Using Birth Cohort Data to Estimate Prenatal Exposures for All Births around the New Bedford Harbor Superfund Site in Massachusetts

Abstract

Introduction: Children born near New Bedford, Massachusetts, USA have been prenatally exposed to environmental chemicals, in part due to an older housing stock, high fish consumption rates, and proximity to the New Bedford Harbor (NBH) Superfund site. Chemical exposure measures are not available for the general population, limiting epidemiologic investigations and potential interventions. Our objective was to leverage available sociodemographic and biomonitoring data from the New Bedford Cohort (NBC) to retrospectively estimate prenatal exposures for all 10,273 births between 1993-1998 in the 4 towns neighboring the NBH Superfund site.Methods: The NBC, a population-based cohort of 788 mother-infant pairs born between 1993-1998 near the NBH, collected questionnaire data and prenatal exposure biomarkers including cord serum polychlorinated biphenyls (PCB), ρ,ρ′-dichlorodiphenyl dichloroethylene (DDE), hexachlorobenzene (HCB), cord blood lead (Pb), and maternal hair mercury (Hg). We used bootstrapped samples of the NBC data to build prenatal exposure models, with multivariable smooths of birth location, birth year, maternal age at birth, and other NBC maternal sociodemographic characteristics as predictors.Results: Maternal country of birth was the strongest exposure predictor across all exposures with women from the Azores and Cape Verde having the highest levels. Maternal age, education, and marital status also were important predictors. The PCB, DDE, HCB, Pb, and Hg exposure models explained 55%, 53%, 41%, 43% and 38% of the variance, respectively.Conclusions: Our analyses suggest that multiple prenatal exposures can be estimated at the population level by modeling available data for a subset of the population. Predictive models using multivariable smoothing explained reasonable amounts of variance. Other exposure models with comparable performance have been used successfully in epidemiologic investigations to characterize exposure-outcome associations.