Abstract
BACKGROUND: Coal-fired power plants are major contributors of ambient sulfur dioxide (SO(2)) air pollution. Epidemiological literature suggests an adverse association between SO(2) exposure during gestation and preterm birth (PTB; <37 weeks completed gestation). PTB is strongly associated with infant mortality and increased risk for later life morbidities. OBJECTIVE: We investigated associations between SO(2) and PTB in North Carolina and evaluated whether the associations were modified by race/ethnicity. METHODS: We assembled a retrospective, administrative cohort of singleton births in North Carolina from 2003-2015. We used US EPA EQUATES data to assign long-term SO(2) gestational exposures to eligible births for the entire pregnancy and by trimester. We used multivariable generalized linear regression to estimate risk differences (RD (95%CI)) per 1-ppb increase in SO(2), adjusted for gestational parent education, Medicaid status, marital status, and season of conception. Multi-pollutant models were additionally adjusted for other criteria air co-pollutants (O(3), PM(2.5), NO(2)). RESULTS: The median SO(2) (24-hour average) across exposure windows was ~1.5 (IQR: 1.8) ppb. The overall baseline risk for PTB was 8,756 per 100,000 live births. When stratified by race/ethnicity, the baseline risk for PTB was 12215, 7824, and 7187 per 100,000 live births among non-Hispanic Black, non-Hispanic white, and Hispanic births, respectively. RDs per 1-ppb increase in SO(2) averaged across the entire pregnancy were 317.0 (95%CI: 279.4, 354.5) and 568.2 (95%CI: 500.3, 636.1) per 100,000 live births for single- and multi-pollutant models, respectively. For the PTB multi-pollutant models, we observed similar RDs for non-Hispanic Black participants (669.6 [95%CI: 573.9, 765.2]) and non-Hispanic white participants (635.4 [95%CI: 557.2, 713.6]) with smaller RDs for Hispanic participants (336.8 [95%CI: 241.3, 432.2]). SIGNIFICANCE: The results for our adjusted single- and multi-pollutant models showed adverse associations between SO(2) and PTB, with some evidence of effect measure modification by race/ethnicity within subcategories of PTB.