Method-Informed Mixture Assessment of Metals Exposure and Socioeconomic Contributions to Neurodevelopmental Delay

• In Event: The Effects of Early Life Environmental Toxicants on Neurodevelopment From Early to Middle Childhood

Abstract

Background. Exposures to complex mixtures are increasingly acknowledged to contribute to adverse neurodevelopmental outcomes through complex joint or interactive effects. Emerging advanced statistical approaches facilitate analyses of complex mixtures by incorporating high correlations, nonlinearities, and interactions among individual exposures. This presentation compares different analytic approaches to determine relative pre- and peri-natal predictors of risks for delayed verbal development in a pediatric cohort.
Methods. Multivariable regression and Random Forest clustering methods were used to understand the relative importance of pre/peri-natal exposures to metal mixtures and sociodemographic variables as contributors to subsequent verbal developmental delays. Bayesian Kernel Machine Regression (BKMR) was used to parse the dose-response relationships and co-exposure profiles.
Results. Multivariable analyses supported multiple metals in the mixtures were predictive of verbal IQ at age 4-5 (N=134). Random Forest analysis demonstrated that those risks were outweighed by sociodemographic characteristics such as mother’s employment, age, and household income. PReMiuM clustering showed no consistently strong relationships between multiple metal clusters and outcomes. BKMR profiles of dose-response relationships clarified the net interactions of multiple metals when exposure occurred against the background micronutrient sufficiency.
Unemployment status was associated with decreased verbal IQ (β=- 7.9, 95% CI: -14, -2.1) when not offset by household income; older maternal age with increased verbal IQ (β= 0.48, 95% CI: 0.01, 0.94); and household income <20,000/yr was associated with decreased verbal IQ (β= -11, 95% CI: -17,-4). Verbal IQ was significantly increased with doubling of zinc (β=6.0, 95% CI: 1.4, 10.5), manganese (β= 4.2, 95% CI: 0.9, 7.5), mercury (β= 2.0, 95% CI: 0.5, 3.6), iodine (β= 1.4, 95% CI: 0.3, 2.5), but decreased by lead (β= -1.9, 95% CI: -3.6, -0.1).
Conclusions: Using distinct analytic strategies to parse effects of complex exposures clarifies the importance of baseline conditions associated with disparities in understanding the response to exposures. Comparison of the outcome of approaches and their unique characteristics informs the choice of method in specific cases, and interpretation and limitations of results. Our data indicate multiple metal mixture components can produce comparable outcomes through exacerbation of underlying effects associated with nutritional deficiency and poverty.

Authors

• Johnnye Lewis
  Presenting Author

• Li Luo, University of New Mexico

• Debra MacKenzie