Acute effects of air pollution on ischaemic stroke onset and deaths: a time-series study in Changzhou, China

Abstract

Objective

To investigate the acute effect of air pollutants on ischaemic stroke (IS) and IS-related death.

Setting

Five urban districts in Changzhou, China, between 9 January 2015 and 31 December 2016.

Participants

A total of 32 840 IS cases and 4028 IS deaths were enrolled.

Main outcome measures

A time-series design, generalised additive model and multivariable regression model were used to examine the percentage change (95% CI) in daily IS counts and deaths with an IQR increase in air pollutant levels for different single or multiple lag days in single-pollutant and two-pollutant models.

Results

Daily IS counts increased 0.208% (95% CI 0.036% to 0.381%) with an IQR increment in the levels of nitrogen dioxide (NO₂). The estimated risk of NO₂ was more robust in males and in the cold season. For daily IS counts, the estimated effects of NO₂ and sulfur dioxide (SO₂) were more significant when adjusted for particulate matter with aerodynamic diameters <2.5 µm (PM_2.5) and PM₁₀. An IQR increment in the concentration of PM₁₀, SO₂ and NO₂ significantly increased IS deaths with 6 days of cumulative effects (0.268%, 95% CI 0.007% to 1.528%; 0.34%, 0.088% to 0.592%; and 0.263%, 0.004% to 0.522%, respectively). Young individuals (<65 years old) had a higher IS mortality risk for PM_2.5, PM₁₀, NO₂ and CO. For IS death, the effect estimates of SO₂ in the elderly, females and the cold season were more pronounced; statistical significance was also identified for SO₂ when adjusted for carbon monoxide (CO).

Conclusions

This study suggested that short-term exposure to ambient NO₂ was associated with increased IS risk. In addition, SO₂ was associated with increased IS onset and death.

Strengths and limitations of this study.

• This is the first comprehensive research in China that explores the short-term effects of air pollutants on the morbidity and mortality of stroke concurrently.

• The statistical power to detect the association was sufficient because of the substantial numbers of ischaemic stroke (IS) cases and IS deaths recorded in our work.

• This time-series study evaluated the association of modifiers between air pollution and daily IS counts or deaths which may provide clues for the primary prevention of IS and IS-related death.

• The exposure measurement of this study was performed at the community level, and we could not precisely quantify the exposure of individuals to air pollutants using the fixed-site monitoring data; this may lead to exposure measurement error.

• This study considers a single city, and the results should be extrapolated with caution.

Introduction

Air pollution is considered a primary risk factor for a mounting total of pollution-related diseases and deaths.¹ Moreover, as a modifiable risk factor, air pollution accounts for more than 7 million deaths annually.² Studies have extensively explored the association between air pollution and the risk of respiratory, reproductive and cardiovascular diseases.^3–5 Epidemiological studies have suggested that exposure to air pollutants on hazy days may increase the risk of cardiovascular diseases.^6–8 For instance, a study conducted in Stockholm suggested that short-term exposure to particulate matter with aerodynamic diameters <2.5 µm (PM_2.5) was associated with out-of-hospital cardiac arrest. Increments in PM_2.5, PM₁₀, nitrogen dioxide (NO₂), sulfur dioxide (SO₂) and carbon monoxide (CO) were associated with higher risk of heart failure hospitalisations and death. A study conducted in Stockholm suggested that short-term exposure to PM_2.5 was associated with out-of-hospital cardiac arrest. A meta-analysis indicated that long-term exposure to PM_2.5 was related to a dramatic increase in cardiovascular mortality.⁹

Stroke is a persistent neurological deficit caused by cerebrovascular damage and includes ischaemic stroke (IS) and haemorrhagic stroke.^10–12 Stroke is one of the leading causes of death and disability-adjusted life years worldwide.^{13 14} In 2010, the estimated number of individuals who suffered from stroke was 50 million, and among them, approximately 17 million cases were newly identified. China bears an enormous stroke burden in the world, having the highest stroke prevalence, incidence and mortality among low-income and middle-income countries worldwide.¹⁵ As the major subtype of stroke in China, IS accounts for 43%–79% of all strokes and has recently prompted growing concerns.¹⁶ Given the tremendous disease burden of stroke, the identification and prevention of risk factors are critical and effective public health strategies.¹⁷ However, the effect of air pollutants remains a matter of debate which might be, at least in part, explained by the heterogeneity of different stroke subtypes.^{18 19} Thus, there is a pressing need to systematically explore the association between air pollutants and specific subtypes of stroke-related incidence and mortality.

Changzhou stands on the southern bank of the Yangtze River. The city is part of the Suzhou-Wuxi-Changzhou metropolitan area, and by the year 2010, there were approximately 4 million residents in five urban districts (Tianning district, Wujin district, Jintan district, Xinbei district and Liyang city).²⁰ Moreover, because of the rapid industrialisation and economic development of the past years, Changzhou is now experiencing severe air pollution. The concentrations of several air pollutants (such as PM_2.5, PM₁₀, SO₂ and NO₂) were equal or higher than the national ambient air quality in the recent years.^{21 22} Therefore, Changzhou was chosen as a suitable place to study the effect of air pollution on the incidence and mortality of IS. In this research, a comprehensive time-series study was conducted in Changzhou to examine the association between main air pollutants and daily IS counts and deaths. To explore the modification effect, the association of modifiers between air pollution and daily IS counts and deaths was also evaluated. This work will help provide insights into the relationship between stroke and air pollutants and have implications for the prevention of IS onset and IS-related deaths.

Materials and methods

Data collection

Air pollution data, including daily 24 hours average levels of PM_2.5, PM₁₀, NO₂, SO₂ and CO between 9 January 2015 and 31 December 2016, were obtained from 10 air quality monitoring stations run by the Changzhou Environmental Monitoring Center. The daily mean levels for air pollutants were averaged from all the stations, and approximately all the patients whose cases were recorded in this study resided less than 40 km from the nearest monitoring station. Thus, the monitoring data could be used as an appropriate proxy for personal exposure.^23–25 Daily meteorological data on temperature (°C) and relative humidity (%) were collected from the Changzhou Meteorological Bureau.

Daily IS and IS mortality data from 9 January 2015 to 31 December 2016 were obtained from the database of the Changzhou Center for Disease Control and Prevention. This database belonged to the government-controlled network reporting system for chronic diseases which was established in Changzhou in 2012. Cardiovascular physicians were responsible for the diagnosis of IS according to the cerebrovascular 10th revision of the International statistical classification of diseases (ICD-10) codes for IS (I63). Demographic data were collected from the Changzhou Municipal Bureau of Statistics. Daily IS counts and deaths were also stratified into groups by sex (male and female) and age group (<65 and ≥65 years) to explore the effect modification of individual characteristics.

Statistical analysis

Daily IS counts, daily IS death counts, air pollutant levels and weather data were linked by date for the subsequent time-series study which was used to analyse the effects of each air pollutant on IS counts and deaths. After evaluating the distribution patterns of daily IS counts and deaths, Poisson distributions were identified. Poisson regression was used in a generalised additive model to analyse the data. A multivariable regression model was used to detect the air pollutant-related associations and control for the potential confounding factors, such as calendar time, day of week (DOW), daily average relative humidity and temperature. The five df was used to adjust for all the confounding factors mentioned above except for DOW (df=7). Furthermore, to explore the potential modifications, data were stratified by sex, age (<65 as young individuals and ≥65 as elderly individuals) and season (warm season as 1 May to 31 October and cold season as 1 November to 30 April). Associations between air pollutants and IS and IS-related deaths were separately examined according to the following equation:

$${\displaystyle \begin{array}{ll}Log[E(T/D_n)]& =intercept+{\beta }_{1}AirPollutio{n}_{n-i}+{\beta }_{2}DOW\\ & +ps(calendartime,df=7+ps(Tem{p}_{n-i},df=5)\\ & +ps(RelativeHumidit{y}_{n-i},df=5)\end{array}}$$

E(T/D_n) represents the estimated numbers of IS onset or death on day n; AirPollution_n-i represents the average level of different air pollutants on day n and i is the day lag; β is the vector of the coefficients; DOW is the day of week; ps () represents a penalised spline function; Temp_n-i and Relative Humidity_n-i represent average temperature and relative humidity on day n and i is the day lag.

In addition, for both daily IS counts and deaths, single-pollutant models were fitted with different single-lag days (lag 0, lag 1, lag 3 and lag 5) to detect the effects of air pollution on IS. As underestimation may occur in single-day models, multiple-day lag models (lag 0–1, lag 0–3 and lag 0–5) were also applied in this study. The smoothing function was used in the analysis of the exposure–response relationship between the log-relative risk of daily IS counts or deaths and air pollutant concentrations using a five df in a single-pollutant model.

Statistical software R V.3.2.3 (R Foundation for Statistical Computing, Austria) was used for data analysis and result output. The baseline data was presented as the mean±SD for continuous variables. All tests were two-sided, and p<0.05 was considered as statistically significant.

Public and patient involvement

Public and patient involvement (PPI) was central and integral in the present study. The PPI group consisted of three male and four female representative patients who were recruited and chaired by HD. The members of the existing group were involved throughout all stages of this research to make contributions. At the outset, they helped researchers with choosing topics and designing the study by providing constructive suggestions. Furthermore, they took part in the conference presentations and group discussions to help filter and interpret data collected from the database mentioned above. They also contributed to discussions about the implications of the results by drawing on their experiences. Findings were disseminated to all PPI members at the end of the study.

Results

Descriptive analysis

During this study period, there were 32 840 IS cases and 4028 IS deaths recorded. For IS, on average, 45.4 cases were identified each day (online supplementary table S1). Of these, 52.2% were males and 81.3% were the elderly (≥65). In addition, for IS-related death, there were 5.6 deaths recorded on average each day (online supplementary table S1). Of these, 48.2% were males and 96.4% were the elderly.

The daily average meteorological data were 17.0°C for temperature and 75% for relative humidity. The daily average levels were 51.8 µg/m³ for PM_2.5, 85.3 µg/m³ for PM₁₀, 22.3 µg/m³ for SO₂, 38.7 µg/m³ for NO₂ and 1 mg/ m³ for CO (table 1). According to WHO, air quality guidelines (25 µg/m³ for PM_2.5, 50 µg/m³ for PM₁₀ and 20 µg/m³ for SO₂), the corresponding average levels of PM_2.5, PM₁₀ and SO₂ in Changzhou were far beyond the WHO standard.

Table 1.

The means, medians and IQRs for daily air pollutants and weather conditions variables in Changzhou, 2015–2016

Variables	Mean±SD	Min	25th	Median	75th	Max	IQR
Weather Condition
Temperature (°C)	17.1±8.9	−6.6	9.4	18.3	24.1	34.6	5.8
Humidity (%)	75.3±14.2	35.0	66.0	76.0	86.0	100.0	10.0
Air Pollutions
PM2.5 (μg/m3)	51.8±30.6	6.8	29.4	44.4	64.4	181.1	20.0
PM10 (μg/m3)	85.3±43.3	10.5	51.5	76.6	108.6	289.1	32.0
NO2 (μg/m3)	38.7±16	10.9	26.5	35.8	47.8	117.3	12.0
SO2 (μg/m3)	22.3±10	6.9	14.7	20.2	27.0	67.8	6.9
CO (mg/m3)	1±0.3	0.4	0.8	1.0	1.2	2.6	0.2

CO, carbon monoxide; NO₂, nitrogen dioxide; PM_2.5, particulate matter <2.5 µm in diameter; PM₁₀, particulate matter <10 µm in diameter; SO₂, sulfur dioxide.

Estimated effects of air pollutants

Figure 1 summarises the percentage change (95% CI) in daily IS counts and deaths with an IQR increase in air pollutant levels for different single-lag or multiple-lag days in single-pollutant models. For daily IS counts, a statistically positive association was observed only for an IQR increment of NO₂ levels, and this association was more significant for multi-day exposure than for single-day exposure. The highest association was observed for lag 0–5 (0.208%, 0.036% to 0.381%) and was used in the subsequent study. For daily IS deaths, statistically positive associations were observed for the different IQRs increments of PM₁₀, SO₂ and NO₂. All associations were the most significant at lag 5 (0.268%, 0.007% to 0.528% for PM₁₀; 0.34%, 0.088% to 0.592% for SO₂; and 0.263%, 0.004% to 0.522% for NO₂) than at other days and thus a 5-day lag model was used in the subsequent analysis.

Figure 1.

The percentage change in daily IS counts and daily IS deaths with an IQR increase in air pollutant levels on different lag days in single-pollutant models in Changzhou, 2015–2016. The data are expressed as the mean with a 95% CI. CO, carbon monoxide; IS, ischaemic stroke; NO₂, nitrogen dioxide; PM_2.5, particulate matter <2.5 µm in diameter; PM₁₀, particulate matter <10 µm in diameter; SO₂, sulfur dioxide.

Concentration–response relationships of air pollutants with daily IS counts or deaths are presented in figure 2. For IS counts, the curve of NO₂ was linear positive and flat at higher concentrations. For daily IS deaths, the concentration–response curves of PM₁₀ and SO₂ were similar, that is, flat at low levels and dramatically linear positive at high levels, whereas the curve for NO₂ suggest a linear rise in daily IS deaths.

Figure 2.

The concentration–response relationships of daily IS counts, daily IS deaths with air pollutants on different lag days (lag 0–5 for daily IS counts and lag 5 for daily IS deaths) in single-pollutant models in Changzhou, 2015–2016. IS, ischaemic stroke; NO ₂, nitrogen dioxide; PM₁₀, particulate matter <10 µm in diameter; RR, risk ratio; SO₂, sulfur dioxide.

The effect of NO₂ on the increase of daily IS counts was more pronounced among males than among females, and the effects of NO₂ and SO₂ on daily IS counts were stronger in the cold season than in the warm season. No significant association was observed when modifying by age (table 2). For daily IS deaths, the estimated effect of PM₁₀ was more pronounced among males, whereas the effects of NO₂ and SO₂ were more robust among females. Moreover, PM_2.5, PM₁₀, NO₂ and CO were found to have greater deleterious effects among young individuals, and SO₂ showed greater risk in the elderly. In addition, when modifying by seasons, statistical significance was noted only for SO₂ in the cold season (table 2).

Table 2.

The percentage change in daily IS counts and deaths with an IQR increase in air pollutant levels modifying by age, sex and seasons in Changzhou, 2015–2016

Variable	Total	Age		Sex		Seasons
		≤65	>65	Female	Male	Warm	Cold
Daily IS counts*
PM2.5	−0.063 (−0.215 to 0.088)	−0.246 (−0.568 to 0.076)	−0.042 (−0.21 to 0.127)	−0.097 (−0.311 to 0.119)	−0.08 (−0.289 to 0.128)	−0.149 (−0.525 to 0.228)	−0.098 (−0.267 to 0.071)
PM10	−0.053 (−0.22 to 0.114)	−0.219 (−0.572 to 0.136)	−0.004 (−0.191 to 0.184)	−0.086 (−0.324 to 0.153)	−0.047 (−0.279 to 0.185)	−0.057 (−0.41 to 0.297)	−0.088 (−0.282 to 0.107)
NO2	0.208 (0.036 to 0.381)†	0.258 (−0.075 to 0.592)	0.166 (−0.025 to 0.357)	0.168 (−0.055 to 0.391)	0.238 (0.002 to 0.475)†	−0.136 (−0.481 to 0.21)	0.382 (0.169 to 0.596)†
SO2	0.125 (−0.034 to 0.284)	−0.044 (−0.364 to 0.277)	0.113 (−0.062 to 0.289)	0.15 (−0.068 to 0.369)	0.056 (−0.162 to 0.275)	-0.309 (−0.651 to 0.033)	0.298 (0.099 to 0.498)†
CO	−0.078 (−0.22 to 0.064)	−0.213 (−0.511 to 0.086)	−0.061 (−0.222 to 0.1)	−0.151 (−0.355 to 0.053)	−0.007 (−0.206 to 0.192)	−0.175 (−0.432 to 0.083)	−0.053 (−0.237 to 0.131)
Daily IS deaths‡
PM2.5	0.228 (−0.005 to 0.462)	1.144 (0.329 to 1.965)†	0.152 (−0.091 to 0.395)	0.124 (−0.19 to 0.439)	0.29 (−0.043 to 0.625)	−0.199 (−0.77 to 0.375)	0.184 (−0.088 to 0.456)
PM10	0.268 (0.007 to 0.528)†	1.52 (0.656 to 2.391)†	0.152 (−0.12 to 0.424)	0.093 (−0.261 to 0.448)	0.404 (0.03 to 0.78)†	-0.012 (−0.53 to 0.509)	0.247 (−0.075 to 0.57)
NO2	0.263 (0.004 to 0.522)†	0.894 (0.058 to 1.736)†	0.195 (−0.073 to 0.463)	0.37 (0.027 to 0.715)†	0.132 (−0.247 to 0.512)	0.275 (−0.222 to 0.775)	0.198 (−0.13 to 0.528)
SO2	0.34 (0.088 to 0.592)†	0.67 (−0.175 to 1.523)	0.307 (0.045 to 0.57)†	0.434 (0.096 to 0.772)†	0.215 (−0.15 to 0.58)	0.017 (−0.493 to 0.531)	0.388 (0.081 to 0.696)†
CO	0.175 (−0.042 to 0.392)	0.9 (0.166 to 1.639)†	0.108 (−0.118 to 0.335)	0.189 (−0.11 to 0.49)	0.154 (−0.162 to 0.471)	0.02 (−0.401 to 0.443)	0.169 (−0.104 to 0.443)

The data are expressed as the mean with 95% CI.

*The multivariable regression model fits the pollutants for daily IS count data at lag 0–5 and adjusts for calendar time (df=7), DOW (df=5), daily average humidity (df=5) and average temperature (df=5).

†Statistically positive association between air pollutants and daily IS counts or deaths.

‡The multivariable regression model fits the pollutants for daily IS death data at lag 5 and adjusts for calendar time (df=7), DOW (df=5), daily average humidity (df=5) and average temperature (df=5).

CO, carbon monoxide; DOW, day of week; IS, ischaemic stroke; NO₂, nitrogen dioxide; PM_2.5, particulate matter <2.5 µm in diameter; PM₁₀, particulate matter <10 µm in diameter; SO₂, sulfur dioxide.

Table 3 shows the percentage increases in daily IS counts and daily IS deaths with each IQR increment in air pollutant levels in the two-pollutant models. For daily IS counts, the estimated effects of NO₂ and SO₂ were more serious when adjusted for PM_2.5 and PM₁₀. For daily IS deaths, statistical significance was identified for SO₂ when adjusted for CO.

Table 3.

The per cent increases in daily IS counts and deaths with each IQR increment in air pollutant levels in two-pollutant models in Changzhou, 2015–2016

Variables	PM2.5	PM10	NO2	SO2	CO
Daily IS counts*
PM2.5	−0.063 (−0.215 to 0.088)	0.25 (−0.314 to 0.817)	0.497 (0.259 to 0.735)†	0.305 (0.083 to 0.527) †	−0.059 (−0.274 to 0.157)
PM10	−0.423 (−0.881 to 0.038)	−0.053 (−0.22 to 0.114)	0.539 (0.29 to 0.788)†	0.262 (0.034 to 0.491) †	−0.065 (−0.293 to 0.163)
NO2	−0.346 (−0.558 to 0.134)	−0.438 (−0.691 to 0.185)	0.208 (0.036 to 0.381)†	−0.007 (−0.267 to 0.253)	−0.341 (−0.556 to 0.124)
SO2	−0.24 (−0.46 to 0.019)	−0.244 (−0.49 to 0.003)	0.22 (−0.067 to 0.507)	0.125 (−0.034 to 0.284)	−0.166 (−0.356 to 0.023)
CO	−0.089 (−0.314 to 0.137)	0.029 (−0.223 to 0.281)	0.45 (0.215 to 0.685)†	0.172 (−0.016 to 0.361)	−0.078 (−0.22 to 0.064)
Daily IS deaths‡
PM2.5	0.228 (−0.005 to 0.462)	0.436 (−0.36 to 1.238)	0.22 (−0.127 to 0.567)	0.28 (−0.071 to 0.632)	0.065 (−0.314 to 0.446)
PM10	0.211 (−0.463 to 0.89)	0.268 (0.007 to 0.528)†	0.218 (−0.135 to 0.573)	0.295 (−0.055 to 0.646)	0.137 (−0.23 to 0.504)
NO2	0.106 (−0.201 to 0.414)	0.149 (−0.21 to 0.509)	0.263 (0.004 to 0.522)†	0.351 (−0.096 to 0.8)	0.057 (−0.269 to 0.384)
SO2	0.066 (−0.25 to 0.383)	0.126 (−0.242 to 0.496)	0.066 (−0.387 to 0.521)	0.34 (0.088 to 0.592)†	0.035 (−0.281 to 0.351)
CO	0.186 (−0.199 to 0.571)	0.233 (−0.187 to 0.656)	0.23 (−0.14 to 0.601)	0.346 (0.002 to 0.691)†	0.175 (−0.042 to 0.392)

The data are expressed as the mean with 95% CI.

*The multivariable regression model fits the pollutants for daily IS counts data at lag 0-5 and adjusts for the other pollutants, calendar time (df=7), DOW (df=5), daily average humidity (df=5) and average temperature (df=5).

†Statistically positive association between air pollutants and daily IS counts or deaths.

‡The multivariable regression model fits the pollutants for daily IS deaths data at lag 5 and adjusts for the other pollutants, calendar time (df=7), DOW (df=5), daily average humidity (df=5) and average temperature (df=5).

CO, carbon monoxide; DOW, day of week; IS, ischaemic stroke; NO₂, nitrogen dioxide; PM_2.5, particulate matter <2.5 µm in diameter; PM₁₀, particulate matter <10 µm in diameter; SO₂, sulfur dioxide.

Discussion

In this population-based time-series study, a total of 32 840 IS cases and 4028 IS deaths were identified in Changzhou from 2015 to 2016. A significant association of NO₂ with daily IS counts was observed even when adjusted by PM_2.5 and PM₁₀. The estimated IS risk of NO₂ was more robust in the elderly and in the cold season. In addition, PM₁₀, SO₂ and NO₂ were positively associated with daily IS deaths. Young individuals had a higher IS mortality risk for PM_2.5, PM₁₀, NO₂ and CO. In addition, the IS mortality risk of SO₂ was more robust in the elderly, females and cold season; statistical significance was also identified for SO₂ when adjusted for CO. To the best of our knowledge, in China, this is the first comprehensive study to explore the acute effect of air pollutants on the morbidity and mortality of stroke concurrently.

PM₁₀ can cause endothelial dysfunction, inflammatory response and neurofunctional impairment and, subsequently, induce ischaemia-like injuries in the brain.²⁶ In the current study, an IQR increment in PM₁₀ concentrations was associated with a 0.268% increment in daily IS deaths which was consistent with previous investigations.^{27 28} Moreover, the function curve for PM₁₀/daily IS deaths was flat at <150 µg/m³ and became dramatically sharp at high levels. A previous multicity time-series study indicated an acute, deleterious and significant effect of PM₁₀ on IS mortality, and the estimated rate ratio was 1.016.²⁹ Long-term exposure to PM₁₀ was positively associated with IS-relate death and the estimated relative risk was 1.37.³⁰ Moreover, in this study, a greater estimated IS mortality risk was observed among males and young individuals. It is plausible that males and young people spent more time outdoors and were exposed to higher levels of PM₁₀ than females and the elderly.³¹

NO₂, a representative pollutant of vehicle exhaust, has been found to induce excitotoxicity and endothelial and inflammatory responses and to damage synaptic plasticity in the brain.^32–34 In China, a multicity case-crossover study demonstrated that an IQR increase (25.4 µg/m³) in NO₂ corresponded to a 2.6% increment in IS emissions. Stroke mortality was also positively associated with ambient NO₂ levels.³⁵ In the present study, we indicated that IQR increments in NO₂ (12 µg/m³) were associated with 0.208% and 0.263% increases in daily IS counts and IS deaths, respectively. An approximately linear concentration– response relationship was identified for NO₂ and daily IS counts or daily IS deaths which was consistent with the previous studies.^{36 37} The positive association for NO₂ and daily IS counts could be observed even below the National Ambient Air Quality Standards of China (80 µg/m³). For daily IS counts, the effects of NO₂ were statistically positive and stronger when adjusted for other pollutants, except for SO₂ (no statistical association was detected, but a stronger association was observed), suggesting that the adverse effect of NO₂ on IS onset was stable and that NO₂ contributed most to the increased risk of IS. Moreover, the effects of NO₂ were more serious in the cold season and in males. For IS deaths, young females were found to experience higher risk. An underlying reason may be that the exposure levels of NO₂ were much higher at these conditions which was similar to that of PM₁₀. The risk factors for stroke, such as hypertension, atrial fibrillation and diabetes mellitus are more frequent, and the estimated effects are stronger among females.³⁸ Females are considered to be more sensitive to stroke onset and deaths. Thus, our findings regarding the higher risk of IS deaths observed among females were reasonable.

SO₂ was considered a strong ambient, gaseous pollutant and irritant that could elevate the expression of vasoregulatory pathways and proinflammatory enzymes such as ET-1, iNOS, COX-2, ICAM-1 and TNF-α.³⁹ In addition, exposure to it was demonstrated to contribute to the development and progression of IS.¹⁸ In the present study, SO₂ was significantly associated with daily IS counts when adjusted by PM_2.5 or PM₁₀. In addition, exposure to SO₂ also had statistical associations with daily IS deaths (the percentage increase was 0.34%) which agreed with previous studies.^{19 40} The function curve for SO₂ and daily IS deaths was flat at <35 µg/m³ and became dramatically sharp at high levels, suggesting that acute and robust increments of SO₂ might exacerbate IS deaths. Furthermore, the associations were more pronounced among females, suggesting that females are more susceptible to IS-related deaths when exposed to gaseous pollutants (NO₂ and SO₂). In addition, SO₂ exhibited a higher risk of IS deaths in the elderly.⁴¹ Thus, it can be deduced that the elderly are more likely to have pre-existing respiratory or cardiovascular burdens, and these vulnerable conditions could influence the effect of SO₂ on stroke.⁴² We also found that exposure to SO₂ had more serious effects on both IS onset and deaths in the cold season than in the warm season.

Examining the deleterious effect of air pollutants on the incidence and outcomes of stroke is of significance, and previous studies mainly have focused on one aspect of IS in China. This study is one of the first to comprehensively explore the relationship between air pollutant levels and morbidity or mortality of IS. As numerous individuals were included to explore the associations, the results of this work strongly support the conclusion and may help provide insights into the relationship between air pollutants and IS. Furthermore, due to the severe air pollution and enormous stroke burden that Changzhou is now experiencing, the findings of this study may provide clues for the primary prevention of the onset or death related to IS, especially for sensitive populations. However, limitations should be noted in this study. Because individuals were identified retrospectively and not all hospitals were included, diagnostic errors and selection bias may exist. We relied on routine measurements from 10 fixed-site monitoring stations instead of more accurate measurement based on the individuals’ residence and indoor exposure. Thus, we were not able to quantify the exposure of individuals precisely. Moreover, the data regarding IS cases and air pollutant levels were collected from only one city, and it is difficult to extrapolate the results to other areas in China. Further national or multicity studies are required.

In conclusion, our results suggested that short-term exposure to ambient NO₂ was associated with increased IS risk. In addition, SO₂ was associated with increased IS counts and deaths. These findings may have significant public health implications for the prevention of IS and IS-related deaths. Further studies are recommended to validate our research.