Mixture of air pollution, brominated flame retardants, polychlorinated biphenyls, per- and polyfluoroalkyl substances, and organochlorine pesticides in relation to vitamin D concentrations in pregnancy

Over two-thirds of pregnant women in the U.S. have insufficient 25(OH)D (Vitamin D) concentrations, which can adversely impact fetal health. Several pollutants have been associated with 25(OH)D, but have not been considered in the context of chemical co-exposures. We aimed to determine associations between a broad mixture of prenatal environmental chemical exposures and 25(OH)D concentrations in mid-pregnancy. Stored mid-pregnancy serum samples were assayed from 421 women delivering live births in Southern California in 2000–2003. 25(OH)D, six BFRs, eleven polychlorinated biphenyls (PCBs), six per- and polyfluoroalkyl substances, and two organochlorine pesticides were detected in ≥60% of specimens. Gestational exposures to airborne particulate matter ≤ 10 μm (PM10) and ≤ 2.5 μm (PM2.5), nitrogen monoxide (NO), nitrogen dioxide (NO2), and ozone concentrations were derived from monitoring station data. Bayesian Hierarchical Modeling (BHM) and Bayesian Kernel Machine Regression (BKMR) analyses estimated overall mixture and individual chemical associations accounting for co-exposures and covariates with mean 25(OH)D levels, and BHM was used to estimate associations with insufficient (<75 nMol/L) 25(OH)D levels. Non-mixture associations for each chemical were estimated with linear and logistic models. PM10 [BHM estimate: −0.133 nmol/l 95% Credible Interval (−0.240, −0.026)] was associated with lower 25(OH)D in BHM and BKMR. Higher quantiles of combined exposures were associated with lower 25(OH)D, though with wide credible intervals. In non-mixture models, PM10, PM2.5, NO, and NO2 were associated with lower concentrations, while O3 and PBDE153 were associated with higher 25(OH)D and/or lower insufficiency. While some chemicals were associated with increased and others with decreased 25(OH)D concentrations, the overall mixture was associated with lower concentrations. Mixture analyses differed from non-mixture regressions, highlighting the importance of mixtures approaches for estimating real-world associations.