Stroke remains one of the leading causes of death among Chinese residents, and air pollution has long been recognized as a significant environmental risk factor. Yet most research on the air pollution–stroke link has focused on heavily polluted or economically developed cities, leaving a critical gap in our understanding of how pollution affects health in regions with relatively clean air. A new study from Guiyang City, the capital of Guizhou Province in Southwest China, now provides the first systematic evidence from a low-pollution karst topography city, and the findings are striking: even when pollutant concentrations fall below national standards, short-term exposure to six major air pollutants is significantly associated with increased stroke risk.
The research, published in Frontiers in Public Health, analyzed daily stroke incidence, air pollutant concentrations, and meteorological conditions for the permanent residents of Guiyang from January 1, 2016, to December 31, 2021. Over the six-year study period, a total of 11,517 stroke cases were reported, averaging 5.25 cases per day. Of these, 6,657 cases (57.80%) occurred in males and 4,860 cases (42.20%) in females, while 7,325 cases (63.60%) occurred in individuals aged 65 years or older and 4,192 cases (36.40%) in those under 65. The overall incidence of stroke showed an upward trend during the study period.
Guiyang presents a unique geographical setting for this kind of investigation. The city features well-developed karst topography, characterized by the coexistence of mountains and valley basins, with an average altitude of 1,000 meters in the study area. This type of landform may lead to temperature inversions and reduced atmospheric ventilation, thereby limiting the dispersion of air pollutants. Compared with plain cities, this distinctive environment may alter the health effects of air pollution. The climate is subtropical, humid, and temperate, with additional features of a plateau monsoon climate, an average annual precipitation of 1,173.9 millimeters, an annual mean temperature of 15.16 degrees Celsius, and an average annual humidity of 80 percent.
To conduct the analysis, the research team collected environmental air pollutant data from twelve monitoring stations established across the urban area, including Tongmuling, Biyunwo, Zhongyuan Village, Xinhua Road, the Municipal Environmental Protection Station, Taici Bridge, Ma’anshan, Hongbianmen, Yanzi Chong, Jianhu Road, Shapo Station, and Yunfeng Station. Daily average concentrations of sulfur dioxide (SO₂), nitrogen dioxide (NO₂), fine particulate matter (PM₂.₅), inhalable particulate matter (PM₁₀), carbon monoxide (CO), and ozone (O₃) were calculated across all valid monitoring stations, with O₃ using an 8-hour average and all other pollutants employing 24-hour averages. Concurrent meteorological data, including daily average temperature, atmospheric pressure, relative humidity, and wind speed, were obtained from the Guiyang Meteorological Bureau.
The statistical approach centered on a time-series generalized additive model (GAM) based on a quasi-Poisson distribution, a framework well suited to modeling daily counts of a low-probability event such as stroke onset relative to the permanent population of roughly six million. The model incorporated smoothing functions to account for nonlinear effects while controlling for confounders such as average temperature, average relative humidity, long-term trends, public holiday effects, and day-of-week effects. Degrees of freedom for time trends (7 per year), temperature (3), and humidity (3) were selected based on the Akaike Information Criterion. Overdispersion was assessed by examining the dispersion parameter in each quasi-Poisson model, and residual autocorrelation was examined using partial autocorrelation function plots. Because the health effects of ambient air pollutants are persistent and delayed, potentially manifesting over several days or weeks following exposure, the team calculated relative risks (RR) and 95% confidence intervals (CI) for single-day lag effects (lag0 through lag7) and cumulative lag effects (lag01 through lag07), with a maximum lag duration of 7 days selected based on existing research.
The results revealed significant positive associations between short-term exposure to all six pollutants and stroke risk. Per 10 micrograms per cubic meter increase in pollutant concentration (CO per 1 milligram per cubic meter), the relative risks at maximum lag were: PM₂.₅ 1.026 (95% CI: 1.005–1.048), PM₁₀ 1.015 (95% CI: 1.001–1.030), SO₂ 1.075 (95% CI: 1.016–1.137), NO₂ 1.057 (95% CI: 1.017–1.098), O₃ 1.013 (95% CI: 1.003–1.023), and CO 1.346 (95% CI: 1.072–1.691). Among these pollutants, CO showed the strongest association, followed by SO₂. In the multi-day cumulative lag effects, PM₂.₅, PM₁₀, SO₂, and NO₂ exhibited the strongest effects at lags ranging from 0 to 4 days, with corresponding RR values of 1.033, 1.019, 1.098, and 1.076, respectively. For O₃, the lagged effect at 0 to 1 days showed the strongest impact on the entire population, with an RR of 1.015. After applying the Benjamini-Hochberg false discovery rate correction for multiple testing, PM₂.₅, PM₁₀, SO₂, NO₂, and CO remained statistically significant, indicating that the associations for the major pollutants are robust to concerns regarding multiple comparisons.
Notably, the daily average concentrations of all six pollutants in Guiyang during 2016–2021 did not exceed the national Grade II concentration limits under the Environmental Air Quality Standards (GB 3095–2012). The daily average concentrations were 29.29 micrograms per cubic meter for PM₂.₅, 50.12 for PM₁₀, 11.29 for SO₂, 21.92 for NO₂, 75.81 for O₃, and 0.63 milligrams per cubic meter for CO. During the study period, all air pollutants except O₃ exhibited higher concentrations in winter and spring and lower concentrations in summer and autumn, while O₃ showed the opposite seasonal pattern. Spearman correlation analysis revealed statistically significant correlations among daily concentrations of the six pollutants, with PM₂.₅ exhibiting the strongest correlation with PM₁₀ (r = 0.948).
Stratified analyses uncovered important patterns of susceptibility. Women were more sensitive to all six pollutants, with CO demonstrating a lag-0 RR as high as 1.548 (95% CI: 1.116–2.147), whereas men showed significant effects only at specific lag times for NO₂ and CO. All six pollutants showed statistically significant effects on female residents. Regarding age, individuals aged 65 years and older exhibited immediate or cumulative delayed effects for the vast majority of pollutants, whereas those under 65 were only affected by short-term impacts of NO₂ and PM₁₀. The RR value for SO₂ exposure among individuals aged 65 and older (1.130, 95% CI: 1.060–1.205) was significantly higher than that for the general population. These findings suggest that women and older adults may be particularly vulnerable to the cerebrovascular effects of air pollution, possibly related to physiological differences such as smaller airway diameters and thinner vascular walls in women, and the natural decline in vascular function, increased incidence of atherosclerosis, and higher prevalence of comorbidities such as hypertension and diabetes among older adults.
To validate the robustness of the findings, the team conducted sensitivity analyses by adjusting the degrees of freedom for time trends, temperature, and humidity within the single-pollutant model, observing minimal changes in the relative risk values for primary pollutants. Additionally, to assess the potential impact of the COVID-19 pandemic, a period-specific sensitivity analysis stratified the study period into pre-pandemic (2016–2019) and pandemic (2020–2021) phases. During the pre-pandemic period, PM₂.₅, PM₁₀, NO₂, and SO₂ were significantly associated with increased stroke risk, with effect estimates generally consistent with those from the full study period. During the pandemic period, PM₂.₅ and NO₂ remained statistically significant, while the associations for PM₁₀, SO₂, CO, and O₃ were attenuated, likely due to reduced statistical power from the shorter observation period. Overall, these findings suggest that the COVID-19 pandemic did not materially alter the main conclusions of the study.
The two-pollutant model analyses provided additional nuance. When PM₂.₅ and NO₂ were adjusted for each other, the estimated effects of these pollutants on stroke incidence in the general population were attenuated and no longer statistically significant, suggesting that their effects are not entirely independent, likely because they share emission sources such as traffic combustion and involve similar biological pathways related to oxidative stress and systemic inflammation. The effect of CO showed a marked attenuation after adjustment for PM₂.₅ and NO₂, with the RR decreasing by more than 20 percent, from a significant 1.346 to a non-significant 1.068. Given that CO had the lowest variance inflation factor across all tested models, this sharp decline suggests that the significant association with CO observed in single-pollutant models likely serves as a proxy for a broader mixture of traffic-related pollutants. Variance inflation factors for all two-pollutant models were below 3, indicating that the observed attenuation was not primarily driven by statistical multicollinearity.
The biological mechanisms underlying these associations are supported by extensive toxicological and epidemiological research. Particulate matter and gaseous pollutants like SO₂ and NO₂ may rapidly induce vascular events by promoting hypercoagulability and triggering acute inflammation. Studies indicate that even low-level PM exposure negatively impacts vascular endothelial function, increases systemic inflammation and platelet activation, and reduces antioxidant enzyme activity. These changes may promote blood coagulation and thrombosis, impair vascular function and blood flow, and lead to elevated blood pressure, ultimately potentially contributing to stroke occurrence. In contrast, O₃, as a potent oxidant, may require prolonged inflammatory cascades to damage the cardiovascular system, explaining its delayed effects. Carbon monoxide directly damages cerebral vascular homeostasis through pathways such as inhibiting platelet aggregation, disrupting neurotransmission, and impairing vasodilation, with its acute exposure exhibiting more pronounced toxic effects.
This study fills a significant research gap regarding the health effects of air pollution at low contamination levels. It aligns with conclusions from studies in non-industrial areas of Poland, confirming that the harmful effects of air pollution on stroke are not confined to highly polluted regions. Even in relatively clean environments like Guiyang City, attention must be paid to the potential impact on public health. The findings are consistent with a growing body of evidence from low-pollution regions worldwide, suggesting that air pollution-induced stroke through mechanisms such as oxidative stress and inflammation may have biological universality across populations. However, the karst topography may lead to frequent temperature inversions and poor atmospheric dispersion conditions, so caution is warranted when extrapolating these findings to plains or regions with markedly different meteorological conditions.
Despite the significant public health implications, several limitations exist. As the study employed a time-series analysis within an ecological framework, the subjects were based on populations rather than individuals, which may lead to ecological fallacy. Ambient monitoring data may not capture intra-urban variability in pollutant concentrations or individual-level exposure patterns influenced by time-activity budgets, indoor air pollution sources such as cooking, heating, and tobacco smoke, and proximity to traffic or industrial sources. Such non-differential exposure misclassification typically biases effect estimates towards the null, potentially leading to an underestimation of the true associations. Additionally, the lack of stratified analysis by disease subtype prevents identification of potential differences in pollutant effects between ischemic and hemorrhagic stroke, as a substantial proportion of cases were recorded under the broad category of cerebrovascular disease without mandatory subtype specification during the study period. Furthermore, potential confounders such as influenza outbreaks and socioeconomic factors were not adjusted for, though the smooth time trend in the model partially captures seasonal influenza patterns.
The findings carry important implications for public health policy. The research team recommends strengthening air quality management and implementing targeted interventions during periods of elevated pollutant concentrations, particularly for high-risk populations such as women and older adults. Allocating healthcare resources appropriately and ensuring that high-risk groups take effective protective measures on days with high pollution levels may help reduce the burden of stroke and enhance public health standards. Future multicenter studies encompassing diverse geographic environments, pollution levels, and climatic conditions should be conducted to further validate and quantify the generalizability of these findings, and future studies incorporating spatial interpolation or satellite-based models could better capture intra-urban exposure contrasts.
Subject of Research: Environmental Health
Article Title: Time-series analysis of the impact of short-term exposure to atmospheric pollutants on stroke risk among residents in Guiyang City, 2016–2021
Article References: Zhang, L., Su, M., Zhai, S., Jiang, M., Chen, X., Liu, R., Chen, G., & Zhang, J. (2026). Time-series analysis of the impact of short-term exposure to atmospheric pollutants on stroke risk among residents in Guiyang City, 2016–2021. Frontiers in Public Health, 14, Article 1777772. https://doi.org/10.3389/fpubh.2026.1777772
Image Credits: AI Generated
DOI: 10.3389/fpubh.2026.1777772
Keywords: air pollution, stroke, PM2.5, time-series analysis, public health, Guiyang, environmental health, Time-series, analysis, impact, short-term, exposure
News Source: Cassandra Pierce. (October 17, 2026). Even Clean Air Cities Carry a Hidden Stroke Risk, Six-Year Study Warns. Scienmag.
