Bayesian Varying Coefficient Kernel Machine Regression to Assess Cognitive Trajectories Associated with Exposure to Complex Mixtures

Abstract

Background/Aim: Exposure to heavy metal mixtures during early life may exert wide-ranging effects on children's neurodevelopment and cognitive growth trajectories. However, there is a lack of statistical methods that can simultaneously accommodate the complex exposure-response relationship between metal mixtures and neurodevelopment, while estimating cognitive trajectories. Methods: We introduce Bayesian Varying Coefficient Kernel Machine Regression (BVCKMR), a hierarchical model that estimates how mixture exposures at a given time point are associated with baseline cognition and cognitive trajectories. BVCKMR flexibly captures the exposure-response relationship, incorporates prior knowledge, and accounts for non-linear and non-additive effects of individual exposures. Using contour plots and cross-sectional plots, BVCKMR provides information about interaction and effect modification between complex mixture components. BVCKMR was applied to data from PROGRESS, a prospective birth cohort study in Mexico City on metal mixture exposures and temporal changes in neurodevelopment. The metal mixture exposures included manganese, arsenic, cobalt, chromium, cesium, copper, lead, cadmium and antimony. Results: The simulation study considered three exposure-response relationship scenarios: linear additive, linear with interaction, and quadratic with interaction. Under each scenario, the baseline cognition and cognitive trajectories for each individual were estimated. A regression comparing the true exposure-response function terms and the predicted estimates using BVCKMR yielded R2 values of 0.74 – 0.97. Results from a subset of PROGRESS (N = 665) provide evidence of significant negative associations between second trimester lead and Bayley Scales of Infant and Toddler Development trajectories across 6 – 24 months (effect size -0.10 [-0.16, -0.04] per interquartile range increase in Pb exposure). We detected an interaction effect between second trimester lead and manganese exposures with 24-month Bayley. Conclusions: BVCKMR is a promising statistical approach for investigating the effects of exposure to complex mixtures on cognitive growth trajectories.