Abstract / Summary
Abstract Background Household air pollution from solid-fuel combustion remains common in China, but prospective evidence for individual stroke subtypes and the role of household ventilation is limited. We examined associations of cooking fuel and stove ventilation with incident hemorrhagic stroke. Methods We analyzed 390197 China Kadoorie Biobank participants without baseline stroke or transient ischemic attack. Cooking fuel was classified as clean (gas or electricity) or polluting (coal or wood), using information from the present residence and, when unavailable, up to two previous residences. The primary outcome was first hemorrhagic stroke (ICD-10 I60-I61) before a competing stroke subtype or censoring. Cause-specific Cox models compared participants within region and adjusted for demographic, socioeconomic, lifestyle, and medical factors. Ventilation was examined only among polluting-fuel users. Results During 3774336 person-years (mean 9.67 years), 5782 hemorrhagic strokes occurred. Compared with clean fuel, polluting fuel was associated with hemorrhagic stroke after full adjustment (hazard ratio [HR] 1.146, 95% confidence interval [CI] 1.043–1.259; P = 0.004). Among 194904 polluting-fuel users, partial ventilation was not associated with risk relative to ventilation of all stoves (HR 0.981, 95% CI 0.896–1.074), whereas no ventilation was associated with higher risk (HR 1.135, 95% CI 1.018–1.264). Current-fuel-only, one-year landmark, and post hoc region-overlap analyses were consistent. Adding blood pressure reduced the fuel log-HR by 33.2%; this was covariate attenuation, not causal mediation. Polluting fuel was inversely associated with ischemic stroke (HR 0.949, 95% CI 0.911–0.989). Conclusions Polluting cooking fuel was associated with a modestly higher hazard of hemorrhagic stroke. Among polluting-fuel users, no ventilation identified a higher-risk group, but the observational design does not establish an intervention effect. The inverse ischemic-stroke association should not be interpreted as protective and warrants investigation of residual confounding and outcome ascertainment.