Abstract
BACKGROUND: The intricate interplay of various sleep characteristics may influence dementia risk through different pathogenic pathways. However, few studies have examined multi-dimensional sleep health in relation to dementia risk or explored potential etiologic heterogeneity by dementia subtypes. METHODS: Our study included 313,248 UK Biobank participants aged ≥ 50 years who were dementia-free in 2006-2010. Incident dementia was identified using validated algorithms through primary care, hospital admissions, or death records through 2022. Multi-dimensional sleep health was evaluated based on seven self-reported sleep-related factors and assessed in two ways: (1) using an a priori sleep health score (SHS) ranging from 0 to 7, with higher scores indicating healthier sleep, and (2) through data-driven sleep health patterns identified by latent class analysis. We used Cox proportional hazards models to estimate the associations between multi-dimensional sleep health and risk of all-cause dementia, vascular dementia (VaD), and Alzheimer's disease (AD). RESULTS: There were 7458 incident all-cause dementia cases (1636 VaD, 3376 AD) after 4,165,352 person-years of follow-up. After adjusting for potential confounders, the hazard ratio (95% CI) comparing participants with SHS of 0-2 (worst sleep) vs 6-7 (best sleep) was 1.76 (1.52, 2.05) for all-cause dementia (p-trend < 0.0001), 2.13 (1.61, 2.83) for VaD (p-trend < 0.0001), and 1.55 (1.22, 1.97) for AD (p-trend < 0.57). We identified six multi-dimensional sleep health patterns, including relatively healthy sleep, insomnia with short sleep duration, non-restorative sleep with evening chronotype, insomnia with non-restorative sleep, snoring with daytime sleepiness and napping, and severely disturbed sleep with multiple symptoms and daytime impairment. Compared with the healthy sleep pattern, all other five sleep patterns were significantly associated with 8-85% higher all-cause dementia risk and 11-148% higher VaD risk, whereas only the severely disturbed sleep pattern was associated with 56% higher AD risk (95% CI: 1.21, 2.01). CONCLUSIONS: Poor multi-dimensional sleep health, either assessed by a simple SHS or characterized by sleep clusters, was associated with higher incident dementia risk. There is substantial heterogeneity in multi-dimensional sleep health patterns and their associations with different dementia outcomes. Understanding the specific sleep health profiles associated with dementia risk may help to identify high-risk populations and inform more targeted interventions.