Review Question

The overall objective of this systematic review is to answer the question what is the effect of in utero exposure to ortho-phthalates on anogenital distance, hypospadias, or fetal testosterone in nonhuman male mammals?

The specific aims of the review are to:

• Identify literature reporting the effects of in utero phthalate exposure on male anogenital distance, hypospadias, or fetal testosterone in nonhuman mammals.
• Extract data on male effects of in utero phthalate exposure on anogenital distance, hypospadias, or fetal testosterone from relevant studies.
• Assess the internal validity (risk of bias) of individual studies.
• Summarize the extent of evidence available.
• Synthesize the evidence using a narrative approach or meta-analysis (if appropriate) considering limitations on data integration, such as study-design heterogeneity.
• Rate the confidence in the body of evidence for studies in nonhuman mammals according to one of five statements: (1) high, (2) moderate, (3) low, (4) very low/no evidence available, or (5) evidence of lack of effects on male reproductive-tract development.

PECO Statement

A PECO (Population, Exposure, Comparator, and Outcome) statement was developed by the review team as an aid to identify search terms and inclusion/exclusion criteria as appropriate for addressing the review question for the systematic review.

Population: Nonhuman male mammals

Exposure:

• In utero exposure to any of the following ortho-phthalates or the corresponding monoester or oxidative metabolites: benzylbutyl phthalate (CAS no. 85-68-7), dibutyl phthalate (CAS no. 84-74-2), diethyl phthalate (CAS 84-66-2), diethylhexyl phthalate (CAS no. 117-81-7), diisobutyl phthalate (CAS no. 84-69-5), diisononyl phthalate (CAS no. 28553-12-0), diisooctyl phthalate (CAS no. 27554-26-3), dimethyl phthalate (CAS no. 131-11-3), di-n-octyl phthalate (CAS no. 117-84-0), diisodecyl phthalate (CAS no. 26761-40-0), and/or dipentyl phthalate (CAS no. 131-18-0).
• Oral route of exposure.

Comparator: Male nonhuman mammals exposed in utero to different doses of phthalates or vehicle-only treatment.

Outcomes:

• Anogenital distance (AGD): the measured distance between the anus and the genitals. Typically measured from the anus to the base of the scrotum or the base of the phallus. Other measures that might be used:
  • Anogenital index (AGI): AGD measurement divided by body weight or by the cube root of body weight
  • Anoscrotal distance (ASD): the measured distance between the anus and base of the scrotum
  • Anopenile distance (APD): the measured distance from the anus to the base of the penis
• Hypospadias (incidence and severity/grade)
• Fetal testosterone concentration (e.g., measured from testes, serum, or plasma taken in utero)

Problem Formulation and Protocol Development

The review question and specific aims were developed and refined through a series of problem formulation steps. The committee considered review articles on endocrine disruptors in surveying the types of chemicals that might make good case examples and held a workshop to explore potential case examples, including phthalates. The committee sought an example of a chemical for which the human and the animal evidence on effects appears to be associated with different exposure levels of that chemical and due to perturbation of the estrogen or androgen hormone system. Phthalates appear to fit this case criterion, and positive feedback was received at the committee's workshop.

Alterations in male reproductive-tract development are the most sensitive effects from exposure to phthalates (NRC 2008). Because the period during in utero sexual differentiation (i.e., the masculinization programming window) is the most sensitive life stage, the exposure period of interest for the systematic review is in utero. Animal studies have illustrated a spectrum of effects in male reproductive development after in utero exposure to phthalates, including under developed or absent reproductive organs, malformed external genitalia (hypospadias), undescended testicles (cryptorchidism), decreased AGD, retained nipples, and decreased sperm production (NRC 2008). The systematic review will focus on end points reflecting androgen-dependent adverse effects (AGD and hypospadias), an adverse effect that occurs at relatively low doses (AGD), and a key event in the adverse outcome pathway leading to reduced AGD and hypospadias (fetal testosterone).

Consideration was given to including cryptorchidism as an end point, but the committee decided against it. The mode of action for phthalate-induced cryptorchidism involves reductions in INSL-3 levels in addition to androgen-dependent mechanisms. Important for the committee's charge, there are few, if any, human studies on dose-response relationships between phthalate exposure and cryptorchidism to compare to animal data. Furthermore, studies have shown that rats exposed to phthalates have similar sensitivity to decreased fetal testosterone and AGD as they do for decreased INSL-3, and that cryptorchidism is a less sensitive end point compared to reductions in AGD. Because the overall objective of the committee is to use this systematic review with the one being conducted on the human evidence to evaluate the coherence between effects and dose-response relationships, the committee judged that it would not be useful to include cryptorchidism in the systematic reviews on phthalates.

The protocol will be peer reviewed by subject-matter and systematic-review experts in accordance with standard report-review practices of the National Academies of Sciences, Engineering, and Medicine. The protocols will be revised in response to peer review comments and will subsequently be published as appendices to the committee's final report. The identity of the peer reviewers will remain anonymous to the committee until the publication of the final report, when their names and affiliations are disclosed in the Preface.

Committee and Staff

There are 11 committee members, supported by two staff members of the National Academies. The committee members were appointed in accordance with the standard policies and practices of the National Academies on the basis of their expertise in general toxicology, reproductive toxicology, developmental toxicology, endocrinology, neurotoxicology, epidemiology, risk assessment, biostatistics, and systematic-review methods. The membership of the committee and the staff was determined before the topic of the systematic review was selected. It was known, however, that each case study would be on an endocrine-disrupting chemical, so committee members who have relevant expertise were specifically recruited and appointed.

Review Team

The review team for this case study will be a subgroup of the committee (DD, KJ), two National Academies staff members (EM, SM), and an information specialist (JB). If a member of the review team was a coauthor of a study under review, that member will recuse himself or herself from the evaluation of the quality of that study.

The review team will be responsible for performing all aspects of the review, including conducting the literature searches; applying inclusion/exclusion criteria to screen studies; extracting data; assessing risk of bias for included studies; and analyzing and synthesizing data. The roles and responsibilities of the team members will be documented throughout the protocol. Throughout the course of its work, the review team will also engage other members of the committee to provide consultation as needed. The involvement of those individuals will be documented and acknowledged.

Biographical information on the review team is presented in Section C-1a.

Search Methods

Literature Search for Independent Systematic Review

The review team will collaborate with an information specialist (JB) who has training, expertise, and familiarity with developing and performing systematic review literature searches. A variety of methods will be used to identify relevant data (see below). Literature searches will not be limited by publication date.

Study Selection

All search results will be imported or manually entered into EndNote (Version x7) reference management software. EndNote will be used to eliminate any duplicate citations before evaluating the eligibility of the citations.

Screening Process

References retrieved from the literature search will be screened for relevance and eligibility against the evidence selection criteria using DistillerSR (Evidence Partners; https://www.evidencepartners.com/). Screeners from the review team will be trained with an initial pilot phase on 25 studies undertaken to improve clarity of the evidence selection criteria and to improve accuracy and consistency among screeners. Screening forms are presented in Section C-1c.

Data Extraction

Data will be collected and recorded (i.e., extracted) from included studies by one member of the review team and checked by a second member for completeness and accuracy. Any discrepancies in data extraction will be resolved through discussion. The extracted data will be used to summarize study designs and findings and/or to conduct statistical analyses. Section C-1d presents the data extraction elements that will be used.

The review team will attempt to contact authors of included studies to obtain missing data considered important for evaluating key study findings (e.g., level of data required to conduct a meta-analysis). The study extraction files will note whether an attempt was made to contact study authors by email for missing data considered important for evaluating key study findings (and whether or not a response was received).

Multiple publications with overlapping data for the same study (e.g., publications reporting subgroups, additional outcomes or exposures outside the scope of an evaluation, or longer follow-up) are identified by examining author affiliations, study designs, cohort name, enrollment criteria, and enrollment dates. If necessary, study authors will be contacted to clarify any uncertainty about the independence of two or more articles. The review will include all publications on the study, select one publication to use as the primary publication, and consider the others as secondary publications with annotation as being related to the primary record during data extraction. The primary study will generally be the publication with the longest follow-up or, for studies with equivalent follow-up periods, the study with the largest number of cases or the most recent publication date. The review will include relevant data from all publications of the study, although if the same outcome is reported in more than one report, the review team will include a single instance of the data (and avoid more than one—that is, duplicate instances of the data).

Data extraction will be completed using the Health Assessment Workspace Collaborative (HAWC) software, an open source and freely available Web-based interface application, for visualization and warehousing.²

Risk of Bias (Quality) Assessment of Indiviual Studies

Risk of bias is related to the internal validity of a study and reflects study-design characteristics that can introduce a systematic error (or deviation from the true effect) that might affect the magnitude and even the direction of the apparent effect. Internal validity or risk of bias will be assessed for individual studies using a tool developed by the National Toxicology Program's Office of Health Assessment and Translation (OHAT) that outlines an approach to evaluating risk of bias for experimental animal studies. The risk of bias domains and questions for experimental animal studies are based on established guidance for experimental human studies (randomized clinical trials) (Higgins and Green 2011; Viswanathan et al. 2012, 2013; Sterne et al. 2014) and recent tools for animal studies (Hooijmans et al. 2014; Koustas et al. 2014). The riskof bias tool includes a common set of questions (Section C-1e) that are answered based on the specific details of individual studies to develop risk of bias ratings (using the four options: definitely low risk of bias; probably low risk of bias; probably high risk of bias; or definitely high risk of bias). Information or study procedures that were not reported are assumed not to have been conducted, resulting in an assessment of “probably high” risk of bias. Study design determines the subset of questions that should be used to assess risk of bias for an individual study (see Table C1-1).

Studies are independently assessed by two assessors (DD, KJ) who answer all applicable risk of bias questions with one of four options (see Table C1-2) following prespecified criteria detailed in Section C-1e. The criteria describe aspects of study design, conduct, and reporting required to reach risk of bias ratings for each question and specify factors that can distinguish among ratings (e.g., what separates “definitely low” from “probably low” risk of bias). The instructions and detailed criteria are tailored to the specific type of human study designs. Risk of bias will be assessed at the outcome level because study design or method specifics may increase the risk of bias for some outcomes and not others within the same study. Information or study procedures that were not reported are assumed not to have been conducted, resulting in an assessment of “probably high” risk of bias. Authors will be queried by email to obtain missing information, and responses received will be used to update riskof bias ratings.

Assessors will be trained using the criteria in an initial pilot phase undertaken to improve clarity of criteria that distinguish between adjacent ratings and to improve consistency among assessors. All team members involved in the risk of bias assessment will be trained on the same set of studies and asked to identify potential ambiguities in the criteria used to assign ratings for each question. Any ambiguities and rating conflicts will be discussed relative to opportunities to refine the criteria to more clearly distinguish between adjacent ratings. If major changes to the risk of bias criteria are made based on the pilot phase (i.e., those that would likely result in revision of response), they will be documented in a protocol amendment along with the date modifications were made and the logic for the changes. It is also expected that information about confounding, exposure characterization, outcome assessment, and other important issues may be identified during or after data extraction, which can lead to further refinement of the risk of bias criteria.

After assessors have independently made risk of bias determinations for a study across all risk of bias questions, the two assessors will compare their results to identify discrepancies and attempt to resolve them. Any remaining discrepancies will be considered and resolved with the review team. The final riskof bias rating for each question will be recorded along with a statement of the basis for that rating.

Data Analysis and Evidence Synthesis

The review team will qualitatively synthesize the body of evidence for each outcome and, where appropriate, a meta-analysis will be performed. If a meta-analysis is performed, summaries of main characteristics for each included study will be compiled and reviewed by two team members to determine comparability between studies, to identify data transformations necessary to ensure comparability, and to determine whether heterogeneity is a concern. The main characteristics considered across all eligible studies include the following:

• Experimental design (e.g., acute, chronic, multigenerational)
• Animal model used (e.g., species, strain, sex, genetic background)
• Age of animals (e.g., at start of treatment, mating, and/or pregnancy status)
• Developmental stage of animals at treatment and outcome assessment
• Dose levels, frequency of treatment, timing, duration, and exposure route
• Health outcome(s) reported
• Type of data (e.g., continuous or dichotomous), statistics presented in paper, access to raw data
• Variation in degree of risk of bias at individual study level

The review team expects to require input from subject-matter experts to help assess the heterogeneity of the studies. Subgroup analyses to examine the extent to which risk of bias contributes to heterogeneity will be performed. If there is evidence of species differences, the review team will consider stratifying by species and performing separate meta-analyses by species. Situations where it may not be appropriate to include a study are when data on exposure or outcome are too different to be combined or other circumstances that may indicate that averaging study results would not produce meaningful results. When considering outcome measures for conducting meta-analyses, benchmark dose (BMD) estimates (and their associated confidence intervals) with a benchmark response (BMR) set to a common percent of control (for continuous outcomes) or extra risk (for dichotomous outcomes) are preferred. A secondary alternative, when there are more than two groups, is the conduct of BMD modeling and the use of the derived BMD estimates. Meta-analyses are not possible with lowest-observed-adverse-effect levels or no-observed-adverse-effect levels, since no confidence interval can be derived for these measures.

If a meta-analysis is conducted, a random effects model will be used for the analysis. Heterogeneity will be assessed using the I-squared statistic. Interpretation of I-squared will be based on the Cochrane Handbook: 0% to 40% (might not be important); 30% to 60% (may represent moderate heterogeneity); 50% to 90% (may represent substantial heterogeneity); and 75% to 100% (considerable heterogeneity). Additionally, as described in the Cochrane Handbook, for the last three categories, the importance of the I-squared will be interpreted considering not only the magnitude of effects but also the strength of the evidence (90% two-tailed confidence interval).

The review team will also perform sensitivity analyses on the exclusion of individual studies in succession.

If sufficient studies are available, subgroup analyses will be performed based on the following characteristics described above: experimental design, animal model used (e.g., species and/or strain), age of animals, and developmental stage of animals at treatment and outcome assessment.

In the event that these proposed methods for data analysis are altered to tailor to the evidence base from included studies, the protocol will be amended accordingly, and the reasons for change will be justified in the documentation.

Confidence Rating: Assessment of the Body of Evidence

The quality of evidence for each male reproductive outcome will be evaluated using the GRADE system for rating the confidence in the body of evidence (Guyatt et al. 2011; Rooney et al. 2014). More detailed guidance on reaching confidence ratings in the body of evidence as “high,” “moderate,” “low,” or “very low” is provided in NTP (2015, see Step 5). In brief, available studies on a particular outcome are initially grouped by key study-design features, and each grouping of studies is given an initial confidence rating by those features.

The initial rating is downgraded for factors that decrease confidence in the results, including

• risk of bias
• unexplained inconsistency
• indirectness or lack of applicability
• imprecision
• publication bias

The initial rating is upgraded for factors that increase confidence in the results, including

• large magnitude of effect
• dose response
• consistency across study designs/populations/animal models or species
• consideration of residual confounding
• other factors that increase confidence in the association or effect (e.g., particularly rare outcomes)

TABLE C1-1

OHAT Risk of Bias Tool.

TABLE C1-2

Answers to the Risk of Bias Questions.

The reasons for downgrading (or upgrading) confidence may not be due to a single domain of the body of evidence. If a decision to downgrade is borderline for two domains, the body of evidence is downgraded once in a single domain to account for both partial concerns based on considering the key drivers of the strengths or weaknesses. Similarly, the body of evidence is not downgraded twice for what is essentially the same limitation (or upgraded twice for the same asset) that could be considered applicable to more than one domain of the body of evidence. Consideration of consistency across study designs, human populations, or animal species is not included in the GRADE guidance (Guyatt et al. 2011); however, it is considered in the modified version of GRADE used by OHAT (Rooney et al. 2014).

Confidence ratings are independently assessed by members of the review team, and discrepancies will be resolved by consensus and consultation with technical advisors as needed. Confidence ratings will be summarized in evidence profile tables.

Electronic Searches

Search Strategies

Title and Abstract Screening Form

INSTRUCTIONS: When a citation is excluded, reason should be specified.

Full-Text Screening Form

INSTRUCTIONS: When a citation is excluded, reason should be specified.

Addition of an Exclusion Criteron for Full-Text Screening

The following criterion for excluding studies at the full-text screening level was added on September 15, 2016, after the protocol was peer reviewed:

Study involved rodents exposed to a single high dose (≥500 mg/kg/day).

The committee judged that a study testing only a single high dose level would not be useful for a systematic review intended to address the low-dose toxicity of phthalates.

Additions to the Review Team

The following committee members were added to the review team to supplement expertise:

• Weihsueh Chiu is a professor in the Department of Veterinary Integrative Biosciences in the College of Veterinary Medicine and Biomedical Sciences at Texas A&M University. Before joining the university, he worked at the US Environmental Protection Agency (EPA) for more than 14 years, most recently as chief of the Toxicity Pathways Branch in the Integrated Risk Information System (IRIS) Division of the National Center for Environmental Assessment. His research has focused on human health risk assessment, particularly with respect to toxicokinetics, mechanisms of toxicity, physiologically based pharmacokinetic modeling, dose-response assessment, and characterizing uncertainty and variability. He led the development of EPA's 2011 IRIS assessment of trichloroethylene, which pioneered the use of probabilistic methods for characterizing uncertainty and variability in toxicokinetics and dose-response. He is currently a member of the NRC's Committee on Predictive-Toxicology Approaches for Military Assessments of Acute Exposures. Dr. Chiu received a PhD in physics from Princeton University.
• Katrina Waters is deputy director of the Biological Sciences Division at the Pacific Northwest National Laboratory. Her research interests are focused on the integration of genomics, proteomics, metabolomics and high-throughput screening data to enable predictive mechanistic modeling of disease and toxicity pathways. She currently serves on EPA's Board of Scientific Counselors Subcommittee on Chemical Safety for Sustainability and the US Food and Drug Administration's Scientific Advisory Board to the National Center for Toxicological Research. She recently served on the NRC's Committee on Predictive-Toxicology Approaches for Military Assessments of Acute Exposures. Dr. Waters received a PhD in biochemistry from the University of Wisconsin–Madison, and did a postdoctoral fellowship on endocrine disruptors at the Chemical Industry Institute of Toxicology.

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. Two of the three mouse studies did not control or account for litter effects, and the studies had issues with outcome assessment and lack of blinding of the researchers to the study groups during outcome assessment (see Figure C4-1). Six of 16 rat studies did not account for litter effects, and most of the studies also had issues with outcome assessment and blinding of the researchers (see Figure C4-2).
• Unexplained inconsistencies: No downgrade. Although there appeared to be heterogeneity in the results (see Figure C4-3), most of it could be explained by dose, species, or strain differences. Meta-analyses of the data found no important heterogeneity in the rat or the mouse data (see Appendix C, Section C-5), further supporting the decision not to downgrade.
• Indirectness: No downgrade.
• Imprecision: No downgrade. Mean versus standard deviation for most studies reflects reasonable precision (see Figure C4-3). Meta-analyses of the data found a statistically significant summary overall estimate for rats but not for mice (see Appendix C, Section C-5). Because the mouse studies account for a small percentage of the overall body of evidence (3 of 19 studies), confidence was not downgraded for imprecision.
• Publication bias: No downgrade (see Appendix C, Section C-3).

FIGURE C4-1

Risk of bias heatmap of studies of DEHP and AGD in mice. In HAWC: https://hawcproject.org/summary/visual/302/.

FIGURE C4-2

Risk of bias heatmap of studies of DEHP and AGD in rats. In HAWC: https://hawcproject.org/summary/visual/319/.

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Meta-analysis of the data showed that, in rats, the effects could be considered large and robust, with overall summary estimates having z-scores of ≥7.0 (see Appendix C, Section C-5). The effect sizes were robust to multiple sensitivity analyses.
• Dose-response: Upgraded. Although the visualization in Figure C4-4 suggests some inconsistency in dose response across studies, an upgrade is supported by the meta-analysis of the rat data, which found statistically significant linear trends in log₁₀ (dose) or dose. The results were robust to multiple sensitivity analyses.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade.

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. Most studies accounted for litter effects and used reliable methods of measuring fetal testosterone. See Figure C4-5.
• Unexplained inconsistencies: No downgrade. Consistent decrease in fetal testosterone across studies, with a few exceptions that can be explained by study design features (e.g., examining testosterone in fetal plasma, which might have technical difficulties). See Figure C4-6. A meta-analysis of the data also supported the decision not to downgrade (see Appendix C, Section C-5).
• Indirectness: No downgrade.
• Imprecision: No downgrade. A meta-analysis of the data found a statistically significant summary overall estimate, linear trend in log₁₀(dose), and linear trend in dose, which were robust to multiple sensitivity analyses.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. High dose groups reflect a relatively large magnitude of change (about 75-90%) across several studies. See Figure C4-6. A meta-analysis of the data also supported the decision to upgrade (see Appendix C, Section C-5).
• Dose-response: Upgraded. Several studies reflect a dose response in the same dose ranges (see Figure C4-7). A meta-analysis of the data also supported the decision to upgrade (see Appendix C, Section C-5).
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only one mouse study was available so cross-species consistency could not be evaluated. Results are generally consistent across studies in the high dose range.

FIGURE C4-5

Risk of bias heatmap of studies of DEHP and fetal testosterone in rodents. In HAWC: https://hawcproject.org/summary/visual/362/.

FIGURE C4-6

Data pivot of animal studies of DEHP and fetal testosterone sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dehp-effect-agd/.

FIGURE C4-7

Data pivot of animal studies of DEHP and fetal testosterone sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dehp-effect-testosterone-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. Over half of the studies had a probably high risk of bias rating because they lacked reporting on the outcome assessment. Other concerns were related to whether the researchers were blinded to the study groups during outcome assessment and not controlling for litter effects. See Figure C4-8.
• Unexplained inconsistencies: Downgraded. Incidence of hypospadias is not consistent across studies within similar dose ranges (e.g., Christiansen et al. [2009, 2010] and Jarfelt et al. [2005] show no increased incidence at doses of 750 mg/kg-day or higher). See Figure C4-9.
• Indirectness: No downgrade.
• Imprecision: No downgrade. No confidence intervals for incidence data, but no hypospadias in control groups. See Figure C4-9.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: No upgrade. Incidence of hypospadias is not consistently large in magnitude across studies for high dose groups. See Figure C4-9.
• Dose-response: No upgrade. Dose response noted for a couple of the studies but not consistently across studies. See Figure C4-10.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgraded. Only one mouse study was available so cross-species consistency could not be evaluated. Results are generally not consistent across studies.
• Rare outcome: Upgraded. Background control incidence of hypospadias was reported as zero across all studies, so any positive finding was considered treatment related.

FIGURE C4-8

Risk of bias heatmap of studies of DEHP and hypospadias in rodents. In HAWC: https://hawcproject.org/summary/visual/360/.

FIGURE C4-9

Data pivot of animal studies of DEHP and hypospadias (% animals affected) sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dehp-effect-hypospadias-animals-affected/. The following links have additional visualizations (more...)

FIGURE C4-10

Data pivot of animal studies of DEHP and hypospadias (% animals affected) sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dehp-effect-hypospadias-animals-affected-dose-resp/. The following links have additional visualizations (more...)

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. All the studies had ratings of probably high risk of bias or definitely high risk of bias in at least one of the key issues considered, and all had multiple risk of bias issues. See Figure C4-11.
• Unexplained inconsistencies: No downgrade. Consistent dose response across most studies with the exception of the study by Aso et al. (2005), which could be explained by study design features. See Figure C4-12.
• Indirectness: No downgrade.
• Imprecision: No downgrade. Mean versus standard deviation for the studies reflects reasonable precision. See Figure C4-12.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Three studies reflect relatively large magnitude of change (about 20-40%) in the same dose range. See Figure C4-12.
• Dose-response: Upgraded. Most studies reflect a dose response in the same dose range. See Figure C4-13.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-11

Risk of bias heatmap of studies of BzBP and AGD in rats. In HAWC: https://hawcproject.org/summary/visual/323/.

FIGURE C4-12

Data pivot of animal studies of BzBP and AGD in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/bzbp-effect-agd/.

FIGURE C4-13

Data pivot of animal studies of BzBP and AGD in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/bzbp-effect-agd-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. Both studies accounted for litter effects. See Figure C4-14.
• Unexplained inconsistencies: No downgrade. See Figure C4-15.
• Indirectness: No downgrade.
• Imprecision: No downgrade. Mean versus standard deviation for the studies reflects reasonable precision. See Figure C4-15.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Higher dose groups reflect a relatively large magnitude of change (about 80% in both studies). See Figure C4-15.
• Dose-response: Upgraded. Both studies reflect a dose response in the same dose range. See Figure C4-16.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-14

Risk of bias heatmap of studies of BzBP and fetal testosterone in rats. In HAWC: https://hawcproject.org/summary/visual/328/.

FIGURE C4-15

Data pivot of animal studies of BzBP and fetal testosterone in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/bzbp-effect-testosterone/.

FIGURE C4-16

Data pivot of animal studies of BzBP and fetal testosterone in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/bzbp-effect-testosterone-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. Both of the studies had probably high risk of bias ratings because of concerns about whether the researchers were blinded to the treatment groups and concerns about the outcome measures. One study did not control for litter effects, but it reported no hypospadias. See Figure C4-17.
• Unexplained inconsistencies: No downgrade. Little response seen in either study. See Figure C4-18.
• Indirectness: No downgrade.
• Imprecision: No downgrade. See Figure C4-18.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: No upgrade. Only a single hypospadias case was reported in the highest dose group in one study. See Figure C4-18.
• Dose-response: No upgrade. See Figure C4-18.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.
• Rare outcome: Because the data are limited and there were risk of bias concerns regarding the outcome measure, confidence was not upgraded for the finding of a rare effect as was done for other phthalates and hypospadias.

FIGURE C4-17

Risk of bias heatmap of studies of BzBP and hypospadias in rats. In HAWC: https://hawcproject.org/summary/visual/335/.

FIGURE C4-18

Data pivot of animal studies of BzBP and hypospadias (% animals affected) in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/bzbp-effect-hypospadias-animals-affected/. The following links have additional visualizations (more...)

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. All studies had ratings of probably high risk of bias or definitely high risk of bias in at least one of the key issues considered, and most of the studies had multiple risk of bias issues. See Figure C4-19.
• Unexplained inconsistencies: No downgrade. Consistent effects observed across multiple studies. Inconsistencies could be explained by study design features. See Figure C4-20.
• Imprecision: No downgrade. Mean versus standard deviation for most studies reflects reasonable precision, with the exception of the study by Struve et al. (2009). See Figure C4-20.
• Indirectness: No downgrade
• Publication bias: No downgrade (see Appendix C, Section C-3).

FIGURE C4-19

Risk of bias heatmap of studies of DBP and AGD in rats. In HAWC: https://hawcproject.org/summary/visual/322/.

FIGURE C4-20

Data pivot of animal studies of DBP and AGD in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-agd/.

FIGURE C4-21

Data pivot of animal studies of DBP and AGD in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-agd-dose-response/.

Factors Considered for Upgrading Confidence

• Large magnitude: No upgrade. Only a few studies demonstrate a large effect (40%) even at higher doses. See Figure C4-20.
• Dose-response: Upgraded. Several studies reflect a dose response. See Figure C4-21.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. See Figure C4-22.
• Unexplained inconsistencies: No downgrade. Consistent effects observed across multiple studies. Inconsistencies could be explained by study design features. See Figure C4-23.
• Indirectness: No downgrade
• Imprecision: No downgrade. Mean versus standard deviation for most studies reflects reasonable precision. See Figure C4-23.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Several studies demonstrate large effects (about 80%) in high dose groups. See Figure C4-23.
• Dose-response: Upgraded. Several studies reflect a dose response. See Figure C4-24.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-22

Risk of bias heatmap of studies of DBP and fetal testosterone in rats. In HAWC: https://hawcproject.org/summary/visual/329/.

FIGURE C4-23

Data pivot of animal studies of DBP and fetal testosterone in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-testosterone/.

FIGURE C4-24

Data pivot of animal studies of DBP and fetal testosterone in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-testosterone-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. Risk of bias concerns included confidence in the outcome assessment and whether the researchers were blinded to the treatment groups. See Figure C4-25.
• Unexplained inconsistencies: No downgrade. Incidence of hypospadias appeared to be consistent across studies within similar dose ranges. See Figure C4-26.
• Indirectness: No downgrade.
• Imprecision: No downgrade. No confidence intervals for incidence data, but no hypospadias in control groups. See Figure C4-26.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. High incidence (about 40%) of hypospadias was found in different studies in high dose groups. See Figure C4-26.
• Dose-response: Upgraded. Dose response was noted for studies and there was general agreement across studies. See Figure C4-27.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.
• Rare outcome: Upgraded. Background control incidence of hypospadias was reported as zero across all studies, so any positive finding was considered treatment related.

FIGURE C4-25

Risk of bias heatmap of studies of DBP and hypospadias in rats. In HAWC: https://hawcproject.org/summary/visual/338/.

FIGURE C4-26

Data pivot of animal studies of DBP and hypospadias in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-hypospadias-animals-affected/. The following links have additional visualizations presenting data (more...)

FIGURE C4-27

Data pivot of animal studies of DBP and hypospadias in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dbp-effect-hypospadias-animals-affected-dose-respo/. The following links have additional visualizations presenting (more...)

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. See Figure C4-28.
• Unexplained inconsistencies: No downgrade. Studies are relatively consistent. See Figure C4-29.
• Indirectness: No downgrade.
• Imprecision: Downgraded. Mean versus standard deviation reflects variable precision across studies, including overlapping error bars between control and significant treatment groups. See Figure C4-29. Meta-analysis of the data supports the downgrade (see Appendix C, Section C-6).
• Publication bias: No downgrade (see Appendix C, Section C-3).

FIGURE C4-28

Risk of bias heatmap of studies of DIBP and fetal testosterone in rats. In HAWC: https://hawcproject.org/summary/visual/332/.

FIGURE C4-29

Data pivot of animal studies of DIBP and fetal testosterone in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dibp-effect-testosterone/.

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Consistently large effects of more than 50% are seen in both studies. See Figure C4-29.
• Dose-response: Upgraded. Dose response is evident in both studies. See Figure C4-30.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-30

Data pivot of animal studies of DIBP and fetal testosterone in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dibp-effect-testosterone-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: Downgraded. Two of the studies had a probably high risk of bias rating in two key areas (whether researchers were blinded to the treatment groups or how outcomes were assessed), and one had a probably high risk of bias rating for not controlling for litter effects. See Figure C4-31.
• Unexplained inconsistencies: Downgraded. Responses are inconsistent across the studies and not explained by methodology or other factors. One would expect a treatment-related decrease in AGD at these dose levels given the increase in fetal testosterone; however, only one of the four studies showed a clear treatment-related decrease in AGD (Boberg et al. 2011). The study by Clewell et al. (2013) did not find decreased AGD, nor did L. Li et al. (2015), although the error was huge in that study. Masutomi et al. (2003) data appear to be equivocal; the mean is lower but the error is large. See Figure C4-32.
• Indirectness: No downgrade
• Imprecision: Downgraded. The L. Li et al. (2015) and Masutomi et al. (2003) studies had larger standard deviations than did the effect measured by Clewell et al. (2013) and Boberg et al. (2011). See Figure C4-32.
• Publication bias: No downgrade (see Appendix C, Section C-3).

FIGURE C4-31

Risk of bias heatmap of studies of DINP and AGD in rats. In HAWC: https://hawcproject.org/summary/visual/324/.

FIGURE C4-32

Data pivot of animal studies of DINP and AGD in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dinp-effect-agd/.

Factors Considered for Upgrading Confidence

• Large magnitude: No upgrade. A large magnitude of effect (less than 20%) was not observed consistently across multiple studies. See Figure C4-32.
• Dose-response: No upgrade. Dose response is not consistent across studies. Only two of four studies show a dose response. The data from Clewell et al. (2013) are internally inconsistent; no effect at PND 2 or PND 49-50, but a statistically-identified decrease in AGD was found at PND 14. See Figure C4-33.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-33

Data pivot of animal studies of DINP and AGD in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dinp-effect-agd-dose-response/.

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. See Figure C4-34.
• Unexplained inconsistencies: No downgrade. Inconsistencies can be explained by study design features (exposure window) and differences in measurements (testosterone in plasma is different from testosterone in the testes). See Figure C4-35.
• Indirectness: No downgrade.
• Imprecision: Downgraded. Mean versus standard deviation reflects variable precision across studies, particularly testosterone production and testosterone measurements in plasma. See Figure C4-35. Meta-analysis of the data also supported a downgrade (see Appendix C, Section C-6).
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgraded. Studies show large effects of more than 50% (see Figure C4-35), and meta-analysis of the data found an overall effect that was large in magnitude (see Appendix C, Section C-6).
• Dose-response: Upgrade. Dose response is evident in most studies, although not statistically significant in most cases. See Figure C4-36.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-34

Risk of bias heatmap of studies of DINP and fetal testosterone in rats. In HAWC: https://hawcproject.org/summary/visual/333/.

FIGURE C4-35

Data pivot of animal studies of DINP and fetal testosterone in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dinp-effect-testosterone/.

Factors Considered for Downgrading Confidence

• Risk of bias: No downgrade. See Figure C4-37.
• Unexplained inconsistencies: No downgrade. Data are relatively consistent across studies, and inconsistencies can be explained by study design or measurement features (incubation time). See Figure C4-38.
• Indirectness: No downgrade.
• Imprecision: No downgrade. Mean versus standard deviation reflects reasonable precision across studies. See Figure C4-38.
• Publication bias: No downgrade (see Appendix C, Section C-3).

Factors Considered for Upgrading Confidence

• Large magnitude: Upgrade. Consistently large effects of more than 60% are seen in several studies within the same dose ranges. See Figure C4-38.
• Dose-response: Upgrade. Dose response is evident in most studies. See Figure C4-39.
• Residual confounding: Not applicable.
• Cross-species consistency: No upgrade. Only studies in rats were available so cross-species consistency could not be evaluated.

FIGURE C4-36

Data pivot of animal studies of DINP and fetal testosterone in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dinp-effect-testosterone-dose-response/.

FIGURE C4-37

Risk of bias heatmap of studies of DPP and fetal testosterone in rats. In HAWC: https://hawcproject.org/summary/visual/334/.

FIGURE C4-38

Data pivot of animal studies of DPP and fetal testosterone in rats sorted by dose. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dpp-effect-testosterone/.

FIGURE C4-39

Data pivot of animal studies of DPP and fetal testosterone in rats sorted by study. In HAWC: https://hawcproject.org/summary/data-pivot/assessment/351/dpp-effect-testosterone-dose-response/.