An analysis of nearly 96,000 UK Biobank participants found that longer REM and deep sleep, more regular sleep and roughly six to eight hours of nightly rest were associated with lower risks across a broad range of diseases, although researchers cautioned that the observational study cannot establish that sleep patterns directly prevent illness.
By yourNEWS Media Newsroom.
A large study tracking the sleep patterns and subsequent health of nearly 96,000 adults has found that people who spent more time in rapid eye movement sleep had lower risks of 83 diseases, while those sleeping fewer than five hours per night showed elevated risks for dozens of conditions.
The research, published Sept. 17 in the open-access journal PLOS Medicine, examined wrist-monitor data from 95,559 participants in the UK Biobank and followed them for a median of 8.9 years. Researchers compared several aspects of sleep — including REM sleep, deep sleep, light sleep, total sleep duration, nighttime wakefulness and irregularity — with the later occurrence of 1,049 health conditions.
The study identified 156 statistically significant associations between sleep characteristics and newly diagnosed diseases after researchers applied the stringent Bonferroni correction for multiple comparisons.
“Phenome-wide association analysis identified 156 significant associations between sleep patterns and incident diseases after Bonferroni correction,” the authors reported.
The strongest pattern involved REM sleep.
For every 47.6-minute increase representing the study’s interquartile range in REM duration, researchers found statistically significant associations with lower risk across 83 diseases spanning 12 categories. Among them were heart failure, atrial fibrillation, ischemic heart disease, multiple sclerosis, dementia, Alzheimer’s disease and parkinsonism.
The hazard ratio was 0.74 for heart failure, indicating a 26% lower relative rate associated with the measured difference in REM sleep after statistical adjustment. The corresponding hazard ratio was 0.54 for dementia and 0.20 for parkinsonism.
Those figures describe statistical associations rather than proof that adding a certain number of minutes of REM sleep would itself reduce an individual’s disease risk by those amounts.
Deep sleep showed a narrower but still significant pattern.
An additional 47.5 minutes across the study’s interquartile range of deep sleep was associated with lower risk for seven diseases across five diagnostic categories. Those included type 2 diabetes, major depressive disorder, sleep apnea and Parkinson’s disease. The hazard ratios included 0.89 for type 2 diabetes, 0.86 for major depressive disorder, 0.78 for sleep apnea and 0.70 for Parkinson’s disease.
The results suggest that sleep architecture — how a night’s sleep is distributed among different physiological stages — may contain health information beyond the total number of hours spent asleep.
The researchers also found that the proportion of REM sleep remained significant after accounting for total sleep time. After that adjustment, a higher REM proportion was associated with lower risk for 54 disease phenotypes, while a larger proportion of light sleep was associated with higher risk for 28 conditions.
The investigators obtained the sleep data without bringing nearly 100,000 people into sleep laboratories.
Each participant wore a wrist accelerometer continuously for seven days and nights. Researchers then used the SleepNet deep-learning algorithm to estimate REM, deep and light sleep, total sleep duration, wakefulness after sleep onset and variations in sleep timing from night to night.
The approach allowed researchers to measure sleep under ordinary living conditions rather than relying on participants to remember and report how long or how well they slept.
That was important because the authors noted that self-reported sleep often correlates poorly with objective measurements. Large-scale objective studies of sleep architecture and future disease have also historically been difficult because conventional polysomnography is expensive and generally measures only one or several nights at a time.
Wrist accelerometers, however, are not equivalent to laboratory polysomnography.
Polysomnography directly measures brain electrical activity along with eye movements, muscle activity and other physiological signals. The SleepNet system infers sleep stages largely from movement-related physiological information.
The study authors specifically cautioned that the algorithm systematically differs from gold-standard polysomnography, tending to underestimate REM and overestimate non-REM sleep. They said the measurements therefore should be understood as “movement-based estimates rather than exact physiological sleep stages.”
The participants were middle-aged and older adults. Their average baseline age was 56.2 years, 43.7% were men and 96.9% identified as White. Median nightly sleep duration was about 392 minutes, or approximately 6.5 hours. Median REM sleep was about 81.7 minutes, deep sleep about 102.6 minutes and light sleep about 193.3 minutes.
Total sleep duration produced another major finding.
Across the 1,049 health outcomes studied, researchers identified 86 conditions with statistically significant nonlinear relationships to sleep duration.
For 69 of those conditions, the calculated point of minimum risk fell between six and eight hours of sleep per night. Examples included dementia at approximately 7.53 hours, hypotension at 7.43 hours, chronic obstructive pulmonary disease at 7.38 hours, type 2 diabetes at 7.34 hours and chronic kidney disease at 7.10 hours.
“Sleep duration exhibited significant non-linear associations with 86 disease phenotypes, with the minimum-risk duration for the majority of these conditions precisely concentrated within a 6-8 hour window,” the authors stated.
Short sleep showed particularly broad associations.
When participants were divided into those sleeping fewer than five hours, five to six hours, six to eight hours, eight to nine hours and at least nine hours, researchers found 41 significant associations with greater disease risk compared with the six-to-eight-hour reference group.
Thirty-seven of those 41 associations occurred among people sleeping fewer than five hours.
Using a broader cutoff, people sleeping less than six hours showed significantly higher rates of 24 conditions after Bonferroni correction, including heart failure, type 2 diabetes and chronic obstructive pulmonary disease. Compared with six-to-eight-hour sleepers, the corresponding hazard ratios were 1.30 for heart failure, 1.18 for type 2 diabetes and 1.24 for COPD.
Long sleep did not show the same breadth of associations.
Among those sleeping more than eight hours, major depressive disorder was the only condition remaining significantly elevated after the most stringent correction in that comparison, with a hazard ratio of 1.60. Researchers cautioned that long sleep can sometimes be a consequence or early symptom of disease rather than its cause — a problem known as reverse causation.
“The findings provide additional evidence supporting the role of a 6-8 hours’ sleep duration as a health safeguard for middle-aged and older adults, likely attributable to more favorable distributions of sleep stages,” the authors stated.
The finding is slightly different from widely used general recommendations for adults.
The National Sleep Foundation recommends seven to nine hours for adults ages 18 to 64 and seven to eight hours for adults 65 and older. The National Heart, Lung, and Blood Institute similarly says most adults should obtain seven to nine hours nightly.
The PLOS researchers themselves noted that their calculated six-to-eight-hour window was slightly shorter than the commonly recommended seven-to-nine-hour range and specifically applies to the middle-aged and older UK Biobank population studied.
Sleep continuity also mattered.
Participants who spent more time awake after initially falling asleep had higher risks for six diseases after the stringent statistical correction. Increased wakefulness after sleep onset was associated with conditions including psychoactive substance dependence, alcohol abuse and osteoarthritis.
Greater night-to-night sleep irregularity was associated with higher risks of abdominal pain, anxiety disorders and major depressive disorder. For anxiety, the hazard ratio associated with the study’s 93-minute interquartile difference in irregularity was 1.23; for major depressive disorder it was 1.26.
The results add to research suggesting that the quality, timing and internal structure of sleep may be relevant to health rather than total duration alone.
A separate study involving 270 middle-aged and older participants in the Atherosclerosis Risk in Communities study used laboratory polysomnography and later brain MRI examinations. Researchers found that lower proportions of slow-wave and REM sleep were associated with greater atrophy in brain regions considered vulnerable to Alzheimer’s disease 13 to 17 years later.
Another large 2026 analysis published in Nature found a U-shaped association between sleep duration and biological aging across brain and body systems. The lowest biological-age gaps occurred between approximately 6.4 and 7.8 hours of sleep, depending on the organ system and sex examined.
Those results broadly overlap with the six-to-eight-hour range identified in the new PLOS Medicine analysis, although the studies used different methods and measured different outcomes.
REM and deep sleep also serve different physiological roles.
Modern sleep classification divides sleep into three non-REM stages — N1, N2 and N3 — plus REM sleep. N3 is commonly referred to as deep or slow-wave sleep.
REM sleep is associated with vivid dreaming, memory processing and emotional regulation, while slow-wave sleep is associated with physical restoration and several metabolic and neurological processes.
Research has also examined whether sleep influences the brain’s glymphatic system, which is thought to participate in cerebrospinal-fluid circulation and removal of metabolic waste. A systematic review found substantial evidence connecting sleep with cerebrospinal-fluid and glymphatic-related processes, but researchers cautioned that the human evidence remains complex and inconsistent across populations and diseases.
The authors of the PLOS study emphasized that their findings should not be interpreted as proof that increasing REM sleep or sleeping a particular number of hours will prevent disease.
This was an observational cohort study. Researchers measured sleep patterns, adjusted their statistical models for numerous factors and then observed which illnesses occurred later. Such a design can uncover associations but cannot establish cause and effect.
“As this is an observational study, the main limitations are that causality cannot be inferred and residual confounding may remain,” the researchers wrote.
The team attempted to address the possibility that undiagnosed illnesses were already altering participants’ sleep before they received a diagnosis.
Cases occurring during the first six months after sleep monitoring were excluded from the primary analysis. Researchers then repeated the analysis using a two-year exclusion period.
Only 95 of the original 156 significant associations remained statistically significant after the two-year washout. The authors said that result suggests early or undiagnosed disease may have influenced sleep patterns in some of the associations.
Other limitations included measuring sleep for only seven days, relying on inpatient diagnosis records rather than complete primary-care and outpatient data and using participant characteristics that in some instances had been collected years before the accelerometer monitoring.
Generalizability is another concern.
UK Biobank participants tend to be healthier and have higher socioeconomic status than the general population, and nearly 97% of participants in this analysis were White. The authors said those characteristics could influence the observed associations and limit how confidently the results can be extended to other populations.
The research also provides context for growing concern about inadequate and irregular sleep in China, where several of the study investigators are based.
The 2026 China Sleep Health Research White Paper reported that surveyed residents averaged 6.97 hours of nighttime sleep in 2025. The report also found widespread sleep difficulties, with about 65% of respondents reporting one or two sleep disturbances per week.
PLOS’ summary of the report said 23% of respondents slept less than six hours per night and only 26% reported maintaining a consistent sleep schedule.
The study was conducted by researchers including Shengzhi Sun of Capital Medical University, Jingsong Luo, Ruiyi Liu, Jie Yin, Wangnan Cao and Rui Chen. Funding included support from the National Natural Science Foundation of China, Beijing Municipal Science and Technology Commission and Chinese Institutes for Medical Research programs. The authors declared no competing interests, and PLOS reported that the funders had no role in the study’s design, analysis, publication decision or preparation.
The researchers said their findings support paying attention to the composition and regularity of sleep in addition to its duration.
“Maintaining a sleep duration of 6-8 hours can effectively reduce the risk of multiple diseases, providing new insights for health promotion and preventive practice,” the authors stated.
Because the evidence remains observational, the strongest conclusion is narrower: among the nearly 96,000 middle-aged and older adults studied, more REM sleep, greater amounts of deep sleep, more consistent sleep and approximately six to eight hours of total sleep were associated with generally more favorable long-term health outcomes.
Whether deliberately changing those sleep characteristics can prevent the diseases identified in the analysis will require controlled interventional research.
Comments