Long-term ozone exposure and all-cause mortality: Cohort evidence in China and global heterogeneity by region

Background: Cohort evidence linking long-term ozone (O3) exposure to mortality remained largely mixed worldwide and was extensively deficient in densely-populated Asia. This study aimed to assess the long-term effects of O3 exposure on all-cause mortality among Chinese adults, as well as to examine potential regional heterogeneity across the globe. Methods: A national dynamic cohort of 42153 adults aged 16+ years were recruited from 25 provinces across Chinese mainland and followed up during 2010–2018. Annual warm-season (April–September) O3 and year-round co-pollutants (i.e., nitrogen dioxide [NO2] and fine particulate matter [PM2.5]) were simulated through validated spatial-temporal prediction models and were assigned to each enrollee in each calendar year. Cox proportional hazards models with time-varying exposures were employed to assess the O3-mortality association. Concentration-response (C-R) curves were fitted by natural cubic spline function to investigate the potential nonlinear association. Both single-pollutant model and co-pollutant models additionally adjusting for PM2.5 and/or NO2 were employed to examine the robustness of the estimated association. The random-effect meta-analysis was adopted to pool effect estimates from the current and prior population-based cohorts (n = 29), and pooled C-R curves were fitted through the meta-smoothing approach by regions. Results: The study population comprised of 42153 participants who contributed 258921.5 person-years at risk (median 6.4 years), of whom 2382 death events occurred during study period. Participants were exposed to an annual average of 51.4 ppb (range: 22.7−74.4 ppb) of warm-season O3 concentration. In the single-pollutant model, a significantly increased hazard ratio (HR) of 1.098 (95% confidence interval [CI]: 1.023−1.179) was associated with a 10-ppb rise in O3 exposure. Associations remained robust to additional adjustments of co-pollutants, with HRs of 1.099 (95% CI: 1.023−1.180) in bi-pollutant model (+PM2.5) and 1.093 (95% CI: 1.018−1.174) in tri-pollutant model (+PM2.5+NO2), respectively. A J-shaped C-R relationship was identified among Chinese general population, suggesting significant excess mortality risk at high ozone exposure only. The combined C-R curves from Asia (n = 4) and North America (n = 17) demonstrated an overall increased risk of all-cause mortality with O3 exposure, with pooled HRs of 1.124 (95% CI: 0.966−1.307) and 1.023 (95% CI: 1.007−1.039) per 10-ppb rise, respectively. Conversely, an opposite association was observed in Europe (n = 8, HR: 0.914 [95% CI: 0.860−0.972]), suggesting significant heterogeneity across regions (P < 0.01). Conclusions: This study provided national evidence that high O3 exposure may curtail long-term survival of Chinese general population. Great between-region heterogeneity of pooled O3-mortality was identified across North America, Europe, and Asia.