Citation:

Kapraun, Dustin F., John F. Wambaugh, C. Ring, R. Tornero-Velez, AND R. Woodrow Setzer. A Method for Identifying Prevalent Chemical Combinations in the US Population. ENVIRONMENTAL HEALTH PERSPECTIVES. National Institute of Environmental Health Sciences (NIEHS), Research Triangle Park, NC, 125(8):1-16, (2017).

Impact/Purpose:

• Agency Problem: Congressional acts, including the Federal Insecticide, Rodenticide, and Fungicide Act, the Safe Drinking Water, the Clean Air Act, the Food Quality Protection Act, and the Toxic Substances Control Act, direct the Agency to ensure the safety of chemicals. With approximately 84,000 chemical substances in commercial use, humans are exposed to thousands of chemicals through the air, water, food, and products. Although humans are exposed to chemical mixtures, human health risk assessments have traditionally focused on single chemicals, limiting the Agency’s ability to characterize risk to real world exposures. Therefore, the National Research Council has recommended using toxicity data on mixtures of concern. Identifying chemicals mixtures to which humans are likely exposed is the first step to meeting the Administrator’s goal of ensuring chemical safety and preventing pollution. • Approach: A cross-Lab/Center partnership of ORD scientists identified relevant chemical combinations using a novel approach. The ORD scientists analyzed available biomonitoring data using an analysis technique called frequent item set mining. This method was originally used to analyzed consumer purchasing behavior, identifying which items are frequently purchased together (e.g., toothpaste and floss are purchased together more frequently than toothpaste and frozen peas). By adapting and applying frequent item set mining to chemical biomonitoring data, the scientists identified chemical combinations that frequently co-occur in the United States. The most prevalent combinations represent chemical mixtures that should be prioritized for human health risk evaluation. • Results: Starting with nearly a decillion (1 x 1033; for reference, a trillion is 1 x 1012) potential mixtures, ORD scientists identified 90 prevalent chemical mixtures in the US population. Chemicals identified in prevalent mixtures included pesticides, endocrine disrupting compounds, and perfluorinated compounds. These 90 chemical mixtures were present >30% of the entire US population, however, certain chemical combinations were observed more frequently in specific demographic groups. For instance, the scientists identified chemical mixtures that were >90% prevalent in children. These analyses identified a suite of chemical mixtures for testing, as well as highlighting potential lifestage and populations differences when evaluating the effects of chemical exposures. • Impact to the Agency: The identification of chemical mixtures has two immediate impacts to the Agency. First, this work developed a novel methodology to identify chemical mixtures from biomonitoring data. As exposure characterization data increase, frequent item set mining will become increasing important to identify chemical mixtures relevant for human health risk assessments. Secondly, this research successfully overcame a long-standing difficulty in chemical mixtures research: the overwhelming large number of potential mixtures. Even with high throughput approaches, testing all potential chemical mixtures is impossible. By identifying the most relevant subset of chemical mixtures, this work enables focused and impactful research ensuring chemical safety to the US population.

Description:

Through the food and water they ingest, the air they breathe, and the consumer products with which they interact at home and at work, humans are exposed to tens of thousands of chemicals, many of which have not been evaluated to determine their potential toxicities. In recent years, high-throughput screening (HTS) methods have been developed that begin to address the need for more efficient toxicity assessment, but testing still tends to focus on individual chemicals. Meanwhile, the exposures that people actually experience involve mixtures of chemicals. Unfortunately, the number of mixtures that can be formed from the thousands of environmental chemicals is huge, and testing all of them would be impossible, even using HTS techniques. We therefore need tools for identifying those mixtures that are most relevant. We propose that frequent itemset mining (FIM), a technique used for finding patterns in consumer purchasing behavior, can be applied to data from large-scale biomonitoring studies to identify combinations of chemicals that frequently co-occur in people. As a proof of concept, we applied FIM to biomonitoring data from the National Health and Nutrition Examination Survey. In this way, we identified 90 chemical combinations consisting of relatively few chemicals that occur in at least 30% of the US population, as well as 3 super-combinations consisting of relatively many chemicals that occur in a small but non-negligible proportion of the US population. Thus, we have demonstrated a technique for narrowing a large number of possible chemical combinations down to a much smaller collection of prevalent chemical combinations.

URLs/Downloads:

Record Details: