Using Environmental Risk Score to study a mixture of phenols and parabens in association with gestational age in a Puerto Rican cohort

Abstract

Background: Most studies of associations between exposure to environmental chemicals and adverse health outcomes have focused on individual chemicals in isolation. However, in reality, people are exposed to multiple chemicals simultaneously, and assessing chemical mixtures may provide more realistic risk estimates. Various phenols, parabens and triclocarban have been associated with gestational age. We explored the association of phenols and parabens as a mixture with gestational age using the Environmental Risk Score (ERS), a weighted sum of chemical concentrations. Methods: We examined 11 combined phenol, paraben and triclocarban urinary biomarkers in relation to gestational age in the Puerto Rico Testsite for Exploring Contamination Threats (PROTECT) cohort (N=515). We randomly split the dataset into training and validation sets and used elastic net regression coefficients for predicting gestational age from the training set to obtain coefficients for the ERS calculation. Individual biomarkers, biomarker-biomarker interactions and squared biomarker measures were included in the joint elastic net model. ERS was examined in association with gestational age in the validation set, while controlling for specific gravity, maternal age and socioeconomic status. Results: The biomarkers driving the ERS scores were 2,4-dichlorophenol, 2,5-dichlorophenol, benzophenone-3, triclosan, triclocarban, ethyl-, methyl- and propyl- paraben. There was a statistically significant association between ERS and gestational age, such that an increase in an ERS standard deviation was associated with a 0.45 day increase in gestational age (95% CI: 0.29, 0.62). This increase in gestational age was consistent with findings from individual chemical biomarker analyses from the same cohort. Conclusions: This study shows the utilization of ERS powered by modern machine learning tools to study multi-pollutants, while allowing for chemical interactions and correlations.