UK study of 95,000+ people: more REM sleep linked to lower risk of 83 diseases, deep sleep just 7

83 versus 7. Those are the two numbers from a large UK sleep study: people with more rapid eye movement (REM) sleep had a lower risk of 83 diseases, while more deep sleep corresponded to only 7. The figures are striking, but read on and you will find that the limitations the study lists itself are what is most worth remembering.

The paper, "Accelerometer-derived real-world sleep stages and risk of incident diseases", was written by six authors led by Jingsong Luo and published in PLOS Medicine on 17 September 2026. From 502,129 participants in UK Biobank, the team analysed data from 95,559 people, with a mean age of 56.2 years and a median follow-up of 8.9 years.

The study shows association, not causation.

Seven days on the wrist, nearly nine years of follow-up

Participants wore an Axivity AX3 triaxial accelerometer on the wrist for seven days. Researchers then followed them until their first diagnosis, death, or 1 April 2024.

Crucially, sleep stages were not measured in a sleep laboratory. The team used a deep learning model called SleepNet to estimate each night's sleep stages from wrist activity. The model was trained on data from more than 1,100 nights of polysomnography (PSG, the multi-channel physiological monitoring commonly used in sleep labs).

The paper itself reports an F1 score of 0.49 for the model's classification of REM, non-REM and wake, and an average difference of 17.1 minutes in REM time compared with PSG. The authors also acknowledge that the agreement between the two is only modest.

1,049 diseases compared one by one

The study used a phenome-wide association analysis, comparing sleep measures against 1,049 disease phenotypes one by one. After Bonferroni correction for multiple comparisons, 156 associations were statistically significant: 97 involving sleep stages, 50 total sleep duration, 6 wake time after sleep onset and 3 sleep irregularity.

More REM sleep was linked to lower risk of 83 diseases, and more deep sleep to 7. According to Medical News Today on 18 September 2026, the former include heart failure, high blood pressure, atrial fibrillation, dementia and multiple sclerosis; the latter include type 2 diabetes, major depressive disorder, sleep apnoea and Parkinson's disease.

Association is not causation, and the authors say so

This is an observational study. The authors state that the design cannot establish causation and that confounding may not have been fully excluded. The medical news site Healio, in its report of 23 September 2026, likewise quoted the authors as saying that residual confounding and reverse causation cannot be fully ruled out.

Reverse causation means that poor sleep may not be causing illness; an undiagnosed disease may already have been disturbing sleep. The team tested this: after excluding cases that arose in the first two years of follow-up, only 95 of the original 156 associations remained significant, suggesting some results may be affected by reverse causation.

Other limitations listed by the authors include:

• Seven days of measurement may not represent a person's long-term sleep habits.
• Sleep stages are estimated from wrist activity and are not equivalent to PSG measurement.
• Disease outcomes were drawn only from hospital records.
• Participants show healthy-volunteer bias and are mainly white Europeans of higher socioeconomic status.

Two independent readings from experts

On the day the paper was published, the UK's Science Media Centre gathered expert comments. Dr Nina Rzechorzek of the University of Cambridge noted that sleep stages were inferred by a deep learning model from wrist activity rather than measured by PSG, and that reverse causation is an important consideration. She also said the design cannot show that increasing REM or deep sleep prevents disease; and she considered the press release's claim that sleeping 6 to 8 hours effectively lowers disease risk to go beyond what the design can support.

Dr Greg Elder of Northumbria University offered a two-sided assessment. On one hand, he said wrist-worn activity devices cannot yet accurately determine which sleep stage a person is in overnight, so the study can only show association; on the other, he considered seven days of objective measurement better than most sleep studies, which rely on participants' self-reports.

Read the study as a lifestyle cue, not a prescription

The study used a wrist-worn accelerometer, and the authors themselves concede that agreement between wrist-based sleep-stage estimates and PSG is limited. Nor did the study show that lengthening REM time reduces disease. A single night's wrist-estimated figure should therefore not be over-interpreted.

Notably, sleep irregularity also appears on the list of associations. According to Medical News Today, it was linked to abdominal pain, anxiety and major depressive disorder. In the same report, outside expert Scott Nass offered two simple suggestions:

• Get up at the same time every day, weekends included.
• Get more daylight in the morning.

Medical News Today also quoted the study authors' conclusion that 6 to 8 hours of sleep a night is ideal; however, this is precisely the part that Rzechorzek considers beyond what the design can support, so treat it as a rough guide.

This article provides general health information only and is not a diagnosis. If you have long-term insomnia, snoring or unusual daytime tiredness, consult a doctor. Sleep-stage numbers are worth a glance, but a regular routine is yours to control.

Prices and details correct as of October 2026

This article is general health information, not medical advice. For any health concern, consult a doctor or registered health professional.