Differential Mortality Risks Associated With PM2.5 Components: A Multi-Country, Multi-City Study

Abstract

Background: The association between fine particulate matter (PM2.5) and mortality widely differs between as well as within countries. Differences in PM2.5 composition can play a role in modifying the effect estimates, but there is little evidence about which components have higher impacts on mortality.

Methods: We applied a 2-stage analysis on data collected from 210 locations in 16 countries. In the first stage, we estimated location-specific relative risks (RR) for mortality associated with daily total PM2.5 through time series regression analysis. We then pooled these estimates in a meta-regression model that included city-specific logratio-transformed proportions of seven PM2.5 components as well as meta-predictors derived from city-specific socio-economic and environmental indicators.

Results: We found associations between RR and several PM2.5 components. Increasing the ammonium (NH4+) proportion from 1% to 22%, while keeping a relative average proportion of other components, increased the RR from 1.0063 (95% confidence interval [95% CI] = 1.0030, 1.0097) to 1.0102 (95% CI = 1.0070, 1.0135). Conversely, an increase in nitrate (NO3-) from 1% to 71% resulted in a reduced RR, from 1.0100 (95% CI = 1.0067, 1.0133) to 1.0037 (95% CI = 0.9998, 1.0077). Differences in composition explained a substantial part of the heterogeneity in PM2.5 risk.

Conclusions: These findings contribute to the identification of more hazardous emission sources. Further work is needed to understand the health impacts of PM2.5 components and sources given the overlapping sources and correlations among many components.

Figure 1. Locations used in the study with their mean PM_2.5 concentration and best linear unbiased predictions (BLUPs) of relative risks (RRs) per 10 μg/m ³ increase in PM_2.5.

Figure 2. Annual geometrical mean of the PM_2.5 composition in each country.

Figure 3

Effect modification from each PM_2.5 component. A: relative excess risk (RER) associated to doubling the relative proportion of each component with 95% confidence interval. B: predicted relative risks (RRs) for different values of each component while keeping the other constituents constant. The predicted RR is associated with an increase of 10μg/m³ in PM_2.5. Thick lines indicate the range of observed values for each component, while thin dashed lines indicate extrapolations. Colored bands represent 95% confidence regions.