The Lancet's 45% dementia figure is a population estimate, not a personal promise

The most quoted number in dementia prevention is that 45% of cases could, in principle, be prevented or delayed by addressing 14 modifiable risk factors [s2]. It is a real estimate from a serious body of work — the 2024 report of the Lancet standing Commission on dementia [s1]. It is also one of the most misread figures in health, because it describes a whole population and is routinely reported as though it were a promise to an individual. It is not. Nobody can lower their own dementia risk by 45%, and the Commission does not claim they can.

What the Commission actually did

The Lancet Commission, first convened in 2017, updates its account of what is known about dementia prevention roughly every few years. Its 2020 report attributed about 40% of dementia to 12 modifiable risk factors [s3]. The 2024 update added two more — high LDL cholesterol in midlife and untreated vision loss in later life — bringing the count to 14 and the headline estimate to 45% [s2] [s3].

The 12 factors carried forward are less education, hearing loss, high blood pressure, smoking, obesity, depression, physical inactivity, diabetes, excessive alcohol, traumatic brain injury, air pollution, and social isolation [s2]. The two new ones were added on the strength of large cohort studies and genetic analyses for cholesterol, and of two large meta-analyses for vision [s2]. High LDL cholesterol was assigned a population-attributable fraction of 7%, untreated vision loss 2% [s2].

What a "population-attributable fraction" is, and is not

The 45% figure is the sum of population-attributable fractions — a specific epidemiological quantity, calculated in the 2024 report using data from the Norwegian Trøndelag Health Study [s2]. A population-attributable fraction estimates the share of cases in a population that would not occur if a risk factor were completely eliminated from that population, assuming the association is causal and everything else stays fixed.

Three things follow from that definition, and each one weakens the everyday reading of the number.

First, it is a statement about a population, not a person. Saying 7% of dementia is attributable to high LDL cholesterol does not mean any given person with high cholesterol has a 7% chance of getting dementia, or that lowering their cholesterol removes 7% of their risk. It means that if nobody in the population had high midlife cholesterol, the total case count might fall by roughly that share.

Second, the fractions assume the associations are causal and fully reversible — that eliminating a factor removes exactly the dementia attributed to it. For most of the 14 factors, the evidence is observational. People with more education, better hearing, and richer social lives differ from those without in many ways that are hard to fully adjust for, and reverse causation is a persistent problem: the earliest, silent changes of dementia can themselves reduce activity, socialising and self-care years before diagnosis.

Third, the fractions cannot simply be added without double-counting. The risk factors overlap heavily — obesity, diabetes, hypertension, physical inactivity and cholesterol travel together — so the Commission adjusts for what they call "communality" between factors. The 45% is already a weighted, deflated total, not a straight sum, which is one reason adding a 7% factor and a 2% factor moved the headline by only five points rather than nine [s2] [s3].

Where the evidence is stronger and where it is thin

The individual risk factors are not equally well established. The clearest causal case in the whole list is arguably the one for treating cardiovascular risk in midlife — blood pressure, smoking, diabetes — because those factors have decades of trial evidence in their own right for other outcomes, and because vascular damage is a plausible mechanism for a meaningful share of dementia.

The weakest link between association and action is where a widely publicised factor has been tested directly and underperformed. Hearing loss is the clearest example: it carries a large attributable fraction in the Commission's accounting, but the largest randomised trial of hearing aids to slow cognitive decline, ACHIEVE, found no overall effect across its full study population, with benefit appearing only in one pre-specified subgroup [s4]. An attributable fraction quantifies an association; it does not guarantee that the obvious intervention will deliver the modelled benefit.

What the number is good for

None of this makes the 45% figure useless — it makes it a public-health number rather than a personal one. As a statement about where a society might invest to reduce its future dementia burden, it is defensible and useful: it points policy toward education, clean air, blood-pressure control, hearing and vision care, and injury prevention, several of which are worth doing regardless of their dementia effect. As a promise that a motivated individual can prevent nearly half their own risk, it collapses. Age and genetics, the two largest drivers of dementia, are nowhere in the 14 factors, because they are not modifiable [s1].

The honest version of the headline is narrower and less quotable: across an entire population, a large share of dementia is statistically linked to conditions that are, in principle, avoidable, and reducing those conditions would probably lower the total number of cases by some amount that is smaller and less certain than the modelled 45%. That is still one of the more hopeful statements in the field. It is just not the personal guarantee the number is usually made to carry.

This article is informational and is not medical advice.