Variations in accelerometer-derived real-world sleep patterns were associated with 156 incident diseases across multiple systems, with higher REM sleep linked to lower risk of 83 diseases and extreme short sleep (<5 hours) accounting for 37 of 41 significant adverse associations compared to a 6–8 hour reference group.
Key Findings
Results
Higher amount of REM sleep was associated with lower risks of 83 diseases across multiple organ systems.
Data were derived from wrist-worn accelerometer recordings in 95,559 UK Biobank participants.
Sleep stages (REM, N1, N2, N3) were estimated using the SleepNet algorithm applied to accelerometer data.
Median follow-up was 8.9 years, with follow-up ending at first disease occurrence, death, or April 1, 2024.
Cox proportional hazard regression was used with adjustment for demographic characteristics, lifestyle factors, and environmental exposures.
A phenome-wide association analysis (PheWAS) was conducted across 1,049 health outcomes.
Results
Higher amount of deep (N3) sleep was associated with lower risks of 7 diseases.
Deep sleep (N3) showed a more limited but still significant protective association compared to REM sleep.
Analysis covered all 1,049 health outcomes in the PheWAS framework.
Adjustments included demographic characteristics, lifestyle factors, and environmental exposures.
The SleepNet algorithm was used to derive N3 stage durations from wrist accelerometer data.
Results
Greater sleep irregularity was associated with elevated risks of 3 diseases, and increased wakefulness after sleep onset (WASO) was linked to elevated risks of 6 diseases.
Sleep irregularity and WASO are measures of sleep fragmentation and disruption derived from accelerometer data.
These associations were identified within the same PheWAS framework covering 1,049 outcomes.
Cox proportional hazard regression was adjusted for demographic, lifestyle, and environmental factors.
Overall, variations in sleep patterns were associated with 156 incident diseases in total.
Results
Significant non-linear relationships between sleep duration and 86 disease phenotypes were identified, with minimum-risk hours for 69 phenotypes concentrated within the 6–8 hour window.
Restricted cubic spline (RCS) analyses were used to assess non-linear dose-response relationships.
Non-linearity was determined with a significance threshold of P for nonlinear <0.05.
86 of the 1,049 examined phenotypes showed significant non-linear associations with sleep duration.
For 69 of those 86 phenotypes, the minimum disease risk fell predominantly within the 6–8 hours sleep duration range.
Results
Individuals with extreme short sleep (<5 hours) showed the most widespread clinical vulnerabilities, accounting for 37 of 41 significant adverse associations compared to the 6–8 hour reference group.
Category-specific analyses compared sleep duration groups against a 6–8 hour reference group.
Extreme short sleepers (<5 hours) accounted for 37 out of 41 identified significant adverse associations.
This finding highlights <5 hours as the sleep duration category with the broadest disease risk profile.
Analyses used Cox proportional hazard regression with full covariate adjustment.
Methods
The study used the SleepNet algorithm applied to wrist-worn accelerometer data from 95,559 UK Biobank participants to derive objective, real-world sleep stage metrics.
Key metrics derived included REM, N1, N2, and N3 stage durations, total sleep duration, sleep irregularity, and WASO.
Participants were middle-aged and older adults from the UK Biobank cohort.
Median follow-up length was 8.9 years.
The PheWAS framework examined associations with 1,049 health outcomes across multiple disease categories.
The observational design was noted as the main limitation, precluding causal inference and remaining susceptible to residual confounding.
Conclusions
Maintaining 6–8 hours of sleep was associated with a more favorable sleep architecture and lower disease risk across multiple systems.
This conclusion was drawn from both the PheWAS and RCS analyses combined.
Favorable sleep architecture in this range was inferred from the concentration of minimum-risk phenotypes within the 6–8 hour window.
The finding applied across multiple organ systems and disease categories.
The authors note this offers insights for prevention and health promotion.
What This Means
This research suggests that the way we sleep — not just how long, but how deeply and how regularly — is linked to the risk of developing a wide range of diseases. Using wrist-worn activity trackers on nearly 96,000 adults from the UK Biobank, researchers used a sophisticated algorithm to estimate sleep stages (including light sleep, deep sleep, and REM or 'dreaming' sleep) and tracked participants' health outcomes over nearly nine years. They then systematically tested whether these sleep measurements were associated with the onset of over 1,000 different health conditions. Overall, variations in sleep patterns were linked to 156 different diseases.
The strongest protective pattern came from REM sleep: people with more REM sleep had lower risks for 83 diseases. Deep sleep also showed protective associations, though for fewer conditions. On the other side, fragmented sleep — measured by wakefulness after falling asleep and irregular sleep schedules — was associated with higher risks of several diseases. For sleep duration, the research found that 6–8 hours per night appeared to be the 'sweet spot' for the lowest disease risk across most health conditions examined. People sleeping fewer than 5 hours per night had the most widespread health vulnerabilities, being linked to significantly higher risks for 37 out of 41 identified adverse health associations.
This research matters because it goes beyond earlier studies that relied on people's own estimates of their sleep and instead uses objective, real-world measurements of sleep stages — something that has been difficult to study at large scale outside of a sleep clinic. This suggests that both the quality of sleep (especially getting enough REM and deep sleep) and the quantity (aiming for 6–8 hours) may be important for long-term health across many different body systems. Because this was an observational study, it cannot prove that poor sleep directly causes these diseases, and other unmeasured factors could play a role. Nonetheless, the findings support the value of prioritizing consistent, sufficient, and restful sleep as part of broader health promotion efforts.
Check Your Own Numbers
Upload your bloodwork. We'll cross-reference your results against this study and 4,700 others.
Luo J, Liu R, Yin J, Cao W, Sun S, Chen R. (2026). Accelerometer-derived real-world sleep stages and risk of incident diseases: A UK Biobank cohort study and phenome-wide association analysis.. PLoS medicine. https://doi.org/10.1371/journal.pmed.1005213