Selection of the critical effect size alters hazard characterization - a retrospective analysis of key studies used for risk assessments of PFAS.

Abstract

Regulatory values for per- and polyfluoroalkyl substances (PFAS) vary widely across agencies, creating inconsistencies that challenge effective risk management and public health communication. These differences often stem from methodological choices in determining points of departure (PoDs), the selection of critical effect size (CES) and the modeling framework for benchmark dose (BMD) analysis. This study investigates the impact of CES selection on hazard characterization by analyzing how variations in CES influence resulting PoDs and health-based guidance values. A retrospective analysis of key studies from four regulatory PFAS risk assessments was conducted, covering both animal and epidemiological data (thyroid hormone, cholesterol, and vaccine response). CES options compared included 5%, 10%, one standard deviation from background, and a generalized effect size theory, using both frequentist and Bayesian statistics. The findings show that CES selection and statistical approach substantially affect BMD estimates such as the lower bound BMD (BMDL) of the respective confidence interval or credible interval; with larger CES values and Bayesian modeling yielding more biologically relevant, stable results. For instance, Bayesian methods provided narrower credible intervals, compared to frequentist methods at lower CES levels, minimizing overly conservative assessments. However, in comparison to the PoD previously derived by the European Food Safety Authority the results generally suggest lower values. In conclusion, this study supports the use of a flexible, endpoint-specific CES with Bayesian model averaging, which may enhance the accuracy and consistency of PFAS guidance values, offering a more robust foundation for regulatory risk assessments.