1. Why this chapter exists separately from the sleep chapter
The sleep chapter covers duration, regularity, apnoea and insomnia, and it is the one to read if you want to sleep better. This chapter answers a narrower question that gets asked constantly and answered badly: does sleep clear something out of the brain, and does missing it cause dementia?
The short version is that the mechanism is real, most of what is quoted about it was measured in mice, and the human evidence supports something meaningful but considerably smaller than the headlines.
2. What is mouse and what is human
The 60% figure is mouse data. The 2013 Science paper that founded this field found that during natural sleep the brain's interstitial space expanded by roughly 60%, allowing cerebrospinal fluid to flush through and carry away metabolic waste including beta-amyloid. That was measured in mice, by two-photon imaging through a cranial window — a technique that cannot be performed on a person. It is quoted as a fact about human brains more or less universally, including by people who know better.
The human study is smaller and says less. A 2018 PET study found that a single night of sleep deprivation produced a measurable increase in beta-amyloid, notably in the hippocampus and thalamus. That is a real human finding and it is about twenty people, one night, and a marker rather than an outcome.
Both things can be worth knowing. What they do not jointly establish is that a poor week's sleep causes Alzheimer's disease, which is the inference most coverage invites.
3. The direction-of-causation problem, which is the hard part
Poor sleep is associated with later dementia in cohort after cohort. The obvious reading is that sleep loss damages the brain.
The other reading is at least as plausible. Sleep architecture changes years before dementia is diagnosed, because the neurodegenerative process affects the brain regions that generate sleep. So disturbed sleep in someone's sixties may be an early symptom of a disease already underway rather than a cause of one to come. Both are probably happening. Nobody can currently say in what proportion, and any source that tells you confidently is going beyond the data.
This is the same shape as the argument in the sleep chapter: illness causes bad sleep at least as readily as bad sleep causes illness, and the epidemiology cannot separate them.
4. What is worth doing anyway
The practical advice survives the uncertainty, which is the useful part — you do not have to resolve the mechanism to act sensibly.
- Regularity appears to matter more than duration. In over 10 million hours of accelerometry from 60,977 UK Biobank participants, the more regular sleepers had substantially lower all-cause mortality, and regularity outperformed duration as a predictor. Same time to bed, same time up, including weekends. Covered in full in the sleep chapter.
- Untreated sleep apnoea is the one with a diagnosis and a treatment. If you snore, wake unrefreshed, or have resistant high blood pressure, that is a referral rather than a lifestyle adjustment — and it is under-recognised in women.
- Alcohol fragments the second half of the night even when it shortens the time taken to fall asleep, and people consistently misjudge that trade.
- CBT for insomnia before hypnotics. First-line in the UK, better in the long run, and it makes things worse for a fortnight before it works, which is why people abandon it.
And one thing not to do. If a sleep tracker is making you anxious about your sleep, stop wearing it. Anxiety about sleep is itself a well-described cause of insomnia, and a device that scores your night out of 100 is a good way to manufacture it. That advice is simultaneously a joke and clinical.