Background Disturbed sleep is common throughout the community and is associated with an increase in daytime sleepiness, both of which, in turn are associated with an increased risk of ischaemic vascular disease. The hypothesis that sleep disturbances are predictive of dementia, and in particular vascular dementia was tested in a large community-based cohort of older men.

Methods A questionnaire on sleep disturbances was administered to 1986 men aged 55–69 years in the Caerphilly Cohort Study and 10 years later the men were examined clinically for evidence of dementia or cognitive impairment with no dementia (CIND).

Findings Approximately 20% of the men reported disturbed sleep and 30% reported ‘severe’ daytime sleepiness. Ten years later 1225 men (75% of the surviving men in the cohort) were tested and 268 (22%) were found to be cognitively impaired with 93 (7.6%) showing clear evidence of dementia and the remaining 175 (14.3%) showing evidence of CIND. After adjustment for possible confounding, including cognitive function and the taking of sleeping tablets at baseline, sleep disturbances appeared to be predictive of dementia and CIND of vascular origin, while there was no suggestion of prediction of non-vascular cognitive impairment by sleep. Prediction of vascular dementia appeared to be particularly strong for daytime sleepiness, with an adjusted OR of 4.44 (95% CI 2.05 to 9.61). Further adjustments for psychological distress at baseline reduced the size of the relationships, but the ORs remain large, consistent with a direct positive effect of sleep disturbance on vascular dementia.

Interpretation Sleep disturbances, and in particular severe daytime sleepiness, appear to be strongly predictive of vascular dementia, but have no predictive power for non vascular dementia.