Association between exposure to ambient air pollution and hospital admission, incidence, and mortality of stroke: an updated systematic review and meta-analysis of more than 23 million participants

Background Previous studies have suggested that exposure to air pollution may increase stroke risk, but the results remain inconsistent. Evidence of more recent studies is highly warranted, especially gas air pollutants. Methods We searched PubMed, Embase, and Web of Science to identify studies till February 2020 and conducted a meta-analysis on the association between air pollution (PM2.5, particulate matter with aerodynamic diameter less than 2.5 μm; PM10, particulate matter with aerodynamic diameter less than 10 μm; NO2, nitrogen dioxide; SO2, sulfur dioxide; CO, carbon monoxide; O3, ozone) and stroke (hospital admission, incidence, and mortality). Fixed- or random-effects model was used to calculate pooled odds ratios (OR)/hazard ratio (HR) and their 95% confidence intervals (CI) for a 10 μg/m3 increase in air pollutant concentration. Results A total of 68 studies conducted from more than 23 million participants were included in our meta-analysis. Meta-analyses showed significant associations of all six air pollutants and stroke hospital admission (e.g., PM2.5: OR = 1.008 (95% CI 1.005, 1.011); NO2: OR = 1.023 (95% CI 1.015, 1.030), per 10 μg/m3 increases in air pollutant concentration). Exposure to PM2.5, SO2, and NO2 was associated with increased risks of stroke incidence (PM2.5: HR = 1.048 (95% CI 1.020, 1.076); SO2: HR = 1.002 (95% CI 1.000, 1.003); NO2: HR = 1.002 (95% CI 1.000, 1.003), respectively). However, no significant differences were found in associations of PM10, CO, O3, and stroke incidence. Except for CO and O3, we found that higher level of air pollution (PM2.5, PM10, SO2, and NO2) exposure was associated with higher stroke mortality (e.g., PM10: OR = 1.006 (95% CI 1.003, 1.010), SO2: OR = 1.006 (95% CI 1.005, 1.008). Conclusions Exposure to air pollution was positively associated with an increased risk of stroke hospital admission (PM2.5, PM10, SO2, NO2, CO, and O3), incidence (PM2.5, SO2, and NO2), and mortality (PM2.5, PM10, SO2, and NO2). Our study would provide a more comprehensive evidence of air pollution and stroke, especially SO2 and NO2. Supplementary Information The online version contains supplementary material available at 10.1186/s12199-021-00937-1.


Introduction
Stroke, characterized by acute cerebral blood circulation disorder, is caused by artery stenosis, occlusion, or rupture caused by various inducing factors in patients with cerebrovascular diseases [1]. Stroke has become a leading contributor to the global burden of disease and the second leading cause of death worldwide [2,3]. According to the Global Burden of Disease Study (GBD) report, there were approximately 80.1 million stroke patients, and 5.5 million deaths were attributed to stroke in 2016 globally [4]. Considering stroke is characterized with high incidence, high mortality, and contribute to severe burden disease, identifying potential risk factor of stroke is of great significance for public health. In parallel, air pollution has also been regarded as one of the major environmental problems and a risk factor of many cardiovascular diseases (CVD), including stroke [5]. GBD 2019 showed that air pollution was globally the sixth leading cause of stroke death during 1990 to 2017, and 28.1% disability-adjusted life years (DALYs) of stroke attribute to environmental factors exposure [6,7].
Air pollution is the most significant environmental risk factor for all-cause mortality [8]. Increasing number of human epidemiologic studies has been conducted to assess the potential association between air pollution exposure and stroke admission, incidence, and mortality in recent years. However, the results were inconsistent, and the associations between exposure to air pollution and stroke have not been fully understood. Some studies reported positive association between air pollution exposure and stroke hospital admission/incidence/mortality, whereas others did not [4,[9][10][11][12][13][14]. For example, Huang et al. 2019 indicated that exposure to PM 2.5 was associated with increased stroke incidence and the adjusted risk ratio (RR) was 1.130 (95%CI: 1.090, 82 1.170) for each increase of 10 μg/m 3 in n PM 2.5 concentration [4]. The adjusted risk ratio (RR) was 1.130 (95% CI 1.090, 1.170) for each increase of 10 μg/m 3 in PM 2.5 concentration, while Wing et al. suggested no association was found between PM 2.5 exposure and stroke incidence (RR = 0.950, 95% CI 0.710, 1.280) [11]. Previous metaanalyses have explored the associations between air pollution exposure and stroke [15][16][17][18][19]. However, these studies were mainly focused on the studies of particulate matter (PM 2.5 , particulate matter with aerodynamic diameter less than 2.5 μm; PM 10 , particulate matter with aerodynamic diameter less than 10 μm) and stroke outcomes [16][17][18][19]; results of gas air pollutants (NO 2 , nitrogen dioxide; SO 2 , sulfur dioxide; CO, carbon monoxide; O 3 , ozone) were scarce. Moreover, to the best of our knowledge, more than 30 studies exploring the association between air pollution exposure and stroke, especially conducted from the multi-city level and with large sample sizes, were published after the most recent metaanalysis. The more recent and comprehensive studies should be included in the meta-analysis to conclude an updated pooled effect estimate.
We therefore conducted an updated systematic review and meta-analysis to assess the association between 6 main air pollutants (PM 2.5 , PM 10 , NO 2 , SO 2 , CO, and O 3 ) and 3 stroke outcomes (hospital admission, incidence, and mortality). This systematic review and metaanalysis was performed according to the guidelines of the Preferred Reporting Items for Systematic Review and Meta-analyses (PRISMA) criteria (Table S1).

Search strategy
Literature was searched in three databases (PubMed, Embase, and Web of Science), with published date until 1 February 2020. The search strategy was pairwise of combinations of terms concerning air pollution (e.g., air pollution, particulate matter, particles, PM 2.5 , PM 10 , nitrogen oxides (NOx), NO 2 , SO 2 , CO, and O 3 ) and stroke (e.g., stroke, cerebrovascular disease, cerebrovascular disorder, cerebral hemorrhage, cerebral infarction, subarachnoid hemorrhage).
We first selected articles by screening titles and abstracts and then the full texts of potentially eligible studies were further evaluated. Reference lists of all the included studies were also manually searched. Literature selection was finished by two independent authors (ZP N and FF L), and conflicts between the two authors were resolved by discussing with an arbitrator (H X).

Inclusion and exclusion criteria
Articles that met the following criteria were included: (1) provided quantitative measure of the associations between air pollution exposure with stroke admission, incidence, and/or mortality (relative risk (RR), odds ratio (OR), or hazard ratio (HR), and their 95% confidence interval (95%CI); (2) cohort, cross-sectional, time series, cross-sectional, case-control, case-crossover, or panel studies; (3) focused on outdoor (ambient) air pollution exposure but not indoor air pollution; (4) original peerreviewed human subject research studies; (5) published in English. Studies were excluded if they were (1) toxicological studies, summaries, or reviews, and (2) articles without effect estimates after contacting the authors. In addition, for more than one article conducted from the same population, only the most recent studies were included.

Data extraction
Data were extracted from all eligible studies, including the following: (1) study characteristics (first author, published year, study location, and period); (2) study population (sample size, proportion of males, range of age, mean age); (3) outcome (type of stroke and outcome was admission, incidence, and/or mortality); (4) air pollution assessment method and increment of air pollution used in effect estimates (per interquartile range (IQR), standard deviation (SD), or per 10 μg/m 3 ); (5) effect estimates of the association between air pollution and stroke risk (OR, RR, HR with 95% CI). The effect estimates of single-pollutant model, generally called "main model" or "fully adjusted model," were extracted [20].

Quality assessment
Two authors (ZP N and FF L) worked independently, and inconsistencies in quality assessment were resolved through discussion. We employed the Newcastle-Ottawa Scale (NOS) to evaluate the quality of included studies. The NOS Tool has designed 8 items to assess the critical appraisal of the potential risk of bias. Total score of NOS ranged from 0-9. Study score higher than or equal to 7 was regarded as high-quality; otherwise, the study was regarded as "low quality" [21].

Statistical analyses
This meta-analysis focused on examining the association between air pollution and three stroke outcomes, including admission, incidence, and/or mortality. We extracted effect estimates (OR, HR, RR, and 95%CI) from individual studies and then converted them into a standardized form of per 10 μg/m 3 increases in air pollution. The significance of the pooled OR, RR, or HR was determined by the Z test [22], and p value less than 0.05 was considered statistically significant. Standard error (SE) for each effect estimate was calculated by using the formula: (upper limit − lower limit)/3.92.
Heterogeneity among studies was evaluated using I 2 statistics and Q test [23]. If the values of I 2 > 50% or p < 0.01, the heterogeneity was "high" and random effect model was used to pool estimates. Otherwise, heterogeneity was considered as "low or moderate," and fixed-effect model was used to pool estimates.
Begg's test and Egger's test were conducted to assess publication bias. The influence of individual studies on the pooled estimates was examined by removing each study from the analysis one by one. Moreover, we also performed sensitivity analysis and subgroup analysis to evaluate if the exposure period would change the significance of the pooled results. Because long-term studies were limited, sensitivity analysis was conducted by omitting long-term exposure (cohort) studies. Subgroup analysis was only performed if the number of short-term exposure studies or long-term exposure studies was more than 3. Publication bias and sensitivity analysis were only performed if the number of included studies was more than 5. All statistical analysis was performed in Stata version 15.0 (StataCorp, College Station, TX, USA).

Literature search and characteristics of included studies
After removing duplicates, 737 records were identified in the initial literature search. By reviewing title and abstracts, 93 studies were downloaded for full-text reading. According to the inclusion and exclusion criteria, a total of 68 studies were included in our meta-analysis (Fig. 1). Table 1 provides the characteristics of 68 studies included in meta-analysis. As for air pollution involved in the study, there were 26 studies that reported the association between air pollution exposure and stroke hospital admission, 19 reported air pollution exposure and stroke incidence, 19 reported air pollution exposure and stroke mortality, and 3 reported both stroke incidence and mortality. The sample size of included studies ranged between 407 and 8,834,533; more than 23 million participants were included in meta-analysis eventually. Furthermore, the studies included were conducted from 18 countries. Time-series and cross-sectional were the most commonly adopted study designs. In our meta-analysis, all 68 included studies were considered as "high quality," and the average NOS score was 8.26 for all studies (Table S2).

Air pollution and stroke hospital admission
A total of 29 studies were performed to assess the association for air pollution and stroke hospital admission, and the results were inconsistent. Most studies showed a positive correlation between exposure to air pollution and the risk of hospital admission for stroke. In metaanalysis, we enrolled 13 studies on PM 2.5 , 11 studies on PM 10 and NO 2 , 10 studies on SO 2 and O 3 , and 6 studies on CO with stroke hospital admission and suggested an increased stroke hospital admission risk after air pollution exposure. The pooled odds ratio (OR) of stroke with a 10 μg/m 3 increase in PM 2.5 , PM 10 , SO 2 , NO 2 , CO, and O 3 was 1.008 (95% CI 1.005, 1.011), 1.004 (95% CI 1.001, 1.006), 1.013 (95% CI 1.007, 1.020), 1.023 (95% CI 1.015, 1.030), 1.000 (95% CI 1.000, 1.001), and 1.002 (95% CI 1.000, 1.003), respectively (Table 2, Figure S1-S6). Heterogeneity among studies was significant (I 2 ≥ 50%, p < 0.001).

Air pollution and stroke incidence
Twenty-three studies have investigated the association of air pollution on stroke incidence (Table 1). For metaanalysis, we extracted 18 studies on PM 2.5 , 13 studies on PM 10 , 10 studies on O 3 , 7 studies on NO 2 , and 4 studies on SO 2 and CO. Ten of these studies suggested increased risks for stroke incidence for at least one of the investigated pollutants. Meta-analysis showed that exposure to PM 2.5 , SO 2 , and NO 2 was associated with increased risks of stroke incidence, and the pooled HR with a 10 μg/m 3 increase was 1.048 (95% CI 1.020, 1.076), 1.002 (95% CI 1.000, 1.003), 1.002 (95% CI 1.000, 1.003), respectively. However, no significant differences were found in associations of PM 10 , CO, O 3 , and stroke incidence (Table 3, Figure S7-S12).

Publication bias and sensitivity analysis
Publication bias of studies on PM 10 exposure and stroke hospital admission may exist, since p values of Begg's      Figure S19-34). Sensitivity analysis showed that the relation of exposure to CO and stroke hospital admission might be influenced by Tian et al.'s study [10]. And the association between exposure to NO 2 and stroke incidence may be influenced by Dong et al.'s study [1]. The pooled OR of exposure to air pollution and stroke mortality might be influenced by some studies (PM 2.5 : Wang et al.'s study [33]; O 3 : Yin et al.'s study [37]). We recalculated the pooled OR/HR and 95% CI after removing those studies (Table S3). Due to limited studies after excluding those studies, the pooled estimated effects of SO 2 and stroke incidence and O 3 and stroke mortality were not recalculated. Other sensitivity analyses indicated that excluding each individual study did not change the results, suggesting the results of the meta-analysis were stable (Table S4, Figure S35-50). Sensitivity analyses by exposure period found that the pooled effect estimates were not changed significantly after excluding the long-term (cohort) studies (Table S5). Subgroup analysis suggested that both short-term and long-term exposure to air pollution would increase the risk of stroke incidence (PM 2.5 , PM 10 , and NO 2 ) and mortality (NO 2 ) (Table S6).

Discussion
We conducted a systematic review and meta-analysis of 68 epidemiological studies and performed a comprehensive evaluation on exposure ambient air pollution and stroke, which were conducted from more than 23 million participants. Most studies suggested that exposure to a higher level of air pollution was associated with increased stroke risk. Meta-analysis showed that exposures to air pollutants were associated with increased risk of stroke hospital admission (PM 2.5 , PM 10 , SO 2 , NO 2, CO,   and O 3 ), incidence (PM 2.5 , SO 2 , and NO 2 ), and mortality (PM 2.5 , PM 10 , SO 2 , and NO 2 ). Although the high heterogeneity may reduce the credibility of the pooled evidence to some extent, the large number of studies included and the consistency of the results indicated that our conclusions were credible to some extent. The positive associations between exposure to PM 2.5 , PM 10 , SO 2 , NO 2 , CO, and O 3, and stroke hospital admission were observed in other meta-analysis. Yang et al. meta-analyzed 34 case-crossover and time series studies and reported significant associations for PM 10 (per 10 μg/m 3 increment: RR = 1.007, 95% CI 1.001, 1.013) and O 3 (per 10 ppb increment: RR = 1.036, 95% CI 1.016, 1.056), but non-significant association for PM 2.5 , SO 2 , NO 2 , and CO [15]. The meta-analysis performed by Yang et al. was not consistent with our current study completely, which might be caused by the different number of the included studies. To our knowledge, more than 16 studies have been published after 2014, and studies included in Yang et al.'s study were mainly conducted in Europe and North America. Data from more recent studies, especially low-and middle-income countries were not considered. Moreover, many studies conducted from the multi-city level and with large sample sizes have been published in recent years, which were more likely to find a significant association between air pollution and stroke hospital admission. For example, Tian et al. performed a time-series of more than 2 million hospital admissions for ischemic stroke in 172 cities in China and suggested that elevated incidence of ischemic stroke hospital admissions was associated with exposure to higher level of PM 2.5 (RR = 1.003, 95% CI 1.002, 1.005, per 10 μg/m 3 increment), SO 2 (RR = 1.013, 95% CI 1.011, 1.017, per 10 μg/m 3 increment), and NO 2 (RR = 1.018, 95% CI 1.015, 1.022, per 10 μg/m 3 increment) [24].
Three meta-analyses were conducted to examine the association between exposure to particulate matter (PM 2.5 and PM 10 ) and stroke incidence, whereas no meta-analysis of gas air pollutants was published before the current study. Li et al. performed a meta-analysis to explore the association between PM 10 and stroke incidence in time-series studies and case-crossover studies. These studies indicated that PM 10 was not associated with stroke incidence in the time-series design (HR = 1.002, 95% CI 0.999, 1.005, per 10 μg/m 3 increment), but significantly associated in case-crossover studies (HR = 1.028, 95% CI 1.001, 1.057, per 10 μg/m 3 increment). Meanwhile, PM 2.5 exposure was related to an increased risk of stroke incidence in time-series design (HR = 1.006, 95% CI 1.002, 1.010, per 10 μg/m 3 increment), but no significant association in case-crossover studies (HR = 1.016, 95% CI 0.937, 1.097, per 10 μg/m 3 increment) [16]. Only 12 studies published before 2010 were included in Li et al.'s study. We updated the literature search up to 2020, which generated more than 10 studies. Moreover, Li et al. separately analyzed the data from time-series and case-crossover studies, which would reduce the number of studies calculated the pooled estimates. These might explain the inconsistency in between our study and Li et al.'s study. Yu et al. updated the literature search before 2012 and identified 19 studies [19]. Yu et al. found that exposure to PM 10 was associated with an increased risk of stroke incidence (HR = 1.004, 95% CI 1.001, 1.008, per 10 μg/m 3 increment), but exposure to PM 2.5 was not significantly associated with stroke incidence (HR = 0.999, 95% CI 0.994, 1.003, per 10 μg/m 3 increment) [19]. The results of these published meta-analyses were not exactly the same as our study, which might be due to more than 15 studies published after Yu et al.'s study. Moreover, we conducted a metaanalysis of gas air pollutants and stroke incidence and found that exposure to a higher level of SO 2 and NO 2 was associated with higher risk of stroke incidence, which may fill the gap of meta-analysis of gas air pollutants and stroke incidence. We also found that compared to short-term exposure, long-term exposure to air pollution may be associated with a higher risk of stroke incidence (PM 2.5 , PM 10 , and NO 2 ), which may be explained by different pathophysiological pathways.
Studies investigating the association between exposure to air pollution and stroke mortality have been partly analyzed in two meta-analysis [15,17]. Yang et al. evaluated the association between all 6 pollutants and suggested that stroke mortality increased 1.34% (95% CI 0.27, 2.42) per 10 μg/m 3 increase in PM 2.5 , 0.65% (95% CI 0.54, 0.77) per 10 μg/m 3 increase in PM 10 , 2.45% (95% CI 1.83, 3.07) per 10 parts per billion (ppb) increase in SO 2 , 7.78% (95% CI 4.49, 11.60) per 1 ppm increase in CO, and 1.50% (95% CI 0.37, 2.63) per 10 ppb increase in NO 2 , respectively [15]. Consistent with Yang et al.'s study, our meta-analysis also indicated that exposure to a higher level of PM 2.5 , PM 10 , SO 2 , and NO 2  [1,9,15,82]. Vascular function injury may be central to mechanisms for air pollutionrelated stroke, which could lead to raised level of blood pressure and plasma viscosity [26]. It has been showed that exposure to air pollution was associated with increased thrombosis and vascular endothelial dysfunction by provoking oxidative stress and releasing systemic inflammatory cytokines [83]. Moreover, evidence also suggested that exposure to air pollution can lead to dysfunction of the autonomic system, which has been found as the major pathway that could result in air pollution-related adverse cardiovascular outcomes, such as stroke [84]. In addition, stroke status may aggravate the susceptibility of population to air pollution and increase the adverse cardiovascular effects of air pollution circularly [62].
A major strength of our meta-analysis is that our systematic review and meta-analysis covered six main air pollutants (PM 2.5 , PM 10, NO 2 , SO 2, CO, O 3 ) and a rich set of stroke outcomes (hospital admission, incidence, and mortality), which may be difficult to obtain from individual studies or isolated reviews or meta-analyses. However, some limitations should be acknowledged. Firstly, high heterogeneity existed in some meta-analysis, which may be due to different study designs, difference in exposure assessment method and population demographics, and the varied covariable adjustment strategies in different studies. Secondly, our study failed to perform the association between different subtypes of stroke (ischemic stroke, hemorrhagic stroke) and air pollution exposure separately because most included studies (48 out of 68 articles) did not report subtypes of stroke or results of ischemic stroke and hemorrhagic stroke specifically. Finally, the correlation between different air pollutants was not examined in our study because different air pollutants were controlled in different studies, and the results of those studies could not be pooled directly.

Conclusion
Our study demonstrated that exposure to air pollution was positively associated with an increased risk of stroke hospital admission (PM 2.5 , PM 10 , SO 2 , NO 2 , CO, and O 3 ), incidence (PM 2.5 , SO 2 , and NO 2 ), and mortality (PM 2.5 , PM 10 , SO 2 , and NO 2 ). Given the great global burden of stroke and air pollution, our findings could provide some scientific evidence to accurate prevention and treatment of stroke and air pollution exposure.

Additional file 1.
Abbreviations PM 2.5 : Particulate matter with aerodynamic diameter less than 2.5 μm; PM 10 : Particulate matter with aerodynamic diameter less than 10 μm;