Sleep duration in relation to dementia risk in the UK Million Women Study
WONG TSZ., REEVES G., BERAL V., FLOUD S.
Background: Sleep is often disrupted in people with dementia. To help assess whether any associationbetween sleep duration and dementia may be the consequence of pre-clinical disease or causal weexamined dementia risk in relation to follow-up time after reporting sleep duration in a large UKprospective study of women. Method: About 1.3 million women were recruited in median year 1998 (IQR 1997-1999) in England andScotland; in 2001 (IQR 2000-2003), ~830,000 reported their usual hours of sleep per 24 hours. Theywere followed for any mention of dementia in hospital records, the first mention of which was themain outcome. Cox proportional hazards regression was used to estimate dementia detection riskratios (RRs and 95% confidence intervals [CI]) for <6, 6, 9, and >9 hours of sleep, compared to 7-8hours, respectively, during follow-up periods of 0-4, 5-9, 10-14 and 15+ years after sleep duration wasreported. Adjustment was made for sociodemographic factors, lifestyle factors, and othercharacteristics. Result: Women most commonly reported sleeping for 7-8 hours (67%), with substantially fewerwomen reporting short sleep duration (<6 hours [5%], 6 hours [18%]) or long sleep duration (9 hours[7%], >9 hours [3%]). Over a mean 17 years of follow-up (with <2% lost to follow-up) ~34,000 womenhad at least one hospital record of mentioning dementia. Compared to 7-8 hours, there was no clearand consistent association between short sleep duration (<6, 6 hours) and dementia detection riskthroughout the follow-up periods. Long sleep duration (9, >9 hours) was associated with higherdementia detection risk in the first 5 years of follow-up (e.g. RR for >9 vs 7-8 hours sleep: 2.78 [95% CI,2.22-3.48]), but these RRs declined rapidly as follow-up time increased and had largely disappeared15+ years of follow-up (1.18 [1.08-1.28]), suggesting that the short-term associations are theconsequence of pre-clinical disease. Associations by follow-up period did not notably differ forAlzheimer’s disease, vascular dementia, and unspecified dementia. Conclusion: Increased sleep duration appears to be a consequence of pre-clinical dementia but thereis little to suggest a long-term causal association of sleep duration with dementia risk in women.