You are here:
Bayesian multinomial probit modeling of daily windows of susceptibility for maternal PM2.5 exposure and congenital heart defects
Citation:
Warren, J., J. Stingone, Tom Luben, A. Olshan, A. Herring, AND M. Fuentes. Bayesian multinomial probit modeling of daily windows of susceptibility for maternal PM2.5 exposure and congenital heart defects. STATISTICS IN MEDICINE. John Wiley & Sons, Ltd., Indianapolis, IN, (2016).
Impact/Purpose:
To jointly identify daily critical periods of development during weeks 2-8 of pregnancy with respect to maternal particulate matter less than 2.5 micrometers (PM2.5) exposure and the development of 12 individual types of congenital heart defects (CHD).
Description:
Past epidemiologic studies suggest maternal ambient air pollution exposure during critical periods of the pregnancy is associated with fetal development. We introduce a multinomial probit model that allows for the joint identification of susceptible daily periods during the pregnancy for 12 individual types of CHDs with respect to maternal PM2.5 exposure. We apply the model to a dataset of mothers from the National Birth Defect Prevention Study where daily PM2.5 exposures from weeks 2-8 of pregnancy are assigned (specific to each location and pregnancy date) using predictions from the downscaler pollution model. Results are compared to an aggregated exposure model which defines exposure as the average value over pregnancy weeks 2-8. Increased PM2.5 exposure during pregnancy days 53 and 50-51 for pulmonary valve stenosis and tetralogy of Fallot, respectively, are associated with an increased probability of development of each CHD. The largest estimated effect is seen for atrioventricular septal defects on pregnancy day 14. The aggregated exposure model fails to identify any significant windows of susceptibility during pregnancy weeks 2-8 for the considered CHDs. Considering daily PM2.5 exposures in a new modeling framework revealed positive associations for defects that the standard aggregated exposure model was unable to identify. Disclaimer: The views expressed in this manuscript are those of the authors and do not necessarily represent the views or policies of the U.S. Environmental Protection Agency
