Time-varying association between fetal death and gestational exposure to ambient fine particles: a nationwide epidemiological study of 49 million fetuses in the contiguous US from 1989 to 2004

Background Gestational exposure to fine particulate matter (PM2.5) has been reported to be associated with an increased risk of fetal death in recent studies, but earlier studies in the past century have usually reported a non-significant association. As such, it remains unknown whether this adverse effect of PM2.5 exposure varies with time. Methods Nearly 49.2 million eligible birth and fetal death records from 1989 to 2004 were selected from the United States (US) birth and fetal death certificate datasets. For each record, the level of prenatal exposure to PM2.5 was taken as the average concentration in the mother’s residential county during the entire gestational period, according to well-established estimates of monthly levels across the contiguous US. We first stratified the dataset by the month of the last menstrual period (LMP) and then independently evaluated the nationwide association between PM2.5 exposure and fetal death within each stratum using five typical logit models: unadjusted, covariate-adjusted, propensity-score, double robust, and diagnostic-score models. Finally, we conducted a meta-analysis to pool estimated LMP-specific associations and explored how the overall association varied by LMP month. Results Different models showed temporal heterogeneity in the estimated association between PM2.5 exposure and fetal death. According to the meta-analysis, double robust model estimates were more homogeneous than the rest, and thus the model outcome was recognized as the main result. For each 1-µg/m3 increase in prenatal exposure to PM2.5, the pooled odds ratio (OR) of fetal death was estimated to be 1.08 [95% confidence interval (CI): 1.05, 1.10]. The LMP-specific ORs exhibited a slightly increasing trend and a significant seasonal pattern. Compared with the pooled OR among samples with the LMP in spring, the estimates for summer, fall and winter were higher by 11.1% (95% CI: 6.2%, 16.3%), 27.8% (95% CI: 22.1%, 33.8%) and 28.8% (95% CI: 23.7%, 34.1%), respectively. We also found that temporal patterns in the association between PM2.5 exposure and fetal death could be explained by several population-level indicators or modifiers (i.e. ethnicity, maternal age, gestational weight gain, previous pregnancy of abnormal termination and diabetes). Conclusions Prenatal exposure to PM2.5 can increase the risk of fetal death. The effects of PM2.5 exposure may be modified by complex factors, which leads to a time-varying association.