September 17, 2026

expert reaction to REM/deep sleep, not just sleep duration, linked with risk of several diseases

A study published in PLOS Medicine looks at sleep types (REM/deep sleep) and risk of diseases. 

Dr Nina Rzechorzek, MRC Clinician Scientist | Brain Temperature & Neural Circadian Rhythms, University of Cambridge, said:

“This is a large, carefully analysed prospective study, with important strengths including nearly 96,000 participants, almost nine years of follow-up, stringent correction for testing more than 1,000 health outcomes, and several sensitivity analyses. However, the headline findings about REM and deep sleep need an important qualification: these sleep stages were inferred by a deep-learning model from wrist movement, rather than measured directly using gold-standard polysomnography. The paper reports only moderate performance for the model’s REM/NREM/wake classification, and the authors themselves emphasise that these should be regarded as movement-based estimates rather than exact physiological sleep stages.

Reverse causation also remains a significant concern. When the authors extended the exclusion period to two years, only 95 of the original 156 significant associations remained, suggesting that early or undiagnosed disease may already have been influencing sleep before a clinical diagnosis was made.

Seven consecutive days of wearable monitoring is valuable for studying sleep at population scale, but it cannot fully capture an individual’s habitual sleep over many years. The authors acknowledge this, alongside the possibility of residual confounding and the limited representativeness of the UK Biobank cohort.

Overall, the results support sleep patterns as potentially informative markers of future disease risk, but they do not show that increasing REM or deep sleep, or simply forcing sleep into a 6–8-hour window, will prevent disease. The press release does state that this is an observational study, but the quoted claim that maintaining 6–8 hours of sleep ‘can effectively reduce’ disease risk goes beyond what this study design can establish.”

Dr Greg Elder, Assistant Professor in Psychology, Associate Director of Northumbria Centre for Sleep Research, Northumbria University, said:

“The press release is accurate although it perhaps does not reflect the main limitation of the study – that movement-based watches can’t yet accurately determine what sleep stage we are actually in overnight.  I think this research is good quality because it adds to the increasing number of studies that show that how much sleep we actually objectively get and not just how much sleep we say that we get (which is often different) is very important from a health perspective. Most real-world population studies of this nature tend to look only at self-report sleep duration, rather than actual sleep duration, or how we actually sleep – sleep is a very complex behaviour. The fact that sleep was measured over seven consecutive days is a strength as this is generally quite representative of our habitual sleep patterns.

I think there are two main limitations to this study. The first is that the study can only show that these aspects of sleep are associated with health conditions – it can’t tell us whether sleep alterations can cause all of these health conditions. Although sleep disturbances (such as short sleep) are often seen in a wide range of health conditions, we don’t know if these sleep disturbances cause, or are simply associated with, these conditions. This is normal for this type of study and is acknowledged in the press release. Future work needs to look at where sleep alterations fit into the underlying mechanisms of each of these health conditions. The second is that we can’t say for sure whether people were actually in rapid eye movement (REM) sleep, or deep sleep, based on wristwatch accelerometer information. Although wristwatches can do a good job of working out whether or not we are asleep, or awake, they don’t explicitly measure eye movements, or brain activity, and we need that information to know for sure whether or not people were actually in REM sleep, or slow-wave sleep (deep sleep). In this study, the research team used a machine-learning algorithm to estimate this, which isn’t the same as actually knowing when people are in light sleep, REM sleep or deep sleep. This is most likely because the data were collected from the large-scale UK Biobank study which is not specifically focused on sleep and didn’t have that information available.

To me, the real-world implications are that the sleep that we actually get is just as important for our health as the sleep that we think we get – in terms of the duration, and timing. This study adds to the evidence to show that a regular sleep pattern, where we get enough sleep, is important for our physical and psychological health, as we get older. Future work needs to understand if sleep causes, or reflects, these health alterations, using reliable, validated sleep measurement tools.”

‘Accelerometer-derived real-world sleep stages and risk of incident diseases: A UK Biobank cohort study and phenome-wide association analysis’ by Jingsong Luo et al. was published in PLOS Medicine at 19:00 UK time on Thursday 17th September. 

DOI: 10.1371/journal.pmed.1005213

Declared interests

Dr Greg Elder: “I have no conflicts of interest to declare.”

For all other experts, no reply to our request for DOIs was received.