Long-term Exposure to Photochemical Oxidants and All-Cause Mortality: A Meta-Analysis of Two-Pollutant Models with...

Photochemical oxidants, represented by nitrogen dioxide (NO₂) and ozone (O₃), are recognized health hazards. However, due to the complex nonlinear photochemical coupling between them in the atmosphere, strong collinearity makes it difficult for traditional epidemiological models to estimate the respective associations of NO₂ and O₃ with health outcomes after accounting for co-pollutant dependence. Single-pollutant models may be affected by omitted variable bias (OVB) when a correlated co-pollutant is not included, potentially influencing the estimated associations. Meanwhile, traditional two-pollutant models often lack statistical power due to scarce sample sizes and variance inflation.

We applied an established OVB-based meta-analytic framework based on the OVB equation to systematically integrate results from both single-pollutant and two-pollutant models. We searched cohort studies until 2024, and used the OVB-corrected meta-analysis to combine the effect estimates from single- and two-pollutant models. Population-weighted RRs were used as the primary summaries of the model-based associations. Geographic and population coverage was additionally evaluated as an exploratory analysis using a pointwise grid-cell probability criterion.

Traditional two-pollutant models indicated a risk ratio (RR) of 1.120 for all-cause mortality per 10 ppbv increment in NO₂, but with a wide confidence interval (95% CI: 1.057-1.187). After applying the OVB correction and incorporating single-pollutant data, the effect estimate for NO₂ was adjusted to 1.076 (95% CI: 1.040-1.112). In contrast, the corrected effect size for O₃ was 1.015 (95% CI: 0.984-1.048). We showed a positive association between NO₂ and all-cause mortality, conditional on the modeled dependence between NO₂ and O₃ under the stated assumptions.

Under the stated assumptions, long-term NO₂ exposure was associated with increased all-cause mortality after adjustment for the modeled dependence between NO₂ and O₃. These findings highlight the relevance of accounting for spatial variation in NO₂-O₃ dependence when interpreting long-term exposure-mortality associations.