Associations between Coarse Particulate Matter Air Pollution and Cause-Specific Mortality: A Nationwide Analysis in 272 Chinese Cities

Background: Coarse particulate matter with aerodynamic diameter between 2.5 and 10μm (PM2.5–10) air pollution is a severe environmental problem in developing countries, but its challenges to public health were rarely evaluated. Objective: We aimed to investigate the associations between day-to-day changes in PM2.5–10 and cause-specific mortality in China. Methods: We conducted a nationwide daily time-series analysis in 272 main Chinese cities from 2013 to 2015. The associations between PM2.5–10 concentrations and mortality were analyzed in each city using overdispersed generalized additive models. Two-stage Bayesian hierarchical models were used to estimate national and regional average associations, and random-effect models were used to pool city-specific concentration–response curves. Two-pollutant models were adjusted for fine particles with aerodynamic diameter ≤2.5μm (PM2.5) or gaseous pollutants. Results: Overall, we observed positive and approximately linear concentration–response associations between PM2.5–10 and daily mortality. A 10-μg/m3 increase in PM2.5–10 was associated with higher mortality due to nonaccidental causes [0.23%; 95% posterior interval (PI): 0.13, 0.33], cardiovascular diseases (CVDs; 0.25%; 95% PI: 0.13, 0.37), coronary heart disease (CHD; 0.21%; 95% PI: 0.05, 0.36), stroke (0.21%; 95% PI: 0.08, 0.35), respiratory diseases (0.26%; 95% PI: 0.07, 0.46), and chronic obstructive pulmonary disease (COPD; 0.34%; 95% PI: 0.12, 0.57). Associations were stronger for cities in southern vs. northern China, with significant differences for total and cardiovascular mortality. Associations with PM2.5–10 were of similar magnitude to those for PM2.5 in both single- and two-pollutant models with mutual adjustment. Associations were robust to adjustment for gaseous pollutants other than nitrogen dioxide and sulfur dioxide. Meta-regression indicated that a larger positive correlation between PM2.5–10 and PM2.5 predicted stronger city-specific associations between PM2.5–10 and total mortality. Conclusions: This analysis showed significant associations between short-term PM2.5–10 exposure and daily nonaccidental and cardiopulmonary mortality based on data from 272 cities located throughout China. Associations appeared to be independent of exposure to PM2.5, carbon monoxide, and ozone. https://doi.org/10.1289/EHP2711

. Percentage difference (posterior mean and 95% posterior intervals) in daily mortality per 10 μg/m 3 increase in 2-day moving average PM 2.5-10 concentrations in 272 Chinese cities, classified by regions and causes. Table S2. Percentage difference (posterior mean and 95% posterior intervals) in daily mortality per city-specific interquartile range increase in 2-day moving average PM 2.5-10 concentrations in 272 Chinese cities, classified by regions and causes. Table S3. The association of PM 2.5-10 and PM 2.5 with cause-specific mortality in single-pollutant and two-pollutant models in 272 Chinese cities. Table S4. Percentage difference (posterior mean and 95% posterior intervals) in daily mortality per 10 μg/m 3 increase in 2-day moving average PM 2.5-10 concentrations in single-pollutant and two-pollutant models with gaseous pollutants. Table S5. Percentage difference (posterior mean and 95% posterior intervals) in daily mortality per 10 μg/m 3 increase in 2-day moving average PM 2.5-10 concentrations, stratified by subgroups in 272 Chinese cities. Table S6. The impacts of annual-mean levels of city characteristics on the association between PM 2.5-10 and total non-accidental mortality in single-variable and combined meta-regression models. Figure S1. Percentage difference (posterior mean and 95% posterior interval) in daily total mortality per 10 μg/m 3 increase in 2-day moving average concentrations of coarse PM, using different lag days of coarse PM, df per year in smoothness of time and lag days for controlling daily-mean temperature and relative humidity. Overdispersed generalized additive models were used to derive city-specific estimates adjusted for time trends, day of week, temperature and humidity and Bayesian hierarchical models were used to pool the estimates.
Abbreviations: PM, particulate matter; df, degree of freedom. Table S1. Percentage difference (posterior mean and 95% posterior intervals) in daily mortality per 10 μg/m 3 increase in 2-day moving average PM 2.5-10 concentrations in 272 Chinese cities, classified by regions and causes.

Mortality
Nationwide ( Note: Overdispersed generalized additive models were used to derive city-specific estimates adjusted for time trends, day of week, temperature, and humidity and Bayesian hierarchical models were used to pool the estimates. The p-value was derived by examining the difference of effect estimates between the north and south by virtue of meta-regression analysis. These data are also reported in Figure 2.
Abbreviations: PIs, posterior intervals; PM 2.5-10 , particulate matter with an aerodynamic diameter between 2.5 and 10μm; CVD, cardiovascular diseases; CHD, coronary heart diseases; RD, respiratory diseases; COPD, chronic obstructive pulmonary disease. Note: Overdispersed generalized additive models were used to derive city-specific estimates adjusted for time trends, day of week, temperature and humidity and Bayesian hierarchical models were used to pool the estimates. P-values are derived by examining the difference of effect estimates between the north and south by virtue of meta-regression analysis.
Abbreviations: PM 2.5-10 , particulate matter with an aerodynamic diameter between 2.5 and 10 μm; CVD, cardiovascular diseases; CHD, coronary heart diseases; RD, respiratory diseases; COPD, chronic obstructive pulmonary disease. Note: The associations were expressed as percentage difference (posterior mean and 95% posterior intervals) in daily mortality per 10 μg/m 3 increase in 2-day moving average concentrations, which were firstly estimated by overdispersed generalized additive models adjusted for time trends, day of week, temperature and humidity in each city and were pooled by Bayesian hierarchical models. The adjustment was performed by adding the second pollutant to the first-stage models. P-values were derived by examining statistical significance of the dichotomous co-pollutant variable in a meta-regression analysis with both single-and two-pollutant model estimates. These data are also reported in Figure 3.
Abbreviations: PM 2.5-10 , particulate matter with an aerodynamic diameter between 2.5 and 10 μm; PM 2.5 , particulate matter with an aerodynamic diameter ≤ 2.5 μm; CVD, cardiovascular diseases; CHD, coronary heart diseases; RD, respiratory diseases; COPD, chronic obstructive pulmonary disease. Note: Overdispersed generalized additive models were used to derive city-specific estimates adjusted for time trends, day of week, temperature and humidity and Bayesian hierarchical models were used to pool the estimates. The adjustment was performed by adding the second pollutant to the first-stage models. P-values were derived by examining statistical significance of the dichotomous co-pollutant variable in a meta-regression analysis with both single-and two-pollutant model estimates. These data are also reported in Figure 4.
Abbreviations: PM 2.5-10 , particulate matter with an aerodynamic diameter between 2.5 and 10μm; SO 2 , sulfur dioxide; NO 2 , nitrogen dioxide; CO, carbon monoxide; O 3 , ozone; CVD, cardiovascular diseases; CHD, coronary heart diseases; RD, respiratory diseases; COPD, chronic obstructive pulmonary disease. Note: Overdispersed generalized additive models were used to derive city-specific estimates adjusted for time trends, day of week, temperature and humidity and Bayesian hierarchical models were used to pool the estimates. P-values were derived by conducting likelihood ratio tests comparing the goodness-of-fit of a meta-regression model with the potential modifier to the simple meta-analysis model without this variable. These data are also reported in Figure 6.
Abbreviations: PM 2.5-10 , particulate matter with an aerodynamic diameter between 2.5 and 10μm; CVD, cardiovascular diseases; RD, respiratory diseases. Note: a It refers to Pearson correlation coefficients for annual mean PM 2.5-10 and PM 2.5 concentrations; b They are percentage differences (posterior mean and 95% posterior intervals) in daily total non-accidental mortality per 10 μg/m 3 increase in 2-day moving average PM 2.5-10 concentrations in association with a 1-unit increase in annual-mean levels of city characteristics.