Prenatal exposure to environmental mixtures and asthma risk among children

Abstract

Introduction Exposures to various environmental contaminants have been associated with increased asthma risk among children. Prenatal exposures can trigger changes to the immune system in the developing fetus which can lead to an onset of childhood allergies including asthma. Children exposed to multiple environmental stressors may be more susceptible. We investigated the association between asthma risk and mixtures of environmental stressors among children born 2000-2005 to mothers living near a PCB-contaminated Superfund Site in New Bedford Harbor, Massachusetts. Methods We identified 10,517 births from four towns surrounding the Superfund Site who were followed until 2010. There were 428 asthma cases identified from emergency department and hospital discharge records. We examined the joint effects of modeled PCBs, DDE, distance to road, and maternal age on asthma risk by smoothing multiple exposures using generalized additive models (GAMs) adjusted for covariates including maternal education, prenatal smoking and alcohol use. Advantages of this approach include the ability to smooth variables of different units, generalizability to multiple covariates, and an overall test of the importance of the mixture. Results DDE, distance to road, and maternal age at birth were included in a multi-dimensional smooth, and the model was further adjusted parametrically for confounders. Results suggest that children of younger mothers who were also exposed to the highest levels of DDE and lived closest to major roads were at greatest risk of developing asthma (OR= 1.6). This distance to road and DDE pattern was not apparent for older mothers. PCBs, which were highly correlated with DDE, produced similar results in separate models. Conclusions This example illustrates the utility of using GAMs for visualizing mixtures; the importance of both chemical and non-chemical stressors in determining asthma risk would not likely have been discovered using more traditional epidemiologic models.