Tiny but Mighty: Association between fine particulate matter (PM2.5) and infant mortality in a North Carolina Birth Cohort (2003-2015)

Many studies have quantified the association between fine particulate matter (PM2.5) and adult mortality. Several studies have evaluated particulate matter (PM10) and postneonatal mortality, but there are relatively few studies about PM2.5 and infant mortality. We investigated the association between long-term PM2.5 exposure and all infant mortality, neonatal mortality and postneonatal mortality.Methods: We conducted an unmatched case-control study of infant mortality sampling 10 controls per case from a North Carolina birth cohort (2003-2015). PM2.5 exposure was estimated at the birthing parent’s residence based on spatiotemporal modeling of PM2.5 concentrations at the census block averaged over 2-week periods. We performed logistic regression with adjustment for birthing parent characteristics, year of birth, infant sex assigned at birth, urbanicity and neighborhood deprivation index (NDI). Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for an increase equal to the interquartile range of PM2.5 (4.0 µg/m3) for five critical exposure periods (each trimester, pregnancy, and first month of life). We conducted additional stratification analyses and a ‘leave-one-out’ analysis to identify potential confounders. Results: Our analytical dataset included 1,315,691 infants from which we identified 5,992 cases of infant mortality and sampled 60,000 non-cases to serve as controls for our analytical sample. After adjusting for covariates, the odds of infant mortality increased by 5% (OR: 1.05, 95% CI: 0.95, 1.17) for each 4.0 µg/m3 increase in PM2.5 exposure averaged over the entire pregnancy. We observed a slightly higher increase in odds for postneonatal mortality (OR: 1.09, 95% CI: 0.94, 1.27) and a modest increase in odds for neonatal mortality (OR: 1.03, 95% CI: 0.90. 1.18), although confidence intervals included null values. All adjusted odds included null values but revealed positive point estimates for the association between infant and postneonatal mortality and PM2.5 exposure during trimester 1 and trimester 2, and neonatal mortality and PM2.5 exposure during trimester 3 and revealed negative or null point estimates for all other exposure periods and outcomes. Leaving out year of birth from the adjusted analysis in pregnancy PM2.5 exposure resulted in estimates similar to the unadjusted model, and stratifying by year of birth resulted in highly varied effect estimates by year. Conclusions: Our results suggest that long-term PM2.5 exposure over the full pregnancy is modestly associated with infant mortality and postneonatal mortality with effects across other exposure periods near the null value. Decisions about adjusting for temporal variables appear to greatly influence effect estimates. Future work in this field will be important to inform policies that aim to protect infant’s environmental health.