Exposome Burden Scores to Summarize Environmental Chemical Mixtures: Creating a Fair and Common Scale for Cross-study...

Environmental health researchers are increasingly concerned about characterizing exposure to environmental chemical mixtures (co-exposure to multiple chemicals simultaneously). We discuss approaches for quantifying an overall summary score or index that reflects an individual's total exposure burden to components of the mixture. We focus on unsupervised methods, in which the summary score is not computed in relation to a pre-specified health outcome.

Sum-scores and principal components analysis (PCA) are common approaches for quantifying a total exposure burden metric but have several limitations: 1) they require imputation when using exposure biomarkers with high frequency of non-detection, 2) they do not account for exposure heterogeneity, 3) sum-scores assume the same measurement error for all people, while there is no error term inherent to the PCA model as its primary purpose is dimension reduction, and 4) in pooled analyses, both approaches are limited to analyzing the set of exposure variables that are in common across all studies, potentially discarding valuable information. Meanwhile, item response theory (IRT) is a novel and promising alternative to calculate an exposure burden score that addresses the above limitations. It allows for the inclusion of exposure analytes with high frequency of non-detects without the need for imputation. It can account for exposure heterogeneity to calculate fair metrics for all people, through assessment of differential item functioning and mixture IRT. IRT also quantifies measurement errors of the exposure burden score that are individual-specific, such that it appropriately assigns a larger standard error to an individual who has missing data on one or more exposure variables. Lastly, IRT enhances cross-study harmonization by enabling the creation of exposure burden calculators to set a common scale across studies, and allows for the inclusion of all exposure variables within a chemical class, even if they were only measured in a subset of participants. Summarizing total exposure burden, through the creation of fair and informative index scores, is a promising tool for environmental health research as environmental exposures are increasingly used for biomonitoring and clinical recommendations.