We usually judge a good night’s sleep by how many hours we get. But the time we spend in different stages of sleep might be just as important.
In a study involving nearly 100,000 UK adults, researchers found that people who spent more time in rapid eye movement (REM) sleep — the stage most closely associated with vivid dreaming — were less likely to develop 83 different health conditions over the following years. These ranged from heart failure to dementia.
But there’s a catch.
A week of wrist movements, followed by years of health records
Measuring sleep on a large scale is very challenging. The most established method, called polysomnography, records brain waves, eye movements and muscle activity to distinguish sleep stages.
However, the equipment and overnight laboratory setup make it difficult to use with tens of thousands of people over extended periods.
Sleep questionnaires offer a simpler alternative, but people may misremember how long they slept or fail to notice brief awakenings, so the method isn’t exactly reliable. People also can’t tell how much time they spent in different sleep stages.
To get a more objective picture, researchers from Peking University and Capital Medical University in Beijing used data from the UK Biobank, a long-running health study. Between 2013 and 2015, participants wore wrist-mounted accelerometers for seven consecutive days. These devices recorded movement, which researchers could use to estimate sleep patterns.
Of the 103,611 participants who returned usable recordings, the team retained data from 95,559 people (the rest failed quality checks or did not meet the study’s completeness requirements). Their average age of participants was 56.2 years.
The researchers then used a machine-learning system called SleepNet to estimate participants’ sleep stages from their wrist movements. The algorithm identified patterns associated with REM sleep, non-REM sleep and wakefulness.
REM sleep is when the brain becomes particularly active and most vivid dreaming occurs. Non-REM sleep includes lighter stages and deep sleep, which is associated with physical restoration and recovery. Throughout a typical night, we cycle between these stages several times.
The team compared these measures with hospital inpatient records covering 1,049 health outcomes. Participants were followed for a median period of 8.9 years, with follow-up ending at the first recorded occurrence of the relevant disease, death or April 1, 2024.
Researchers also accounted for factors including age, sex, body mass index, smoking, alcohol consumption, physical activity and environmental exposures.
REM sleep stood out across dozens of diseases
The analysis identified 156 significant associations between sleep patterns and newly recorded diseases.
People who spent more time in REM sleep had a lower risk of developing 83 different diseases. These included cardiovascular conditions such as heart failure, high blood pressure and ischemic heart disease, as well as neurological conditions including dementia, Parkinson’s, and multiple sclerosis.
Some of the associations were substantial. Comparing participants who differed by roughly 48 minutes in estimated REM sleep per night, those with more REM sleep had a 26% lower relative hazard of heart failure and a 46% lower relative hazard of dementia during follow-up. These figures describe statistical associations, not the benefits someone could expect from deliberately increasing their REM sleep.
Importantly, it’s not just about how much people slept in general; REM seems to matter regardless. When researchers accounted for total sleep duration and examined REM sleep as a proportion of time asleep, associations with 54 conditions remained statistically significant.
Deep sleep also appeared to matter, although across fewer diseases. Greater amounts were associated with a lower risk of seven conditions, including type 2 diabetes, major depressive disorder, sleep apnea and Parkinson’s disease.
This being said, sleep duration also mattered. The researchers found relationships between sleep duration and 86 disease outcomes, meaning that the relationship did not follow a simple, steadily rising or falling pattern. For 69 of these conditions, the estimated lowest-risk sleep duration fell within a window of six to eight hours.
People sleeping fewer than five hours showed the broadest pattern of adverse associations: they accounted for 37 of the 41 significant adverse associations identified when extreme short sleepers were compared with the six-to-eight-hour reference group. Longer sleep duration was also associated with a lower risk of 50 conditions.
Better sleep measurements, but no proof of cause and effect
The study’s scale and long follow-up offer valuable clues about the relationship between sleep and health.
However, it cannot establish that changing someone’s sleep stages will prevent disease. Existing health problems could themselves disrupt sleep, while other unmeasured factors (such as stress and diet) may influence both sleep and later diagnoses.
There is also a technical limitation. Wrist movements are not the same as directly measuring brain activity. Although SleepNet was trained and validated against laboratory sleep recordings, its ability to distinguish REM sleep, non-REM sleep and wakefulness from movement data was only moderate.
The algorithm’s REM estimates differed from laboratory measurements by an average of about 17 minutes, with other systematic differences also observed. And because participants wore the devices for just one week, the recordings may not fully represent their typical sleep patterns over many years. This is an important limitation of the study.
The study also relied on hospital inpatient diagnoses rather than all medical records, and its participants were overwhelmingly White, potentially limiting how broadly the findings apply.
Still, the research adds to growing evidence that sleep has more dimensions than simply the number of hours spent in bed.
For now, though, there’s no reason to start obsessing over the REM percentages displayed by a smartwatch. The more useful message remains familiar: get enough sleep, keep a reasonably consistent schedule and seek medical advice if sleep problems persist.
The study is published in the journal PLOS.
