Hypothesis-driven weight of evidence evaluation indicates styrene lacks endocrine disruption potential

Abstract

Styrene is among the U.S. EPA’s List 2 chemicals for Tier 1 endocrine screening subject to the agency’s two-tiered Endocrine Disruptor Screening Program (EDSP). Both U.S. EPA and OECD guidelines require a Weight of Evidence (WoE) to evaluate a chemical’s potential for disrupting the endocrine system. Styrene was evaluated for its potential to disrupt estrogen, androgen, thyroid, and steroidogenic (EATS) pathways using a rigorous WoE methodology that included problem formulation, systematic literature search and selection, data quality evaluation, relevance weighting of endpoint data, and application of specific interpretive criteria. Sufficient data were available to assess the endocrine disruptive potential of styrene based on endpoints that would respond to EATS modes of action in some Tier 1-type and many Tier 2-type reproductive, developmental, and repeat dose toxicity studies. Responses to styrene were inconsistent with patterns of responses expected for chemicals and hormones known to operate via EATS MoAs, and thus, styrene cannot be deemed an endocrine disruptor, a potential endocrine disruptor, or to exhibit endocrine disruptive properties. Because Tier 1 EDSP screening results would trigger Tier 2 studies, like those evaluated here, subjecting styrene to further endocrine screening would produce no additional useful information and would be unjustified from animal welfare perspectives.

Keywords:

• OSRI
• styrene
• endocrine disruptor
• endocrine activity
• data quality
• mode of action
• estrogen agonist
• estrogen antagonist
• androgen agonist
• androgen antagonist
• thyroid inhibition
• steroidogenesis
• weight of evidence

Introduction

Styrene is among the second list of pesticides and industrial chemicals (“List 2”) that might undergo Tier 1 screening for the potential to interact with the endocrine system subject to the two-tiered Endocrine Disruptor Screening Program (EDSP) as initially implemented by the U.S. EPA [June 14, 2013, 78 FR 35922]. Tier 1 of the EDSP screens a chemical’s potential to interact with estrogen, androgen, thyroid and steroidogenic (EATS) components of the endocrine system in mammals and wildlife using short-term in vitro and in vivo assays (U.S. EPA 2009a). Chemicals that are not deprioritized for further evaluation based on Tier 1 screening may be required to undergo Tier 2 testing for adverse effects and dose-response characterization in intact organisms (U.S. EPA 2011). Tier 2 provides the basis for risk assessment to evaluate the possibility of adverse effects under conditions of exposure relevant to humans and wildlife. The EDSP differs from regulatory programs in the European Union (e.g. European Parliament 2006, 2012) that seek to affix qualitative “hazards” to chemicals for classification and labeling. Because the U.S. EPA regulates on risks of adverse effects from exposure to chemicals irrespective of the type of toxicity that might occur, the EDSP does not attempt to establish a causal link between endocrine modes of action (MoAs) identified in Tier 1 and adverse effects observed in Tier 2, as would be required to label a chemical an “endocrine disruptor” under the World Health Organization [WHO] International Program on Chemical Safety [IPCS] (WHO/IPCS 2002) definition.

Since its first application to screen 52 chemicals in 2009 (List 1) the U.S. EPA has modified the EDSP Tier 1 screening battery to include high-throughput in vitro assays, in silico methods, and computational models (Browne et al. 2015; Kleinstreuer et al. 2017), which may now be used in lieu of some of the original EDSP screens. The U.S. EPA has stated an intent to further increase the speed and efficiency of endocrine assessments by combining the results of these high-throughput bioactivity screens with the results of exposure assessment models to prioritize chemicals for endocrine screening (Rotroff et al. 2013), a program referred to as “EDSP-21”.

The Organization for Economic Co-operation and Development (OECD 2012) has developed a slightly different approach, using a 5-tiered Conceptual Framework consisting of in vitro and in vivo screens and tests that may be combined in various ways to evaluate chemical effects on endocrine function. Levels 2 through 5 of the Conceptual Framework contain the 11 screens and tests that compose the EDSP, as well as general toxicology and carcinogenicity studies that evaluate endpoints that could be affected by an endocrine MoA (OECD 2012). Both the U.S. EPA (U.S. EPA 2011) and OECD (OECD 2012) programs require the application of a Weight of Evidence (WoE) procedure to evaluate the results of all available data regarding a chemical’s potential for interaction with the endocrine system. The evaluation of styrene reported here follows the method of Borgert, Mihaich, Ortego, et al. (2011), which is one of the published WoE approaches cited by OECD (2012) as appropriate for evaluating the available scientific data relevant to endocrine disruption.

As was done to generate List 1 of the EDSP, the U.S. EPA developed List 2 using exposure criteria and a public comment process. List 2 comprises 109 chemicals selected from an initial list of over 200 candidates that included the pesticides scheduled for Registration Review during fiscal years 2007 and 2008, the industrial chemicals with a national primary drinking water regulation or the unregulated contaminants listed on the third Contaminant Candidate List (CCL3) (U.S. EPA 2009b). With implementation of the TSCA reform bill (U.S. Public Law 114-182 2016) and inclusion of components of the EDSP in existing programs for pesticides and industrial chemicals, it is currently unclear whether test orders will be given to manufacturers of chemicals on List 2 and, if so, what combination of Tier 1 screening assays (U.S. EPA Series 890) and EDSP-21 bioactivity screening results might be required for a regulatory determination of the potential for a chemical to interact with EATS pathways.

Scope and purpose

Irrespective of the assays that might be required, it is anticipated that test orders for List 2 would include an opportunity for recipients to submit “other scientifically-relevant information” (OSRI) to assist the U.S. EPA in determining the final requirements of endocrine screening for each chemical, as was the case for the test orders released for List 1 chemicals in 2009. Pursuant to that expectation, this report documents a WoE evaluation of the potential for styrene to act via EATS pathways and produce adverse effects via those modes of action. The WoE analysis included the following general steps, consistent with the methodology outline by Borgert, Mihaich, Ortego, et al. (2011):

1. Literature search and selection

2. Literature evaluation and data compilation

3. Weight of Evidence (WoE) evaluation for each mode of action addressed by the EDSP (estrogen and anti-estrogen, androgen and anti-androgen, thyroid modulation (stimulation and/or inhibition), aromatase inhibition, steroidogenesis modulation)

Organization of document

The main body of this document first presents the methods used for the WoE evaluation of styrene, including sections on literature search and selection and literature evaluation and data compilation. Following the descriptions of methods, results are explained for ToxRTool^® evaluations of the studies selected and for endocrine screening results from the EDSP Dashboard and ToxCast^®. Studies containing data used to evaluate EATS hypotheses are listed by the number used to identify them in Supplementary Tables 1–6. Following that list is a detailed explanation of the endpoint responses relevant for evaluating each of the six endocrine MoAs as listed in Supplementary Tables 1–6. The review concludes with a discussion section that explores the implications of the results, which are also summarized in Supplementary Table 7. Supplemental Materials A–D provide additional information and details relevant to various aspects of this WoE analysis.

Methods

Literature search and selection

Literature search

The literature search strategy was designed to connect the specific chemical name and CASRN (styrene/100-42-5) to the sets of key terms, phrases and assay search strings listed below, which aimed at capturing literature on putative endocrine-disruptive effects, the endocrine pathways addressed by the EDSP (estrogen, androgen, thyroid, and steroidogenesis, often referred to “EATS”), specific types of assays and endpoints used in the EDSP, and general types of toxicity that might be produced by endocrine modes of action (e.g. Reproductive and Developmental toxicity). The following search terms were used to query the primary citation databases PUBMED and Web of Science as well as governmental sources of literature (e.g. ATSDR, NTP, HSDB, DART), which also captured literature on endocrine-related cancers:

Endocrine disruption (or disruptor(s)), Endocrine effects, toxicity, Estrogen(s), estrogenic, anti-estrogenic, estrogen antagonist, androgen(s), androgenic, anti-androgenic, thyroid, hypothyroid, hyperthyroid, anti-thyroid, steroidogen* (* for variations steroidogenesis or steroidogenic), Hypothalamic-pituitary (or more specific: HPT axis, (hypothalamic-pituitary-thyroid axis), HPG axis (hypothalamic-pituitary-gonodal axis).

• The following assay search strings were used:

• (estrogen OR androgen OR (uterine AND cytosol) OR (prostate AND cytosol)) AND

• receptor AND binding AND assay*

• estrogen AND receptor AND alpha AND transcription* AND activation AND assay*

• aromatase AND assay

• steroidogenesis AND assay

• uterotrophic AND assay

• Hershberger AND assay

• (Male OR female) AND pubertal AND assay

• male AND intact AND assay

• fish AND reproduction AND assay

• amphibian AND metamorphosis AND assay

• Reproductive (associated keywords with truncation for variation: sperm; spermatozoa; spermatogen*; infertility; fertility; ovary, ovaries; ovarian; semen; testic*; testis; testes; gonad*)

• Developmental (associated keywords with truncation for variation:abnormalities/drug-induced; fetal; fetal; fetus; embryo*; prenatal; teratogen*)

Literature and data selection

An initial triage of this literature was conducted to separate studies into three categories: Apparently Relevant, Possibly Relevant, or Apparently Not Relevant. Apparently Relevant and Possibly Relevant studies were reviewed, and additional literature identified from the reference lists of those articles was collected and triaged by the same process. Usable data from all relevant studies were then compiled into tables as described in the following section, entitled “Literature evaluation & data compilation.” In May 2018 a literature search update was conducted and 9 additional studies were evaluated for a grand total of 181 publications and reports reviewed. A comprehensive list of literature evaluated is provided in Supplemental Material A.

Publications that met all of the following criteria were considered for inclusion in the WoE evaluation. Only those studies that failed to meet these inclusion criteria were excluded from consideration and no study was excluded based solely on data quality assessments.

• Publication in English or English translation available;

• Measurement of in vivo or in vitro endpoints relevant to one or more of the hypotheses evaluated by the WoE methodology used herein;

• Endpoints measured in a study type or experimental model (e.g. regulatory guideline; experimental investigative) amenable to the WoE methodology used herein;

• Appropriate control groups against which endpoint responses to styrene could be compared;

• Exposure levels and endpoint responses quantified according to objective or standardized methods (i.e. not reported simply as “positive” or “negative”);

• A statistical method was applied to provide perspective on numerical differences.

It is important to appreciate that the goals of data selection for this WoE evaluation are different than the goals of data selection for risk assessments, safety assessments, or for derivation of exposure limits such as reference doses (RfDs) and workplace environmental exposure limits (WEELs). In keeping with the scope of U.S. EPA WOE evaluations for endocrine activity, the purpose of this WoE is to determine whether, based on all available scientific data, styrene has the potential to interact via EATS MoAs. To meet that goal, selection criteria were broad in order to ensure adequate information to evaluate potential EATS endocrine MoAs irrespective of whether the data are suitable for other types of assessments. Consequently, many studies were used in this evaluation that would not be sufficiently reliable, relevant, or useful for other purposes.

For example, route of exposure was not an exclusion or inclusion criterion, even though some routes of exposure are more relevant than others for risk assessment. Similarly, studies conducted at excessively high doses were not excluded, even though endocrine MoAs may exhibit dose-dependency and can be overwhelmed by systemic toxicity at high doses, as is well established generally (Slikker et al. 2004; Marty et al. 2018; Borgert et al. 2021). The high doses used in many toxicity studies may trigger responses that would have no relevance whatsoever for endocrine MoAs at foreseeable levels of human exposure. For this WOE evaluation, an endpoint was generally considered to not have responded if the authors deemed the change in an endpoint to be unrelated to the test article or to have occurred secondary to maternal or systemic toxicity.

Finally, for purposes of this WoE evaluation, endpoints were deemed to have responded to the test chemical when they met the requirements of statistical significance even though, in many instances, the magnitude of the response may not have been biologically significant. This is a critically important point because the homeostatic function of the endocrine system allows for wide ranges of endocrine organ weights and hormone levels that are within the norm of healthy physiology. Consequently, values that are historically within normal ranges for the test species, and thus, not indicative of adverse effects may nonetheless be statistically significant compared to concurrent controls within a particular study.

All of these factors would tend to bias this evaluation toward a false positive conclusion. Hence, any conclusion of potential endocrine activity would have to be considered provisional. On the other hand, this approach to data selection resulted in the largest possible dataset upon which to evaluate the potential for styrene to act via EATS MoAs. Data were available on which to assess the effect of styrene on most of the endpoints that would respond to EATS modes of action in reproductive toxicity, developmental toxicity, and repeat dose toxicity studies. Consequently, a high level of confidence can be placed in negative conclusions regarding potential EATS activity because few data were excluded, and few data gaps were created as a result of this literature search and data selection strategy.

Literature evaluation and data compilation

Data quality assessment

Data selected for testing the MoA hypotheses were evaluated according to their primary, secondary, and tertiary validity as recommended previously (see Supplemental Material in Borgert, Mihaich, Ortego, et al. 2011), which is consistent with the U.S. EPA’s guidance on data quality assessment (U.S. EPA 2011). Primary validity (the identity and integrity of the measurements) was assumed for studies conducted according to international (e.g. OECD; U.S. EPA) toxicological guidelines and for studies reported in the peer-reviewed literature that employed similar measurement techniques (Borgert et al. 2016). Secondary validity (adequate controls and the reliability of the recorded measurements) was evaluated by subjecting each study to Klimisch et al. (1997) criteria using the ToxRTool scoring system created by the European Center for the Validation of Alternative Methods (Schneider et al. 2009). Tertiary validity (probative nature of the study design for evidence of causation) was evaluated according to the ability of the study to provide a causal link between the endpoint measured and the hypothesized MoA. All studies and endpoints utilized in this evaluation are listed and the endpoints used for each hypothesis are briefly described in Supplemental Material B.

Because our literature search attempted to exhaustively document all informative data regarding an endocrine MoA for outcomes related to exposure to styrene, several studies were found that cited endocrine-related outcomes, but with such severe limitations of either study design or data evaluation that the relevance of endpoints measured in those studies could not be ranked according to our methodology, and thus, the data were not used in this evaluation. Those publications are nonetheless referenced to assure the reader that the data were considered but determined to be noninformative for an endocrine WoE evaluation of styrene (Soto et al. 1995; Ohtani et al. 2001; Kuwada et al. 2002; Tatarazako et al. 2002; Jarry et al. 2004; Ruiz et al. 2012, Ruiz et al 2014). It is important to note that of these seven studies, only one (Tatarazako et al. 2002) reported a biologically significant effect of styrene, and the relevance of reduced reproduction of Ceriodaphnia dubia observed in that study is unknown for any specific endocrine mode of action. Thus, exclusion of those studies did not bias this WoE evaluation toward a false negative interpretation. The rationale for excluding each of these studies is explained briefly in Supplemental Material D.

Weight-of-evidence methodology

The methodology used for this WoE analysis is based on a broadly-applicable, general WoE framework whose utility was illustrated by application to the results of the U.S. EPA’s EDSP Tier 1 screening assays (Borgert, Mihaich, Ortego, et al. 2011; Borgert et al. 2014). This approach defines specific hypotheses related to interactions with specific MoAs – in this case, EATS pathways – and weights the importance of endpoints as Rank 1, 2, or 3 according to their mechanistic relevance for each hypothesized MoA. It then applies an interpretive algorithm that sequentially considers responses produced by a chemical in Rank 1, 2, and 3 endpoints, in order of their importance for evaluating the hypothesis. The rankings are described as follows (Borgert et al. 2014):

• Rank 1 endpoints are specific and sensitive for the hypothesis being evaluated. These can be interpreted without clarification from other endpoints and are rarely confounded by nonspecific activity. Rank 1 endpoints are in vivo measurements only, because in vitro responses are rarely able to identify a relevant biological effect.

• Rank 2 endpoints are also specific and sensitive for the hypothesis being evaluated but are less informative than Rank 1 as these are often subject to confounding influences or other modes of action. Rank 2 endpoints include both in vitro and in vivo data.

• Rank 3 endpoints are relevant for the hypothesis being evaluated but only when corroborative of Rank 1 and 2 endpoints. Rank 3 endpoints are not specific for a particular hypothesis and include some in vitro and many apical in vivo endpoints.

The weighting scheme and interpretive algorithm used herein follows the same logic but was adapted to accommodate the types of endpoints assessed in long-term toxicity tests, consistent with previously published data and approaches (Biegel et al. 1998; Andrews et al. 2002; Delclos et al. 2009; National Toxicology Program [NTP] 2010; Mihaich et al. 2017; Neal et al. 2017; Mihaich and Borgert 2018; Afarinesh et al. 2020). Adaptations are necessary because the intent and design of toxicity testing differs greatly from endocrine screening. Toxicity tests evaluate apical endpoints intended to reflect adverse effects, irrespective of the MoA that produces the adverse effect, whereas endocrine screening assays measure endpoints intended to reflect interaction with a very specific pathway within the endocrine system irrespective of whether that interaction portends adverse effects. Results of screening assays cannot determine whether a chemical has endocrine disruptive properties because they do not identify adverse effects. According to the World Health Organization/International Programme on Chemical Safety definition, endocrine disruptive properties are identified when the mode(s) of endocrine activity identified in endocrine screens are shown to be causal for the adverse effects observed in toxicity tests (WHO/IPCS 2002).

In keeping with their purpose as mechanistic screening assays, several endpoints evaluated in the U.S. EPA EDSP were assigned a Rank 1 relevance, meaning that they are “… specific and sensitive for the hypothesis, are interpretable without knowing the response of other endpoints, and are in vivo measurements rarely confounded by artifacts or non-specific activity” (Borgert et al. 2014). However, few endpoints measured in assays with styrene achieve such a high level of mechanistic specificity for an endocrine-mediated effect; consequently, there are several data gaps for Rank 1 endpoints measured following styrene exposure. Endpoints measured in toxicity tests were assigned to relevance Ranks 2 and 3 according to their specificity for each MoA hypothesis. Rank 2 endpoints are like Rank 1 endpoints except that they are less informative because they may be subject to confounding influences that make interpretation difficult. Rank 3 endpoints are only corroborative of other endpoint responses in the overall WoE but cannot be used alone to support a hypothesis (Borgert et al. 2014).

Because the rankings were developed for application within but not across hypotheses, the ranking of an endpoint may differ from hypothesis to hypothesis, even if the endpoint is measured within the same type of study (e.g. reproduction study). Also, because different types of studies may measure an endpoint under different conditions (e.g. life-stage; exposure initiation and duration, etc.), the same endpoint may be ranked differently among different types of studies in which it is measured (e.g. repeat dose toxicity versus reproductive toxicity). The rankings are based on empirical observations conducted in studies with known hormonal agonists and antagonists, albeit limited in number and scope, and so these rankings will need to be updated to reflect new knowledge as it develops.

A general principal of endocrine screening and testing is that the results of in vivo tests supersede the results of screening assays (Endocrine Disruptor Screening and Testing Advisory Committee [EDSTAC] 1998; Borgert, Mihaich, Ortego, et al. 2011; Borgert, Mihaich, Quill, et al. 2011; U.S. EPA 2011) because a potential endocrine MoA (i.e. a potential for endocrine activity in vivo) is meaningless unless the chemical produces an adverse effect via that activity. Thus, a lack of response in toxicity endpoints is a strong indication that the chemical lacks endocrine disruptive properties; however, an observed response is ambiguous because these tests measure apical endpoints that can be affected by both endocrine and non-endocrine MoAs. Positive results in toxicity tests cannot be assumed to reflect endocrine disruptive properties unless a causal linkage is made to an underlying endocrine MoA. A second general principle of endocrine screening and testing is that no single assay or endpoint should be considered in isolation. Hence, this WoE evaluation considered whether styrene produces a pattern of adverse effects consistent with EATS agonist or antagonist activity via estrogen, androgen, or thyroid pathways, or consistent with induction or inhibition of steroidogenic enzymes. Although complete consistency of response across all endpoints relevant for a particular MoA might occur only with a highly potent prototype chemical (e.g. ethinyl estradiol for the estrogen agonist MoA), this would not be required for inferring potential endocrine activity. However, a preponderance of negative responses among the relevant endpoints would suggest the absence of activity, especially when the pattern is inconsistent among different studies of similar design.

It is important to appreciate that Rank 2 endpoints may be less specific than Rank 1 endpoints with respect to MoA, particularly in sub-chronic and chronic toxicity studies such as repeat dose toxicity, developmental toxicity, developmental neurotoxicity, and reproductive toxicity studies. The apical nature of the endpoints measured in those types of studies can render them incapable of distinguishing between agonists and antagonists of a particular pathway. Consequently, agonists and antagonists may be expected to produce the same direction of response in some endpoints. One underlying reason is that many endocrine-responsive tissues and processes are maintained by normal hormone levels but may be altered in the same direction when levels of a hormone are either sub- or supra-normal. Parameters that reflect reproductive function, such as fertility and litter size, are prime examples, as are the histological appearance of endocrine glands. These are optimal when normal levels of hormones are present but decrease with either insufficient or excess hormonal activity. Atrophy of some tissues can occur due to reduced levels of a stimulatory hormone whose release is depressed by negative feedback control following long-term administration of an agonist and may also atrophy due to blockade of the receptor in that tissue by antagonists of the same pathway. Hypertrophic responses can follow a similar pattern in other organs and tissues, and hormone levels themselves can change uniformly in response to agonists or antagonists, depending on dose and duration of administration. For these reasons, it is essential to consider the consistency of response expected across several endpoints relevant to each MoA.

In general, results published as statistically significant differences from concurrent controls were considered to be responses for purposes of this WoE evaluation, even though in several instances, authors of the published studies deemed the results to be unrelated to styrene treatment. This practice ensured that the WoE evaluation utilized all available data and was not biased toward negative interpretations due to preemptory exclusion of results. Nonetheless, the authors’ conclusions were considered and used in the WoE evaluation to interpret whether observations were consistent with EATS MoAs. Results that were not statistically distinguishable from concurrent controls were considered to be negative for this WoE evaluation, regardless of whether the authors noted a dose-responsive trend or a trend toward statistical significance. This was considered a reasonable practice given the high doses of styrene typically administered in toxicology studies and the sensitivity of endocrine endpoints to true agonists and antagonists. Although some might contend that such trends could represent low-dose nonmonotonic endocrine-mediated activity, two lines of reasoning argue against that possibility for the endpoints evaluated herein. First, this WoE evaluation did not exclude statistically significant differences that simply failed to be evident at all doses. Therefore, statistically significant nonmonotonicity would not be discounted. Second, placing mechanistic interpretations on apparent “trends” that occur within the normal biological range for hormone-sensitive endpoints, which may result from natural hormonal fluctuation, would be extremely tenuous.

The magnitude of the responses at individual endpoints was generally not informative of the potency by which styrene might affect an endocrine pathway because the data available for this WoE evaluation comprise apical endpoints that can be affected by various MoAs other than EATS. A low magnitude of response in such assays would not necessarily negate an endocrine MoA, nor would a high magnitude of response ensure one. On the other hand, magnitude of response was indeed considered, but in a more general context. Authors often noted when the magnitude of the response was within the normal range for the test species, i.e. within the range of historical controls, and this was taken into consideration in the interpretation, as was when the magnitude of the response was so low as to be of questionable biological relevance. Since known agonists and antagonists of EATS pathways produce characteristic patterns of response, this WoE focused on the pattern of endpoint responses affected by styrene rather than the relying primarily on an evaluation of the magnitude of the responses.

Results – data selection, data quality evaluation, and high-throughput screening

Twenty-two studies contained relevant data for evaluating the potential for styrene to interact via EATS pathways. Each study was assigned a number, as listed in Appendix A, for use in identifying endpoint responses in Supplementary Tables 1–6. To avoid confusion, the reader must understand that in Supplementary Tables 1–6, the second column, “Assay” refers to the general types of toxicological studies from which endpoint responses were extracted, but the endpoints themselves may not be unique or specific to that type of study or to the category of toxicity it addresses. Rather, each endpoint is included because of its relevance for evaluating the underlying endocrine MoA(s) that can affect it. The same understanding must extend to the discussion of Supplementary Tables 1–6 in the text. Only three of these 22 studies differed significantly from guideline regulatory toxicology studies. Two evaluated male reproductive tract effects in Wistar rats. Chamkhia et al. (2006), listed as [3] in Supplementary Tables 1–6, evaluated male endpoints following 10 days of styrene administration that are similar to endpoints evaluated in the Hershberger and other guideline toxicology studies, and Srivastava et al. (1989), [20], evaluated testicular effects and sperm counts following 60 days of styrene administration. Lindbohm et al. (1985), [17], evaluated pregnancy outcomes among styrene-exposed workers.

ToxRTool summary

All 22 studies were evaluated using the Toxicological data Reliability Tool, knowns as ToxRTool. The tool comprises a list of 21 criteria for in vivo studies and 18 criteria for in vitro studies. Each criterion is assigned a value of either “1” (criterion met) or “0” (criterion not met); thus, a maximum reliability score of 21 is possible. Criteria are divided into five groups: I. Test substance identification, II. Test system characterization, III. Study design description, IV. Study results documentation and V. Plausibility of study design and results. Eleven of the 22 studies used in this evaluation (1, 2, 3. 4, 5, 12, 15, 16, 18, 19, 22) met all ToxRTool reliability criteria (21 for in vivo and 18 for in vitro studies). Appendix A provides a summary of ToxRTool evaluations for all 22 studies, including the numerical designation for each study used in Supplementary Tables 1–6, and an explanation of deficiencies for studies that did not meet all criteria.

ToxRTool scores were not used to eliminate studies from consideration nor to discount the results of any studies for relevance to the hypothesis or strength (magnitude) of response weighting. ToxRTool was used in the interpretation of the overall quality of the data selected for WoE evaluation, which is deemed to be adequate to strong, and in the interpretation of conflicting results between studies. Most interpretations were not affected because the ToxRTool scores did not differ significantly for studies that reported an effect of styrene versus those that reported no effect of styrene, although, as discussed in subsequent sections, for most endpoints, the majority of studies found no response.

Results for styrene in U.S. EPA’s ToxCast^® high throughput and EDSP 21 assays

Styrene has been evaluated in the U.S. EPA’s ToxCast^® suite of high-throughput in vitro and in silico assays. Styrene showed no activity in any EDSP21 assay, and very little activity in ToxCast^® assays generally. Styrene showed no activity in the six assays that probed potential activity (agonism or antagonism) via estrogen receptors; no activity in any of the eight assays that probed potential activity (agonism or antagonism) via androgen receptors; no activity in the six assays that probed potential activity (agonism or antagonism) via thyroid hormone or thyroid stimulating hormone receptors, and; no activity in either assay for aromatase. It is not possible at this time to assign a relevance weight ranking for these data, however, due to uncertainties raised by quality control checks on the ToxCast^® data for styrene that have been conducted since the EDSP21 Dashboard and original ToxCast^® Chemicals Dashboard were relocated to EPA’s current Actor website. Supplemental Material C shows graphs and tables extracted from the U.S. EPA’s original EDSP21 Dashboard and its CompTox Chemicals Dashboard, last accessed on 20 September 2020, as well as graphical results from the EPA’s latest algorithm for evaluating ToxCast^® data accessed 25 February 2022.

Results – endocrine WoE evaluation for styrene

Evaluation of estrogen agonist MoA

Supplementary Table 1 shows that of the two possible Rank 1 endpoints for the estrogen agonist hypothesis, styrene failed to increase uterus weight in the rat uterotrophic assay [21] but was not tested for vitellogenin production in male fish. Thus, the Rank 1 endpoints evaluated suggest that styrene lacks potential estrogen agonist activity. That conclusion is supported by the Rank 2 endpoints relevant to the estrogen agonist mode of action that were evaluated in studies with styrene.

A total of 29 of a possible 53 Rank 2 endpoints relevant for evaluating the estrogen agonist MoA were measured among one estrogen receptor transactivation (ERTA) study [22], 10 repeat dose toxicity studies [1, 2, 3, 9, 10, 11, 15, 18, 19, 20], two developmental toxicity studies [12, 16], one developmental neurotoxicity study [5], and two reproductive toxicity studies [2, 4] conducted with styrene. None of those 29 endpoints consistently responded to styrene exposure in all studies in which it was measured, and 26 were unaffected in any study in which it was measured. Only three endpoints responded to styrene in the direction expected for the estrogen agonist mode of action – testis weight, testis histopathology (atrophy), and post-implantation loss – but responded inconsistently among studies in which these endpoints were measured. Styrene increased testis weight in one repeat dose toxicity study [3], reduced it in another [9], and had no effect on testis weight in eight others in which it was measured [1, 2, 10, 11, 15, 18, 19, 20]. Styrene altered testis histopathology in two repeat dose toxicity studies [3, 20] but atrophy was evident only in the latter [20]. Styrene had no effect on testis histopathology in seven other repeat dose studies in which it was evaluated [1, 2, 10, 11, 15, 18, 19]. No study observed both reduced testis weight and altered testis histopathology, which would be expected of a true estrogen agonist effect based on the testicular response to both 17β-estradiol and ethinylestradiol (Biegel et al. 1998; NTP 2010). Only one of the three studies in which testicular effects were observed [3] obtained a ToxRTool^® score of 21. In that study, the increase in testis weight was slight and due to the type of histopathological changes noted, was interpreted by the authors as an indirect effect secondary to high dose toxicity rather than to an endocrine MoA, a conclusion consistent with that of other authors who noted histopathological effects in the testes secondary to alteration of several testicular enzyme activities [20]. Thus, the response to styrene in these two endpoints is neither consistent across studies nor consistent with potential estrogen agonist activity. No repeat dose toxicity study with styrene produced the effect expected of an estrogen agonist on histopathology of the epididymis, mammary gland, ovaries, uterus, or vagina, or on weight of epididymis, ovaries or uterus (Supplementary Table 1).

Supplementary Table 1 shows that time to vaginal patency was unaffected by styrene in a developmental neurotoxicity study [5], and that post-implantation loss was increased in mice and hamsters in one developmental toxicity study with styrene [16] but not in rats or rabbits in another [12]. The number of corpora lutea and pre-implantation losses were not evaluated in those studies. Supplementary Table 1 also shows that styrene produced no changes among 16 of a possible 26 Rank 2 endpoints measurable in reproductive toxicity studies, including number of corpora lutea, epididymal sperm counts, estrus cyclicity, fertility, gestational length, number of implantations, litter size, mating index, ovarian follicle counts, ovary weights and histopathology, testis weight and histopathology, time to mating, and uterus weights and histopathology. Only two of a possible 15 Rank 3 endpoints (estrogen receptor binding – ERBA; gross pathology) were measured in response to styrene. Estrogen receptor binding was not observed in either assay [21, 22], and gross pathology was observed in [18] but not in other studies [1, 15], consistent with the general lack of responses observed in Rank 1 and Rank 2 endpoints.

In summary, the endpoint considered to be the most dispositive for the estrogen agonist MoA – uterotrophic activity in female rats – was not affected by exposure to styrene. Even if the three responses to styrene among 29 Rank 2 endpoints were to be considered biologically significant, which seems highly unlikely given the inconsistency of the observations, the overall pattern of endpoints that responded to administration of styrene is strongly inconsistent with a pattern of responses expected of a chemical with the potential to act via an estrogen agonist MoA. Although 38 of the possible 70 endpoints relevant to the estrogen agonist MoA were not measured in the available studies, 32 endpoints were measured and of those, only three appeared to respond to styrene exposure, but inconsistently across studies. Despite some data gaps, the data nonetheless provide sufficient evidence to conclude that the pattern of endpoint responses elicited by styrene is inconsistent with activity via the estrogen agonist MoA and, thus, that styrene lacks the potential to act via this endocrine MoA. That conclusion is strengthened by the very low proportion of endpoints that responded in any study, their questionable biological relevance due to the small magnitude of responses observed and the fact that some were within the normal range for the rodent test strain. Those factors and the lack of consistency of responses across studies confers very high confidence to the conclusion that Styrene lacks the potential to act via the estrogen agonist MoA

Evaluation of estrogen antagonist MoA

Supplementary Table 2 shows that no Rank 1 endpoints were measured relevant to the estrogen antagonist hypothesis for styrene. Seventeen of a possible 26 Rank 2 endpoints for the estrogen antagonist mode of action were evaluated in 10 repeat dose toxicity studies [1, 2, 3, 9, 10, 11, 15, 18, 19, 20], one developmental neurotoxicity study [5], and two reproductive toxicity studies [2, 4] conducted with styrene. Among those 17 Rank 2 endpoints that were measured, 15 did not respond to styrene, indicating a lack of potential anti-estrogenic activity.

Among U.S. EPA’s (U.S. EPA 2009a) EDSP Tier 1 screening-level assays (ERBA, FSTRA, Female Pubertal), only one of five relevant Rank 2 endpoints was measured; styrene did not compete for binding at the estrogen receptor in either study in which binding was evaluated [21, 22]. Responses were measured in all six endpoints relevant to the estrogen antagonist MoA in repeat dose toxicity studies. One repeat dose toxicity study reported that styrene increased testis weight [3], another reported a reduction [9], and eight other repeat dose toxicity studies reported that testis weight was unaffected by styrene [1, 2, 10, 11, 15, 18, 19, 20]. Testis histopathology was altered by styrene in two repeat dose toxicity studies [3, 20] but atrophy, the expected change, was evident only in the latter [20]. Styrene had no effect on testis histopathology in seven other repeat dose studies in which it was evaluated [1, 2, 10, 11, 15, 18, 19]. No study observed both reduced testis weight and altered testis histopathology, which would be expected of an estrogen antagonist. Of the three studies in which testicular effects were observed, only one [3] obtained a ToxRTool^® score of 21, and due to the type of histopathological changes noted, the authors of that study interpreted the slight increase in testis weight as an indirect effect secondary to high dose toxicity rather than to an endocrine MoA. That interpretation is consistent with other authors who noted histopathological effects in the testes secondary to the alteration of several testicular enzyme activities [20]. Thus, the response to styrene in these two endpoints is neither consistent across studies nor consistent with potential estrogen antagonist activity. No repeat dose toxicity study with styrene produced the effects expected of an estrogen antagonist on histopathology of the epididymis, ovary, prostate, or seminal vesicle (Supplementary Table 2).

Responses were measured in 9 of the 13 endpoints relevant to the estrogen antagonist hypothesis in two reproductive toxicity studies [2, 4], none of which were altered by administration of styrene, including number of corpora lutea, epididymal sperm count, estrus cyclicity, fertility, litter size, ovary or testis histopathology, testis weight, or time to mating. Time to vaginal patency was unaltered by styrene in one neurodevelopmental toxicity study [5]. Only one of a possible 12 Rank 3 endpoints relevant to estrogen antagonism was measured – gross pathology – and was altered in one study [18] but not in other studies [1, 15].

In summary, among various types of studies, responses to styrene were measured in 18 of a possible 39 Rank 1, 2, and 3 endpoints relevant to the estrogen antagonist MoA. Styrene altered responses in only three of those 18 endpoints, but those three endpoints were not consistently altered among the studies that measured them, and the testicular effects observed were slight and not internally consistent within the studies that observed them. The lack of response to styrene in 15 of the 18 endpoints in which it was measured and the equivocal nature of the response in just three other endpoints provides a sufficient dataset on which to conclude that styrene lacks the potential to act as an estrogen antagonist.

Evaluation of androgen agonist MoA

Supplementary Table 3 shows that responses to styrene were not measured among two possible Rank 1 endpoints for the androgen agonist hypothesis. The Hershberger assay was counted as one Rank 1 response for lack of concordance of five endpoints. Responses to styrene were measured in 27 of a possible 47 Rank 2 endpoints relevant for evaluating the androgen agonist MoA among 10 repeat dose toxicity studies [1, 2, 3, 9, 10, 11, 15, 18, 19, 20], four developmental toxicity studies [6, 7, 12, 13], two reproductive toxicity studies [2, 4] and one developmental neurotoxicity study [5] conducted with styrene. The Hershberger assay was counted as one Rank 2 response for lack of concordance of 2–4 endpoints Among repeat dose toxicity studies conducted with styrene, no study reported alterations of ovary histopathology or weight, prostate weight, seminal vesicle weight, or uterus weight. Only three Rank 2 endpoints responded to styrene among 18 that were measured in ten repeat dose toxicity studies. In repeat dose toxicity studies, styrene showed opposite effects in two studies, increased [3] or reduced [9], but had no effect on testis weight in eight others in which it was measured [1, 2, 10, 11, 15, 18, 19, 20]. Although styrene altered testis histopathology in two repeat dose toxicity studies [3, 20] only the latter [20] showed the expected response (atrophy) for this single endpoint. Styrene did not affect testis histopathology in seven other repeat dose studies [1, 2, 10, 11, 15, 18, 19]. No study showed the expected response for an androgen agonist: reduced testis weight and altered testis histopathology. Testicular effects were observed in three studies, only one [3] of which obtained a ToxRTool^® score of 21. Because the type of histopathological changes were inconsistent with an (anti-)androgenic action, the authors interpreted the slight increase in testes weight as an indirect effect of high dose histopathology rather than to an endocrine MoA, a conclusion consistent with that of other authors who noted histopathological effects in the testes secondary to alteration of several testicular enzyme activities [20]. Thus, the response to styrene in these two endpoints is neither consistent across studies nor consistent with potential androgen agonist activity.

Supplementary Table 3 shows that one Rank 2 endpoint – sperm count – was consistently altered by the administration of styrene among four repeat dose studies that measured this endpoint [3, 9, 10, 20]. The reduction in sperm count was attributed to high-dose toxicity to specific cell types rather than to an endocrine mechanism by the authors of those studies [3, 9, 20]. An important component of that mechanism may be alteration of enzymes critical to energy production and maintenance of normal cellular physiology during critical stages of spermatogenesis [9, 10]. Only one of these four studies, [3], achieved the maximum ToxRTool score of 21; the others received scores of 19, 15, and 16, respectively. Of note, the reproduction study that found no change in sperm count with exposure to styrene, [4], also achieved the maximum ToxRTool score of 21.

Styrene failed to alter any Rank 2 endpoint relevant to the androgen agonist MoA that was measured in developmental or reproductive toxicity tests. Styrene elicited no change in litter size in four developmental toxicity studies that evaluated it [6, 7, 12, 13], and no change in the number of implantations [12] or sex ratio [6]. Among two reproductive toxicity studies [2, 4] and one neurodevelopmental toxicity study [5], styrene did not alter estrous cyclicity [4], fertility [2, 4], number of implantations [4], litter size [2, 4], mating index [4], ovarian follicle count [4], ovary histopathology [2, 4], ovary weight increase [2], sex ratio [2, 4], sperm count [4], testis histopathology [2], testis weight [2, 4 F], time to balano-preputial separation [5], time to mating [4], or time to vaginal opening [5]. Among Rank 3 endpoints, styrene altered gross pathology in one repeat dose toxicity study [8] but not in two other studies [1, 15].

Although many endpoints were not measured that would have provided relevant information for evaluating the potential for styrene to act via an androgen agonist MoA, the available data provide sufficient evidence to conclude that the pattern of endpoint responses elicited by styrene is inconsistent with activity via the androgen agonist MoA. More than one-third (27 of a possible 76) of the endpoints relevant for evaluating the androgen agonist MoA were measured following administration of styrene, and of those, only one endpoint – sperm counts – was consistently affected across the repeat dose toxicity studies that measured it. However, that effect was thought by the study authors to be secondary to high-dose cellular toxicity of styrene rather than due to an endocrine MoA, and sperm counts were not affected in either the parents or offspring in a 2-generation reproduction study [4]. Only three other endpoints were affected by styrene, and none of those were consistently altered in studies that measured them. Therefore, there is high confidence in the conclusion that styrene shows no potential to act via the androgen agonist pathway.

Evaluation of androgen antagonist MoA

Supplementary Table 4 shows that styrene was not assessed in a sufficient number of Hershberger endpoints to evaluate Rank 1 for the androgen antagonist MoA. Styrene failed to alter LABC, seminal vesicle or ventral prostate weights in testosterone-propionate-treated castrate male rats [21], constituting a Rank 2 response for a concordant response among three of five possible Hershberger endpoints. Thus, although an insufficient number of endpoints were measured to evaluate a Rank 1 response (concordance of five of five endpoints required), styrene nonetheless showed a lack of Rank 2 response (concordance of at least two of five required) in the Hershberger assay for the androgen antagonist MoA. A lack of Rank 3 response (one of five endpoints responsive in the Hershberger assay) would then be obligate, but was not counted to avoid overemphasis of one assay. Responses to styrene were measured in 22 of the remaining 44 Rank 2 endpoints relevant for evaluating the androgen antagonist MoA among one androgen receptor binding assay [21] and 10 repeat dose toxicity studies [1, 2, 3, 9, 10, 11, 15, 18, 19, 20], two reproductive toxicity studies [2, 4] and one neurodevelopmental toxicity study [5]. Styrene did not bind to the androgen receptor [21] and produced no changes in any repeat dose toxicity study that measured epididymis weight [20], epididymis histopathology [1, 15, 18, 19], ovary histopathology [1, 2, 15, 18, 19], prostate histopathology [1, 15, 18, 19], prostate weight [3], seminal vesicle histopathology [1, 15, 18, 19], seminal vesicle weight [3], uterus histopathology [1, 2, 15, 18, 19].

Styrene produced responses in only two Rank 2 endpoints measured in repeat dose toxicity studies. Styrene had no effect on testis weight in eight repeat dose toxicity studies in which it was measured [1, 2, 10, 11, 15, 18, 19, 20], but showed opposite effects in two others, increased [3] and decreased [9]. No repeat dose toxicity study observed both reduced testis weight and altered testis histopathology, which would be expected of an androgen antagonist. Testis histopathology was unaffected in seven repeat dose studies in which it was evaluated [1, 2, 11, 15, 18, 19], and in the two studies that reported altered testis histopathology [3, 20], atrophy was evident only in the latter [20]. Only one of the three studies that reported testicular effects [3] obtained a ToxRTool^® score of 21, and in that study, the increase in testis weight was slight and was not accompanied by the histopathological changes expected of androgen agonists or antagonists. The authors interpreted the observations to be an indirect effect secondary to high dose toxicity rather than to an endocrine MoA, consistent with the conclusions of other authors who noted histopathological effects in the testes secondary to alteration of several testicular enzyme activities [20]. Thus, the response to styrene in these two endpoints is neither consistent across studies nor consistent with potential androgen antagonist activity.

No Rank 2 endpoint measured in two reproductive toxicity studies [2, 4] and one neurodevelopmental toxicity study [5] responded to styrene administration, including estrous cyclicity [4], fertility [2, 4], litter size [2, 4], ovary histopathology [2, 4], sperm count [4], sperm motility [4], testis histopathology (atrophy) [2], testis weight [2, 4], time to balano-preputial separation [5], time to mating [4], and uterus histopathology [2]. Among Rank 3 endpoints, styrene altered gross pathology in one repeat dose toxicity study [8] but not in other studies [1, 15].

Although two endpoints were not measured in the Hershberger assay [Ranks 1, 2, and 3] that would have provided relevant information for evaluating styrene’s potential to act via an androgen antagonist MoA, and responses were not measured in 22 of the 44 other Rank 2 assays, the available data provide sufficient evidence to conclude that the pattern of endpoint responses elicited by styrene is inconsistent with activity via the androgen antagonist MoA. Twenty-three of a possible 44 Rank 2 endpoints were measured following the administration of styrene that are relevant for evaluating the androgen antagonist MoA, and only two of those responded. Testis histopathology and weight responded inconsistently among studies that measured them, and the endpoints showed a lack of internal consistency within studies. These effects were interpreted by the study authors to be secondary to high-dose cellular toxicity of styrene rather than due to an endocrine MoA. Therefore, there is high confidence in the conclusion that styrene shows no potential to act via the androgen antagonist pathway.

Evaluation of thyroid inhibition MoA

Supplementary Table 5 shows that no Rank 1 endpoints relevant for evaluating the thyroid inhibition hypothesis were evaluated following the administration of styrene. Seven of a possible 20 Rank 2 endpoints for the thyroid inhibition MoA were measured among five repeat dose toxicity studies [2, 8, 15, 18, 19], three developmental toxicity studies [12, 16, 17], and two reproductive toxicity studies [2, 4]. Responses to styrene exposure were measured for 8 of 19 possible Rank 3 endpoints relevant to a thyroid inhibition MoA in two reproductive toxicity study [2, 4], four developmental toxicity studies [6, 7, 13, 14], one neurodevelopmental toxicity study [5].

Among the Rank 2 endpoints measured, styrene elicited no alteration of thyroid follicular cell histopathology in two reproductive [2, 4] and four repeat dose [2, 15, 18, 19] studies, no change in thyroid hormone levels in one repeat dose toxicity study [8], no change in fetal weight in one developmental study [12], and no decrease in pup survival in two reproductive studies [2, 4]. In developmental toxicity studies, fetal survival was reduced in mice and hamsters [16] but not in rats, rabbits or in pregnant women exposed to styrene [12, 17]. Given the lack of consistency among studies and the fact that the incidence of spontaneous abortion does not appear to be increased among pregnant women exposed to styrene, it is reasonable to interpret this endpoint as non-responsive. Among reproductive toxicity studies, pup growth was reduced in the F2 generation, but not in the F1 generation in one study [4] and not in F1, F2, or F3 in another study [2]. Thus, this endpoint should also be considered non-responsive to styrene.

Styrene produced mixed responses among Rank 3 endpoints. Liver weights were increased in males but not females in one reproduction study [4] but not in either sex in another [2]. Behavioral ontogeny and learning and memory were not affected in any study in which they were measured [5, 13]. Pup survival was affected in one neurodevelopmental study [6] but was not affected in two other studies of the same type [5, 13]. Pup survival was not affected, however, in Reproductive Toxicity studies [2, 4], which is a Rank 2 endpoint and therefore of higher relevance for the thyroid inhibition MoA than pup survival in neurodevelopmental studies. Pup growth was reduced in three Developmental studies [6, 7, 13] in which it was measured but not in another [14], and was affected only in the F2 in one reproduction study, but not in F1 of that study [4], or in F1, F2, and F3 of another study [2], which are Rank 2 for the thyroid inhibition MoA. Motor activity was affected in one developmental study [13] but not in two others [5, 14]; brain morphometry was affected in one developmental study [6] but not in two others [5, 7], and auditory startle response was affected in one developmental study [13] but not in another [5]. Since Rank 3 endpoints are not viewed as being interpretable on their own, but only as corroborative evidence for Rank 1 and Rank 2 endpoints (Borgert et al. 2014), and since no evidence for thyroid inhibition was provided by Rank 2 endpoints, these Rank 3 responses provide no evidence to suggest that styrene has the potential to inhibit thyroid hormone pathways.

In summary, although fewer than half of the endpoints relevant for thyroid inhibition were evaluated in response to styrene, those that were evaluated provide no convincing evidence that styrene has the potential to act via inhibition of thyroid pathways.

Evaluation of steroidogenic enzymes MoA

Supplementary Table 6 shows that no Rank 1 endpoints were measured relevant to evaluating the potential for styrene to interact with steroidogenic enzymes. Sixteen of 25 Rank 2 endpoints for the steroidogenesis MoA were evaluated in studies on styrene in nine repeat dose [1, 2, 3, 10, 11, 15, 18, 19, 20], one developmental toxicity study [6], and three reproductive toxicity studies [2, 4, 17]. To capture live births in pregnant women occupationally exposed to styrene, study [17] was listed under reproductive toxicity studies here, but it was listed under developmental toxicity studies in Supplementary Table 5 under fetal survival. This seemed reasonable since a human study in workers is not unambiguously categorizable under either label. In no repeat dose toxicity study in which they were measured did styrene affect ovary weight [1, 2, 15, 19], ovary or uterus histopathology [1, 2, 15, 18, 19], or uterus weight [2, 19]. Styrene altered testis histopathology in two repeat dose toxicity studies [3, 20], but the expected effect, atrophy, was observed in just one of those [20]. Testicular changes were interpreted by the study authors to be due to high dose cellular toxicity rather than to an endocrine MoA. Testis histopathology was unaffected by styrene in seven other studies [1, 2, 10, 11, 15, 18, 19]. Styrene did not alter sex ratio in the developmental toxicity study that evaluated it [6].

Styrene reduced the number of live births in F2 generation of one reproductive toxicity study [2], but not the F1 or F3 generations of that study, nor in the F1 and F2 generations of another reproductive toxicity study [4]. All other endpoints measured in Reproduction studies with styrene were unchanged relative to controls, including estrous cyclicity [4], fertility [2,4], mating index [4], ovary histopathology [2], sex ratio [2,4], sperm count [4], testis histopathology [2], uterus histopathology [2,4], or uterus weight [2]. The only Rank 3 endpoint measured relevant to the steroidogenesis MoA – gross pathology – was unaltered in one repeat dose toxicity study [18] but not in others [1, 15].

Taken together, of the 37 endpoints relevant for evaluating potential interaction with steroidogenic enzymes, 17 were evaluated subsequent to styrene exposure, only three of those endpoints responded in any study, and none of them consistently across studies. Although 20 endpoints were left unmeasured, the large preponderance of endpoints unresponsive to styrene among the 17 that were measured provides strong evidence that styrene lacks the potential to interact with the steroidogenic pathway.

Discussion

This WoE evaluation was conducted by a recognized methodology and included elements critical for an objective, transparent and rigorous analysis of the potential for styrene to act via EATS modes of action. These elements include a clear problem formulation, a systematic search and selection of literature according to specific criteria, an evaluation of data quality by published methods, the weighting of data relevance by consistent criteria, and interpretation of results according to patterns of responses expected for chemicals and hormones known to operate via these MoAs. The approach to data selection applied here produced a dataset sufficient for a robust WoE evaluation, despite some data gaps. Data were available on which to assess the effect of styrene on many of the endpoints that would respond to EATS MoA in reproductive toxicity, developmental toxicity, and repeat dose toxicity studies.

Data quality was assessed in part by the selection of studies that met the inclusion criteria, which requires data amenable to a relevance weighting of endpoints consistent with the WoE methodology employed. Included studies were also subjected to the familiar ToxRTool scoring system. Although several studies included in the WoE evaluation had relatively low ToxRTool scores (10–16 of 21 maximum; see Appendix A), the most serious deficiencies involved information on the purity of the test substance, ambiguities regarding the dose received or the timing of dosing, or inappropriate statistical analysis based on individual animals rather than litters. Those types of deficiencies would tend to bias the studies toward reporting false positive results due to effects of unrecognized impurities, assumptions that effects were caused by lower doses than were actually administered, and assumptions of statistical significance based on artificially inflated sample sizes. Thus, the body of literature evaluated included in this WoE evaluation of styrene is of sufficient quality to ensure a credible determination, with reasonable certainty that the results were not biased toward a false-negative conclusion. Due to the broad parameters used to conduct the literature and data selection used in this WoE evaluation, the paucity of data gaps in the extant literature, and the objective WoE methodology herein, the conclusions of this evaluation carry a high degree of confidence. Supplementary Table 7 provides a summary of the results of endpoint responses for all six MoAs evaluated.

The WoE methodology employed here is unique in that the endpoints evaluated are ranked according to their relevance for testing each hypothesis. Ideally, relevance rankings would be based on empirical evidence sufficient to calculate positive and negative predictive values. Absent such evidence, the rankings employed here are the interpretations of an expert panel based on empirical observations of endpoint responses to known positive and negative controls for each MoA (Borgert et al. 2014). The rankings differ primarily with respect to the specificity and sensitivity of the endpoints for the hypothesis under consideration. Specificity was judged on two factors: the degree to which the endpoint reflects a response of the physiological system to the MoA under evaluation, and second, the degree to which the endpoint can respond to MoAs other than the one being tested, especially to non-endocrine MoAs. This is an important consideration for any evaluation of endocrine MoAs because the endocrine system retains a homeostatic role in nearly every system in which it controls primary biological functions, such as growth, development, reproduction, or metabolism. Consequently, endocrine pathways not only directly affect, but are also indirectly affected by, many non-endocrine physiological and biochemical processes. Discerning direct from indirect effects relies primarily on understanding the specificity of the response.

In this WoE methodology, specificity is also addressed by the requirement to evaluate the pattern of the responses relevant to each MoA and to interpret patterns inconsistent with the response expected of known effectors and inhibitors as unlikely to indicate activity via that pathway. Although it is theoretically possible for an agonist or antagonist of a particular hormonal pathway to exhibit a novel pattern of endpoint responses, consistent with the “selective response modifier” concept of hormone receptor interactions, a finite number of response types have been identified for the various hormone receptor systems, each consistent with ligand-receptor affinities, potencies, and the tissue distribution of hormone receptors (see Borgert et al. 2018 and references therein). Assessment approaches that fail to appreciate this basic feature of hormonal action, such as the key characteristic approach proposed by La Merrill et al. (2020), have little utility for identifying potential endocrine disruptors.

One might consider how the methodology would perform for a chemical that produces responses in more endpoints relevant to a particular hormonal MoA than does styrene, particularly in Rank 1 and Rank 2 endpoints. For such chemicals, the pattern of responses expected for each MoA would require more nuanced scrutiny, as would the potential for competing MoAs to have affected the endpoints, including hormonal MoAs other than the one under scrutiny. It is important to recognize that there is some potential for ligand-receptor interactions across different hormonal receptors, however, receptor specificity is nonetheless a prominent operative feature of the endocrine system. Androgens are not potent estrogens, etc., and so one would not expect a chemical to exhibit several endocrine MoAs simultaneously. When responses occur across various hormonal MoAs, care must be taken to exclude the potential for systemic toxicity to underly them all.

Although not discussed extensively here because the lack of a pattern of relevant responses to styrene obviates the need to do so, a careful consideration of response weightings, as proposed by Borgert et al. (2014), would likely be critical to the evaluation. For example, the use of response weighting to clarify the endocrine-disruptive potential of a chemical that produces weak responses in Rank 1 and Rank 2 endpoints has been described previously for octamethylcyclotetrasiloxane (Borgert et al. 2018; Matthews 2021). Response weighting is related to the concept of mechanistic potency, a fundamental of receptor, enzyme and transport kinetics applicable to the interactions of biological macromolecules with all small molecules, whether endogenous or exogenous.

The data relevant to potential endocrine activity of styrene evaluated here reveal that styrene does not produce a pattern of results consistent with EATS modes of action and shows no or exceedingly low potential to interact with EATS pathways. Styrene’s lack of potential to act via EATS MoAs is indicated by the lack of consistent responses in endpoints relevant for each EATS hypothesis and the lack of a pattern of responses indicative of any MoA. Supplementary Tables 1–6 show generally mixed results for endpoints relevant to potential EATS activity. In only a few instances are endpoint responses in one or a few studies unopposed by a lack of responses in the same endpoints in other studies, and in those instances, the responses were observed at high doses confounded by systemic toxicity and by other MoAs. As shown in Supplementary Tables 1–6, the pattern(s) of endpoint responses to styrene are inconsistent with EATS activity. The lack of a pattern indicative of an endocrine response by the MoAs evaluated here is important because, although some variation between endocrine modulators is expected due to the possibility for selective endocrine response modifiers (e.g. Kuiper et al. 1999), even selective responses are not random and contradictory to an endocrine mechanism, as are the responses observed with styrene.

Data utilized in this WoE evaluation of styrene derive largely, although not exclusively, from repeat-dose, sub-chronic and chronic toxicity studies, i.e. OSRI, rather than from the types of mechanistic screening assays composing the U.S. EPA’s original EDSP Tier 1 and lower levels of the OECD Toolbox (U.S. EPA 2009a; OECD 2012). This WoE evaluation for styrene suggests that conducting additional EDSP Tier Screens or OECD screening-level assays would yield no additional useful information because responses observed in such screening would only trigger the types of studies and measurements of endpoints evaluated herein, which reveal a lack of potential endocrine activity via EATS MoAs. Furthermore, since much of the available reproductive and developmental studies on styrene are recent (post 2001) and incorporate endpoints sensitive for adverse effects that could be produced via endocrine mechanisms, the Tier 2 data are not lacking or obsolete. The extant reproductive, developmental and repeat-dose toxicology studies have assessed the life-stages identified as most relevant for evaluating adverse effects of styrene that could arise from any MoA, including endocrine. Ecological effects are not anticipated due to the uses of styrene, its physical-chemical properties, and its environmental fate and degradation. That little useful information might be gained from further endocrine screening and testing of styrene would seem to preclude a justification for the use of additional animals to pursue such information.

The fact that the repeat-dose, sub-chronic and chronic reproductive toxicity studies with styrene measure apical endpoints (i.e. many Rank 2 endpoints evaluated herein) carries particular significance. Whereas responses in highly specific endocrine screening assays (Rank 1) are not definitive but indicate potential activity that needs to be tested in definitive, longer-term studies, apical endpoints reflect potential adverse effects that could occur by either endocrine or non-endocrine pathways. Therefore, a lack of response in apical endpoints is more informative, and should be given more weight, than a response. A lack of response in apical endpoints that would be expected to respond to endocrine modulators acting via EATS pathways precludes an endocrine MoA and a non-endocrine MoA that indirectly impacts the endocrine system, whereas a response merely opens the possibility that an endocrine MoA underlies the effect. The predominate lack of response to styrene in these apical endpoints further underscores the compelling nature of the evidence that styrene lacks endocrine activity and endocrine disruptive potential.

Although this evaluation for styrene focused on OSRI as defined by the U.S. EPA’s EDSP, the results of this evaluation are also relevant to regulatory evaluation of endocrine disruptors in the EU, where the focus is on identifying chemicals that can be labeled “endocrine disruptors” based on satisfying the WHO/IPCS definition. Because styrene shows no potential to act via EATS pathways, as shown here, it is not biologically plausible that any adverse effect of styrene occurs consequent to these endocrine MoAs. Although many other endocrine MoAs exist, compelling evidence is lacking that styrene operates by endocrine modes of action not evaluated herein. Therefore, styrene cannot be deemed to be an endocrine disruptor, a potential endocrine disruptor, or to exhibit endocrine disruptive properties based on an objective evaluation of the extant data.

Abbreviations
AMA	=	Amphibian Metamorphosis Assay
ARBA	=	Androgen Receptor Binding Assay
ATSDR	=	Agency for Toxic Substances and Disease Registry
CASRN	=	Chemical Abstracts Service Registry Number
CCL3	=	Third Contaminant Candidate List
CompTox	=	Computational Toxicology Dashboard
DART	=	Developmental and Reproductive Toxicology
EATS	=	Estrogen, Androgen, Thyroid, and Steroidogenesis
EDSP	=	Endocrine Disruptor Screening Program
EDSP21	=	Endocrine Disruptor Screening Program for the 21st Century Dashboard
ERBA	=	Estrogen Receptor Binding Assay
ERTA	=	Estrogen Receptor Transactivation Assay
EU	=	European Union
FSTRA	=	Fish Short-Term Reproduction Assay
HSDB	=	Hazardous Substances Data Bank
IC25	=	Inhibitory Concentration 25%
IPCS	=	International Programme on Chemical Safety
LABC	=	Levator ani bulbocavernosus muscle
LOEC	=	Lowest observable effect concentration
NOEC	=	No observable effect concentration
MoAs	=	Modes of Action
NTP	=	National Toxicology Program
OECD	=	Organization for Economic Cooperation and Development
OSRI	=	Other Scientifically Relevant Information
RfDs	=	Reference Doses
TSCA	=	Toxic Substances Control Act
TSH	=	Thyroid Stimulating Hormone
T4	=	Thyroxin
U.S. EPA	=	United States Environmental Protection Agency
WEELs	=	Workplace Environmental Exposure Limits
WHO	=	World Health Organization
WoE	=	Weight of Evidence

Acknowledgments

The author expresses sincere appreciation for the exceptionally thorough and insightful comments provided by the peer-reviewers of this paper. Their attention to detail is unsurpassed and their reviews improved the manuscript immensely.

Declaration of interest

C. J. Borgert received funding from the Styrene Information and Research Center (SIRC) to conduct this WoE evaluation. SIRC reviewed and commented on drafts of the manuscript. The methodologies, analyses, interpretations, content of the manuscript, and the decision to submit it for publication were made solely by the author and did not depend on approval from SIRC. C. J. Borgert also receives funding from the Endocrine Policy Forum (EPF) to evaluate issues related to endocrine disruption. EPF did not fund or otherwise contribute to this evaluation. The author has no financial interests or affiliations that influenced the interpretations of the science presented and evaluated in this manuscript.

Supplemental material

Supplemental material for this article is available online at https://doi.org/10.1080/10408444.2022.2112652.

Appendix A:

ToxRTool summary

Table

Study number in Supplementary Tables 1–6 and text	Study	ToxRTool ccore (maximum score) in vivo: (21) in vitro: (18)	Deficiency/Comment
[1]	Cruzan et al. 2001	21 (21)
[2]	Beliles 1985	21 (21)
[3]	Chamkhia et al. 2006	21 (21)
[4]	Cruzan et al. 2005a	21 (21)
[5]	Cruzan et al. 2005b	21 (21)
[6]	Katakura et al. 2001	18 (21)	The purity or source of the substance not provided; physico-chemical properties not provided.
[7]	Kishi et al. 1992	17 (21)	The purity or source of the substance not provided. The frequency and duration of exposure not well defined; exposure of dams defined as the length of gestation, but that data is provided as mean length of gestation such that length of exposure varied. Does not state that pregnant dams were actually exposed. Authors only state that rats were exposed, they did not specify if the rats were pregnant dams, or if they were pregnant at what number of days of gestation the exposure was started.
[8]	Santini et al. 2008	13 (21)	Purity of the substance was not provided; source or origin of the substance was not provided. Little information concerning exposure details was provided such as hours worked/day, protection used. Not possible to determine doses administered or concentrations in application media given. Sufficient details of the administration scheme was not given; defined as 38 exposed workers; frequency and duration of exposure was not well explained; exposure was indirectly inferred from mean number of years of working. No estimations of vapor concentrations associated with exposure were provided.
[9]	Srivastava et al. 1992a	19 (21)	The purity of the substance was not quantified. The route was insufficiently described as "oral". This could mean any number of things – the way it is written one could assume gavage, but this was not adequately described.
[10]	Srivastava et al. 1992b	15 (21)	Purity of the substance not provided; source or origin of the substance not provided. Since there was no information about the source or purity of the styrene and the route of administration was oral, there is not sufficient evidence that other factors could be influencing the experimental outcomes. Species information limited to albino rats from Industrial Toxicology Research Centre breeding colony. Age of male offspring were known but not dams treated during gestation. Sufficient details of the administration scheme were not given; doses were described as mg/kg bwt but bwts were not given.
[11]	Takao et al. 2000	14 (21)	Purity of the substance was not provided. Although the concentration of styrene in drinking water was provided; the purity of the styrene would affect the overall concentration. Concentration of styrene in drinking water (ad libitum) is provided but volumes of drinking water consumed by subjects is averaged and therefore experimental doses are not known. Sufficient details of administration scheme were not given. The number of test animals was provided in the figures (n = 7 consistently). The dose/animal cannot be determined due to the use of mean ± SEM for the drinking water intake. The use of means ± SEM for exposure is circumspect.
[12]	Murray et al. 1978	21 (21)
[13]	Kishi et al. 1995	18 (21)	The purity or source of the substance was not provided; physico-chemical properties also was not provided. Cannot rule out oral exposure as styrene was introduced into the cage as a vapor.
[14]	Khanna et al. 1991	17 (21)	The purity or source of the substance was not provided; physico-chemical properties not provided. If the purity is not known, then the exposure in oral or inhalation studies cannot be effectively determined. The authors did not specify the particular type of oral route being used, they simply said oral route. The authors needed to be clearer in specifying if the oral route is via gavage, food, water, etc. The authors' study design is inadequate. The authors report and analyze their data with respect to individual animals, when they should actually be performing the analyses and reporting at the level of the litter. The authors are violating the independence assumption required of their statistical methods.
[15]	Cruzan et al. 1998	21 (21)
[16]	Kankaanpää et al. 1980	21 (21)
[17]	Lindbohm et al. 1985	10 (21)	Female workers were enrolled in a case-control study to evaluate spontaneous abortions among women who handled various types of plastics, many of whom handled more than one type of plastic and about one-third of whom may have been exposed to thermal degradation products of plastics. No estimate of styrene exposure was made, and thus, if exposure to styrene occurred, it was likely to have been confounded by exposure to other chemicals, including thermal degradation products of plastics. Exposure information was collected by written or telephone surveys or occupational physicians or occupational nurses; there is a risk of inaccurate or incomplete information, especially if the factory worked with multiple plastic types. Authors identified low statistical power as a weakness of the study that limited interpretations. Due to the use of hospital registries and occupational physician/nurse surveys, it was not possible for the authors to obtain additional confounding factors such as prior obstetric history, alcohol use, tobacco use, and additional potential confounding or explanatory factors.
[18]	Jersey 1978	21 (21)
[19]	Cruzan et al. 1997	21 (21)
[20]	Srivastava et al. 1989	16 (21)	The authors do not specify the purity of styrene, they only state that it is of "the highest purity." The authors do not state if animals were co-housed; also, do not state room conditions. The authors only state that the animals were treated orally – there is no indication of what type of oral exposure was given.
[21]	Date et al. 2002	20 (21)	Strain information was given for all assays except the Hershberger. Housing conditions were not given, including if animals were co-housed or housed individually.
[22]	Ohno et al. 2001	18 (18)	In vitro study