Potential health effects of chemicals that disrupt endocrine function pose an environmental health concern about their ability to interfere with normal hormone function in human and wildlife populations. The endocrine system regulates biologic processes throughout the body and can be sensitive to small changes in hormone concentrations. Endocrine-disruptor research has focused primarily on chemicals that affect three hormone pathways that play important roles in reproduction and development—the estrogen, androgen, and thyroid hormone pathways. Some of the research has identified dose–response relationships that have nonmonotonic curves. Nonmonotonic dose–response (NMDR) curves are of concern because they do not follow the usual assumption made in toxicology that toxic response decreases as dose decreases. The slope of a NMDR curve changes sign, and the function can take on the shape of a U, the shape of an inverted U, or another shape that has more than one inflection point.
The existence of NMDR curves has been controversial for decades, and there has been considerable debate about their implications for the testing of chemicals and the assessment of risks posed by chemicals. Toxicity tests are designed to identify hazards and to characterize dose–response relationships, so they are aimed at finding a dose that is high enough to elicit a response and exploring dose–response relationships by spacing lower doses to identify a no-observed-adverse-effect level (NOAEL) or a lowest observed-adverse-effect level. One concern raised by NMDR relationships is that such studies as currently designed might not detect critical points (such as peaks and valleys) along a dose–response curve if only a few doses are tested or if the inflection point occurs below the doses tested. Another concern is that some NMDR relationships are found in connection with biologic effects that are not usually evaluated in toxicity tests. If current testing strategies are inadequate to account for NMDR relationships, changes in risk-assessment practices might be necessary.
To address these concerns, the US Environmental Protection Agency (EPA) has developed a draft report, State of the Science Evaluation: Nonmonotonic Dose Responses as They Apply to Estrogen, Androgen, and Thyroid Pathways and EPA Testing and Assessment Procedures (SOTS evaluation). The SOTS evaluation presents a collection of the evidence on NMDR relationships to answer three central scientific questions:
1. Do [NMDRs] exist for chemicals and if so under what conditions do they occur?
2. Do NMDRs capture adverse effects that are not captured using [EPA’s] current chemical testing strategies (i.e., false negatives)?
3. Do NMDRs provide key information that would alter EPA’s current weight of evidence [WOE] conclusions and risk assessment determinations, either qualitatively or quantitatively?
Given the important role that the SOTS evaluation could play in making decisions about conventional toxicity testing and risk-assessment practices, EPA asked the National Research Council to evaluate whether it presents a scientifically sound and high-quality analysis of the literature on NMDR curves. To the committee’s knowledge, this is one of the few attempts to evaluate specifically the implications of the evidence on NMDR curves for testing strategies and risk-assessment practices, and the agency is commended for undertaking such a challenging task.
This report presents the findings of the committee convened by the National Research Council. The committee focused on whether EPA fairly and soundly evaluated the evidence from diverse sources (in vitro, animal, mode-of-action, and epidemiologic studies) and on whether the SOTS evaluation provided a robust, objective, and repeatable analysis. The committee reviewed how clearly EPA describes how its assessment was performed, whether consistent methods and criteria were applied in the analysis of different evidence streams, and whether appropriate methods were applied in evaluating the evidence. Thus, the focus was on the process of the SOTS evaluation and documentation that resulted. The committee organized its evaluation according to the three central scientific questions that EPA sought to answer. Question 1 is considered first, and Questions 2 and 3 are addressed together later.
STATE-OF-THE-SCIENCE APPROACH TO ANSWERING QUESTION 1
To answer the question of whether NMDR curves exist for chemicals and, if so, under what conditions, EPA focused on evidence of NMDR relationships in studies of chemicals that have effects on the estrogen, androgen, and thyroid hormone pathways. EPA’s SOTS evaluation made a distinction between NMDR relationships and the related issue of low-dose effects and successfully avoided conflating the two concepts. It presented evidence from in vitro studies and in vivo studies of aquatic species and mammalian models. It concluded that exposure to endocrine disruptors can result in NMDR curves for specific end points and that such curves were found more often in in vitro studies, at high doses, and for exposures of short duration.
Because of time constraints, independent groups wrote the sections on the estrogen, androgen, and thyroid hormone pathways. No plan was established in
advance for the writing groups to follow. Rather, the groups determined independently how to perform their analyses. The committee sought to determine how the groups approached their evaluations by looking for documentation of the literature-search strategies, criteria for selecting studies, methods for assessing study quality, presentation of the evidence, and methods used for synthesizing the evidence. Documentation of some of these elements was difficult to find and in most cases had to be inferred. The exception was the section on the mammalian evidence of NMDR curves in studies of thyroid disruptors, which provided enough description of how the literature was searched and filtered, of the modes of action (MOAs) that were considered, and of how the findings were synthesized for the committee to understand the process that was used. EPA acknowledged that such a process was not followed by the groups evaluating the data on the estrogen- and androgen-hormone pathways and that an expert-driven approach was used instead. Such an approach might be appropriate as an internal scoping exercise for the agency, but a higher standard of documentation and analysis is needed to provide the foundational support necessary to make decisions about the agency’s toxicity-testing strategies and risk-assessment practices. A lack of transparency and a lack of consistency were identified as two critical limitations of EPA’s evaluation. EPA has acknowledged such limitations and has indicated that it plans to conduct more systematic literature searches, data extraction, and evaluations of the evidence on NMDR curves. However, the results of these endeavors were not available to the committee, so findings and recommendations are restricted to what is presented in the SOTS evaluation.
Recommendation: An analytic plan should be developed and applied consistently to the evidence on the three hormone pathways. Important elements of the plan include predefining and documenting the literature-search strategies and their results, establishing criteria for selecting studies for analysis, establishing criteria for determining study quality, using templates for presenting evidence consistently in tabular and graphic form, and documenting approaches to integration of evidence. Guidance on these elements is provided below.
Scoping and Framing Questions
The committee supports EPA’s approach of posing questions as a means of focusing the evaluation of the evidence on NMDR curves. The three central scientific questions are critical and are framed broadly. The first question is whether endocrine disruptors have NMDR curves and, if so, under what conditions. Identifying the “conditions” under which NMDR curves occur requires analysis of the different types of evidence that are available. However, EPA restricted its analysis to in vitro studies, studies of aquatic species, and studies of mammalian models; epidemiologic and other types of human studies were excluded from consideration. That exclusion raises a problem in that the SOTS
evaluation will ultimately inform decisions about human health risk-assessment practices. In addition, for the estrogen and androgen hormone pathways, it seems that evidence-gathering encompassed only a subset of MOAs by which chemicals and endogenous hormones can act in the body. That restricted the array of chemicals considered to be disruptive of the estrogen and androgen pathways and limited the ability to detect conditions under which NMDR curves might occur as a result of different modes of action.
Recommendation: If EPA wishes to determine the conditions under which NMDR curves occur, it should expand its evaluation to include evidence from epidemiologic and clinical studies and include chemicals that have a variety of potential MOAs for the different hormone pathways. If such a broad analysis is not feasible in light of the agency’s immediate needs, consideration should be given to narrowing the questions and their answers.
Methods of Analysis
Systematic approaches to evaluating the scientific literature have been recommended by other National Research Council committees and are being developed by government agencies and incorporated into environmental health assessments. Emphasis is placed on specifying and documenting the methods that will be used to answer a study question. Such methods include determining the literature-search strategy, defining criteria to select studies for analysis, evaluating study quality, presenting data, and integrating evidence. It is noteworthy that EPA’s SOTS evaluation of NMDR curves related to chemicals that affect the thyroid hormone pathway contained elements of that type of structured review, but the sections on the estrogen and androgen pathways did not. EPA has indicated that it has since started incorporating elements of systematic review into its literature evaluation in connection with endocrine disruptors.
An issue that should be further explored by EPA is whether the available data lend themselves to answering questions about NMDR curves in a definitive manner. That will require establishing the study-design criteria that would be necessary to characterize whether an NMDR relationship exists and identifying studies that meet the criteria. Conventional toxicity tests have design limitations that can make it difficult for them to detect or characterize NMDR relationships, such as the number of doses, the number of test animals, and the variety of end points considered. Statistical methods are available for conducting post hoc analyses of the data and for combining evidence from multiple studies, including meta-analytic approaches. Such methods could be adapted to evaluating the evidence on NMDR curves. However, that would need further research and development before implementation, inasmuch as current methods have been designed to examine single measures and outcomes and do not have the capacity to explore more complex relationships.
Recommendations: EPA should consistently use a more systematic approach to evaluating the literature on NMDR curves for all three hormone pathways. Guidance for such approaches is available from clinical guidelines, other National Research Council reports, and guidelines being developed in other government agencies. In developing an approach for the literature on NMDR relationships, special consideration should be given to the following:
• The methodologic features that would be necessary for a study to be able to detect an NMDR relationship should be identified. Ideally, multiple dose groups would be spaced across a defined exposure domain, including doses below those typically tested. Statistical design, biologic plausibility, and replicability should be factored into interpreting and weighing the evidence from such studies.
• Study exclusion and inclusion criteria should be established. Although statistical significance is an important consideration, it should not be an absolute criterion for including or excluding studies, inasmuch as standard toxicity-testing strategies generally do not have sufficient sensitivity and statistical power to detect NMDR curves.
• Study quality criteria should be established. Statistical criteria should be given particular attention. It will be important to balance study quality criteria that are based on statistical significance and those based on biologic plausibility.
• Secondary analyses of other studies may be necessary. Current methods for performing post hoc analysis of data and for combining evidence from multiple studies might be adapted for such purposes but would require research and development before implementation. EPA should consider soliciting input from the biostatistics community on the best methods to pursue in the long term and on what measures to take to complete the SOTS evaluation.
• Justifying the use of definitions that are not consistent with those used by EPA programs or that are controversial, such as definitions of low-dose effect, resilience, and adverse effect.
STATE-OF-THE SCIENCE APPROACH TO ANSWERING QUESTIONS 2 AND 3
EPA’s SOTS evaluation concluded that current testing strategies include assays that detect chemicals that interfere with the estrogen, androgen, and thyroid signaling pathways. Assertions were made that the evidence is insufficient to show that NMDR curves for adverse effects occur below particular thresholds—NOAELs or benchmark doses (BMDs)—derived from current testing strategies, that current testing assays are sensitive in detecting chemicals that interfere with the hormone pathways, and that NMDR curves occur mainly at high doses. However, it is acknowledged that traditional toxicity testing may not be sufficient for defining the shapes of dose–response curves. These conclusions appear to be based on the expert opinions of the authors of the document. The
present committee recognizes that expert judgment is integral to scientific analyses, but EPA’s evaluation provided insufficient documentation of the analyses that led to the conclusions.
The “testing strategies” that were under consideration were not specified, so it was unclear to the committee whether they included EPA’s toxicity-testing studies, testing associated with the Endocrine Disruptor Screening Program, testing as part of the ToxCast program, or other types of testing that are important for framing MOA or WOE analysis. Each of those categories of studies has different goals, so EPA’s conclusions about the adequacy of toxicity testing should be rooted in specific reference to and demonstration of its own testing strategies in a systematic manner for the estrogen, androgen, and thyroid pathways.
With respect to the question of whether NMDR curves provide information that would alter EPA’s current WOE conclusions and risk-assessment determinations, EPA concludes that NMDR curves can have both qualitative and quantitative effects and that current risk-assessment practices will consider the evidence appropriately. However, no analyses of how NMDR curves might affect considerations of hazard identification and dose–response relationships or of how current practices are equipped to handle evidence of NMDR relationships are provided in support of those conclusions. Thus, whether EPA’s conclusions are supported by the evidence presented in the SOTS evaluation could not be determined by the committee.
• EPA should specify the toxicity-testing strategies under consideration and evaluate them transparently for their ability to detect NMDR relationships mediated by the estrogen, androgen, and thyroid hormone pathways. The tests should be evaluated in terms of the design criteria that would be necessary to detect such relationships and the adverse effects of interest.
• The concept of “resilience” or adaptation is controversial and not adequately addressed by EPA in the SOTS evaluation. EPA should distinguish between end points that are adverse and ones that are adaptive. Because effects that are adaptive in some people are adverse in others, consideration should be given to potential windows of susceptibility (for example, during fetal development), sensitive populations (for example, those with pre-existing health conditions), and other factors (such as multiple chemical exposures) in making these distinctions.
• EPA’s evaluation should indicate how NMDR relationships for estrogen, androgen, and thyroid pathways would be addressed under the agency’s current risk-assessment guidelines and practices, including statistical considerations, uncertainty analyses, life-stage or susceptibility issues, and MOAs.