The short-term effect of ozone on pregnancy loss modified by temperature: Findings from a nationwide epidemiological study in the contiguous United States.

BACKGROUND

Pregnancy loss, a major health issue that affects human sustainability, has been linked to short-term exposure to ground-surface ozone (O). However, the association is inconsistent, possibly because of the co-occurrence of O and heat episodes, as increased temperature is a risk factor for pregnancy loss. To explain this inconsistency, the effect of O on pregnancy loss needs to be examined jointly with that of high temperature.

METHODS

A total of 247,305 pregnancy losses during the warm season were extracted from fetal death certificates from the 386 counties in contiguous United States from 1989 to 2005. We assessed environmental exposure based on the daily maximum 8 h average of O from Air Quality System monitors and the 24 h average temperature from the North American Regional Reanalysis product. We conducted a bidirectional, time-stratified case-crossover study of the association between pregnancy loss and exposures to O and temperature and their multiplicative interaction. The main time window for the exposure assessment was the day of case occurrence and the preceding 3 days. To estimate the association, we used conditional logistic regression with adjustment for relative humidity, height of the planetary boundary layer, and holidays. Sensitivity analyses were performed on the lagged structure, nonlinearity, and between-subpopulation heterogeneity of the estimated joint effect.

RESULTS

The joint effect was first estimated by the regression against categorical exposure by tertile. Compared to the low-low exposure group (O ≤ 78 μg/m and temperature ≤ 18 °C), the odds of pregnancy loss was significantly higher by 6.0 % (95 % confidence interval [CI] 2.4-9.7 %), 9.8 % (6.1-13.8 %), and 7.5 % (4.7-10.3 %) in the high-low (>104 μg/m and ≤18 °C), low-high (≤78 μg/m and >23 °C), and high-high (>104 μg/m and >23 °C) groups. The model of linear exposure and the multiplicative interaction yielded similar results. Each increment of 10 μg/m in O and 1 °C in temperature was associated with a 3.0 % (2.0 %-4.0 %) and 3.9 % (3.5 %-4.3 %), respectively, increase in the odds of pregnancy loss. A decrease in odds of 0.2 % (0.1 %-0.2 %) was associated with the temperature × O interaction. The finding of an antagonistic interaction between temperature and O was confirmed by models parametrizing the joint exposure as alternative nonlinear terms (i.e., a two-dimensional spline term or a varying-coefficient term) and was robust to a variety of exposure lags and stratifications. Therefore, the marginal effect of O was estimated to vary by climate zone. A significant association between O and pregnancy loss was observed in the northern, but not southern, United States.

CONCLUSION

Joint exposure to O and high temperature can increase the risk for pregnancy loss. The adverse effect of O is potentially modified by ambient temperature. In high-latitude cities, controlling for O pollution could protect maternal health.