Ask most people to name the things that raise the risk of dementia and you will hear about genes, head injuries, maybe blood pressure. Almost nobody says hearing. Almost nobody says vision.
Both are on the list, and vision was added only recently. That tells you something. This list is not settled ancient knowledge. It is a working document that has changed twice in six years, and it will change again.
It is also the most misreported subject in consumer health writing. A number built to describe entire populations gets printed as though it describes you. Statistical association gets printed as cause. Research that found nothing in its main result gets printed as a breakthrough.
So this article does two things: tells you what the evidence says, and shows you how to read the next headline about it.
Where the nearly-half figure comes from
The Lancet standing Commission on dementia prevention, intervention and care published its 2024 report on July 31 of that year, led by Gill Livingston of University College London. The Commission is a standing group of researchers that periodically reviews the world’s evidence and issues a consolidated assessment. It is not a single study, and nobody should call it one.
That report identified 14 risk factors that are at least partly changeable. Its conclusion, in the Commission’s own words, is that those factors “account for nearly half of dementias which theoretically can be prevented by eliminating these risks.”
Read the two hedges in that sentence, because coverage almost always deletes them. The first is theoretically. The second is eliminating, meaning removing every one of these exposures from every person alive, completely, for life. Nobody proposes that is achievable. It is a ceiling, not a forecast.
The list has also grown. The Commission’s previous report, in 2020, named 12 factors and put the figure at 40 percent of cases. Four years later it named 14 and said nearly half. The newcomers were uncorrected vision impairment and high LDL cholesterol in midlife, and smoking, diabetes and depression were reclassified as midlife rather than late-life risks.
A figure that moves from 40 percent to nearly half in four years, on the strength of two added factors, is not a law of nature. It is a current best estimate from a specific group reading a specific body of evidence.
A risk factor is not a cause, and a population number is not your number
Most of what feeds these estimates is observational research. Investigators follow large groups for years, record who was exposed to what, note who develops dementia, and adjust statistically for the other things they happened to measure. That design can show two things travel together. On its own it cannot show one produced the other.
The Commission says so directly, writing that “some associations may be only partly causal.” For factors measured close to the time dementia appears, the report flags that the relationship could run the other way, or in both directions at once.
Depression is the cleanest illustration. The Commission reports a pooled relative risk around 2.25 for dementia over 10 to 14 years of follow-up, which sounds decisive. But early dementia can alter mood years before anyone notices a memory problem. Some of the depression in those datasets may be the first symptom rather than a cause. The Commission also reports that treated depression carried lower risk than untreated, which is encouraging and still not proof of cause.
Now the second half, which is where readers get hurt. A population-attributable fraction is a modeling estimate. It answers one question: if this exposure vanished from an entire population, and if the association is genuinely causal, and if everything else held still, what share of cases would not occur?
Three conditions, none of which describes your individual life.
The arithmetic has a feature worth understanding. An attributable fraction depends on how common an exposure is at least as much as on how dangerous it is. A factor that doubles risk but affects 2 percent of people shifts few cases. A factor that raises risk by a quarter but affects half the population shifts many. So the ranking of risk factors by population impact does not match the ranking by how much any one person’s odds would change.
The Commission’s own text contains a good example. It cites a United States study of 16,690 participants estimating that about 1.8 percent of dementia in that population was attributable to vision impairment. That is a small-sounding percentage attached to a very large number of real Americans. Neither figure tells you your personal probability if your cataracts go untreated.
The arithmetic also refuses to let you add the factors up. The Commission notes that risk clusters within individuals and says it corrected for that overlap when combining factors. In plain terms, the same person often carries five of these at once, so the percentages cannot be stacked.
A more useful way to group the fourteen
The Commission presents its factors in its own order for its own reasons. Here is a different cut, organized by what you would do about each one and whom you would see.
Things you perceive the world through. Hearing loss and uncorrected vision impairment. Both are found by routine testing and both have ordinary treatments.
Things moving through your blood vessels. High blood pressure in midlife, high LDL cholesterol in midlife, diabetes, obesity and physical inactivity. These show up on standard lab work and standard checkups, and they are managed with the same tools used for heart health. What is good for the arteries feeding the heart appears good for the arteries feeding the brain.
Things that damage tissue directly. Head injury, smoking, excessive alcohol and air pollution. Three are exposures you can sometimes reduce. The fourth, air pollution, is decided mostly by where you live and by policy rather than by personal choice.
Things that shape how the brain is used and supported. Less education, depression and infrequent social contact. These are the hardest to act on alone and the most tangled with income, geography and circumstance.
Four of these reward a closer look, and three are below. The fourth, social connection, has an article of its own in this library, so here is the single sentence: the Commission reports reviews finding relative risks roughly between 1.18 and 1.57 for infrequent social contact and about 1.26 for loneliness, and notes that reverse causation may explain part of it.
Hearing and vision, the two nobody guesses
Start with how common hearing loss is, because the numbers reframe the topic. According to the National Institute on Deafness and Other Communication Disorders, part of NIH, disabling hearing loss affects about 22 percent of US adults aged 65 to 74 and 55 percent of adults 75 and older. Roughly 28.8 million American adults could benefit from hearing aids. Among adults 70 and older with hearing loss who could benefit, fewer than one in three, about 30 percent, have ever used one.
That gap is the story. More than half of the oldest Americans have meaningful hearing loss, and most are not treating it.
You can picture what untreated hearing loss does without any neuroscience. You stop going to the restaurant because you cannot follow the table. You say yes to things you did not catch. You answer less, so people ask less. Two plausible mechanisms sit underneath that: straining to decode speech consumes effort that would otherwise go elsewhere, and withdrawal strips away the social contact separately on the risk list.
Then comes the part that gets reported badly. The ACHIEVE trial, published in The Lancet in 2023 and funded by NIH, randomly assigned 977 adults aged 70 to 84 with untreated hearing loss and no substantial cognitive impairment, at four US sites, to hearing aids with audiological counseling or to a health education program. After three years, the main result was nothing. Cognitive change was essentially identical in both groups, p value 0.96.
A prespecified sensitivity analysis, meaning one planned before the data were examined rather than hunted for afterward, found the effect differed between the two populations the trial recruited from. One was the ARIC study, a long-running US cardiovascular cohort whose members were older and carried more risk factors. The other was healthy community volunteers. In the ARIC group, cognitive decline was 48 percent slower with the hearing intervention. A 2025 secondary analysis in Alzheimer’s and Dementia went further: among participants in the top quarter of predicted risk, decline was about 62 percent slower, confidence interval 33.7 to 94.1 percent.
Here is how to hold that honestly. A trial that misses its primary endpoint and then finds an effect in a subgroup has generated a strong hypothesis, not confirmed a conclusion. The tier is preliminary. It needs its own trial to become established. None of which changes the fact that hearing aids have well-proven reasons to exist already. They let you hear. That argument does not need dementia attached to it.
Vision is the newer addition, and the Commission reports a pooled relative risk around 1.47 for dementia with vision impairment. The operative word in the factor’s name is uncorrected. Cataract surgery and a current eyeglass prescription are among the most routine interventions in medicine, which makes this one of the few items on the list where the fix is cheap and finished.
Cholesterol as a brain number
The Commission added high LDL cholesterol in midlife in 2024 and described the evidence as high quality, consistent and biologically plausible. Each 1 mmol per litre increase in LDL was associated with 8 percent higher incidence of all-cause dementia, a relative risk of 1.08 with a confidence interval of 1.03 to 1.14.
That modest per-unit association matters for two reasons. The first is timing: the finding is about midlife, not late life, so a lab report at 50 may carry information about a decade that feels impossibly far away. The second is that it is the same number your clinician is already discussing with you for cardiovascular reasons. Nobody needs a new regimen for this. The brain argument raises the stakes of a conversation most middle-aged adults are already having and often postponing.
Education works differently, because nobody can go back and acquire more schooling at 72. The concept underneath it is cognitive reserve: brains worked harder across a lifetime appear to tolerate more underlying damage before function visibly slips. Reserve does not make the disease absent. It appears to delay the point at which symptoms surface. That is a real benefit and a bounded one, and it is the honest reason to keep learning difficult things late in life: the activity is worth having, not functioning as insurance.
Why brain training products oversell
The commercial promise is transfer. Train on the game, get better at life. Researchers split that into near transfer, meaning improvement on tasks resembling the training, and far transfer, meaning improvement in genuinely different abilities or in day-to-day function.
A 2021 meta-analysis in Neuropsychology Review by Lan Nguyen, Karen Murphy and Glenda Andrews pooled 43 studies covering 2,636 participants across seven commercially available programs, in both healthy older adults and people with mild cognitive impairment. Near transfer was small but statistically significant in both groups. Far transfer to objective measures of everyday functioning was not significant. In the healthy group it appeared on subjective measures, meaning participants felt sharper, but not on objective ones. After adjusting for publication bias, the tendency of positive findings to reach print more readily than null ones, only the processing-speed effect survived.
The authors’ own summary is that these programs are reasonable as entertainment, and that current evidence does not support claims that they improve memory, general cognition or everyday functioning.
File that as insufficient evidence for the functional claims rather than as disproven. The practical reading: if you enjoy them, play them, and treat the advertising as advertising. A crossword, a language class and a hard piece of music carry the same evidence status and cost less.
Normal forgetting and the kind that deserves a look
MedlinePlus, published by the National Library of Medicine at NIH and updated September 29, 2026, draws this line in plain language, and the examples are what make it usable.
Ordinary forgetting, per that page: retrieval takes longer as you age, and forgetting a name, where you put your keys, or whether you locked the front door is normal.
A different category: forgetting how to use your phone, or how to find your way home.
Carry that distinction, because it beats any reassurance. Misplacing keys is a retrieval hiccup, and everyone has them under stress and fatigue. Losing the procedure for a device you have operated for years, or a route you have driven a thousand times, is a different sort of failure. It is a reason to make an appointment, not a diagnosis.
The same page lists causes of memory problems that are not dementia at all and that a clinician can look for: stroke, depression, head injury, blood clots or tumors in the brain, problems with the kidneys, liver or thyroid, and reactions to certain medicines. Some are treatable, several are easily missed, and the medication angle has its own article in this library.
This article will not give you a checklist to score yourself against, for a substantive reason rather than a cautious one. Telling normal aging from mild cognitive impairment from dementia takes history, examination and usually formal testing, and the people who most need evaluation are often the ones a self-quiz reassures. If you or the people around you have noticed a change, the next step is a clinician.
Four things worth bringing to that appointment
- A specific example with a date attached. On the 12th I could not work out how to use the remote tells a clinician far more than my memory is bad.
- Whoever has noticed it. Family members often see a pattern the person living it does not.
- Your complete medication and supplement list, prescription and over-the-counter alike.
- When your hearing and vision were last tested, and the results if you have them.
The framework is not settled, and the evidence promises nothing
In August 2026 The Lancet published a correspondence piece titled “Rethinking risk factors for global dementia prevention,” by Agustin Ibanez of the Global Brain Health Institute at Trinity College Dublin and Hernando Santamaria-Garcia of Pontificia Universidad Javeriana in Bogota. Livingston and four co-authors published a reply in the same issue, volume 408, pages 787 to 790.
Both pieces sit behind a paywall and neither carries a public abstract. This article verified that the exchange exists and who wrote it, could not read either one, and is therefore not going to characterize what they argued. What the exchange does tell you is that two years on, the framework is being contested in the journal that published it, with a response from the Commission’s own authors.
One related 2026 paper is readable. In the Bulletin of the World Health Organization, Thomas Daly, Andrea Slachevsky, Dominic Trépel, Sebastian Walsh and Agustin Ibáñez raise an objection about framing rather than arithmetic. They argue the Commission presents risk in a linear, visual way detached from structural conditions such as income and the built environment, and that prevention advice aimed at individual behavior tends to benefit better-resourced people most, widening existing gaps. They also note that declines in age-adjusted dementia incidence seen in high-income countries have not been reproduced in low- and middle-income ones.
That has a practical edge. Whether you can act on items from this list depends heavily on money, time, insurance and where you live. A list of modifiable risk factors is not the same as a list of things you personally can modify.
Finally, the hard part. Nothing described here prevents dementia, and the strongest trial evidence shows why the Commission reached for the word theoretically.
SPRINT MIND is the clearest case. It randomly assigned 9,361 adults with high blood pressure, across 102 sites in the United States and Puerto Rico, to a systolic target below 120 or below 140. Its primary cognitive outcome was probable dementia, and it missed: 7.2 versus 8.6 cases per 1,000 person-years, hazard ratio 0.83, confidence interval 0.67 to 1.04, which crosses the line of no effect. Its secondary outcomes did reach significance. Mild cognitive impairment fell, hazard ratio 0.81, as did the combined measure of impairment or dementia, hazard ratio 0.85. The authors note the trial stopped early because of benefit on its cardiovascular outcome, leaving it possibly underpowered for dementia.
That is close to the high-water mark for this field: a large, well-run, federally funded American trial that moved a secondary endpoint and did not move its primary one. It is a reason to treat high blood pressure, which was already a reason.
So use the list as a map of where attention is worth spending, not a scorecard you can pass. Get hearing and vision tested and corrected, because both are under-treated, fixable, and improve your life this year regardless of what they do in twenty. Know your blood pressure and your LDL in midlife, when the association looks strongest. And take a noticed change to a clinician rather than a search engine, because several of the conditions that look like dementia are not dementia, and some of them can be treated.
Sources
- Livingston G, Huntley J, Liu KY, et al. The Lancet, 2024;404(10452):572-628. Accepted version, UCL Discovery. DOI: 10.1016/S0140-6736(24)01296-0. “Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission.” discovery.ucl.ac.uk
- Livingston G, Huntley J, Liu KY, et al. The Lancet, 2024. Second readable copy used to cross-check the quoted wording. “Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission (hosted PDF).” chronicdisease.org
- Ibanez A, Santamaria-Garcia H. The Lancet, 2026;408(10557):787-789. PubMed. DOI: 10.1016/S0140-6736(26)01440-6. Bibliographic record verified; full text paywalled and not read. “Rethinking risk factors for global dementia prevention.” doi.org
- Livingston G, Sommerlad A, Mukadam N, Kivimaki M, Ferri CP. The Lancet, 2026;408(10557):789-790. PubMed. DOI: 10.1016/S0140-6736(26)01441-8. Bibliographic record verified; full text paywalled and not read. “Rethinking risk factors for global dementia prevention – Authors' reply.” doi.org
- Daly T, Slachevsky A, Trepel D, Walsh S, Ibanez A. Bulletin of the World Health Organization, 2026. PubMed. DOI: 10.2471/BLT.25.293220. “Accounting for inequality in dementia prevention programmes.” pmc.ncbi.nlm.nih.gov
- Lin FR, Pike JR, Albert MS, et al. The Lancet, 2023;402(10404):786-797. PubMed. DOI: 10.1016/S0140-6736(23)01406-X. “Hearing intervention versus health education control to reduce cognitive decline in older adults with hearing loss in the USA (ACHIEVE): a multicentre, randomised controlled trial.” doi.org
- Pike JR, Huang AR, Reed NS, et al. Alzheimer's & Dementia, 2025;21(5):e70156. PubMed. DOI: 10.1002/alz.70156. “Cognitive benefits of hearing intervention vary by risk of cognitive decline: A secondary analysis of the ACHIEVE trial.” doi.org
- Nguyen L, Murphy K, Andrews G. Neuropsychology Review, 2021;32(3):601-630. PubMed. DOI: 10.1007/s11065-021-09515-2. “A Game a Day Keeps Cognitive Decline Away? A Systematic Review and Meta-Analysis of Commercially-Available Brain Training Programs in Healthy and Cognitively Impaired Older Adults.” doi.org
- Williamson JD, Pajewski NM, Auchus AP, et al. JAMA, 2019;321(6):553-561. PubMed. DOI: 10.1001/jama.2018.21442. “Effect of Intensive vs Standard Blood Pressure Control on Probable Dementia: A Randomized Clinical Trial (SPRINT MIND).” doi.org
- National Library of Medicine, National Institutes of Health (MedlinePlus). “Memory (updated September 29, 2026).” medlineplus.gov
- National Institute on Deafness and Other Communication Disorders, National Institutes of Health. “Quick Statistics About Hearing, Balance, and Dizziness (updated September 20, 2024).” nidcd.nih.gov