Cross-Cohort Mixture Analysis: A Data Integration Approach With Applications on Gestational Age and...

Area of Science:

• Environmental epidemiology
• Statistical modeling
• Public health research

Background:

• Integrating multiple studies enhances statistical power for environmental exposure mixture research.
• Existing methods often overlook crucial differences across study populations.
• Analyzing complex environmental exposures requires advanced statistical techniques.

Purpose of the Study:

• To extend the Bayesian Weighted Quantile Sum (BWQS) regression into a hierarchical framework (HBWQS) for analyzing environmental exposure mixtures across multiple cohorts.
• To identify the most harmful co-occurring environmental exposures and their associations with health outcomes across different populations.
• To account for inter-cohort variability in mixture analyses.

Main Methods:

• Developed and applied the hierarchical Bayesian Weighted Quantile Sum (HBWQS) regression model.
• Utilized 10 simulated scenarios with varying mixture components and populations.
• Applied the model to real-world data on prenatal metal mixture exposure (arsenic, cadmium, lead) and gestational age metrics.

Main Results:

• Simulated scenarios demonstrated good empirical coverage and minimal bias for HBWQS-estimated parameters.
• HBWQS regression showed better average performance than standard BWQS using the Watanabe-Akaike information criterion.
• Analysis of Environmental influences on Child Health Outcomes (ECHO) program data indicated a negative association between an environmental mixture and gestational age at one site.

Conclusions:

• The HBWQS approach effectively combines multiple cohorts while accounting for individual cohort differences in mixture analyses.
• This method enhances the identification of harmful environmental exposures and their health impacts.
• HBWQS findings can inform regulations, policies, and interventions for managing co-occurring environmental exposures and maximize data utilization.