The short version
Cognitive decline is not simply a matter of time passing. Around 45% of dementia cases worldwide are associated with factors that can in principle be changed, and the largest of them are the same things this site spends most of its time on: blood pressure, cholesterol, hearing, movement, smoking, alcohol, sleep.
And almost none of what is sold for brain health is on that list. The gap between the two is what this topic is about.
Where the effort is best spent
Ordered by how much difference it makes across a population, which is not the order these appear in when the subject comes up.
| What | Roughly how much | Where it is covered |
|---|---|---|
| Hearing loss, treated | 7% of cases — joint largest, and aids are free on the NHS | Dementia risk |
| Raised LDL in midlife | 7% — joint largest, and added to the list only in 2024 | Cholesterol |
| Education, and staying mentally active | 5% — largely structural, and largely decided early | Dementia risk |
| Social contact | 5% — real, hard to intervene on, easy to underrate | Dementia risk |
| Blood pressure from midlife | The vascular cluster as a whole dominates, and hypertension leads it for stroke | Blood pressure |
| Depression, treated | 3% | Dementia risk |
| Head injury, avoided | 3% | Dementia risk |
| Air pollution | 3% | Air pollution |
| Not smoking | 2% of dementia cases — low here only because UK smoking prevalence is now low. It remains the largest single lever for overall mortality | Vaping & smoking |
| Physical activity | Moderate, and the only thing here with a randomised structural brain finding | Exercise and the hippocampus |
| Sleep, and treating apnoea | Moderate, with the direction of causation genuinely unsettled | Sleep and the glymphatic evidence |
| Vision, and having the cataract done | About 2%, and among the easiest to act on | Dementia risk |
| Alcohol | Moderate, and the safe-level story has changed | Alcohol |
| Staying socially connected | Real, hard to intervene on, easy to underrate | Dementia risk |
| Diet pattern | Smaller than the above, and it is the same pattern as for the heart | Eating for the heart |
| Supplements | Smallest of anything on this table, and one of them has real trial evidence | The supplement claims |
The percentages are the Lancet Commission's 2024 population attributable fractions — the share of dementia cases associated with each factor across a whole population, assuming causality and complete elimination. An earlier version of this table put smoking first and said the gap was “not close”. That was wrong for this question. Smoking leads the all-cause mortality table this one was modelled on; for dementia specifically it is 2%, because so few people in the UK now smoke. The correction is recorded here rather than made silently.
Five caveats, because a table like this invites more precision than it can carry. A population attributable fraction is not your personal risk reduction, and it is population-level and therefore nobody's personal order. The magnitudes are not additive and not independent — blood pressure and LDL act partly through the same vessels. The bands are orders of magnitude rather than effect estimates. And any source giving you days-gained figures for these is inventing precision that does not exist.
Two errors, and most people make one of them
The first is believing it is settled at birth and nothing helps. Twenty years ago dementia was widely presented as bad luck. The direction of the evidence has changed substantially since, and the change is genuinely new information rather than a shift in fashion.
The second is believing a capsule will do it. The brain-health supplement market is large, the mechanisms it advertises are frequently real, and the outcomes it implies have mostly never been tested. In 2026 a very large trial of a genuinely promising drug moved the biomarkers and changed nothing, which is the clearest demonstration yet of how far a plausible mechanism can travel without arriving.
If you are here because something feels wrong
Forgetting names and losing words is extremely common at every adult age, and much more so when you are tired, stressed, or in your forties and fifties.
- If you are a woman in midlife, the pattern has a well-described cause and an unexpectedly reassuring evidence base. Start at Menopause and cognition.
- If it is new, progressive, or other people have noticed it, that is a GP conversation rather than a reading exercise. Thyroid disease, B12 and iron deficiency, sleep apnoea, depression and alcohol all present this way, all are common, and all are treatable.
- If a relative's memory has changed, the useful information is what has changed and over what period — not a score. The dementia chapter sets out what is known about risk and what the new drugs can and cannot do.
What this topic will not do. It will not give you a brain age, score your cognition, or sell you a protocol. It grades what is claimed, says which studies were done in mice, and points at the small number of things that are worth the effort. Several of those live in other topics on this site, and this page links to them rather than restating them, because two pages on one subject drift apart.