News · Brain & Mental Health
Nearly 4 in 10 dementia cases tied to risks you can change
A Lancet Public Health study of 1.75 million Danes put 37.8% of dementia down to 16 modifiable risk factors, with infections and depression topping the list, not the usual suspects.
Based on a peer-reviewed study in The Lancet Public Health
- Researchers in Copenhagen followed 1,753,515 people from age 65, using national registries, and published in The Lancet Public Health.
- Across 16 modifiable risk factors, the overall population attributable fraction was 37.8% (95% CI, 35.6-40.0), meaning that share of dementia is statistically linked to factors that could in principle be changed.
- The largest single contributors were hospital-diagnosed infections (9.4%), depression (8.7%) and cardiovascular disease (7.6%), ahead of more familiar targets.
- After correcting for reverse causation, infections fell to 7.0% and depression to 5.7%, so part of the raw signal was early dementia causing the risk factor rather than the reverse.
- The apparent protective effect of obesity disappeared entirely under the same correction: it was an artifact, not a benefit.
- The burden was higher in men, in those born 1948 to 1957, and in people with low education.
- Attributable fractions describe populations, not individuals. Avoiding every listed factor does not cut any one person's risk by 38%.
Dementia prevention advice tends to arrive as a familiar list: hearing aids, blood pressure, exercise, education. A study in The Lancet Public Health has rebuilt that list from the ground up for one entire country, and the order came out differently.
Researchers in Copenhagen followed 1 753 515 individuals from the age of 65 and calculated how much dementia could be attributed to each of 16 potentially modifiable risk factors. The total came to 37.8%.
Why do it country by country
Global estimates already exist, and they are widely quoted. The authors’ argument is that they may not transfer: although global estimates for dementia risk reduction are available, risk profiles might differ in a high-income, welfare-state setting.
There was also a methodological gap. Few studies have been conducted that simultaneously estimate the prevalence of dementia risk factors and their associated risk of dementia within the same nationwide population. Most stitch together risk from one source and prevalence from another. This one had both in the same registry.
What came top
The ranking is the news. The largest attributable fractions were for hospital-diagnosed infections, depression, and cardiovascular disease, in that order, at roughly 9%, 9% and 8%.
Until now the field’s attention has been elsewhere: infections have never topped a dementia risk table. Neither is depression, which has long sat in the awkward position of being possibly a cause, possibly an early symptom, and possibly both.
The researchers took that ambiguity seriously enough to test it. They accounted for reverse causation by splitting the risk time into less than 5 years and more than 5 years since exposure and the figures fell: after accounting for reverse causation, the attributable fraction for hospital-diagnosed infections was 7.0%, and 5.7% for depression.
Lower, but not gone. Some of the signal was early dementia driving the risk factor. Not all of it.
The obesity mirage
The most instructive result concerns a finding the paper demolishes rather than reports. Obesity had appeared mildly protective against dementia, a recurring oddity in this literature.
It was not real. The apparently protective effect of obesity on dementia was also shown to be driven by reverse causation. People sliding toward dementia lose weight years before diagnosis, which leaves the still-obese group looking deceptively healthy.
Anyone tempted to read a comforting message into old headlines about weight and the brain should stop there.
Who carries the burden
The overall attributable fraction was higher in males, 1948-57 birth cohorts, and individuals with low education. The same national advice therefore has very different force depending on who is receiving it.
The interactions matter too. Risk factor combinations involving depression, hospital-diagnosed infections, or cardiovascular disease showed the strongest superadditive effects, meaning the pair did more harm than the sum of its parts.
The number to be careful with
An attributable fraction is a population statistic. It estimates what might be avoided if a risk factor were removed from everyone, assuming the association is causal. It does not promise any individual a 38% reduction, and no one should read it that way.
Registry data also captures what got diagnosed and coded, which favors conditions serious enough to reach a hospital. Milder infections and undiagnosed depression are invisible here.
What the study does supply is a country-specific map rather than a borrowed one. The authors’ own framing is that these findings reinforce the need for tailored dementia interventions building on country-specific evidence.
The general lesson travels even if the numbers do not: the biggest modifiable contributors to dementia in a wealthy country may be the illnesses people already see a doctor about, not the lifestyle items on the poster.
People also ask
Does this mean I can cut my dementia risk by 38%?
No, and this is the most common misreading of this kind of figure. A population attributable fraction estimates how much dementia in a whole country is statistically associated with a set of risk factors, assuming those associations are causal and the factors were entirely removed. It is a public-health planning number, not a personal guarantee. Individual risk depends on genetics, age and much else that no one can change.
Why do infections come out on top?
That was the surprise. Hospital-diagnosed infections carried the largest single attributable fraction at 9.4%, ahead of depression and cardiovascular disease. Part of that is reverse causation, since people developing dementia are more prone to serious infections before diagnosis, and correcting for it dropped the figure to 7.0%. What survives suggests infections severe enough to need hospital care are worth taking seriously as a brain-health matter, not just an acute illness.
What happened with obesity?
In the raw analysis obesity looked mildly protective against dementia, an oddity seen in other studies too. This paper shows that apparent effect was driven by reverse causation: people in the early, undiagnosed stages of dementia tend to lose weight, which makes the remaining obese group look healthier. Once the researchers accounted for it, the protection vanished. It is a useful demonstration that a counterintuitive finding is often a measurement artifact.
How is this different from the Lancet Commission's global dementia estimates?
The Commission's widely cited figure is a global synthesis drawing on studies from many countries. This one estimates prevalence and risk within a single nationwide population using linked registry data, which is why the authors argue risk profiles might differ in a high-income, welfare-state setting. It is less generalizable, and far more internally consistent.
What are 'superadditive' risk factor combinations?
It means two risks together did more damage than adding their separate effects would predict. Combinations involving depression, hospital-diagnosed infections or cardiovascular disease showed the strongest such effects. The practical implication is that treating one factor in someone who has several may pay off more than the individual numbers suggest.