7-value for beta coefficient corresponding C ovariate to covariate Plateau (SE) AIC Base model (only CHEEC) -- 0.5577 (0.3568) 74.0 Time since first exposure 0.9973 0.5580 (0.3634) 76.0 Age at X-ray 0.8734 0.5707 (0.3793) 76.0 Gender 0.5544 0.6167 (0.4138) 75.7 Smoking history 0.0841 0.5927 (0.3779) 72.3 BMI* 0.3806 0.4622 (0.2810) 55.8 17 18 *Note that only 97 observations used due to missing values (AIC not comparable). 19 20 21 To evaluate the assumption of a 1% background rate of LPT, the best-fitting model (i.e., 22 Michaelis-Menten with 10-year lagged exposure) was rerun, allowing the background rate to be 23 estimated as a parameter rather than fixed at 1%. The resulting estimated background rate was 24 quite close to the assumed rate of 1%, at 3.12% (SE = 2.84%). Both the fixed and estimated 25 values are in the range of estimates from previous studies (see Section 5.2.3.3.), and the This document is a draftfor review purposes only and does not constitute Agency policy. E-6 DRAFT--DO NOT CITE OR QUOTE 1 difference in the BMCL when the background rate is fixed at 1% versus when it is estimated is 2 -15% (0.1177 compared to 0.1349 fiber/cc-year). 3 4 E.1.3. Derivation of the Candidate Point of Departure (POD) and Reference Concentration 5 (RfC) for Localized Pleural Thickening Using the Michaelis-Menten Model 6 The candidate point of departure (POD) is 0.1177 fiber/cc-year, the BMCLio for this data 7 set. The reference concentration (RfC) is derived from the POD using the duration of exposure 8 of 70 years, lagged by 10 years, and a total uncertainty factor of 100. See Section 5.2.4. 9 RfC = [0.1177 (fiber/cc) x (year)] x 1 - (70 - 10) years x 1/100 = 1.96 x io -5 = 2 x 10"5 10 fibers/cc (rounded to 1 significant digit). 11 12 E.2. SENSITIVITY ANALYSIS FOR EFFECT OF SMOKING IN THE 2004 DATA SET 13 FOR WORKERS HIRED IN 1972 AND LATER 14 Due to the lack of statistical significance, smoking was not included in further analyses 15 for derivation of the RfC. However, based on the literature suggesting that smoking may play a 16 role in determining risk of LPT (see Section 5.3.6), the role of smoking was investigated further 17 for these sensitivity analyses. 18 The prevalence of any smoking history was 75.0% (n = 9) among cases, and 51.9% 19 (n = 55) among noncases. As noted above, the smoking variable was not significant at the 20 alpha = 0.05 level in the best-fitting (i.e., Michaelis-Menten) regression model controlling for 21 CHEEC lagged by 10 years (p = 0.08), but inclusion of the smoking variable did decrease the 22 AIC (AIC of 72.3 compared to 74.0 for the base model; see Table E-4). These results 23 (borderline statistical significance of the term but nontrivial improvement in model fit) may 24 indicate that smoking is associated with another variable that is associated with the outcome, or 25 that the variable is too poorly measured to accurately reflect the effect of smoking. 26 To evaluate whether smoking may modify the effect measure for the association between 27 Libby Amphibole asbestos exposure and risk of LPT, a third model was fit, which added an 28 interaction term between the exposure metric and smoking; in this model, neither the smoking 29 variable by itself nor the interaction term were significant (p = 0.2278 and p = 0.6598, 30 respectively), and the AIC increased from the base model (i.e., AIC of 74.1). Therefore, only 31 smoking (no interaction term) was retained for further sensitivity analyses. 32 This document is a draftfor review purposes only and does not constitute Agency policy. E-7 DRAFT--DO NOT CITE OR QUOTE 1 Table E-4. Evaluation of smoking in the best-fitting model 2 M odel3 AIC 1 74.0 2 72.3 Variable (None) Smoke Beta -- 1.8232 /j-value -- 0.0841 3 74.1 Smoke 2.5401 0.2278 Ln(CHEEC, lag 10)* Smoke 0.2182 0.6598 3 4 aThe following model forms were used for statistical analysis: 5 (1) P(LPT) =bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10))] 6 (2) P(LPT) = bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10) - beta*Smoke)] 7 (3) P(LPT) =bkg + (Plateau - bkg) / [1 + exp(-a - ln(CHEEC, lag 10) - beta*Smoke - beta2*ln(CHEEC, lag 8 10)* Smoke)] 9 10 11 The preferred model for RfC derivation (i.e., Model 1) yielded a BMC and a BMCL of 12 0.26 and 0.12 ftber/cc-year, respectively (see Table E-5). Model 2, which includes the smoking 13 variable, was used to derive estimates for smokers and nonsmokers separately. The BMC and 14 BMCL were derived by setting the beta coefficient for smoking to zero for nonsmokers, and to 15 the maximum likelihood (MLE)-estimated value (1.82) for smokers. The BMCL for nonsmokers 16 was about twice as high (0.25 ftber/cc-year) as that for the full cohort, while the POD for 17 smokers was about 1/3 that of the full cohort (0.04 ftber/cc-year). 18 19 20 Table E-5. Evaluation of smoking on estimated BMCs and BMCLs 21 M odel Group BM C (fiber/cc-year) BMCL (fiber/cc-year) 1 All 0.2642 0.1177 2 Nonsmokers 0.9344 0.2463 2 Smokers 0.1509 0.0398 22 23 24 The lower BMCL among smokers compared to nonsmokers may indicate that smoking 25 increases risk for development of LPT among individuals exposed to Libby Amphibole asbestos; 26 another possibility is that smoking may affect the timing and progression of LPT development. 27 If LPT develops sooner among smokers compared to nonsmokers, this could lead to a higher This document is a draftfor review purposes only and does not constitute Agency policy. E-8 DRAFT--DO NOT CITE OR QUOTE 1 prevalence of LPT among smokers at a given observation time, and subsequently higher 2 estimated risk. The lack of detailed smoking information in this cohort (such as pack-years) 3 limits the ability to explore the effect of smoking on LPT risk among individuals exposed to 4 Libby Amphibole asbestos, but these sensitivity analyses indicate that smoking should be 5 considered when evaluating risk of respiratory health outcomes in this group. 6 7 E.3. STATISTICAL ANALYSIS OF THE FULL DATA SET 8 E.3.1. Identification of Key Explanatory Variables 9 In order to begin modeling the data, key explanatory variables were identified using 10 logistic regression to analyze the data of Rohs et al. (2008). Logistic regression was performed 11 using the R statistical software, version 2.11.1. All fitting was performed using individual data, 12 without any grouping. The dependent variable was localized pleural thickening (n = 59) noted 13 on chest X-rays of former workers in the Marysvill e, OH facility (// = 252) and no reported 14 history of exposure to commercial asbestos at other locations. The available potential 15 explanatory variables included CE at the time of X-ray, fiber/cc-year (equivalent to CHEEC used 16 in the University of Cinci nnati report); time since first exposure (T; defined as time between first 17 exposure and date of X-ray in years); age at time of X-ray; gender; smoking status (i.e., ever, 18 never); and BMI. The BMI variable was missing for 34 individuals. 19 Initial analysis showed that CHEEC was a significant explanatory variable using both 20 CHEEC and ln(CHEEC). The strategy used to determine what other explanatory variables were 21 influential consisted of including CHEEC and then adding one additional explanatory variable at 22 a time. Explanatory variables havingp >0.2 were dropped from further consideration. 23 Explanatory variables havingp <0.2 were given further consideration. 24 BMI was investigated as a potential explanatory variable because fat pads can sometimes 25 be misdiagnosed as pleural thickening. Thus, there might be a positive relation between BMI 26 and pleural thickening. Analysis of a model with CHEEC or ln(CEIEEC) plus BMI (// = 218) 27 showed that BMI was not a significant explanatory variable. Two subsequent models using BMI 28 cutoffs of 25 and 30 also showed that BMI was not a significant explanatory variable. Analysis 29 of a model with CHEEC or ln(CHEEC) plus smoking indicated smoking was not a significant 30 explanatory variable. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. E-9 DRAFT--DO NOT CITE OR QUOTE 1 Analysis of a model of CHEEC plus gender indicated gender was a potential contributing 2 explanatory variable (p = 0.18). However, it should be noted that the worker cohort was highly 3 imbalanced with 236 males and 16 females. Only three females have a cumulative human 4 equivalent exposure greater than 0.15 fiber/cc-year. These considerations indicated that the 5 potential relevance of gender as an explanatory variable should be viewed with caution. 6 Analysis of ln(CHEEC) plus gender showed that gender was not a significant explanatory 7 variable. Accordingly, gender was eliminated as an explanatory variable. 8 The importance of T (time since first exposure) is clearly illustrated by comparing the 9 results of Lockey et al. (1984) with the results of Rohs et al. (2008). These two studies were 10 conducted in the same occupational cohort 24 years apart. In the initial study (Lockey et ah. 11 1984). only 2% of the individuals showed pleural changes; in the follow-up study (Rohs et al.. 12 2008). 28% of the individuals showed pleural changes. Logistic fitting of a model including 13 CHEEC or ln(CHEEC) plus T showed that T was a highly significant explanatory variable with p 14 <0.0005. This result is consistent with findings in other occupational cohorts exposed to various 15 forms of asbestos fibers that the time since first exposure is a significant explanatory vari able, 16 even in the absence of continued exposure (Ehrlich et al.. 1992; Jarvholm. 1992). T was retained 17 as an explanatory variable. However, an important point of clarification is that the T variable is 18 not the same as time of event. The LPT could have formed at any time before the X-ray was 19 taken (e.g., LPT detected in 2004 could have been present in 1990). 20 Analysis of a model of CE1EEC plus age at X-ray indicated that age was a significant 21 explanatory variable with p = 0.032. Analysis of a model of ln(CElEEC) plus age at X-ray 22 showed that age at X-ray was a potentially significant explanatory variable with p = 0.14. It 23 should be noted that this result does not mean that age is an independent risk factor for the 24 development of localized pleural thickening. In fact, there is no biological evidence that age is 25 an independent predictor of the development of localized pleural thi ckening without a history of 26 previous exposure to durable mineral fibers such as amphibole fibers. With a history of exposure 27 to amphibole fibers, age has been shown to be related to pleural thickening (Amandus et al.. 28 1987). However, it is quite possible that the association between age and prevalence is because 29 age at X-ray is related to T from first exposure, which is clearly one of the key explanatory 30 variables. Therefore, age at X-ray was not included as an explanatory variable. 31 This document is a draftfor re\'iew purposes only and does not constitute Agency policy. E-10 DRAFT--DO NOT CITE OR QUOTE 1 E.3.2. Selection of Model Form 2 Figure E-2 (see Panel A) presents a plot of prevalence of LPT as a function of T, 3 stratified by CE (CHEEC). As seen, the prevalence appears to be low (i.e., close to zero) until 4 about 15-20 years after first exposure and then appears to rise in a nonlinear fashion. Figure E-2 5 (see Panel B) presents a plot of prevalence as a function of CE, stratified according to time since 6 first exposure. As seen, prevalence appears to rise rapidly with increasing CE but then tends to 7 flatten out (plateau). Based on these attributes of the base data set, the objective was to select a 8 model that included a plateau term whose value depended on T. Several alternative model forms 9 were investigated, using the dichotomous Hill model as the starting point: 10 11 12 /i(CHEEC) = bkg + (Plateau - bkg) ^ [1 + exp {-a - b x ln(CHEEC)}] 13 14 15 In the dichotomous Hill model, the plateau term is a constant, with a value bounded 16 between background and 1.0. In order to be consistent with the data, this model was modified so 17 that the plateau term was a function of T. Several different nonlinear equations for the plateau 18 function were tested, including the following: 19 20 21 Plateau = MIN[1, bkg + (1 - bkg) x kl x 7] 22 Plateau = MIN[1, bkg + (1 - bkg) x kl x T2] 23 Plateau = MIN[1, bkg + (1 - bkg) x kl x J 3] 24 Plateau = bkg + (1-bkg) x 0(7]m,s),where 0(7]m,s) = cumulative normal probability 25 function 26 Plateau = bkg + (1-bkg) x G(7]a,P), where G(7]a,P), = cumulative gamma probability 27 function 28 Plateau = bkg + (1-bkg) x W(7]a,P), where W(7]a,P) = cumulative Weibull probability 29 function 30 31 This document is a draftfor review purposes only and does not constitute Agency policy. E -11 DRAFT--DO NOT CITE OR QUOTE This document is a draftfo r review purposes only and does not constitute Agency policy. E-12 Pane! A. Prevalence vs. Time From First Exposure (Grouped by CHEEC) Panel B. Prevalence vs. CHEEC (Grouped by Time From First Exposure) Figure E-2. Raw data plots. DRAFT--DO NOT CITE OR QUOTE CHEEC Bins (f/cc-yrs) Bin No. Min Max CHEEC 1 0 0.05 CHEEC 2 0.05 0.1 CHEEC 3 0.1 0.2 CHEEC 4 0.2 1 CHEEC 5 1 35 Mean 0.021 0.071 0.145 0.452 9.728 N Cases Prev 67 2 3.0% 44 1 2.3% 108 10 9.3% 101 20 19.8% 114 28 24.6% T Bins (years) Index Min Max T1 0 10 T2 10 20 T3 20 30 T4 30 40 T5 40 50 Mean 4.39 12.69 25.41 34.50 45.76 N Cases Prev 87 1 1.1% 53 0 0.0% 123 8 6.5% 118 27 22.9% 53 25 47.2% 1 AIC values when the plateau term is T, '/'2, T3, cumulative normal, cumulative gamma, or 2 cumulative Weibull are 293.97, 279.21, 276.12, 277.30, 277.07, and 276.98, respectively. The 3 plateau term based on T was not chosen because the curve reaches a plateau of 1when T is 4 about 50 years. Of those that have a plateau less than 1 at high T, the plateau term based on the 5 cumulative normal function was chosen because of its ease of use and familiarity. 6 Combining this equation for the plateau term with the basic probability model yields: 7 8 9 p (CHEEC,7) = bkg + (1 - bkg) x <fr(T\m,s) [1 + exp{-a - b x ln(CHEEC)}] 10 11 12 Further testing indicated that the lowest AIC was achieved when the b term was set to 13 1.0, resulting in a modified version of the discrete Michaelis-Menten equation: 14 15 16 p (CHEEC,7) = bkg + (1 - bkg) x <&(T\m,s) [1 + exp{-a -ln(CHEEC)}] 17 18 19 This equation can also be written as: 20 21 22 p (CHEEC,7) = bkg + (1 - bkg) x <&(l\m,s) x { CHEEC/[CHEEC + exp(-a)]} 23 24 25 This equation was selected as the preferred model for fitting to the data. In this model, T 26 (years) and CHEEC (fiber/cc-year) are explanatory variables. Fitting parameters of the 27 cumulative normal function are m (mid-point) and 5 (steepness). The a term is the intercept of 28 the exponential term when CHEEC equals 1 (ln(CHEEC) equals zero). Background is assumed 29 to be a constant (0.01) (see Section 5.2.3). 30 31 E.3.3. Parameterization 32 Fitting of the model to selected data sets was performed using the method of MLE, using 33 individual data without binning. The BMC for any specified value of T is calculated from the 34 MLE parameters and the specified value of T as follows: 35 This document is a draftfor review purposes only and does not constitute Agency policy. E -13 DRAFT--DO NOT CITE OR QUOTE 1 BMCt = exp [-a - In{Q x <b{T\m,s) - 1}] 2 3 Where: 4 Q = 0 ~ bkg) (BMR - bkg) 5 6 7 For a BMR of 10% extra risk, the value Q is 0.10. 8 9 E.3.4. Model-Fitting Results 10 Table E-6 provides the model-fitting results for each of the three data sets evaluated for 11 each of 5 lags of CHEEC and for each of 5 values of T. In all cases, the BMR is 10% extra risk. 12 Based on a background rate of 0.01, this BMR corresponds to a probability of LPT of 0.109. 13 Inspection of this table reveals that, for each of the three data sets evaluated, there is 14 relatively little effect of CHEEC lag over the interval 0-15 years. For the full data set and the 15 sub-cohort of workers hired in 1972 and later, the lowest AIC is achieved for a lag of 10 years. 16 It should be noted that the time since first exposure in the full cohort ranged up to 47.4 years; 17 therefore, estimates for values of T greater than 47.4 years represent extrapolation outside the 18 range of observed data, and should be interpreted with caution. 19 Figure E-3 presents a graph comparing the observed data to the predicted values from the 20 model (no lag) for the full data set. As above, this requires grouping the observed data into bins, 21 even though fitting was performed using the individual data. Because the choice of bins is 22 arbitrary, the appearance of the graphs would likely be changed somewhat if different bins were 23 chosen. Nevertheless, it seems apparent that the model predictions are in good accord with the 24 data. 25 26 This document is a draftfor review purposes only and does not constitute Agency policy. E-14 DRAFT--DO NOT CITE OR QUOTE This document is a draftfo r review purposes only and does not constitute Agency policy. Table E-6. Model-fitting results for the full data set if --O4 Study 1980+ 2004 1980+ 2004 1980+ 2004 Year of Hire All > 1972 < 1972 N Cases 434 61 198 13 236 48 CHEEC Lag 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 MLE Parameters m s a 42.38 13.30 1.977 42.44 13.54 2.000 42.58 14.10 2.061 42.86 15.16 2.167 43.28 16.06 2.395 31.41 10.47 -0.015 31.58 11.81 0.095 3.5E+05 3.0E+06 0.162 1.4E+06 5.4E+06 6.5E-01 2.1E+06 4.2E+06 1.5E+00 43.15 13.33 2.259 43.22 13.46 2.331 43.44 13.88 2.472 43.71 14.83 2.625 44.18 15.84 2.903 AIC 278.02 277.87 277.61 277.67 279.11 88.85 88.43 87.81 88.29 91.23 192.77 192.86 193.04 193.34 194.27 f =30 BMC BMCL 0.1822 0.0709 0.1711 0.0666 0.1477 0.0580 0.1166 0.0486 0.0876 0.0349 0.2930 0.1023 0.2623 0.0956 0.2402 0.0905 0.1766 0.0643 0.1036 0.0220 0.1689 0.0227 0.1540 0.0190 0.1270 0.0092 0.0934 0.0642 - T= 35 BMC BMCL 0.0731 0.0260 0.0707 0.0253 0.0651 0.0235 0.0567 0.0219 0.0449 0.0159 0.1900 0.0399 0.1770 0.0399 0.2402 0.0432 0.1766 0.0315 0.1036 0.0059 0.0613 0.0071 0.0569 0.0059 0.0492 0.0028 0.0406 0.0303 - BMC 0.0421 0.0412 0.0389 0.0352 0.0286 0.1462 0.1374 0.2402 0.1766 0.1036 0.0341 0.0318 0.0279 0.0241 0.0185 O-r II p-i BMCL 0.0138 0.0136 0.0129 0.0124 0.0091 0.0227 0.0232 0.0262 0.0185 0.0029 0.0037 0.0031 0.0015 - J =50 BMC BMCL 0.0224 0.0067 0.0221 0.0066 0.0212 0.0064 0.0197 0.0062 0.0162 0.0045 0.1177 0.0136 0.1082 0.0142 0.2402 0.0162 0.1766 0.0107 0.1036 0.0013 0.0175 0.0017 0.0164 0.0014 0.0145 0.0007 0.0128 0.0101 - BMC 0.0157 0.0154 0.0146 0.0133 0.0107 0.1128 0.1011 0.2402 0.1766 0.1036 0.0119 0.0111 0.0097 0.0084 0.0065 BMCL 0.0042 0.0042 0.0040 0.0038 0.0028 0.0109 0.0112 0.0123 0.0075 0.0007 0.0010 0.0008 0.0004 - E - 15 DRAFT--DO NOT CITE OR QUOTE The BMC for any specified value of T is calculated from the model parameter estimates and the specified value of T as follows: BMC = exp[-a - ln{(l - bkg) - (BMR - bkg) x <SXJ\m,s) - 1}]. The BMCL is estimated by rewriting the model so that BMC appears as an explicit term in the model, for a specified T of interest: BMR = bkg + (1-bkg) x dH'Ifi.s) + (1 + exp(-a - ln(BMC)], Solving for a yields: -a = ln[Q x (I>(71/m,.v)-1| + ln(BMC), Substituting yields: /TCHEEC.7) = bkg + (1 - bkg)*<b(7|m.s) + 11 + exp(z')], Where: z ' = ln((9*<D(7|m..v) - 1) + ln(BMC) - ln(CHEEC), Simplifying yields: /TCHEEC. / ) = bkg + (1 - bkg) x (T\m,s) + [1 + Q x (T\m,s) - 1) x BMC - CHEEC] Using this equation, a trial value of the BMC is selected and treated as a constant, and the equation is refit to the data to find the MLE values of the remaining parameters (m,s). After optimization, the value of the log-likelihood is recorded for the specified trial value of the BMC, and the process is repeated for other trial values of the BMC. The BMCL is the trial value of the BMC where the log-likelihood decreases from the MLE log-likelihood value by an amount equal to CHIDIST (2a, 1) - 2. For a = 0.05, the decrease is 1.3528. Panel A. Observed vs Predicted Prevalence as a Function of Time Since First Exposure (Grouped by CELEEC) CHEEC Bins (fiber/ccyears) Index Min Max Mean jp Cases Prev CHEEC 1 0 0.05 0.021 67 2 3.0% CHEEC 2 0.05 0.1 0.071 44 1 2.3% CHEEC 3 0.1 0.2 0.145 108 10 9.3% CHEEC 4 0.2 1 0.452 101 20 19.8% 0 10 20 30 40 50 Time From First Exposure (years) CHEEC 5 1 35 9.728 114 28 24,6% Panel B. Observed vs Predicted Prevalence as Function of CiTEEC (Grouped by Time Since First Exposure) Index T1 T2 T3 T4 T5 Min (year) 0 10 20 30 40 Max (year) 10 20 30 40 50 Mean (year) 4.39 12.69 25.41 34.50 45.76 W Cases Prev 87 1 1.1% 53 0 0.0% 123 8 6.5% 118 27 22,9% 53 25 47.2% 2 Figure E-3. Observed versus predicted for base-case data set. 3 ~ 4 E.3.5. Derivation of the POD and RfC for Localized Pleural Thickening Using the 5 Cumulative Normal Michaelis-Menten Model 6 For comparison with the primary analysis, a POD and RfC are derived for LPT from the 7 combined 1980 + 2004 data set as it provides the widest distribution of L-values (see 8 Section 5.2.3.2). A lag period of 5 years is used because Larson et al. (2010) showed that LPT 9 could be observed much earlier than previously thought. 10 Because the RfC is intended to provide protection for a lifetime of exposure (exposure 11 begins at birth and continues to age 70), the POD is the BMCLio with T= 70 years of 12 0.0042 fiber/cc-year calculated with the cumulative normal Michaeli s-Menten model (from 13 Table E-6). The POD is divided by duration of exposure of 70 years, lagged by 5 years, and then This document is a draftfor w fe * purposes only and does not constitute Agency policy. E-16 DRAFT--DO NOT CITE OR QUOTE 1 divided by an uncertainty factor (see Section 5.2.4). In this case, as the model accounts for the 2 full lifetime of exposure of 70 years, the uncertainty factor of 100 is reduced to 30. 3 RfC = [0.0042 (fiber/cc) x (year)] x l -h(70 - 5) years x 1/30 = 2 x 10 6fibers/cc 4 (rounded to one significant digit). 5 To provide a frame of reference, the calculation above was repeated with the data set 6 restricted to those hired in 1972 or later, when industrial hygiene data were collected in the 7 facility (from Table E-6). 8 RfC = [0.0112 (fiber/cc) x (year)] x l -h(70 - 5) years x 1/30 = 7 x 10 6fibers/cc 9 (rounded to one significant digit). 10 The reasonably good correlation in the calculated RfCs with the two different data sets 11 (2 x 1O'6versus 7 x 10-6 fiber/cc) provides some confidence in the exposure reconstruction 12 pre-1972. 13 An alternative candidate POD is the BMCLio with 7 = 40 years of 0.0136 fiber/cc-year 14 calculated with the Cumulative Normal Michaelis-Menten model (from Table E-6). The 15 BMCLio with T= 40 years is used because it is near the upper end of the range of T-values 16 available in the data set (Tmax= 47.375 years). A lag time of 5 years and a total uncertainty 17 factor of 100 are used(see Section 5.2.5). 18 RfC = [0.0136 (fiber/cc) x (year)] x 1 -h(40 - 5) years x 1 -h 100 = 4 x 10 6 fibers/cc 19 (rounded to one significant digit). 20 21 E.3.6. Sensitivity Analysis 22 The University of Cincinnati increased the exposure metric by a factor of 2 between 1972 23 and 1967 to account for conditions in the facility before engineering controls were added. For 24 the purposes of comparison, the CE was also calculated without this doubling. Plots of 25 prevalence of LPT with these two different exposure metrics are virtually identical (not shown). 26 One worker in the 1980 study was exposed only 5 months before X-ray and showed LPT. 27 Excluding this worker from the analysis did not change the calculated RfC. 28 Figure E-4 shows a plot of the PODs (fiber/cc-year) versus time since first exposure 29 (years) calculated from the Michaelis-Menten model using the 2004, data for workers hired in 30 1972 and later (see Section E.l), and from the Cumulative Normal Michaelis-Menten model 31 using the full data set (see Table E-6). Because the Michaelis-Menten model is independent of This document is a draftfo r review purposes only and does not constitute Agency policy. E -17 DRAFT--DO NOT CITE OR QUOTE 1 time since first exposure, the mean value of T for the data set is used. As there are few 2 individuals with long T (maximum of 47 years) and low CE, it is not clear whether the apparent 3 plateau with the Cumulative Normal Michaelis-Menten model is a reflection of the limitation of 4 the data or an expression of the underlying biology. 5 6 Time From First Exposure (years) 7 8 Figure E-4. PODs (fiber/cc-year) versus time since first exposure (years). 9 ' ' ` ` 10 11 This document is a draftfor w ife * purposes only and does not constitute Agency policy. E -18 DRAFT--DO NOT CITE OR QUOTE 1 E.4. REFERENCES Amandus. HE; Altliouse. R; Morgan. WKC: Sargent. EN; Jones. R. (1987). The morbidity and mortality o f venniculite miners and millers exposed to tremolite-actinolite: Part III. Radiographic findings. Am JInd Med 11: 27-37. Crump, KS: Howe. RB, (1985) A review o f methods for calculating statistical confidence limits in low dose extrapolation. In DB Clayson; D Krewski; I Munro (Eds.). Toxicological risk assessment: Vol 1 Biological and statistical criteria (pp. 187-203). Boca Raton, FL: CRC Press, Inc. Ehrlich. R: Lilis, R; Chan. E: Nicholson. WJ: Selikoff, IJ. (1992). Long term radiological effects o f short term exposure to amosite asbestos among factory workers. Br J Ind Med 49: 268-275. Hosmer, DW. Jr: Lameshow, S. (2000). Applied logistic regression. In. New York, N Y : John Wiley & Sons. Jarvholm, B. (1992). Pleural plaques and exposure to asbestos: a mathematical model. Int J Epidemiol 21: 1180-1184. Larson, T; Mever, C: Kapil. V: Gurnev. J: Tarver. R: Black, C: Lockev. J. (2010). Workers with Libby amphibole exposure: retrospective identification and progression of radiographic changes. Radiology 255: 924-933. http://dx.doi.org/10.1148/radiol.10091447. Lockev, JE: Brooks, SM: Jarabek, AM: Khourv, PR: McKay, RT: Carson. A: Morrison. JA: Wiot, JF: Spitz, HB, (1984). Pulmonary changes after exposure to venniculite contaminated with fibrous tremolite. Am Rev Respir Dis 129: 952-958. Robs, A: Lockev, J: Dunning. K: Shukla, R: Faa H: Hilbert, T: Boiton, E: Wiot, J: Mever. C: Shipley, R: Lemasters, G: Kapil, V. (2008). Low-level fiber-induced radiographic changes caused by Libby venniculite: A 25-year follow-up study. Am J Respir Crit Care Med 177: 630-637. http://dx.doi.org/10.1164/rccm.200706-841QC. Wheeler. MW. (2005). Benchmark dose estimation using SAS. In Proceedings o f the Thirtieth Annual SAS Users Group International Conference. Cary, NC: SAS Institute. http://www2.sas.eom/proceedings/sugi30/201-30.pdf. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. E -19 DRAFT--DO NOT CITE OR QUOTE 1 APPENDIX F. MARYSVILLE, OH WORKER OCCUPATIONAL EXPOSURE 2 RECONSTRUCTION 3 The Development of a Cumulative Human Equivalent Exposure Concentration 4 5 6 7 8 BY: 9 James E. Lockey, MD, MS 10 Carol Rice, PhD 11 Eric Borton, BS 12 Timothy Hilbert, MS 13 Grace LeMasters, PhD 14 15 16 17 18 19 20 University of Cincinnati 21 Department of Environmental Health 22 3223 Eden Ave., ML 0056 23 Cincinnati, OH 45267 24 This document is a draftfor review purposes only and does not constitute Agency policy. F-l DRAFT--DO NOT CITE OR QUOTE 1 F.l. INTRODUCTION 2 This project builds on the previous work of Dr. James Lockey et al. investigating possible 3 effects of exposures to dust containing Libby Amphiboles at a plant in Marysville, OH (Rohs et 4 al.. 2008; Lockey et al.. 1984). The data used in the original exposure reconstruction and as 5 reported in the published manuscripts, was based on the exposures measurements available at 6 that time (Lockey et al.. 1984). This exposure reconstruction is based on approximately five 7 times additional occupational fiber exposure data than was previously utilized in 1980. These 8 exposure measurements were recently obtained by the U.S. Environmental Protection Agency 9 (HPA) from the company and through trial transcripts from the United States of America vs. 10 W.R. Grace, et al., as well as the archived data used in the 1980 exposure reconstruction. Four 11 steps were undertaken to construct an exposure matrix describing exposure over each year from 12 1957 to 2000. In a final fifth step, this matrix was used to calculate an exposure metric for 13 workers. 14 15 16 1. Data searches, requests, and document selection 17 2. Document evaluation, data entry, cleaning, editing and standardization 18 3. Completeness and trends in measurements 19 4. Decisions relevant to the exposure matrix 20 5. Development of a cumulative human equivalent exposure concentration 21 22 23 F.2. DATA SEARCHES, REQUESTS, AND DOCUMENT SELECTION 24 Three sources of paper records were identified. First, sampling reports from OM Scott 25 that included measurements at the facility pre- and post-1980 were received via the EPA. These 26 reports contained both measurement results and information about the plant. OM Scott was also 27 contacted with a request for available maps of the plant layout prior to 1980. Secondly, archived 28 files from the Lockey et al. (1984) study were identified. Lastly, as a result of the recent W.R. 29 Grace trial, there was additional discovery of material relevant to the OM Scott plant. The 30 Department of Justice (DOJ) was contacted for the release of these data. There were seven 31 4" binders available for review and every page (approximately 3,150 pages) was scanned This document is a draftfor review purposes only and does not constitute Agency policy. F-2 DRAFT--DO NOT CITE OR QUOTE 1 visually to identify pages relevant to the current project. Aspects of particular interest included 2 the manufacturing process, usage and source of raw materials, engineering and design changes in 3 the plant, work practices and exposure assessment methodology. Approval was received from 4 the DOJ to utilize the relevant data for this project. 5 6 F.3. DOCUMENT EVALUATION, DATA ENTRY (QUALITATIVE AND 7 QUANTITATIVE), CLEANING, EDITING AND STANDARDIZATION 8 All of the records--both the qualitative and quantitative--were reviewed in this second 9 phase. 10 11 F.3.1. Qualitative Information 12 Written reports, letters, memos, and notes contained background information on plant 13 operations. A total of 1,489 pages were read for potentially useful and pertinent information 14 regarding OM Scott and abstracted into a data file. From these records, we obtained: 15 16 17 Plant layout, including changes over time. This allowed us to associate the 18 descriptions used on air sampling data forms/reports with jobs or departments 19 within the plant. A limited number of aerial images were available to identify 20 major structures. 21 Process descriptions were derived including workers per shift, workers per 22 department, sources of raw materials, and raw material volume in number of 23 railroad cars received, tonnage of railroad cars from Libby and South Carolina, 24 and tonnage of unexpanded vermiculite received. 25 For each department a list ofjob titles and tasks. 26 27 28 Gaps in understanding were filled-in with information gathered from the focus groups, 29 specifically regarding: 30 31 32 Plant lay-out and changes over time, including engineering controls, 33 Historical pattern of job rotations within department from 1957 to 1980, 34 Time spent in work locations at the plant site, This document is a draftfor review purposes only and does not constitute Agency policy. F-3 DRAFT--DO NOT CITE OR QUOTE 1 Overtime associated with departments and season, 2 Use/nonuse of respirators. 3 4 5 F.3.2. Quantitative Data 6 Air sampling reports include quantitative measurement of airborne dust and fiber 7 concentration associated with a department job. These records were computerized following the 8 data entry scheme provided on June 1, 2009 and approved. Records were double entered and 9 verified. 10 Two identical Microsoft Access databases were created for initial and duplicate entry of 11 the quantitative data. Each individual performing data entry had a unique and separate database 12 to avoid possible data entry confusion. Variables to be entered have been previously provided. 13 A random 10% check of entered data was conducted throughout the data entry process to 14 maintain quality of data, to address data entry questions and to resolve potential database issues. 15 Data entry differences were below 5% throughout the entry process. 16 Each record was assigned a document and record identification (ID) number. The 17 document ID variable was based on data source. For example, if the data were provided by the 18 EPA from OM Scott then the EPA document ID was used. Data hardcopies from the EPA, 19 Department of Justice and 1980 University of Cincinnati (UC) data were each numbered starting 20 from 1. The document ID variable states EPA, DOJ or UC followed by the document number. 21 Record IDs were generated by using a unique identifier like a sample number for each document. 22 If a unique identifier was unable to be discerned then the entry personnel was instructed to 23 consecutively number each sample per document starting from one. 24 A final verification of data entry used SAS version 9.2 PROC COMPARE to import the 25 initial and duplicate Access tables. Discrepancies were below 5% as a result of the 10% random 26 checks throughout the entry process. All discrepancies were addressed by reviewing the original 27 document. The initial and duplicate Access databases were archived. A copy of the initial 28 database was converted to Microsoft Excel format for ease of standardization and analyses. 29 30 F.3.3. Process of Standardization 31 The standardization process included categorizing entered data into appropriate variable 32 fields, spell checking, identifying duplicate record entry from duplicate documents, merging This document is a draftfo r review purposes only and does not constitute Agency policy. F-4 DRAFT--DO NOT CITE OR QUOTE 1 records for the same sample or measurement, evaluating data for completeness, and categorizing 2 groups of data based on type of sample or measurement. 3 Data were reviewed and edited to ensure the information was entered into the appropriate 4 data field. A frequency of the data fields using SAS 9.2 PROC FREQ identified spelling 5 differences and patterns to ensure correct labeling of the data. Additional data variables were 6 created depending on recognized need to distinguish important pieces of data. 7 A new variable called group ID was created to identify, track, and consolidate partial 8 and/or complete duplicate data into one unique sample. Partial data were identified on a 9 combination of sample date, sample record ID, sample result, volume, sampling time and/or 10 document patterns. A document pattern would include instances where only a group of sample 11 results were available in one document and another document(s) would match the exact sequence 12 of sample results. 13 Data were further categorized based on the type of sample. Categories include dust 14 samples, bulk samples, personal and area fiber samples, limit of detection (LOD) or 15 quantification (LOQ) samples, off-site locations, and time weighted average samples. Some 16 samples were collected with a direct reading fibrous aerosol monitor, but these were not used as 17 there was no calibration information included in the records. Thus, only the fiber count data 18 collected with a sampling pump were used. In addition, group IDs lacking a sample result, 19 sample year or department were excluded. 20 Personal and area samples were plotted by year and department and found to be visually 21 similar. In addition the range, means, and standard deviations were approximately equal. 22 Therefore, personal and area sample data sets were merged and both utilized for the development 23 of the Exposure Matrix. Group IDs with only LOD or LOQ values were grouped by year and 24 categorized as trionize or background. In order to assign an estimate for the LOD or LOQ the 25 median value of each group was divided by two and assigned to all samples in that group. Given 26 the small number of LOD and LOQ samples (n = 35), it is unlikely any detectable bias was 27 introduced using this method. Time weighted average (TWA) values were not utilized when the 28 individual measurements that comprised the TWA were already available. 29 Sample analysis did not specify the type of fibers identified in the fiber counts. Counting 30 rules used included any fiber with the proper dimensions and not specifically Libby Amphibole 31 fibers. Attempts in other studies to convert from total dust to fiber count have relied on This document is a draftfo r review purposes only and does not constitute Agency policy. F-5 DRAFT--DO NOT CITE OR QUOTE 1 similarities in equipment or process where side-by-side samples were collected. We did not 2 identify any `pairs' of dust/fiber data from this plant. Moreover, fibers are a minor component of 3 the dust exposure, limiting an ability to find a relationship over time. Therefore, total dust 4 measurements were not converted to fiber counts and were not used as part of the fiber exposure 5 estimation. 6 7 F.3.4. Completeness and Trends in Measurements 8 From the paper records, we concluded that additional information would be helpful from 9 workers in order to obtain descriptions of work organization and practices. Focus groups 10 discussions were conducted with long-term OM Scott workers (n = 15) in 2010. These focus 11 groups provided valuable qualitative data in order to fill gaps regarding work plant operations, 12 especially during the earlier years. 13 As described earlier, the data used for exposure reconstruction was obtained from three 14 sources: UC archived records (reported previously by Lockey et al. (1984)). information 15 obtained by the EPA from the company, and from the DOJ documents. Table F-l shows that a 16 total of 914 IH fiber measurements were available for this analysis. Of this total, only 180 17 (19.6%) of the IH fiber measurements were available from the UC archived records. The yearly 18 number of samples collected was not uniform. As shown in Table F-2, the first fiber count 19 measurements were available in 1972 and the last in 1994. About 26% of the samples were 20 collected in 1978. Focus group participants reported working in the summer. Summer activities, 21 however, involved fewer work hours and included clean-up and repair activities in addition to 22 production. Since less than 6% of the fiber samples were collected during the summer months, 23 no seasonal trend analysis was possible. 24 25 F.4. DECISIONS RELEVANT TO THE EXPOSURE MATRIX 26 F.4.1. General Issues 27 A graphical display of fiber count results indicated that all samples in various trionizing 28 jobs generally followed the same pattern: higher in the early years of industrial hygiene This document is a draftfor re\'iew purposes only and does not constitute Agency policy. F-6 DRAFT--DO NOT CITE OR QUOTE 1 Table F-l. Industrial hygiene fiber measurements by document source 2 Document source Trionize Background Total (%) DOJ 38 0 38 (4.16) EPA 398 122 520 (56.89) UC 135 45 180(19.69) COMBINED 172 4 176(19.26) Total (%) 743 (81.29) 171 (18.71) 914 3 4 5 sampling, and declining gradually over time. Further, from the focus groups, we learned that no 6 one, single engineering change resulted in a dramatic reduction in the perception of dustiness in 7 the plant. Thus, the workers' recollections supported the findings from the industrial hygiene 8 data demonstrating a gradual decline in levels of exposure rather than a dramatic step-wise drop 9 due to any one engineering change. 10 Changes in work practices such as the use of compressed air and brooms for clean-up 11 versus the use of wet vacuuming may result in marked decreases in exposure. We discussed 12 work practices in the focus groups, and no remarkable changes were documented. Participants 13 did note that during some years, sampling practices included leaving pumps in control rooms 14 during high-dust activities. High-dust activities included the use of compressed air to remove 15 particulate from surface areas. We did not find any documentation that high exposure work was 16 excluded from the sampling effort in the industrial hygiene reports. In fact, in the early years, 17 some activities recorded in the sampling record included reference to compressed air "blow 18 down", one of the activities associated with potentially high exposures. Consequently, no 19 adjustment was made for any potentially unsampled periods from 1972 through 1994 when 20 industrial hygiene measurements were available. 21 Per the focus groups, workers reported very sporadic usage of respirators due to heat and 22 discomfort. Because of the heat, the workers preferred paper masks, and reported reusing them 23 from day to day. There was no documentation of fit-testing of the paper masks. Paper masks 24 may provide some protection against the larger particles, but likely provided little reduction in 25 respirable particles, particularly when reused. Therefore, no adjustment was made to lower the 26 exposure estimates due to respirator use. This document is a draftfor review purposes only and does not constitute Agency policy. F-7 DRAFT--DO NOT CITE OR QUOTE Table F-2. Industrial hygiene fiber measurements by department and year This document is a draftfor review purposes only and does not constitute Agency policy. Dept. Background Trionize Total (Year %) 1972 1973 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1993 1994 Total (Dept. %) 3 0 2 0 10 54 2 0 12 7 3 11 5 23 13 16 0 10 171 (18.71) 9 40 20 115 68 183 26 23 38 24 8 27 14 52 33 31 3 29 743 (81.29) 12 40 22 115 78 237 28 23 50 31 11 38 19 75 46 47 3 39 914 (1.31) (4.38) (2.41) (12.58) (8.53) (25.93) (3.06) (2.52) (5.47) (3.39) (1.20) (4.16) (2.08) (8.21) (5.03) (5.14) (0.33) (4.27) (100.00) Dept. = department. F-8 DRAFT--DO NOT CITE OR QUOTE 1 F.4.2. Vermiculite Raw M aterial Sources 2 Libby vermiculite usage ended in 1980 per shipping records obtained from B. Benson 3 and an Agency for Toxic Substances and Disease Registry (ATSDR) report (Benson. 2009; 4 ATSDR. 20051. Post 1980 usage included African/Virginia/South Carolina vermiculite until 5 2000. In 2000, corn cobs were introduced as an inert carrier of lawn care chemicals, and 6 vermiculite usage ended. There were two primary sources of information regarding vermiculite 7 sources: 8 9 10 An internal UC document from the 1980 study with estimates of railroad car loads 11 delivered to the plant per year. Documents indicate railroad cars from Libby were 12 100 ton cars and from South Carolina 70 ton cars. 13 The Chamberlain memo provides information regarding vermiculite sources for 14 1964-1972 in railroad car loads per year. 15 16 17 Per the UC document, 100% South Carolina vermiculite was estimated to be used from 18 1957-1960. Per the Chamberlain memo, Libby vermiculite began arriving in 1960. Focus 19 groups placed it earlier, in 1958 or 1959. We believe there is sufficient evidence to support a 20 1959 start date for Libby vermiculite with 1957 and 1958 assumed to be 100% South Carolina 21 vermiculite 22 Documentation was found from the original 1980 UC documents indicating an estimated 23 Libby tonnage contribution of 32% from 1959-1963. These percentages for 1959-1963 were 24 adopted for use in this project. After adjusting for the difference in rail car sizes, the 25 Chamberlain memo indicates that Libby tonnage usage increased from 57% in 1964 to 73% in 26 1965 to 92% in 1966. Table F-3 illustrates the distribution of unexpanded vermiculite sources 27 received at the plant between 1957 and 1971. From 1959 until 1971 fiber level adjustments were 28 made based on the percent Libby versus South Carolina vermiculite tonnage received at the 29 plant. The estimates were derived from 1972 when the earliest industrial hygiene samples were 30 available and 93% of the vermiculite was Libby. 31 To devel op the relati onship of fiber levels between South Carolina and Libby 32 vermiculite, samples that recorded a 100% of either source for vermiculite were identified. Two 33 jobs with a higher number of samples from the same year from each source were used to This document is a draftfor review purposes only and does not constitute Agency policy. F-9 DRAFT--DO NOT CITE OR QUOTE 1 Table F-3. Tonnage by year and vermiculite source 2 Year % Tonnage Libby % Tonnage SC Comment 1957 100 No confirmation of Libby usage 1958 100 No confirmation of Libby usage 1959 32 68 Libby usage began per focus groups; Chamberlain says 1960 1960 32 68 Chamberlain memo and 1980 chart 1961 32 68 Chamberlain memo and 1980 chart 1962 32 68 Chamberlain memo and 1980 chart 1963 32 68 Chamberlain memo and 1980 chart 1964 57 43 Chamberlain memo 1965 73 27 Chamberlain memo 1966 92 8 Chamberlain memo 1967 87 13 Chamberlain memo 1968 79 21 Chamberlain memo 1969 82 18 Chamberlain memo 1970 90 10 Chamberlain memo 1971 95 5 Chamberlain memo 3 4 5 establish the relationship: track-unload for 1977 and expander for 1978. The samples used 6 included 22 Libby track-unload, 8 Libby expander, 17 South Carolina track-unload, and 7 South 7 Carolina expander. A weighted average of these samples generated a 10:1 fiber count ratio for 8 Libby: South Carolina vermiculite. This ratio was used for estimating the proportion of Libby 9 versus South Carolina fiber exposure levels from 1959 to 1971. From 1972 and beyond, IH 10 measurements were available and no adjustment in the IH data was made based on vermiculite 11 source. Tonnage records demonstrate that Libby was the primary source of vermiculite from 12 1972 until 1979, supplemented by African vermiculite, and that Libby vermiculite usage ended 13 in 1980. 14 The 100% Libby samples were compared to samples labeled as 50% Libby. The 15 resultant measurements were accordingly lower, demonstrating internal consistency within the 16 data. This document is a draftfor review purposes only and does not constitute Agency policy. F-10 DRAFT--DO NOT CITE OR QUOTE 1 Assessment of exposure in 1977 during application of the final, expanded product that 2 included a mix of South African and Libby vermiculite showed no fibers. Therefore, fiber 3 exposure estimation was restricted to jobs in the plant areas where expanding was conducted. 4 5 F.4.3. Exposure Estimates by Time Period for the Trionizing Department 6 For this project, exposures of interest were from 1957 through 2000. Exposure 7 measurements in the plant where vermiculite was used were initiated in 1972. For prior years, it 8 was necessary to estimate exposure from the measurements collected in 1972 and later and with 9 supporting qualitative information. Important changes occurred in production due to increasing 10 use of engineering controls to reduce airborne particulate. In addition, the source of vermiculite 11 changed over the years. Therefore, the exposure estimation process was divided into two efforts: 12 1972 and later when industrial hygiene measurements were available; and 1957 to 1971, when 13 no industrial hygiene measurements were available. The exposure estimation process is 14 described below, first for Trionizing where vermiculite was expanded and then for other 15 departments where either no or expanded vermiculite was used. 16 17 F.4.3.1. Trionizing Department Exposure Estimation >1972--2000 18 For the years with exposure measurements, fiber exposure level was estimated from the 19 measurement data. This was done by department. 20 21 F.4.3.1.1. Trionizing department 22 The trionizing department included jobs from the entry of vermiculite into the plant, 23 through final product. These were: track at raw material entry and production jobs of 24 screen/mill, dryer, expander, blender, resin, and clean-up, Workers rotated through the various 25 jobs within the department. Overall rotation among jobs reported in the 1980 Lockey et al. study 26 was verified by the focus groups. 27 Plots of the measurements over time were made for individual trionizing jobs. Based on 28 these plots, it was determined that all industrial hygiene sample results from the various 29 trionizing production jobs (screen/mill through clean-up) followed the same general distribution 30 and should be combined. The track job included two very different work activities: unloading 31 rail cars containing vermiculite (track unload) and general track work such as bringing in the rail This document is a draftfor review purposes only and does not constitute Agency policy. F-l 1 DRAFT--DO NOT CITE OR QUOTE 1 cars, and monitoring discharge (track other). The two track job activities (unload and other) had 2 a substantially larger range of sampling results and were treated separately. 3 The following steps were followed: 4 5 6 1. The data were log-transformed. 7 2. For all exposure values for the combined trionizing jobs from 1972-1979, a curve 8 was drawn connecting the mean values of years having at least 40 exposure 9 measurements (1973, 1976, and 1978). This criteria was chosen to assure that 10 stable means were used to define the curve over this time period. For each year, 11 the annual exposure estimate was determined by exponentiation of the value from 12 the curve. The sharp decline seen in exposures throughout this time period 13 parallels the addition of engineering controls including dust collection, enclosing 14 vibrating conveyors, adding ventilators, erecting a wall between track and 15 trionizing, and sealing leaks in the system. As values for 1980-1994 were similar 16 and near the level of detection, the mean value for all the samples was used and 17 then extended until 2000. 18 3. The measurement results for track unload and track other were plotted and a 19 straight line produced to best fit the data points. An estimate of exposure at each 20 year was determined by exponentiation of the value on the line for that year. 21 4. For the trionizing department, it was estimated that 11% of work time was spent 22 in track and 89% in all other jobs. This is consistent with the previous weights 23 used in the 1980 Lockey study and confirmed by the focus group. 24 5. The Focus groups reported that when working track, track unload required about 25 25% of the time and track other comprised about 75% of the track job time. 26 Therefore, a weighted average for exposure at track within the trionizing 27 department was derived. This 25% time estimate for track unload is higher than 28 that previously published (Lockey et ak, 1984). 29 30 31 Figure F-l illustrates on a log scale a fitted line of all usable industrial hygiene 32 measurements across all jobs (except track) within the trionizing department. This document is a draftfor w ife * purposes only and does not constitute Agency policy. F -l2 DRAFT--DO NOT CITE OR QUOTE ISll 1952 19 $ 4 4 3i 1- 19|724 iN - < .^ 1982 1987 4 4 1 4 4 <> V fl 4> <: < | I> < 4 :4 ; 4 4 4 4 4 < 4 4 1992 19il 4 TrionizingexcludingTrack A YearlyGeometricMeans I Log of fiber count (fibers/cc) 2 3 Figure F-l. Illustrates on a log scale a fitted line of all usable industrial 4 hygiene measurements across all jobs (except track) within the trionizing 5 department. 6 7 8 F.4.3.2. Trionizing Department Exposure Estimation 195 7-19 71 9 There are no industrial hygiene measurements available prior to 1972. Engineering 10 changes did not result in "step-function" decreases in exposures based on focus group reports. 11 Rather a more gradual decline in exposure occurred beginning with improvements in 1968, when 12 two dust collectors were added. Focus group workers report that dust exposures in trionizing 13 were at least two times higher in the 1960's. Track jobs, however, were outdoors and likely 14 unaffected by plant engineering controls. Flence, estimates for fiber exposure levels for track 15 duties were adjusted by type of vermiculite only. 16 For trionizing employees, excluding outdoor track duties, the estimate from the focus 17 group of "twice as high" was generated beginning from 1972 and increasing until 1967. The 18 year 1972 was used as the start of the "gradual" retrospective increase in exposure back to 1967 This document is a draftfo r review purposes only and does not constitute Agency policy. F-l3 DRAFT--DO NOT CITE OR QUOTE 1 as 1972 was the first year when industrial hygiene measurements were available, and the percent 2 Libby vermiculite utilized was 93%. The year 1967 was selected as this was the year preceding 3 engineering controls. A line was drawn to connect these two points and then the adjustment was 4 made for the percent yearly Libby and South Carolina vermiculite utilized from 1967 through 5 1971. Prior to 1967, exposure was extended backward in time, assuming no change from the 6 1967 value except for a yearly adjustment for percent Libby and South Carolina usage. As 7 described above and shown in Figure F-l, after 1980 when Libby vermiculite was no longer used 8 and major environmental controls had been implemented, fiber exposure levels remained near 9 the level of detection (0.01) through the last available industrial hygiene information in 1994. 10 The levels were estimated to be the same from 1994 forward until 2000 when vermiculite was no 11 longer used. 12 13 F.4.4. Exposure Estimates for Nontrionizing Departments 14 Departments using only expanded vermiculite or no vermiculite were defined as having 15 "plant background" exposure. These included the departments of polyform, plant maintenance, 16 office, research, pilot plant, warehouse, central maintenance, and packaging. This decision was 17 based on plots of available sampling data showing similar levels, and qualitative reports 18 documenting that there were not fibers in the finished product. Plant background prior to 1972 19 was calculated using similar methodology as for trionizing. Although the background level was 20 not affected by engineering control as in trionizing, exposures would be affected by the percent 21 of Libby vermiculite used. Therefore, for the years prior to 1972, the measured plant 22 background rate in 1972 of 0.02 was adjusted by the yearly percent Libby vermiculite utilized. 23 The two years prior to Libby vermiculite usage, 1956 and 1957, were assigned level of 24 detection (0.01). This is in line with industrial hygiene measurements post Libby vermiculite 25 usage through 1994. 26 Polyform began in 1969, and no unexpanded vermiculite was used there. The 27 background exposure level was used for any time in Polyform. 28 Plant Maintenance--Although there were some differences of opinion in the focus group 29 regarding where plant maintenance spent their time, the consensus reached was to assign 30 approximately 50% of time in trionizing and 50% in areas defined as plant background for their 31 work in shop and other departments. This document is a draftfo r review purposes only and does not constitute Agency policy. F-l4 DRAFT--DO NOT CITE OR QUOTE 1 Office--Assigned plant background. 2 Research--Assigned plant background. 3 Pilot plant--Per the focus group participants, the pilot plant did not have its own 4 expander, and used only expanded vermiculite in test and run simulations. Plant 5 background levels were thus assigned to the pilot plant. 6 Warehouse--Only expanded vermiculite was in this area. Although bags did 7 break, the exposure was to final product, not unexpanded vermiculite. 8 Central Maintenance--According to the focus group, these employees worked 9 outside of trionizing for about 90% time (background) and 10% (trionizing) for 10 installation of new equipment/parts. Around 1982 central maintenance 11 department was discontinued, and the work was contracted to outside personnel. 12 Packaging--Assigned plant background. 13 14 15 Table F-4 illustrates the fiber exposure matrix from 1957 to 2000 using this methodology. 16 17 F.4.5. Decisions Related to Break Periods and Hours Worked 18 Cumulative exposure is the product over time of the level of exposure and duration. 19 Level of exposure is derived from the exposure matrix and duration from the work history. 20 However, in this workforce, work time is complicated by breaks where exposure is at a lower 21 level and seasonal changes resulting in extra hours worked beyond the usual 40 hour week. Each 22 of these factors is described below: 23 According to the focus group data there was approximately a 30-minute break for lunch 24 and two 15-minute breaks during the day. Therefore, every worker was considered to have at 25 least one hour of background exposure daily. There was no documentation that a third 15minute 26 break was provided when working longer than 8 hours in a day. 27 Employees in some departments frequently worked extra hours each day, and weekends 28 as well, depending on the production needs and season. Decisions regarding this work 29 organization are summarized below: 30 31 This document is a draftfor review purposes only and does not constitute Agency policy. F-l 5 DRAFT--DO NOT CITE OR QUOTE Table F-4. Exposure matrix assuming doubling of fiber levels from 1972 to 1967 but with adjustment for vermiculite source from 1957-1971 This document is a draftfo r re\'iew purposes only and does not constitute A gency policy. Department Trionizing Plant maint (50/50) Central maint (90/10) Background8 Trionizing Plant maint (50/50) Central maint (90/10) Background8 1957 0.729 0.369 0.082 0.010 1972 3.674 1.847 0.385 0.020 1958 0.729 0.369 0.082 0.010 1973 3.007 1.513 0.319 0.020 1959 2.825 1.416 0.289 0.008 1974 2.464 1.242 0.264 0.020 1960 2.825 1.416 0.289 0.008 1975 2.019 1.020 0.220 0.020 1961 2.825 1.416 0.289 0.008 1976 1.391 0.705 0.157 0.020 1962 2.825 1.416 0.289 0.008 1977 0.150 0.090 0.030 0.020 1963 2.825 1.416 0.289 0.008 1978 0.086 0.053 0.027 0.020 1964 4.462 2.237 0.457 0.012 1979 0.077 0.044 0.017 0.010 1965 5.510 2.763 0.565 0.015 1980 0.063 0.036 0.015 0.010 1966 6.755 3.387 0.692 0.019 1981 0.063 0.036 0.015 0.010 1967 6.427 3.222 0.659 0.018 1982 0.060 0.035 0.015 0.010 1968 5.542 2.779 0.569 0.016 1983 0.060 0.035 0.015 0.010 1969 5.279 2.648 0.543 0.017 1984 0.055 0.032 0.010 1970 4.923 2.470 0.509 0.018 1985 0.055 0.032 0.010 1971 4.316 2.168 0.449 0.019 1986-2000 0.052 0.031 0.010 background applies to Pilot Plant, Research, Polyform, Office, Packaging, Warehouse. 16 DRAFT--DO NOT CITE OR QUOTE 1 1. Extra hours--Were defined as hours worked in excess of 8 hours per day. 2 2. Four departments worked no extra hours--office, pilot plant, research, and central 3 maintenance. According to focus group data, the only departments that worked 4 extra hours outside of their own department were trionizing and polyform. Thus, 5 a decision was needed as to how to appropriate the amount of overtime spent 6 outside trionizing and polyform. 7 3. Extra hours for polyform workers--According to the focus groups, polyform 8 workers first worked in their own department, and went to trionizing to work 9 extra hours. According to workers, about 75% of the daily overtime was in their 10 own department. Therefore, for each 4 hours worked beyond the normal 8 hour 11 day, it is estimated that they spent 3 hours in polyform and 1 in trionizing. This 12 rule was not applied to 8-hour weekend days worked. 13 4. Extra hours for trionizing workers--As for polyform workers, above, it is 14 estimated that trionizing workers spent three hours in trionizing and one hour in 15 polyform as a daily average. 16 17 18 Schedules by season differed due to production rate: 19 20 21 For trionizing, plant maintenance, polyform, warehouse, and packaging the spring 22 schedule was from January through May--7 days at 12 hours. 23 For trionizing, plant maintenance, polyform, warehouse, and packaging the 24 summer schedule was from June through August--5 days at 8 hours. Due to the 25 difficulty that heat and humidity brought to the process, polyform was shut down 26 during summer. During the summer, polyform workers did outside jobs. As 27 these jobs have the same exposure level as polyform (background rate), no 28 adjustment was made for the summer polyform shutdowns. The trionizing 29 department more typically slowed down production in the summer, and this is 30 reflected in the number of hours worked from June through August. 31 For trionizing, plant maintenance, polyform, warehouse, and packaging the fall 32 schedule was from September through December--5 days at 12 hours and 33 2 weekend days at 8 hours. 34 35 36 In light of these extra hours, exposure values by department and season were modified 37 for use in the cumulative equivalent human equivalent exposure concentration estimations. 38 This document is a draftfor review purposes only and does not constitute Agency policy. F-17 DRAFT--DO NOT CITE OR QUOTE 1 F.5. DEVELOPMENT OF A CUMULATIVE HUMAN EQUIVALENT EXPOSURE 2 CONCENTRATION 3 An EPA adjustment of cumulative occupational exposure to fibers to continuous human 4 exposure to fibers (24 hours/day; 7 days/week) was provided by B. Benson. This adjustment 5 was accepted as provided for the development of a cumulative human equivalent exposure 6 concentration (CHEEC) for the Marysville, OH occupational cohort. 7 8 F.5.1. Seasonal Schedule Correction Factor 9 For this project the Correction Factor was adjusted for the specific information on work 10 schedules related to the seasonal changes to meet production demands as described above in 11 Section 4.4. UC applied these correction factors supplied by the EPA (B. Benson) to the work 12 history data obtained by UC during 1980 and updated in 2004. 13 14 F.5.2. Decision Rules to address Department Changes Occurring Within Seasons 15 Decision rules were implemented to systematically standardize each worker's 16 occupational history to a format that corresponded directly with the seasonal changes that 17 occurred at the plant. Previous decisions related to department exposure levels and seasonal 18 work resulted in six unique exposure categories: trionizing, plant maintenance, central 19 maintenance, polyform, background (office, research, pilot plant), and background with extra 20 time (warehouse, packaging). The date of any job change by a worker between these six 21 categories was adjusted so the change occurred at the starting month for the nearest season. 22 23 F.5.3. Development of CHEEC 24 In preparation for creating the CHEEC, the exposure matrix was converted to a seasonal 25 (spring, summer, fall) exposure value. This value is the estimate of the amount of exposure 26 occurring by department for each season of each year. With the worker's occupational histories 27 standardized to the same seasons, the CHEEC for each worker was then calculated as the sum of 28 exposure values for all seasons worked between 1957-2000. The correction factors used in 29 derivation of the CHEEC are outlined below. 30 31 32 This document is a draftfor review purposes only and does not constitute Agency policy. F-l 8 DRAFT--DO NOT CITE OR QUOTE 1 General Procedure 2 (Cumulative F ib e r s ) o c c u p x Correction Factor = (Cumulative F ib e rs)n E C - 3 OCCUP = Occupational Exposure. 4 HEC = Human Equivalent Concentration for exposure of 24 hours/day, 7 5 days/week. 6 The Correction Factor usually used with an occupational study is 7 5 days = 7 days x 10 m3 ^ 20 m3. 8 9 10 UC Procedure 11 CHEEC= (Exposure Est year-dept-season i x Correction FactorseaSon 1 12 x Seasonal Duration Factor) + (Exposure Est year-dept-season 2 13 x Correction Factorseason2 x Seasonal Duration Factor) 14 + ... (Exposure Est year-dept-seasonx x Correction Factorseason 15 x Seasonal Duration Factor). 16 17 Where the Seasonal Duration Factor for the Spring is 5/12 year; the Summer is 3/12 year; 18 the Fall is 4/12 year. 19 20 F.5.3.1. Detailed Calculations Follow 21 F.5.3.1.1. Work schedulefor trionizing, plant maintenance, polyform, warehouse, and 22 packaging 23 F.5.3.1.1.1. Spring 24 January 1 to May 31:7 days/week, 12 hours/day, with New Years' Day off, and 25 accounting for leap years: 26 27 28 151.25 - 1 = 150.25 days 29 Breathing rate, working = 1.25 m3/hour x 12 hours = 15 m3 30 Breathing rate, not working = 0.625 m /hour x 12 hours = 7.5 m 31 Total breathing rate =15 + 7.5 = 22.5 m3/day 32 Correction Factor Spring = 150.25 = 151.25 x 15 = 22.5 = 0.662259 33 This document is a draftfor review purposes only and does not constitute Agency policy. F-19 DRAFT--DO NOT CITE OR QUOTE 1 F.5.3.1.1.2. Summer 2 June 1 to August 31:5 days/week, 8 hours/day, 2 week summer vacation: 3 4 5 (9 2 - 14) x 5 - 7 = 55.714286 days 6 Breathing rate, working = 1.25 m3/hour x 8 hours = 10 m3 7 Breathing rate, not working = 0.625 m /hour x 16 hours = 10 m 8 Total breathing rate = 10 + 10 = 20 m3/day 9 Correction Factor Summer = 55.714286 = 92 x 10 = 20 = 0.302795 10 11 12 F.5.3.1.1.3. Fall 13 September 1 to December 31: 5 days/week, 12 hours/day and 2 days/week, 8 hours/day, 14 with Christmas Day off: 15 16 17 122-1 = 121 days 18 Breathing rate, working, 12 hour day = 1.25 m3/hour x 12 hours = 15 m3 19 Breathing rate, working, 8 hour day = 1.25 m3/hour x 8 hours = 10 m3 20 Breathing rate, not working = 0.625 m3/hour x 16 hours = 10 m3 21 Total breathing rate, 12 hour work day =15 + 7.5 = 22.5m /day 22 Total breathing rate, 8 hour work day =10 + 10 = 20 m /day 23 Correction Factor Fall = 121 = 122 x (86.42857 x 15 + 22.5 + 34.57143 x 24 10 = 20) = 121 =0.613973 25 26 27 F.5.3.1.2. Work schedule for office, pilot plant, research, and central maintenance 28 No extra days or extra hours. 29 30 F.5.3.1.2.1. Spring 31 January 1 to May 31:5 days/week, 8 hours/day, with New Years' Day off, and 32 accounting for leap years. 33 34 This document is a draftfor review purposes only and does not constitute Agency policy. F-20 DRAFT--DO NOT CITE OR QUOTE 1 (151.25 - 1) x 5 days - 7 days = 107.321429 2 Breathing rate, working = 1.25 m /hour x 8 hours = 10 nr 3 Breathing rate, not working = 0.625 nr/hour x 16 hours = 10 m' 4 Total breathing rate = 10 + 10 = 20 m '/day 5 Correction Factor Spring = 107.321429 = 151.25 x 1 0 -2 0 = 0.354782 6 7 8 F.5.3.1.2.2. Summer 9 June 1 to August 31:5 days/week, 8 hours/day, 2 week summer vacation. 10 11 12 (92 - 14) x 5 = 7 = 55.714286 days 13 Breathing rate, working = 1.25 m /hour x 8 hours = 10 nr 14 Breathing rate, not working = 0.625 m' /hour x 16 hours = 10 m 15 Total breathing rate = 10 + 10 = 20 mVday 16 Correction Factor Summer = 55.714286 = 92 SK10 = 20 = 0.302795 17 18 19 F.5.3.1.2.3. Fall 20 September 1 to December 31: 5 days/week, 8 hours/day, with Christmas Day off. 21 22 23 (1 2 2 - l) x 5 - 1 = 86.428571 days 24 Breathing rate, working, 8 hour day = 1.25 m /hour * 8 hours = 10 m 25 Breathing rate, not working = 0.625 nr/hour x 16 hours = 10 nr 26 Total breathing rate = 10 + 10 = 20 m '/day 27 Correction Factor Fall = 86.428571 - 122 x 10 = 20 = 0.354215 28 29 30 F.5.4. Results of the Cumulative Human Equivalent Exposure Concentration (CHEEC) 31 To verify the accuracy of the CHEEC calculations, several quality control checks were 32 conducted. The distribution was evaluated by reviewing the mean, median, standard deviation, 33 highest 10 values, and lowest 10 values. Several workers were also randomly selected and their 34 values hand-calculated to ensure all programming was correct. Tables 5-7 provide a list of all 35 280 subjects participating in the 2004 Marysville health update (Rohs et aL 2008). These tables This document is a draftfo r review purposes only and does not constitute Agency policy. F-21 DRAFT--DO NOT CITE OR QUOTE 1 describe each subject's identification number, job start and stop date, date of radiograph, age, 2 gender, body mass index, smoking history, asbestos exposures, health outcomes, and the 3 cumulative human equivalent exposure concentration (CHEEC) for all departmental exposures 4 they reported while employed at the OM Scott Marysville, OH plant. 5 6 F.6. STRENGTHS AND LIMITATIONS 7 There are major strengths in this exposure reconstruction project: 8 9 10 1. Data were gathered from court records, federal sources and archived files, totaling 11 over 3,000 pages. These data were reviewed and both qualitative and quantitative 12 data were abstracted to aid in this reconstruction. 13 2. Approximately five times more fiber measurements became available than had 14 been used in the original studies. 15 3. Two focus groups were conducted in 2010 with long term workers who provided 16 input regarding exposure and production process changes. 17 4. There were sufficient data available to examine exposure intensity over time for 18 jobs within the trionizing department as well as for other departments. These data 19 enhanced exposure estimates for all departments from 1972 to 1994. 20 5. Industrial hygiene data were available allowing for comparisons of fiber counts 21 when 100% Libby or 100% South Carolina vermiculite was used in order to 22 calculate a ratio of fibers in each. 23 6. There were data available from archived records, Scott memos, and worker 24 information that allowed for exposure estimates to be adjusted for type of 25 vermiculite used from 1957 until 1971 when no industrial hygiene data were 26 available. 27 7. Worker report data were available that provided documentation for increased 28 dustiness before industrial hygiene data were available, compared with years 29 when measurements were available. 30 8. Based on past and current data gathered in the focus group, exposures were 31 adjusted to account for seasonal work schedules by departments. 32 9. All decisions based on level of exposure by year were data driven. 33 34 35 The limitations for this project are also recognized: This document is a draftfor review purposes only and does not constitute Agency policy. F-22 DRAFT--DO NOT CITE OR QUOTE 1 1. The exposure metric used (fibers/cc) results from an analytical method that is a 2 count of fibers (defined as any viewed elongated particle in excess of 5 pm in 3 length and with a length to width ratio of 3:1) collected on a filter and viewed at 4 400* with light microscopy. The composition of the fiber is not known. Also, a 5 fiber with diameter less than a limit of resoluti on of 0.2 pm cannot be viewed 6 with this method. 7 2. It is unknown if other sampling results exist. If any are found in the future, these 8 can be incorporated into a future exposure assessment. 9 3. Some dusty activities may not have been sampled or rarely sampled e.g., summer 10 cleanup. We have no way of estimating the effect of these activities on overall 11 exposure estimates. 12 4. We did not reduce exposure estimates due to possible use of respiratory 13 protection. Substantially more documentation regarding enforced usage, fit 14 testing and cleaning/storage protocols would be needed for meaningful reduction 15 in exposure estimates. 16 5. By combining all individual trionizing job duties into one department exposure, 17 the nonexpander trionizing exposure estimates may have been overestimated as 18 there were more expander measurements, and these were somewhat higher than 19 for other job duties. 20 6. From 1980 forward, Libby vermiculite was not used. Thus for any individual 21 year during this period, exposure from a qualitative and quantitative perspective 22 does not reflect Libby Amphibole exposure. 23 7. Seasonal work schedule adjustments were based on recall of focus group 24 participants and may over or under estimate true durations and location of 25 additional work hours. 26 27 28 F.7. REFERENCES 29 ATSDR (Agency for Toxic Substances and Disease Registry). (2005). Health Consultation: Fonner O.C. Scott and 30 Sons Company, 14111 Scottslawn Road, Marysville, Union County, Ohio. EPA Facility ID: 31 OHD990834483. (PB2008-104070). Atlanta, GA. 32 Benson, R, (2009). Use of Libby Amphibole in Marysville, Ohio. Denver, CO: U.S. Environmental Protection 33 Agency. 34 Lockev. JE: Brooks, SM: Jarabek, AM: Khourv. PR: McKay, RT: Carson. A: Morrison. JA: Wiot, JF: Spitz, HB, 35 (1984). Pulmonary changes after exposure to venniculite contaminated with fibrous tremolite. Am Rev 36 RespirDis 129: 952-958. 37 Rohs. A: Lockev. J: Dunning, K: Shukla. R: Fan, H; Hilbert. T: Borton. E: Wiot. J: Mever, C: Shipley, R: 38 Lemasters. G: Kapil, V. (2008). Low-level fiber-induced radiographic changes caused by Libby 39 vermiculite: A 25-year follow-up study. Am J Respir Crit Care Med 177: 630-637. 40 http://dx.doi.org/10.1164/rccm.200706-8410C. This document is a draftfo r review purposes only and does not constitute Agency policy. F-23 DRAFT--DO NOT CITE OR QUOTE 1 APPENDIX G. EXTRA RISK AND UNIT RISK CALCULATION 2 G.l. MESOTHELIOMA MORTALITY 3 The increased risk of mesothelioma mortality attributable to continuous fiber exposure 4 was estimated using a life-table procedure based on the general U.S. population. The life-table 5 procedure involved the application of the estimated Libby Amphibole asbestos'-specific toxicity 6 to a structured representation of the general U.S. population in such a manner as to yield 7 age-specific risk estimates for mesothelioma mortality in the absence and presence of exposure 8 to Libby Amphibole asbestos. Baseline all-cause mortality rates were included in the life-table 9 in such a way as to enable computation of the specific absolute risk of mesothelioma mortality 10 while accounting for other competing causes of mortality. For each age-interval in the life-table, 11 the effect estimates of the Poisson regression model analysis (the absolute risk) were used to 12 estimate mesothelioma mortality at a particular exposure level. These age-specific absolute risks 13 can then be summed over a lifetime. Different exposure levels are evaluated to ascertain what 14 magnitude of exposure would be expected to produce 1% absolute risk of mesothelioma 15 mortality. By this method, the exposure-response relationship determined in the Libby worker 16 cohort is used to estimate mesothelioma mortality in the general U.S. population that would be 17 expected from continuous lifetime environmental exposure to various concentrations of Libby 18 Amphibole asbestos. 19 Assuming no background risk for mesothelioma, extra risk is the same as absolute risk. 20 Absolute risk estimates were calculated using the effect estimates derived from the modeling of 21 the mesothelioma mortality risk and a life-table analysis program that accounts for competing 22 causes of death. The unit risk of mesothelioma is computed using the 95% upper bound to 23 estimate an upper bound for extra risk of mesothelioma due to Libby Amphibole asbestos 24 exposure. The upper bound calculation is specific to the exposure metric parameters; the effect 'The term "Libby Amphibole asbestos" is used in this document to identify the mixture o f amphibole mineral fibers o f varying elemental composition (e.g., winchite, richterite, tremolite, etc.) that have been identified in the Rainy Creek complex near Libby. MT. It is further described in Section 2.2. 2This program is an adaptation o f the approach previously used by the Committee on the Biological Effects of Ionizing Radiation (BEIR, 1988). Compared to life-table methods based on full life exposures from birth, the method used here yielded unit risk differences between full life exposure to scaled adult-only exposure between -3% to -2% for the mesothelioma mortality unit risks for the tun mesothelioma models (see Tables G -l and G-2). A spreadsheet containing the extra risk calculation for the derivation o f the LEC0i for mesothelioma mortality is presented in Tables G -l and G-2. This document is a draftfo r re\'iew purposes only and does not constitute utgencv policy. G-l DRAFT--DO NOT CITE OR QUOTE 1 of metric uncertainty in these values is discussed in Section 5.4.5.3. Because this human health 2 assessment derived a combined inhalation unit risk (IUR) for both mesothelioma and lung cancer 3 mortality, an interim value based on the central effect estimate (rather than the upper bound) is 4 also computed to avoid statistical concerns regarding the combination of upper bounds. Details 5 are shown in Section 5.4.5.3. This current assessment does not directly apply life-table 6 calculations to estimate partial lifetime risk scenarios; the use of the IUR for partial lifetime 7 extrapolations is discussed in Section 5.4.5.4. 8 U.S. age-specific all-cause mortality rates from the 2010 National Vital Statistics Report 9 (NVSR) for deaths in 2007 among all race and gender groups combined (Xu et aU 2010) were 10 used to specify the all-cause background mortality rates (R0) in the life-table analysis. The risk 11 with exposure (Rr) was computed up to age 85 years, assuming continuous environmental 12 exposure to Libby Amphibole asbestos. Conversions between occupational Libby Amphibole 13 asbestos exposures and continuous environmental asbestos exposures were made to account only 14 for differences in the amount of air inhaled per day during a higher effort occupational shift 15 (8 hours; 10 m ) compared to a standard 24-hour (20 m ) day (U.S. EPA. 1994) because results 16 were already based on a 365-day calendar year. The computation of the unit risk involved three 17 steps. The first step was to compute the unit risk for adults. This was achieved by initiating 18 exposure at age 16 years and maintaining continuous exposure throughout the remainder of life 19 while allowing for the incremental mathematical decay of previously accumulated exposure.4 20 An age of 16 years was used because it roughly matched the youngest age of a worker in the 21 subcohort and was consistent with the application of a similar life-table methodology when the 22 age-dependent adjustment factors (ADAFs) are applied; however, the application of age- 23 dependent adjustment factors was not recommended in this case (see Section 4.6.2.2). An 24 adjustment was also made in the life-table for the lag period, so that the age-specific risk 25 calculation s began at 16+ (the length of the lag peri od) years of age. The standard assumption 26 used by the U.S. Environmental Protection Agency (EPA) is that the average lifetime spans 27 70 years. Because the adult-only-exposure unit risk excluded the first 16 years, the 28 adult-only-exposure unit risk based on 54 years was then rescaled for an entire lifetime of 'Note that 85 years is not employed here as an average lifespan b u t rather, as a cut-off point for the life-table analysis, which uses actual age-specific mortality rates. Exposures in the life-tables were computed at the mid-point of each age interv al and appropriately lagged. This document is a draftfo r re\'iew purposes only and does not constitute utgenev policy. G-2 DRAFT--DO NOT CITE OR QUOTE 1 continuous exposure by multiplying the interim value for adult-only-exposure by 70/54 to cover 2 the childhood years (<16 years) to compute the "adult-based" unit risk. After rescaling, the 3 resulting "adult-based" lifetime unit risk estimate (in contrast to the unsealed 4 "adult-only-exposure" unit risk estimate obtained from the life-table calculations) may be 5 prorated for less-than-lifetime exposure scenarios in the same manner as would be used for an 6 "adult-based" unit risk estimate derived from a rodent bioassay (see Section 5.4.5.4). 7 Consistent with the Guidelinesfor Carcinogen Risk Assessment (U.S. EPA. 2005). the 8 same data and methodology were also used to estimate the exposure level effective concentration 9 (EC,:) and the associated 95% lower confidence limit of that exposure level effective 10 concentration (LECT) corresponding to an absolute risk of 1% (x = 0.01). A 1%-risk level is 11 commonly used for the determination of the point of departure (POD) for low-dose extrapolation 12 from epidemiological data, and the LEC value corresponding to that risk level was used as the 13 actual POD. 14 The following tables illustrate the computational details of the unit risks for 15 mesothelioma mortality (see Tables G-l and G-2). The results of Tables G-l and G-2 are shown 16 in Table 5-16 and are not adjusted for the underascertainment of mesothelioma described in 17 Section 5.4.5.1.1. The unit risks adjusted for underascertainment are shown in Table 5-17. 18 19 20 Column Definitions for Tables G-l and G-2: 21 Column A: Age interval up to age 85. 22 Column B: All-cause mortality rate for interval /' (x 10'Vyear) (Xu et aU 2010) 23 Column C: All-cause hazard rate for interval i (jh*,) (= all-cause mortality rate x number of 24 years in age interval). 25 Column D: Probability of surviving interval i (q,) [= exp(-/?*,)]. 26 Column E: Probability of surviving up to interval i (St) (S\ = 1; St =Dm x $ ri, for i > 1). 27 Column F: Lagged exposure at mid-interval (x dose) assuming constant exposure was initiated 28 at age 16. 29 Column G: Mesothelioma mortality hazard rate in exposed people for interval. To estimate the 30 EECoi, i e., the 95% lower bound on the continuous exposure giving an extra risk of 31 1%, the 95% upper bound on the regression coefficient is used. This document is a draftfor re\'iew purposes only and does not constitute Agency policy. G-3 DRAFT--DO NOT CITE OR QUOTE 1 Column H: All-cause hazard rate in exposed people for interval i (h*x,) [= h*t + (hx, - hi)]. 2 Column I: Probability of surviving interval i without dying from mesothelioma for exposed 3 people (qXi) [= exp(-/?*x,)]. 4 Column J: Probability of surviving up to interval i without dying from mesothelioma for 5 exposed people (Sx) (S.X] = 1; Sx, = Sx, i x qx, \, for i > 1). 6 Column K: Conditional probability of dying from mesothelioma in interval i for exposed people 7 [= (hx, x- h*x,) / Sx, x (1 - qx,)] (Rxx the lifetime probability of dying from 8 mesothelioma for exposed people = the sum of the conditional probabilities across 9 the intervals). 10 11 12 Note that the life-tables for mesothelioma mortality estimate the extra risk as the absolute 13 risk as there is no assumption of a background risk in the absence of exposure. In each of the 14 life-tables, inhalation exposure commences at age 16 years and continues at the same exposure 15 concentration for the duration of the life-table. This allows for the computation of an 16 "adult-only-exposure" occupational lifetime unit risk, which is then scaled by a ratio of 70:54 to 17 account for risk over the standard 70-year lifetime. While exposure is initiated in the life-table at 18 age 16 years, this exposure is lagged to match the corresponding exposure-response models, 19 which provide the hazard rates per unit of exposure. For example, in Table G-l, Column F 20 shows exposure lagged by 10 years so that no lagged exposure appears in the table prior to age 21 26 years (16+10). In Table G-2, Column F shows exposure lagged by 15 years so that no 22 lagged exposure appears in the table prior to age 31 years (16 + 15). Note that risks are initially 23 shown in 1-year intervals because children's risk intervals can be smaller, and there was a need 24 to be able to begin exposures at 16 years. This document is a draftfor review purposes only and does not constitute Agency policy. G-4 DRAFT--DO NOT CITE OR QUOTE Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the reasonable upper bound This document is a draftfor review purposes only and does not constitute Agency policy. G-5 A Age int. <1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 B A ll-ca u se m ortality (xlO 5/ year) 684.5 28.6 28.6 28.6 29.9 13.7 13.7 13.7 13.7 13.7 18.7 18.7 18.7 18.7 18.7 61.9 61.9 C D A ll-cau se hazard rate (h*) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0002 0.0006 0.0006 Prob. of surviving interval (?) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 0.9994 E Prob. of surviving up to interval (S) 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 0.9898 F Lagged exp. mid. int. (Adosc) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 G Exposed meso. hazard rate (hx) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 H Exposed all-cause haz. rate (h*x) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0002 0.0006 0.0006 I Exposed prob. of surviving interval (qx) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 0.9994 J Exposed prob. of surviving up to int. (Sx) 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 0.9898 K Exposed cond. prob. of meso. in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 DRAFT--DO NOT CITE OR QUOTE ^ Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the reasonable upper bound (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-6 A Age int. 17 18 19 20 21 22 23 24 25 26 27 28 29 30-34 35-39 40-44 45-49 50-54 B A ll-ca u se m ortality (xlO 5/ year) 61.9 61.9 61.9 98.3 98.3 98.3 98.3 98.3 99.4 99.4 99.4 99.4 99.4 110.8 145.8 221.6 340.0 509.0 C D A ll-cau se hazard rate (h *) 0.0006 0.0006 0.0006 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0055 0.0073 0.0111 0.0170 0.0255 Prob. of surviving interval (?) 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9945 0.9927 0.9890 0.9831 0.9749 E Prob. of surviving up to interval (S) 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9806 0.9796 0.9786 0.9777 0.9723 0.9652 0.9546 0.9385 F Lagged exp. mid. int. (Adose) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.144 0.401 0.626 0.821 1.268 1.701 1.918 2.026 2.080 G Exposed meso. hazard rate (h x ) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0002 0.0003 0.0004 0.0006 0.0009 0.0010 0.0010 0.0011 H Exposed all-cause haz. rate (h*x) 0.0006 0.0006 0.0006 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0011 0.0012 0.0013 0.0014 0.0062 0.0082 0.0121 0.0180 0.0265 I Exposed prob. of surviving interval (qx) 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9989 0.9988 0.9987 0.9986 0.9938 0.9919 0.9880 0.9821 0.9738 J Exposed prob. of surviving up to int. (Sx) 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9805 0.9793 0.9780 0.9767 0.9706 0.9628 0.9512 0.9342 K Exposed cond. prob. of meso. in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0002 0.0003 0.0004 0.0006 0.0008 0.0009 0.0010 0.0010 DRAFT--DO NOT CITE OR QUOTE Table G-l. Mesothelioma extra risk calculation for environmental exposure to 0.1479 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the reasonable upper bound (continued) This document is a draftfor review purposes only and does not constitute Agency policy. A Age int. 55-59 60-64 65-69 70-74 75-79 80-84 B A ll-ca u se m ortality (xlO 5/ year) 726.3 1,068.3 1,627.5 2,491.3 3,945.9 6,381.4 C D A ll-cau se hazard rate (h *) 0.0363 0.0534 0.0814 0.1246 0.1973 0.3191 Prob. of surviving interval (?) 0.9643 0.9480 0.9218 0.8829 0.8209 0.7268 E Prob. of surviving up to interval (S) 0.9149 0.8823 0.8364 0.7710 0.6807 0.5588 F Lagged exp. mid. int. (Xdose) 2.107 2.121 2.127 2.131 2.132 2.133 G Exposed meso. hazard rate (h x ) 0.0011 0.0011 0.0011 0.0011 0.0011 0.0011 H Exposed all-cause haz. rate (h*x) 0.0374 0.0545 0.0825 0.1256 0.1984 0.3202 I Exposed prob. of surviving interval (qx) 0.9633 0.9470 0.9209 0.8819 0.8201 0.7260 J K Exposed prob. of surviving up to int. (Sx) 0.9098 Exposed cond. prob. of meso. in interval (Rv) 0.0010 0.8764 0.0009 0.8299 0.0009 0.7642 0.0008 0.6740 0.0007 0.5527 0.0005 Absolute Rx = 0.0100 exp. = exposure, haz. = hazard, int. = interval, meso. = mesothelioma, mid. = mid-interval, Prob. = probability. Absolute risk = 0.01000, exp. level = 0.1479; occupational lifetime unit risk = 0.01/0.1479 = 0.0676 (based on occupational exposures beginning at age 16 years); scaled occupational lifetime unit risk = 0.0876 (scaled by ratio o f 70:54 to account for risk over 70-year lifetime). G-7 DRAFT--DO NOT CITE OR QUOTE Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion This document is a draftfor review purposes only and does not constitute Agency policy. G-8 A B C A ge int. <1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 A ll-ca u se m ortality (xlO 5/ year) 684.5 28.6 28.6 28.6 29.9 13.7 13.7 13.7 13.7 13.7 18.7 18.7 18.7 18.7 18.7 61.9 61.9 A ll-cau se hazard rate (h *) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0002 0.0006 0.0006 D Prob. of surviving interval (?) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 0.9994 E Prob. of surviving up to interval (N 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 0.9898 F Lagged exp. mid. int. (Ado sc) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 G Exposed meso. hazard rate (h x ) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 H Exposed all-cause haz. rate (h*x) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0002 0.0006 0.0006 I Exposed prob. of surviving interval (qx) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 0.9994 J Exposed prob. of surviving up to int. {Sx) 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 0.9898 K Exposed cond. prob. of meso. in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 DRAFT--DO NOT CITE OR QUOTE Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-9 A B C A ge int. 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 A ll-ca u se m ortality ( x l 0 5/year) 61.9 61.9 61.9 98.3 98.3 98.3 98.3 98.3 99.4 99.4 99.4 99.4 99.4 110.8 110.8 110.8 110.8 110.8 A ll-cau se hazard rate (h *) 0.0006 0.0006 0.0006 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0055 0.0055 0.0055 0.0055 0.0055 D Prob. of surviving interval (</) 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9945 0.9945 0.9945 0.9945 0.9945 E Prob. of surviving up to interval (S) 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9806 0.9796 0.9786 0.9777 0.9777 0.9777 0.9777 0.9777 F Lagged exp. mid. int. (Ado sc) 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.238 0.664 1.035 1.357 G Exposed meso. hazard rate (hx) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0002 0.0004 0.0005 H Exposed all-cause haz. rate (h*x) 0.0006 0.0006 0.0006 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0011 0.0012 0.0013 0.0015 0.0016 I Exposed prob. of surviving interval (qx) 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9989 0.9988 0.9987 0.9985 0.9984 J Exposed prob. of surviving up to int. (Sx) 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9806 0.9796 0.9786 0.9777 0.9766 0.9754 0.9741 0.9727 K Exposed cond. prob. of meso. in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 0.0002 0.0003 0.0005 DRAFT--DO NOT CITE OR QUOTE Table G-2. Mesothelioma extra risk calculation for environmental exposure to 0.2446 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 15-year exposure lag and a 5-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-10 DRAFT--DO NOT CITE OR QUOTE A B C A ge int. 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 A ll-ca u se m ortality ( x l 0 5/year) 145.8 221.6 340.0 509.0 726.3 1,068.3 1,627.5 2,491.3 3,945.9 6,381.4 A ll-cau se hazard rate (h *) 0.0073 0.0111 0.0170 0.0255 0.0363 0.0534 0.0814 0.1246 0.1973 0.3191 D Prob. of surviving interval (</) 0.9927 0.9890 0.9831 0.9749 0.9643 0.9480 0.9218 0.8829 0.8209 0.7268 E Prob. of surviving up to interval (S) 0.9723 0.9652 0.9546 F Lagged exp. mid. int. (Ado sc) 2.097 2.813 3.171 0.9385 0.9149 0.8823 0.8364 0.7710 0.6807 0.5588 3.350 3.440 3.485 3.507 3.518 3.524 3.527 G Exposed meso. hazard rate (h x ) 0.0007 0.0010 0.0011 0.0011 0.0012 0.0012 0.0012 0.0012 0.0012 0.0012 H Exposed all-cause haz. rate (h*x) 0.0080 0.0120 0.0181 0.0266 0.0375 0.0546 0.0826 0.1258 0.1985 0.3203 I Exposed prob. of surviving interval (qx) 0.9920 0.9880 0.9821 0.9738 0.9632 0.9469 0.9207 0.8818 0.8200 0.7259 J Exposed prob. of surviving up to int. (Sx) 0.9712 0.9634 0.9519 K Exposed cond. prob. of meso. in interval (Rv) 0.0007 0.0009 0.0010 0.9348 0.0011 0.9103 0.0011 0.8768 0.0010 0.8302 0.0010 0.7644 0.0009 0.6740 0.0007 0.5527 0.0006 Absolute Rx = 0.0100 exp. = exposure, haz. = hazard, int. = interval, meso. = mesothelioma, mid. = mid-interval, Prob. = probability. Absolute risk = 0.01000; exp. level = 0.2446; Occupational lifetime unit risk = 0.01/0.2446 = 0.0409 (Based on occupational exposures beginning at age 16 years); Scaled occupational lifetime unit risk = 0.0530 (Scaled by ratio o f 70:54 to account for risk over 70-year lifetime). 1 G.2. LUNG CANCER MORTALITY 2 Lung cancer mortality risk computations are very similar to mesothelioma mortality 3 computations above (see Section G.l), with one important difference that extra risk is used for 4 lung cancer. Extra risk is defined as equaling (RT- R,,) (1 - RQ, where Rxis the lifetime lung 5 cancer mortality risk in the exposed population and R0is the lifetime lung cancer mortality risk 6 in an unexposed population (i.e., the background risk). U.S. age-specific all-cause mortality 7 rates from the 2010 National Vital Statistics Report (Xu et ah, 2010) for deaths in 2007 among 8 all race and gender groups combined were used to specify the all-cause background mortality 9 rates (R0) in the life-table analysis. Cause-specific background mortality rates for cancers of the 10 lung, trachea, and bronchus were obtained from a Surveillance, Epidemiology, and End Results 11 (SEER) report on mortality during 2003-2007 (2003-2007 Surveillance Epidemiology and End 12 Results Table 15.10. age-specific U.S. death rates). 13 The following tables show details of the computations of the unit risks for lung-cancer 14 mortality (see Tables G-3 and G-4). The results of Tables G-3 and G-4 are shown in Table 5-19. 15 16 17 Column Definitions for Tables G-3 and G-4: 18 Column A: Age interval up to age 85. 19 Column B: All-cause mortality rate for interval f (xlOVyear) (Xu et ah, 2010) 20 Column C: Lung-cancer mortality rate for interval /' (' lO'/year) (2003-2007 Surveillance. 21 Epidemiology and End Results Table 15,10, age-specific U.S. death rates). 22 Column D: All-cause hazard rate for interval i (//*,) (= all-cause mortality rate number of 23 years in age interval). 24 Column E: Probability of surviving interval i {qj) [= exp(-/7*,)]. 25 Column F: Probability of surviving up to interval / (S,) (A = 1; Si = S: i x q, for /' > 1). 26 Column G: Lung-cancer mortality hazard rate for interval i (h,) 27 (= lung-cancer mortality rate x number of years in interval). 28 Column Ft: Conditional probability of dying from lung cancer in interval /' 29 i = (h, x- /?*,) mSi x (1 - ft)], i.e., conditional upon surviving up to interval i (Ra, the 30 background lifetime probability of dying from lung cancer = the sum of the 31 conditional probabilities across the intervals). This document is a draftfo r re\'iew purposes only and does not constitute A gency policy. G-11 DRAFT--DO NOT CITE OR QUOTE 1 Column I: Lagged exposure at mid-interval (x dose) assuming constant exposure was initiated 2 at age 16. 3 Column J: Lung-cancer mortality hazard rate in exposed people for interval. To estimate the 4 LECoi, i.e., the 95% lower bound on the continuous exposure giving an extra risk of 5 1%, the 95% upper bound on the regression coefficient is used, i.e., 6 Maximum Likelihood Estimate + 1.645 x standard error. 7 Column K: All-cause hazard rate in exposed people for interval i (h*xt) [= /?*, + (hx, - /?,)]. 8 Column L: Probability of surviving interval i without dying from lung cancer for exposed 9 people (qx,) [= exp(-/?*x,)]. 10 Column M: Probability of surviving up to interval i without dying from lung cancer for exposed 11 people (Sxj) (Sx] = 1; Sx, = Sx, i x qx, \, for i > 1). 12 Column N: Conditional probability of dying from lung cancer in interval i for exposed people 13 [ = (hx, x- h*Xj) x Sxj x (1 - qx,f\ (Rx, the lifetime probability of dying from lung 14 cancer for exposed people = the sum of the conditional probabilities across the 15 intervals). 16 17 18 In each of the life-tables, inhalation exposure commences at age 16 years and continues 19 at the same exposure concentration for the duration of the life-table. This allows for the 20 computation of an "adult-only-exposure" occupational lifetime unit risk, which is then scaled by 21 a ratio of 70:54 to account for risk over the standard 70-year lifetime. While exposure is initiated 22 at age 16 years, this exposure is lagged to match the corresponding exposure-response models, 23 which provide the hazard rates per unit of exposure. For example, in Tables G-3 and G-4, 24 Column I shows exposure lagged by 10 years so that no lagged exposure appears prior to age 25 26 years. This document is a draftfo r review purposes only and does not constitute Agency policy. G-12 DRAFT--DO NOT CITE OR QUOTE Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/cc Libby Amphibole asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year exposure lag, as described in Section 5.4.5.3 as the reasonable upper bound This document is a draftfor review purposes only and does not constitute Agency policy. G-13 A Age Int. <1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 B A ll-cau se m ortality (xlO 5/ year) 684.5 28.6 28.6 28.6 29.9 13.7 13.7 13.7 13.7 13.7 18.7 18.7 18.7 18.7 18.7 61.9 C D E Lung CA m ortality ( x l 0 5/year) A ll cause hazard rate (ft*) Prob. of surviving interval (?) 0 0.0068 0.9932 0 0.0003 0.9997 0 0.0003 0.9997 0 0.0003 0.9997 0 0.0003 0.9997 0 0.0001 0.9999 0 0.0001 0.9999 0 0.0001 0.9999 0 0.0001 0.9999 0 0.0001 0.9999 0 0.0002 0.9998 0 0.0002 0.9998 0 0.0002 0.9998 0 0.0002 0.9998 0 0.0002 0.9998 0 0.0006 0.9994 F G H I J K L Prob. of surviving up to interval (S) 1.0000 Lung CA hazard rate (ft) 0.0000 Cond. prob. of lung CA m ortality in interval (Ro) 0.0000 Lagged exp. mid. int. (Adosc) 0.00 Exposed lung CA hazard rate (hx) 0.0000 Exposed all-cause haz. rate (h*x) 0.0068 Exposed prob. of surviving interval (qx) 0.9932 0.9932 0.0000 0.0000 0.00 0.0000 0.0003 0.9997 0.9929 0.0000 0.0000 0.00 0.0000 0.0003 0.9997 0.9926 0.0000 0.0000 0.00 0.0000 0.0003 0.9997 0.9923 0.0000 0.0000 0.00 0.0000 0.0003 0.9997 0.9920 0.0000 0.0000 0.00 0.0000 0.0001 0.9999 0.9919 0.0000 0.0000 0.00 0.0000 0.0001 0.9999 0.9918 0.0000 0.0000 0.00 0.0000 0.0001 0.9999 0.9916 0.0000 0.0000 0.00 0.0000 0.0001 0.9999 0.9915 0.0000 0.0000 0.00 0.0000 0.0001 0.9999 0.9914 0.0000 0.0000 0.00 0.0000 0.0002 0.9998 0.9912 0.0000 0.0000 0.00 0.0000 0.0002 0.9998 0.9910 0.0000 0.0000 0.00 0.0000 0.0002 0.9998 0.9908 0.0000 0.0000 0.00 0.0000 0.0002 0.9998 0.9906 0.0000 0.0000 0.00 0.0000 0.0002 0.9998 0.9904 0.0000 0.0000 0.00 0.0000 0.0006 0.9994 M N Exposed prob. of surviving up to int. (Sx) Exposed cond. prob. of lung CA in interval (Rv) 1.0000 0.0000 0.9932 0.0000 0.9929 0.0000 0.9926 0.0000 0.9923 0.0000 0.9920 0.0000 0.9919 0.0000 0.9918 0.0000 0.9916 0.0000 0.9915 0.0000 0.9914 0.0000 0.9912 0.0000 0.9910 0.0000 0.9908 0.0000 0.9906 0.0000 0.9904 0.0000 DRAFT--DO NOT CITE OR QUOTE Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/cc Libby Amphibole asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year exposure lag, as described in Section 5.4.5.3 as the reasonable upper bound (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-14 A Age Int. 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30-34 35-39 40-44 B C D A ll-ca u se m ortality ( x l 0 5/year) Lung CA m ortality ( x l 0 5/year) A ll cause hazard rate (h*) 61.9 0 0.0006 61.9 0 0.0006 61.9 0 0.0006 61.9 0 0.0006 98.3 0.1 0.0010 98.3 0.1 0.0010 98.3 0.1 0.0010 98.3 0.1 0.0010 98.3 0.1 0.0010 99.4 0.2 0.0010 99.4 0.2 0.0010 99.4 0.2 0.0010 99.4 0.2 0.0010 99.4 0.2 0.0010 110.8 0.5 0.0055 145.8 2.1 0.0073 221.6 7.9 0.0111 E Prob. of surviving interval (?) 0.9994 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9945 0.9927 0.9890 F G H I J K L Prob. of surviving up to interval (S) Lung CA hazard rate (h) Cond. prob. of lung CA m ortality in interval ( R 0) Lagged exp. mid. int. (Adose) Exposed lung CA hazard rate (hx) Exposed all-cause haz. rate (h*x) Exposed prob. of surviving interval (qx) 0.9898 0.0000 0.0000 0.00 0.0000 0.0006 0.9994 0.9892 0.0000 0.0000 0.00 0.0000 0.0006 0.9994 0.9886 0.0000 0.0000 0.00 0.0000 0.0006 0.9994 0.9880 0.0000 0.0000 0.00 0.0000 0.0006 0.9994 0.9874 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9864 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9854 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9845 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9835 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9825 0.0000 0.0000 0.00 0.0000 0.0010 0.9990 0.9815 0.0000 0.0000 0.10 0.0000 0.0010 0.9990 0.9806 0.0000 0.0000 0.29 0.0000 0.0010 0.9990 0.9796 0.0000 0.0000 0.48 0.0000 0.0010 0.9990 0.9786 0.0000 0.0000 0.67 0.0000 0.0010 0.9990 0.9777 0.0000 0.0000 1.24 0.0000 0.0055 0.9945 0.9723 0.0001 0.0001 2.20 0.0001 0.0073 0.9927 0.9652 0.0004 0.0004 3.15 0.0004 0.0111 0.9890 M N Exposed prob. of surviving up to int. (Sx) Exposed cond. prob. of lung CA in interval (Rx) 0.9898 0.0000 0.9892 0.0000 0.9886 0.0000 0.9880 0.0000 0.9874 0.0000 0.9864 0.0000 0.9854 0.0000 0.9845 0.0000 0.9835 0.0000 0.9825 0.0000 0.9815 0.0000 0.9806 0.0000 0.9796 0.0000 0.9786 0.0000 0.9777 0.0000 0.9722 0.0001 0.9652 0.0004 DRAFT--DO NOT CITE OR QUOTE Table G-3. Lung cancer extra risk calculation for environmental exposure to 0.191 fibers/cc Libby Amphibole asbestos using a linear exposure-response model based on the metric of cumulative exposure with a 10-year exposure lag, as described in Section 5.4.5.3 as the reasonable upper bound (continued) This document is a draftfor review purposes only and does not constitute Agency policy. A Age Int. 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 B C D A ll-ca u se m ortality ( x l 0 5/year) Lung CA m ortality ( x l 0 5/year) A ll cause hazard rate (h*) 340.0 20.2 0.0170 509.0 39.8 0.0255 726.3 74.7 0.0363 1,068.3 139.8 0.0534 1,627.5 220.9 0.0814 2,491.3 304.3 0.1246 3,945.9 369.5 0.1973 6,381.4 379.4 0.3191 E Prob. of surviving interval (?) 0.9831 0.9749 0.9643 0.9480 0.9218 0.8829 0.8209 0.7268 F G H I J K L Prob. of surviving up to interval (S) 0.9546 Lung CA hazard rate (h) 0.0010 Cond. prob. of lung CA m ortality in interval ( R 0) 0.0010 Lagged exp. mid. int. (Xdose) 4.11 Exposed lung CA hazard rate (hx) 0.0011 Exposed all-cause haz. rate (h*x) 0.0171 Exposed prob. of surviving interval (qx) 0.9831 0.9385 0.0020 0.0018 5.06 0.0022 0.0257 0.9747 0.9149 0.0037 0.0034 6.02 0.0042 0.0368 0.9639 0.8823 0.0070 0.0060 6.97 0.0080 0.0544 0.9470 0.8364 0.0110 0.0089 7.93 0.0129 0.0832 0.9201 0.7710 0.0152 0.0110 8.88 0.0181 0.1275 0.8803 0.6807 0.0185 0.0114 9.84 0.0224 0.2013 0.8177 0.5588 0.0190 0.0091 10.79 0.0235 0.3236 0.7236 R0 = 0.0531 M N Exposed prob. of surviving up to int. (Sx) 0.9545 Exposed cond. prob. of lung CA in interval (Rx) 0.0010 0.9384 0.0020 0.9146 0.0038 0.8815 0.0069 0.8348 0.0103 0.7682 0.0131 0.6762 0.0137 0.5529 0.0111 Rx = 0.0625 CA = cancer, cond. = conditional, exp. = exposure, haz. = hazard, int. = interval, mid. = mid-interval, Prob. = probability. Extra risk = 0.01001; exp. level = 0.191; occupational lifetime unit = 0.01/0.191 = 0.0524 (based on occupational exposures beginning at age 16 years); scaled occupational lifetime unit = 0.0679 (scaled by ratio o f 70:54 to account for risk over 70-year lifetime). G-15 DRAFT--DO NOT CITE OR QUOTE Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion This document is a draftfor review purposes only and does not constitute Agency policy. G-16 DRAFT--DO NOT CITE OR QUOTE A Age int. <1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 B C D A ll-cau se m ortality ( x l 0 5/year) 684.5 28.6 28.6 28.6 29.9 13.7 13.7 13.7 13.7 13.7 Lung CA m ortality ( x l 0 5/year) 0 0 0 0 0 0 0 0 0 0 A ll-ca u se hazard rate (ft*) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 18.7 0 0.0002 18.7 0 0.0002 18.7 0 0.0002 18.7 0 0.0002 18.7 0 0.0002 61.9 0 0.0006 E Prob. of surviving interval (?) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 F Prob. of surviving up to interval (S) 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 G H Lung CA hazard rate (ft) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Cond. prob. of lung CA m ortality in interval (Ro) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 I Lagged exp. m id. int. (Ado sc) 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 J Exposed lung CA hazard rate (h x ) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 K Exposed all-cause hazard rate (h*x) 0.0068 0.0003 0.0003 0.0003 0.0003 0.0001 0.0001 0.0001 0.0001 0.0001 0.0002 0.0002 0.0002 0.0002 0.0002 0.0006 L Exposed prob. of surviving interval (qx) 0.9932 0.9997 0.9997 0.9997 0.9997 0.9999 0.9999 0.9999 0.9999 0.9999 0.9998 0.9998 0.9998 0.9998 0.9998 0.9994 M Exposed prob. of surviving up to int. (Sx) 1.0000 0.9932 0.9929 0.9926 0.9923 0.9920 0.9919 0.9918 0.9916 0.9915 0.9914 0.9912 0.9910 0.9908 0.9906 0.9904 N Exposed cond. prob. of lung CA in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-17 DRAFT--DO NOT CITE OR QUOTE A Age int. 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30-34 35-39 B All-cause mortality (x l0 5/year) 61.9 61.9 61.9 61.9 98.3 98.3 98.3 98.3 98.3 99.4 99.4 99.4 99.4 99.4 110.8 145.8 C Lung CA mortality (x l0 5/year) 0 0 0 0 0.1 0.1 0.1 0.1 0.1 0.2 0.2 0.2 0.2 0.2 0.5 2.1 D E All cause hazard rate (h*) Prob. of surviving interval (?) 0.0006 0.9994 0.0006 0.9994 0.0006 0.9994 0.0006 0.9994 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0010 0.9990 0.0055 0.9945 0.0073 0.9927 F Prob. of surviving up to interval (S) 0.9898 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9806 0.9796 0.9786 0.9777 0.9723 G H I J K L Lung CA hazard rate (h) 0.0000 Cond. prob. of lung CA mortality in interval (Ro) 0.0000 Lagged exp. mid. int. (Ado sc) 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.00 0.0000 0.0000 0.16 0.0000 0.0000 0.48 0.0000 0.0000 0.77 0.0000 0.0000 1.04 0.0000 0.0000 1.74 0.0001 0.0001 2.64 Exposed lung CA hazard rate (hx) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 Exposed all-cause hazard rate (h*x) 0.0006 0.0006 0.0006 0.0006 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0010 0.0055 0.0073 Exposed prob. of surviving interval (qx) 0.9994 0.9994 0.9994 0.9994 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9990 0.9945 0.9927 M Exposed prob. of surviving up to int. (Sx) 0.9898 0.9892 0.9886 0.9880 0.9874 0.9864 0.9854 0.9845 0.9835 0.9825 0.9815 0.9806 0.9796 0.9786 0.9777 0.9722 N Exposed cond. prob. of lung CA in interval (Rv) 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0000 0.0001 Table G-4. Lung cancer extra risk calculation for environmental exposure to 0.333 fibers/cc Libby Amphibole asbestos using the metric of cumulative exposure with a 10-year exposure lag and a 10-year half-life of exposure, as described in Section 5.4.5.3 as the lowest information criterion (continued) This document is a draftfor review purposes only and does not constitute Agency policy. G-l 8 DRAFT--DO NOT CITE OR QUOTE A B Age in t 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-- 84 A ll-cau se m ortality ( x l 0 5/year) 221.6 340.0 509.0 726.3 1,068.3 1,627.5 2,491.3 3,945.9 6,381.4 C Lung CA m ortality ( x l 0 5/year) 7.9 20.2 39.8 74.7 139.8 220.9 304.3 369.5 379.4 D E A ll cause hazard rate (h*) Prob. of surviving interval (?) 0.0111 0.9890 0.0170 0.9831 0.0255 0.9749 0.0363 0.9643 0.0534 0.9480 0.0814 0.9218 0.1246 0.8829 0.1973 0.8209 0.3191 0.7268 F Prob. of surviving up to interval (S) 0.9652 0.9546 0.9385 0.9149 0.8823 0.8364 0.7710 0.6807 0.5588 G H I J K L Lung CA hazard rate (h) 0.0004 Cond. prob. of lung CA m ortality in interval (Ro) 0.0004 Lagged exp. m id. int. (Ado sc) 3.27 Exposed lung CA hazard rate (hx) 0.0004 Exposed all-cause hazard rate (h*x) 0.0111 Exposed prob. of surviving interval (qx) 0.9889 0.0010 0.0010 3.72 0.0012 0.0172 0.9830 0.0020 0.0018 4.04 0.0023 0.0258 0.9746 0.0037 0.0034 4.26 0.0044 0.0370 0.9637 0.0070 0.0060 4.42 0.0083 0.0547 0.9468 0.0110 0.0089 4.53 0.0131 0.0834 0.9200 0.0152 0.0110 4.61 0.0181 0.1274 0.8803 0.0185 0.0114 4.67 0.0220 0.2008 0.8180 0.0190 0.0091 4.71 0.0226 0.3227 0.7242 M Exposed prob. of surviving up to int. (Sx) 0.9652 0.9545 0.9383 0.9144 0.8812 0.8343 0.7675 0.6757 0.5527 N Exposed cond. prob. of lung CA in interval (Rv) 0.0004 0.0011 0.0021 0.0039 0.0071 0.0105 0.0130 0.0135 0.0107 1^ = 0.0531 Rx = 0.0626 CA = cancer, cond. = conditional, exp. = exposure, haz. = hazard, int. = interval, mid. = mid-interval, Prob. = probability. Extra risk = 0.01001; exp. level = 0.333; occupational lifetime unit risk = 0.01/0.333 = 0.0300 (based on occupational exposures beginning at age 16 years); scaled occupational lifetime unit = 0.0389 (scaled by ratio of 70:54 to account for risk over 70-year lifetime). 1 G.3. REFERENCES 2 6EIR (Committee on the Biological Effects o f Ionizing Radiation). (1988). Health risks o f radon and other 3 internally deposited alpha-emitters: BEIRIV. Washington, DC: National Academy Press. 4 http://www.nap.edu/openbook.r)hp?isbn=0309037972. 5 U.S. EPA (U.S. Environmental Protection Agency). (1994). Methods for derivation of inhalation reference 6 concentrations and application o f inhalation dosimetry. (EPA/600/8-90/066F). Research Triangle Park, NC: 7 U.S. Environmental Protection Agency, Office o f Research and Development, Office o f Health and 8 Enviromnental Assessment, Environmental Criteria and Assessment Office. 9 http://cfpub.epa. gov/ncea/cfm/recordisplav.cfm?deid=71993. 10 U.S. EPA (U.S. Enviromnental Protection Agency). (2005). Guidelines for carcinogen risk assessment. (EPA/630/P- 11 03/00IF). Washington, DC. http://www.epa.gov/cancerguidelines/. 12 Xu, JO: Kochanek, KD: Murphy, SL: Teiada-Vera, B, (2010). Deaths: Final Data for 2007. Hyattsville, MD: 13 National Center for Health Statistics, http://www.cdc. gov/nchs/data/nvsr/nvsr58/nvsr58 19.pdf. 14 This document is a draftfor w fe * purposes only and does not constitute Agency policy. G-19 DRAFT--DO NOT CITE OR QUOTE 1 APPENDIX H. GLOSSARY OF ASBESTOS TERMINOLOGY 2 Acicular: The very long and very thin, often needle-like shape that characterizes some prismatic 3 crystals. (Prismatic crystals have one elongated dimension and two other dimensions that 4 are approximately equal). Acicular crystals or fragments do not have the strength, 5 flexibility, or other properties often associated with asbestiform fibers. 6 7 Actinolite: An amphibole mineral in the tremolite-ferroactinolite series. Actinolite can occur in 8 both asbestiform and nonasbestiform mineral habits. The asbestiform variety is often 9 referred to as actinolite asbestos. 10 11 Amosite: An amphibole mineral in the cummingtonite-grunerite series that occurs in the 12 asbestiform habit. The name amosite is a commercial term derived from the acronym for 13 "Asbestos Mines of South Africa." Amosite is sometimes referred to as "brown 14 asbestos." 15 16 Amphibole: A group of minerals composed of double-chain SiC>4 tetrahedra linked at the vertices 17 and generally containing ions of iron and/or magnesium in their structures. Amphibole 18 minerals are of either igneous or metamorphic origin. Amphiboles can occur in a variety 19 of mineral habits including asbestiform and nonasbestiform. 20 21 Anthophyllite: An amphibole mineral that can occur in both the asbestiform and nonasbestiform 22 mineral habits. The asbestiform variety is referred to as anthophyllite asbestos. 23 24 Asbestiform: A specific type of mineral fibrosity in which crystal growth is primarily in one 25 dimension, and the crystals form as long, flexible fibers. In minerals occurring in 26 asbestiform habit, fibers form in bundles that can be separated into smaller bundles and 27 ultimately into fibrils. 28 This document is a draftfor review purposes only and does not constitute Agency policy. H -1 DRAFT--DO NOT CITE OR QUOTE 1 Asbestos: A generic term for silicate minerals occurring in the asbestiform habit, usually used to 2 refer to those minerals that have been commercially exploited as asbestos, including 3 chrysotile in the serpentine mineral group and tremolite asbestos, actinolite asbestos, 4 anthophyllite asbestos, cummingtonite-grunerite asbestos (amosite), and riebeckite 5 asbestos (crocidolite) in the amphibole mineral group. 6 7 Asbestos Structure: A term applied to any connected or overlapping grouping of asbestos fibers 8 or bundles, with or without other particles. 9 10 Aspect Ratio: The ratio of the length of a particle to its diameter. 11 12 Biopersistence: The ability to remain in the lung or other tissue. Biopersistence of mineral fibers 13 is a function of their fragility, solubility, and clearance. 14 15 Chrysotile: A mineral in the serpentine mineral group that occurs in the asbestiform habit. 16 Chrysotile generally occurs segregated as parallel fibers in veins or veinlets and can be 17 easily separated into individual fibers or bundles. Often referred to as "white asbestos," 18 chrysotile is used commercially in cement or friction products and for its good 19 spinnability in the making of textile products. 20 21 Cleavase Fragment: A particle, formed by comminution (i.e., crushing, grinding, or breaking) 22 of minerals, often characterized by parallel sides. In contrast to fibers from an asbestos 23 mineral; elongate mineral particles in a population of cleavage fragments are generally 24 wider and shorter, generally have a lower aspect ratio, and do not exhibit fibrillar 25 bundling at any level of examination. 26 27 Countable Particle: A particle that meets specified dimensional criteria and is (to be) counted 28 according to an established protocol. A countable particle under the National Institute for 29 Occupational Safety and Health asbestos fiber definition is any acicular crystal, 30 asbestiform fiber, prismatic crystal, or cleavage fragment of a covered mineral that is This document is a draftfor review purposes only and does not constitute Agency policy. H-2 DRAFT--DO NOT CITE OR QUOTE 1 longer than 5 pm and has a minimum aspect ratio of 3:1 based on a microscopic analysis 2 of an airborne sample using NIOSH Method 7400 or an equivalent method. 3 4 Crocidolite: An asbestiform amphibole mineral in the glaucophane-riebeckite series. 5 Crocidolite, commonly referred to as "blue asbestos," is a varietal name for the 6 asbestiform habit of the mineral riebeckite. 7 8 Durability: The tendency of particles to resist degradation in body fluids. 9 10 Elongated Mineral Particle (EMP): Any particle or fragment of a mineral (e.g., fibril or bundle 11 of fibrils: acicular, prismatic, or cleavage fragment) with a minimum aspect ratio of 3:1, 12 based on a microscopic analysis of an airborne sample using NIOSH Method 7400 or an 13 equivalent method. 14 15 Elongated Particle (EP): A particle with a minimum aspect ratio of 3:1, based on a microscopic 16 analysis of an airborne sample using NIOSH Method 7400 or an equivalent method. 17 18 Fiber: "Fiber" can be used in a regulatory context or in a mineralogical context. 19 20 In the regulatory context, a fiber is an elongated particle equal to or longer than 5 pm 21 with a minimum aspect ratio of 3:1. The dimensional determination is made based on a 22 microscopic analysis of an air sample using NIOSH Method 7400 or an equivalent 23 method. 24 25 In the mineralogical context, a fiber is an elongated crystalline unit that resembles an 26 organic fiber and that can be separated from a bundle or appears to have grown 27 individually in that shape. 28 29 Fibril: A single fiber of asbestos that cannot be further separated longitudinally into thinner 30 components without losing its fibrous properties or appearances. This document is a draftfor review purposes only and does not constitute Agency policy. H-3 DRAFT--DO NOT CITE OR QUOTE 1 Fibrous: A descriptive characteristic of a mineral composed of parallel, radiating, or interlaced 2 aggregates of fibers, from which the fibers are sometimes separable. 3 4 Fragility: The tendency of particles to break into smaller particles. 5 6 Libby Amyhibole Asbestos: The term used in this document to identify the mixture of amphibole 7 mineral fibers of varying elemental composition (e.g., winchite, richterite, tremolite, etc.) 8 that have been identified in the Rainy Creek complex near Libby, MT as described in 9 Section 2.2. 10 11 Nonasbestiform: The term used to describe fibers not having an asbestiform habit. The massive 12 nonfibrous forms of the asbestos minerals have the same chemical formula and internal 13 crystal structure as the asbestiform variety but have crystal habits where growth is more 14 equivalent in two or three dimensions instead of primarily one dimension. When milled 15 or crushed, nonasbestiform minerals generally do not break into fibers/fibrils but rather 16 into fragments resulting from cleavage along the two or three growth planes. Often, 17 cleavage fragments can appear fibrous. 18 19 Primary Structure: A fibrous structure that is a separate entity in the transmission electron 20 microscope image. 21 22 Refractory Ceramic Fiber (RCF): An amorphous, synthetic fiber produced by melting and 23 blowing or spinning calcined kaolin clay or a combination of alumina (AI2 O3 ) and silicon 24 dioxide (Si02). Oxides (such as zirconia, ferric oxide, titanium oxide, magnesium oxide, 25 and calcium oxide) and alkalis may be added. 26 27 Solid Solution Series: A grouping of minerals that includes two or more minerals in which the 28 cations in secondary structural position are similar in chemical properties and size and 29 can be present in variable but frequently limited ratios. 30 This document is a draftfor review purposes only and does not constitute Agency policy. H-4 DRAFT--DO NOT CITE OR QUOTE 1 Structure: A single fiber, fiber bundle, cluster, or matrix. 2 3 Synthetic Vitreous Fiber (SVF): Any of a number of manufactured fibers produced by the 4 melting and subsequent fiberization of kaolin clay, sand, rock, slag, etc. Fibrous glass, 5 mineral wool, ceramic fibers, and alkaline earth silicate wools are the major types of 6 SVF, also called man-made mineral fiber (MMMF) or man-made vitreous fiber 7 (MMVF). 8 9 Thoracic-size Particle: A particle with an aerodynamic equivalent diameter that enables it to be 10 deposited in the airways of the lung or the gas exchange region of the lung when inhaled. 11 12 Tremolite: An amphibole mineral in the series tremolite-ferroactinolite. Tremolite can occur in 13 both fibrous and nonfibrous mineral habits. The asbestiform variety is often referred to 14 as tremolite asbestos. Due only to changes in the International Mineralogical 15 Association's amphibole nomenclature, subsets of what was formerly referred to as 16 tremolite asbestos are now mineralogically specified as asbestiform winchite and 17 asbestiform richterite. This document is a draftfor review purposes only and does not constitute Agency policy. H-5 DRAFT--DO NOT CITE OR QUOTE