RiskAnalysis, Vol. 12, No. 2, 1992
Comment
Interspecies Extrapolation of Toxicity Data1 Michael J. Goddard2 and Daniel Krewski2 Toxicological tests using nonhuman test systems remain an important source of information on potential human health hazards(’). Extrapolation of toxicological test results to the human situation is based on the premise that mammalian species respond in a similar fashion to toxic agents. All 53 agents known to cause cancer in humans have also been shown to cause cancer in one or more animal species.(2) Chemicals that cause cancer in mice tend to cause cancer in rats with concordance for chemicals tested in the U.S. National Toxicological Program of about 74%.(SLave et al.(4)suggested that this level of concordance between mice and rats represents an upper bound on concordance between either species and humans. Recent results by Piegorsch ef al.@)have shown that estimates of species concordance depends on carcinogenic potency, and, for low potency carcinogens, the maximum possible species concordance may only be about 80%. Imperfect qualitative agreement between species suggests the need for caution in extrapolating quantitatively between species. In order to address biological differences between species, predictions of human health risk based on the results of toxicological findings in laboratory animals need to take into consideration interspecies differences in physiological, pharmacokinetic, and pharmacodynamic factors associated with the induction of toxic effects. Allometric relationships among physiological parameters in different mammalian species have been observed by a number of investigators. For example, heat loss appears to be related to the surface area of mammals while metabolism is more closely related to body weight. To the extent that such factors mediate toxicological response, they should be adjusted for in interspecies extrapolation. l
Watanabe, Bois and Zeise,c6) hereafter referred to as WBZ, consider the allometric equation
P=axBW’, where P represents a measure of toxic potency in a given species, BW denotes the body weight of that species, and a and b are allometric constants. The U. S. Food & Drug Administration currently uses a value of b = 1, corresponding to interspecies extrapolation on the basis of body weight, whereas the U.S. Environmental Protection A g e n d a employsb = 2/3, correspondingroughly to interspecies extrapolation on the basis of body surface area. In order that b = 213 corresponds precisely to body surface area, it is necessary to assume that toxicological reactions occur on surfaces and different mammalian species have roughly the same shape and density.(*)Surface area extrapolation is more conservative than body weight extrapolation, with rats and mice leading to predictions of human risk that are approximately 6 and 13 times lower based on surface area than body weight. There also appears to be some correlation between the potency (as measured by the TD50, or the dose that induces an excess tumor risk of 50%) of these chemicals as expressed in these two species.(9)A subsequent empirical cornparison(’O) of the carcinogenic potency of 23 chemicals for which both human and animal data are available suggested that carcinogenic potency in animals is correlated with carcinogenic potency in humans. While demonstrating correlation coefficients ranging as high as r = 0.9, the uncertainty associated with predictions of carcinogenicpotency in humans was large. Moreover, the use of both body weight and surface area to convert between species yielded comparable correlation coefficients. Kaldor ef al. (11) obtained estimates of the carcinogenic potency of a small number of antineoplastic drugs that increase the risk of secondary tumours (acute nonlymphocytic leukemia) in humans, along with corresponding estimates for animals. The authors found the results encouraging in terms of using animal data to predict human potency rankings, but cautioned against
Comment on “Interspecies Extrapolation: A Reexaminationof Acute Toxicity Data,” by Watanabe, Bois, and Zeise in Risk Analysis, Vol. 12, No. 2 (June 1992), pp. 301-310. Environmental Health Directorate, Health Protection Branch, Health and Welfare Canada, Ottawa, Ontario, Canada K1A OL2.
315
02724332/92/0600031.5$M.~0/10 1992 Society for Risk Analyuo
316
quantitative prediction of human carcinogenic potency on the basis of these data. Travis and White(12)re-examined the original data reported by Frierich et al. (13)and Schein et al. (14)on the toxicity of anticancer drugs in animals and humans, and concluded that a value of b = 0.75 provides an overall best fit to these data. WBZ note that .this corresponds roughly to species extrapolation on the basis of metabolic rate (c.f Schmidt-Niel~en).('~)This value is intermediate between the value of b = 1 used by the Food and Drug Administration and the value of b = 2/3 used by the Environmental Protection Agency, and has been suggested as a compromise value towards which both agencies may possibly converge. Based on an analysis of the anesthetic potency of 11volatile anesthetics, Travis and Bower@) offer further support for the use of a scaling factor of b = 3/4 with respect to acute toxic effects. WBZ have conducted a further re-analysis of the data by Freirich et al. (13) and Schein et al. (14) in order to determine the most appropriate value of b to use in species extrapolation of toxicity. The novelty of this analysis is the provision for measurement error in the original data; whereas Travis and White(12)assumed that the LD,, and the MTD values were known without error, WBZ allowed for uncertainty in these estimate both by reviewing the variability in the original data and by allowing for an additional uncertainty factor U in the range 1-3 fold. Although the best estimate of the value of b generally remained close to the original estimate of b = 314 obtained by Travis and White,('*) the introduction of additional uncertainty in the analysis led to wider confidence limits on b which now include b = 2/3. WBZ conclude that surface area scaling should thus be used in the interests of prudence. It is illuminating to consider this conclusion in light of Popper's('') paradigm for scientific discovery. Popper's approach is that of firstly embracing a general theory and then testing this by attempting to "falsify" very specific hypotheses. Herein, the general theory might be that equation (3) of WBZ with a specific value of b is valid for all species and chemicals. We add to our knowledge by testing a particular instance. For example, we might study the hypothesis that, for amethopterin, the allometric constant b = 0.75 is valid for mice. (Yet more specifics are required even here including, among others, study duration, route of exposure, and dosing regimen.) Our goal would be to find those values of b which we can reject and thus use to focus our general theory. In this approach, we require much more specific information and extensive testing to feel comfortable with so general a theory. We view the paper by WBZ as a
Goddard and Krewski
cautionary note in this vein: we do not have sufficient detailed information to safely exclude some important aspects of the general allometric law, notably either b = 0.75 o r b = 0.67. The main source of variability in the approach used by WBZ is the standard deviation of a log-normal distribution, the SDL.In standardizing the data, the authors (of necessity, and following the practice of many) make numerous conversions and assumptions employing constants. For example, body weights are constant, as are dose conversions to the five day dosing regimen. To allow for extra variability, several analyses are performed with SDLmultiplied by a factor of up to threefold, which is justified by an ad hoc rather than a strictly scientific explanation. We observe that predictions are based on multiplying factors, and in such situations, variability propagates exponentially. Thus, it is possible that the true variability may exceed the values presented by WBZ,when allowance can be made for the variability of each component in the prediction. The paper can also be viewed as a contribution to the emerging field of meta-analy~is,('~J~) a science involving the combination of results from several experiments. The aim of many meta-analyses is to combine several investigations of a single hypothesis into a larger investigation of the same hypothesis. In this study, as with those of Freireich et al. ,(13) and Schein et al. ,(14) the purpose is to study a separate hypothesis that is not found in any of the constituent reports. Combining similar studies of one hypothesis into a large test of the same hypothesis involves many assumptions and corrections to yield results that one would expect from experiments using a common protocol. Combining reports to study an issue differing from the hypotheses of the component studies requires equally stringent comparability along with the assumptions and corrections to create comparable models from components that are logically or physically distinct. For example, one needs to make certain assumptions or corrections about dosing regimens and study durations to combine information from several investigations into the LD,, of mice to FUDR. Additional assumptions and corrections (all of which contribute potential variability) are required if one attempts to combine information from both rat and mouse studies. We were struck by the variation in LD,,s between Tables I and I11 for the 11 chemicals in the latter table. Presumably, the values in Table I are drawn from published tables, and values in Table 111were obtained from reanalysis of the Swiss mouse data. Some differences were strikingly large, notably myleran with 15.0 mgkg and 66.9 m a g . The authors use Table 111 as evidence in support of the log-normal assumption of LD,,s. It
Interspecies Extrapolation
may also provide evidence for variability due to the estimation procedure used to estimate the LDl0. In conclusion, we are generally sympathetic to the thrust of the argument put forth by WBZ. Given the paucity and variability of data, we are not very far along in our understanding of allometry in the context of interspecies extrapolation. The paper represents an important cautionary note about possibly rejecting specific hypothesis with the data to hand. At the same time, arguments can be made against the tendency to invoke conservative assumptions whenever uncertainty arises is risk assessment. Conservatism is prudent from the point of view of public health, but has been criticized(20)as leading to excessive economic penalties. If a best estimate of the interspecies scaling factor b is desired, two analyses by WBZ leads to essentially the same value (b = 3/4) as was obtained earlier by Travis and White(12); however, the third WBZ analysis, that for the “reliable data”, yields an estimate of 0.62, lower than the most conservative value (b = 2/3) currently in use. It is clear that further work is required to make progress in allometric scaling of toxicity data between species. Empirical studies such as the present are useful, but are limited in that they do not directly address all of the factors responsible for species differences in toxic response.(21.22)In this regard, physiologically based pharmacokinetic models that allow for the use of species specific metabolic parameters (some of which can be determined by direct measurement in humans) may prove useful in predicting the dose of toxic metabolites reaching target tissues in different species.c’) Other important differences such as tissue growth rates and cell kinetics also need to be considered in extrapolating estimates of carcinogenic potency between species.(=) Because of such differences, Health and Welfare Canada(=) conclude that “major advances in the biology, and especially, the molecular biology of the carcinogenic process are needed before accurate prediction of human-active agents will be feasible”.
REFERENCES 1. D. B. Clayson and D. Krewski. “Interpretation and Extrapolation of Toxicological Data,” In Handbook of In Hvo Tadciiy Testing (D. L. Arnold, H. Grice and D. Krewski, eds.). Academic Press, New York, pp. 643-663 (1990). 2. Tomatis, L., Aitio, A., Wilbourn, J. and Shuker, L. Human carcinogens so far identified, Japanese Journal of Cancer Research, 80, 795-807 (1989). 3. J. K. Haseman and J. E. Huff, “Species Correlation in Longterm Carcinogenicity Studies,” Cancer Letters 37, 125-132 (1987).
317 4. Lave, L. B., Ennever, F. K., Rosenkranz, H. S. and Omenn, G. S. Information value of the rodent bioassay, Nature, 336, 631-3 (1988). 5. Piegorsch, W. W., Carr, G. J., Portier, C. J. and Hoel, D. G. “Concordance of carcinogenic response between rodent species: Potency dependence and potential underestimation”, Risk Analysis, 12, 115-121 (1992). 6. K. Watanabe, F. Y. Bois, and L. Zeise, “Interspecies Extrapolation: A Reexamination of Acute Toxicity Data,” Risk Analysis 12,301-310 (1992). 7. U.S. Environmental Protection Agency, “Guidelines for Carcinogenic Risk Assessment,” Federal Regkter 51, 33992-34003 (1986). 8. Krewski, D., Goddard, M. J., and Withey, J. Carcinogenic potency and interspecies extrapolation in Mutation and the Environment Par? D: Carcinogenesis, Mendelsohn, M. L. and Albertini, R. J. (Eds), Wiley-Liss, New York, NY,pp 323-334 (1990). 9. D. W. Gaylor and J. J. Chen, “Relative Potency of Chemical Carcinogens in Rodents,” Risk Analysis 6, 283-290 (1986). 10. B. C. Allen, K. S. Crump, and A. Shipp, “Correlation Between Carcinogenic Potency of Chemicals in Animals and Humans,” Risk Analysis 8,531-544 (1988). 11. Kaldor, J. M., Day, N. E., and Hemminki, K. Quantifying the carcinogenicityof antineoplasticdrugs. European Journal of Cancer 24, 703-711 (1988). 12. C. C. Travis and R. K.White, “Interspecific (SPECIES) Scaling of Toxicity Data,” Risk Analysis 8, 119-125 (1988). 13. E. J. Frierich, E. A. Gehan, D. P. Rall, L. H. Schmidt, and H. E. Skipper, “Quantitative Comparison of Toxicity of Anticancer Agents in Mouse, Rat, Hamster, Dog, Monkey, and Man,” Cancer Chemotherapy Reports 50, 21!3-244 (1966). 14. P. S. Schein, R. Davis, S. Carter, J. Newman, D. R. Schein, and D. P. Rall, “The Evaluation of Anticancer Drugs in Dogs and Monkeys for the Prediction of QualitativeToxicities in Man,” Clinical P h a m and Ther. 11, 340 (1979). 15. K. Schmidt-Nielsen, “Scaling,” Why is animal size so important? Cambridge University Press, Cambridge (1984). 16. C. C. Travis and J. C. Bowers, “Interspecies Scaling of Anesthetic Potency,” Toxicology and Industrial Health 7 , 249-260 (1991). 17. K. Popper, The Logic of Scientific Knowledge, Harper Row, New York, N.Y. (1968). 18. C. Mann, “Meta-Analysis in the Breech,” Science, 249, 476480 (1990). 19. L. Hedges and I. Olkin, “Statistical Methods for Meta-Analysis,” Academic Press, New York, N.Y. (1985). 20. Executive Office of the President of the United States (1990) Current regulatory issues in risk assessment and risk management, in Regulatory Program of the United States Government, April I,
1990 to March 31, 1991. 21. I. F. H. Purchase, “Carcinogenic Risk Assessment: Are Animals Good Surrogatesfor Man?,” In Cancer Rish: Strategiesfor Elimination (P. Bannasch, ed.). Springer-Verlag, Berlin, pp. 65-79 (1987). 22. Oser, B. The rat as a model for human toxicological evaluation, Journal of Taxicology and Environmental Health, 8, 521-542 (1981). 23. D. Krewski, J. Withey, L. F. Ku, and C. C. Travis, “Physiologically Based Pharmacokinetic Models. Applications in Carcinogenic Risk Assessment,” In New Trends in Phannacokinetics, (A. Rescigno and A. K. Thakur eds), Plenum Press, New York (1991). ,-- -,24. Krewski, D., Goddard, M. J., and Zielinski, J. Dose-response relationships in carcinogenesis, in Mechanism of Carcinogenesis in Identijkation of Rish, IARC Monographs, Lyon, to appear (1992). 25. Health and Welfare, Canada Carcinogen Assessment, Supply and Services, Canada (1992).
Comment
Interspecies Extrapolation of Toxicity Data1 Michael J. Goddard2 and Daniel Krewski2 Toxicological tests using nonhuman test systems remain an important source of information on potential human health hazards(’). Extrapolation of toxicological test results to the human situation is based on the premise that mammalian species respond in a similar fashion to toxic agents. All 53 agents known to cause cancer in humans have also been shown to cause cancer in one or more animal species.(2) Chemicals that cause cancer in mice tend to cause cancer in rats with concordance for chemicals tested in the U.S. National Toxicological Program of about 74%.(SLave et al.(4)suggested that this level of concordance between mice and rats represents an upper bound on concordance between either species and humans. Recent results by Piegorsch ef al.@)have shown that estimates of species concordance depends on carcinogenic potency, and, for low potency carcinogens, the maximum possible species concordance may only be about 80%. Imperfect qualitative agreement between species suggests the need for caution in extrapolating quantitatively between species. In order to address biological differences between species, predictions of human health risk based on the results of toxicological findings in laboratory animals need to take into consideration interspecies differences in physiological, pharmacokinetic, and pharmacodynamic factors associated with the induction of toxic effects. Allometric relationships among physiological parameters in different mammalian species have been observed by a number of investigators. For example, heat loss appears to be related to the surface area of mammals while metabolism is more closely related to body weight. To the extent that such factors mediate toxicological response, they should be adjusted for in interspecies extrapolation. l
Watanabe, Bois and Zeise,c6) hereafter referred to as WBZ, consider the allometric equation
P=axBW’, where P represents a measure of toxic potency in a given species, BW denotes the body weight of that species, and a and b are allometric constants. The U. S. Food & Drug Administration currently uses a value of b = 1, corresponding to interspecies extrapolation on the basis of body weight, whereas the U.S. Environmental Protection A g e n d a employsb = 2/3, correspondingroughly to interspecies extrapolation on the basis of body surface area. In order that b = 213 corresponds precisely to body surface area, it is necessary to assume that toxicological reactions occur on surfaces and different mammalian species have roughly the same shape and density.(*)Surface area extrapolation is more conservative than body weight extrapolation, with rats and mice leading to predictions of human risk that are approximately 6 and 13 times lower based on surface area than body weight. There also appears to be some correlation between the potency (as measured by the TD50, or the dose that induces an excess tumor risk of 50%) of these chemicals as expressed in these two species.(9)A subsequent empirical cornparison(’O) of the carcinogenic potency of 23 chemicals for which both human and animal data are available suggested that carcinogenic potency in animals is correlated with carcinogenic potency in humans. While demonstrating correlation coefficients ranging as high as r = 0.9, the uncertainty associated with predictions of carcinogenicpotency in humans was large. Moreover, the use of both body weight and surface area to convert between species yielded comparable correlation coefficients. Kaldor ef al. (11) obtained estimates of the carcinogenic potency of a small number of antineoplastic drugs that increase the risk of secondary tumours (acute nonlymphocytic leukemia) in humans, along with corresponding estimates for animals. The authors found the results encouraging in terms of using animal data to predict human potency rankings, but cautioned against
Comment on “Interspecies Extrapolation: A Reexaminationof Acute Toxicity Data,” by Watanabe, Bois, and Zeise in Risk Analysis, Vol. 12, No. 2 (June 1992), pp. 301-310. Environmental Health Directorate, Health Protection Branch, Health and Welfare Canada, Ottawa, Ontario, Canada K1A OL2.
315
02724332/92/0600031.5$M.~0/10 1992 Society for Risk Analyuo
316
quantitative prediction of human carcinogenic potency on the basis of these data. Travis and White(12)re-examined the original data reported by Frierich et al. (13)and Schein et al. (14)on the toxicity of anticancer drugs in animals and humans, and concluded that a value of b = 0.75 provides an overall best fit to these data. WBZ note that .this corresponds roughly to species extrapolation on the basis of metabolic rate (c.f Schmidt-Niel~en).('~)This value is intermediate between the value of b = 1 used by the Food and Drug Administration and the value of b = 2/3 used by the Environmental Protection Agency, and has been suggested as a compromise value towards which both agencies may possibly converge. Based on an analysis of the anesthetic potency of 11volatile anesthetics, Travis and Bower@) offer further support for the use of a scaling factor of b = 3/4 with respect to acute toxic effects. WBZ have conducted a further re-analysis of the data by Freirich et al. (13) and Schein et al. (14) in order to determine the most appropriate value of b to use in species extrapolation of toxicity. The novelty of this analysis is the provision for measurement error in the original data; whereas Travis and White(12)assumed that the LD,, and the MTD values were known without error, WBZ allowed for uncertainty in these estimate both by reviewing the variability in the original data and by allowing for an additional uncertainty factor U in the range 1-3 fold. Although the best estimate of the value of b generally remained close to the original estimate of b = 314 obtained by Travis and White,('*) the introduction of additional uncertainty in the analysis led to wider confidence limits on b which now include b = 2/3. WBZ conclude that surface area scaling should thus be used in the interests of prudence. It is illuminating to consider this conclusion in light of Popper's('') paradigm for scientific discovery. Popper's approach is that of firstly embracing a general theory and then testing this by attempting to "falsify" very specific hypotheses. Herein, the general theory might be that equation (3) of WBZ with a specific value of b is valid for all species and chemicals. We add to our knowledge by testing a particular instance. For example, we might study the hypothesis that, for amethopterin, the allometric constant b = 0.75 is valid for mice. (Yet more specifics are required even here including, among others, study duration, route of exposure, and dosing regimen.) Our goal would be to find those values of b which we can reject and thus use to focus our general theory. In this approach, we require much more specific information and extensive testing to feel comfortable with so general a theory. We view the paper by WBZ as a
Goddard and Krewski
cautionary note in this vein: we do not have sufficient detailed information to safely exclude some important aspects of the general allometric law, notably either b = 0.75 o r b = 0.67. The main source of variability in the approach used by WBZ is the standard deviation of a log-normal distribution, the SDL.In standardizing the data, the authors (of necessity, and following the practice of many) make numerous conversions and assumptions employing constants. For example, body weights are constant, as are dose conversions to the five day dosing regimen. To allow for extra variability, several analyses are performed with SDLmultiplied by a factor of up to threefold, which is justified by an ad hoc rather than a strictly scientific explanation. We observe that predictions are based on multiplying factors, and in such situations, variability propagates exponentially. Thus, it is possible that the true variability may exceed the values presented by WBZ,when allowance can be made for the variability of each component in the prediction. The paper can also be viewed as a contribution to the emerging field of meta-analy~is,('~J~) a science involving the combination of results from several experiments. The aim of many meta-analyses is to combine several investigations of a single hypothesis into a larger investigation of the same hypothesis. In this study, as with those of Freireich et al. ,(13) and Schein et al. ,(14) the purpose is to study a separate hypothesis that is not found in any of the constituent reports. Combining similar studies of one hypothesis into a large test of the same hypothesis involves many assumptions and corrections to yield results that one would expect from experiments using a common protocol. Combining reports to study an issue differing from the hypotheses of the component studies requires equally stringent comparability along with the assumptions and corrections to create comparable models from components that are logically or physically distinct. For example, one needs to make certain assumptions or corrections about dosing regimens and study durations to combine information from several investigations into the LD,, of mice to FUDR. Additional assumptions and corrections (all of which contribute potential variability) are required if one attempts to combine information from both rat and mouse studies. We were struck by the variation in LD,,s between Tables I and I11 for the 11 chemicals in the latter table. Presumably, the values in Table I are drawn from published tables, and values in Table 111were obtained from reanalysis of the Swiss mouse data. Some differences were strikingly large, notably myleran with 15.0 mgkg and 66.9 m a g . The authors use Table 111 as evidence in support of the log-normal assumption of LD,,s. It
Interspecies Extrapolation
may also provide evidence for variability due to the estimation procedure used to estimate the LDl0. In conclusion, we are generally sympathetic to the thrust of the argument put forth by WBZ. Given the paucity and variability of data, we are not very far along in our understanding of allometry in the context of interspecies extrapolation. The paper represents an important cautionary note about possibly rejecting specific hypothesis with the data to hand. At the same time, arguments can be made against the tendency to invoke conservative assumptions whenever uncertainty arises is risk assessment. Conservatism is prudent from the point of view of public health, but has been criticized(20)as leading to excessive economic penalties. If a best estimate of the interspecies scaling factor b is desired, two analyses by WBZ leads to essentially the same value (b = 3/4) as was obtained earlier by Travis and White(12); however, the third WBZ analysis, that for the “reliable data”, yields an estimate of 0.62, lower than the most conservative value (b = 2/3) currently in use. It is clear that further work is required to make progress in allometric scaling of toxicity data between species. Empirical studies such as the present are useful, but are limited in that they do not directly address all of the factors responsible for species differences in toxic response.(21.22)In this regard, physiologically based pharmacokinetic models that allow for the use of species specific metabolic parameters (some of which can be determined by direct measurement in humans) may prove useful in predicting the dose of toxic metabolites reaching target tissues in different species.c’) Other important differences such as tissue growth rates and cell kinetics also need to be considered in extrapolating estimates of carcinogenic potency between species.(=) Because of such differences, Health and Welfare Canada(=) conclude that “major advances in the biology, and especially, the molecular biology of the carcinogenic process are needed before accurate prediction of human-active agents will be feasible”.
REFERENCES 1. D. B. Clayson and D. Krewski. “Interpretation and Extrapolation of Toxicological Data,” In Handbook of In Hvo Tadciiy Testing (D. L. Arnold, H. Grice and D. Krewski, eds.). Academic Press, New York, pp. 643-663 (1990). 2. Tomatis, L., Aitio, A., Wilbourn, J. and Shuker, L. Human carcinogens so far identified, Japanese Journal of Cancer Research, 80, 795-807 (1989). 3. J. K. Haseman and J. E. Huff, “Species Correlation in Longterm Carcinogenicity Studies,” Cancer Letters 37, 125-132 (1987).
317 4. Lave, L. B., Ennever, F. K., Rosenkranz, H. S. and Omenn, G. S. Information value of the rodent bioassay, Nature, 336, 631-3 (1988). 5. Piegorsch, W. W., Carr, G. J., Portier, C. J. and Hoel, D. G. “Concordance of carcinogenic response between rodent species: Potency dependence and potential underestimation”, Risk Analysis, 12, 115-121 (1992). 6. K. Watanabe, F. Y. Bois, and L. Zeise, “Interspecies Extrapolation: A Reexamination of Acute Toxicity Data,” Risk Analysis 12,301-310 (1992). 7. U.S. Environmental Protection Agency, “Guidelines for Carcinogenic Risk Assessment,” Federal Regkter 51, 33992-34003 (1986). 8. Krewski, D., Goddard, M. J., and Withey, J. Carcinogenic potency and interspecies extrapolation in Mutation and the Environment Par? D: Carcinogenesis, Mendelsohn, M. L. and Albertini, R. J. (Eds), Wiley-Liss, New York, NY,pp 323-334 (1990). 9. D. W. Gaylor and J. J. Chen, “Relative Potency of Chemical Carcinogens in Rodents,” Risk Analysis 6, 283-290 (1986). 10. B. C. Allen, K. S. Crump, and A. Shipp, “Correlation Between Carcinogenic Potency of Chemicals in Animals and Humans,” Risk Analysis 8,531-544 (1988). 11. Kaldor, J. M., Day, N. E., and Hemminki, K. Quantifying the carcinogenicityof antineoplasticdrugs. European Journal of Cancer 24, 703-711 (1988). 12. C. C. Travis and R. K.White, “Interspecific (SPECIES) Scaling of Toxicity Data,” Risk Analysis 8, 119-125 (1988). 13. E. J. Frierich, E. A. Gehan, D. P. Rall, L. H. Schmidt, and H. E. Skipper, “Quantitative Comparison of Toxicity of Anticancer Agents in Mouse, Rat, Hamster, Dog, Monkey, and Man,” Cancer Chemotherapy Reports 50, 21!3-244 (1966). 14. P. S. Schein, R. Davis, S. Carter, J. Newman, D. R. Schein, and D. P. Rall, “The Evaluation of Anticancer Drugs in Dogs and Monkeys for the Prediction of QualitativeToxicities in Man,” Clinical P h a m and Ther. 11, 340 (1979). 15. K. Schmidt-Nielsen, “Scaling,” Why is animal size so important? Cambridge University Press, Cambridge (1984). 16. C. C. Travis and J. C. Bowers, “Interspecies Scaling of Anesthetic Potency,” Toxicology and Industrial Health 7 , 249-260 (1991). 17. K. Popper, The Logic of Scientific Knowledge, Harper Row, New York, N.Y. (1968). 18. C. Mann, “Meta-Analysis in the Breech,” Science, 249, 476480 (1990). 19. L. Hedges and I. Olkin, “Statistical Methods for Meta-Analysis,” Academic Press, New York, N.Y. (1985). 20. Executive Office of the President of the United States (1990) Current regulatory issues in risk assessment and risk management, in Regulatory Program of the United States Government, April I,
1990 to March 31, 1991. 21. I. F. H. Purchase, “Carcinogenic Risk Assessment: Are Animals Good Surrogatesfor Man?,” In Cancer Rish: Strategiesfor Elimination (P. Bannasch, ed.). Springer-Verlag, Berlin, pp. 65-79 (1987). 22. Oser, B. The rat as a model for human toxicological evaluation, Journal of Taxicology and Environmental Health, 8, 521-542 (1981). 23. D. Krewski, J. Withey, L. F. Ku, and C. C. Travis, “Physiologically Based Pharmacokinetic Models. Applications in Carcinogenic Risk Assessment,” In New Trends in Phannacokinetics, (A. Rescigno and A. K. Thakur eds), Plenum Press, New York (1991). ,-- -,24. Krewski, D., Goddard, M. J., and Zielinski, J. Dose-response relationships in carcinogenesis, in Mechanism of Carcinogenesis in Identijkation of Rish, IARC Monographs, Lyon, to appear (1992). 25. Health and Welfare, Canada Carcinogen Assessment, Supply and Services, Canada (1992).
