More REM and deep sleep tracked with lower risk of dozens of diseases in 95,559 people

Advice about sleep almost always comes down to a number of hours. But a night is not one thing: it cycles through light sleep, deep sleep and REM, the stage where most dreaming happens. A study of 95,559 UK adults who wore wrist monitors for a week, then were followed for nearly nine years, asked whether the makeup of a night matters as much as its length.

It looks like it might. People who got more REM sleep went on to be diagnosed with fewer illnesses across a striking range of conditions, and those who got more deep sleep did better on a smaller set. Both findings come with a large caveat about what an observational study can show.

Why look at sleep stages rather than sleep hours?

The gold standard for measuring sleep stages is polysomnography, an overnight lab test wired to the scalp, face and chest. As the authors note, its high cost and technical complexity make it impractical for studies of tens of thousands of people.

Wrist accelerometers, the research-grade cousins of fitness trackers, are cheap enough for big cohorts but measure movement rather than brain waves. The workaround here is an algorithm called SleepNet, which infers sleep stages from the movement signal and was validated against 1,113 nights of laboratory-based polysomnography recordings. That makes the stages in this study an educated estimate, not a direct reading.

How was the sleep study run?

Writing in PLOS Medicine, the researchers took a week of wrist-monitor data from 95,559 UK Biobank participants, middle-aged and older adults. From it they derived REM, the three stages of non-REM sleep, total sleep duration, how irregular sleep timing was, and how much time people spent awake after first falling asleep.

Then they waited. Participants were followed from the date they wore the monitor until they were diagnosed with something, died, or reached April 2024, with a median follow-up of 8.9 years. Rather than picking a few diseases in advance, the team screened 1,049 health outcomes at once, an approach that casts a wide net but also means some associations will turn up by chance.

Which diseases tracked with more REM sleep?

Of the 1,049 conditions, sleep patterns were linked to 156. REM sleep carried by far the most associations: more of it went with lower risk of 83 conditions.

Some of the strongest were neurological. Comparing people at the top of the REM range with those at the bottom, risk was lower for parkinsonism, multiple sclerosis, dementias and Alzheimer’s disease, and also for atrial fibrillation, an irregular heart rhythm, and ischemic heart disease.

Deep sleep, measured per extra 47.5 minutes, was linked to lower risk of 7 conditions, including type 2 diabetes, major depressive disorder, sleep apnea and Parkinson’s disease. Light sleep cut both ways: more of it went with higher risk of depression but lower risk of heart failure.

What about sleep timing and total hours?

Two patterns pointed the other way. Greater sleep irregularity, meaning bedtimes and wake times that moved around, was linked to higher risk of 3 conditions, and more time awake after falling asleep to higher risk of 6.

Total sleep duration mattered too, and not in a straight line. For 86 conditions the relationship curved, with the lowest-risk point usually falling between 6 and 8 hours. The extremes were where the trouble clustered: people sleeping under 5 hours accounted for 37 of the 41 harmful associations the researchers found when they compared duration groups against a 6 to 8 hour reference.

Could the illness be changing the sleep, rather than the reverse?

This is the question that decides how much the headline is worth, and the honest answer is that this study cannot separate the two. The authors are direct about it: the main limitation is the observational design, which remains susceptible to residual confounding and precludes causal inference.

The concern is sharpest for the brain findings. Conditions such as Parkinson’s disease and dementia can disturb sleep years before anyone is diagnosed, so low REM could be an early symptom rather than a cause. The same logic applies to sleep apnea, which by definition fragments the night.

There is also the measurement question. A week of wrist data scored by an algorithm is a snapshot, and people’s sleep changes with seasons, work and life.

So what does this mean for your own sleep?

Nothing here tells you how to get more REM, and no product can sell it to you. Sleep stages are not something you can choose; they emerge from getting enough sleep, regularly, without interruption. What this study adds is a reason to care about the shape of the night and not just the hours in bed.

The practical advice is the familiar kind. MedlinePlus recommends 7 to 9 hours a day for adults and notes that it’s not just the number of hours of sleep you get that matters, because interrupted sleep can cost you stages you need. Keeping a consistent schedule is the one lever in this study’s list that is genuinely in most people’s hands.