News · Sleep
Both short and long sleep tracked with Alzheimer-type brain change
Using brain scans from 38,816 UK Biobank participants, researchers found a U-shaped pattern: five to six hours and nine to ten hours both lined up with more Alzheimer-like shrinkage and worse test scores.
Based on brain imaging in 38,816 UK Biobank participants plus an electronic health record analysis
- 38,816 UK Biobank participants with brain scans, plus separate health-record analysis.
- The curve was U-shaped: 5 to 6 hours and 9 to 10 hours both looked worse than the middle.
- The measure was a machine-learning score for Alzheimer-type shrinkage, not a diagnosis.
- Insomnia and hypersomnia both carried higher 10-year risk in the records analysis.
- Observational and self-reported sleep, so it shows a link and cannot show direction.
Sleep advice usually runs in one direction: most people need more. A study in Alzheimer’s & Dementia is a reminder that the relationship bends at both ends.
Among 38,816 participants in the UK Biobank, the researchers compared how long people said they slept against a brain scan measure built to detect the specific pattern of shrinkage seen in early Alzheimer’s disease.
What was measured
The imaging outcome is a machine-learning-derived magnetic resonance imaging (MRI) signature of AD-related atrophy, where AD is Alzheimer’s disease. The measure is named SPARE-AD, and it condenses a whole brain scan into one number describing how closely that brain resembles the early-Alzheimer’s pattern.
It is a resemblance score, not a diagnosis. Plenty of people score high and stay well.
Alongside it, participants took three standard cognitive tests covering processing speed, attention and the ability to switch between tasks. Prior research had linked sleep to dementia diagnoses; what was unclear was whether duration tracked the brain shrinkage itself.
What they found
Sleep duration exhibited a U-shaped association with SPARE-AD, and the same shape appeared on all three cognitive tests.
Both tails looked worse than the middle. Short (5-6 hours) and long (9-10 hours) sleep duration were associated with poorer cognitive performance.
That is the whole finding, and its shape is the reason it is interesting. A straight line would suggest more sleep is always better. A U says there is a middle, and that both ends of the range mark something.
The second dataset
The researchers then ran a separate analysis on electronic health records, using recorded diagnoses instead of what people said about themselves.
Both insomnia and hypersomnia showed a high risk for Alzheimer’s disease over ten years, hypersomnia meaning excessive sleepiness or sleeping far longer than usual.
Two different datasets and two different ways of measuring, pointing the same way, is a stronger position than either alone.
What this does not settle
The direction of the arrow, which is the question everyone actually wants answered.
Alzheimer’s disease begins changing the brain a decade or more before anyone notices symptoms, and those changes disrupt sleep. So poor sleep in this data may be an early consequence of disease rather than a cause of it. Nothing in an observational study can separate those.
The long-sleep half is especially exposed to that problem. Very few people choose to sleep ten hours. It usually reflects depression, illness, medication or sleep so broken that it takes longer to get enough, and any of those could be doing the work.
Self-reported sleep is the other soft spot. People estimate their own sleep badly, and those who sleep worst tend to underestimate most.
How to read it
Not as an instruction to set an alarm differently. The study did not test whether changing sleep changes brains, and no observational study can.
What it supports is narrower and still useful. If your sleep sits well outside the middle of the range and you did not choose that, it is worth treating as a symptom worth investigating rather than a habit to manage. And for anyone already sleeping seven or eight hours, this is not a finding that asks anything of you.
People also ask
Does bad sleep cause Alzheimer's, or does Alzheimer's cause bad sleep?
This study cannot tell you, and that is the central problem in the field. Brain changes of Alzheimer's disease begin a decade or more before symptoms, and they disturb sleep. So a person sleeping badly at 60 may be sleeping badly because of early disease rather than heading toward it. The U-shape is consistent with both stories, and probably with both being partly true.
What is a brain atrophy signature?
It is a single score derived by machine learning from a whole brain scan, trained to recognize the particular pattern of shrinkage seen in early Alzheimer's disease rather than any shrinkage at all. The one used here is called SPARE-AD. It is a research measure of resemblance to a pattern, not a diagnosis, and a person can score high and never develop the disease.
So what is the right amount of sleep?
This study points at the middle of the range rather than a number, and it did not test whether changing your sleep changes anything. Short sleep here meant five to six hours and long sleep nine to ten. If your sleep sits well outside that middle and you did not choose it, that is worth raising with a doctor. It is not a reason for anyone sleeping seven or eight hours to act.
Is long sleep actually harmful?
Long sleep is the half of the U that is most often misread. Sleeping ten hours is rarely a habit someone adopts; it usually reflects something else, including depression, illness, medication or disrupted sleep that needs more hours to compensate. That makes the long-sleep arm particularly vulnerable to the disease driving the sleep rather than the reverse.
How was sleep measured?
By self-report, which is the weak point. People are poor at estimating their own sleep, and the error is not random: those who sleep badly tend to underestimate. A wrist monitor would be firmer. The health-record half of the study sidesteps this by using recorded diagnoses of insomnia and hypersomnia instead, and found higher Alzheimer's risk with both.