The Use and Misuse of Historical Controls in Regulatory Toxicology: Lessons from the CLARITY-BPA Study.

For many endocrine-disrupting chemicals (EDCs) including Bisphenol A (BPA), animal studies show that environmentally relevant exposures cause harm; human studies are consistent with these findings. Yet, regulatory agencies charged with protecting public health continue to conclude that human exposures to these EDCs pose no risk. One reason for the disconnect between the scientific consensus on EDCs in the endocrinology community and the failure to act in the regulatory community is the dependence of the latter on so-called "guideline studies" to evaluate hazards, and the inability to incorporate independent scientific studies in risk assessment. The Consortium Linking Academic and Regulatory Insights on Toxicity (CLARITY) study was intended to bridge this gap, combining a "guideline" study with independent hypothesis-driven studies designed to be more appropriate to evaluate EDCs. Here we examined an aspect of "guideline" studies, the use of so-called "historical controls," which are essentially control data borrowed from prior studies to aid in the interpretation of current findings. The US Food and Drug Administration authors used historical controls to question the plausibility of statistically significant BPA-related effects in the CLARITY study. We examined the use of historical controls on 5 outcomes in the CLARITY "guideline" study: mammary neoplasms, pituitary neoplasms, kidney nephropathy, prostate inflammation and adenomas, and body weight. Using US Food and Drug Administration-proposed historical control data, our evaluation revealed that endpoints used in "guideline" studies are not as reproducible as previously held. Combined with other data comparing the effects of ethinyl estradiol in 2 "guideline" studies including CLARITY-BPA, we conclude that near-exclusive reliance on "guideline" studies can result in scientifically invalid conclusions. © Endocrine Society 2019. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.