Figure 1: Pooled association between sleep macro-architecture measures and incident dementia.

Figure depicts the meta-analysis with forest plot. All results were adjusted for age (years), sex (men vs. women), BMI (kg/m²), anti-depressant use (yes vs no), sedative use (yes vs no), and APOE e4 status (non e4 carrier vs. at least one copy of e4). Cohort studies included: ARIC, Atherosclerosis Risk in Communities study; CHS, Cardiovascular Health Study; FHS, Framingham Heart Study; MrOS, Osteoporotic Fractures in Men Study; SOF, Study of Osteoporotic Fractures. The sleep exposures in each model included: N1, non-rapid eye movement sleep stage 1; N2, non-rapid eye movement sleep stage 2; N3, non-rapid eye movement sleep stage 3; REM, rapid eye movement sleep; WASO, Wake after sleep onset; SME, sleep maintenance efficiency. Note that, square root transformation was applied to N1% and N3% and natural log transformation was applied to SME%, WASO and the AHI due to skewed distributions of these sleep metrics. Dementia case numbers are presented for each cohort with hazard ratio (HR) and 95% confidence intervals (95% CI) for dementia risk. Heterogeneity in effect sizes was determined via the Higgins I² test. Statistical significance, p<0.05.