You will see the figure quoted constantly: around 45% of dementia cases could potentially be prevented or delayed by addressing 14 modifiable risk factors. It comes from the Lancet Commission on dementia prevention, intervention and care, and it is the most useful single number in the field.
It is also widely misread. This article is about what the number means, what it does not mean, and what to do first — rather than a walk through all fourteen factors, which we've covered in detail separately.
If you want the full factor-by-factor breakdown — all 14, grouped by life stage, with the Commission's own recommended interventions — that's here, and it's the more complete piece. Come back for this one if the question you actually have is "what does 45% mean for me?"
What a population attributable fraction is
The 45% is a population attributable fraction, and almost every misunderstanding of it comes from not knowing that.
A PAF answers a population-level question: if this set of risk factors were entirely removed from the whole population, what share of cases would not occur? It is a modelled estimate, built from observed associations and from how common each factor is in the population. It is not a personal probability, and it is not a promise.
Three consequences follow, and all three are routinely lost in the retelling.
One: it describes populations, not people
An attributable share for hearing loss does not mean treating your hearing cuts your personal risk by that amount. It means that if nobody in the population had untreated hearing loss, dementia cases across that population would be lower by roughly that share.
Your own risk depends on your own particular combination of genetics, age, health history and exposures, and no population figure can be read off onto an individual. This matters in both directions: it means the number is less reassuring than it sounds if you fix one thing, and less damning than it sounds if you can't.
Two: the factors overlap, so they don't simply add up
This is the error that produces bad content, and we made it ourselves in an earlier version of this article.
Obesity, diabetes, physical inactivity and high LDL cholesterol travel together in the same people. So do smoking and alcohol. So do low education and several downstream exposures. When factors cluster like that, their individual attributable shares overlap — the same case of dementia can be "attributable" to more than one of them.
The Commission accounts for that overlap when it produces the combined figure, which is why the total is 45% rather than the sum of the individual parts. Anyone presenting the fourteen as fourteen clean slices of a pie chart has misunderstood the arithmetic — and a table of per-factor percentages that adds up to more than 45% is a sign the overlap has been dropped.
Three: "prevented or delayed" is doing real work
Those three words are not padding. Dementia incidence rises steeply with age, so pushing onset back by several years is a substantial gain even where it is not prevention — both for the person and, at population scale, for the number of people living with the condition at any given time.
It also means the honest promise is smaller and more achievable than "prevent dementia." Later is a real outcome.
Why we do not print per-factor figures
The Commission's 2020 report identified 12 factors accounting for an estimated 40% of cases. The 2024 update added two — high LDL cholesterol in midlife and untreated vision loss in later life — and revised the total to 45%.
We are not going to print the individual shares. Published summaries of the 2024 report give per-factor figures inconsistently, and the numbers shifted between the 2020 and 2024 analyses. Our factor-by-factor piece removed its per-factor percentages in September for exactly that reason — and this page went on carrying two of them in prose for weeks afterwards, which meant two live Proco articles disagreed about the same report. They are now gone from both.
The 45% combined total is the robust number. Treat any per-factor figure you see elsewhere — including in an older version of this page — as an indication of relative importance at best, and be suspicious of anyone presenting them to the decimal point.
The part that makes this immediately useful: the vascular thread
Look at the list and a pattern appears. Hypertension, diabetes, high LDL cholesterol, smoking, obesity and physical inactivity are all cardiovascular risk factors. They are on a dementia list because what damages the arteries feeding your heart damages the arteries feeding your brain.
That has a practical consequence worth more than any individual percentage. It is the same cluster that drives stroke risk — and in stroke the modifiable share is even larger, with ten factors accounting for roughly 90% of population-attributable risk, hypertension alone for about a third.
So you are not choosing between working on dementia risk and working on stroke risk. One set of actions addresses both of the largest causes of neurological disability there are, and the overlap is the single strongest argument for bothering.
Alcohol is on both lists too, and the picture there has shifted — the "moderate drinking is protective" story has not held up well.
Where to start, if you want one thing
Two appointments cover more of this list than any other action available to you.
- Get your blood pressure measured, and your lipids checked if you are over 40 and haven't had them done. Usually the same visit. Hypertension is the most treatable large item on the list and produces no symptoms whatsoever, which is exactly why it goes unaddressed for years.
- Get your hearing and your vision checked if you are over 50, or sooner if you find yourself asking people to repeat themselves or holding things further away to read. Both of the sensory factors specify untreated loss, and both are usually correctable.
On hearing specifically, it's worth knowing that the observational association is strong while the evidence that treating it reduces dementia risk is more mixed — the large randomised trial had a null primary result and a positive finding in one higher-risk subgroup. Hearing aids are worth having for hearing's own sake regardless; the dementia argument is a bonus reason rather than the reason.
The fourteen, one by one
Each factor now has its own piece, written to the same standard: the actual effect estimates, the confidence intervals, the nulls, and an explicit statement of where we sell nothing. Thirteen of the fourteen now have a post of their own; the fourteenth is noted honestly below.
- Less education — what cognitive reserve actually buys you, and why the factor is about a lifetime of cognitive demand rather than a certificate.
- Hearing loss — the strong association and the null randomised trial.
- High LDL cholesterol — why midlife is the word that matters, and why late-life readings mislead.
- Depression — what actually works, and the cause-or-consequence problem.
- Traumatic brain injury — our concussion piece is written but held pending clinical review, so there is no link here yet. We would rather say that than point you at nothing.
- Physical inactivity — the strongest recommendation on the whole list, and why it still rests on weaker trial evidence than its grading suggests.
- Diabetes — insulin resistance and cognition, including why the glucose-lowering trials did not improve it.
- Smoking — what quitting gives back and how long it takes, plus the nicotine paradox and where vaping data simply does not exist.
- Hypertension — blood pressure and the brain, the most treatable large item here.
- Obesity — why the timing changes the answer, and why high late-life BMI only looks protective.
- Excessive alcohol — how badly the "moderate drinking is protective" story has aged.
- Social isolation — loneliness and cognitive decline, and why isolation and loneliness are not the same variable.
- Air pollution — what the PM2.5 evidence actually shows, and the one factor that is mostly not an individual choice.
- Untreated vision loss — the newest factor on the list, where the action is an eye test and glaucoma is the instructive null.
What this figure doesn't say
It doesn't say dementia is your fault. The largest risk factors for dementia are not on the list at all, because they are not modifiable: age above everything else, and genetics — APOE variants in particular. People do everything on the list and still develop dementia. A framework for reducing population risk is not a framework for assigning blame, and the Commission is explicit about that.
It doesn't say a worrying memory means dementia. A great deal of what sends people to a doctor about their memory turns out to be normal ageing, stress, poor sleep, depression, medication effects or a treatable deficiency. If that's why you're here, this is the more useful article, and a handful of cheap blood tests — thyroid included — cover the reversible causes.
And it doesn't say any supplement prevents dementia. There is no supplement anywhere on the list of fourteen. Ours are not on it and we are not going to imply otherwise. Omega-3 and cognitive ageing has a real research literature worth reading on its own terms — we've covered the dose-response evidence separately — and it is a different question from the one this Commission answers. Treating the two as the same thing would be the easiest way for a company like ours to mislead you.
A note on this article
An earlier version printed a ranked table of percentages for all fourteen factors. We removed it, for two reasons: the individual figures are reported inconsistently across published summaries, and our table summed to 47% while the text claimed it summed to 45% — which is precisely the overlap error described above. The correction is left visible rather than quietly swapped.
Updated 30 September 2026. Removing the table did not finish the job. This page went on carrying two per-factor figures in ordinary prose — a share for hearing loss and LDL cholesterol, and another for vision loss — after our factor-by-factor article had already deleted the same numbers as unsourceable and published a correction saying so. So for several weeks two live Proco articles took opposite positions on the same report, and this was the one in the wrong. Both now carry the same position: fourteen factors, 45% of cases, no per-factor figures.
Figures from the 2024 report of the Lancet standing Commission on dementia prevention, intervention and care. This article is not medical advice and cannot assess anyone's individual risk.
Sources: Livingston G, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. Lancet. 2024;404(10452):572-628 · The Lancet — risk factors for dementia (2024 infographic)