Bayesian Weighted Sums: A Flexible Approach to Estimate Summed Mixture Effects

Area of Science:

• Environmental Health
• Biostatistics
• Epidemiology

Background:

• Existing methods for studying exposure mixtures have limitations.
• A novel approach is needed to estimate both the cumulative impact and individual contributions of complex environmental exposures.

Purpose of the Study:

• To introduce and validate Bayesian Weighted Sums (BWS), a new statistical method for analyzing exposure mixtures.
• To apply BWS to assess the association between polybrominated diphenyl ether (PBDE) congeners and Autism Spectrum Disorder (ASD) and Social Responsiveness Scores (SRS).

Main Methods:

• Developed the Bayesian Weighted Sums (BWS) approach to estimate summed effects and individual weights of exposure mixtures.
• Applied BWS to simulated datasets and real-world data from the Early Autism Risk Longitudinal Investigation (EARLI) study.
• Analyzed correlated exposures, specifically focusing on five PBDE congeners (28, 47, 99, 100, and 153).

Main Results:

• Simulations confirmed the reliability of the BWS method.
• In analyses of EARLI data, a 1-unit increase in the weighted sum of PBDEs was associated with increased odds of ASD (Odds Ratio: 1.41).
• A 1-unit increase in the weighted sum of PBDEs was associated with a 0.15 change in standardized SRS scores.

Conclusions:

• Bayesian Weighted Sums (BWS) offers a flexible and distinct method for estimating the effects of individual components within exposure mixtures.
• BWS allows for the specification of multiple exposure groups based on existing epidemiological or toxicological knowledge.
• This approach advances the analysis of complex environmental exposures and their health implications.