国家神经系统疾病临床医学研究中心数据管理与共享系统

发表杂志：Environmental Science & Technology

发表类型: Article

影响因子(IF): 11.3

PMID: 41757637 Pubmed原文链接 全文下载

Stroke is the third leading cause of death globally. Ischemic stroke is the most common type of stroke and it should be noted that approximately 17% of stroke burden can be attributed to ambient PM2.5 exposure1–3. In China, where population-weighted PM2.5 concentrations significantly exceed global average, PM2.5-attributed stroke burden is even higher (~25%). To mitigate such health threats, the Chinese government has implemented multiple air quality control policies, and recent years had witnessed significant reduction in fine particulate level. It is estimated that the improvements in air quality may translate to the aversion of more than 1.76 million of premature deaths4,5. However, these evaluations were largely based on the assumption that the health effect of PM2.5 is constant, while such effect has been found to be time-variant, with varied patterns for different locations and outcomes. For instance, studies from South Korea reported an increasing toxicity of PM2.5 on mortality with the association increased from 0.35% to 0.48% for per 10 µg/m3 increase6, while evidences from the Japan, Germany suggested a declining trend7,8. Moreover, the study conducted in the US Medicare dataset showed different temporal trend for cardiovascular and respiratory outcomes regarding short-term exposure to PM2.59. However, there is currently a lack of studies on the temporal variation of PM2.5 in China, which could lead to biased estimation on the impact of air quality control policies, nevertheless, rarely did studies focus on the outcome of ischemic stroke onset. It should also be noted that PM2.5 is a complex mixture of multiple organic and inorganic species, as different mixture of PM2.5 components with varied health effect may contribute to the observed temporal variations in the overall toxicity of PM2.5, the association between PM2.5 components and health outcomes may also change over time. Emerging research indicated that temporal changes in PM2.5 components extend beyond variations in concentration. For instance, Wu et al. documented significant interannual variations in the physicochemical properties of black carbon, suggesting potential modifications in composition10. Epidemiological studies also observed significantly different health effects of carboniferous components at different periods11,12. Nevertheless, there is currently a lack of evidence on the temporal changed in the association between short-term PM2.5 components exposure and ischemic stroke onset. Therefore, based on the individual-level incidence data collected from a national stroke screening program in China, we conducted a case-crossover study to assess the temporal changes in the association between PM2.5 and its components and the risk of ischemic stroke onset from 2015 to 2022.

Prior to analysis, we performed data imputation using the mice package in R, employing a machine learning-based random forest algorithm to fill in the missing values in demographic and health-related variables (gender, age, BMI, smoking status and history of hypertension or diabetes). Conditional logistic regression was used to assess the association between different lag days (lag 0 to lag 2) exposure of PM2.5 total mass and five PM2.5 components and stroke onset. To control for potential confounding from time-varying factors. The 3-day average ambient temperature and relative humidity were both included using natural cubic splines with six and three degrees of freedom separately(see Supplementary Materials)23. The lag time with the strongest estimated association for PM2.5 and its components were used for the subsequent analyses. To evaluate the temporal trend, we estimated and compared separate associations between exposure to PM2.5 and PM2.5 components and stroke risk for two time periods (Period 1: 2015-2017, Period 2: 2018-2022) regarding the implementation timeline of air pollution policies: the Air Pollution Prevention and Control Action Plan (2013-2017) and the Blue Sky Defense War (2018-2020). We further evaluated the annual variation in health effects using two different models (linear interaction models and nonlinear interaction models) (see Supplementary Materials). Consider the non-linear health effect of exposure to PM2.5 and its components24 and their decreasing concentration during the study period, the difference in estimated linear effects across different periods may partially due to its change in concentration distribution. Therefore, we constructed non-linear models by incorporating restricted cubic splines of PM2.5 and its components with knots placed at the 10th, 50th, and 90th percentiles in separate conditional logistic regression models. To calculate the point estimates and 95% confidence intervals (CIs) of the difference characterized by average linear effect between different periods. Subgroup analyses were performed to identify the pattern of temporal trend across two time period in subgroup population stratified by region (North, Northeast, East, Central, South, Southwest, and Northwest China), season (spring, summer, autumn and winter), gender (male, female), age (<65, ≥65 y), BMI (normal, overweight), smoking status (never smoker, current smoker or previous smoker) and history of hypertension or diabetes (presence, absence). We conducted several sensitivity analyses. First, because short-term exposure was easier affected by extreme values, we trimmed the highest 2.5% of daily PM2.5 and its components concentrations during study period to test the robustness of results. Second, data-driven imputation approach used in this study could introduce bias in the estimates, we restricted the analysis in the observations without missing information in demographic and health-related variables. All statistical analyses were performed using R software version 4.5.0 (R Development Core Team, Vienna, Austria). The associations were estimated by point estimates with 95% CIs, and a two-tailed P value < 0.05 was considered as statistically significant.