Abstract
BACKGROUND AND OBJECTIVES: Sleep disruptions are associated with cognitive aging in older adults. However, little is known about longitudinal sleep changes in the oldest old and whether these changes are linked to cognitive impairment. We aimed to determine whether changes in 24-hour multidimensional sleep-wake activity are associated with mild cognitive impairment (MCI) and dementia in oldest old women. METHODS: We studied cognitively unimpaired women enrolled in the Study of Osteoporotic Fractures who completed wrist actigraphy twice (baseline and follow-up) and had cognitive status evaluated at follow-up using a neuropsychological battery and adjudication. To identify multidimensional sleep-wake change profiles, we performed hierarchical clustering on principal components on the 5-year changes (median 5.0 [range 3.5-6.3] years) in nighttime sleep (sleep duration, sleep efficiency [SE], and wake after sleep onset [WASO]), napping (duration and frequency), and circadian rest-activity rhythms (RARs; acrophase, amplitude, mesor, and robustness). Using multinomial logistic regression, we evaluated the associations between these profiles-and individual parameter changes-and MCI and dementia risk at follow-up. RESULTS: Of 733 participants (mean age 82.5 ± 2.9 years), 164 (22.4%) developed MCI and 93 (12.7%) developed dementia by the follow-up visit. We identified 3 sleep-wake change profiles: stable sleep (SS; n = 321 [43.8%]) was characterized by stability or small improvements; declining nighttime sleep (n = 256 [34.9%]) showed decreases in nighttime sleep quality and duration, moderate napping increases, and worsening circadian RARs; and increasing sleepiness (IS; n = 156 [21.3%]) exhibited large increases in daytime and nighttime sleep duration and quality, and worsening circadian RARs. After adjustment for age, education, race, body mass index, diabetes, hypertension, myocardial infarction, antidepressant use, and baseline cognition, women with IS had approximately double the risk of dementia (odds ratio 2.21, 95% CI 1.14-4.26) compared with those with SS. SE, WASO, nap duration, and nap frequency were individually associated with dementia. Neither sleep-wake change profiles nor individual parameters were associated with MCI. DISCUSSION: Among community-dwelling women in their 80s, those with increasing 24-hour sleepiness over 5 years had doubled dementia risk during that time. Change in multidimensional 24-hour sleep-wake activity may serve as an early marker or risk factor for dementia in oldest old women.