Habitual sleep duration and spontaneous intracerebral hemorrhage: a prospective UK Biobank cohort study with supporting hospital-based neurosurgical data

Background Abnormal sleep duration has been associated with stroke, but evidence specific to spontaneous intracerebral hemorrhage (sICH) remains limited. Methods We conducted a two-part observational study. The prospective UK Biobank analysis included 314,236 participants free of sICH at baseline, among whom 1,387 developed incident sICH. The supporting cross-sectional hospital analysis included 312 patients with imaging-confirmed sICH and 515 non-ICH neurosurgical controls from Jinjiang Municipal Hospital. Habitual sleep duration was categorized as short (<7 h/day), normal (7–9 h/day), or long (>9 h/day). Cox proportional hazards models [hazard ratios (HRs)] and logistic regression [odds ratios (ORs)] were used, together with restricted cubic splines, additional adjustment for medications, blood pressure, and indicators of prevalent ill health, and multiple imputation of missing data. A systemic inflammatory marker score and interaction analyses with false-discovery-rate correction were examined as secondary, exploratory analyses. Results Compared with normal sleep, short and long sleep were associated with higher incident sICH risk in the UK Biobank [adjusted HRs 1.14 (95% CI 1.03–1.26) and 1.55 (1.15–2.09)]; the short-sleep association was modest and less robust in sensitivity analyses. Corresponding cross-sectional ORs in the hospital dataset were 1.49 (1.04–2.13) and 2.05 (1.17–3.59), respectively. Restricted cubic splines suggested U-shaped associations in both datasets, with the lowest estimated hazard or odds at approximately 7–8 h/day. Estimates were similar after additional confounder adjustment (HRs 1.10 and 1.43 for short and long sleep, respectively) and multiple imputation (HRs 1.13 and 1.50, respectively). No interaction, including with inflammatory marker burden, remained significant after false-discovery-rate correction, and adjustment for the marker score changed the estimates only modestly. Conclusion Both short and long sleep duration were associated with higher incident sICH risk in the UK Biobank and higher odds of sICH in the hospital dataset, with stronger and more consistent associations observed for long sleep. Systemic inflammatory markers neither explained nor reliably modified these associations, and because long sleep may partly reflect prevalent ill health, the long-sleep association warrants cautious interpretation. Sleep duration may be a useful marker of sICH susceptibility.