More REM sleep is linked to a lower risk of 83 diseases, study finds

A study published in PLOS Medicine on September 17, 2026, found that people who got more REM sleep faced a lower risk of 83 different diseases, based on wrist-accelerometer data from 95,559 UK Biobank participants tracked for a median of 8.9 years. The research, led by Shengzhi Sun of Capital Medical University in Beijing along with coauthors from Peking University’s School of Public Health, is among the largest efforts yet to connect specific sleep stages, rather than just total sleep time, to long-term disease outcomes.

The size of the dataset is what let the researchers run what they called a phenome-wide association analysis, cross-referencing sleep-stage data against more than a thousand distinct health outcomes recorded in participants’ medical records rather than testing a handful of preselected conditions.

Wrist Sensors Replaced Self-Reported Sleep Logs

Participants wore accelerometers on their wrists for seven consecutive days and nights, and the research team ran that raw motion data through a deep-learning algorithm to classify each night into REM sleep, deep sleep, lighter sleep stages, total duration and periods of wakefulness after initially falling asleep. That approach sidesteps a known weakness in sleep research: people are notoriously inaccurate at estimating how long they slept or how well, and clinical sleep labs that can measure stages directly are too expensive and disruptive to run on nearly 100,000 people for the better part of a decade.

The tradeoff is that wrist accelerometers cannot distinguish sleep stages as finely as a lab-grade EEG-based sleep study, which reads brainwave activity directly rather than inferring stage from movement patterns. The authors treated their sleep-stage classifications as estimates validated against smaller EEG-based samples rather than a diagnostic-grade measurement, a limitation that applies to any accelerometer study operating at this scale.

Eighty-Three Diseases Tracked Back to One Sleep Stage

The study’s central result involves REM sleep specifically: greater REM sleep, measured per a 47.6-minute increase, was associated with a lower risk of 83 diseases across the tracked health outcomes. Three examples illustrate the range of what showed up — heart failure risk fell with a hazard ratio of 0.74, dementia risk fell with a hazard ratio of 0.54, and Parkinson’s disease risk fell with a hazard ratio of 0.20, meaning higher REM sleep was associated with roughly a fifth the risk in that last comparison. A hazard ratio below 1.0 indicates lower risk in the group with more of the measured trait, so the further below 1.0 the ratio falls, the stronger the association the researchers found.

Sun and coauthors described the pattern as evidence that sleep architecture, not just how many hours someone spends in bed, carries independent information about disease risk. The published paper, titled “Accelerometer-derived real-world sleep stages and risk of incident diseases,” frames REM sleep’s association with dementia and Parkinson’s as consistent with existing neuroscience research suggesting REM sleep plays a role in clearing metabolic byproducts from the brain and consolidating memory, though the study itself is observational and cannot establish that low REM sleep causes those diseases rather than reflecting an early symptom of them.

Deep Sleep and Total Duration Add a Second Layer

Deep sleep showed a narrower but still measurable association, linked to lower risk of 7 conditions, including type 2 diabetes and major depressive disorder. That is a much shorter list than REM sleep’s 83, and the EurekAlert release accompanying the study attributed the difference to REM sleep simply showing a stronger and more widespread statistical signal across the phenome-wide analysis than deep sleep did, not to deep sleep being unimportant.

Total sleep duration behaved differently from either individual stage, showing what the researchers called a non-linear relationship with disease risk. The lowest risk concentrated in a 6-to-8-hour window, while people sleeping less than 5 hours a night faced elevated risk across 37 conditions — the most widespread pattern of any sleep measure in the study. Medical Xpress, summarizing the same findings, noted that the short-sleep group’s risk elevation spanned cardiovascular, metabolic and psychiatric conditions rather than clustering in one body system, which is part of why the authors framed 6 to 8 hours as a broad health target rather than a niche recommendation for one condition.

The study leaves open how much of the REM-sleep association reflects something modifiable through behavior — alcohol use, medication timing and stress are all known to suppress REM sleep — versus something driven by underlying health conditions that both reduce REM sleep and independently raise disease risk. Because the UK Biobank cohort was recruited from people already willing to wear a research accelerometer and share years of follow-up health data, the sample also skews toward people healthier and more health-engaged than the general population, which can understate how strong these associations might look in a broader group. An independent research-literacy review of the paper flagged that distinction as the central open question for follow-up studies, since UK Biobank’s largely middle-aged and older, predominantly healthy-volunteer population may not generalize cleanly to younger adults or people with existing sleep disorders.

This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.

More from Morning Overview