Post-Discharge Ambient Sulfur Dioxide Exposure and Stroke Recurrence After Acute Ischemic Stroke

Article information

J Stroke. 2026;28(3):451-461
Publication date (electronic) : 2026 September 30
doi : https://doi.org/10.5853/jos.2026.03265
1Department of Neurology, Korea University College of Medicine, Korea University Guro Hospital, Seoul, Korea
2Department of Medical IT Engineering, College of Software Convergence, Soonchunhyang University, Asan, Korea
3Department of Neurology, Seoul National University College of Medicine, Seoul National University Bundang Hospital, Seongnam, Korea
4Department of Neurology, Chonnam National University Hospital, Gwangju, Korea
5Department of Neurology, Chungbuk National University Hospital, Chungbuk National University College of Medicine, Cheongju, Korea
6Department of Neurology, Dong-A University Hospital, Busan, Korea
7Department of Neurology, Dongguk University Ilsan Hospital, Goyang, Korea
8Department of Neurology, Uijeongbu Eulji Hospital, Eulji University School of Medicine, Uijeongbu, Korea
9Department of Neurology, Nowon Eulji Medical Center, Eulji University School of Medicine, Seoul, Korea
10Department of Neurology, Eulji University Hospital, Eulji University School of Medicine, Daejeon, Korea
11Department of Neurology, Hallym University Sacred Heart Hospital, Anyang, Korea
12Department of Neurology, Inje University Ilsan Paik Hospital, Goyang, Korea
13Department of Neurology, Jeju National University Hospital, Jeju, Korea
14Department of Neurology, Seoul Medical Center, Seoul, Korea
15Department of Neurology, Ulsan University Hospital, Ulsan, Korea
16Department of Neurology, Yeungnam University Medical Center, Daegu, Korea
17Department of Neurology, Soonchunhyang University Seoul Hospital, Seoul, Korea
18Department of Neurology, Keimyung University Dongsan Medical Center, Daegu, Korea
19Department of Neurology, Chung-Ang University Hospital, Seoul, Korea
20Department of Neurology, Hallym University Chuncheon Sacred Heart Hospital, Chuncheon, Korea
21Department of Biostatistics, Korea University, Seoul, Korea
Correspondence: Keon-Joo Lee Department of Neurology, Korea University College of Medicine, Korea University Guro Hospital, 148 Gurodong-ro, Guro-gu, Seoul 08308, Korea Tel: +82-2-2626-3169 E-mail: gooday19@gmail.com
Co-correspondence: Hee-Joon Bae Department of Neurology, Cerebrovascular Center, Seoul National University Bundang Hospital, 82 Gumi-ro 173 beon-gil, Bundang-gu, Seongnam 13620, Korea Tel: +82-31-787-7464 E-mail: braindoc@snu.ac.kr
Received 2026 May 22; Revised 2026 September 5; Accepted 2026 September 7.

Abstract

Background and Purpose

Ambient air pollution is an established risk factor for incident stroke, but whether post-discharge pollutant exposure influences stroke recurrence remains unclear. We investigated the association between post-discharge exposure to six ambient air pollutants and stroke recurrence after acute ischemic stroke.

Methods

We analyzed 27,340 patients in the Clinical Research Collaboration for Stroke in Korea–National Institutes of Health nationwide multicenter registry (2014–2021). The 3-month post-discharge district-level mean concentrations of particulate matter ≤10 μm (PM10), ≤2.5 μm (PM2.5), nitrogen dioxide, sulfur dioxide (SO2), carbon monoxide, and ozone were estimated by inverse-distance-weighted interpolation. The primary outcome was stroke recurrence from 3 to 15 months post-discharge. Cause-specific Cox models with cluster-robust standard errors were used, treating death as a competing risk; non-linear dose-response was assessed by restricted cubic splines.

Results

During follow-up (median 364.8 days), 764 patients had recurrence and 542 died. Only SO2 showed a significant association with recurrence (P for overall association in the spline model= 0.024), with a graded dose-response; the non-linear (J-shaped) pattern did not reach significance (P for non-linearity=0.095), and quartile contrasts were individually non-significant (Q4 vs. Q1: hazard ratio 1.19, 95% confidence interval 0.94–1.50). The association was numerically stronger among older adults (≥75 years; P for interaction=0.051) and women (P for interaction=0.062). The highest SO2 concentrations were observed in harbor cities (Incheon, Ulsan, Busan). No significant associations were observed for the other five pollutants.

Conclusions

Elevated post-discharge SO2 exposure was associated with increased stroke recurrence, with a graded dose-response, supporting consideration of ambient air quality in secondary stroke prevention.

Introduction

Ambient air pollution is increasingly recognized as a major modifiable environmental risk factor for cardiovascular and cerebrovascular disease [1,2], with nearly 20% of cardiovascular mortality worldwide attributable to air pollution [2,3]. Fine particulate matter (PM2.5), nitrogen dioxide (NO2), sulfur dioxide (SO2), carbon monoxide (CO), and ozone (O3) each show independent associations with incident ischemic stroke [4-6], supporting integration of air quality standards into cardiovascular prevention frameworks [7]. However, most evidence pertains to first-ever stroke. Whether ambient pollutant exposure after an initial ischemic event modifies the risk of subsequent cerebrovascular outcomes—particularly stroke recurrence—remains largely uninvestigated.

Stroke recurrence is among the most clinically significant complications after an index ischemic event, affecting approximately 5% of survivors within the first year and disproportionately increasing cumulative disability, dependence, and mortality [8-10]. Current secondary prevention centers on pharmacotherapy and lifestyle modification [11]; environmental exposures are not yet systematically incorporated into post-stroke risk stratification. After an acute ischemic event, the cerebral and systemic vasculature exhibits endothelial dysfunction, persistent inflammation, oxidative stress, and a prothrombotic state [12,13]—conditions that may amplify the vascular toxicity of inhaled pollutants via systemic inflammation, autonomic imbalance, and oxidative damage [14,15]. Despite this biological plausibility, prospective evidence on post-discharge ambient exposure and stroke recurrence is essentially absent.

We addressed this gap using the Clinical Research Collaboration for Stroke in Korea–National Institutes of Health (CRCS-K-NIH) nationwide prospective registry, evaluating whether postdischarge exposure to six major ambient pollutants—PM10, PM2.5, NO2, SO2, CO, and O3—is independently associated with stroke recurrence in patients with acute ischemic stroke.

Methods

Study population and data source

Data were drawn from the CRCS-K-NIH registry, a nationwide, multicenter, prospective cohort currently enrolling consecutive patients with acute stroke across 20 centers in South Korea [16-18]. Between March 2014 and April 2021, 58,370 patients from 17 centers were registered. The registry captures standardized demographics, vascular risk factors, stroke characteristics, in-hospital treatment, and scheduled follow-up outcomes.

Eligibility criteria are detailed in Supplementary Figure 1. Patients were included if they (1) were admitted within 7 days of symptom onset, (2) had a confirmed ischemic stroke with an acute infarct on brain computed tomography or magnetic resonance imaging (MRI), (3) had provided informed consent for residential address collection, and (4) had valid ambient air quality data matched to their residential district. Exclusions were: missing discharge date or modified Rankin Scale (mRS); duplicate enrollment; discharge mRS >3 (as patients with severe disability are largely indoors and outdoor concentrations poorly proxy actual exposure); follow-up <3 months; or any vascular event within 3 months of discharge. The 3-month post-discharge exclusion window temporally separated exposure and outcome periods to minimize reverse-causality bias. Those with valid district-level measurements covering less than 50% of the 90-day post-discharge window were also excluded (n=6), because a representative window average could not be computed. After applying all criteria, 27,340 patients constituted the analytic cohort. The geographic distribution of residential addresses is shown in Supplementary Figure 2.

The study protocol was approved by the institutional review boards of all participating centers, including the coordinating center, Seoul National University Bundang Hospital (IRB No. B-2306-835-101), and all patients provided written informed consent. Individual patient-level data are available on reasonable request to the corresponding author, subject to legal and ethical restrictions.

Exposure assessment

The primary exposure was the mean ambient concentration of six pollutants during the first 90 days after discharge: particulate matter ≤10 μm (PM10, μg/m3) and ≤2.5 μm (PM2.5, μg/m3), NO2 (ppb), SO2 (ppb), CO (ppb), and O3 (ppb). Hourly concentration data were obtained from the Korean National Ambient Air Monitoring Network of the Ministry of Environment.

Because monitoring stations are not uniformly distributed across administrative districts (si/gun/gu), hourly pollutant and meteorological data from 332 stations were spatially interpolated to 250 districts using inverse-distance weighting (station locations are shown in Supplementary Figure 2), in which the estimate at each district centroid is a weighted average of surrounding station values with weights inversely proportional to squared distance [19]. District-level hourly values were then averaged over the 90-day post-discharge window. PM2.5 monitoring was unavailable in some districts because certain stations were established after enrollment began, so PM2.5 data were available for 26,340 of 27,340 patients (96.3%); the 1,000 patients missing PM2.5 were excluded only from PM2.5 models (complete-case analysis), with no imputation. Distributions of pollutant concentrations are shown in Supplementary Figure 3.

Outcome definition

The primary outcome was stroke recurrence, defined as a new ischemic or hemorrhagic stroke event occurring >3 months and ≤15 months after discharge. This 12-month outcome window commenced after the 3-month exposure assessment period to avoid temporal overlap. Events were prospectively ascertained at scheduled follow-up via in-person visits or structured telephone interviews by trained research coordinators [16]. All-cause mortality during follow-up was treated as a competing risk in the primary analysis.

Covariates

A comprehensive set of covariates was specified a priori. Patient-level variables included age, sex, onset-to-arrival time, initial National Institutes of Health Stroke Scale (NIHSS) score, premorbid mRS, stroke subtype by Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria using a pre-specified MRI-based algorithm [20], prior stroke or transient ischemic attack, coronary heart disease, hypertension, diabetes, dyslipidemia, atrial fibrillation, current smoking, intravenous thrombolysis, endovascular thrombectomy, discharge antiplatelet/anticoagulant/statin prescriptions, initial fasting glucose, and initial systolic blood pressure. District-level variables (si/gun/gu unit) included gross regional domestic product per capita, marriage rate, proportion of highly educated residents, working-age population rate, green area proportion, and mean temperature, relative humidity, and atmospheric pressure during the 90-day post-discharge period. Admitting hospital was included to account for center-level confounding.

Statistical analysis

Analyses were performed in SAS 9.4 (SAS Institute Inc., Cary, NC, USA); graphical outputs were generated in R 4.5.1 (R Foundation for Statistical Computing, Vienna, Austria). The primary analytic framework was a marginal Cox proportional hazards model with cause-specific hazard specification, treating stroke recurrence as the event of interest and all-cause mortality as a competing risk [21]. Cluster-robust standard errors accounted for the multilevel structure (patients nested within districts). The cause-specific approach was preferred over the Fine-Gray subdistribution model to evaluate the etiologic impact of exposures [21]. Continuous exposures were modeled with restricted cubic splines (RCS). Each pollutant was evaluated in a separate single-pollutant model to avoid multicollinearity among correlated ambient pollutants; pairwise correlations are shown in Supplementary Figure 4.

Baseline characteristics were compared across pollutant quartiles using the chi-squared test for categorical variables and analysis of variance or the Kruskal-Wallis test for continuous variables. Cumulative incidence functions were estimated by the Aalen-Johansen estimator and compared across quartiles using Gray’s test (recurrence) and the log-rank test (mortality). Each pollutant was modeled both as a categorical (quartiles; Q1 reference) and continuous variable; adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were derived from the marginal Cox model.

RCS models used 3 knots at the 10th, 50th, and 90th percentiles of each pollutant distribution. Omnibus Wald tests for overall association and tests for departure from linearity were reported. For pollutants with a significant overall association, an exploratory inflection point was identified as the value at which the lower bound of the 95% CI first exceeded HR=1.0. The proportional hazards assumption was evaluated by scaled Schoenfeld residuals and log-log survival plots, with no substantive violations identified.

Predefined subgroup analyses stratified on age (<75 vs. ≥75 years), sex, stroke subtype, atrial fibrillation, symptomatic steno-occlusion, and discharge mRS (0–1 vs. 2–3), hypothesizing a stronger association in patients with greater outdoor exposure. Effect modification was assessed by multiplicative interaction terms. All tests were two-sided; P<0.05 was considered significant for primary analyses, with a more lenient threshold of P<0.10 for interaction terms given conventionally reduced subgroup power. No formal correction for multiple comparisons was applied across the six pollutants, their quartile and spline specifications, or the interaction terms; nominally significant results are therefore descriptive rather than confirmatory. Sensitivity analyses examined robustness across 1-, 2-, and 3-month exposure windows. To test whether the SO2 association depended on the highest-exposure port regions, we repeated the cumulative incidence and RCS analyses excluding residents of the three major industrial-maritime cities (Incheon, Ulsan, Busan) simultaneously and in leave-one-region-out fashion. Inverse-distance weighting was used solely to derive each patient’s district-level exposure value; all subsequent inference was conducted at the individual level using cause-specific Cox models, with district-level socioeconomic and meteorological covariates included to account for area-level confounding.

Results

Of 58,370 patients initially registered, 27,340 met all eligibility criteria (Supplementary Figure 1). Mean age was 66.7±12.9 years; 17,406 (63.7%) were male. Hypertension was the most prevalent vascular risk factor (63.9%), followed by diabetes (31.1%), dyslipidemia (30.3%), and atrial fibrillation (16.5%); 16.1% had prior stroke or transient ischemic attack. By TOAST, large artery atherosclerosis predominated (34.5%), then small vessel occlusion (23.2%) and cardioembolism (17.3%). Symptomatic steno-occlusion was documented in 38.0%. Median admission NIHSS was 2.0 (interquartile range [IQR] 1.0–5.0) and discharge mRS 1.0 (IQR 1.0–2.0), with 55.8% mRS 0–1. At discharge, 82.4% received antiplatelets, 18.5% anticoagulants, and 92.5% statins. Baseline characteristics appear in Table 1 and, by pollutant quartile, in Supplementary Tables 1-6; differences across quartiles were confined mainly to district-level meteorological and socioeconomic variables.

Baseline characteristics of the study cohort (n=27,340)

During follow-up (median 364.8 days; IQR 364.8–365.8), 764 patients had stroke recurrence and 542 died. The 1-year cumulative incidence was 2.79% (95% CI 2.60–2.99) for recurrence and 1.98% (1.82–2.15) for mortality, each a competing risk for the other; quartile-stratified estimates are in Supplementary Tables 7 and 8. Recurrence curves by pollutant quartile are shown in Figure 1. Gray’s test indicated significant differences across quartiles for SO2 and PM10 (P=0.031 and 0.034), with no significant separation for NO2, CO, O3, or PM2.5. For mortality, no significant differences were observed for any pollutant (all log-rank P>0.05) (Supplementary Table 8 and Supplementary Figure 5).

Figure 1.

Cumulative incidence functions for stroke recurrence by quartiles of air pollutant exposure. Cumulative incidence of stroke recurrence within 1 year after discharge by quartiles of 3-month average post-discharge concentrations: (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, (F) PM2.5. All-cause mortality was a competing risk; P-values from Gray's test. SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter.

In the multivariable marginal Cox model, SO2 was the only pollutant with a statistically significant overall association with stroke recurrence (omnibus Wald test across quartiles, P=0.049) (Supplementary Table 9). Compared with Q1, adjusted HRs for Q2, Q3, and Q4 were 0.96 (95% CI, 0.78–1.19), 0.91 (0.73–1.13), and 1.19 (0.94–1.50), respectively. While no individual pairwise quartile comparison reached significance, the omnibus test was significant, with increased risk concentrated in the highest quartile. The remaining five pollutants showed no significant overall associations (Supplementary Table 9).

RCS analysis modeling SO2 continuously revealed a graded, positive dose-response, with the hazard relatively flat at lower concentrations and rising progressively at higher exposure (Figure 2). The omnibus Wald test was significant (P=0.024) and the non-linearity test approached significance (P=0.095). The concentration at which the lower 95% CI first exceeded HR=1.0 was approximately 8 ppb; because this inflection point was identified post hoc and the non-linearity test did not reach conventional significance, it should be regarded as exploratory rather than a validated clinical threshold. RCS analyses for the other five pollutants showed no significant overall associations, with generally flat curves (Figure 2).

Figure 2.

Restricted cubic spline plots for air pollutants and stroke recurrence. Restricted cubic spline curves for each pollutant, modeled continuously, and the cause-specific hazard of stroke recurrence: (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, (F) PM2.5. The reference is the median exposure (SO2, NO2, O3, CO in ppb; PM10, PM2.5 in μg/m³). The dashed line indicates HR=1.0. P-values for overall association and non-linearity are shown in each panel. SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter; HR, hazard ratio.

Predefined subgroup analyses for SO2 are summarized in Supplementary Figure 6. In RCS models stratified by predefined subgroups, the dose-response association was numerically stronger among adults ≥75 years (P for interaction=0.051) and women (P for interaction=0.062), although neither interaction reached conventional significance. No meaningful effect modification was observed by stroke subtype, atrial fibrillation, symptomatic steno-occlusion, or discharge mRS.

Post-discharge SO2 concentrations varied substantially across districts (Figure 3A). The highest mean levels during the study period were concentrated in major port cities: Incheon Jung-gu (9.5 ppb), Ulsan Nam-gu (8.9 ppb), Busan Yeongdo-gu (7.3 ppb), Ulsan Dong-gu (7.2 ppb), and Busan Jung-gu (7.1 ppb). This distribution is consistent with known SO2 sources from maritime shipping and petrochemical industries. For descriptive reference, district-level SO2 concentrations and age- and sex-adjusted cumulative incidence of recurrence are mapped in Figure 3; these maps are presented for visualization only and were not used for statistical inference.

Figure 3.

District-level distribution of average SO2 concentration and age- and sex-adjusted cumulative incidence of stroke recurrence across Korean districts (2014–2021). (A) Average ambient SO2 concentration by district (si/gun/gu) over the study period; the two highest-exposure districts—Incheon Jung-gu (9.5 ppb) and Ulsan Nam-gu (8.9 ppb)—are annotated with green arrowheads. (B) Age- and sex-adjusted cumulative incidence of stroke recurrence in the same districts. Because patient numbers within individual districts were small, cumulative incidence was estimated within 70 medium-level care areas and then mapped to their constituent districts. Asterisks denote areas with fewer than 50 patients or fewer than three recurrence events, for which estimates are unstable. These maps are provided as descriptive visualization of the distribution of exposure and outcome and were not used for statistical inference; all inference in this study was conducted at the individual level. The juxtaposition of panels A and B is presented for orientation only and is not intended to imply a spatial correspondence between exposure and outcome; no statistical comparison between the two maps was performed. SO2, sulfur dioxide.

Sensitivity analyses with shorter exposure windows showed a pattern consistent with a cumulative dose-response (Figure 4): the SO2–recurrence association was non-significant with the 1-month average, with the P-value decreasing in the 2-month window and reaching significance with the 3-month window. This progressive strengthening suggests that the cerebrovascular effect of SO2 may require sustained, cumulative exposure to appreciably elevate recurrence risk. The overall dose-response pattern was consistent across all three windows.

Figure 4.

Sensitivity analysis: SO2 dose-response by exposure duration. Restricted cubic spline plots for SO2 as a continuous variable, repeated for 1-month (A), 2-month (B), and 3-month (C) post-discharge exposure windows. SO2 is expressed in ppb. SO2, sulfur dioxide.

The SO2–recurrence association was robust to exclusion of the major port cities. Excluding Incheon, Ulsan, and Busan simultaneously (n=23,511) strengthened rather than attenuated the overall spline association (P=0.004; non-linearity P=0.028). Leave-one-region-out results were consistent (Incheon excluded, P= 0.017; Ulsan excluded, P=0.002; Busan excluded, P=0.088) (Supplementary Figures 7 and 8).

Discussion

In this nationwide multicenter study of 27,340 patients with acute ischemic stroke, among six major ambient air pollutants assessed during the first 3 months after discharge, only SO2 was significantly associated with stroke recurrence within 1 year, with a graded dose-response that persisted after excluding the major port cities. The association appeared numerically stronger among older adults and women, and the highest SO2 concentrations were observed in major port cities. To our knowledge, this is among the first large multicenter stroke-registry studies to evaluate post-discharge ambient pollutant exposure and recurrence within a competing-risk framework. Statistical support derived from the overall spline rather than any single quartile contrast, indicating a graded rather than threshold relationship.

That SO2—rather than particulate matter or other gaseous pollutants—was the sole pollutant significantly associated with recurrence merits careful interpretation. SO2 is a highly water-soluble gas that is absorbed by the respiratory tract and enters systemic circulation as sulfite and bisulfite derivatives [22,23]. In rat middle cerebral artery occlusion models, sustained SO2 inhalation upregulated cortical endothelin-1, inducible nitric oxide synthase, cyclooxygenase-2, and intercellular adhesion molecule-1, promoting endothelial dysfunction and vascular inflammation [22]. These pathways are especially relevant in post-stroke vasculature already marked by endothelial injury, inflammation, and a prothrombotic milieu.

Epidemiologic data are consistent with these mechanisms. Both short- and long-term SO2 exposures are associated with increased hospital admissions for cardiovascular disease and ischemic stroke [5,24,25]. A Korean nationwide study reported a strong SO2–cardioembolic stroke association (adjusted odds ratio 1.57 per 10-ppb increment), raising the possibility that SO2 triggers atrial fibrillation via autonomic dysregulation and atrial inflammatory remodeling [26]; ambient pollution has likewise been linked to incident atrial fibrillation [27]. Our subgroup analysis showed no significant effect modification by stroke subtype, although power to detect subtype-specific associations was limited.

The concentration of high SO2 in harbor regions creates a subpopulation with persistently elevated exposure from maritime shipping [28]. Because urban PM2.5 and nitrogen oxides are also driven by traffic and residential combustion, the harbor–nonharbor contrast is diluted for those pollutants [29], whereas the SO2 arises from a narrower set of sources, producing a wider range of exposure values across the cohort and correspondingly greater power to detect an association. Importantly, excluding the three major port cities did not attenuate the association—it was in fact stronger in the reduced cohort—making it less likely that the signal is solely attributable to the industrial-maritime profile of these regions.

The absence of significant associations for the other five pollutants should not be interpreted as definitive evidence of no effect; the modest number of recurrence events (n=764) may have limited power, particularly for pollutants with narrower interquartile ranges. Notably, PM2.5 showed a borderline overall association (P=0.098) with a suggestion of non-linearity (P=0.052); however, the direction was inverse—counterintuitive from a biological standpoint. This likely reflects residual confounding by urbanization: in South Korea, PM2.5 is highest in dense metropolitan areas [30] where residents also have better access to specialized stroke care, shorter transport times, higher socioeconomic status, and better adherence—all of which lower recurrence risk [31-33]. District-level socioeconomic adjustment may not fully capture these advantages.

The highest SO2 concentrations were observed in Incheon, Ulsan, and Busan—Korea’s principal maritime and petrochemical hubs. The International Maritime Organization implemented a global sulfur cap in January 2020, cutting permissible marine-fuel sulfur from 3.5% to 0.5% and substantially reducing SO2 in designated emission control areas [34]. Our 2014–2021 period largely predates full implementation, so harbor SO2 levels may reflect a transitional environment; whether the post-2020 decline reduces cerebrovascular risk warrants surveillance. SO2 also remains a concern where fossil-fuel combustion dominates energy supply and ambient levels far exceed those in Korea [35,36].

In national statistics, age-standardized incidence of stroke and myocardial infarction in the major port cities was comparable to the national average (stroke 112.3–113.5 vs. 113.2 per 100,000; myocardial infarction 34.5–38.5 vs. 37.1 per 100,000, 2023), indicating these are not areas of unusually high baseline vascular disease [37]. Because recurrence arises in already-injured, prothrombotic vasculature, such patients may be more susceptible to ambient pollutants than the general population, so the individual-level association is not explained by ecological incidence patterns. District-level rates should nonetheless be read cautiously, being shaped by hospital catchment and referral patterns.

A stronger SO2–recurrence association in older adults and women is biologically plausible in the post-stroke context. Aging entails reduced antioxidant defenses and endothelial dysfunction with lower nitric oxide bioavailability [38], which may lower the threshold at which oxidative stressors provoke recurrent injury; older individuals show heightened cardiovascular susceptibility to pollution [39]. In postmenopausal women, loss of estrogen-mediated vascular protection may heighten susceptibility to pollution-induced endothelial activation [40,41], and women appear more susceptible than men to cardiovascular effects of air pollution [42-44]. These subgroup findings are hypothesis-generating.

Clinical implications are twofold. Individually, the graded dose-response suggests that patients in high-SO2 districts may benefit from awareness of local air quality, limiting outdoor activity on high-SO2 days, and rigorous secondary prevention; our data do not, however, support a specific concentration as a clinical cutoff. At the population level, these findings strengthen the rationale for air quality monitoring in settings where SO2 emissions from shipping and fossil-fuel combustion remain substantial, supporting integration of environmental context into secondary stroke prevention [7].

Strengths include a large nationwide prospective cohort with standardized data across 17 centers [8], a cause-specific marginal Cox framework accounting for the competing risk of death [21], district-level socioeconomic and meteorological covariates often overlooked in air pollution studies [45], and consistency across exposure-window sensitivity analyses.

Several limitations should be acknowledged. First, residual confounding cannot be excluded; unmeasured indoor pollution, proximity to specific emission sources, and individual activity patterns may have influenced results. Second, exposure was assigned at the district level using values interpolated from monitoring stations rather than personal monitoring, which likely introduces non-differential measurement error and attenuates true effects toward the null [46]. This is especially relevant for SO2, a source-specific gas whose concentrations can vary appreciably over short distances near industrial stacks and shipping lanes, so that a single district-level value may poorly represent an individual’s actual exposure. Our metric reflects ambient rather than internal dose; biomarker approaches (e.g., urinary sulfate) would be more precise but are impractical at registry scale. Relatedly, although cluster-robust standard errors accounted for correlation among patients within the same district, we did not formally model spatial autocorrelation between neighboring districts; because exposure values were derived by interpolation and are therefore smoothed across adjacent areas, this could understate the true uncertainty of the SO2 association. Our analysis was not designed to characterize spatial structure, and the estimates should be interpreted accordingly. Third, residential mobility during the post-discharge period was not accounted for. Fourth, single-pollutant modeling does not capture synergistic interactions, and the SO2 association may partly proxy co-emitted pollutants not independently measured (e.g., ultrafine particles, volatile organic compounds) [47]. Fifth, outcome ascertainment through clinical follow-up may not have captured all events, particularly mild strokes managed at non-participating hospitals. Sixth, the 2014–2021 period overlaps the COVID-19 pandemic; lockdown-related emission reductions and presentation delays may have affected 2020–2021 data [48-50]. Seventh, the cohort was restricted to patients discharged with mRS ≤3 who remained event-free for 3 months, so the findings apply to ambulatory survivors of mild-to-moderate stroke rather than to all stroke patients. Finally, adherence at the 3-month visit (available for 26,448 patients, 96.7%) was high (median 100% [IQR 99–100] of prescribed doses taken in the preceding month), reducing concern that early adherence differences alone explain the association; however, adherence was self-reported and ascertained only to 3 months, so residual confounding by the completeness of secondary prevention thereafter cannot be excluded.

Conclusions

In this nationwide multicenter registry study, post-discharge exposure to ambient SO2 was associated with increased stroke recurrence within one year, with a graded dose-response that was robust to exclusion of the major port cities and numerically stronger (hypothesis-generating) associations among older adults and women. The highest SO2 concentrations were observed in harbor regions with active maritime shipping. These findings suggest that ambient SO2 may be an environmental determinant of stroke recurrence and support integrating air quality considerations into secondary prevention.

Supplementary materials

Supplementary materials related to this article can be found online at https://doi.org/10.5853/jos.2026.03265.

Supplementary Table 1.

Baseline characteristics by SO2 quartiles (n=27,340)

jos-2026-03265-Supplementary-Table-1.pdf
Supplementary Table 2.

Baseline characteristics by NO2 quartiles (n=27,340)

jos-2026-03265-Supplementary-Table-2.pdf
Supplementary Table 3.

Baseline characteristics by CO quartiles (n=27,340)

jos-2026-03265-Supplementary-Table-3.pdf
Supplementary Table 4.

Baseline characteristics by O3 quartiles (n=27,340)

jos-2026-03265-Supplementary-Table-4.pdf
Supplementary Table 5.

Baseline characteristics by PM10 quartiles (n=27,340)

jos-2026-03265-Supplementary-Table-5.pdf
Supplementary Table 6.

Baseline characteristics by PM2.5 quartiles (n=26,340)

jos-2026-03265-Supplementary-Table-6.pdf
Supplementary Table 7.

Cumulative incidence of stroke recurrence by air pollutant quartiles

jos-2026-03265-Supplementary-Table-7.pdf
Supplementary Table 8.

Cumulative incidence of all-cause mortality by air pollutant quartiles

jos-2026-03265-Supplementary-Table-8.pdf
Supplementary Table 9.

Multivariable-adjusted hazard ratios (95% CI) for stroke recurrence by air pollutant quartiles

jos-2026-03265-Supplementary-Table-9.pdf
Supplementary Figure 1.

Study flowchart. CT, computed tomography; MRI, magnetic resonance imaging; mRS, modified Rankin Scale.

jos-2026-03265-Supplementary-Figure-1.pdf
Supplementary Figure 2.

Geographic distribution of patient residential addresses and air-quality monitoring stations. (A) Geographic distribution of patients’ residential districts. (B) Locations of the 332 ambient air-quality monitoring stations used for inverse-distance-weighted interpolation. Station density was higher in metropolitan areas and lower in mountainous and rural regions.

jos-2026-03265-Supplementary-Figure-2.pdf
Supplementary Figure 3.

Distributions of the 3-month average post-discharge air pollutant concentrations. Histograms of the 3-month average post-discharge concentrations of (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, and (F) PM2.5. Each panel displays the median and IQR of the corresponding pollutant. SO2, NO2, O3, and CO are expressed in ppb; PM10 and PM2.5 are expressed in μg/m3. IQR, interquartile range; SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter.

jos-2026-03265-Supplementary-Figure-3.pdf
Supplementary Figure 4.

Correlation matrix among the six ambient air pollutants. Pearson correlation coefficients among the 3-month average post-discharge concentrations of the six ambient air pollutants. PM10 and PM2.5 (r=0.89) and CO and NO2 (r=0.85) were strongly correlated, whereas SO2 showed only low-to-moderate correlations with the other pollutants (|r|=0.25–0.61). O3, ozone; CO, carbon monoxide; SO2, sulfur dioxide; PM, particulate matter; NO2, nitrogen dioxide.

jos-2026-03265-Supplementary-Figure-4.pdf
Supplementary Figure 5.

Cumulative incidence functions for all-cause mortality by air pollutant quartiles. Cumulative incidence curves for all-cause mortality within 1 year after discharge according to quartiles of the 3-month average post-discharge concentrations of (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, and (F) PM2.5. P-values were derived using log-rank test. SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter.

jos-2026-03265-Supplementary-Figure-5.pdf
Supplementary Figure 6.

Subgroup analyses: SO2 and stroke recurrence. Plots showing hazard ratios (95% CI) for the association between SO2 exposure and stroke recurrence across predefined subgroups: (A) age (<75 vs. ≥75 years), (B) sex (male vs. female), (C) stroke subtype, (D) atrial fibrillation, (E) symptomatic steno-occlusion, and (F) discharge mRS. P-values for interaction are reported for each stratification variable. SO2 is expressed in ppb. SO2, sulfur dioxide; TOAST, Trial of Org 10172 in Acute Stroke Treatment; LAA, large artery atherosclerosis; SVO, small vessel occlusion; CE, cardioembolism; mRS, modified Rankin Scale; CI, confidence interval.

jos-2026-03265-Supplementary-Figure-6.pdf
Supplementary Figure 7.

Cumulative incidence of stroke recurrence by SO2 quartile after exclusion of major port cities. Cumulative incidence curves for stroke recurrence within 1 year after discharge by quartiles of the 3-month average post-discharge SO2 concentration: (A) excluding Incheon, Ulsan, and Busan (n=23,511); (B) excluding Incheon (n=27,208); (C) excluding Busan (n=25,041); (D) excluding Ulsan (n=25,954). All-cause mortality was treated as a competing risk. P-values were derived using Gray’s test. SO2, sulfur dioxide.

jos-2026-03265-Supplementary-Figure-7.pdf
Supplementary Figure 8.

Restricted cubic spline curves for SO2 and stroke recurrence after exclusion of major port cities. Restricted cubic spline curves showing the association between continuously modeled SO2 concentration and the cause-specific hazard of stroke recurrence under the same exclusion scenarios: (A) excluding Incheon, Ulsan, and Busan; (B) excluding Incheon; (C) excluding Busan; (D) excluding Ulsan. P-values for overall association and non-linearity are shown in each panel. SO2, sulfur dioxide.

jos-2026-03265-Supplementary-Figure-8.pdf

Notes

Funding statement

This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF), funded by the Ministry of Education (grant number RS-2023-00238058), and by a research grant from the Korea National Institute of Health (project no. 2023-ER-1006-02). The funding sources had no role in the design, conduct, analysis, interpretation, or reporting of the study.

Conflicts of interest

Hee-Joon Bae reports grants from Amgen Korea Inc., Bayer Korea, Bristol Myers Squibb Korea, Celltrion, Dong-A ST, Otsuka Korea, Samjin Pharm, and Takeda Pharmaceuticals Korea Co., Ltd., and personal fees from Amgen Korea, Bayer, Daewoong Pharmaceutical Co., Ltd., Daiichi Sankyo, Eisai Korea, Inc., JW Pharmaceutical, and SK Chemicals, outside the submitted work. The other authors report no disclosures.

Author contribution

Conceptualization: Keon-Joo Lee. Study design: Keon-Joo Lee, Jeongeun Hwang, Seong-Eun Kim. Methodology: Jeongeun Hwang, Seong-Eun Kim. Data collection: Keon-Joo Lee, Beom Joon Kim, Moon-Ku Han, Hyunsoo Kim, Joon-Tae Kim, Kang-Ho Choi, Kyu Sun Yum, Dong-Ick Shin, Jae-Kwan Cha, Dae-Hyun Kim, Dong-Seok Gwak, Dong-Eog Kim, Jong-Moo Park, Kyusik Kang, Soo Joo Lee, Jae Guk Kim, Minwoo Lee, Mi-Sun Oh, Kyung-Ho Yu, Hong-Kyun Park, Keun-Sik Hong, Yong-Jin Cho, Joong-Goo Kim, Jay Chol Choi, Tai Hwan Park, Sang-Soon Park, Jee-Hyun Kwon, Wook-Joo Kim, Doo Hyuk Kwon, Jun Lee, Kyungbok Lee, Jeong-Yoon Lee, Sung Il Sohn, Jeong-Ho Hong, Kwang-Yeol Park, Hae-Bong Jeong, Chulho Kim, Sang-Hwa Lee, Hee-Joon Bae. Investigation: all authors. Statistical analysis: Seong-Eun Kim. Writing—original draft: Keon-Joo Lee. Writing—review & editing: Keon-Joo Lee, Jeongeun Hwang, Seong-Eun Kim, Hee-Joon Bae. Funding acquisition: Keon-Joo Lee, Hee-Joon Bae. Approval of final manuscript: all authors.

References

1. GBD 2019 Stroke Collaborators. Global, regional, and national burden of stroke and its risk factors, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet Neurol 2021;20:795–820.
2. Brauer M, Casadei B, Harrington RA, Kovacs R, Sliwa K, ; WHF Air Pollution Expert Group. Taking a stand against air pollution-the impact on cardiovascular disease: a joint opinion from the World Heart Federation, American College of Cardiology, American Heart Association, and the European Society of Cardiology. Circulation 2021;143:e800–e804.
3. Cohen AJ, Brauer M, Burnett R, Anderson HR, Frostad J, Estep K, et al. Estimates and 25-year trends of the global burden of disease attributable to ambient air pollution: an analysis of data from the Global Burden of Diseases Study 2015. Lancet 2017;389:1907–1918.
4. Verhoeven JI, Allach Y, Vaartjes ICH, Klijn CJM, de Leeuw FE. Ambient air pollution and the risk of ischaemic and haemorrhagic stroke. Lancet Planet Health 2021;5:e542–e552.
5. Shah AS, Lee KK, McAllister DA, Hunter A, Nair H, Whiteley W, et al. Short term exposure to air pollution and stroke: systematic review and meta-analysis. BMJ 2015;350:h1295.
6. Toubasi A, Al-Sayegh TN. Short-term exposure to air pollution and ischemic stroke: a systematic review and meta-analysis. Neurology 2023;101:e1922–e1932.
7. Kaufman JD, Elkind MSV, Bhatnagar A, Koehler K, Balmes JR, Sidney S, et al. Guidance to reduce the cardiovascular burden of ambient air pollutants: a policy statement from the American Heart Association. Circulation 2020;142:e432–e447.
8. Kim DY, Park TH, Cho YJ, Park JM, Lee K, Lee M, et al. Contemporary statistics of acute ischemic stroke and transient ischemic attack in 2021: insights from the CRCS-K-NIH registry. J Korean Med Sci 2024;39e278.
9. Rohweder G, Ellekjær H, Salvesen Ø, Naalsund E, Indredavik B. Functional outcome after common poststroke complications occurring in the first 90 days. Stroke 2015;46:65–70.
10. Khanevski AN, Bjerkreim AT, Novotny V, Naess H, Thomassen L, Logallo N, et al. Recurrent ischemic stroke: incidence, predictors, and impact on mortality. Acta Neurol Scand 2019;140:3–8.
11. Kleindorfer DO, Towfighi A, Chaturvedi S, Cockroft KM, Gutierrez J, Lombardi-Hill D, et al. 2021 Guideline for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline from the American Heart Association/American Stroke Association. Stroke 2021;52:e364–e467.
12. Iadecola C, Buckwalter MS, Anrather J. Immune responses to stroke: mechanisms, modulation, and therapeutic potential. J Clin Invest 2020;130:2777–2788.
13. Esenwa CC, Elkind MS. Inflammatory risk factors, biomarkers and associated therapy in ischaemic stroke. Nat Rev Neurol 2016;12:594–604.
14. Rajagopalan S, Al-Kindi SG, Brook RD. Air pollution and cardiovascular disease: JACC state-of-the-art review. J Am Coll Cardiol 2018;72:2054–2070.
15. Miller MR, Newby DE. Air pollution and cardiovascular disease: car sick. Cardiovasc Res 2020;116:279–294.
16. Kim BJ, Park JM, Kang K, Lee SJ, Ko Y, Kim JG, et al. Case characteristics, hyperacute treatment, and outcome information from the clinical research center for stroke-fifth division registry in South Korea. J Stroke 2015;17:38–53.
17. Kim BJ, Han MK, Park TH, Park SS, Lee KB, Lee BC, et al. Current status of acute stroke management in Korea: a report on a multicenter, comprehensive acute stroke registry. Int J Stroke 2014;9:514–518.
18. Bae HJ, ; CRCS-K Investigators. David G. Sherman Lecture Award: 15-year experience of the nationwide multicenter stroke registry in Korea. Stroke 2022;53:2976–2987.
19. Hwang J, Yi H, Jang M, Kim JG, Kwon SU, Kim N, et al. Air pollution and subarachnoid hemorrhage mortality: a stronger association in women than in men. J Stroke 2022;24:429–432.
20. Ko Y, Lee S, Chung JW, Han MK, Park JM, Kang K, et al. MRI-based algorithm for acute ischemic stroke subtype classification. J Stroke 2014;16:161–172.
21. Austin PC, Lee DS, Fine JP. Introduction to the analysis of survival data in the presence of competing risks. Circulation 2016;133:601–609.
22. Sang N, Yun Y, Li H, Hou L, Han M, Li G. SO2 inhalation contributes to the development and progression of ischemic stroke in the brain. Toxicol Sci 2010;114:226–236.
23. Wang XB, Jin HF, Tang CS, Du JB. The biological effect of endogenous sulfur dioxide in the cardiovascular system. Eur J Pharmacol 2011;670:1–6.
24. Sunyer J, Ballester F, Tertre AL, Atkinson R, Ayres JG, Forastiere F, et al. The association of daily sulfur dioxide air pollution levels with hospital admissions for cardiovascular diseases in Europe (the Aphea-II study). Eur Heart J 2003;24:752–760.
25. Niu Z, Liu F, Yu H, Wu S, Xiang H. 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. Environ Health Prev Med 2021;26:15.
26. Chung JW, Bang OY, Ahn K, Park SS, Park TH, Kim JG, et al. Air pollution is associated with ischemic stroke via cardiogenic embolism. Stroke 2017;48:17–23.
27. Kim IS, Yang PS, Lee J, Yu HT, Kim TH, Uhm JS, et al. Long-term exposure of fine particulate matter air pollution and incident atrial fibrillation in the general population: a nationwide cohort study. Int J Cardiol 2019;283:178–183.
28. Kim Y, Moon N, Chung Y, Seo J. Impact of IMO sulfur regulations on air quality in Busan, Republic of Korea. Atmosphere 2022;13:1631.
29. Viana M, Hammingh P, Colette A, Querol X, Degraeuwe B, de Vlieger I, et al. Impact of maritime transport emissions on coastal air quality in Europe. Atmos Environ 2014;90:96–105.
30. Kim HC, Kim S, Kim BU, Jin CS, Hong S, Park R, et al. Recent increase of surface particulate matter concentrations in the Seoul Metropolitan Area, Korea. Sci Rep 2017;7:4710.
31. Lee EJ, Jeong HY, Kim J, Park NH, Kang MK, Lee D, et al. Regional disparities in prehospital delay of acute ischemic stroke: the Korean Stroke Registry. Eur Stroke J 2024;9:1063–1073.
32. Al AlShaikh S, Quinn T, Dunn W, Walters M, Dawson J. Predictive factors of non-adherence to secondary preventative medication after stroke or transient ischaemic attack: a systematic review and meta-analyses. Eur Stroke J 2016;1:65–75.
33. Pantoja-Ruiz C, Akinyemi R, Lucumi-Cuesta DI, Youkee D, Emmett E, Soley-Bori M, et al. Socioeconomic status and stroke: a review of the latest evidence on inequalities and their drivers. Stroke 2025;56:794–805.
34. Sofiev M, Winebrake JJ, Johansson L, Carr EW, Prank M, Soares J, et al. Cleaner fuels for ships provide public health benefits with climate tradeoffs. Nat Commun 2018;9:406.
35. Li C, McLinden C, Fioletov V, Krotkov N, Carn S, Joiner J, et al. India is overtaking China as the world’s largest emitter of anthropogenic sulfur dioxide. Sci Rep 2017;7:14304.
36. Lin J, Zhou C, Chen L, Huang G, Lamarque JF, Nie J, et al. Sulfur emissions from consumption by developed and developing countries produce comparable climate impacts. Nat Geosci 2022;15:184–189.
37. Korea Disease Control and Prevention Agency. Korea National Cardio-cerebrovascular Disease Statistics 2023 Cheongju: Korea Disease Control and Prevention Agency; 2025.
38. Donato AJ, Machin DR, Lesniewski LA. Mechanisms of dysfunction in the aging vasculature and role in age-related disease. Circ Res 2018;123:825–848.
39. Cesaroni G, Forastiere F, Stafoggia M, Andersen ZJ, Badaloni C, Beelen R, et al. Long term exposure to ambient air pollution and incidence of acute coronary events: prospective cohort study and meta-analysis in 11 European cohorts from the ESCAPE Project. BMJ 2014;348:f7412.
40. Moreau KL, Hildreth KL, Klawitter J, Blatchford P, Kohrt WM. Decline in endothelial function across the menopause transition in healthy women is related to decreased estradiol and increased oxidative stress. Geroscience 2020;42:1699–1714.
41. Iorga A, Cunningham CM, Moazeni S, Ruffenach G, Umar S, Eghbali M. The protective role of estrogen and estrogen receptors in cardiovascular disease and the controversial use of estrogen therapy. Biol Sex Differ 2017;8:33.
42. Miller KA, Siscovick DS, Sheppard L, Shepherd K, Sullivan JH, Anderson GL, et al. Long-term exposure to air pollution and incidence of cardiovascular events in women. N Engl J Med 2007;356:447–458.
43. Chen LH, Knutsen SF, Shavlik D, Beeson WL, Petersen F, Ghamsary M, et al. The association between fatal coronary heart disease and ambient particulate air pollution: are females at greater risk? Environ Health Perspect 2005;113:1723–1729.
44. Zhang J, Wang X, Yan M, Shan A, Wang C, Yang X, et al. Sex differences in cardiovascular risk associated with long-term PM2.5 exposure: a systematic review and meta-analysis of cohort studies. Front Public Health 2022;10:802167.
45. Hajat A, MacLehose RF, Rosofsky A, Walker KD, Clougherty JE. Confounding by socioeconomic status in epidemiological studies of air pollution and health: challenges and opportunities. Environ Health Perspect 2021;129:065001.
46. Goldman GT, Mulholland JA, Russell AG, Strickland MJ, Klein M, Waller LA, et al. Impact of exposure measurement error in air pollution epidemiology: effect of error type in time-series studies. Environ Health 2011;10:61.
47. Jang E, Choi S, Yoo E, Hyun S, An J. Impact of shipping emissions regulation on urban aerosol composition changes revealed by receptor and numerical modelling. NPJ Clim Atmos Sci 2023;6:52.
48. Ju MJ, Oh J, Choi YH. Changes in air pollution levels after COVID-19 outbreak in Korea. Sci Total Environ 2021;750:141521.
49. Na SK, Kim JH, Lee WY, Oh MR. Impact of the COVID-19 pandemic on admission and mortality among patients with severe emergency diseases at emergency departments in Korea in 2020: registry data from the National Emergency Department Information System. J Korean Med Sci 2023;38e243.
50. Kansagra AP, Goyal MS, Hamilton S, Albers GW. Collateral effect of Covid-19 on stroke evaluation in the United States. N Engl J Med 2020;383:400–401.

Article information Continued

Figure 1.

Cumulative incidence functions for stroke recurrence by quartiles of air pollutant exposure. Cumulative incidence of stroke recurrence within 1 year after discharge by quartiles of 3-month average post-discharge concentrations: (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, (F) PM2.5. All-cause mortality was a competing risk; P-values from Gray's test. SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter.

Figure 2.

Restricted cubic spline plots for air pollutants and stroke recurrence. Restricted cubic spline curves for each pollutant, modeled continuously, and the cause-specific hazard of stroke recurrence: (A) SO2, (B) NO2, (C) CO, (D) O3, (E) PM10, (F) PM2.5. The reference is the median exposure (SO2, NO2, O3, CO in ppb; PM10, PM2.5 in μg/m³). The dashed line indicates HR=1.0. P-values for overall association and non-linearity are shown in each panel. SO2, sulfur dioxide; NO2, nitrogen dioxide; CO, carbon monoxide; O3, ozone; PM, particulate matter; HR, hazard ratio.

Figure 3.

District-level distribution of average SO2 concentration and age- and sex-adjusted cumulative incidence of stroke recurrence across Korean districts (2014–2021). (A) Average ambient SO2 concentration by district (si/gun/gu) over the study period; the two highest-exposure districts—Incheon Jung-gu (9.5 ppb) and Ulsan Nam-gu (8.9 ppb)—are annotated with green arrowheads. (B) Age- and sex-adjusted cumulative incidence of stroke recurrence in the same districts. Because patient numbers within individual districts were small, cumulative incidence was estimated within 70 medium-level care areas and then mapped to their constituent districts. Asterisks denote areas with fewer than 50 patients or fewer than three recurrence events, for which estimates are unstable. These maps are provided as descriptive visualization of the distribution of exposure and outcome and were not used for statistical inference; all inference in this study was conducted at the individual level. The juxtaposition of panels A and B is presented for orientation only and is not intended to imply a spatial correspondence between exposure and outcome; no statistical comparison between the two maps was performed. SO2, sulfur dioxide.

Figure 4.

Sensitivity analysis: SO2 dose-response by exposure duration. Restricted cubic spline plots for SO2 as a continuous variable, repeated for 1-month (A), 2-month (B), and 3-month (C) post-discharge exposure windows. SO2 is expressed in ppb. SO2, sulfur dioxide.

Table 1.

Baseline characteristics of the study cohort (n=27,340)

Variable Value
Demographics
 Age (yr) 66.7±12.9
 Male sex 17,406 (63.7)
 Premorbid mRS 0 (0–0)
Vascular risk factors
 Hypertension 17,457 (63.9)
 Diabetes mellitus 8,494 (31.1)
 Dyslipidemia 8,297 (30.3)
 Atrial fibrillation 4,498 (16.5)
 Coronary heart disease 2,283 (8.4)
 Current smoking 7,161 (26.2)
 Previous stroke or TIA 4,396 (16.1)
Stroke characteristics
 Onset to arrival (hr) 9.7 (2.3–35.6)
 NIHSS at admission 2.0 (1.0–5.0)
Stroke subtype (TOAST)
 Large artery atherosclerosis 9,440 (34.5)
 Small vessel occlusion 6,329 (23.2)
 Cardioembolism 4,729 (17.3)
 Other determined 886 (3.2)
 Undetermined 5,956 (21.8)
 Symptomatic steno-occlusion 10,390 (38.0)
Treatment
 Intravenous thrombolysis 3,055 (11.2)
 Endovascular thrombectomy 1,724 (6.3)
 Discharge antiplatelet therapy 22,527 (82.4)
 Discharge anticoagulant therapy 5,052 (18.5)
 Discharge statin therapy 25,297 (92.5)
Physiologic variables
 Initial fasting glucose (mg/dL) 143.6±62.3
 Initial SBP (mm Hg) 148.3±27.2
Discharge functional status
 mRS at discharge 1.0 (1.0–2.0)
 mRS 0–1 15,269 (55.8)
 mRS 2–3 12,071 (44.2)
District-level variables
 GRDP per capita (thousands KRW) 35,086±31,266
 Marriage rate (%) 56.3±3.5
 High education rate (%) 48.1±10.8
 Working population rate (%) 58.2±3.4
 Green area proportion (%) 67.3±18.7
Meteorological data (90-day post-discharge)
 Mean temperature (°C) 13.3±8.2
 Mean relative humidity (%) 68.4±7.5
 Mean atmospheric pressure (hPa) 1,005.3±5.9

Values are mean±SD, n (%), or median (IQR).

mRS, modified Rankin Scale; TIA, transient ischemic attack; NIHSS, National Institutes of Health Stroke Scale; TOAST, Trial of Org 10172 in Acute Stroke Treatment; SBP, systolic blood pressure; GRDP, gross regional domestic product; KRW, Korean won; SD, standard deviation; IQR, interquartile range.