Relevance and challenges of exposome studies for environmental health research.

The exposome encompasses all environmental exposures humans undergo from conception. We frame exposome studies according to four broad aims essential for environmental health research. First, a descriptive aim, consisting in assessing exposure patterns, including correlations between exposures and their within-subject variability. This descriptive aim includes "environmental justice" studies of associations between the exposome and sociodemographic factors. A second - etiologic - aim is to describe the subclinical and clinical effects of the exposome (hazard identification). A third aim is the quantification of the exposome health impact (e.g., in life years lost at the population level) and the ranking of exposures in terms of population disease burden (exposome disease burden or risk assessment). A fourth (intervention) aim corresponds to the identification of approaches to modify the exposome, as a way to improve health. With the large increase in the number of chemicals quantifiable in a small volume of a biospecimen, the main challenge of descriptive exposome studies is population representativeness. Regarding the etiologic aim, by simultaneously showing associations of a given biological parameter with hundreds of exposures, exposome studies effectively limit publication bias and selective reporting of results, which are a strong concern in single-exposure studies. They nonetheless face several challenges, related to the curse of dimensionality, the correlation between exposures, the breadth vs. depth tension… Increasing the number of exposures considered implies to simultaneously increase study population sizes, aiming for cohorts of 100,000 subjects or more. The accuracy of exposure assessment should simultaneously be improved, e.g., by collecting repeated biospecimens within each subject, assessing exposures at various ages and decreasing limits of quantification. Classical exposome statistical designs such as ExWAS (exposome-wide association studies) are subject to a high false positive rate. Models adapted to sparse data allowing to control for confounding by co-exposures appear more efficient. The results from exposome descriptive and etiologic studies can be combined to feed exposome disease burden assessments. These can in turn help prioritize exposures for which efficient interventions need to be identified. The approaches outlined in this work could help exposome research contribute more strongly to environmental health research and to the associated risk management decisions.