Evidence from prospective studies on heterogeneity in responses to ambient air pollutants across stroke clinical subgroups is limited. We conducted an exploratory, region-wide ecological time-series analysis to estimate lag–response associations, focusing on stroke clinical-phenotype subgroups. Daily stroke admissions (14 February 2024–31 August 2025) were modelled using single-pollutant, log-linear quasi-Poisson generalised additive models with distributed-lag non-linear cross-basis functions (penalised splines), adjusting for potential confounders. Cumulative relative risk (RR) estimates were summarised for two exposure contrasts: the interquartile range (IQR) and the 97.5th versus 2.5th percentiles (adverse-episode), across predefined lag windows (days). Cumulative RRs were estimated for clinical subgroups. Across pollutants, adverse-episode contrasts (RR97.5th) generally yielded larger effects than day-to-day contrasts (RRIQR). Nitrogen dioxide (NO2) showed the largest positive cumulative effect at lag 0–5 days (RR97.5th: 1.24). Under the day-to-day contrast, NO2 showed marginally significant increases in cumulative RRs at lag 0–3 among older adults (≥ 65 years), male patients, and those with diabetes or hypertension. Week-scale lag effects for NO2 were statistically significant among patients with macroangiopathy (RRIQR: 1.49; and RR97.5th: 3.01). Ozone showed negative associations across lag windows for most clinical subgroups. For particulate matter, elevations were most evident at lag 0–3 and attenuated by lag 0–7. Among female patients, PM10–stroke associations at lag 0–3 were statistically significant under both contrasts (RRIQR: 1.08; and RR97.5th: 1.41). Under the day-to-day contrast, PM10 showed a statistically significant association among patients undergoing thrombectomy across all lag windows; no significant association was observed under the adverse-episode contrast. This study leverages data from one of the first large, prospective registries of stroke admissions, with detailed clinical phenotyping, to examine pollution–stroke associations by clinical subgroup. Although exploratory, these findings suggest clinically relevant links that warrant validation in multicentre cohorts.
Zhou et al. (2026) studied this question.