Comparing exposure assessment methods for traffic-related air pollution in an adverse pregnancy outcome study

Abstract

Background Previous studies reported adverse impacts of traffic-related air pollution exposure on pregnancy outcomes. Yet, little information exists on how effect estimates are impacted by the different exposure assessment methods employed in these studies. Objectives To compare effect estimates for traffic-related air pollution exposure and preeclampsia, preterm birth (gestational age less than 37 weeks), and very preterm birth (gestational age less than 30 weeks) based on four commonly used exposure assessment methods. Methods We identified 81,186 singleton births during 1997–2006 at four hospitals in Los Angeles and Orange Counties, California. Exposures were assigned to individual subjects based on residential address at delivery using the nearest ambient monitoring station data [carbon monoxide (CO), nitrogen dioxide (NO 2), nitric oxide (NO), nitrogen oxides (NO x ), ozone (O 3), and particulate matter less than 2.5 (PM 2.5) or less than 10 (PM 10) μm in aerodynamic diameter], both unadjusted and temporally adjusted land-use regression (LUR) model estimates (NO, NO 2, and NO x ), CALINE4 line-source air dispersion model estimates (NO x and PM 2.5), and a simple traffic-density measure. We employed unconditional logistic regression to analyze preeclampsia in our birth cohort, while for gestational age-matched risk sets with preterm and very preterm birth we employed conditional logistic regression. Results We observed elevated risks for preeclampsia, preterm birth, and very preterm birth from maternal exposures to traffic air pollutants measured at ambient stations (CO, NO, NO 2, and NO x ) and modeled through CALINE4 (NO x and PM 2.5) and LUR (NO 2 and NO x ). Increased risk of preterm birth and very preterm birth were also positively associated with PM 10 and PM 2.5 air pollution measured at ambient stations. For LUR-modeled NO 2 and NO x exposures, elevated risks for all the outcomes were observed in Los Angeles only—the region for which the LUR models were initially developed. Unadjusted LUR models often produced odds ratios somewhat larger in size than temporally adjusted models. The size of effect estimates was smaller for exposures based on simpler traffic density measures than the other exposure assessment methods. Conclusion We generally confirmed that traffic-related air pollution was associated with adverse reproductive outcomes regardless of the exposure assessment method employed, yet the size of the estimated effect depended on how both temporal and spatial variations were incorporated into exposure assessment. The LUR model was not transferable even between two contiguous areas within the same large metropolitan area in Southern California.