- Email: [email protected]
A hierarchical approach to the evaluation of chemicals for estrogenic and other endocrine-disrupting properties
Environmental Toxicology and Pharmacology 3 (1997) 87 – 90
A hierarchical approach to the evaluation of chemicals for estrogenic and other endocrine-disrupting properties John Ashby * Zeneca Central Toxicology Laboratory, Alderley Park, Cheshire, UK
Abstract The emergence of estrogenicity/endocrine-disruption as an important endpoint in the toxicological assessment of chemicals presents a series of problems to overcome before regulatory control of such agents can be enacted. A framework is presented by which progress in this endeavour can be expedited. A hierarchial approach to testing is proposed, together with consideration of the types of information required to transform test data into human risk estimations. The approach is based broadly on current methods for defining potential human carcinogens and mutagens, and if found acceptable, would dramatically accelerate regulatory progress on this subject. However, several questions must be answered, using focused data, before the approach can be endorsed or transferred into a regulatory context. The importance of early consideration of all aspects of this complex new toxicity, including the unexpected observation of synergism between synthetic estrogens, is emphasized. © 1997 Elsevier Science B.V.
Concern that environmental chemicals may be affecting the reproductive capacity or sexual development of wildlife has arisen from events such as the feminization of alligators and fish living in polluted waters. Links have also been speculated to exist between reports of falling human sperm counts and exposure of humans to estrogenic environmental chemicals, and between increases in the incidences of human mammary, prostate and testicular cancer and exposure to unknown environmental endocrine disrupting (ED) chemicals (Colborn and Clement, 1992; McLachlan and Korach, 1995; Chapin et al., 1996). These concerns have been crystalized by the demonstration that exposure of pregnant rats to alkyl phenol polyethoxylate detergents and the plasticizer butyl benzyl phthalate leads to a reduction in testis size and sperm counts in the offspring (Sharpe et al., 1995). At present, the potential of chemicals to elicit such toxicities is only assessed for major environmental chemicals via the conduct of resource consuming, albeit definitive, rodent multi-generation reproductive studies (Brown et al., 1995). Extrapolation of the limited amount of data currently available indi* Corresponding author. Fax: +44 0 1625590249. 1382-6689/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S 1 3 8 2 - 6 6 8 9 ( 9 7 ) 0 0 0 9 2 - 6
cates that a large number of naturally occurring and synthetic chemicals will elicit some level of ED effects in at least some experimental systems. Further, an unexpected complication to this already complex picture was recently provided by the demonstration of synergism between synthetic estrogens when evaluated in vitro (Arnold et al., 1996). Given all of this, the challenge becomes to expedite the identification and control of human and wildlife exposure to the most important environmental ED agents and to potentially synergizing mixtures. This will not be an easy task given the large number of chemicals involved and the many levels at which chemicals may interfere with normal endocrine function. A framework is suggested here for how this daunting task can be approached (Scheme 1). The major principle underpinning the Scheme is based on the current approach to the assessment of chemicals for genetic toxicity/carcinogenic potential. Namely, that the evaluation process should commence in vitro, using the minimum required number of assays, followed by assessment in vivo of activities observed in vitro (Ashby and Morrod, 1991). If this approach can be transferred to the study of ED activities it will greatly simplify the task ahead. Equally, any
88
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
major defects in this adapted approach are best recognized at an early stage. Whichever outcome emerges, it is best to approach this new area mindful of the results of 20 years of endeavour in the former area. The first quartile of the Scheme concerns the identification in vitro of any intrinsic ED activities that a chemical may possess. Many biological events/changes could be adapted to serve as in vitro assays for ED activity, ranging from the binding of chemicals to isolated receptors through to the stimulation into division of estrogen-dependent MCF-7 cells (Soto et al., 1995). Initial work in this area has tended to concentrate on elaboration of a series of receptor assays based in a range of different cells. McLachlan (1993) has identified a panel of at least nine relevant biological receptors that could be linked to a reporter gene and transfected to yield receptor-specific assays. Exactly how many such independent assays will be required is uncertain, but the need to consider at least the estrogen receptor and the androgen receptor is already evident (Kelse et al., 1996). However, two problems with this approach are already evident. The first is that incorporation of a source of auxiliary metabolic enzymes (e.g. rat liver S9 mix) has yet to be described for any of the current in vitro assays. This may prove to be a difficult requirement to meet due to the need to remove natural steroid hormones from the activating tissue homogenate (cf the need for charcoal stripping of the culture medium in the MCF-7 assay). The transfection of appropriate receptors into metabolically competent cells may sometimes offer a solution to this problem. The second problem is that the complexity of the reproductive process in mammals may make it difficult to model certain processes in vitro. Examples are provided by the chemical induction in male rodent pups of hypo- or hyper-thyroidism leading to interference with the regulation by thyroid hormone receptors of Sertoli cell division in the developing testes (Cooke and Hess, 1994; Cooke et al., 1996). Likewise, modulation of sex hormone metabolism in vivo may be impossible to anticipate based on tests conducted in vitro (Majdic et al., 1996). The basic knowledge of whether or not rodent assays are obligatory for the primary screening of chemicals for ED activities is best acquired at this early stage by the generation of appropriately focused data. Whatever battery of assays is eventually agreed upon for the primary evaluation of ED activities, it is relevant to note that over 100 in vitro assays were initially developed to assess the mutagenicity of chemicals, and that these were eventually culled to four or five robust complementary assays following the demonstration of massive redundancy. That error, caused by an initial excess of unfocused enthusiasm, would be best avoided during the development of assays for ED activity.
Experience to date indicates that a substantial number of structurally diverse chemicals will possess ED activities in vitro. However, a chasm exists between showing that a chemical can induce MCF-7 cells to divide in culture, and the conclusion that it poses a significant hazard to the human endocrine system at expected levels of exposure. Again, by analogy, it is now generally accepted that although genetic toxicity in vitro is relatively common among synthetic and naturally occurring chemicals, not all such agents pose a carcinogenic or mutagenic hazard to humans. That principle influenced the design of the remaining three quartiles of Scheme 1. At the present time, the evaluation of a chemical for ED activities in vivo may be triggered either by the need to assess ED effects observed in vitro, or in order to reveal disturbances to the catabolism or anabolism of natural hormones, or other indirect ED effects that may be unique to whole animals. It will be necessary to accept, at least initially, that both genders of rodent are worthy of study—i.e., the inability of a chemical to affect the uterus may not preclude affects on testicular development, and vice versa. An important aspect of the study of ED activities in vivo is that the time of exposure of the embryo or neonate to the test chemical may prove to be critical to the outcome of the study. The list of experimentally observable changes to the rodent endocrine system shown in the second quartile of Scheme 1 is not exhaustive, and no standardized protocols currently exist for their measurement. For
Scheme 1. Framework for the development of a sequential approach to the evaluation of chemicals for endocrine disrupting potential in humans. SAR, structure activity relationships; QSAR, quantitarive SAR; ER and AR, estrogen and androgen receptors; PK, pharmacokinetics.
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
90
Cooke, P.S. and R.A. Hess, et al., 1994, in: Function of the Somatic Cells in the Testis, ed. A. Bartke (Springer Verlag, NY) pp. 400 – 407. Cooke, P.S., Y. Zhao and L.G. Hansen, 1996, Tox. Appl. Pharmacol. 136, 112 – 117. Kelse, W.R. and C. Wong, et al., 1996, Fund. Appl. Tox., 29, 9 – 11. Majdic, G. and R.M. Sharpe, et al., 1996, Endocrinology 137, 1063 – 1070. McLachlan, J.A., 1993, Environ. Health Perspect. 101, 386 – 387. McLachlan, J.A. and K.S. Korach, 1995, eds. Environ. Health Per-
.
.
spect. 103 (Suppl.7), pp. 3 – 178. Mellanen, P. and T. Peta´nen, et al., 1996, Tox. Appl. Pharmacol. 136, 381 – 388. Sharpe, R.M. and J.S. Fisher, et al., 1995, Environ. Health Perspect. 103, 1136 – 1143. Soto, A.M., C. Sonenschein, et al., 1995, Environ. Health Perspect. 103 (Suppl.7), 113 – 122. Sumpter, J.P. and S. Jobling, 1995, Environ. Health Perspect. 103, (Suppl.7), 173 – 178. Wakeling, A.E., 1995, Mutat. Res. 333, 45 – 50.
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
example, the rodent uterotrophic assay, which measures the ability of a chemical to promote uterine growth in estrogen-deprived animals, can be conducted using any of a wide variety of test protocols. The major variables encountered are the use of either rats or mice, oral or subcutaneous injection of the test agent, and ovariectomized, hypophysectomized, or immature animals. Use of immature animals is practically the most convenient, but this precludes the study of other estrogen-dependent changes, e.g., to peripheral lipid levels or bone density (Black et al., 1994). It is also important to note that at present there are no agreed criteria for activity in these assays and that different biological events may trigger the same increase in uterine weight (e.g., water retention versus induced cell division; Wakeling, 1995). Such a range of uncertainties will probably beset the use of all ED assays, and this indicates that conclusions of the activity or inactivity of a chemical should currently be drawn with caution. These several uncertainties also suggest that studies to modify or refine the basic testing strategy shown in Scheme 1 should be commissioned with a view to gaining agreement on a core set of reliable in vitro and in vivo assays and agreed criteria for their conduct and interpretation. Chemicals shown to possess significant ED activities in vivo should next be assessed for the hazard they might pose to humans. The major factors listed on the left side of Scheme 1 are known to affect critically the prediction of human hazard for rodent carcinogens, and they are likely to apply equally to the assessment of the hazard posed to humans by a chemical possessed of ED activities. At this point, in the evaluation, it would also be appropriate to consider the potential relevance to wildlife of the effects observed in rodents. This can be achieved by the use of oviparous test species in conjunction with assessment of the environmental fate/ bioaccumulation of the test agent (Sumpter and Jobling, 1995). Some agents may pose a unique ED hazard to either mammals or oviparous species, but there seems to be no case for introducing this further complexity at the present time. This means that a unified approach to the prediction and assessment of ED activities for both humans and wildlife can be adopted for the present. The hierarchical testing strategy implied in Scheme 1 was implicitly employed in a recent study to detect the active estrogenic agent in wood debarking effluent (Mellanen et al., 1996). Eleven suspect chemicals were tested in the MCF-7 in vitro assay, and seven were found to be active. Of these, only two were found to be active in an in vivo fish vitellogenin assay, one of which (b-sitosterol) was concluded to be the active contaminant based on the potency of its activity in vivo and on the independence of this activity from the route of administration to the fish.
89
The final quartile of Scheme 1 is concerned with human risk assessment. This is a complex task involving the description and quantitation of exposure and the extrapolation across species of dose-response data derived from rodent studies. In appropriate cases the human risk projected might warrant the conduct of human epidemiological studies, but that will be rare. In most cases estimates of human hazard will be made based on an integrated assessment of the totality of the available data linked to an appreciation that the mechanism of action postulated to explain the effects seen in the experimental systems will apply to humans. This final stage of the hazard assessment process, ending in a risk assessment for humans or wildlife, is the single goal to pursue. However, it’s intrinsic difficulty may lead to it being neglected in favour of the easier task of testing new chemicals for ED activities in vitro. The recent demonstration of synergism between a range of different ED agents (Arnold et al., 1996) brings with it the potential for limitless speculations and studies. It is suggested that while the unexpected observation of synergism is of undoubted importance, initial attention should be focused on the derivation of an agreed approach to the definition and assessment of ED activity for individual agents. If synergism is to be studied in it’s own right at this early stage, it would be best to concentrate on the activities of representative environmental samples, rather than on the limitless number of artificial mixtures that could be made in the laboratory. Finally, it is relevant to take into consideration that the US National Toxicology Program has invested hundreds of millions of dollars over the past 20 years to establish the carcinogenicity to rodents of only 400 chemicals. That rate of progress indicates that a more streamlined approach must be developed for the prediction and assessment of ED hazards to humans. Acceptance of the magnitude of the task faced, as indicated in Scheme 1, is suggested to form the first critical step towards achievement of that goal. In the absence of such an holistic approach, uninterpretable indications of ED activity will accumulate for an increasing number of chemicals, coupled to irrational reactions to politically sensitive issues, such as with recent concerns over the safety of infant food formulae. References Arnold, S.F. and D.M. Klotz, et al., 1996, Science 272, 1489–1492. Ashby, J. and R.S. Morrod, 1991, Nature 352, 185 – 186. Black, L.J. and M. Sato, et al., 1994, J. Clin. Invest. 93, 63–39. Brown, N.A. and H. Speilmann, et al., 1995, ATLA, 23, 868–882. Chapin, R.E. and J.T. Stevens, et al., 1996, Fund. Appl. Toxicol. 29, 1 – 17. Colborn, T. and C. Clement, 1992, eds., in: Chemically-induced Alterations in Sexual and Functional Development (Princetown Scientific Publishing).
A hierarchical approach to the evaluation of chemicals for estrogenic and other endocrine-disrupting properties John Ashby * Zeneca Central Toxicology Laboratory, Alderley Park, Cheshire, UK
Abstract The emergence of estrogenicity/endocrine-disruption as an important endpoint in the toxicological assessment of chemicals presents a series of problems to overcome before regulatory control of such agents can be enacted. A framework is presented by which progress in this endeavour can be expedited. A hierarchial approach to testing is proposed, together with consideration of the types of information required to transform test data into human risk estimations. The approach is based broadly on current methods for defining potential human carcinogens and mutagens, and if found acceptable, would dramatically accelerate regulatory progress on this subject. However, several questions must be answered, using focused data, before the approach can be endorsed or transferred into a regulatory context. The importance of early consideration of all aspects of this complex new toxicity, including the unexpected observation of synergism between synthetic estrogens, is emphasized. © 1997 Elsevier Science B.V.
Concern that environmental chemicals may be affecting the reproductive capacity or sexual development of wildlife has arisen from events such as the feminization of alligators and fish living in polluted waters. Links have also been speculated to exist between reports of falling human sperm counts and exposure of humans to estrogenic environmental chemicals, and between increases in the incidences of human mammary, prostate and testicular cancer and exposure to unknown environmental endocrine disrupting (ED) chemicals (Colborn and Clement, 1992; McLachlan and Korach, 1995; Chapin et al., 1996). These concerns have been crystalized by the demonstration that exposure of pregnant rats to alkyl phenol polyethoxylate detergents and the plasticizer butyl benzyl phthalate leads to a reduction in testis size and sperm counts in the offspring (Sharpe et al., 1995). At present, the potential of chemicals to elicit such toxicities is only assessed for major environmental chemicals via the conduct of resource consuming, albeit definitive, rodent multi-generation reproductive studies (Brown et al., 1995). Extrapolation of the limited amount of data currently available indi* Corresponding author. Fax: +44 0 1625590249. 1382-6689/97/$17.00 © 1997 Elsevier Science B.V. All rights reserved. PII S 1 3 8 2 - 6 6 8 9 ( 9 7 ) 0 0 0 9 2 - 6
cates that a large number of naturally occurring and synthetic chemicals will elicit some level of ED effects in at least some experimental systems. Further, an unexpected complication to this already complex picture was recently provided by the demonstration of synergism between synthetic estrogens when evaluated in vitro (Arnold et al., 1996). Given all of this, the challenge becomes to expedite the identification and control of human and wildlife exposure to the most important environmental ED agents and to potentially synergizing mixtures. This will not be an easy task given the large number of chemicals involved and the many levels at which chemicals may interfere with normal endocrine function. A framework is suggested here for how this daunting task can be approached (Scheme 1). The major principle underpinning the Scheme is based on the current approach to the assessment of chemicals for genetic toxicity/carcinogenic potential. Namely, that the evaluation process should commence in vitro, using the minimum required number of assays, followed by assessment in vivo of activities observed in vitro (Ashby and Morrod, 1991). If this approach can be transferred to the study of ED activities it will greatly simplify the task ahead. Equally, any
88
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
major defects in this adapted approach are best recognized at an early stage. Whichever outcome emerges, it is best to approach this new area mindful of the results of 20 years of endeavour in the former area. The first quartile of the Scheme concerns the identification in vitro of any intrinsic ED activities that a chemical may possess. Many biological events/changes could be adapted to serve as in vitro assays for ED activity, ranging from the binding of chemicals to isolated receptors through to the stimulation into division of estrogen-dependent MCF-7 cells (Soto et al., 1995). Initial work in this area has tended to concentrate on elaboration of a series of receptor assays based in a range of different cells. McLachlan (1993) has identified a panel of at least nine relevant biological receptors that could be linked to a reporter gene and transfected to yield receptor-specific assays. Exactly how many such independent assays will be required is uncertain, but the need to consider at least the estrogen receptor and the androgen receptor is already evident (Kelse et al., 1996). However, two problems with this approach are already evident. The first is that incorporation of a source of auxiliary metabolic enzymes (e.g. rat liver S9 mix) has yet to be described for any of the current in vitro assays. This may prove to be a difficult requirement to meet due to the need to remove natural steroid hormones from the activating tissue homogenate (cf the need for charcoal stripping of the culture medium in the MCF-7 assay). The transfection of appropriate receptors into metabolically competent cells may sometimes offer a solution to this problem. The second problem is that the complexity of the reproductive process in mammals may make it difficult to model certain processes in vitro. Examples are provided by the chemical induction in male rodent pups of hypo- or hyper-thyroidism leading to interference with the regulation by thyroid hormone receptors of Sertoli cell division in the developing testes (Cooke and Hess, 1994; Cooke et al., 1996). Likewise, modulation of sex hormone metabolism in vivo may be impossible to anticipate based on tests conducted in vitro (Majdic et al., 1996). The basic knowledge of whether or not rodent assays are obligatory for the primary screening of chemicals for ED activities is best acquired at this early stage by the generation of appropriately focused data. Whatever battery of assays is eventually agreed upon for the primary evaluation of ED activities, it is relevant to note that over 100 in vitro assays were initially developed to assess the mutagenicity of chemicals, and that these were eventually culled to four or five robust complementary assays following the demonstration of massive redundancy. That error, caused by an initial excess of unfocused enthusiasm, would be best avoided during the development of assays for ED activity.
Experience to date indicates that a substantial number of structurally diverse chemicals will possess ED activities in vitro. However, a chasm exists between showing that a chemical can induce MCF-7 cells to divide in culture, and the conclusion that it poses a significant hazard to the human endocrine system at expected levels of exposure. Again, by analogy, it is now generally accepted that although genetic toxicity in vitro is relatively common among synthetic and naturally occurring chemicals, not all such agents pose a carcinogenic or mutagenic hazard to humans. That principle influenced the design of the remaining three quartiles of Scheme 1. At the present time, the evaluation of a chemical for ED activities in vivo may be triggered either by the need to assess ED effects observed in vitro, or in order to reveal disturbances to the catabolism or anabolism of natural hormones, or other indirect ED effects that may be unique to whole animals. It will be necessary to accept, at least initially, that both genders of rodent are worthy of study—i.e., the inability of a chemical to affect the uterus may not preclude affects on testicular development, and vice versa. An important aspect of the study of ED activities in vivo is that the time of exposure of the embryo or neonate to the test chemical may prove to be critical to the outcome of the study. The list of experimentally observable changes to the rodent endocrine system shown in the second quartile of Scheme 1 is not exhaustive, and no standardized protocols currently exist for their measurement. For
Scheme 1. Framework for the development of a sequential approach to the evaluation of chemicals for endocrine disrupting potential in humans. SAR, structure activity relationships; QSAR, quantitarive SAR; ER and AR, estrogen and androgen receptors; PK, pharmacokinetics.
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
90
Cooke, P.S. and R.A. Hess, et al., 1994, in: Function of the Somatic Cells in the Testis, ed. A. Bartke (Springer Verlag, NY) pp. 400 – 407. Cooke, P.S., Y. Zhao and L.G. Hansen, 1996, Tox. Appl. Pharmacol. 136, 112 – 117. Kelse, W.R. and C. Wong, et al., 1996, Fund. Appl. Tox., 29, 9 – 11. Majdic, G. and R.M. Sharpe, et al., 1996, Endocrinology 137, 1063 – 1070. McLachlan, J.A., 1993, Environ. Health Perspect. 101, 386 – 387. McLachlan, J.A. and K.S. Korach, 1995, eds. Environ. Health Per-
.
.
spect. 103 (Suppl.7), pp. 3 – 178. Mellanen, P. and T. Peta´nen, et al., 1996, Tox. Appl. Pharmacol. 136, 381 – 388. Sharpe, R.M. and J.S. Fisher, et al., 1995, Environ. Health Perspect. 103, 1136 – 1143. Soto, A.M., C. Sonenschein, et al., 1995, Environ. Health Perspect. 103 (Suppl.7), 113 – 122. Sumpter, J.P. and S. Jobling, 1995, Environ. Health Perspect. 103, (Suppl.7), 173 – 178. Wakeling, A.E., 1995, Mutat. Res. 333, 45 – 50.
J. Ashby / En6ironmental Toxicology and Pharmacology 3 (1997) 87–90
example, the rodent uterotrophic assay, which measures the ability of a chemical to promote uterine growth in estrogen-deprived animals, can be conducted using any of a wide variety of test protocols. The major variables encountered are the use of either rats or mice, oral or subcutaneous injection of the test agent, and ovariectomized, hypophysectomized, or immature animals. Use of immature animals is practically the most convenient, but this precludes the study of other estrogen-dependent changes, e.g., to peripheral lipid levels or bone density (Black et al., 1994). It is also important to note that at present there are no agreed criteria for activity in these assays and that different biological events may trigger the same increase in uterine weight (e.g., water retention versus induced cell division; Wakeling, 1995). Such a range of uncertainties will probably beset the use of all ED assays, and this indicates that conclusions of the activity or inactivity of a chemical should currently be drawn with caution. These several uncertainties also suggest that studies to modify or refine the basic testing strategy shown in Scheme 1 should be commissioned with a view to gaining agreement on a core set of reliable in vitro and in vivo assays and agreed criteria for their conduct and interpretation. Chemicals shown to possess significant ED activities in vivo should next be assessed for the hazard they might pose to humans. The major factors listed on the left side of Scheme 1 are known to affect critically the prediction of human hazard for rodent carcinogens, and they are likely to apply equally to the assessment of the hazard posed to humans by a chemical possessed of ED activities. At this point, in the evaluation, it would also be appropriate to consider the potential relevance to wildlife of the effects observed in rodents. This can be achieved by the use of oviparous test species in conjunction with assessment of the environmental fate/ bioaccumulation of the test agent (Sumpter and Jobling, 1995). Some agents may pose a unique ED hazard to either mammals or oviparous species, but there seems to be no case for introducing this further complexity at the present time. This means that a unified approach to the prediction and assessment of ED activities for both humans and wildlife can be adopted for the present. The hierarchical testing strategy implied in Scheme 1 was implicitly employed in a recent study to detect the active estrogenic agent in wood debarking effluent (Mellanen et al., 1996). Eleven suspect chemicals were tested in the MCF-7 in vitro assay, and seven were found to be active. Of these, only two were found to be active in an in vivo fish vitellogenin assay, one of which (b-sitosterol) was concluded to be the active contaminant based on the potency of its activity in vivo and on the independence of this activity from the route of administration to the fish.
89
The final quartile of Scheme 1 is concerned with human risk assessment. This is a complex task involving the description and quantitation of exposure and the extrapolation across species of dose-response data derived from rodent studies. In appropriate cases the human risk projected might warrant the conduct of human epidemiological studies, but that will be rare. In most cases estimates of human hazard will be made based on an integrated assessment of the totality of the available data linked to an appreciation that the mechanism of action postulated to explain the effects seen in the experimental systems will apply to humans. This final stage of the hazard assessment process, ending in a risk assessment for humans or wildlife, is the single goal to pursue. However, it’s intrinsic difficulty may lead to it being neglected in favour of the easier task of testing new chemicals for ED activities in vitro. The recent demonstration of synergism between a range of different ED agents (Arnold et al., 1996) brings with it the potential for limitless speculations and studies. It is suggested that while the unexpected observation of synergism is of undoubted importance, initial attention should be focused on the derivation of an agreed approach to the definition and assessment of ED activity for individual agents. If synergism is to be studied in it’s own right at this early stage, it would be best to concentrate on the activities of representative environmental samples, rather than on the limitless number of artificial mixtures that could be made in the laboratory. Finally, it is relevant to take into consideration that the US National Toxicology Program has invested hundreds of millions of dollars over the past 20 years to establish the carcinogenicity to rodents of only 400 chemicals. That rate of progress indicates that a more streamlined approach must be developed for the prediction and assessment of ED hazards to humans. Acceptance of the magnitude of the task faced, as indicated in Scheme 1, is suggested to form the first critical step towards achievement of that goal. In the absence of such an holistic approach, uninterpretable indications of ED activity will accumulate for an increasing number of chemicals, coupled to irrational reactions to politically sensitive issues, such as with recent concerns over the safety of infant food formulae. References Arnold, S.F. and D.M. Klotz, et al., 1996, Science 272, 1489–1492. Ashby, J. and R.S. Morrod, 1991, Nature 352, 185 – 186. Black, L.J. and M. Sato, et al., 1994, J. Clin. Invest. 93, 63–39. Brown, N.A. and H. Speilmann, et al., 1995, ATLA, 23, 868–882. Chapin, R.E. and J.T. Stevens, et al., 1996, Fund. Appl. Toxicol. 29, 1 – 17. Colborn, T. and C. Clement, 1992, eds., in: Chemically-induced Alterations in Sexual and Functional Development (Princetown Scientific Publishing).