a. Three studies are not displayed on the figure because they used a different effect estimate type (Yang et al., 2016, Liu et al., 2018) or only presented quantitative results for significant associations (Berg et al., 2016). None reported an association between PFDA exposure and T4.

b. Studies are sorted by median exposure level.

c. Regression coefficients are expressed as the increase in T4 levels per 1 ng/mL unit increase in exposure unless otherwise noted. Units for the outcome varied across studies.

d. If a study presented results for both free T4 and total T4 and the effect estimates were similar, both were extracted, but only free T4 is presented on the figure.

e. Effect estimates for Preston et al., 2018 and Lewis et al., 2015 were recalculated from percent difference to regression coefficient for consistency with the other studies.