Cognitive flexibility, the ability to smoothly adapt to changing circumstances, is a skill that is vital to higher-level executive functions such as problem-solving, planning, and reasoning. As it undergoes substantial development during adolescence, decrements in cognitive flexibility may not become apparent until this time. There is evidence that prenatal exposure to individual chemicals may adversely impact executive functions in children, but few studies have explored the association of co-exposure to multiple chemicals with cognitive flexibility specifically among adolescents. We investigated this association among a diverse group of adolescents living near a Superfund site in New Bedford, Massachusetts. Specifically, using Bayesian kernel machine regression (BKMR) and multivariable regression analyses, we investigated the association of biomarkers of prenatal exposure to organochlorines (DDE, HCB, PCBs) and metals (lead, manganese) with cognitive flexibility, measured with four subtests of the Delis-Kaplan Executive Function System. In BKMR models, we observed adverse joint associations of the chemical mixture with two of the four cognitive flexibility subtests. In covariate-adjusted linear regression models, a two-fold increase in cord blood Mn was associated with poorer performance on two of the subtests: Trail-Making (scaled score difference = −0.60; 95% CI: −1.16, −0.05 points) and Color-Word Interference (scaled score difference = −0.53; 95% CI: −1.08, 0.01 points). These adverse Mn-cognitive flexibility associations were supported by the results of the BKMR. There was little evidence of effect modification by sex and some evidence of effect modification by a measure of social disadvantage, particularly for the associations between HCB and cognitive flexibility. This study is among the first to provide evidence of an adverse association of prenatal exposure to a chemical mixture with cognitive flexibility in adolescence.
Machine Learning Enhanced Spectroscopic Analysis: Towards Autonomous Chemical Mixture Characterization for Rapid Process Optimization
