In recent years, there has been increasing concern regarding potential adverse human health effects of various environmental contaminants designated by some scientists as ''endocrine disruptors" (Thomas and Colborn 1992; Colborn et al. 1993) and referred to in this report as hormonally active agents (HAAs). These concerns have originated, in part, from observations of developmental and reproductive derangements in wildlife populations exposed to a wide range of synthetic chemicals and their by-products. Most notable are the adverse reproductive and developmental effects that have been observed in birds such as cormorants, herring gulls, Caspian terns, and bald eagles that feed on contaminated fish, which have led to drastically lowered reproductive success and population declines (Fox 1992; Giesy et al. 1994a).
Some of the environmental chemicals associated with adverse reproductive and developmental effects in animals mimic the actions of the female sex hormone estradiol and it has been hypothesized that human exposure to these compounds, generically referred to as xenoestrogens, may produce similar adverse effects on reproduction and development and be involved in the increasing incidence of breast cancer in human populations (Davis et al. 1993). It has also been hypothesized that environmental agents with estrogenic, anti-estrogenic, or antiandrogenic effects, or those affecting aryl hydrocarbon receptors may play a role in reported declines in sperm counts, increases in rates of testicular and prostatic cancer, and other urogenital tract anomalies in men (Sharpe and Skakkebaek 1993; Sharpe et al. 1995).
Historically, the focus on xenoestrogens as disruptors of reproductive function is based on an iatrogenic epidemic precipitated by administration of large doses of a potent synthetic estrogen, diethylstilbestrol (DES), to vast numbers ofcontinue
pregnant women throughout the world. The intent of this treatment was to rectify hormonal imbalances perceived to be causal in the premature termination of pregnancy in women at risk because of histories of habitual spontaneous abortion. The physiologic rationale for this therapy, initiated some 50 yr ago when reproductive endocrinology was still in its infancy, is difficult to reconstruct. It was based, in part, on the perception that estrogen and progesterone production was decreased near the time of normal or premature labor and that estrogens, under some circumstances, can stimulate ovarian function in experimental animals (Smith et al. 1946; Smith 1948). Using the dosage regimen prescribed by the originators of the therapy, and followed by hundreds of physicians, increasingly large doses of DES were administered from early pregnancy until week 35; women who followed this regimen may have received a total dose as high as 1.625 kg of DES orally. Although the dosage protocols varied somewhat from clinic to clinic, millions of women were subjected to this hormonal assault beginning in the 1940s. It has been estimated that in the United States alone some 4 million pregnant women and their fetuses were exposed in this manner between 1947 and 1971 (Mittendorf 1995) and perhaps millions more elsewhere in the world.
In the United States, the use of DES in pregnancy was abruptly halted in 1971 by action of the Food and Drug Administration (FDA) following the finding by Herbst and his colleagues of the appearance of an otherwise rare tumor, clear cell adenocarcinoma (CCA), of the vagina in eight young women whose mothers were treated with DES during pregnancy (Herbst et al. 1971). This study, which identified more cases of CCA than had been previously reported in the medical literature in this age group, marked the initiation of intensive studies of the offspring of women treated with DES during pregnancy. As part of this effort, a Registry for Clear Cell Adenocarcinoma of the Genital Tract in Young Females was established to identify cases, regardless of exposure. As of April 1995, out of a total of 622 identified cases in the United States, ranging in age from 15 to 35 yr, prenatal DES exposure could be documented in 367 cases (Mittendorf 1995).
Adverse consequences of prenatal DES exposure on the human female genital tract, including structural abnormalities and epithelial changes, have been reviewed in detail by Herbst and Bern (1981) and by Mittendorf (1995), and are the subject of continued, intensive investigation. In addition to alterations of the female genital tract caused by DES, the fertility and reproductive performance of DES daughters have been impaired, and the risks of prematurity, spontaneous abortion, and ectopic pregnancy have increased, particularly in DES daughters, with genital tract changes, that were exposed in early pregnancy (Herbst and Bern 1981; Stillman 1982; Mittendorf 1995). Nevertheless, more than 80% of DES daughters who wished to become pregnant delivered at least one live baby (Herbst 1992). Appendix A of this report summarizes these additional findings.
While an increased incidence of malignancies or decreased fertility has not been consistently reported in the male offspring of DES exposed mothers ("DEScontinue
sons"), increased rates of genital tract abnormalities including cryptorchidism and hypospadias have been documented (Gill et al. 1976, 1979; Rothman and Louik 1978; Depue 1984; Mittendorf 1995; Wilcox et al. 1995). The available data on sperm concentrations in "DES sons" are inconsistent. The numbers involved, however, are small and, therefore, these studies have limited statistical power. Adverse reproductive outcomes in males exposed to DES during development are reviewed in more detail in Appendix A.
The effects of DES administration on the offspring of treated women and their replication in mice and rats (see Newbold 1995) focused attention on the estrogenic properties of HAAs as potential carcinogens and teratogens. Studies on mice in the early 1960s demonstrated the induction of long-term irreversible changes from exposure to estrogen during a critical developmental period (see Herbst and Bern 1981). These early findings demonstrate, in dramatic fashion, the vulnerability of the fetus during critical stages of development to unphysiologic perturbations in its environment with unforeseen outcomes later in postnatal life, and have focused attention on the fetus as a unique target for the potential disruptive effects of environmental agents with hormonal activity (Bern 1992a). It must be recognized, however, that DES, which is a potent synthetic estrogen, was given to pregnant women and experimental animals in very high doses relative to physiologic estrogenic activity and that great care must be taken in extrapolating the DES experience to other estrogenic substances and dosing regimens. 1 Additionally, laboratory studies indicate that nonestrogenic agents, such as antiestrogens, androgens, and progestins (Bern 1992a,b), may induce some of the same developmental changes that are seen after perinatal estrogen exposure.
The finding that in utero exposure to an estrogen could cause cancer and reproductive anomalies in adult offspring has served as the basis for hypotheses linking the effects of environmental contaminants with estrogenic and other endocrine-like activities to reproductive and developmental anomalies in wildlife and human populations (Colborn et al. 1993).
Although HAAs in high concentrations can affect humans and wildlife, whether environmental exposures to them are responsible for a variety of widespread adverse effects on the health of humans and wildlife remains a topic of debate. In response to considerable public and congressional interest in this matter, the U.S. Environmental Protection Agency (EPA), the U.S. Department of the Interior (DOI), the U.S. Centers for Disease Control and Prevention (CDC), and the U.S. Congress requested a National Research Council (NRC) study ofcontinue
1 Because DES. while not an "environmental estrogen." has been considered by some workers in the field as a model for the action of estrogens generally, its effects on reproductive organs are summarized in Appendix A.
this topic. In response to this request, the NRC, the working arm of NAS, assembled a multidisciplinary expert committee, the Committee on Hormonally Active Agents in the Environment, under NRC procedures to "review critically the literature on hormone-related toxicants in the environment; identify the known and suspected toxicologic mechanisms and impacts on fish, wildlife, and humans; identify significant uncertainties, limitations of knowledge, and weaknesses in the available evidence; develop a science-based conceptual framework for assessing observed phenomena; and recommend research, monitoring, and testing priorities. To the extent practicable with available information and study resources, the committee [was asked to] identify particular chemical substances, geographic areas, contaminant sources, human subpopulations, and fish and wildlife populations of special concern with respect to hormone-related toxicants in the environment. In addition, if possible and warranted, the committee [should] suggest general approaches for identifying and mitigating these toxicological problems" (Committee Statement of Task).
This committee (like many groups before it) had difficulty deciding on the proper descriptors and definitions of its subject. Historically, the conceptual construct underlying the charge to the committee centered on the finding that the compounds under consideration have estrogenic activity and the hypothesis that this activity disrupts normal developmental and reproductive processes by interacting with estrogen receptors, the so-called "estrogen hypothesis." When the committee began its work, however, it was clearly recognized that the compounds to be evaluated also had antiestrogenic and antiandrogenic properties as well as some other hormonal activities, such as effects on thyroid function, and that not all of their actions are mediated by known hormone receptors. Indeed, the mechanisms underlying many of the effects of environmental contaminants on a variety of systems in animals and humans have not been elucidated. It is for this reason that the widely used descriptor of these compounds, "endocrine disruptors," was considered by some members of the committee to be too restrictive and to imply modes of action that are in fact unknown.
Because the compounds in the "endocrine disruptor" inventory gleaned from the literature-the starting point of the committee's work-by definition have hormone-like activity in at least some test systems, we have chosen the term "hormonally active agents" to describe them without regard to their mode or mechanism of action. This new and unfamiliar terminology was arrived at following extensive discussion, and some discomfort with its use remains (see discussion at the end of this Introduction).
For the historical reasons already mentioned, compounds with estrogenic activity predominate in the list of HAAs but substances in the environment with other demonstrable endocrine activities have not been excluded. Chemicals in the environment that do not conform to the broad definition of HAA-that is, that do not have any hormone-like activity in the usual test systems-were excluded from our study.break
It became clear as the work progressed that the same data could be approached with different viewpoints leading to different judgments. These differences are discussed in the next section.
The NRC's charge to the present committee did not include a request to assess risk management policy options, so the committee has not done so in this report.2 However, the committee hopes that the critical evaluation of the data presented in this report will be helpful in informing those who must make the decisions about policy options.
The report consists of 11 chapters, one appendix, and one addendum. Chapter 2 presents the HAAs that are discussed in this report and describes the current state of knowledge regarding the different modes of action of HAAs. In this chapter, a discussion of the biology of estrogen action is presented as a model for understanding the mechanisms by which HAAs with estrogenic activity could interfere with normal endocrine processes. It has been assumed that lessons learned from the study of estrogenic HAAs might be relevant to understanding the mechanisms of action of other types of HAAs that act by binding to intracellular hormone receptors. Chapter 3 describes what is currently known with regard to exposures and monitoring of HAAs. Chapter 4 reviews the information on dosimetryhow the uptake, distribution, metabolism, and elimination of hormones and HAAs might be involved in the mechanisms leading to altered hormonal processes and their consequences. In Chapters 5-7, the committee identifies a number of vertebrate biologic systems for which there were data relating alterations in these systems to HAA exposure, and separate chapters were devoted to each system: reproductive (Chapter 5), neurobehavioral (Chapter 6), and immune (Chapter 7). Chapters 8 and 9 review what is known about HAAs and cancer in animals and humans, respectively. This is followed by a chapter oncontinue
2 The National Research Council's policy on reports related to human-health risks (unpublished material) emphasizes the need for its study committees to separate matters of scientific fact and analysis from views and judgments about what public policies or measures, if any, should be taken to avoid or limit risk in a given situation. The NRC takes no position on any risk-management philosophy (e.g., the precautionary principle), but asks its study committees to avoid explicitly or implicitly advocating a position on the degree of caution that should be exercised in risk management. Weighing the relative importance of protecting public health and economic interests, for example, in the face of uncertainty is a public policy judgment, not a scientific one. The NRC does not discourage its study committees from addressing risk management questions when the charge calls for it, but recognizes that its proper role is to inform public-policy choices to the extent practicable by describing and interpreting relevant scientific facts and uncertainties for the benefit of citizens and their representatives who must make the policy judgments. NRC committees must carefully separate their proper role of informing policy choices from the different role of recommending policy. When a committee is asked to discuss policy options, the NRC recommends that various options be presented and their implications be explained in a neutral manner. This is not an easy task, but examples of such treatment are illustrated in Science and Judgement in Risk Assessment (NRC 1994) and Science and the Endangered Species Act NRC 1995).
ecologic effects (Chapter 10). Chapter 11 describes techniques for screening and monitoring for HAAs. Appendix A provides a summary of reproductive disturbances in humans and experimental animals exposed to DES. In the addendum that follows the References, at the request of EPA, the committee comments on the recommendations made by the Endocrine Disruptor Screening and Testing Advisory Committee (EDSTAC). Because the NRC committee was unable to reach consensus on some important matters, the remainder of this chapter describes these difficult areas.
Issues that Divided the Committee
In its report, the committee's approach comprised three steps: ( 1 ) identification of chemicals in the environment with hormonal activity; (2) evaluation of the scientific literature on the effects caused by those chemicals in laboratory animals, humans, and wildlife; and (3) consideration of whether observed effects can be attributed to the hormonal properties of the chemicals and to observed environmental exposures.
There are many areas of consensus among the committee members, reflected throughout the report and in the Executive Summary. However, it became clear as the work of the committee progressed that the same data could be approached from different viewpoints. Those different views led to different judgments among the committee members about the significance of the threat posed by HAAs and are described in this section. Many of the differences reflect areas where additional research would be helpful, and the differences sometime illustrate the kinds of research that are needed. In other cases, the differences do not reflect the need for research but reflect differing judgments about the significance of information. The differences are not confined to the members of this committee but are also reflected in the scientific community at large and in the comments received during review.
Much of the division among committee members appears to stem from different views of how we come to know what we know. How we understand the natural world and how we decide among conflicting hypotheses about the natural world is the province of epistemology. Committee members seemed to differ on some basic epistemologic issues, which led to different interpretations and conclusions on the issues of HAAs in the environment.
In part, difficulty in agreeing on various aspects of the problem arises because many of the terms used are imprecise, with possible differences in meaning that lead to differences in interpretation. For example, the charge to the committee, most broadly and simply stated, was to evaluate the endocrine-disruptor hypothesis. Each term is open for interpretation. What does it mean to undertake an evaluation? What do we mean by endocrine disruptor? And, what exactly is the hypothesis that is to be evaluated? The charge to the committee required more than simply compiling facts and presenting a review of the current literature.break
To fulfill its charge, the committee had to (1) interpret the hypothesis, (2) decide on which facts were germane to evaluating that hypothesis, (3) decide on allowable sources of information, (4) assign different weights to different kinds of evidence when data or interpretations were in conflict, (5) establish and evaluate alternative explanations that might account for observed patterns in the data, (6) decide on what type of errors were acceptable and which were not, (7) establish criteria for arriving at meaningful conclusions and recommendations based on understanding of the hypothesis and the status of the data pertinent to the hypothesis, and (8) reach consensus on all of these points. Unfortunately, the committee did not achieve consensus on all these tasks and related issues. In particular, the following issues were the focus of disagreements among the committee members or issues that need some explanation here: ( ) in their identification of and approach to the endocrine-disruptor hypothesis; (2) how much to emphasize other hormonal systems (i.e., those not involving sex steroids), (3) identifying allowable sources of information; (4) how to evaluate the evidence, including how to deal with Type I and Type II errors; (5) how to define HAAs. including the importance of mechanism of action of toxicants and their relevance to the endocrine-disruptor hypothesis; (6) the importance of low-dose effects and the shape of the dose-response curves; (7) how to treat periods of susceptibility; (8) how to extrapolate information on one species to others, including humans; and (9) the definition and quantification of environmentally relevant doses, concentrations, and exposures.
The Endocrine Disruptor Hypothesis
The charge to evaluate the endocrine-disruptor hypothesis begins with ambiguity about what the exact hypothesisand hence the charge to the committeereally is. In its simplest form, the hypothesis is that some chemicals in the environment mimic estrogens (and other sex-hormones) and hence interfere with (disrupt) endogenous endocrine systems, with adverse effects. Interpreted narrowly, it would pertain to one or more species in at least one place. Interpreted broadly, it would mean that the impacts of HAAs are uniform and global; this is probably not meaningful, and can certainly be falsified by a single counterexample. The more interesting alternative lies somewhere between these two extremes. The committee was divided as to the degree to which available data support an interpretation of the hypothesis between the two extremes. An analysis of the hypothesis helps to understand some of the differing judgments among committee members.
The hypothesis must be viewed in the context in which this committee was formed. Public and scientific concern has mounted over a perceived correlation between the presence of certain classes of chemicals in the environment and certain biologic effects. Often, the effects in question-for example, defects in development, declines in fertility, incidence of various cancers, possible popula-soft
tion declines in wildlife speciesare alarming. In some cases, the effect is in question; for example, is human sperm concentration really in decline? In other cases, an effect is quite clear, for example, developmental deformities of organ systems in wildlife species, including reproductive organs, but the cause is in question. Because there is simultaneous ambiguity at the level of both causes and effects, it is hard to identify the real nature of the question to be reviewed, and the committee decided first to clarify the nature of the hypothesis under scrutiny. Clarifying the hypothesis is related to clarifying the definition of "endocrine disruptor" and agreeing on an appropriate term; that problem is discussed later in this chapter.
For the endocrine-disruptor hypothesis to be understandable, some effect of some chemical on or through the endocrine system must be documented. The endocrine system can be disrupted in many ways that would lead almost every chemical with a disruptive effect on the organism to be classified as an endocrine disruptor; however most proponents of the hypothesis have a narrower definition in mind. Their formulation of the hypothesis suggests that some chemical inputs can act analogously to specific hormones, and the chemicals either overmodulate or undermodulate specific activities of hormones, producing specific effects that are hormonally mediated. Even the term "disrupt" is subject to differing interpretations: exposure to chemicals can result in physiologic adaptations within the normal range of variation to the death of the organism, and all gradations between those extremes. There is not agreement on how to categorize those responses. The EDSTAC report (EPA 1998) described a similar disagreement among its membership, also involving the difficult problem of an objective judgment of what constitutes an "adverse" effect.
The above is why the committee refers to such chemicals as hormonally active agents rather than as endocrine disruptors. However, there is still ambiguity, because modulation of hormonal activity can occur through a variety of mechanisms.
The committee recognized that it could not undertake a general evaluation of the correlation between specific chemicals and general effects. By restricting its review to information about chemicals that produce effects through hormonal pathways, the committee did not intend to suggest that there are no correlations between a variety of chemicals and effects not mediated through hormonal pathways or that such effects are not important. The committee also made no judgment as to whether such chemicals produce effects mediated through other mechanisms. An alternative interpretation of the hypothesis would probably lead to a different review and interpretation of evidence.
Other Hormonal Systems
The early focus of the endocrine disruptor literature on compounds with gonadal steroid activity and the need to limit the scope of the study prompted thecontinue
committee to emphasize these activities and their consequences. It was recognized that interference by HAAs with other hormonal and non-hormonal physiologic systems may have major impacts on development and other functions but the literature on these was not abundant and there was some disagreement about the committee's availability of time and resources to further extend its purview. In addition, various HAAs have been reported to have profound effects on the biosynthesis and metabolism of thyroid hormones with consequent reductions in circulating thyroid hormone levels (Brouwer et al. 1998). Whether this disruption of thyroid hormone homeostasis mediates some of the developmental effects associated with exposure to these HAAs remains to be determined.
As an example, the retinoic acid system is another pathway the committee did not consider in detail that could result in serious developmental defects if disrupted. Retinoids and retinoid receptors are key factors in vertebrate development. TCDD, planar PCBs, and a variety of other compounds affect development through processes governing cell fate and organ development and function. Phenotypically, these effects in some species resemble effects resulting from excess or from deficient levels of retinoic acid (Birnbaum et al. 1989). That suggests that at least some developmental abnormalities could result from chemical effects on retinoid synthesis, retinoid inactivation, or on retinoid receptors or function.
Interactions between different hormonal systems can link effects on one system to outcomes through another system. For example, thyroid receptors form functional heterodimers with 9-cis retinoic acid receptors (RXRs) (Puzianowska-Kuznicka et al. 1997). Linkage of the thyroid system to estrogen-active compounds also is suggested, as the estrogen receptor-related orphan receptor alpha 1 stimulates the expression of the thyroid hormone receptor alpha (Tra) (Vanacker et al. 1998).
Allowable Sources of Information
The literature related to endocrine disruption is large and often controversial. Articles, reports, abstracts, and circulating manuscripts appear at an accelerating rate. The committee had to choose what to base its conclusions and recommendations on. The foremost problem of the committee then was to provide meaningful information with respect to this broad and highly controversial topic and yet keep the scope of its review manageable.
Many members of the committee are involved actively in research in this area, and so they often had access to unpublished results or manuscripts. Although such information could alert the committee to possible trends, unpublished reports could not be included in this evaluation because they had not undergone independent peer review. Similarly, other unreviewed information, including much literature posted on the Internet, many federal agency reports, and some reports from industry and nongovernmental organizations, were notcontinue
used in evaluating the data. However, such publications, including abstracts, have been cited in the report for informational purposes.
Restricting information sources to published, peer-reviewed material meant that some late-breaking results were not fully evaluated, including results that may be inconsistent with some of the report's conclusions, but that restriction was the clearest and most objective way to deal with the problem of unpublished data. Peer-review, of course, provides no guarantee of accuracy or reliability. In some cases, phenomena reported in peer-reviewed papers could not be observed by others, and some unreviewed reports have been influential and accurate.
The committee evaluated the conclusions of peer-reviewed papers based on the methods used by the investigator. Often, the assessment of a topic was revisited many times because investigations are ongoing on many of the more controversial topics, and so some conclusions are unavoidably tentative, although they are based on the best available information from peer-reviewed sources at the time this report was completed.
Evaluating the Evidence
Under ideal conditions, a critical experiment can be performed with an outcome that can definitively decide between two hypotheses. The Meselson-Stahl demonstration of the semiconservative replication of DNA (Meselson and Stahl 1957) and the Michelson-Morley experiment refuting the existence of a "luminiferous ether" (Michelson and Morley 1887) are classic examples of this sort. Similarly, Koch's postulates (Koch 1876; Yerushalmy and Palmer 1959) have been used as guides for designing critical experiments to establish causes of infectious disease.
Although such critical experiments can give elegant and definitive answers to important theoretic and practical questions, they are limited to hypotheses for which one can construct controlled environmental conditions that essentially isolate the process under investigation from all other processes. Koch's postulates, for example, depend entirely on the ability of the investigator to isolate putative causal agents from alternatives.
Many important hypotheses about causal mechanisms cannot be tested through critical experiments simply because the factors generating responses cannot be isolated from each other. This is often true when attempting to resolve environmental effects on biologic processes outside the laboratory setting. For example, potentially harmful chemicals seldom exist in isolation from other chemicals in the environment, and they usually interact with other physical environmental factors. The ability to deal with potential effects due to mixtures of chemicals continues to plague toxicologists, and these difficulties apply to HAAs. Many reports have recommended further research on mixtures of chemicals, and this committee agrees with that recommendation. However, the committee hascontinue
no illusion that the problem of how to apportion cause among the members of a mixture of environmental chemicals will soon be easily solved.
In systems where definitive critical experiments are difficult or impossible or are dominated by complex interactions under natural conditions, one can either withhold judgment or take a weight-of-evidence approach, in which experimental evidence is only one of several lines of evidence used in determining likely causal mechanisms.
The committee agreed that lack of evidence could not be taken as an indication that a proposed process does not operate. The greatest disagreement within the committee concerned how much we should rely on different positive lines of evidence in evaluating the endocrine-disruptor hypothesis or on the weight that negative evidence should receive. The committee often failed to agree on the relative importance of each class of evidence in reaching a common judgment on causation, or even whether a specific criterion could be applied to the hypothesis.
To a large degree, differences among committee members could be divided along two perspectives on the weight-of-evidence approach. Some committee members placed almost exclusive weight on experimental evidence and the establishment of a plausible mechanism of action. Other committee members placed less weight on the mechanism of action and more weight on consistency and coherence of results among studies and an analogy with other compounds in test systems, especially endogenous sex-steroid hormones. Because the weight-of-evidence approach does not by itself dictate how weights should be assigned to the factors, both of these positions are equally valid.
Type 1 and Type 2 Errors and the Precautionary Principle
Committee members had differing views on how analyses of risks to humans and wildlife should be influenced by the probability of making Type 1 and Type 2 errors. A Type 1 error is the conclusion that an association (between exposure and adverse health effects in this case) exists when in fact it does not. A Type 2 error is the conclusion that there is no association (between exposure and adverse health effects) when in fact there is. Research cannot remove all uncertainties in describing the real world, and so some assumptions must be made in statistical analyses. Specifically, some committee members felt that the precautionary principle requires statistical analysts to assume as the default hypothesis that an environmental agent has adverse effects. This approach, discussed in detail in an NRC report (NRC 1995), would mean that the alternative hypothesis of no effect would be rejected unless the probability of its occurrence by chance was less than 5%. Other committee members felt that the above approach would amount to trying to prove a null hypothesis of no effect for every chemical, a large and difficult task. Also, they considered that choosing the appropriate default hypothesis of an effect would present statistical difficulties, such as deciding how much of an effect to hypothesize. They considered that testing the null hypoth-soft
esis of no effect is statistically sound and widely accepted scientifically, and that choosing a default hypothesis of some effect amounted to a confounding of risk analysis with management.
Simberloff (1990) provided a thoughtful analysis of this matter, concluding that the automatic adoption of a null hypothesis, whether it is that there is no effect or that there is an effect, may not be the best way to analyze every scientific uncertainty. He pointed out that the decision to minimize either Type 1 or Type 2 errors should be based on a careful evaluation of each case, and not be accepted as a default of the analysis. We refer interested readers to that paper and to the NRC's 1995 report for a more detailed discussion of the problem.
Defining a Hormonally Active Agent
Early in its deliberations, the committee decided that in this report the term endocrine disruptor should be replaced, recognizing that the term is fraught with emotional overtones and was tantamount to a prejudgment of potential outcomes. Furthermore, because all the compounds initially considered by the committee possessed hormonelike activity in at least some test system, the committee adopted hormonally active agents (HAAs) as a more neutral mechanistic descriptor. Thus, the use of HAA was meant to describe these agents without regard to the outcome of any critical evaluation of their mode or mechanism of action. Any definition provides a semantic filter of sorts that limits the number of substances reviewed, and the committee members felt varying levels of discomfort with respect to the definition and change in terminology, because any attempt to clarify can either eliminate a major portion of the problem from consideration or expand consideration to compounds not originally contemplated.
Indeed, the problem of defining HAAs is inseparable from the mechanism problem: A single chemical can have multiple effects on an organism that act through several mechanisms, not all of which involve hormone receptors. Thus, some observers object to the characterization of such a chemical as an HAA, and others wonder if calling it an HAA predisposes a conclusion to the question of whether its effects are related to its hormonal activity. The example of DDT's role in the production of thin-shelled eggs by carnivorous birds is illustrative. There is no controversy about the observation that DDT (or its metabolites) causes shell thinning in carnivorous birds' eggs, such as cormorants and eagles. There is also no controversy about the observation that DDT and its derivatives can mimic some actions of estrogens. The difficulty arises because DDT can also have effects that are not directly related to these hormones. The controversy continues because many mechanisms have been proposed to explain how DDT causes eggshells to be thin, but none has been firmly established to be related to the action of estrogen (Chapter 5).
In approaching the issue of HAA mechanisms, therefore, the committee does not believe that HAAs can function only through receptor binding, although thatcontinue
is the particular mechanism that has been most studied. Where mechanisms of HAA action are known they are presented in this report. Where the mechanisms are unknown, they often are not mentioned in this report. Lack of knowledge about a mechanism does not mean that a reported effect is unconfirmed or unimportant; it does suggest that additional studies are called for. This leads again to the question of how HAA should be defined. Is a compound an HAA if it can affect hormonal pathways, or is it an HAA only if in a particular case it does affect a hormonal pathway? If it is the former, then should all the biologic effects of such a compound be considered in this report, even if they are unrelated to hormonal pathways? Is it even theoretically possible to describe a biologic effect of an HAA, indeed of any chemical, as completely unrelated to hormonal pathways? The answers to these questions are not agreed on by all the members of the committee.
In considering the consequences of HAA exposure, it is important to know the shape of the dose-response curve for a particular HAA. In evaluating the data. it became apparent that committee members opinions differed on the extent to which inverted U-shaped curves should be emphasized in this report, in light of the paucity of data regarding the actions of HAAs. All members agreed in principle that this phenomenon is potentially important for evaluating the results of bioassays and for designing toxicologic studies. If an underlying monotonic dose-response function (i.e., a function where response increases as dose increases or at least does not decrease) and a dose below which there is no effect (a threshold dose) are assumed when designing a toxicologic study, there is a risk of failing to detect a contaminant that is inactive at intermediate doses but does have an effect at low doses, in other words, one that does not display a monotonic dose-response function.
There is a published report that DES, a powerful estrogen, when administered prenatally to mice, elicits an inverted U-shaped curve in the prostate gland of the offspring (vom Saal et al. 1997). The higher doses administered reduced prostatic size whereas lower doses increased it; still lower doses had no effect. Other workers using the same strain of mice at the one dose of DES that caused a maximal increase in prostate weight in the vom Saal et al. (1997) study were unable to replicate that finding (Cagen et al. 1999).
Inverted U-shaped dose response curves have been reported in other in vivo studies. vom Saal et al. (1995) reported an inverted U-shaped-dose relationship between maternal doses of DES and territorial marking in male offspring. These investigators also reported an inverted U-shaped dose-response relationship with respect to prostate size when mouse fetuses are exposed to estradiol (vom Saal et al. 1997). Halling and Forsberg (1993) reported that uterine weight as a fraction of body weight following neonatal dosing displayed an inverted-U dose responsecontinue
with DES but not estradiol. Fan et al. (1996) showed that TCDD induced an inverted U-shaped dose-response curve for its effects on cell-mediated immunity in rat; at low doses, TCDD enhanced and at high doses TCDD suppressed the delayed-type hypersensitivity reaction.
Periods of Susceptibility
The committee members agreed that there are important, critical periods during pre- and postnatal life during which an organism is particularly sensitive to exposure to various chemicals, but disagreed on how much emphasis should be placed on such critical periods in the absence of relevant data regarding HAAs. Despite growing information on the timing of events in differentiation in vertebrates, there is not much information that specifically links that timing to the actions of specific HAAs.
The issue is important because molecular events early in development determine phenotype later in development and in the adult. Disruption of these early events can thus produce effects that become apparent only later in development or in juveniles or adults. There are well-documented cases where exposure to chemicals such as PCBs and TCDD early in development causes severe defects later, whereas exposure later in development has no adverse effects (e.g., Vandersea et al. 1998).
Extrapolation Between Species
There were disagreements among committee members about the extrapolation of findings made with one species to others. In some cases, some members argued that if an adverse effect was observed in one species, then one should assume or suspect its occurrence in others. Other committee members considered that it was necessary to establish the adverse effect for each species of concern. In other cases, some committee members argued that the absence of an adverse effect in one or a few species meant that an adverse effect was unlikely in any species, while other members argued that it was necessary to establish the lack of an adverse effect for each species of concern. These disagreements led to disagreements about how to describe cross-species extrapolations and the comparative physiology of various species, and what to conclude based on incomplete information.
This issue is important because concern over particular effects ascribed to HAAs in the environment often is based on observations with one or a few species. The degree to which such observations apply to species other than those they were made on is crucial to any judgment about the seriousness of the risk of HAAs to wildlife, ecosystems, or humans. Although it is usually impossible to extrapolate qualitative dose-response relationships from one species to another, it is often prudent and useful to infer broad similarities in processes involved in toxicitycontinue
between species in the same taxonomic order and sometimes even in the same taxonomic class. The difficulty arises in deciding whether and when to generalize.
The difficulty is compounded because different processes behave differently with respect to inter-species differences. At one extreme, animals as distantly related to humans as nematodes have 40% of their genes in common with human genes. Genes and gene products governing pattern formation are structurally similar and appear to operate similarly even in evolutionarily distant species (e.g., Tomarev et al. 1997). At the other extreme, differences between genetic strains of a single species, due sometimes to a difference at only one genetic locus, are the basis of much of our understanding of development and physiology. An example of susceptibility differences among closely related organisms is the more than 5,000-fold difference in toxicity to TCDD among different mammal species (Pohjanvirta and Tuomisto 1994). A single allelic substitution (e.g., alcohol dehydrogenase, cytochrome P450) can affect responses to a toxicant within a single species (e.g., Meyer et al. 1990). The committee members did agree on the general principle that it is necessary to characterize the nature of interspecies differences.
Environmentally Relevant Doses, Concentrations, and Exposures
Committee members did not agree on the definition of ''environmentally relevant" as applied to HAAs, or on how much information was required to establish environmental relevance. A related problem is the imprecise use of the terms "dose," "concentration," and "exposure" in toxicology literature in general. The two issues are discussed here together.
In this report, as in environmental toxicology in general, there is concern regarding the concentrations of suspect chemicals to which organisms are environmentally exposed. High doses often are used in the laboratory, as, for example, the use of maximum tolerated dose in early phases of evaluating carcinogenicity of chemicals (Goldstein 1994). However, our greatest concern in this report is whether the much lower doses of HAAs that result from environmental exposure cause adverse effects. "Environmentally relevant" is often used to refer to concentration and the resulting doses that organisms might encounter from exposures in nature or in the workplace. Environmentally relevant can apply to the presence of and exposure to chemicals in several contexts. The committee members did not agree on how much and what kinds of information are needed to establish how well laboratory experiments are environmentally relevant or what concentrations of chemicals in the environment are physiologically relevant. The following sections describe the relationship between environmental concentration, environmental exposure, and dose. In much of the published literature the committee reviewed, distinctions between environmental concentration, exposure, and dose often are not made. This failure can reduce the relevance and value of research results.break
Environmentally Relevant Chemical Residues. For a compound to be environmentally relevant, it must exist in the environment. However, in the environment, one almost always finds mixtures. Experiments that use such mixtures, whether extracted from environmental matrices or artificially produced, may be relevant to contemporary environmental exposures and would account for possible interactive effects of chemicals in the mixture. Such experiments would not identify the specific agent or agents involved in effects of those mixtures. Furthermore, there is a lag between exposure and effect. For example, if exposure during development produces an effect only in the adult or if there is a long latent period, then it is usually too late to measure the exposure when the effect is observed, because the causative agent has long since disappeared. This may be of greater concern in long-lived species, including humans, and for conditions known to have a long lag time before diagnosis, such as some cancers.
Environmentally Relevant Chemistry. Chemicals of concern as potential endocrine disruptors or HAAs in the environment most often are organic chemicals that are hydrophobic to varying degrees. Such chemicals seldom are free in the environment but are complexed with organic or particulate matter. The chemistry governing that complexation may determine the bioavailability of the compounds (Farrington 1991). Moreover, different structures may be broken down by bacterial action at rates that depend on the structures.
Environmental Concentrations. This term refers to concentrations as they exist in the environment. The environmental chemistry processes will affect the concentration available. The effect of environmental chemicals depends on their concentration and the degree and method of exposure to them that organisms undergo. Experiments that use such concentrations may produce exposures typical of those experienced by organisms in the real world, but to faithfully reproduce such exposures, all routes of exposure must be considered and integrated.
Environmentally Relevant Routes of Exposure. Exposure refers to contact with an environmental chemical by a particular organism. The amount or degree of exposure is a function of the concentration of the chemical in the environment and the duration (time) of the contact (NRC 1991a; Ott 1995). Although there are many kinds of exposure, such as peak exposure, average exposure, minimum exposure, instantaneous exposure, integrated exposure, and so on (Ott 1995), our main concern here is with the fact of exposure and not with the detailed nature of the exposure, although the details are required to assess quantitative risks. We include all environments an organism experiences in our understanding of exposure, including workplace environments; we exclude the deliberate or accidental administration of chemicals by means of drug-delivery systems. The matrix in which chemicals occur and the lifestyle of an organism will determine the dominant route of exposure. If an objective is to examine the consequences of exposure in the environment, routes of exposure should be considered in experimental design. Exposure by intraperitoneal injection of adult females may be adequate to determine if a chemical or mixture has the capacity to elicit specific changes.continue
But if the major route of exposure is dietary, as it is for many chemicals for most vertebrates, then relating experimental exposure to that in the environment should include a dietary route.
Environmentally Relevant Doses. A dose is the amount of a chemical that enters an organism's body (often expressed as an amount per unit weight). This is also often called the "internal dose." An "environmentally relevant dose" as used in this report means a dose administered in an experimental setting similar to that which would result from exposure to the same chemicals at concentrations that occur in the environment. To determine whether an experimental dose of any given chemical is environmentally relevant requires knowledge of environmental concentrations of the chemical and the resultant doses experienced by organisms that encounter it in the environment.break