Document ba5Y6jzRkpJroOxV2o9ZdMo03

r,t qJ it ^ 3M-514 ^nv | tj |v#'9 Pnftttti <a '"if*i` Vkuik ATTACHMENT V CARCINOGENIC RISK ESTIMATION FOR CHLOROFORM: AN ALTERNATIVE TO ERA'S PROCEDURES Sir.--The Environmental Protection Agency lEPAI has proposed jn amendment to the National Interim Prfmarv Drinking Water Regulations as part of the Safe Drinking Water Act The amend ment proposed by EPA would establish a Maximum Contaminant Level for chloroform land oilier trthalomethancst of 0 1 ppm. The Agency indicates in its proposal that this action is necessary because it is felt that the concentrations of chloroform known to exist in drinking-water constitute an appreci able cancer risk to man tFederal Register 1978. 43, 5755l. The magnitude of the cancer risk from chlorolorm has been estimated by the EPA from data generated at the National Cancer Institute iRepori on the Carcinogenesis Bioassay of Chloroform. Bethesda. MD. 19761. Ir^this study, long-term administration of chloroform at very high doses was found to produce tumours in laboratory rodents However, there is reason to believe (hat the risk extrapolation performed by EPA seriouslv overestimates the actual potential of chloroform to induce cancer in man. There are two major reasons why this particular risk estimate is probably inaccurate. First, the EPA has failed to consider important data concerning the mechanisms through which chloroform exerts its toxicity This has led to a major error in making species species extrapo lations of carcinogenic risk Secondly, the NCI study was carried out with very high doves of chloro form At slightly lower doses, the relative carcinogenicity of chlorolorm falls sharply. This suggests that there may be detoxification mechanisms which effectively protect the animals until ihev are overwhelmed by very large doses. Detailed discussion of ihese two points follows In new of ihc enormous economic impact of regulations such as these, it is imperative that ihe risk estimation be performed as accurately as possible so that the risk henctii ratio can be properly considered. Species.species extrapolation for chloroform toxiatt The Carcinogen Assessment Group of EPA performs risk extrapolations under the assumption that the carcinogenic process does not have a threshold Since it is impossible to prove or disprove this point experimentally, further argument is Futile and we will proceed without resolving this question. The risk extrapolation from animals to nan involves two steps IAI First, the exetss cancer incidence tif statistically different Irom the control) is empirically fined to a mathematical model relating dose and response. The most commonly used models are. (0 Linear P * fid (ill `One hit' P * I - e",J where P is the fractional incidence of excess tumours, d is the average daily dose in me kg. and 0 is a dimensionless parameter to be estimated from the experimental data (Bl Next a species correction factor is applied to /J Tin- has been chosen to he Ihe cube root of the ratio of Ihe body weights lie. adjustment on a dose unit surface areal Assuming an jverace weight of 30g for mice. 500g for rats and "0kg for man. this 'ranslaies as a 13 3-fold increase for a mousc-to-man, extrapolation and a 5 19-fold increase for a rat-to-mjn extrapolation Finally the risk to man is calculated by substitupne the human dose into 'he equation The net result is that man is always considered to be more sensitive to a carcinogen "ban the small animal species. This type of species.species extrapolation was developed hv Rail iF.ntr. Ris i960. J, }Ml for application lo a group of 18 antincoplastic agent'. These chemicals produce direct cytotoxiciiv and. according to the author ". . are generally not involved in the variable drug metabolizing systems In the ease of chlorofcrm, however, metabolism plays a different role in toxicitv This is because the toxicity of chloroform is apparcntlv not due o the parent molecule iiself. but rather to the production of reactive metabohteis) which bind covalently to tissue proteins iBrodie et at Proc natn. Acad Set. 1971. 63. 160: llett et j/. Expl mol rath 19'?. 19. 315: Reid Proc Fifth Im, Congr. Phormoc. Sun Francisco, C.4. 1973: Reid .fc ICrishnx Expl mol Path. !9'3. 18. 80 Retd et at. Am Ret- resp Do. 197?, 1C7, 539) Chloroform ihus appears to belong io a class of chemicals that require metabolic activation for toxicity Con.equcntly, anv species variation in the capacity to metabolize chloroform should dramatically affect the oncogenicity of this compound. In order to assess the relative importance of these factors, wc ha'e attempted to predict the carcinogenicity of chloroform to rats on the basis of the tumour incidence observed in mice iNCt, foe. or.). Since rats were also tested in the NCI vtudv of chloroform the observed incidence of rat tumourj can serve to indicate the validity of th; predictions. First, we used the interspecia extrapolation procedure of the ERA tprocedure Vi The r.xis in the NCI study had a mean body weight of approx matclv JOOg. while ihe mice had a mean bodv weight of approximately 30g. The cube rooi of the bodv-wetcht ratio (fOtl Jdi is 3 55 The rats should be more sensitive io chloroform by this factor That u. a dose of chloroform that produced 10s, tumours in a group of mice should produce 3,-5', tumours m a croup oi rao Secondly, we have attempted to predict carcinogenicity from the relative rates of production of the toxic chloroform metabolite (procedure Bl. Several investigators have studied the metabolism of chloroform in rats and mice and concluded that mice metabolize chloroform more rapidly than do-rats. The most complete studis are those ol Brown et ul. i Acnnhttoiru 1974, 4. I ? 11. who admtni-- ' 511 CMA. 049494 si: Leuerv 10 ihe EJnor tered a 6G-rng.'kg dose of chloroform orally to rati squirrel monkeys and three mains of mice. They recovered very little unmetabolizcd chloroform from the mice (Table I). In the rat. three times as much unmetabolizcd chloroform was recovered. while in the monkev, 13 times as much unaltered chloroform was exhaled, f rv er j/. [Arcln mf Pharmacodin Thrr. 1972. 196, 9ki adminis tered chloroform to human volunteers and found Ihat 17-66'', of the matcrul was exhaled un changed. However, the human data mav be iomewhat misleading. The dose of chloroform used in man t? mg,kg) was much lower than that given to ihe animals Consequently, the relative amount of chloroform that would be metabolized at higher doses may have been overestimated. Table I. Imrnprciri tiunparimni chhir<ihirm Species Oral dose Img. kg) CHCIt excreted unchanged I", of dose! Mouse Rn Monkey Min 60 60 60 C7 h* :o* *8* P-*6r Brown rt at. Xrnohionco 1974 4, 151 tFry ul. Archt mi. PharmucoJm. Thrr. 1971 196. 98 Although it is obviouslv difficult to make quantitative predictions from this type of metabolic data, it is clear that procedure B frclauve metabolic capacity! leads to the opposite prediction from that given bv the ERA procedure (A) Since the mouse metabolizes chloroform most rapidly, and presumably produces more of ihe ioxic metabolite, ii should be more sensitive to chloroform than the rat. Therefore, the interspecus correction factor is used in reverse in this case. The results of these predictions may be summarized as follows (It Predicied from EP.Vs surface area model (Rail. lac. cir.l (2) Predicted by metabolic rates (Brown ct al. loc. fit ) (3l Observed in NCt bioassay (NCI. loc cn ) Rai > Mouse 12 55-fold) Mouse > Rat (2 55-fold) Meuse > Rat (4 2-fold) For the purpose of evaluating the actual carcinogenicity in the two species, the 0 parameter is , estimated for the most sensitive sex in the NCI bioassay according io the one hit model The relative sensitivity is then approximated for low doses as the ratio of the two /) values Clearlv procedure B is a better predictor than procedure A. for the actual tumour incidence is 4-2-fold higher in the mouse at equivalent doses of chloroform. Thus the application of the ERA technique for mterspecia extrapolation has produced an error of lO'-fold in estimating the sensitivity of the rat from the mouse data. It ts also noteworthy that extrapolation of these data to man with these two procedures gives very different results. The EPA procedure iprocedure Al produces human risk estimates that are inconsistent. The mouse data predict a human risk ll-fold higher than the rat-data prediction. In contrast, with procedure B. the two risk estimations agree very well. Risk predicted from the mouse data is only l-6-fold higher than risk predicted from the rat data. Based on these considerations we would consider man to be the least sensitive of the three species to the carcinogenic action of chloroform. This estimate is based on the data of Fry ei al. [loc. cir.l as well as on the general principle that man normally metabolizes materials much more slowly than the small laboratory animals (Weiss er al. Ini J. elm Pharmac. 1977, 15, 5*;i, Although the absolute magnitude of the risk is dependent upon the particular mathematical model employed in step A. incorrect application of the species correction factor' causes an increase of more than two orders of magnitude in the estimated risk to man. Instead of in excess cancer risk "...on the order of 10`4-10"1 ..." lEPA. Statement of the Basis and Purpose for an Amendment to the National Interim Primary Drinking Water Standards on Triltalomethancs Office of W'ater Supply Criteria and Standards Division. Washington. DC, 1973*. we feel that the risk cannot be higher than about 1*. of this figure. In addition to chloroform, there are many other instances where the relative carcinogenicity of chemicals correlates well with their rate of activation to a reactive species. For instance, vinylidene chlortde |VDC1 is metabolized more extensively in the mouse than m the rat tMcKenna er al. Encir. Hhh Perrpta. I77, 21. 9|, Tumours have been observed in mice exposed to VDC iMaltom, ibid 1977, 21. I) while studies in rats have failed to reveal a discernible oncogenic response iMalmm. loc. cii.i Rampv rr al. Encir. Hhh Pcrsprci. 19 T. 21. 33; Viola & Caputo. Cancerogenicuy Studies on Vinylidene Chlortde: paper presented at N1EHS Conference on Comparative Metabolism and i CMA 049495 Letters io (he Editor M3 Toxiciiv of Vinyl Chloride Related Compounds. Bethesda. MD, 2-1 Mav P771 Similarly 2acetaminofluorene is metabolized u a reactive electrophile and (he carcinogenicity of this material correlates well anti the relative abundance of enzvmcs to earrv out this activation in various species IMtiler. Report to Conference on Environmental Carcinogenesis, Michigan State University f East Lansing. MI. 22-23 May 197SI. Trichloroethvlcne is carcinogenic in BbC3Fl mice but nm in Osborne-Mendel rats iNCl, Report on the Carcinogenesis Bioassay of Trichloroethylene. Bethesda. MD. 19761 Thu correlates well with the greaier capacity of mouse microtomes to catalyse incorpor ation of trichloroethylene metabolites into DNA iB.anerjee & Van Duuren. Cancer Res. 19711. JR, 7t6l, Hujk Joseilav dost turapolauons /N For most risk estimations it is customary to extrapolate from high doses, where experimentally observable rates of tumijur induction exist, to doses in ihc environment several orders of maeniiude lower. For msiance. the doses of chloroform used in the NCI viudy iRcport on the Carcinogenesis Bioassay of Chloroform. Beihesda. MD. 19761 are about 100,000 fold higher than the Maximum Contaminant Level proposed by the EPA. This raises important questions about the validity of such extrapolations. There are many instances where toxicity land presumably also carcinogenicityi increases dispropor tionately once the normal detoxification mechanisms are overwhelmed. An example of this tvpe of behaviour is the work recently reported by McKenna el at. iInc cir I for VDC. It has been postulated that VDC is metabolized in the body to a reactive electrophile which reacts with either endogenous glutathione (detoxification! or various macromolecules Itoxic actioni In the rat. gluiathtone is depleted by exposure to high levels of VDC iJaeger rt uI xpl mol. Path 197a 20, 1 JsT| Furthermore, fasting ihc rats prior to VDC exposure further depleted the glutathione and dramaucallv increased the hepatotoxicity of VDC More importantlv. McKenna et .ii tlm ml noted a sudden disproportionate increase in macromolecul.ir binding of VDC metabolites as glutathione was depleted bv more than 30";, These observations jre particularly important m view of ihe fact that chloroform is also known io deplete glutathione, jnd that chloroform toxicity can be anta gonized by administration of eysieine. a glutathione precursor (Docks ic Krishna, Cxpl muj Path 1976. 24, 131 The validity of using high doses of chloroform to predict the effects of low doves mav be texted by exam.nine the data of the NCI (Report on the Carcinogenesis Bioassay of Chloroform. Bethesda, MD. 19761 and of Roe (Preliminary Report of Long Term Tests of Chloroform in Rais. Mice and Dogs, personal communication. 19761. Data from the highest dose group will he used to estimate the parameters for the one hit model and the linear model Tumour incidence for lower doves will then be predicted and compared to observed incidence rates Bemuse of the different sponuneous tumour rates for the various studies, the most logical w.iv to compare results Irom the studies is through excess tumours calculated from Abbott s correction r O' Thus our models will be: P,,,,,. /Jx ilinear model) P..c... " 1 - '** lone hit modcll Upper 93", confidence limits arc also calculated for the observed number of tumours. These are exact confidence intervals when there is no spontaneous background rate li.e. when Abbott s correction ts not necessarvi or when the observed Irequencv rate in (he ireated croup ts zero. For non-zero frequencies with non-zero spontaneous rates the confidence interval is calculated using ihe normal approximation to the binomial distribution and using the appropriate standard deviation for a ratio from the Taylor series expansion of the ratio The predictions, as well as the actually observed incidences, arc listed in Table 2. Neither the linear model nor the one hit model adequately predict the tumour incidence below 200mg chloro form kg. In most cases, even the upper 95", confidence interval for the observed tumour rates is below rhe predicted values. For instance, the NCI male mice given 13S mekg would be predicted to have 2 4 times more tumours than were actually observed. The tumour incidences observed by Roc Hoc. cir.l in mice exposed to 60mg,kg or less are even further from the predicted values lTable 2). Since Roe utilized four different strains of mive. and none of these results are well desenhed by the model, the failure of the extrapolation is prohablv due to a dettciencv in the model rather than to strain differences. However the only definitive procedure to establish ihe dose-dependency of the carcinogenic effect of chloroform is for the NCI to repeat the carcinogen bioassay over a wider range of doses Until this has been done, there will be a great deal of uncertamtv about the lower end of the dose-response curve. Experience with human exposure to chloroform also appears to be inconsistent with EPVs risk assessment. For example, a study of British workers in a confectionery factors indicated chronic exposure lup io 10 yearsi to concentrations of 50-l2Jppm chloroform vapours iChallen rt </ Hr J. ind. Med. 1938. 15. 243) If we assume that Ihe two routes oi exposure (oral and inhalation) i CMA. 049496 Letter* 10 ihe tuttor Tabic 2. Prrdicted amt ith\eried incidences u( excess liver tumours m ekinrnfnrm. treated mice Predicted tumours ("J Observed tumours Strain B6C3F1 CJ7BI CBA CFT tet - B6C3FI IC1 Dose (me, kg) 277 ns 60 60 60 60 17 477 238 60 17 Linear * One hri* Mala W) 72 20 4: 20 42 20 42 20 42 6 14 Females 47 73 12 31 3 10 Upper 05*. Incidence confidence f,,l interval 92 30 0 0 2 (NS) 7 INS)*, S iNSI 95 SO 0 0 47 6 I 13 IS 20 91 S s NS - Observed tumour incidence not significantly dilTcreni from control incidence P - | - r--' for male and I - e*'4**'* *-- for female tP m 000332 * do*e for male and 000199 dose for female I A 22"., incidence of e*ee*s kulnei tumour* wen in this strain al significantly different from that in the control* are essentially equivalent, thee workers would be predicted to have a lifetime cancer risk of 20--W*, in the group exposed for a mean of 5 * years |EPA one hit model. NCI female mouse data with jurface area correction! These exposures were documented 20 years ago. so that even with a long latency period, increased cancer incidence should be demonstrable tf the EPA risk estimation is valid. Similar reports exist from the chemical industry (Bomski et u/ Areh. Cemerbepath. Cenerhehyg. 196?, la. 127: Lehman Si Schmtdt-Kehl. Arch. H\q. 1936. 116. 131). However, to date, no epidemi ology study convincingly links industrial exposure to chloroform with an increased cancer risk.. In summary, there are several reasons for believing that the current EPA risk estimation for chloroform -is seriously in error We feel that definitive studies of the metabolism of chloroform in various species must be carried out in order to allow a rational species.species extrapolation to be performed. Furthermore, there is evidence of a sudden change in the shape of the chloroform dose-response curve below 200 mg/kg. A complete experiment to evaluate the carcinogenicity of chloroform at lower doses must be carried out before high dose, low dose extrapolations can be performed. A study of the pharmacokinetics of chloroform tn (he selected species would be valuable in selecting appropriate doses for this study. There is good evidence that there are two classes of carcinogens. One class appears to produce its effect directly (eg. 0-propiolactone) and for these agents, at least when given orally, it may be appropriate to apply surface-area correction factors (procedure A). Compounds tn the second class require metabolic activation before they become carcinogenic tea. 2-cetvlaminofluorenei. In this case, the relative rate of metabolic activation appears to be the most reliable means of making interspecies extrapolation of carcinogenic risk Consequently, we strongly suggest that the NCI develop an appropriate technique for distinguishing between these two classes of carcinogens and that EPA then apply appropriate specio<species extrapolation techniques before proposing any regulations based on risk assessment. R. H Reitz and P J. Gehjung. Toticologv Research Laboratory. I MIS Building, Don- Chemical Company. and C. N Paak, C.R. Mathematical .Application Croup. I "0* Buildina, Dan- Chemical Company. Midland. Ml AdoAll. LSA CMA 049497