Last week’s headline claimed that more REM sleep is linked to a lower risk of 83 diseases. Researchers estimated this REM sleep using wrist movement data.
What?
A PLOS Medicine study published on September 17 tracked 95,559 UK Biobank adults for about 9 years. Each person wore a wrist sensor for a week, and an algorithm used their movement to estimate sleep stages. For every extra 47.6 minutes of estimated REM, there was a 46% lower risk of dementia and a 26% lower risk of heart failure. This is just an association, not proof. Also, the 80% drop in risk that some outlets reported for “Parkinson’s” actually refers to parkinsonism, which is a broader diagnosis.
So what?
These numbers have two main problems.
First, consider the wrist measurements. Compared with a sleep lab, the algorithm matched only moderately well and tended to undercount REM by about 17 minutes each night. The authors themselves call these numbers “movement-based estimates.” So, remember that your own wrist REM score is just an estimate too.
Second, there’s the timing. The strongest links are for dementia and parkinsonism, which are diseases that can disrupt sleep years before they are diagnosed. Wrist sensors can pick up more night waking as early as seven years before someone is diagnosed with Parkinson’s. The study authors removed cases from the first two years, and the links with dementia and parkinsonism still showed up. But two years is a short window for diseases that can develop over a decade or more. Also, a 2025 study combined data from five groups whose sleep was measured with electrodes (4,657 people in total) and found no link between sleep stages and dementia.
My read: this study does not show that increasing REM prevents disease. I wouldn’t rearrange a single night to improve a watch score on this evidence. Two things have been tested in sleep labs, and both cut REM: short nights and alcohol.
Now what?
Try to get enough sleep each night. In a 2018 sleep-lab study, 35 adults who were limited to six hours in bed for a week lost about half an hour of REM sleep each night on average, but their deep sleep stayed about the same. REM sleep is longest in the last third of the night, which early alarms often cut short. If you have an early flight, try to go to bed earlier.
Avoid drinking alcohol the night before an important day. A 2025 review of 27 studies found that REM sleep dropped even with low amounts of alcohol, which the authors define as about two standard drinks or less.
Don’t focus on your tracker’s nightly REM number. Instead, focus on your average sleep time over the week and try to wake up at the same time every day.
If your partner notices that you act out your dreams by punching, kicking, or shouting, talk to your doctor. REM sleep behavior disorder can show up years before Parkinson’s or similar conditions. It’s important to get it checked out.
Your watch will keep giving you a REM score, but you don’t have to pay attention to it.
Stay healthy.
Andre
Sources: PLOS Medicine paper · Skorucak, Sleep 2018 (PDF) · Yiallourou, Sleep 2025 · Gardiner, Sleep Med Rev 2025 · Postuma, Brain 2019 · ScienceDaily headline