Methods for Modeling Exposures and Health Risks from Combined Chemical and Non-Chemical Stressors

摘要

Background: Although there is increasing interest in cumulative risk assessment and related approaches to characterize health impacts from multiple stressors, methodological challenges remain. Published epidemiological studies rarely report associations in a manner applicable for cumulative risk assessment, and multi-stressor exposure data are often lacking. We provide two examples that illustrate the strengths and weaknesses of novel methods for characterizing multi-stressor exposures and associated health risks. Methods: In one case, we utilized biomarker and questionnaire data from a prospective birth cohort study to develop exposure regression models that can be used to estimate exposures in a population-based surveillance database. We used generalized additive models (GAMs) to investigate the association between the predicted exposures, other stressors, and teen pregnancy. In the second case, we used sociodemographic Census information, temperature and air pollution predictions from satellite-based models, and other GIS data to characterize exposures for all Massachusetts births, and we applied elastic nets and Bayesian kernel machine regression (BKMR) to select among candidate variables and determine multi-stressor associations with birth weight. Results: Regression models explained exposure variability as a function of sociodemographics, birth address, and birth year. GAMs yielded valuable insights regarding non-linear interactions among stressors that would not have been identified through conventional approaches. While elastic net and BKMR models yielded different subsets of predictors, constrained model-building strategies helped identify the combinations of stressors associated with birth weight. Conclusions: Exposure modeling that leverages population-specific measurements and extensive GIS data, coupled with epidemiological approaches selected with cumulative risk assessments in mind, provide value insight on the health effects of combined exposures.