Invited Perspective: How Can Studies of Chemical Mixtures and Human Health Guide Interventions and Policy?

The use of sophisticated biostatistical methods to estimate the health effects of chemical mixtures in human populations has increased dramatically since 2015, when the National Institute of Environmental Health Sciences hosted the workshop “Statistical Approaches for Studying Environmental Chemical Mixtures in Epidemiology Studies” (Taylor et al. 2016). Since then, investigators have routinely employed Bayesian kernel machine regression (Bobb et al. 2015), weighted quantile sums regression (Carrico et al. 2015), quantile g-computation (Keil et al. 2020), or other methods when estimating the potential health effects of chemical mixtures (Hamra and Buckley 2018).

Using quantile g-computation, van den Dries et al. (2021) estimated the associations of 11 phthalate metabolites, 3 phenols, and 5 organophosphate pesticide metabolites with repeated measures of fetal growth in 776 mother–child pairs from the Generation R cohort. This mixture was associated with lower fetal weight in the second trimester and lower birth weight, but not in a monotonic fashion. Interestingly, the associations for the pesticide mixture were stronger and monotonic for fetal weight in the second trimester compared with weight in the third trimester or at birth, whereas the association for the phthalate mixture was strongest for birth weight, but this association was nonmonotonic.

As more studies estimate the potential health effects of chemical mixtures, we should develop and implement scalable and sustainable interventions, regulations, and policies to reduce chemical mixture exposures at the individual and population levels. Indeed, the authors of this well-conducted study allude to this point, and there are opportunities to use data like those from this study to inform individual-level interventions and population-wide policies that reduce chemical mixture exposures in an effort to improve human health.

In fact, prior studies have demonstrated that behavioral interventions could reduce exposure to mixtures of phthalates, phenols, and pesticides. Rudel et al. (2011) found that decreasing consumption of processed and packaged food reduced both urinary bis(2-ethylhexyl) phthalate (DEHP) metabolite and bisphenol A concentrations. Another study found that switching to personal care products labeled phthalate- and phenol-free caused reductions in urinary concentrations of several phthalate metabolites, triclosan, and benzophenone-3 (Harley et al. 2016). We previously found that a residential intervention designed to reduce household dust was associated with lower urinary concentrations of several phthalates (Sears et al. 2020). Finally, the consumption of organic food can reduce exposure to organophosphate and pyrethroid pesticides (Curl et al. 2019; Lu et al. 2006).

Although behavioral interventions have the potential to reduce environmental chemical exposures, scientific and pragmatic questions remain. First, what factors shape the profiles of chemical mixtures, and how do these profiles differ across populations? For example, the patterns of personal care product use during pregnancy may differ among racial/ethnic groups, which can contribute to different mixture profiles (James-Todd et al. 2012). Because of these different chemical mixture profiles, the impact of an intervention on exposure-related health end points might vary across diverse populations, especially when chemicals have synergistic or antagonistic effects.

Second, does the effect of a chemical mixture on a given health end point vary across populations because of differences in the prevalence of modifiers or other pollutant exposures? For instance, one study reported that the association between urinary DEHP metabolite concentrations and preterm birth was greater among women who reported experiencing one or more stressful life events during pregnancy compared with those who experienced none (Ferguson et al. 2019).

Finally, prior studies were often conducted under unsustainable conditions (e.g., researchers purchasing and providing participants with organic food). Thus, efficacious and durable interventions need to be scalable to larger numbers of people and designed to be financially and logistically sustainable over long time scales.

Although behavioral interventions may reduce exposures, we should ask why individuals are responsible for taking on this burden. Is it reasonable for individuals to navigate the dizzying array of chemical exposures, emerging replacement chemicals, and methods to reduce them in the face of everyday demands? Do individuals consider the potential health risks from chemical exposures to be greater than other risks they face? One study suggested that only 55% of mothers were concerned about the health effects of chemicals in products they buy (Pell et al. 2017). Even when individuals are empowered with knowledge about the health effects of chemical mixtures and the potential sources, can we meaningfully and consistently reduce exposure through behavioral modifications? Prior intervention studies emphasized the difficulty that researchers and participants have identifying chemical-free foods owing to inadequate labeling and the pervasiveness of chemical contamination throughout the food chain (Galloway et al. 2018; Sathyanarayana et al. 2013).

Ultimately, effective policies and regulations must be developed and implemented to reduce population-level exposures to chemicals and chemical mixtures. Indeed, there is evidence suggesting that the Consumer Product Safety Improvement Act of 2008 (U.S. Congress 2008), which restricted the use of some phthalates in child care articles, may have reduced exposure to several phthalates at the population level (Zota et al. 2014). Thus, the methods used in studies like van den Dries et al. (2021) and others can help regulators prioritize chemicals and mixtures for population-level interventions based on their health effects in vulnerable populations.