Reading time: 9 minutes

Overview

1. A new perspective on sleep and health
2. What is REM sleep?
3. How the study was conducted
4. What the researchers found
5. What the findings mean – and what they do not
6. How accurately can wearables measure sleep stages?
7. What you can take away for daily life
8. Frequently asked questions
9. Conclusion

A new perspective on sleep and health

We all know the feeling: you have slept for seven hours, yet you still do not feel properly refreshed in the morning. For many of us, sleep duration is the main measure of how good the night was. But it tells us little about sleep quality, which may be just as important. Sleep quality includes, for example, how uninterrupted your sleep was and how it was structured across the night.

While we sleep, we move through several sleep stages controlled by the brain. Light sleep, deep sleep and rapid eye movement sleep occur in cycles that repeat several times during the night. Two people can both sleep for seven hours and still have very different nights. One may sleep largely undisturbed and wake feeling refreshed; the other may wake repeatedly and feel exhausted the next morning.

A study published in PLOS Medicine in September 2026 investigated how such differences relate to long-term health. The research team analysed sleep patterns from 95,559 UK Biobank participants and compared them with the subsequent occurrence of more than one thousand health conditions.

The most striking finding was that a greater proportion or longer duration of estimated REM sleep was associated with a lower risk of numerous later diagnoses. The analysis also indicated that total sleep duration, regularity and time spent awake during the night may matter.

What is REM sleep?

REM stands for rapid eye movement, referring to the quick eye movements that occur beneath the closed eyelids during this stage of sleep. The brain is highly active, while most skeletal muscles are temporarily inhibited. Particularly vivid dreams are common during REM sleep.

REM sleep is associated with the processing of emotions and experiences, as well as with learning and memory. These functions cannot, however, be assigned exclusively to one sleep stage. Light and deep sleep also matter. This is why trying to maximise only the REM score shown by a sleep tracker is unlikely to be helpful.

REM periods generally become longer later in the night. People who regularly shorten their sleep by getting up earlier may therefore miss or interrupt these later REM-rich cycles. This helps illustrate why allowing sufficient time for sleep as a whole is more useful than chasing one individual sleep-stage value.

How the study was conducted

The research team used data from the UK Biobank, a large British long-term health database. The current analysis included 95,559 predominantly middle-aged and older adults.

Participants wore an accelerometer on their wrist around the clock for seven days. A machine-learning model called SleepNet used the movement data to estimate several sleep characteristics:

• total sleep duration,
• estimated time spent in REM, light and deep sleep,
• time awake after first falling asleep,
• day-to-day variation in sleep duration.

The team then used medical records to establish which new conditions occurred after the measurement period. Participants were followed for a median of 8.9 years, and associations were tested across 1,049 conditions. The researchers also accounted for numerous possible influencing factors, including age, sex, BMI, smoking, alcohol intake, physical activity, socioeconomic characteristics, night-time noise and air pollution. This strengthens the analysis, although unknown or imprecisely measured influences can never be ruled out completely.

What the researchers found

After the research group applied stringent statistical corrections, 156 associations between sleep characteristics and later diagnoses remained. Estimated REM sleep accounted for a particularly large proportion of them.

1. More estimated REM sleep was associated with lower risk

An additional 47.6 minutes of estimated REM sleep was associated with a lower risk of 83 conditions across twelve categories. These included heart failure, atrial fibrillation, ischaemic heart disease, dementia, Alzheimer's disease and Parkinsonian syndromes.

The strength of the statistical association varied by condition. The hazard ratio was 0.83 for atrial fibrillation, 0.74 for heart failure and 0.69 for Alzheimer's disease. In this analysis, a hazard ratio below 1 means that the diagnosis occurred less often during follow-up among people with a longer estimated REM-sleep duration. In relative terms, these values correspond to 17%, 26% and 31% lower statistical risks, respectively.

They do not predict an individual's personal risk and they do not prove that more REM sleep prevents these conditions. A controlled intervention study would be needed to investigate causality.

2. Deep sleep also showed relevant associations

More estimated deep sleep was associated with a lower risk of seven conditions, including type 2 diabetes, major depressive disorder, sleep apnoea and Parkinson's disease. The pattern was less extensive than for REM sleep, but still notable.

3. Six to eight hours was often the lowest-risk range

Total sleep duration also made a difference. For 69 of the conditions studied, statistical risk was lowest among participants who slept for between six and eight hours. Elevated risks appeared particularly often among people sleeping for fewer than five hours.

This does not prove that six hours is optimal for everyone. Sleep requirements, age, health and personal circumstances vary. Rather, the findings suggest that extreme deviations – especially very short sleep – should not automatically be considered harmless.

4. Regularity and uninterrupted sleep also mattered

Greater day-to-day variation in sleep duration and more time awake after first falling asleep were also associated with higher risks for several conditions. This suggests that a stable rhythm and limited sleep fragmentation may be relevant aspects of sleep health.

What the findings mean – and what they do not

The large sample, long follow-up and sensor-based measurement are important strengths. Nevertheless, this was an observational study. It can reveal that particular sleep patterns and later health conditions occur together more or less frequently. It cannot establish that one caused the other.

The findings therefore do not mean that everybody sleeping for five hours will develop several diseases. They describe differences in statistical risk across a large group, not certain outcomes for individual people.

Several explanations are possible:

• A stable sleep architecture might support biological recovery and regulation.
• An emerging but not yet diagnosed condition might already be changing sleep. In that case, poorer sleep could partly be a consequence rather than a cause.
• Other factors – such as medication, psychological strain or existing health problems – could influence both sleep and disease risk.

The researchers adjusted their analysis for many potential influences. They also repeated the calculation after excluding conditions diagnosed within the first two years following sleep measurement. Ninety-five of the original 156 associations remained. This suggests that the results cannot be explained solely by illnesses already developing at the time of measurement. It still does not prove that the sleep patterns themselves were the cause.

How accurately can wearables measure sleep stages?

The measurement method is essential when interpreting the study. A wrist sensor records movement, but not brain waves, eye movements or muscle tone. These variables are measured during polysomnography in a sleep laboratory, which remains the gold standard for determining sleep stages.

SleepNet was trained using more than 1,100 nights recorded alongside polysomnography and was validated in several groups. Even so, agreement when distinguishing wakefulness, REM and non-REM stages remained moderate. Compared with polysomnography, the model underestimated REM duration by approximately 17 minutes on average.

The values in the study are therefore not a minute-by-minute map of each participant's sleep architecture. Yet estimates of this kind can still be useful in very large studies because they capture real-world sleep patterns across several nights rather than one artificial night in a laboratory.

The same caution applies to consumer smartwatches. One low REM reading is not a medical finding. Longer-term patterns, how you feel and any symptoms are more important.

What you can take away for daily life

The study does not reveal a particular “REM hack”. It does, however, reinforce several sensible foundations that support sufficient and relatively uninterrupted sleep.

1. Allow enough time for the whole night

Regularly sleeping for a very short time reduces not only total sleep, but may also cut short later REM-rich cycles. Rather than trying to optimise REM sleep in isolation, begin by creating a realistic window for enough total sleep.

2. Keep wake and bedtimes reasonably consistent

A regular schedule supports your circadian rhythm – your internal body clock. Your times do not need to be identical to the minute, even at weekends. Large and frequent shifts, however, make it harder for the body to maintain reliable timing cues.

3. Do not overlook night-time interruptions

Frequent or prolonged awakenings can disrupt sleep continuity even when your total time in bed appears sufficient. A quiet, dark and comfortably cool bedroom can help create better conditions for uninterrupted sleep.

4. Use your smartwatch as a rough guide, not a diagnosis

Look primarily at trends over several weeks rather than judging individual nights or readings. If chasing the perfect sleep score begins to create pressure, the tracker is no longer helping. Pay attention to how you actually feel during the day.

5. Seek advice for persistent symptoms

Regular loud snoring, witnessed pauses in breathing, morning headaches, severe daytime sleepiness or persistently unrefreshing sleep can indicate a sleep disorder. These symptoms should be discussed with a healthcare professional, regardless of what your watch displays.

Frequently asked questions about REM sleep

How much REM sleep is normal?

We usually pass through the different sleep stages four to six times during a night. In adults, REM sleep averages around 20% to 25% of total sleep time. During eight hours of sleep, that would be approximately 1½ to 2 hours. REM periods generally become longer towards morning.

These are reference values, not a target for every night. The proportion of REM sleep varies between individuals and from one night to the next. Age, medication, alcohol and health can all have an effect. A single REM value from a smartwatch therefore tells you very little, especially because wearables only estimate sleep stages.

Can I deliberately increase my REM sleep?

The study does not provide a reliable method for increasing REM sleep by a particular amount. It observed relationships between sleep patterns and later diagnoses; it did not test interventions designed to change REM sleep.

What you can influence are the general conditions for sleep: allow enough time, aim for a reasonably consistent schedule and pay attention to frequent night-time interruptions. More displayed REM sleep is not automatically better, and your watch does not measure it with minute-by-minute accuracy.

Does little REM sleep mean I will become ill?

No. One low reading does not provide reliable information about your individual risk. The study describes statistical associations across a large population over many years. It is not a prediction model for individual people.

Are the sleep stages shown by my smartwatch reliable?

Wearables can help reveal longer-term trends, but they do not replace an assessment in a sleep laboratory. Their classification of individual sleep stages is less accurate than their basic recording of movement and rest periods.

Are six to eight hours optimal for everyone?

No. This was the range associated with the lowest statistical risk for many conditions in this particular study. It is not a personalised sleep recommendation. Many adults need seven hours or more, and individual requirements vary.

Conclusion: Sleep quality is more than sleep duration

This new UK Biobank study offers an unusually broad view of how sleep patterns relate to long-term health. More estimated REM sleep was associated with a lower risk of 83 health conditions. Deep sleep, regularity and time awake during the night also showed relevant associations.

The findings do not turn REM sleep into a protective treatment or a tracker into a diagnostic device. They do, however, broaden the picture: healthy sleep is not simply about reaching a particular number of hours. It also involves sufficient opportunity, a reasonably stable rhythm and as few disturbances as possible.

Perhaps the most important takeaway is this: do not try to optimise the REM score of one night. Create the conditions that allow your body to move through its natural sleep architecture regularly.

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.

National Heart, Lung, and Blood Institute. Healthy Sleep Habits. NHLBI, National Institutes of Health.

Centers for Disease Control and Prevention. About Sleep. CDC.