Single- and two-pollutant concentration-response functions for PM_2.5 and NO₂ for quantifying mortality burden in...

Health Impact Assessments (HIAs) for air pollutant mixtures are challenging because risk estimates are primarily derived from single-pollutant models. Combining risk estimates from multiple pollutants requires new approaches, as a simple addition of single pollutant risk estimates from correlated air pollutants may result in double counting. We investigated approaches applying concentration-response functions (CRFs) from single- and two-pollutant models in HIAs, focusing on long-term exposure to particulate matter with a diameter less than 2.5 μm (PM_2.5) and nitrogen dioxide (NO₂) and their associations with all-cause mortality.

A systematic literature search of MEDLINE and EMBASE identified cohort studies employing single- and two-pollutant models of long-term exposure to PM_2.5 and NO₂ with all-cause mortality. Pooled CRFs were calculated through random-effects meta-analyses of risk estimates from single- and two-pollutant models. Coefficient differences were calculated by comparing single- and two-pollutant model estimates. Four approaches to estimating population-attributable fractions (PAFs) were compared: PM_2.5 or NO₂ single-pollutant models to represent the mixture, the sum of single-pollutant models, the sum of two-pollutant models and the sum of single-pollutant models from a larger body of evidence adjusted by coefficient difference.

Seventeen papers reported both single and two-pollutant estimates. Pooled hazard ratios (HRs) for mortality from single- and two-pollutant models were 1.053 (95% confidence interval: 1.034-1.071) and 1.035 (1.014-1.057), respectively, for a 5 μg/m³ increase in PM_2.5. HRs for a 10 μg/m³ increase in NO₂ were 1.032 (1.014-1.049) and 1.024 (1.000-1.049) for single- and two-pollutant models, respectively. The average coefficient difference between single- and two-pollutant models was 0.017 for PM_2.5 and 0.007 for NO₂. Combined PAFs for the PM_2.5-NO₂ mixture using joint HRs from single- and two-pollutant model CRFs were 0.09 and 0.06, respectively.

Utilizing CRFs from two-pollutant models or applying the coefficient difference to a more extensive evidence base seems to mitigate the potential overestimation of mixture health impacts from adding single-pollutant CRFs.