Background: The longitudinal associations of sleep timing and time in bed
(TIB) with dementia and cognitive decline in older adults are unclear.
Methods: This population-based cohort study used data from 1982 participants
who were aged ≥60 years, free of dementia, and living in rural communities in
western Shandong, China. At the baseline (2014) and follow-up (2018) examinations, sleep parameters were assessed using standard questionnaires. Cognitive
function was measured using the Mini-Mental State Examination (MMSE).
Dementia was diagnosed following the DSM-IV criteria, and the NIA-AA criteria
for Alzheimer disease (AD). Data were analyzed using restricted cubic splines,
Cox proportional-hazards models, and general linear models.
Results: During the mean follow-up of 3.7 years, dementia was diagnosed in
97 participants (68 with AD). Restricted cubic spline curves showed J-shaped
associations of sleep duration, TIB, and rise time with dementia risk, and a
reverse J-shaped association with mid-sleep time. When sleep parameters were
categorized into tertiles, the multivariable-adjusted hazard ratio (HR) of incident dementia was 1.69 (95%CI 1.01–2.83) for baseline sleep duration >8 hours
(vs. 7–8 h), 2.17 (1.22–3.87) for bedtime before 9 p.m. (vs. 10 p.m. or later), and
2.00 (1.23–3.24) for mid-sleep time before 1 a.m. (vs. 1–1.5 a.m.). Early bedtim

and mid-sleep time were significantly associated with incident AD (HR range:
2.25–2.51; p < 0.05). Among individuals who were free of dementia at followup, baseline long TIB, early bedtime and mid-sleep time, early and late rise
time, and prolonged TIB and advanced bedtime and mid-sleep time from baseline to follow-up were associated with a greater decline in MMSE score
(p < 0.05). These associations with cognitive decline were statistically evident
mainly among men or participants who were aged 60–74 years.

Conclusions: Long TIB and early sleep timing are associated with an
increased risk of dementia, and the associations with greater cognitive decline
are evident only among older people aged 60–74 years and men.

https://agsjournals.onlinelibrary.wiley.com/doi/epdf/10.1111/jgs.18042