Accelerometer Sleep Metrics Linked to Dementia Risk in Older Adults
Key Clinical Summary
- Accelerometer-derived sleep-wake cycle metrics were associated with incident dementia in a cohort of more than 53,000 older adults.
- Two composite sleep-wake components independently predicted dementia risk and modestly improved prediction models alongside established risk factors.
- Predictive improvement from sleep-wake metrics was comparable to that associated with APOE genotype and was validated in the Whitehall II cohort.
Disruptions in sleep-wake cycles may provide scalable early markers for dementia risk, according to findings published in JAMA Neurology. Using data from the UK Biobank and Whitehall II cohort studies, investigators evaluated whether accelerometer-derived activity and sleep measures were associated with incident dementia, and whether these metrics improved dementia prediction beyond established risk factors.
Study Findings
Researchers included 53,448 UK Biobank participants aged 60 years and older (mean age, 67.5 years; 54.2% female) with a mean follow-up of 7.8 years. A separate validation analysis included 3965 Whitehall II participants (mean age, 69.4 years; 25.9% female) followed for a mean of 10.6 years.
The study analyzed data from older adults without dementia who participated in accelerometer substudies within the UK Biobank and Whitehall II cohorts. Incident all-cause dementia was identified through electronic health records during follow-up.
Investigators extracted 36 accelerometer-derived sleep-wake cycle metrics and applied machine learning methods to identify variables most strongly associated with dementia risk. Nine metrics were ultimately grouped into 2 composite components.
Higher values for component 1 reflected shorter and less frequent bouts of moderate-to-vigorous physical activity, increased low-intensity activity, lower diversity of activity intensity, and greater transitions from activity to rest during daytime. Higher values for component 2 reflected more extreme sleep durations, prolonged wake bouts during sleep, lower transitions from wake to sleep, and earlier waking times.
Both components were associated with increased dementia risk. Component 1 was associated with a 43% higher risk of dementia (hazard ratio [HR], 1.43; 95% CI, 1.33-1.54), while component 2 was associated with a 10% higher risk (HR, 1.10; 95% CI, 1.04-1.17).
Adding the sleep-wake components to models containing sociodemographic, behavioral, and health-related variables improved dementia prediction, increasing the C index by 0.018 (95% CI, 0.011-0.025). Findings were replicated in the Whitehall II cohort.
Clinical Implications
The findings suggest that objective sleep-wake cycle measures may offer clinically meaningful information for dementia risk stratification. Wearable accelerometers may provide scalable tools for identifying individuals at elevated risk before cognitive symptoms emerge.
The study also evaluated multidimensional patterns rather than isolated sleep duration alone. Daytime activity fragmentation, reduced physical activity intensity, disturbed nighttime sleep continuity, and altered chronotype collectively contributed to dementia prediction. This broader characterization of circadian and behavioral patterns may better capture early neurodegenerative changes.
Although the improvement in predictive performance was modest, investigators noted that the contribution of sleep-wake metrics was comparable to that associated with APOE genotype when added to age-only prediction models. The findings support growing evidence linking circadian disruption and sleep abnormalities with neurodegenerative disease processes.
Expert Commentary
“Results of this cohort study suggest that accelerometer-derived sleep-wake cycle features, integrating key metrics of daytime activity, sleep, and chronotype, made a modest albeit statistically significant contribution to a dementia prediction model containing age, known risk factors, and an Alzheimer disease (AD) blood biomarker,” wrote Clémence Cavaillès, PhD, Université Paris Cité and Université Sorbonne Paris Nord, Inserm U1153, INRAE, Centre for Research in Epidemiology and Statistics, Epidemiology of Ageing and Neurodegenerative Diseases, Paris, France, and study coauthors.
They concluded that future studies should evaluate the “clinical utility as scalable, noninvasive behavioral markers in combination with established predictors to enable early identification of individuals at higher risk of dementia.”
