Publications : 2026

East A, Klaren W, Wikoff D, Patlewicz G. Identifying potential co-exposures with food additives using data mining and machine learning approaches across EPA’s Consumer Product Database (CPDat). Abstract M-36, Poster Session, International Society of Exposure Science (ISES) 2026 Annual Meeting, Vancouver, BC, Canada, October 5, 2026.

Abstract

Recent U.S. Food and Drug Administration (FDA) and Environmental Protection Agency (EPA) guidance have identified cumulative exposure from multiple pathways as a key focus in the evolution of risk assessment frameworks. As an example, there is growing interest in better characterizing the universe of cumulative exposure scenarios associated with multi-constituent food products, food additives, contaminants, and ultra-processed foods (UPFs). As such, there is an increased demand for tools and methodologies that can readily characterize the totality of exposures across the range of possible exposure scenarios. The objective of this work is to evaluate the utility of a recently updated EPA product database, CPDat 4.0, leveraging Product Use Categories, Functional Use Categories, and Keywords to identify cumulative exposures involving food additives and contaminants. Results indicate that 3,170 chemicals are tagged as food additives in CPDat. Preliminary modeling suggests that pesticides, fragrances, and personal care products most commonly co-occur with food additives. Despite this overlap,  only 51.3% of chemicals tagged both fragrance and food additive had chemical composition data, with only presence data available for the remaining 48.7%. Similarly, 32.5% of chemicals tagged as both food additives and consumer product ingredients had no composition data available. This analysis highlights data paucity as a persistent challenge in the quantification of cumulative and aggregate exposures, while identifying common co-exposure classes for further investigation.  By understanding the universe of chemicals and co-exposures associated with food additives, the findings of this effort provide a foothold in advancing cumulative exposomics evaluations using real world-data.