Gestational exposures to mixtures of multiple chemical classes and autism spectrum disorder in the MARBLES study.

BACKGROUND

Previous epidemiologic studies on gestational chemical exposures and autism spectrum disorder (ASD) often lack analysis of chemical mixtures or are limited to investigating certain chemical classes.

OBJECTIVE

We examined the impact of multi-class chemical mixtures on ASD risk, using data from the MARBLES (Markers of Autism Risks in Babies-Learning Early Signs) cohort.

METHODS

Children were clinically assessed at age 3 and classified as ASD, typical development (TD), or non-TD with other neurodevelopmental concerns. In blood or urine from 105 pregnant mothers, we quantified 42 biomarkers across 5 chemical classes: per- and polyfluoroalkyl substances (PFAS), parabens, phenols, phthalates, and organophosphate esters (OPEs). We only analyzed 30 biomarkers detected in >50 % of the sample. After identifying clusters with similar chemical profiles via hierarchical clustering, we applied linear discriminant analysis (LDA) to compute LDA exposure summary scores. In covariate-adjusted models, we used LDA scores to assess co-adjusted, multipollutant associations (relative risk [RR]) with ASD or non-TD, via quasi-Poisson regression. We further examined overall mixture effect and chemical interactions with Bayesian kernel machine regression.

RESULTS

We identified four distinct clusters: PFAS (Cluster 1), OPEs (Cluster 2), parabens and triclosan (Cluster 3), and phthalates and bisphenol A (Cluster 4). Relative to TD, LDA scores for each cluster were associated with increased risk of ASD (RR [95 % CI]: 1.14 [1.03, 1.25], 1.12 [1.01, 1.24], 1.17 [1.07, 1.29], 1.17 [1.07, 1.28] for Cluster 1-4, respectively), whereas clusters 2 and 4 were associated with non-TD (1.07 [1.01, 1.14] and 1.12 [1.05, 1.19], respectively). Cumulative exposure across the four clusters was linked to increased risk of both ASD and non-TD. Potential interactions within and between clusters were observed.

CONCLUSION

This study shows that considering multiple chemical classes resulted in stronger associations with ASD and non-TD risk, compared to when investigated separately in our previous studies.