A New View of the Things We Use: Using Purchasing Data to Predict Common Mixture Exposures

Abstract

Humans are exposed to a multitude of chemicals,1 typically in mixtures. Testing substances individually using traditional methods can be time consuming and cost prohibitive;2 the number of possible chemical combinations makes this type of testing unrealistic for mixtures.3 To circumvent these challenges, scientists at the U.S. Environmental Protection Agency (EPA) used a data-driven approach to identify and prioritize relevant chemical combinations, as reported recently in Environmental Health Perspectives.3

By identifying patterns in purchasing data, investigators can estimate common ingredient combinations that consumers are exposed to in the products they use regularly. Knowledge of common co-exposures could help direct experimental toxicology assessments. Image: © iStock/Prostock-Studio.

Other efforts to identify high-priority co-exposures have not captured the full spectrum of chemical combinations because of either the lack of ingredient information or insufficient purchase and use data.2^,4^,5 “This study is unique because of the way we were able to integrate multiple data sets to improve our overall understanding of human exposure,” says Zachary Stanfield, a postdoctoral researcher at the U.S. EPA and first author of the study.

Stanfield and colleagues matched consumer purchasing data from a marketing database compiled in 2012 with ingredient data obtained from the U.S. EPA’s Chemical and Product Database.6 The combined data streams included approximately 2.4 million purchases by about 53,000 households of 31,000 products. Using frequent item set mining,7 a well-established method for identifying patterns in behavior, the group sifted through the extremely high number of theoretical chemical combinations to identify the truly relevant ones.

“Our findings show that, from a risk standpoint, we don’t need to concern ourselves with all possible chemical combinations,” notes Kristin Isaacs, senior author of the study. “This ‘top-down’ approach can supplement others that ‘look under the lamppost’ at specific chemicals.”

“This is an important start,” says Julia Brody, executive director of the Silent Spring Institute. “It creates a resource so we can evaluate the combined effects on health [of specific chemicals].” She notes that products’ ingredient data were incomplete and calls for additional ingredient disclosure. “In the U.S., manufacturers can use ingredients without testing first for effects on health, and they don’t have to list all the ingredients on product labels, so we’re always playing catch-up to understand how [products] affect people’s health,” she explains. Brody was not involved in the study.

The study revealed exposure variations across demographic groups. Focused analysis comparing predominant purchases and demographic patterns showed several distinct patterns. For example, chemical mixtures most frequently encountered by households with children, households headed by women of color, and lower-income households diverged from those encountered by the rest of the study households.

“This suggests a need for further study of the product chemical combinations, such as from hair products used by Black women, that may be contributing to health inequalities,” says Brody. More in-depth studies should look at the factors underlying differences in purchasing behavior within households and communities—for example, how choice, brand, or product availability influences chemical mixture exposures. Brody says information like this can better inform public health policy aimed at eliminating racial inequities in environmental health.

Biography

Florencia Pascual, PhD, is a Durham, NC–based freelance science writer.