The Navigation Guide—Evidence-Based Medicine Meets Environmental Health: Systematic Review of Human Evidence for PFOA Effects on Fetal Growth

Background: The Navigation Guide methodology was developed to meet the need for a robust method of systematic and transparent research synthesis in environmental health science. We conducted a case study systematic review to support proof of concept of the method. Objective: We applied the Navigation Guide systematic review methodology to determine whether developmental exposure to perfluorooctanoic acid (PFOA) affects fetal growth in humans. Methods: We applied the first 3 steps of the Navigation Guide methodology to human epidemiological data: 1) specify the study question, 2) select the evidence, and 3) rate the quality and strength of the evidence. We developed a protocol, conducted a comprehensive search of the literature, and identified relevant studies using prespecified criteria. We evaluated each study for risk of bias and conducted meta-analyses on a subset of studies. We rated quality and strength of the entire body of human evidence. Results: We identified 18 human studies that met our inclusion criteria, and 9 of these were combined through meta-analysis. Through meta-analysis, we estimated that a 1-ng/mL increase in serum or plasma PFOA was associated with a –18.9 g (95% CI: –29.8, –7.9) difference in birth weight. We concluded that the risk of bias across studies was low, and we assigned a “moderate” quality rating to the overall body of human evidence. Conclusion: On the basis of this first application of the Navigation Guide systematic review methodology, we concluded that there is “sufficient” human evidence that developmental exposure to PFOA reduces fetal growth. Citation: Johnson PI, Sutton P, Atchley DS, Koustas E, Lam J, Sen S, Robinson KA, Axelrad DA, Woodruff TJ. 2014. The Navigation Guide—evidence-based medicine meets environmental health: systematic review of human evidence for PFOA effects on fetal growth. Environ Health Perspect 122:1028–1039; http://dx.doi.org/10.1289/ehp.1307893


30
References 31 Instructions for Making Risk of Bias Determinations 33 Figure S1. Summary of data extracted from all studies of PFOA exposure with a dichotomous outcome of birth weight. 43

Bias
Rating Support for judgment Recruitment Probably low risk The study lacks description of recruitment methods such as number of individuals approached, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
High risk The authors did not adjust for gestational age or maternal age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting
Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Probably low risk Associated funds and persons appear to be from government and/or academia only and free of financial interests in study results. However, no claim denying conflicts of interest was made. • Design: Cross-sectional study • Participants: Women at Ottawa Hospital during 2005-2008. Majority were scheduled for cesarean schedule.
• Notes: Birth weight was a covariate in model of ln(PFOA )  Table S5. Characteristics of Chen et al. 2012.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Low risk Knowledge of the exposure groups was adequately prevented during the study.

Confounding
Low risk The authors accounted for both maternal age and gestational age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting
Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Cross-sectional study.
• Participants: Non-smoking mothers in Taipei between April 2004 and January 2005. The subjects were from the Taiwan Birth Panel Study.
• Notes: Study was included after contacting authors of a paper that was identified during the systematic search (Wang et al. 2011).
The two papers are written on the same cohort, but the Chen et al paper has fetal growth measurements as outcomes of the analyses (under peer review at time of search).  Fei et al. 2008.

Bias
Rating Support for judgment Recruitment Probably low risk Study lacks complete description of recruitment methods, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Low risk Knowledge of the exposure groups were adequately prevented during the study.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Probably low risk Incomplete description of the quality assurance and control aspects of the exposure measurements as outlined in the protocol, but otherwise no problems relating to exposure misclassification were detected. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting
Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
High risk 3M, a manufacturer of PFOA, provided funding for the study and conducted laboratory analysis • Design: Cross-sectional study.
• Participants: Randomly selected group of singleton births from Danish National Birth Cohort between 1996 and 2002.
• Exposure: PFOA concentrations in maternal plasma collected in the 1 st trimester (Continuous and quartile). n=1399.
• Notes: Ponderal index was also analyzed but only presented as stratified by certain covariates.  Fei et al. 2007.

Bias
Rating Support for judgment Recruitment Probably low risk Study lacks description of recruitment methods, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Low risk Knowledge of the exposure groups were adequately prevented during the study.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Probably low risk Incomplete description of the quality assurance and control aspects of the exposure measurements as outlined in the protocol, but otherwise no problems relating to exposure misclassification were detected. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting
Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
High risk 3M, a manufacturer of PFOA, provided funding for the study and conducted laboratory analysis. • Design: Cross-sectional study.
• Participants: Randomly selected group of singleton births from Danish National Birth Cohort between March 1996 and November
• Outcomes: Birth weight (continuous and dichotomous; beta statistic, difference in means, odds ratio). Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
High risk Gestational age was not measured in the study. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk The number participants reported in the paper is larger than the n provided as raw data, but the discrepancy was explained in correspondence with the author.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Probably low risk Associated funds and persons appear to be from government and/or academia only and free of financial interests in study results. However, no claim denying conflicts of interest was made. • Design: Cross-sectional study.
• Participants: Mothers in Munich, Germany participating in a birthing class between December 2007 and October 2009.
• Notes: Received raw individual data on birth weight and PFOA concentrations from study authors. Table S9. Characteristics of Halldorsson et al. 2012.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
High risk The authors did not adjust for gestational age or maternal age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Cross-sectional study.
• Participants: Singleton pregnancies in Aarhus, Denmark between April 1988 and January 1989.
• Outcomes: Waist circumference and BMI of offspring at age 20.
• Notes: Birth weight was presented by quartiles of maternal PFOA concentration in a table of maternal characteristics.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Probably low risk Because the study population was selected from subjects undergoing a screening procedure typically recommended for older women, the study could be biased towards inclusion of older women and high-risk pregnancies. However, no association between PFOA and age was found, and there was no reason to suspect that comparison groups within the study were different with respect to PFOA exposure.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Cross-sectional study.

Bias
Rating Support for judgment Recruitment Probably low risk Study lacks complete description of exclusion criteria, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Low risk Knowledge of the exposure groups were adequately prevented during the study.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk There was a discrepancy in the sample sizes for data analysis from what was reported as the total number of participants. However, the discrepancy was explained as participant drop-out and missing covariate information upon correspondence with the authors.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Probably low risk Associated funds and persons appear to be from government and/or academia only and free of financial interests in study results. However, no claim denying conflicts of interest was made. • Design: Cross-sectional study.
• Participants: Pregnant women at three hospitals in Seoul, Cheongju, and Gumi, South Korea between August 2008 and March,
• Notes: Received raw individual level data from authors. Table S12. Characteristics of Kim SK et al. 2011.

Bias
Rating Support for judgment Recruitment Probably low risk Study lacks complete description of recruitment criteria, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
High risk The authors did not adjust for gestational age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk Three participants were excluded from the birth weight analysis, but the discrepancy was accounted for in correspondence with the authors. Selective reporting Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
High risk No conflict of interest statement is provided and the study was funded by a plastic additives manufacturer. • Design: Cross-sectional study.
• Exposure: PFOA concentrations in maternal serum collected the day before delivery and cord blood (continuous). n=17.

Bias
Rating Support for judgment Recruitment Probably low risk The population is divided into two groups (menarche <11.5 yrs. and random non-cases >11.5 yrs.) from previous study and randomization is not described. Weighted statistical analysis accounted for sampling scheme. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups. Confounding Low risk The authors only adjusted for gestational age in their analysis because preliminary analysis found that maternal age did not have an effect in this population. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Probably low risk* A number of births was excluded from analysis due to missing covariate data (6% for birth weight, 20% for birth length). Categories for variables missing >10% rate were included in analyses. No reason to suspect that study groups are different with respect to missing covariates; mean values of outcomes and covariates were similar between group in analysis and total cohort. Selective reporting Low risk The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Prospective cohort study; cross-sectional exposure.
• Participants: Pregnant women from the Avon Longitudinal Study for Parents and Children, Great Britain with an expected delivery data between April, 1991 and December, 1992 selected from a previous study of maternal perfluoroalkyl compound concentrations and menarche.
• Notes: Cross-sectional exposure. Received beta estimates on untransformed PFOA scale from authors. *Because the level of missing data was higher (19-20%) for birth length and ponderal index, we assigned "probably high risk" to these outcomes (not presented in Figure 3) for this domain. This is the only instance where a risk of bias rating differed by outcome and therefore risk of bias results are not presented separately by outcome. Table S14. Characteristics of Monroy et al. 2008.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
High risk The authors did not adjust for maternal age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Low risk No other potential biases are suspected.

Conflict of interest
High risk The study was funded by the American Chemistry Council, an industry group. • Design: Cross-sectional study.
• Participants: Women that presented for an obstetrical ultrasound from January 2004 until June 2005 in Ontario Canada. The subjects are part of a larger ongoing cohort study, Family Study.
• Exposure: PFOA concentrations in maternal serum collected in the 2 nd trimester and cord serum (continuous). n=89.
• Notes: Birth weight was modeled as a predictor of PFOA levels.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment High risk Exposure was estimated using water service category and participant resident history. Incomplete outcome data Low risk The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Cross-sectional study.
• Exposure: Estimated PFOA exposure based on water service category (water provided completely, partially, or not at all by PFOA-contaminated source). n=1555.
• Outcomes: Birth weight (continuous and dichotomous; difference in means, beta statistic, odds ratio).

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably high risk The outcome data were self-reported and it was unclear, although suspected, that the participants were aware of being exposed to PFOA because they were part of a publicized study that was funded by a lawsuit.

Confounding
Probably high risk The authors restricted the analysis of birth weight to term births but did not account for gestational age in the analysis of term birth weight. Exposure assessment Probably high risk Exposure was retrospectively estimated using environmental, exposure and pharmacokinetic modeling with several assumptions, although moderate correlation between predicted vs. observed serum PFOA was reported. Modeling used selfreported residential histories and GIS linkage to public water distribution systems. Incomplete outcome data High risk Unclear why number of term and preterm births does not add up to number of live births reported; appears that a subset (n=120) of births was not accounted for in analysis of low birth weight.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Retrospective cohort study.
• Participants: Singleton pregnancies between 1990 and 2006 in the Mid-Ohio Valley. Participants are part of the much larger C8 Health Project. n=10,189 total live births.
• Exposure: PFOA exposure estimated retrospectively based on environmental, exposure and pharmacokinetic modeling in conjunction with self-reported residential histories and GIS linkage to public water distribution systems. The correlation between the predicted and observed serum PFOA (measured at time of C8 Health Project) was r=0.67. PFOA concentrations were assigned to calendar year of pregnancy for present study data analysis (continuous and quintiles of exposure used in analysis).
• Notes: Although excluded from analysis of PFOA and term low birth weight, the relationship between PFOA and pre-term birth was examined separately (findings: non-significant adjusted OR).

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment High risk Exposure was retrospectively estimated using environmental, exposure and pharmacokinetic modeling with several assumptions, although moderate correlation between predicted vs. observed serum PFOA was reported. Modeling used birth residence for 6 yr. residential history. Incomplete outcome data Low risk The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Retrospective cohort study.
• Participants: Singleton pregnancies between 1990 and 2004 in the Mid-Ohio Valley. Participants are part of the much larger C8 Health Project. Case-control sampling identified cases of term low birth weight (n=918) and a 10% sample of term >/= 2500 grams controls (n=3616).
• Exposure: PFOA exposure estimated retrospectively based on environmental, exposure and pharmacokinetic modeling in conjunction with birth records and geo identifiers of birth residence assumed for last 6 yrs. Geocodes determined public vs. private well water source; multiple imputations used for zip code only births (34%) to generate new exposure estimate. (Continuous and quintiles of exposure used in analysis). n=4,534.
• Outcomes: Birth weight (continuous and dichotomous; beta statistic, odds ratio), small for gestational age.
• Notes: Although excluded from analysis of PFOA and term low birth weight, the relationship between PFOA and pre-term birth was examined separately (findings: non-significant adjusted OR).

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Probably high risk Exposure was retrospectively estimated using environmental, exposure and pharmacokinetic modeling with several assumptions, although moderate correlation between predicted vs. observed serum PFOA was reported. Modeling used lifetime residential history and exposure estimates were calibrated with serum levels measured after the study period. Incomplete outcome data

Low risk
The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Retrospective cohort study. concentrations to update estimates and assumed that "higher than expected" serum concentration reflects lifetime (continuous and quintiles of exposure used in analysis). n=4,142.
• Outcomes: Birth weight (continuous and dichotomous; beta statistic, odds ratio), small for gestational age.
• Notes: Although excluded from analysis of PFOA and term low birth weight, the relationship between PFOA and pre-term birth was examined separately (findings: some evidence of elevated adjusted odds). Table S19. Characteristics of Stein et al. 2009.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding High risk The outcome data were self-reported and the participants were aware of being exposed to PFOA because they were part of a publicized study that was funded by a lawsuit.

Confounding
Probably high risk The authors restricted the analysis of birth weight to term births but did not account for gestational age in the analysis of term birth weight. Exposure assessment Probably high risk Serum levels measured up to 5 years after birth were used to estimate exposure. The description of laboratory analysis also lacks detail. Incomplete outcome data Probably low risk One participant was missing from analysis without explanation, but not likely to affect results.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Cross-sectional study; retrospective account of outcome data.
• Participants: Pregnancies among female C8 Health Project participants between 2000 and 2006 with serum PFC measurements.
• Exposure: PFOA concentrations in maternal serum collected up to five years after birth (continuous and percentiles). n=1589 live births.
• Notes: Although excluded from analysis of PFOA and term low birth weight, the relationship between PFOA and pre-term birth was examined separately (findings: non-significant adjusted OR). Table S20. Characteristics of Washino et al. 2009.

Bias
Rating Support for judgment Recruitment Low risk The strategy for recruiting participants was consistent across study groups. Blinding Probably low risk The authors did not discuss blinding but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Low risk The exposure assessment methods were robust and included a detailed description of QA/QC. Incomplete outcome data Low risk The study did not have incomplete outcome data.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported.

Other bias
Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Prospective cohort study; cross-sectional exposure.
• Participants: Pregnant women who delivered at Sapporo Toho Hospital, Japan between July 2002 and October 2005.
• Notes: Received raw individual data from authors. The study lacks description of randomization and the study population is divided into two categories (random, subfecund) from a previous study, but otherwise no reason to suspect there were substantial differences between comparison groups other than PFOA exposure. Blinding Probably low risk The authors described blinding only for the QA/QC but the study design prevents knowledge of exposure groups.

Confounding
Low risk The authors adjusted for both maternal age and gestational age in their analysis. Exposure assessment Probably low risk Incomplete description of the quality assurance and control aspects of the exposure measurements as outlined in the protocol, but otherwise no problems relating to exposure misclassification were detected. Incomplete outcome data Low risk Incomplete outcome data were adequately addressed.

Selective reporting Low risk
The study is free of suggestion of selective outcome reporting. All of the study's specified outcomes were adequately reported. Other bias Low risk No other potential biases are suspected.

Conflict of interest
Low risk The authors report no conflict of interest, and associated funds and persons appear to be from government and/or academia only. • Design: Nested case-control study.
• Participants: Women enrolled in Norwegian Mother and Child Cohort Study between 2003 and 2004 selected for a prior casebased study of PFCs and subfecundity.
• Exposure: PFOA concentrations in maternal serum collected at the time of enrollment (continuous and quartile). n=849.
• Outcomes: Birth weight (continuous data; beta statistic), small for gestational age, large for gestational age.
• Notes: Although excluded from analysis of PFOA and term low birth weight, the relationship between PFOA and pre-term birth was examined separately (findings: non-significant adjusted OR). Table S22. Meta-analysis for association between PFOA exposure and birth length (cm).   Table S24. Meta-analysis for association between PFOA exposure and head circumference (cm). Apelberg et al. 2007 293 -0.21 (-0.41, -0

Criteria for a judgment of 'YES' (i.e. low risk of bias):
Protocols for recruitment and inclusion/exclusion criteria were applied similarly across study groups, and any one of the following: • Study participants were recruited from the same population at the same time frame; or • Study participants were not all recruited from the same population, but proportions of participants from each population in each study group are uniform

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information about participant selection to permit a judgment of 'YES', but there is indirect evidence that suggests that participant recruitment and inclusion/exclusion criteria was consistent, as described by the criteria for a judgment of 'YES'.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
Any one of the following: • Protocols for recruitment or inclusion/exclusion criteria were applied differently across study groups; or • Study participants were recruited at different time frames; or • Study participants were recruited from different populations and proportions of participants from each population in each study group are not uniform

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information about participant selection to permit a judgment of 'NO', but there is indirect evidence that suggests that participant recruitment or inclusion/exclusion criteria was inconsistent, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that participant selection is not an element of study design capable of introducing risk of bias in the study.

Criteria for a judgment of 'YES' (i.e. low risk of bias):
Any one of the following: • No blinding, but the review authors judge that the outcome and the outcome measurement are not likely to be influenced by lack of blinding; or • Blinding of key study personnel ensured, and unlikely that the blinding could have been broken; or • Some key study personnel were not blinded, but outcome assessment was blinded and the non-blinding of others unlikely to introduce bias.

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information about blinding to permit a judgment of 'YES', but there is indirect evidence that suggests the study was adequately blinded, as described by the criteria for a judgment of 'YES'.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
Any one of the following: • No blinding or incomplete blinding, and the outcome or outcome measurement is likely to be influenced by lack of blinding; or • Blinding of key study personnel attempted, but likely that the blinding could have been broken; or • Some key study personnel were not blinded, and the non-blinding of others likely to introduce bias.

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information about blinding to permit a judgment of 'NO', but there is indirect evidence that suggests the study was not adequately blinded, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that blinding is not an element of study design capable of introducing risk of bias in the study.

Criteria for a judgment of 'YES' (i.e. low risk of bias):
The reviewers judge that there is low risk of exposure misclassification and any one of the following: • There is high confidence in the accuracy of the exposure assessment methods; or • Less-established or less direct exposure measurements are validated against wellestablished or direct methods AND if applicable, appropriate QA/QC for methods are described and are satisfactory, with at least three of the following items reported, or at least two of the following items reported plus evidence of satisfactory performance in a high quality inter-laboratory comparison: Limit of detection or quantification; standards recovery; measure of repeatability; investigation and prevention of blanks contamination.

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information about the exposure assessment methods to permit a judgment of 'YES', but there is indirect evidence that suggests that methods were robust, as described by the criteria for a judgment of 'YES'. Studies only reporting that the QA/QC items above were satisfactory but not reporting all of the actual numbers may receive a judgment of "probably yes."

Criteria for the judgment of 'NO' (i.e. high risk of bias):
The reviewers judge that there is high risk of exposure misclassification and any one of the following: • There is low confidence in the accuracy of the exposure assessment methods; or • Less-established or less direct exposure measurements are not validated and are suspected to introduce bias that impacts the outcome assessment (example: participants are asked to report exposure status retrospectively, subject to recall bias) • Uncertain how exposure information was obtained

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information about the exposure assessment methods to permit a judgment of 'NO', but there is indirect evidence that suggests that methods were not robust, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that exposure assessment is not an element of study design capable of introducing risk of bias in the study.

Criteria for a judgment of 'YES' (i.e. low risk of bias):
The study accounted for (i.e., matched, stratified, multivariate analysis or otherwise statistically controlled for) important potential confounders, or reported that potential confounders were evaluated and omitted because inclusion did not substantially affect the results. The determination of specific confounders may be informed by the data, including the studies included in the review.

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
The study accounted for most but not all of the important potential confounders AND this lack of accounting is not expected to introduce substantial bias.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
The study did not account for or evaluate important potential confounders.

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
The study accounted for some but not all of the important potential confounders AND this lack of accounting may have introduced substantial bias.

Criteria for a judgment of 'YES' (i.e. low risk of bias):
Participants were followed long enough to obtain outcome measurements and any one of the following: • No missing outcome data; or • Reasons for missing outcome data unlikely to be related to true outcome (for survival data, censoring unlikely to be introducing bias); or

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information about incomplete outcome data to permit a judgment of 'YES', but there is indirect evidence that suggests incomplete outcome data was adequately addressed, as described by the criteria for a judgment of 'YES'.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
Participants were not followed long enough to obtain outcome measurements OR any one of the following: • Reason for missing outcome data likely to be related to true outcome, with either imbalance in numbers or reasons for missing data across exposure groups; or • For dichotomous outcome data, the proportion of missing outcomes compared with observed event risk enough to induce biologically relevant bias in intervention effect estimate; or • For continuous outcome data, plausible effect size (difference in means or standardized difference in means) among missing outcomes enough to induce biologically relevant bias in observed effect size; or • Potentially inappropriate application of imputation.

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information about incomplete outcome data to permit a judgment of 'NO', but there is indirect evidence that suggests incomplete outcome data was not adequately addressed, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that incomplete outcome data is not capable of introducing risk of bias in the study.
6. Are reports of the study free of suggestion of selective outcome reporting?

Criteria for a judgment of 'YES' (i.e. low risk of bias):
All of the study's pre-specified (primary and secondary) outcomes outlined in the protocol, methods, abstract, and/or introduction that are of interest in the review have been reported in the pre-specified way.

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information about selective outcome reporting to permit a judgment of 'YES', but there is indirect evidence that suggests the study was free of selective reporting, as described by the criteria for a judgment of 'YES'.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
Any one of the following: • Not all of the study's pre-specified primary outcomes (as outlined in the protocol, methods, abstract, and/or introduction) have been reported; or • One or more primary outcomes is reported using measurements, analysis methods or subsets of the data (e.g. subscales) that were not pre-specified; or • One or more reported primary outcomes were not pre-specified (unless clear justification for their reporting is provided, such as an unexpected effect); or One or more outcomes of interest are reported incompletely

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information about selective outcome reporting to permit a judgment of 'NO', but there is indirect evidence that suggests the study was not free of selective reporting, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that selective outcome reporting is not capable of introducing risk of bias in the study.

Criteria for a judgment of 'YES' (i.e. low risk of bias):
The study appears to be free of other sources of bias.

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information to permit a judgment of 'YES', but there is indirect evidence that suggests the study was free of other threats to validity.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
There is at least one important risk of bias. For example, the study: • Had a potential source of bias related to the specific study design used; or • Stopped early due to some data-dependent process (including a formal-stopping rule); or • Had extreme imbalance of characteristics among exposure groups; or • Had differential surveillance for outcome between exposure groups or between exposed/unexposed groups • The conduct of the study is affected by interim results (e.g. recruiting additional participants from a subgroup showing greater or lesser effect); or • An insensitive instrument is used to measure outcomes (which can lead to underestimation of both beneficial and harmful effects); or • Selective reporting of subgroups; or • Has been claimed to have been fraudulent; or • Had some other problem

Criteria for the judgment of 'PROBABLY NO' (i.e. probably high risk of bias):
There is insufficient information to permit a judgment of 'NO', but there is indirect evidence that suggests the study was not free of other threats to validity, as described by the criteria for a judgment of 'NO'.

Criteria for the judgment of 'NOT APPLICABLE' (risk of bias domain is not applicable to study):
There is evidence that other potential threats to validity are not capable of introducing risk of bias in the study.
8. Was the study free of support from a company, study author, or other entity having a financial interest in any of the exposures studied?

Criteria for a judgment of 'YES' (i.e. low risk of bias):
The study did not receive support from a company, study author, or other entity having a financial interest in the outcome of the study. Examples include the following: • Funding source is limited to government, non-profit organizations, or academic grants funded by government, foundations and/or non-profit organizations; • Chemicals or other treatment used in study were purchased from a supplier; • Company affiliated staff are not mentioned in the acknowledgements section; • Authors were not employees of a company with a financial interest in the outcome of the study; • Company with a financial interest in the outcome of the study was not involved in the design, conduct, analysis, or reporting of the study and authors had complete access to the data; • Study authors make a claim denying conflicts of interest; • Study authors are unaffiliated with companies with financial interest, and there is no reason to believe a conflict of interest exists; • All study authors are affiliated with a government agency (are prohibited from involvement in projects for which there is a conflict of interest or an appearance of conflict of interest).

Criteria for the judgment of 'PROBABLY YES' (i.e. probably low risk of bias):
There is insufficient information to permit a judgment of 'YES', but there is indirect evidence that suggests the study was free of support from a company, study author, or other entity having a financial interest in the outcome of the study, as described by the criteria for a judgment of 'YES'.

Criteria for the judgment of 'NO' (i.e. high risk of bias):
The study received support from a company, study author, or other entity having a financial interest in the outcome of the study. Examples of support include: • Research funds; • Chemicals provided at no cost; • Writing services; • Author/staff from study was employee or otherwise affiliated with company with financial interest; • Company limited author access to the data; • Company was involved in the design, conduct, analysis, or reporting of the study; • Study authors claim a conflict of interest

Figure legend
Figure S1. Summary of data extracted from all studies of PFOA exposure with a dichotomous outcome of birth weight. Low birth weight is defined as less than 2500 grams. The PFOA increase is the exposure contrast being compared in each study. Square markers represent data where there was an exposure gradient that can be evaluated when considering dose-response in upgrading the quality of the evidence (Table 6). Savitz et al. 2012b paper presented additional alternative estimates not included in Figure S1 due to space limitations. Nolan et al. 2009, Savitz et al. 2012a, Savitz et al. 2012band Stein et al. 2009 were conducted in the same geographical area and participants may overlap.