Maternal Smoking as a Confounder in Studies of Air Pollution and Infant Mortality

To the Editor: aternal smoking increases the risk of various adverse birth and infant outcomes, including intrauterine growth restriction, preterm delivery, and infant mortality.1–3 Consequently, if maternal smokers are also more likely to be exposed to other environmental toxins, smoking could be an important confounder of the relationship between these environmental exposures and adverse infant outcomes. Based on suggestive evidence of an association between air pollution and postneonatal mortality,3 we recently undertook an investigation using 1999–2002 U.S. linked birth–infant death records4 and county-level ambient pollution monitoring data.5 Postneonatal infant mortality due to respiratory causes was defined as infant death between 1 and 12 months of age with International Classification of Diseases, 10th Revision codes of J000–J984 or P271. Infants were assigned pollution levels based on the ambient levels in their county during their first 2 months of life. After exclusions and adjustments for maternal education, marital status, age, primiparity, county-level poverty indicators, birth region, birth month, and birth year, we found modest elevations in risk of postneonatal respiratory death with exposures to PM10, PM2.5, and carbon monoxide (CO) (results not presented here). To assess whether these relationships were confounded by maternal smoking, we examined the subset of births from our study population with maternal smoking data (yes vs no for the question of smoking during pregnancy). Using this subset, we fit logistic generalized estimating equation models with and without smoking to observe the variability of the pollutant parameter estimates attributable to the presence of smoking in the model. Maternal smoking information was available for 2,469,164 births in the pollution-linked data set, including 453 postneonatal respiratory deaths; this constituted 69% of our study population. Many of the excluded births occurred in California where maternal smoking status is not recorded on the birth certificate. Within this population, the prevalence of reported maternal smoking was 9%. The prevalence varied by region with the highest prevalence of 12% reported in the Midwest. Limiting our analysis to this subset of births with reported maternal smoking status, we re-examined the association between air pollution and postneonatal respiratory death, adding adjustment for maternal smoking. In all models, maternal smoking was a relatively strong independent predictor of infant respiratory mortality (Table 1). However, controlling for smoking in the model had no effect on the parameter estimates for any of the pollutants.TABLE 1: Associations of Smoking and Air Pollution With Postneonatal Respiratory DeathValidation studies comparing birth records with other sources of maternal smoking data demonstrate that maternal smoking is underreported.2,6 Nevertheless, if smoking is indeed a confounder of the association between ambient pollution level and postneonatal respiratory death, we would expect at least some attenuation in the pollutant parameter estimates after controlling for the portion of maternal smoking that was reported. Any assessment of maternal smoking and infant outcomes using the birth record has limitations. Some states have begun to collect more detailed information regarding the timing of smoking during pregnancy (eg, which trimester).2 When these data are more widely available, this study question could be reexamined. In our investigation, we did not find any indication that smoking was a confounder of the relationship between ambient air pollution and postneonatal infant mortality due to respiratory causes. The lack of any attenuation in the pollutant effect estimates after controlling for maternal smoking suggests that residual confounding by unreported smoking in this cohort is unlikely. The views expressed are those of the authors and not of their institutions. Lyndsey A. Darrow Rollins School of Public Health, Emory University, Atlanta, GA, [email protected] Tracey J. Woodruff Office of Policy, Economics, and Innovation, US Environmental Protection Agency, Washington, DC Jennifer D. Parker National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD