Sleep profiles, dementia and the heart: a study of 2,667 men

A study published in 2025 in Communications Medicine, a Nature Portfolio journal, followed 2,667 men aged 65 or over for 12 years. Using unsupervised machine learning, the researchers derived sleep profiles from 24-hour actigraphy recordings and studied their link with dementia and cardiovascular disease.

The three sleep profiles

Three emerged: active healthy sleepers (64.0% of participants), fragmented poor sleepers (14.1%) and long and frequent nappers (21.9%). The main comparison is between the first profile and the second.

The risk data

Compared with active healthy sleepers, fragmented poor sleepers showed a hazard ratio of 1.35 for dementia and 1.32 for cardiovascular events, after accounting for other factors (multivariable adjustment).

In plain terms: over the period observed, the risk was about 35% higher for dementia and 32% higher for cardiovascular events. Long and frequent nappers showed only a marginal association with cardiovascular events, and none with dementia.

Until now most research had studied one sleep characteristic at a time, such as duration.

This study looks at them together: sleep duration and fragmentation, activity across the 24 hours, daytime behaviour. It is an observational study: it shows associations in a group, and the group is made up only of older men. The authors draw two indications: sleep offers possible targets for intervention, and poor sleepers deserve closer assessment.

The Digital Twin and this research

The Digital Twin lines up the data your devices write to your phone, such as steps, heart rate and sleep, and shows you how they change over time. It does not assign your sleep to a profile, it makes no forecasts about dementia and it does not diagnose: this research describes groups of people, not you. If your sleep changes, talk to your doctor.