Abstract

Quantifying the impact of exposure to environmental chemical mixtures is crucial for identifying risk factors for diseases and developing more targeted public health interventions. A core challenge in environmental health studies lies in understanding how and why exposure mixtures affect human health and well-being. Additionally, hurdles such as non-additive exposure-outcome relationships, high-dimensional data, higher-order interactions between exposures, limits of detection, and missing values further complicate the analysis.

In this project, we propose the development of a novel multi-pollutant statistical model incorporating time-varying covariates to evaluate the association of higher-level interaction effects of environmental chemical exposures with cancer incidence. Our approach integrates a multi-pollutant model with time-varying covariates, enabling us to leverage cumulative exposure assessments of pesticides over various time periods alongside cancer incidence data. Unlike conventional methods that often focus solely on baseline exposures, our methodology incorporates repeated measures of exposure.

We employ a Cox regression model tailored for time-varying covariates to assess cancer risk comprehensively. Furthermore, we develop an MCMC algorithm using a shrinkage prior that incorporates the strong heredity principle between the main effects and interaction effects. The performance of our methodology is examined through simulation studies and a real data application in cancer epidemiology.