The Sleep Stage Linked to 83 Better Health Outcomes

Key takeaways

• In 95,559 UK Biobank participants tracked for a median of 8.9 years, more REM sleep was associated with lower risk across 83 health outcomes, while more deep sleep was linked with lower risk across seven.

• The apparent low-risk range for most outcomes clustered around six to eight hours of nightly sleep—not necessarily the often-cited eight-hour target.

• Fragmented sleep and greater night-to-night irregularity were associated with worse long-term health patterns, suggesting that sleep continuity and consistency matter alongside total time asleep.

• This was an observational analysis based on seven days of wrist-based sleep tracking. It identifies associations, not proof that increasing a particular sleep stage will prevent illness.

Sleep is often reduced to a nightly score: how many hours you got, whether you hit a goal, and perhaps how often you woke up. This study takes a more granular view. It asks how the architecture of sleep—REM, deep, light sleep, continuity, and regularity—relates to health years later.

Researchers used seven days of accelerometer data rather than self-reported sleep from nearly 96,000 adults. That is important because people are not especially accurate at estimating either their sleep duration or how often they are awake overnight.

The result is a reminder that two people can spend the same number of hours in bed and have very different sleep exposures.

REM’s broad signal

REM sleep is the stage most closely associated with vivid dreaming, but it is also a period of unusually active brain physiology. In this analysis, participants with about 47.6 more minutes of REM sleep across the distribution had lower risk across 83 outcomes during follow-up.

Deep sleep had a narrower, but still meaningful, pattern: more of it was associated with lower risk across seven outcomes. That distinction is worth noting. REM and deep sleep are both often labeled “restorative,” but they may not map onto health in identical ways.

The practical implication is not to chase a specific REM percentage on a wearable. Consumer devices estimate sleep stages imperfectly, and the study does not show that manipulating REM directly changes future outcomes. It does suggest that a sleep routine supporting normal architecture—not just enough time in bed—may matter.

The six-to-eight-hour zone

Total sleep showed a curved relationship with long-term health. For 69 of the 86 outcomes linked with sleep duration, the lowest observed risk occurred within a six-to-eight-hour range.

The strongest adverse pattern appeared below five hours of sleep, which was associated with higher risk across 37 outcomes. More sleep was not automatically better, either. Very long sleep can reflect underlying health issues, time in bed without actual sleep, or other factors the analysis cannot fully disentangle.

The useful takeaway is not to obsess over landing on a precise number. It is to recognize that routinely sleeping well under six hours is unlikely to be a neutral trade-off.

Consistency has its own role

The study also found that greater night-to-night variability and more wakefulness after falling asleep were associated with less favorable health outcomes.

That matters because sleep timing is often the first thing sacrificed to an otherwise “healthy” routine: early workouts after late nights, variable weekends, travel, long workdays, or social catch-up. But a schedule that repeatedly shifts may create a different biological signal than simply sleeping fewer hours once in a while.

Sleep regularity is less glamorous than sleep tracking, supplements, or elaborate recovery routines. It may also be one of the more actionable variables in the data: keeping wake time reasonably consistent, protecting enough sleep opportunity, and reducing the things that repeatedly fragment the night.

A longevity takeaway 

The main reframe is that sleep quality is not a soft lifestyle metric. It is a multi-part physiological exposure: duration, timing, continuity, and stage distribution all appear to carry information about future health.

Still, the study cannot tell us whether poor sleep drives these outcomes, whether early biological changes disrupt sleep before they are clinically visible, or both. Sleep stages are also influenced by age, medications, alcohol, stress, sleep disorders, and many other variables.

This study reinforces a familiar message with much more detail behind it: a stable six-to-eight-hour sleep window, with relatively uninterrupted nights, appears to be a meaningful long-term health pattern.

The goal is not to optimize a wearable’s REM score. It is to create the conditions that allow normal sleep architecture to emerge—adequate sleep opportunity, a reasonably regular schedule, light exposure during the day, and fewer habitual disruptions close to bedtime.

Reference:
Luo J, Liu R, Yin J, et al. 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.