Characterizing variability in personal chemical exposure to improve exposomics

Understanding how chemical exposure varies within and between people over time is a critical component of characterizing the exposome-the totality of lifetime exposures. However, variability remains an understudied aspect of exposomic research.

Our objective was to investigate trends in variability for chemical exposure data between and within people that appear across differing study designs.

Thirty-five people in Eugene, Oregon, and 46 people in St. Helens, Oregon, wore silicone wristbands over multiple seasons, including a span of heavy wildfire smoke in Eugene. Each participant wore between four and 14 wristbands. We analyzed 586 wristbands for 94 (Eugene) and 58 (St. Helens) chemicals. While analytic tests differed between the studies, the same 43 polycyclic aromatic hydrocarbons (PAHs) were measured in both studies. We also evaluated three environmental variables for their impact on chemical concentrations. We fit generalized mixed effects models to each chemical, and used variance partitioning to understand and quantify sources of variability across environmental factors and inter- and intra-individual variables.

We observed PAHs that were consistent within people across different days. For a subset of these PAHs, results did not agree well between studies, indicating the importance of measuring chemical data at different time points across studies. Environmental variables were not sufficient for explaining data variability for most chemicals. Only 21% and 30% of the modeled chemicals for Eugene and St. Helens, respectively, had a combined environmental variable R² at or above 0.1. Yet, environmental factors still revealed valuable information; we observed higher combined R² values for styrene, o-xylene, ethylbenzene, and phenanthrene in the Eugene detection model, which came from a combination of fine particulate matter and smoke density information.

Our manuscript is the largest investigation of intra- and inter- variability in silicone wristband concentrations, containing over 23,000 chemical data points across two different personal chemical exposure studies. Certain chemicals were consistent within people across different days. For a subset of chemicals, results did not agree well between the two wristband studies. Our findings highlight the importance of measuring chemical data at different time points across studies to better understand the exposome. Environmental variables included in this study were not sufficient for explaining the data variability for most chemicals.