The short answer
Most healthy adults spend about 13 to 25 per cent of the night in deep sleep and 20 to 25 per cent in REM sleep. On an eight-hour night, that works out at roughly 60 to 110 minutes of deep sleep and 90 to 120 minutes of REM. There is no official per-stage target in sleep medicine: the guidance is for total sleep, which for adults is seven to nine hours on a reasonably regular schedule. A healthy brain divides up the stages by itself, provided it gets enough raw material.
That is the backdrop. The new finding is more surprising: in the largest study of its kind, researchers at Peking University linked the composition of your sleep, not just its length, to your chances of developing dozens of diseases years later. REM sleep was the standout, and very short sleep was the clearest warning sign.
The study, in one paragraph
Luo and colleagues at the School of Public Health at Peking University, working with Capital Medical University in Beijing, analysed wrist-worn accelerometer data from 95,559 UK Biobank participants aged 37 to 73. Each person wore an Axivity AX3 research tracker for seven consecutive days and nights. A deep-learning algorithm called SleepNet converted the movement recordings into estimates of REM sleep, deep sleep (N3), light sleep, total sleep duration, night-to-night irregularity, and wakefulness after falling asleep. The team then checked hospital diagnosis records for a median of 8.9 years and tested the links against 1,049 disease phenotypes built from more than 10,000 diagnostic codes. The paper, “Accelerometer-derived real-world sleep stages and risk of incident diseases: a UK Biobank cohort study and phenome-wide association analysis”, was published in PLOS Medicine on 17 September 2026.
What did they find about REM sleep?
REM was the strongest signal in the study. People with more estimated REM sleep developed dozens of diseases at lower rates over the follow-up period. Each interquartile-range increase in REM duration, about 47.6 minutes in this sample, was associated with a lower risk of 83 disease phenotypes, across 12 major disease categories.
Some of the specific figures: heart failure showed a hazard ratio of 0.74, meaning an estimated 26 per cent lower relative risk; atrial fibrillation came in at 0.83; ischaemic heart disease at 0.87; multiple sclerosis at 0.48; dementia at 0.54; and Parkinson’s disease at 0.20, the largest protective association in the analysis.
Two points of context matter before you read too much into those numbers. First, they compare the middle half of participants with each other; they do not mean that adding 48 minutes of REM will cut an individual’s risk by those percentages. Second, diagnoses came from inpatient hospital records, so milder conditions treated outside hospital are undercounted. The study was observational, and it cannot prove that extra REM prevents disease. Early, undetected illness can also degrade sleep, so some of the association may run in the other direction.
What about deep sleep?
Deep sleep showed a narrower but still notable pattern. More estimated deep sleep was associated with a lower risk of seven conditions: type 2 diabetes (hazard ratio 0.89), major depressive disorder (0.86), Parkinson’s disease (0.70), and sleep apnoea (0.78) among them. That is a smaller list than the 83 tied to REM, but the conditions it names, diabetes and depression in particular, affect a huge share of adults.
So how much deep sleep do you need? Sleep physiology references put deep sleep at 13 to 25 per cent of a typical adult night. On an eight-hour night, that is roughly 60 to 115 minutes, broken into several chunks in the first half of the night. Deep sleep is when your body does much of its physical repair: tissue maintenance, bone and muscle building, and immune support. If you consistently feel groggy despite a full night in bed, short or fragmented deep sleep is one of the things a clinician would investigate.
Is six hours of sleep enough?
This is where the study gives its clearest verdict. Total sleep duration showed a non-linear relationship with disease risk. People averaging six to eight hours per night had the lowest risk for 69 conditions, while those sleeping under five hours faced elevated risk for 37 conditions, including dementia, heart failure and type 2 diabetes. The researchers described the six-to-eight-hour window as a health safeguard for middle-aged and older adults, probably because a full window of that length lets the night fit in healthy amounts of both REM and deep sleep.
Six hours sits at the bottom edge of that window. It is markedly better than five, but it leaves little margin. And because REM periods grow longer in the later sleep cycles, a short night cuts REM disproportionately. The last cycle of the night is where the longest REM stretch usually sits, which is also the stretch your alarm is most likely to interrupt.
How are sleep stages actually measured?
One honest limitation: nobody in this study had wires glued to their scalp. Sleep stages were estimated from wrist movement using a machine-learning model, not measured with a polysomnography sleep lab. Movement-based estimates are reasonably good at telling total sleep from wake, but only moderately accurate at distinguishing REM from light sleep. Consumer trackers on your wrist work the same way, only usually with less data. Treat your wearable’s stage breakdown as an estimate, not a diagnosis. The study’s strength is its size and its objective measurement, not the precision of any single night’s readout.
What the study cannot tell you
Because this was an observational study, it cannot answer the question most readers actually want answered: if I change my sleep, will my risk change? The design observes patterns; it does not test interventions. It also captured only one week of sleep data, and sleep measured in middle age may partly reflect health conditions that had already begun. The authors are explicit about this, and any coverage that turns the findings into a prescription is getting ahead of the evidence.
What the study does add is the biggest objective dataset yet showing that how you sleep, not just how long, lines up with a wide range of later diagnoses. That is a reason to take persistent sleep problems seriously, not a reason to buy a gadget that claims to boost your REM.
What can you actually do with this?
If the 6-to-8-hour window and the REM findings sound like a description of your own short, irregular nights, the evidence-backed moves are unglamorous and familiar. Standard sleep guidance, such as the NHS’s sleep recommendations, emphasises a consistent bedtime and wake time, a cool, dark, quiet bedroom, and cutting back on caffeine and alcohol in the hours before bed. Alcohol is worth singling out: it fragments sleep and suppresses REM in the second half of the night, which is exactly when your longest REM stretches would have happened. Daytime exercise and morning daylight exposure help anchor your body clock, which keeps the sleep stages you want in the parts of the night where they belong.
If you are sleeping seven to eight hours, keeping a regular schedule, avoiding alcohol and caffeine late in the day, and still waking with significant brain fog, low mood, or memory problems, talk to your doctor. Sleep apnoea and other treatable disorders steal deep sleep specifically, and a clinical sleep evaluation can find them in a way a wrist tracker cannot.
The bottom line
The question “how much deep sleep do you need” has a duller answer than the internet wants: about one to two hours on a full night, which your brain will arrange for you if you give it seven to nine hours of regular sleep. The interesting part of this study is not the minute counts. It is that a 95,000-person dataset, measured with trackers instead of questionnaires, found the architecture of your sleep, how much REM and deep sleep it contains, and how regular it is, tracking alongside dozens of diseases across nearly a decade. Hours get you in the game. Regularity and full nights are what let the stages do their work.
Frequently asked questions
How much deep sleep do you need per night?
Healthy adults typically spend 13 to 25 per cent of the night in deep sleep, roughly 60 to 115 minutes on an eight-hour night. There is no official per-stage target; sleep medicine sets targets for total sleep, seven to nine hours for adults, and a healthy brain apportions the stages on its own.
Is 6 hours of sleep enough?
In the 2026 PLOS Medicine study of 95,559 adults, six to eight hours of nightly sleep was the window with the lowest risk across 69 conditions. Six hours sits at the bottom of that window: better than five, but with little margin, and it risks cutting the long REM stretches of the final sleep cycle.
Can you get more REM sleep?
Because REM stretches grow longer in later sleep cycles, the most effective way to get more REM is to sleep long enough to complete five to six full cycles, about seven to eight hours, and to wake without an alarm where possible. Keeping a consistent bedtime and avoiding alcohol close to bedtime also protect REM.
Does more REM sleep prevent disease?
No study has shown that. The 2026 Peking University analysis was observational: more REM was associated with lower rates of 83 diagnoses over 8.9 years, but the design cannot show cause and effect, and early illness can itself degrade sleep.
Sources
- Luo J, Liu R, Yin J, Cao W, Sun S, Chen R. Accelerometer-derived real-world sleep stages and risk of incident diseases: a UK Biobank cohort study and phenome-wide association analysis. PLOS Medicine. 2026;23(9):e1005213. DOI: 10.1371/journal.pmed.1005213.
- NHS guidance on sleep: adults need seven to nine hours of sleep a night.
- Sleep stage share norms from NIH physiology references: N1 ~5%, N2 ~45 to 55%, N3 (deep) 13 to 25%, REM 20 to 25% of total sleep.