Document 4Q67b1aDVwB1kqedG7dX8qN41

'Il'o: Mary Paxton lFrom: Tony Cox Re: Status update on cyclophosphamide modeling and comparison to benzene Dafi.e: 4-6-1993 In preparation for our meeting in Denver on April 29, ll am pleased to provide this summary of our current effort on biologically-based risk assessment (BBRA) modeling of cyclophosphamide (CP). This memo also offers one basis for comparison of the CP model ro potential similar work that might be undertaken for benzene later in the year if the BBRA approach for CP proves useful. The following questions are addressed: o What is the overall "conceptual flowchart" for the BBRA model of ClP? o What is the current status of development for each of its modules? o To what extent can the different modules be reused for other chemicals, especially benzene, and to what extent must the various modules and flows be developed (or customized) separately-for each new chemical? o What data are available to populate the different modules for CP? Are comparable data available for benzene? These questions are addressed in the following sections. As discussed in' the original proposal, a primary goal of our cyclophosphamide (CJP} modeling effort is to test the feasibility of applying the full biologically-based risk assessment (BBRA) framework, developed and proposed in previous-years, to a relatively well-studied human leukemogen. If the data, modeling, or't:omputational challenges prove too great even for CP, then it would probably not be prudent to expect useful short-term (1993-1994) benefits from a similar effort directed at benzene. On the other hand, a convincing demonstration that computer simulation modeling: applied to existing data for CP can lead to useful insights and quantitative results about dose-response relations in humans would raise confidence that application of the approach to benzene might also yield valuable results. Thus, as suggested at our August meeting in Denver in 1992, completion of a full BBRA model for CP may help to provide timely information about the attractiveness of the BBRA BP-00021020 F... ' approach (as compared to other, less knowledge-intensive techniques such as Bayesian modeling or neural nets) for achieving practical dose-response models in-the short term. Completing a BBRA model for CP remains the main goal of our current effort. llune still looks realistic as a completion date. A secondary goal that has emerged during the course of the research is to show how realistic scientific knowledge gaps and data uncertainties can be dealt with during the process of preparing and validating a BBRA model. Thus, not only should this effort enable the API to make a mid-year evaluation of the feasibility of carrying out the BBRA approach (i.e., can a plausible madel incorporating biological information about all of the relevant processes in CP-induced leukemogenesis actually be built in a reasonable amount of time?), but also it should shed light on some of the practical aspects-of using partial, incomplete, and possibly inconsistent information to draw sound and useful conclusions while still adequately expressing model and data uncertainties. A full BBRA model for any chemical carcinogen includes the following components or "modules": o Pharmacokinetics and metabolism (converting time series of administered doses to corresponding time series of internal doses of reactive metabolites at target-sites). o Cell kinetics and cytotoxicity (ideally, for target cell populations and other cell populations -ihat regulate or affect them). o Cancer induction (including stochastic inducing mutations and DNA repair). o Latency period (describing the time between creation of a malignant stem cell and its clinical expression as a tumor. This module must also describe spontaneous extinction of malignant populations and the effects of continued exposure on th~ probability distribution of the timeto-tumor). o- Tumor pro_gression and growth. Figure l is a conceptual "flow chart" showing at a high level how these modules can be fi~ together to form-a computational-model of carcinogenesis for use in biologically-based risk assessment. The reactive metabolites (possibly also including the parent compound in some cases) resulting from the administered dose time series are shown as potentially affecting the BP-00021021 !FlllGllJlR!E ll: pharmacoilinelice PBPI( Modu!:l Aggregats VISW Cancer Induction 11/lodul<;l_ Stochastic induciion with repair r:@j Tumor Progression Module malignant population COPIES--> from above Cancer Resp~nse - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - ~-----~~-- ~ BP-00021022 '' carcinogenic process in at least the following nine ways (indicated by the thin "information arrows" running from the "Reactive Metabolites" compartment to various cell flow and transition rates, whkh are represented by circles attached to the thick arrows indicating flows or transitions into or out of compartments. Compartments are indicated in the diagram by rectangular boxes.) \ o Carcinogens (typically. reactive metabolites of the administered compound) may affect proliferation of stem cells. This possibility is indicated in figure 1 by the left-most infonnation arrow directed from the "reactive metabolites" compartment to the "births" inflow arrow into the "normal stem cell-population" compartment. o Carcinogens may alter the rate of stem cell death and aRQPtosis. Since the stem cell compartment helps io regulate its own birth and death rates (as shown by the feedback infonnation arrows from the "normal stem cells" compartment to the two leftmost flow arrows in the cell kinetics/cytotoxicity module), damage to the stem cell compartment may cause very immature hematopoietic progenitor cells (HPCs) to be prematurely recruited into active proliferation. 'li'his response, recently discus-sed in relation to chemical-induced leukemogenesis by Irons and Stillman (1993, forthcoming), is included in ihe current model but can not be seen at the high level of aggregation shown in figure 1. To show such effects, it is necessary to "zoom in" on the stem cell compartment to reveal its subcompartments (e.g., the resting and cycling subpopulations and cells in different stages of the cell cycle), as shown in Figures 2 and 3. o Carcinogens may alter the differentiation and maturation kinetics of normal stem cells. e.g., by interacting with normal growth factors to make cells that wotilO normally not respond differentiate and/or proliferate prematurely. Recent evidence from benzene metabolites and other secondary Jeukemogens suggests that suchJnteractions with_growth Jactors, leading to altered kinetics of hematopoietic progenitor cells (HPCs, contained in the aggregate cell compartments in Figure 1) may be a common feature in chemically-induced secondary leukemogenesis (Irons, 1993; personal correspondence). o Carcinogens may act as alkyiating agents or exert other genotoxic damage, contributing to the flow of "initiating mutations" linking the normal stem cell population compartment in the Cell Kinetics and Cytotoxicity module to the initiated cells compartment in the CancerInduction Module. Thisjs the classical mechanism of carcinogenesis assumed in the multistage model. It is ~likely to be an appropriate description of at least part of CP's BP-00021023 L I IF'HGUJJRIE ~: !DetaJE~ i!Df Une \Gralflll.llion:wiesi G.RlboMoO!~~ n!!ll ftlln~~ IC~DD llGinldDte!i alllli!ll Cy~oao~uday Modlll.llfite actively cycling traction CFU GM proliferation pool blood granulocytes death and removal release to blood 01~-~:==11 LGRF BP-00021024 ---- -- - - - - - - - - - r ,' lFllG\lJ~!E 3\~ lDeftan~ oW Uil~ ftem C~i~ Compar~melnl& nllll &!hill! ([llr~lllillllllOJl)Ollell lUlboNil~i!llte~ dose rate of carcinogen BP-00021025 l mechanism of carcinogenic action since the metabolite that is thought to be responsible for much of Cl?'s therapeutic action (namely, the aziridinium ion of the CP metabolite ph-esphoramide mustard) has a bifunctional alkylating activity that allows it to cross-link guanine residues on the two strands of DNA in a cell, making the cell non-functional in most cases. o Carcinogens may act as mitogens. not only for normal stem cells (which would increase the flow of initiating mutations, other things being equal), but also for initiated cells, i.e., cells at risk of malignant transformation. Direct mitogenic action, e.g., due to synergy between a metabolite and a growth factor in recruiting resting stem cells to cycle ("push") may add to the indirect effect of cytotoxity in down-stream compartments that becomes translated into increased stem cell activity by the feedback control loops in the hematopoietic system ("pull"). The MVK model enables such prolif-erative effects to be accounted for TCDD and other carcinogens. o Carcinogens may affect the death/differentiation rates of initiated stem cells. e-.g., by creating adducts that block the nmmal actions of tumor-suppressing or apoptosis-inducing gene products. o "Completer" carcinogens may increase the rate of malignant transformations of previously initiated cells. o CarCinogens may reduce the effectiveness of the immune surveillance system that normally screens for tumor cells, thus increasing the "effective" transformation rate of malignant transformation. o Many carcinogens may also affect tumor growth rates. changing the time until a malignant cell proliferates to become a clinically detectable tumor. (Indeed, for example, this is precisely why-CP has been used in cancer chemotherapy. Tumors may also affect pharmacokinetics, although that information path is not shown in Figure 1.) These multiple mechanisms of carcinogenesis are no longer purely speculative. Within the past two years, examples and-evidence supportin.g each as-a contributor to some forms of cancer have been presented in the peer-reviewed literature. Our comprehensive-model of chemical carcinogenesis makes it possible to represent and simulate all such mechanisms. BP-00021026 ,', The diagram in Figure 1 represents the first stage of an iterative modeling process for creating and validating a computational dose-response model of carcinogenesis for use in BBRA. The input to the model is a time series of administered dose (upper left-hand corner). Its output is a time-to-occurrence of a clinical endpoint (e.g., time until the fraction of leukocytes in circulating blood exceeds a threshold that defines the leukemic state) (lower right-hand comer). Randomness intervenes between input and output in the Cancer Induction Module (random initiations, births, deaths, and- second transformations) and in the Latency Period Model (random selection of stem cells to undergo clonal expansion). To complete the model, each module must be refmed and populated with empirical data and formulas showing how its input values propagate into its output values. Model uncertainties about the structure of the diagram and the correct value-propagation formulas, as weiLas data uncertainties about the values of specific parameters, must be addressed by using empiricai relations, assumptions, or approximations to close the essential knowledge gaps. Then, sensitivity analyses obtained by running the completed model can be used to show where improved information would be most valuable in increasing the-precision and confidence of the estimated input-output (i.e., dose-response) relation. The general framework in Figure 1 must be specialized for any specific chemical. While the overall structure and the mass-balance flow equations relating the compartments and flows in the model may be re-used for any chemical carcinogen, the par-ameter values and flow rate formulas must be changed for each chemical. More importantly, the detailed structures of the modules (i.e., the detailed diagrams found by zooming in on the aggregate building blocks) mayu be very different for different chemicals. The-remainder of this memo sketches our progress on each of the modules. o Phannacokinerics and metabolism. The pharmacokinetics and metablism of CP have been well studied and excellent data for building-a model with circulating metabolites are available (see e.g., Sladek, 1988). For many compounds,-including benzene and isophosphamide (an oxazaphosphorine closely related to CP), metabolic saturation is achieved at physiological doses, making it-necessary to use nonlinear formulas based on enzyme kinetics to describe the conversion of administered d-oses into time courses of carcinogenic metabolites. For Cl?, matters -are simpler: metabolism and pharmacokinetics of CP and two of its principle metabolites (4-hydroxyCP ~nd phosphoramide mustard) in man appear to be approximately first-order (linear), making it possible to describe the time courses of blood plasma BP-00021027 ~~ ' I :~ ' concentrations by a simple compartmental flow model (Cohen et. al., 1971; Sladek, 1988). lFigure 4 shows the structure.of the model. This module is currently complete, pending any further refinement following model validation. It can be used as a stand-alone module to simulate the time course of metabolite concentrations in response to any administered dose. By contrast, for benzene, a more elaborate PBPK model (or, at best, a three-compartment model) is required to accurately simulate human pharmacokinetics of the parent compound in blood (corresponding to the Parent Compound Module in the upper left hand comer of Figure 4). Circulating metabolites of benzene have-been represented for PBPK models of lFisher rats, but are not yet available for analogous models for humans. i i o Cell kinetics and cytotoxicity. Biotransformation of CE in the liver produces 4- f hydroxycyclophosphamide (a possible transport form), which may be further metabolized by I target cells (via beta-elimination of the highly toxic acrolein) to form the cytostatic and cell- ! killing reactive-metabolite phosphoramide mustard (PM). Data are presented by Sladek (1988) for rodents and humans, showing the cytotoxic potencies of acrolein and lPM in different cell popvlations (percent cell kill per hour at different concentration levels of tile parent compound, with percent-of cell kill attributable to each metabolite). These data have been incorporated into a cytotoxicity and cell kinetics module customized for CP. The linkages between the Cellular Metabolism Module of the pharmacokinetics module shown in Figure 4 and the cell kineti<:s and cytotoxicity are being finalized based on these data. Since cell-kill in each target population is proportional to the AUC of the metabolite (Sladek, 1988), the linkages fmm metabolism to cytotoxicity are relatively easy to make for CP. The task is further simplified by the fact that we can concentrate on damage in the granulocytic lineage (Steinbach et..al., 1980).. By contrast, benzene and its metaboiites. have more confusing celllevel dose-response curves and exert profound effects on the erythroid as well as the granulocytic cell populations. Thus, a more complicated cell kinetics and cytotoxicity model would be required for benzene. :However, some relevant data and a useful start have been pr.o:vided alr-eady by Scheding et. al. (1992) for benzene. o Cancer-induction. A stochastic cancer induction model- adapted from Portier and KoppSchneider (J991) has been completed and is being customized for CP-hased on the simplifying assumption that phosphoramide mustard (PM) is the-direct alkylating agent of principle interest. The potency (initiating mutations per cell per unit time) is-being estimated using heavily processed (and somewhat confusing) data presented by Dedrick and Morrison (1992). Effects of CP metabolites on DNA repair rates are unknown and are assumed to be negligible (a simplification which probably won't hurt are estimates even if it is incorrect, as BP-00021028 lEllGlUJRIE 41~ JD)e~mnll o~ afhle flPll'DarmacolkiiJle~nlt atllTlli l.Weftabo~nmm Mlodllt.D~te? CIUQomnrell ~oil" CydopllH)Jlblbiamnlle (QJ} &J~ Parenl Cempound Module .~-R~.81 -Hepatic metabolism module hydroxyCP and aldophosphamide in blood rlFI\~ ~~ creation carboxyphosphamide Cellular Me1abolism Module aldophosphamide tau1omer traction for aldophosphamide acrolein fractien A simplified scheme of cyclophosphamide (CP) metabolism and pharmacokinetics, including its principle circulating metabolites and showing only the most importani compounds. -r(SOURCE: Adapted from Sladek, 1988) BP-00021029 this would induce a compensating error in the estimated mutation potency). This module is not yet complete, as some of the data sources are still on order; however, it should be finished (in at least preliminary form) by April 29, and may even be integrated into the rest of the model by then. o Latency period and tumor progression. We are experimenting with a model of stressinduced clonal succession (in which a malignant stem cell becomes expressed only if selected for clonal expansion), based on empirical data modeling by Guttorp et. al. (1988). The time until a malignant cell starts to undergo clonal expansion accounts for part of the random latency period. Simple models of leukemia dynamics based on.competition between normal and malignant cells are also being investigated. The progression of a secondary tumor in a patient may be strongly affected by the presence ofa prior tumor, and we are not attempting to model this interaction, since it seems of limited interest in predictive cancer modeling for general populations. The latency peri~d and tumor progression modules will be the last ones completed. The current target date for completion and integration (-w.Plch is expected to be easy, given the lack of strong linkages to previous modules) is May 20th. This will leave us on schedule for an mid-to-late June deliverable of the fmal model and report. BP-00021030 -----~-------------------------------------- ' (Of over 100 articles and papers surveyed in connection with cyclophosphamide and secondary leukemogenesis and MDS, the following few are the most useful. Of these, the single most informative one is by Sladek, 1988.) Cohen L.L., lY. Jao, aand WJ. Jusko, "Pharmacokinetics of cyclophosphamide in man", British Journal ofPharmacology, ~3, 677-680, 1971. Craddock, C.f., et al, "Circulating stem cells in mice treated with cyclophosphamide," Blood, ~(D, 1, 264-269, 1992. Dedrick, R.L., and P.F. Morrison, "Carcinogenic potency of alkylating agents in rodents and humans," Cancer Research, 2, 2464-2467, May, 1992. Graham. M.A., RJ. Riley, DJ. Kerr, "Drug metabolism in carcinogenesis and cancer chemotherapy-", Pharmac. Ther. R, 275-289, 1991. Guttorp, P., et. al., "A stochastic model for hematopoiesis in the cat," lMA Journal of Mathematics Applied in Medicine and Biology, I, 125-143, 1990. Irons, JR., and W.S. Stillman, "Cell proliferation and differentiation in chemical leukemogenesis," forthcoming in Stem Cells, 1993 or 1994~ Patel, J.M., "Metabolism and pulmonary toxicity of cyclophosphamide," Pharmac. Ther. ~'7. 13i-146, 1990. lPortier, CJ., and A. Kopp-Schneider, "A multistage model of c~cin()genesis incorporating DNA damage and repair," Risk Analysis, ll!, 3, 1991, 535-544. Scheding, S., Loeffler, M., et. al., "Hematotox,ic effects of benzene analyzed by mathematical -modeling," Toxicology, n, 265-279, 1992. Sladek, N.lE., "Metabolism of oxazaphosphorines", Pharmac. Ther. '31, 301-355, 1988. Steinbach, K.H., et. al., "A mathematical model of canine granuiocytopoiesis", Journal of Mathematical Biology, 1-12, 1980. ---------- BP-00021031 rli"l' American Automobile Manufacturers Association 7430 Second Avenue, Sutte 300 o Detroit, Michigan 48202 Tel. No. 313-872-4311 o Fax No. 313-872-5400 April 3, 199J Dr. Mary Jaurr JP'uwn Health Scientit Health and Environmental Sciences Departmmt American Petroleum ][nstirure 1220 IL Street, N.W. Washington. DC 20005 forThank you you!l participation mihe meeting last month with EJP>A ro dlis.cuS lheruth meffect of benzene in the context of the Agencti!l study on mobile oourre wric!l. &sOO )l! theour discussions with Human Health Assessment <Group on predicting ris!l: from llow llevcl exposure to benzene, it would seem tile Agency may be receptive wooncep~ for inoorpmating biologically based mechanisms into their dose response model. Since AJP'I has a projecron the biological basis of chemicai carcinogen~:Si~ using ben.wne as~ model compound, t.'lere may be op~rtun.iti.es lfor collaboration between oor assooiatiom. At least one AAMA member company llw conducted e>tW!siv~ ~x~riineni.!l &00 Ma.lyru on metabolic transformation of benzene, -cytotoxic tffects, genotmric effeci.!l of ibeKll.eHle metabolites and epigenetic effects of 'bellmle. Other AA:M.A companies !have oonsidem.bre expertise in mathematical modeling techniques used ro predicR low !evcl risks ~f ~m~sm@ w chemical carcinogens. And all AAM.A companies are interested in better understanding ~e biological basis and mechanisms of action inherent in chemical carcinogen~. Benzene ns of particular concern because i.t is both a mobile and! stationary rource -pollutant which even ~ very low ~evels of emissions is frequently calculated \bly regularoey agencies ro represeni 21 significant level of risk. Jrn addition mJroiential public illealtb concerns, there are very practicaH. issues of control technology, regulatory standard$, oost, ~.~ which make it imperative ro better umderstand Uhe true risk of benzene e~sure. H ~ more~ scientifically credibles biologically based! mechanism can 100 developOO lfor understanding porenti.aR adverse health impacts and risk efbenzene, looth ilie public lllealt:ln concerns tWll regulatory challenges will be se!i'Verll. Chrysler Corporation o Ford Motor Company o General Motors Cmporation BP-00021032 fc ,. ;I ~ -2- Therefore, AA:MA is pr~sing AFK consider Ill collaborative tffort ro develop & more biologically b&SOO, mechanistic mooei for prooicting_ 'benzene risk a~ low Revels of ex~sure. This oollaboratiorn could taJre nhe form of sharoo funding ~md/ort join~ efforts mr~ collaboration among ocienlisis ftn our respective industries~. A fusi step would oo to lllave & meeting of interested $Cientists and represeniatives from our associations and member companies or oonsultmts. Please call me witl'l yow thoughts on ilib pro_p{)Jsal Md perhapS\ we can proceOO ro rmomge & meeting of interestOO inilividml$. I. I RJIP>/ Richard! r. ~w Manager, lEnvironmen~ ~tb Technical Affairs Division BP-00021033