Document pB2vY7ge0xj3z32XeqgepYgL7

, . .... -, f8!2!i/g-3 B9 :39 :28 \1 lA FAX ~OM:API-HESD WASH DC => 21& 586 4858 R. BREGLIA Pa!Je 2 TO: -tJIA XPEDITE Cll-1) OCT 25, 1993 9=14AM n180 P.02 American Petroleum Institute 1220 L -street, Nor1hw.est W........,n,-D.C. 20005 ~ memo Date: From~ To: -Re: October 25. 1993 Mary Paxton /1~ Benzene Ta~ 1Forc(}(BTF) conference Call on cox's Revised Proposal ln response to the questions posed at the last BTF meeting, -Tony Cox has revised his proposal for implementing a full BBRA model for benzene. His revision and a further addendum he wrote after telephone -d~scussions with Pat Beatty and me are attached for your consideration. I need to have paperwork underway by November 1 for any contract we want to init:i::-ate with 1993 funds. Therefore, I am scheduling a conference call for 2~30 pm EDT on Wednesday, October 27, for the purpose of making a decision about this work. The call-in number is (202)-833-877-T. Please return the following f-orm so wa wilJ know who to &xpect on the call. Thank you. att: 10/12 revision 10/21 addendum BP-00011878 I iB'/Z't5/93 B9:39:45 VIA FAX FROM: API -HESD WASH DC Please return to: Roxie Landry HESD, API fax: 202-682-8270 -> 21E. 58. 4858 -R-. B~ECLlA Pa!Je 3 TO: UIA XPEDJTE C11-U OCT 25, 1993 9:15AM ~180 P.03 (name) (company) will particip~te wi~l not participate jn the conference Cal-l Of. the-Benzene Task Force on Cox proposal at 2:30pm EDT on Wednesday, October 27, 1993. BP-00011879  'lEI(Z~/93 H9:4B:EtB VIA FAX ,, ,f FROM=API-HESD WASH DC -> 21& 58& 4Et58 R. BREGLIA Page- 4 TO: UIA XPEDITE C11-1) OCT 25, 1993 9:15AM U180 P.04 To: Dr. Mary Paxton, API From: Tony COx, Cox As&ociates Re: Proposal to develop a full BBRA model for benzene Date: 10-12-1993 Thank you for your comments on my 9-18-93 proposal to develop a full BBRA model for benzene. 1 understand that the conunittce has two concerns. which can be paraphrased as follows: (i) Cost concerns: -1\1y initially proposed budget of $78k is expensive. Is this much work really necessary, and if so, why? (Can't most of the benefits~ achieved instead just-by integrating the software modules already aeated in previous years and changing &~of their parameter values, without undenalcing a huge additional cffort'l) (ii) Data concerns: Are there enough benzene-specific data cwrently available to make it worthwhile to invest now in completing the lJBRA model for benzene? If so. where .are these data? (It is almost certainly not worth spending a large. amount of resca.ch dollars on the problem if the data are not sufficient to produce credible, useful results.) 1 am sympathetic to both concerns. This revised proposaJ tries to address them both. Itcontains details on the work and some proposed reductions in scope and cost. A. Background The computermodel-to be delivered in this project is Intended to achieve the following practical benefits: (i) Make biological/] based correction.v to previous benzene-risk a.vsessments (e.g.~ by...enabling -more detailed and credible corrections for high-to-low concentration extrapolations, mouse-to-man extrapolations. and single dose to repeated <lose extrapolations than has-been possible using previous. less comprehensive models). (ii) Use simulations of biologlcal pro<::esses to explore mechanistic hYf'()theseS and-10 -develop improved risk and potency estimates. Improved quantitative understanding of the implieations of benzene biology for the dose-response relation mBY also lead to useful qualitative insights into regulatory and soien~ policy issues, such as whether the ben1.ene dose-response relation is linear or nonlinear at low doses. I- BP-00011880 ..18(2;:S/93 B9!40!48 VIA FAX FROM=API-HESD WASH DC -> 216 586 485B R. BREGLIA Page 5 TO: VIA XPEDITE C11-1J OCT 25, 1993----g~ 15AM t:llBB P. B5 (iii) Use the simulated experiments to guitk tire design and to interpret the results of emplrlrol (laboratory) experiments being perfonned by Professor Rich Irons's team. The model should help to consolidate a number of findings and partial insights into benzene biology and identify areas where reducing scientific unoenainties would most help to improve risk esthnates. To help achieve these benefits. API and W-SPA have already funded imponant bui1ding blocks and a substantial amount of methodology development. This prior work will greatly facilitate the cost-effecti-ve completion of the Cllll"ent project and will help to assure that we achieve the benefits just outlined. For example. we have already had to confront a number of challenging modeling and methodological challenges. such as those involved in draWing sound~ useful. and robust conclusions from models formulated with incomplete (and often inaccurate or imprecise) scientific knowledge. The CP modeling project provided- an opportunity_to develop practical methods for coping with these challenges. Other key activities completed so far include die following: o In 1990, we developed a PBPK model for ben7.ene metabolism in rodents and man. (This model looked at total benzene metabolites. but did not includo circulating mctaboBtes.) o In 1991, the PBPK model for benzene was applied to address implications for benzene risk assessment of using internal dose surrogates instead of administered dosedn regulatory dose-response modeling; o In 1992, a biomathematical framework was developed for integrating the PBPK modeling with new models of pharmacodynamic intcmc::tions. 'This work produced a relatively detailed framework for representing the effects of ~hernicallcukemogens on cenhernatotoxicity, lcinetics, and stochastic AML induction. o In 1993-;-the BBRA framework was implemented for cyclophosphamide (CP) as a case study. This project also provided a mathematical basis for describing and exploring (via computer simulation experiments) the quantitative-dose-response implications of Dr. lrons' insights into the possible biological mechanisms involved in chemically induced myeloid leukemogenesis. -2- BP-00011881 lH/ZS/93 89:41:38 VIR FAI : FROM:API-HESD WASH DC -> 2Hi 586 485B B BREGLrA Pa!Je 6 TO: VIA XPEDITE C11-1) OCT 25, 19'33 9:16AM I:U80 P.-afi The pwpose of the-new work propo~ ben:: is build a complete BBRA model for ben%ene. lt will be simllar in scope to the CP rnodel. but with much more detailed components. Specifically. it will use a state-of-the-art PBPK modcl of benzene metabolism (compared to the simple compartmental flow model of CP phannacOkinctics) and will118t) a hcmatotoxicity model that includes the granulopoiesis model for CP as one sub-model. Thls project is intended to develop a ben~spccific model that has enough detail and credibility to be useful in addressing-issues important to science advocacy or science policy dcblltes. It will require using the best available components. including revisting the PBPK model (to address circulating metabolites and ttansport forms in-greater detail) and extending the hematotoxicity model developed for CP to include damage to erythrocyteS. The following tasks and estimated resource requirements are proposed to complete the BBRA model for benzene. B. Statement of Work Task 1: metabolite& of benzene. Di$cusslon: A goal of our effort in 1990-1993 was to develop computer modules (i.e. sulrmodels) for animals- and humans that can be extended and refined as additional data become available. New data on benzene pharmacokinetics and metabolism have become available since 1990 that describe cilwlating metabolites in rats. (Our 1990 benzene PBPIC model used data of Medinsk:y et at. and Travis et al. did not describe circulating melabolites.) This task will update our PBPK model of-benzene to include tbe new information on production of HQ and other bone marrow benzene metabolites. lt will require critically revicw.ing and synthesizing the benzene me&abolism and pbarmaco.ldnetics literature published since 1990. Specifically, we will incorporate the benzene-specific PBPK modeling enhancements ftlld metabolite parameter values reported by Smith, Bois, and other recent investigators. For an introduction to recent literature on improved PBPK e.s.models of benzene and its principal metabolites, see -Bois, F.Y., T.J. Woodruff, amUl.C. Spear, "Comparison of three physiologically basal phannacokinelic models-of-benzene disposition," Toxicology and Applied Pharmacology, UO, 79-88, 1991. BP-00011882 1B(ZS/93 a9:4z:a9 VIA FAX { FRDM=API-HESD WASH DC -> 216 586 4858 R. BREGLIA Page 7 TO: VIA XPEDITE (11-1) OCT 25, 1993 9: 16RM *l1SB e. 0'? Note: We recommend this paper only as a guide to progress in PBPK modeling of benzene since 1993. We do not agree with or endorse its approach to unce.r1a.inty and sensitivity analysis, which fails to exploit known constraints on the possible values of quantities.. In developing the human PBPK model for benzene. interspecies comparisons and extrapolations of benzene metabolism (especially from the Fisher rats-to humans) will be made in several ways. For example, a human metabolism model with circulating metaoolites will initially be developed by using enzyme specific activity levels across species. This methodology has been used in other several other PBPK modeling efforts. The resulting "biologicallywbased" estimates of metabolic parameter values for human metabolism of benzene will be critically compared to ~vlously published msul&s{due to Henderson, Medinsky, and co-workers) for human benzene metaboHsm extrapolated from rodent data by allomeaic scaling. They will also be cross-checked with -empirical results on_ human cell metabolism obtained by Rich Irons's team. We intmd that the PBPK model developed in this task will be the most detailed and best supported model fonn human benzene metabolism available. Jt may therefore have applications outside the -scope of this-BBRA projecL For example, it migliniQ used to calculate the intl"nlal doses that would have been produced by past occupational exposure _scenarios. Deliverable: PBPK models for benzene -and its-.key circuiating and bone- marrow metabolites in mice. rats. BOd humans. {The mouse PBPK model will -allow pnwious risk estimates based on administered doses of benzene and responses in B6C3F1 mice to be recalculated using internal dose surrrogates. 'Ibis wiU ref'me our own previous work by supplying better qaantitative estimates of specific internal dose surrogates.) Woddng notes on post-1990 benzene metabolism and pnlli'IlliWOk:inetics literature and recent findings will also be made available to the API upon request. EJtimated Resource Requirement6: 170-hours. 17k, 4 months. We do not see any opponunities to significandy redue~this level of effort. We will need to replicate the PBPK. modeling work already performed for benzene in 19891~ for several ciwulating metabolites, and may have to add uansport forms, red blood ceU::and hematocrit binding, and other details. We will also be adding new compartments for bone marrow rnt;tabo~ BP-00011883 ..1H/Z5/S3 H9:~z:sH VIA FAX ( FROM: API -HESD-WRSH DC -> 216 586 4858 R. BRECLIA Page B TO: UIA XPEDITE C11-1) OCT 25, 1993 9:17AM ~180 P.08 involving at least the same le\'el of detail as previous efforts on benzene. and probably JllOIC (e.g., based on Rich Irons's quantitative findings on cellular metabolism of benzene in the bone marrow). Whereas our benzene PBPK model with aggregate metabolism was validated with data on ben~ne levels in blood. urine. and exhaled breath from multiple human volunteers, it will now be necessary to add sensitivity analyses and perhaps some new mathematical techniques (such as a generalization of the AUC consetvation theorem developed for linear compartmental rnodels in our cyclophosphamide report) to deal with the uneertainties raised in modeling circulating metabolites. We believe that 170 hOUJ"S for this effort reflects great cost-effwiency and significant savings compared to what such a modeling effort would usually require. This savings is based on our existing familiarity with beMene metabolism-and biochemistry and on our extensive experience with PBPK extrapolation methods for chemicals. Moreover. we suspect that a well supponcd, critically examined model of human benzene metabolism and marrow metabolites will be valuable to API in several applications, e.g., in assessing the relative risks associated with various occupational and public health exposure s'lellarios. Ttuk 2: B11Ud a hematoro:dcil] modd. l'lkgnJtc U with the PBPK model. Discussion: The main challenge in this task will be to develop a model of the hematopoietic system that adequately describes (i) Bone marrow suppression and inhibition (or stimulation) of hematopoiesis during and following benzene cxposwe. (ii) Kinetics of-affected-hematopoietic ceH populations. including stem cells (preC-PU-OM and earlier) and proliferating cells of the erythroid and granulocytic lineages. Observed compensating proliferation following cessation of benzene exposure must be explained by the model. (iii) The shift in bone marrow cell composition (e.g., in terms of cumulative stresses experienced by different cell populations) due to..prolonged. repetitive exposures. This model wiJl draw heavily on the hematotoxicity modetde\leloped for CP. which will probably be included as a-sub-model. Whereas the CP model focused on granulopoiesis. however, the benzene hematotoxicity model must also consider in detail bematotoxic effects -s- BP-00011884 1. 6. /ZrS/93 89:43:31 VIA FAX FREII"t= API -HESD WASH DC -> 2.16 5Bi 4858 R. BREGLIA Pa!Je 9 TO: UIR XPEDITE (11-1) OCT25, 1993 9:17AM ~180-P.B9 ool'he crfthroid Ji.n.r.asc. whkb i& profoundly llff~ by benzene expoiUI'C. Tbe feedback interactions of the GM ond erythroid lines in concrontns hematupulctic stem cdl populations must also be represen'led. We amk:ipate that the bcnJt>.nc hcmawoxity mockl will be more thaD twice: as large as the correxl'nnrling ~ model. Jt will pro"idc a solid basis for modeling other hemalotnxic agents (cxcludinc tbosc that act on thrombopoiesis). silloe bemene ~~posure llffoc:t8 so mpy bematopOietkl a:ll popUlations. Are sllfficient ben2e1te-specific data available to justify devdopin& a mod~l of bem.ene hematotoxicity Ill this time? We ooHcve that the ans~ is a strong yes, c:spcclally if we wutinue to 11~ animal data to ~ llOd eheot our modeling. Benzene actua11y compares favorAbly to Cl' in terms of ~vaila'bilhy of relcvaol ceU kinctlc.;s data a..t qD&Dtitative hematotoldcity data. This is especially true if ITUirinc 1b.ta are used. Since murine hematopoie&is and buman-henuuopoiem parameter~ are 11iplificantly different. hOMVc:t' (compared, for CQ:Jnplc, to normal amine a.nd oormnl hlliDIU\ llcn.topoicsis), full use of die IIVaUabJe bc-.nmnc-spccific murine h~.xicity dota to fi1 and cross-vaUdate ow beaza1e lteo~nl!tntoxicity modcl may require implementing a separate model of aonn111 murine hcnllllopoicaia. {Our CP project foeusod on human bemawtu.dcity.) Building a wmcambinoo J".IWK/homa~xicity model lor mi~ P well as fm- bnman11 allow U& to vo.lielale our models with a wealth of detailed IPlhnal data oompilcd by many IUitbotS over many years. These include data from lron.,s inY&:~~tipt:icmiJ 118 well as older sources. such as the dala repmltll by C".mokitc (E1rrironmental Hetlllh Penpeclive&. 1989, pp 99-1"00), who pluUal quantitative data for hematopoietic stem c:e11s and early propnhur a::lls during 1nd following bcn7JCI1C. cxpooure. We have reviewed and presented ~pecific data souroes on bcntene-indueed hernatotoxichy several lime~ ""a !be put. 5 ycar11. A rocen& BO'Ql'()O of QlliDdtadve ~ JciDCIICi ltnd t:ytOioxicity paramctrz mimatcs for benuoe is lbe following papec: s., -s.Scllcclin~. M. Loeffler, Schmltz, H. J. ~del. and H. U. Wiebmana, "Homa1otoxic effect~> otbenzcneanal)"'.cd by mathematical modeling," Toxicology, '72, 26s;i79, ] 992. W~ capDCL 10 nlmantiAllyimprovc upon this previoua: worlc b)' ..datillt IW".matntoxic effects over time 1lot to the parenrcompound, ben:r.me, hnt rather to levels of specific bellao:ene BP-00011885 18(~5/93 B9:44:BB -viA FAX t FROM:API-HESD WASH DC -> 216 5BIJ-4BSB Jl. BREGLIA Pa!Je 18 TO: UIA XPEDITE (11-1) OCT -25,~ 1993 9:18AM **181a P.10 metabolites in-bone marrow, as predicted by the7BPK model. In addition. we will used improved modeling techniques that allow more accurate temporal resolution of the effects of benl:Cnc exposure on hematopoiesis. Deliverable: A combined ben%.Cne PBPK hematotoxicity model. As not! above. it may prove necessary to develop separate versions of the combined model, one for mice and one -for humans. so as to take full advantage of available data. Based on our experience with CP data, for which fewer data on hematotodcity are available, we believe that it is realistic to complete a credible, predictively useful model of benzene hematotoxity based on existing data. We expect this model to be more oomplete and more credible than other published models (such as the one of Scheding et al.) due to its explicit accounting for the roles of benzene metabolites and to its improved level of temporal resolution (hours instead of days). which will permit more extensive comparisons to empirical data. Estima~d Resource Reqwrements: 380 hours, 38/c, 6 months. This will cover the cost of developing a new model of benzene hematotoxicity from original data-and modeling efforts. The new model will incorporate the best feature& of our CP model (e.g. a temporal resolution of hours) and will break new ground. compared to previous models of benzene hematotoxicity, by making toxicity a function of specifro metabolite levels instead of benzene levels. Tuk 3: lmpkme~tl tl slochasllc limultJtion mo4el of benz.e11e~inducetl leulemill indrutiora ad link il to the PBPK aiUl hemiUotozicU, models. Discussion: This task will deliver a stochastic simulation model for leukemogenesis. showing the effects of exposurc-rclatr.d changes in cell kinetics on the resulting age-specific leukemia hazard function. It will be based on tfie model of leukemogenesis developed during the CP project, which extends the MVK twostage stochastic model to allow for the "cytotoxic shunt" mechanism discovered by Irons. The new model will also allow for realistic details (discussed but never implemented by Moolgavkar et al., for reasons of mathematical tractability) such as spontaneous extinction of mai:ignant stem cells as well as of initiated ones. Estimates for initiation rates and other parameters will be developed using the most recent data.--e.g those reponed in G.M. Farris et al., "Carcinogenicity of inhaled benzene in CBA mice." Funt:ltlmental andAppliedTo~cology, 20, 4. S03-S07, 1993. -7- BP-00011886 18125/93 89:44:53 VIA FAX- 41 ., ( FRDM=API-HESD WASH DC -> 216 586 4858 E. BREGLIA TO: UIA XPEDITE CU-D- OCT 25, 1993 9:18AM ~18e ~~~ Since these data (which update earlier experimental investigations by Cronkite) indicate an excess of malignant lymphomas rather than myelogenous-leukemias, we will need to use careful biological interpretation in selecting studies and aspects of responses that may be relevant to leukemogenesis jn humans. However, the mouse data should be useful in calculating the effective rate of carcinogenic damage occurring in mice under different exposure scenarios. With tb-e help of our mouse PBPK and hematotoxieity models. we should be able to back-calculate approximate potency factors for internal doses. Much of the detailed work in this project will consist of developing credible panuneter estimates. ~ing ways to cross-validate them (e.g., by applying the same calculations to multiple species or data sets and comparlng the results) and criticall}' evaluating and coping with the remaining uncertainties. Deliverable: A floppy disk containing a stochastic simulation model for pn:dicting leukemia hazard functions-(-and confidence bands, with enough "petitions) for specified benzene exposure scenarios. The specific deliverables from this task arc as follows: (i) A simulation model that integrates the benzene-specific PBPK, hematotoxicity, and stochastic leukemia induction modules. (This will be delivered on a floppy disk.) (ii) A technical memorandum summarizing the model structure, the data used, theuncertainty management methodology used to overcome data and knowledge gaps (e.g., sensitivity analyses for the transition rate parameten of the leukemia model). key sensitivities and remainins-uncertainties, and the main qualitative insights and quantitative risk estimates obtained by-applying the model to various exposure scenarios. Estimated Resource Requirements: 220 lwurs. 22k. 4 months. This task combines Tasks 3 and 4 of my September 18 proposal. We have streamlined the deliwtabJes (1 tcchnical memorandum instead of2) and increased the scope of the memorandum to make it J1lOie u5eful (by including quantitative risk estimates). Successful completion of this project should produce-a wealth of publishable results. Ollr benzene BBRA risk model will address not only qulilitative and scientific questions of importance to benzene risk assessment (e.g., is cumulative exposure a good predictor of risk?), but also specific, quantitative risk predictions (qctbcr with un~ty BP-00011887 18125/93 89:45:33 VIA FAX r.. e- '~, FROM:API-HESD WASH DC -> 216 5B6 4858 R. BREGLIA Pa!Je 12 TO VIA XPEDITE C11-1) OCT 25, 1~ 9:1~M ~180 P.12 inlciYals) for benzene. Following succe&&ful completion of the scope of work outlined here. we will be pleased, at APrs discretion. to submit a separate propos81 to prepare one or more journal articles based on the results. We expect that the results of the cun-ent effort will help to create and gain acceptance for-a new paradigm of practical, applied BBRA modeling for chemicals of importance to industry and regulators. T-asks 1-3 have a combined estimated resource requirement of $77k, compared to the $78k initially_proposed. The project can be completed within 8 months. We believe that this represents an extremely cost~effective work plan. After careful thought about the steps: that must be taken to complete this. work with high enough quality to withstand skeptical (or even hostile) scrutiny from people who are inclided to resist the BBRA approach to risk assessment, I can not find any clear ways to reduce this effort without sacrificing a significant amount-of its impact. Funher cost reductions could be achieved. especially by uncritically adopting models from the literature Tatbcr than a-eating improved models to replace them. However, the quality of-the .final model would be compromised. Please let me know if I can provide further information-01' change the scope of work to make be more useful to API. 1 anticipate that the project proposed lwre, by demonstrating the current practicality of BBRA modeling, can significantly advance the acceptance of "more science in risk assessment". I believe that it would be a mistake to defer completing this project until more data are available. BBRA modelil!& must be applicable to chemicals with incomplete and uncertain scientific data if it is to be useful as a practical (affordabJc) alternative to the LMS paradigm. Our uncertainty-management techniques are now wen- enough dev.eloped to take on an important real problem such as benzene:. However, if the committee decides that completing this project is premature, J will be happy to submit a different proposal for a smaller proj~-that fOCUIIC8 on using BBRA modeling to learn from and tCI pide OJ180ing research efforts. BP-00011888 . . ,.J ~.B"/lS/93 89:4&:14 VIA FAX FROM:API-HESD WASH DC -> 216 586 4B58 R. BREGLIA Page 13 TO: UIA XPEDITE Cll-1) OCT 25,, 1993 9=19AM ~180 P.13 To: Dr. Patrick Beatty, WSPA Frorn: Tony Cox, Cox Associu.tcs Re: Addendum to proposal to develop a full BBRA model for benzene Date: 10-21-1993 The purpose of this addendum is to sununarize and conf'um the main points we agreed on today in our telephone conversation about the desired outcome of the benzene BBRA modeling effort. After discussing technical strategies for uncertainty management in the BBRA model, we agreed that the benzene model must do the following things; Point out which data elements are most important for benzene ride as~'ilmMt Guide further experimentation by showing how eliminating various uncertaintieR would improve (make more precise or more certain) risk estimates for benzene. Present-(-:.nd link the_ derivation of risk estimates to) te&tablehypotheses. and identify which ones would be most valuable to te..c;t. We agreed that the modeling effort should not only clarify which spe<.:ific facts that we don't know now would be most valuable in improving the confidence in.-and precision of, BBRA risk estimates, but it must also suggest the empirical measurements and experimental tests of hypotheses that would help to resolve these uncertainties. I have viewed this active approach to uncertainty management as an intrinsic part of the project. Suggesting what to do about uncertainties, as opposed to merely documenting them, is part the uncertainty and sensitivity aR-alyses buried it in my brief- description of the second deliverable from Task 3. Such analyses quantify how the reduced parameters and variables. -on which risk dopends, might change in Jight of new data and scientific information. They identify-which changes would most affect risk estimates and suggest experiments and information collection opportunities that will reduce the uncertainties to which risk estimates are most sensitive. (Indeed, part of the BBRA approach is the recognition that inforlllation, however insight-provoking, that does not affect the reduced parameters or variables-is not needed in order to develop confident. accurate estimates of risk.) Rather than leaving this implicit. I commit as part of Task 3 to: (i) Discuss the model's most important data elements, sensitivities, and uncertainties. (ii) Identify those-uncertainties-about facts-that most affect the uncerbUnty in risk estimates. (iii) Reconnnend hypotheses, e,;periments.__and measurements to reduce those uncertainties. Please let me know if I can make this effon more-valuable to you in any other way. BP-00011889