The sleep-wake cycle: a new behavioral marker for dementia risk prediction
Dementia, including Alzheimer’s disease and related dementias, represents a major public health challenge due to population aging. Early detection is essential given the long preclinical phase of this syndrome, as well as the emergence of therapies targeting its earliest stages. Disruptions in the sleep-wake cycle have been reported during the preclinical phase of dementia; however, their contribution to dementia risk prediction remains uncertain.
This study examined the association between objective sleep-wake cycle measures derived from accelerometry and dementia risk in two UK-based prospective cohorts including participants aged 60 years and older: the UK Biobank (53,448 participants; mean follow-up 7.8 years) and Whitehall II (3,965 participants; mean follow-up 10.6 years). Among 36 accelerometry-derived metrics capturing daytime activity, sleep, rest-activity rhythms, and chronotype, nine metrics predictive of dementia risk were identified and combined into two components using a machine learning approach. The first component primarily reflected lower, less frequent, and more fragmented daytime physical activity. The second mainly captured more disrupted sleep profiles, including extreme sleep durations, prolonged nocturnal awakenings, lower probability of falling back asleep once awakened, and earlier waking times.
Both sleep-wake cycle components were associated with an increased risk of dementia and improved the predictive performance of a model including 13 sociodemographic, behavioral, and health-related risk factors for dementia. These findings were replicated and confirmed in the Whitehall II validation cohort, where the addition of these components also improved the predictive performance of a model further incorporating cognitive functioning and the blood biomarker p-tau217, a promising biomarker of Alzheimer’s disease. The incremental contribution of sleep-wake cycle measures was greater than that of sex and education level, comparable to that of APOE genotype and cognitive functioning, and approximately twofold lower than that of the blood biomarker p-tau217.
These results demonstrate that accelerometry-derived sleep-wake cycle metrics contribute significantly to dementia risk prediction over and above known risk factors. In the era of digital biomarkers, this study highlights the potential of sleep-wake cycle measures as scalable, non-invasive behavioral markers that could be deployed at scale, in combination with established risk factors and biomarkers, for the early identification of individuals at increased risk of dementia.
By Clémence Cavaillès
