Dementia Prevention

Nearly half of dementia cases worldwide are tied to 14 risk factors that can be changed, and most of them are the same things that protect the heart. Hearing loss is the biggest single one — mainly because it is so common, not because it is so dangerous.

The 2024 Lancet Commission on dementia prevention estimates that eliminating 14 modifiable risk factors would prevent or delay about 45% of dementia cases globally. Acting on them is the most impactful thing most adults can do for long-term cognitive health — with one honest caveat that runs through this whole page: no trial has yet shown that any of these interventions prevents dementia as an endpoint. What exists is a strong risk-factor map, several trials that move cognition, and one that moves dementia diagnoses in a natural experiment.

The Lancet Commission framework

Moderate — a careful synthesis of observational evidence, but the estimates assume causality and are not trial-derived.

The 2024 Lancet Commission update ranks modifiable risk factors by population attributable fraction — the share of dementia cases that would not occur if the risk factor were eliminated from the population.^[1] Two factors were new in 2024: high LDL ("bad") cholesterol in midlife, and untreated vision loss in late life.

Life stage	Risk factor	Risk multiplier	How common	Share of cases
Early life (<45)	Less education	1.6×	23%	5%
Midlife (45–65)	Hearing loss	1.4×	59%	7%
Midlife	High LDL cholesterol (new 2024)	1.3×	77%	7%
Midlife	Depression	2.2×	7%	3%
Midlife	Traumatic brain injury	1.7×	12%	3%
Midlife	Physical inactivity	1.2×	28%	2%
Midlife	Smoking	1.3×	22%	2%
Midlife	Diabetes	1.7×	9%	2%
Midlife	Hypertension	1.2×	31%	2%
Midlife	Obesity	1.3×	13%	1%
Midlife	Excessive alcohol	1.2×	13%	1%
Late life (65+)	Social isolation	1.6×	24%	5%
Late life	Air pollution (fine particulates)	1.1×	75%	3%
Late life	Untreated vision loss (new 2024)	1.5×	13%	2%
	All 14 combined			45%

Two things in that table are worth more attention than the headline number.

The ranking is driven by how common a factor is, not how dangerous it is. Hearing loss tops the list with a fairly ordinary risk multiplier of 1.4 — it wins because 59% of the relevant population has it. Depression carries by far the largest individual risk of the 14 (2.2×) but affects 7% of people, so it contributes a modest 3%. Hypertension's small share is often misread as "blood pressure doesn't matter much for the brain": in fact its individual effect on dementia risk is genuinely one of the weakest measured (1.2×) despite being present in 31% of adults. The case for treating it hard is cardiovascular, plus a measurable effect on milder cognitive impairment — not a large dementia effect.

Ten of the 14 are midlife factors. Late life contributes only three — social isolation, air pollution and uncorrected vision loss. The window that matters most is roughly ages 45 to 65.

These fractions assume the associations are causal, which for some factors is uncertain. The Commission itself flags that late-life depression may be an early symptom of brewing dementia rather than a cause, and that removing a risk factor across a population does not guarantee any individual avoids dementia (Caution — these are population averages, not personal predictions). When the full 14-factor model was re-tested inside a single long-running cohort of 5,526 Americans, only three factors survived — hearing loss and diabetes as risks, and midlife hypertension in the protective direction — leading those authors to warn that meta-analytic estimates may be inflated by publication bias. That analysis is a conference abstract rather than a peer-reviewed paper, so treat it as a caution flag rather than a refutation.^[2]

The biggest individual levers

1. Treat hearing loss

Moderate for the observational signal; the one randomised trial was null on its primary outcome.

The largest single factor in the Commission's model — tied with high LDL cholesterol at a rounded 7% share — and the one most people leave untreated for years.
The ACHIEVE trial — 977 US adults aged 70–84 with untreated hearing loss, randomised to hearing aids and audiological counselling or to health education — was null on its primary outcome: three-year cognitive decline was essentially identical in the two groups.^[3] The much-quoted 48% slowing came only from a pre-specified higher-risk subgroup (participants drawn from a long-running cardiovascular cohort, who were older and declining faster) — encouraging but hypothesis-generating (Weak/preliminary for the subgroup benefit). The underlying case for treating hearing loss still rests on its large population share and the low harm of the intervention.
A 2026 target trial emulation — an observational analysis designed to mimic a randomised trial, not an actual trial — followed 2,777 Australians aged around 75 with hearing impairment for seven years. Dementia occurred in 5.0% of those prescribed hearing aids versus 7.5% of those not (a relative risk of 0.67, i.e. about a third lower; in absolute terms roughly 25 fewer cases per 1,000 people over seven years), with a clean dose-response by how often the aids were actually worn. Cognitive test scores among survivors barely moved, so the benefit shows up as crossing the diagnostic threshold rather than as sharper thinking day to day.^[4]
Mechanisms: reduced cognitive load from straining to hear; preserved social engagement; reduced isolation.
Action: get audiometry if any decline is suspected. If you pass a standard audiogram but struggle in noise, ask for extended high-frequency audiometry, which tests the range above the standard 8 kHz ceiling; one analysis suggests it explains far more of the age-related difficulty understanding speech in noise than the standard test does, though that specific comparison comes from a single small preprint and the numbers should be treated as provisional. Hearing aids are not vanity, modern devices are small and effective, and the device that lives in a drawer doesn't count. See Hearing for the full picture.

2. Treat hypertension in midlife

Moderate for milder cognitive impairment; Weak/preliminary for dementia specifically.

SPRINT-MIND tested an intensive systolic blood-pressure target (below 120 mmHg) against the standard (below 140) in 9,361 adults. It cut mild cognitive impairment by about a fifth and the combined mild-impairment-or-dementia endpoint by about 15%, but the pre-specified primary endpoint of dementia alone missed significance — a 17% reduction whose confidence interval still included no effect at all (hazard ratio 0.83, 95% confidence interval 0.67–1.04). That bracketed range is where the true effect most plausibly lies; because it crosses 1.0, the result is compatible with no benefit. The trial was stopped early for cardiovascular benefit and was probably underpowered for dementia.^[5] Extended follow-up to a median of about seven years kept the composite endpoint significant while dementia alone remained non-significant.^[6]
The 2024 Commission calls for systolic pressure below 130 mmHg from age 40 — a target justified by the cognitive-impairment benefit and by cardiovascular gains, not by a proven dementia reduction.
Most effective: lifestyle (the blood-pressure-lowering DASH eating pattern, exercise, weight loss) plus medication if needed. See Blood pressure.

3. Manage cardiometabolic health

Moderate — consistent observational signal, no dementia-endpoint trials.

Type 2 diabetes raises dementia risk by roughly 70% in the Commission's pooled estimate — substantial, though short of the "doubles your risk" often quoted. Well-controlled diabetes is much less damaging than uncontrolled.
Visceral obesity specifically — central adiposity correlates with dementia more than total weight does.
High LDL cholesterol — the 2024 addition reflects accumulating evidence that the vascular contribution to dementia is larger than previously recognised. It is now tied with hearing loss for the largest share of cases, mostly because three-quarters of middle-aged adults have elevated levels.
Action: track HbA1c (the three-month blood-sugar average), a lipid panel, blood pressure, and waist circumference. Treat in midlife rather than waiting. See Lipid management.

4. Exercise — both modalities

Moderate for cognition; no trial has shown a dementia-incidence reduction.

A single trial in 120 older adults found aerobic exercise increased anterior hippocampal volume by about 2%, effectively reversing one to two years of age-related shrinkage; the specific finding rests on that one study and should be read as suggestive (Weak/preliminary).^[7]
Resistance training has an independent signal: a 12-month randomised trial in senior women improved one executive-function measure (selective attention and conflict resolution) by 11–13% while a balance-and-tone control group slightly worsened. One trial, one cognitive domain, one population.^[8]
The Finnish FINGER trial — a two-year multi-domain programme of diet, exercise, cognitive training and vascular-risk monitoring in 1,260 at-risk older adults — improved cognition, though the absolute between-group difference on the cognitive composite was small (about 0.022 units per year); the eye-catching "25%/83%/150%" figures circulating online are relative differences on composite and sub-scores.^[9]
US POINTER (2025), the large American successor to FINGER, randomised 2,111 at-risk adults aged 60–79 to a structured versus a self-guided version of the same multidomain programme (exercise, MIND-style diet, cognitive and social engagement, cardiometabolic monitoring). Both arms improved; the more intensive, accountable structured arm improved slightly more on global cognition. The added benefit was the same in carriers and non-carriers of APOE ε4, the main genetic risk variant for Alzheimer's, but was larger in participants who started with weaker cognition.^[10] Two caveats matter: the trial had no do-nothing control arm, so it shows that added structure helps rather than that multidomain living prevents decline; and it was funded and staffed by the Alzheimer's Association, with the structured arm using a commercial brain-training product.
Dose: 150+ minutes of aerobic exercise plus two resistance sessions per week (general exercise guidelines).
The other side of the coin: prolonged sitting is associated with hippocampal atrophy and white-matter damage independent of structured exercise. Highly sedentary older adults show worse trajectories on episodic memory and processing speed even when they meet the exercise guidelines — the active-couch-potato pattern. Walking about 7,000 steps a day is associated with 38% lower dementia incidence than 2,000 steps, in a dose-response meta-analysis that graded its own dementia evidence as moderate certainty.^[11] See Sitting.

5. Sleep — quality, regularity, and screening for apnea

Weak/preliminary for the mechanism; Moderate for the association.

Deep (slow-wave) sleep is when the brain's waste-clearance system appears most active at removing beta-amyloid — but the direct human evidence is one 17-person crossover study that moved amyloid and not tau, the other core Alzheimer's protein, so this is a plausible mechanism rather than an established one, and total sleep duration is the better-supported target than deep sleep specifically. See Sleep architecture.
Chronic short sleep predicts dementia incidence in cohort studies.
Sleep regularity independently predicts cognitive trajectory.
Untreated obstructive sleep apnea (OSA) — repeated breathing interruptions during sleep — is a substantial risk, and continuous positive airway pressure (CPAP) treatment improves cognitive trajectory in some studies. See Sleep-disordered breathing.
Excessive habitual daytime napping (over 60–90 minutes, especially in the morning or with high day-to-day variability) is now read as a behavioural marker of underlying neurodegeneration — see Daytime naps. Long naps don't cause dementia, but they are often an early visible sign of fragmented night-time sleep and amyloid accumulation.
Action: 7–8 hours, regular schedule, screen for sleep apnea if any indicator. If you've started napping noticeably more than you used to, treat it as a signal worth investigating.

6. MIND-pattern eating, and minimise ultra-processed food

Weak for the dementia-specific claim — the observational signal is large, the one randomised trial was null.

The widely cited "~53% lower Alzheimer's risk" comes from an observational cohort of 923 older adults followed 4.5 years: those in the top third for adherence to the MIND dietary pattern had roughly half the Alzheimer's incidence rate (hazard ratio 0.47, i.e. about 53% lower), a design prone to reverse causation.^[12]
The only randomised trial of the MIND diet — 604 older adults at risk, three years — was a clean null: global cognition and brain-imaging outcomes did not differ from the control diet, and both arms improved similarly under mild calorie restriction.^[13] The broader Mediterranean/MIND pattern is still worth following on cardiovascular grounds, but the trial does not support a dementia-specific claim.
A 2026 systematic review found ultra-processed food intake associated with accelerated cognitive decline and higher dementia risk even after adjusting for adherence to an otherwise healthy dietary pattern — suggesting a processing-specific effect on top of nutrient quality.^[14]
See Dietary patterns and Ultra-processed food.

Moderate — one of the three late-life factors, with a consistent cohort signal.

Frequent social contact, being married or partnered, and larger social networks all associate with reduced dementia; social isolation accounts for about 5% of cases in the Commission's model, tied with education for the largest non-sensory share.
The "loneliness epidemic" is a measurable health problem, not a figure of speech.
Depth of network matters more than the binary "lives alone" — composite measures of social integration (how many kinds of tie a person has, and how often they use them) track health outcomes far more strongly than living arrangements alone. See Purpose.
Action: maintain regular real-world relationships; volunteer; participate in groups; for older adults, structured social programmes.

8. Stay cognitively engaged

Weak for far transfer — training reliably improves the trained skill; generalisation is limited.

Educational attainment in early life builds cognitive reserve, and it is the single early-life factor in the Commission's model.
Continued learning in adulthood — language acquisition, music, complex hobbies — has a cohort signal.
The ACTIVE trial randomised 2,832 older adults to memory, reasoning or speed-of-processing training. Ten years later the reasoning and speed groups still outperformed controls on the ability they had trained (memory training did not last), and all three trained groups reported less difficulty with everyday activities — but that everyday-function benefit was self-reported rather than performance-measured, and the trial was not designed to show a dementia-incidence reduction.^[15]
"Brain training" apps have small, narrow effects with limited transfer to general cognition.
Action: real-world complex learning (language, music, dancing, demanding skills) beats digital "brain training". See Cognitive engagement.

9. Defend purpose, especially in late midlife

Moderate for the association; observational and open to reverse causation.

A 28-year prospective analysis of 4,632 adults in the Wisconsin Longitudinal Study measured sense of purpose at ages 52, 63 and 70 and cognition at 80. Purpose at 52 predicted better cognition but not lower dementia; purpose maintained through ages 63–70 predicted lower dementia at 80, with modest effect sizes (median odds ratio 0.85, roughly 15% lower odds). A steeper decline in purpose between 52 and 70 was found in those who had dementia at 80.^[16]
Earlier work in the Rush Memory and Aging Project found that a person scoring high on purpose was about 2.4 times more likely to remain free of Alzheimer's over up to seven years than someone scoring low.^[17]
Late midlife is the critical window — the years when retirement, empty nest, role loss at work and early widowhood erode purpose, with measurable cognitive cost a decade later.
Action: plan the next role before the current one ends; build at least one purpose source independent of paid work; maintain weekly contact with a small set of relationships of substance. See Purpose.

What about specific supplements and drugs?

Modest evidence

Omega-3 (the long-chain fat DHA) — an observational signal, but pooled randomised trials in cognitively healthy older adults show essentially no benefit for global cognition. Reasonable as part of an overall dietary pattern, not as a stand-alone cognitive supplement (Weak). See Omega-3.
Vitamin D — worth repleting if deficient; no clear additional cognitive benefit at higher doses (Weak).
B vitamins (B6, B12, folate) — the VITACOG trial showed slower brain shrinkage specifically in people with mild cognitive impairment and elevated homocysteine — an amino acid that builds up when B-vitamin status is poor — and a re-analysis found even that held only in participants with adequate omega-3 status. Routine use in people with normal homocysteine is not supported (Weak). See B vitamins.
Daily multivitamin — the COSMOS-Mind trial in 2,262 older adults found a daily multivitamin produced a small improvement on a telephone-administered cognitive composite over three years, while cocoa extract did nothing. The endpoint was a cognitive test score, not a dementia diagnosis, and the supplements were industry-donated (Weak/preliminary).^[18] Worth a neutral mention, not a recommendation.
Creatine — once corrected meta-analyses are applied, the cognitive effect in well-rested young adults is small. The benefit concentrates in older adults, vegetarians, and people under metabolic stress (acute sleep deprivation, hypoxia). About 10 g/day is the rough threshold for measurable effects on brain energy metabolism; the standard 3–5 g/day muscle dose is insufficient (Weak). See Creatine.

Weak or negative evidence

Ginkgo biloba — negative in both large prevention trials: 3,069 older Americans over roughly six years showed no reduction in dementia (if anything a non-significant increase),^[19] and 2,854 French adults over five years showed no reduction in Alzheimer's diagnoses.^[20]
Resveratrol — no convincing cognitive trials.
Curcumin — small studies suggestive; not robust. See Curcumin.
Most "memory" formulas — unproven mixtures.

Pharmacological

Anti-amyloid antibodies (lecanemab, donanemab) — modest disease-slowing in early symptomatic Alzheimer's, not prevention in healthy adults. The slowing is real but small in absolute terms: over 18 months lecanemab reduced decline on an 18-point clinical dementia rating scale by 0.45 points — about 27% relative slowing on a score that worsened by 1.66 points in the placebo group.^[21] Donanemab produced a similar 0.7-point difference over 76 weeks.^[22] Both carry a meaningful rate of amyloid-related brain swelling and microbleeds — 13% with lecanemab, 24% with donanemab — and require intensive infusion and monitoring (Moderate for the effect; Caution for the harm profile).
Shingles (herpes-zoster) vaccine — an unusually strong non-drug signal. A natural experiment in Wales, where a birthdate cut-off made the vaccine abruptly available to people born on or after 2 September 1933 but not to those born a week earlier, found new dementia diagnoses over seven years fell by 3.5 percentage points in absolute terms — a one-fifth relative reduction. The birthdate design mimics randomisation and sidesteps the "healthy-vaccinee" bias of ordinary observational studies (Moderate — quasi-experimental, not yet a randomised trial).^[23] An independent Australian natural experiment found the same direction of effect.^[24] The evidence rests on the older live-attenuated vaccine; data for the now-standard recombinant vaccine are still accruing, and a definitive randomised trial is being sought.
GLP-1 receptor agonists (semaglutide, tirzepatide) — an observational signal for reduced dementia incidence in type 2 diabetes, plus a large Swedish register study in which people already diagnosed with depression or anxiety had substantially fewer episodes of psychiatric deterioration while on semaglutide than during their own periods off it. The dedicated phase 3 EVOKE and EVOKE+ trials in early Alzheimer's missed their primary cognitive endpoint but moved markers of Alzheimer's pathology in the spinal fluid in the right direction. The honest read: the strongest dementia-reduction numbers come from observational cohorts (susceptible to confounding), while the head-on trials in established disease were null; the underlying biology can be modulated, but clinical disease isn't being rescued at three years, and the prophylactic-decades-earlier hypothesis is still open (Weak/preliminary; all major trials manufacturer-funded). See GLP-1 receptor agonists.
Statins — a cohort signal for reduced dementia in statin users; not confirmed by any trial with a dementia endpoint (Weak).

What's overrated

"Brain training" apps for transferable cognition.
Mega-doses of B vitamins in people who aren't deficient.
"Anti-inflammatory" supplement protocols without dietary pattern change.
Coconut oil for Alzheimer's — anecdotal claims, no robust trial support.
Benzodiazepine "dementia risk" — the alarming headlines overstate it. In a Dutch cohort of 3,526 older adults followed nearly seven years, benzodiazepine users were no more likely to develop dementia than non-users (Moderate — an overrated risk).^[25]
Anticholinergic drugs — a genuine but heavily confounded signal. In that same cohort, persistent use of strongly anticholinergic drugs roughly doubled dementia risk, but the association vanished entirely once people taking antidepressants and antipsychotics were excluded — so it tracks the underlying condition more than the drug. Sensible to minimise strong anticholinergics where that's easy, but the causal case is weaker than reported (Weak).

A practical brain-health checklist

Get audiometry every 5 years from age 50, sooner if there's any concern.
Treat hypertension in midlife — target below 130 mmHg systolic if tolerated.
Treat type 2 diabetes, prediabetes and metabolic syndrome. Keep HbA1c below 5.7%.
Reduce LDL cholesterol through diet, exercise, and statins if indicated.
Don't smoke. Drink minimally.
150+ minutes of aerobic exercise plus two resistance sessions per week, and defend a floor of around 7,000 daily steps.
7–8 hours of sleep, regular schedule, screen for sleep apnea.
Mediterranean / MIND-pattern eating. Leafy greens daily, berries twice a week or more, fish, nuts, olive oil, whole grains, beans, poultry. Limit red and processed meat, butter, fried foods, sweets, and ultra-processed food.
Stay socially connected. Real-world relationships matter.
Keep learning. Real skills, ideally social and physical — dancing, music, sports, languages.
Wear seatbelts and helmets; avoid repeated head impacts.
Reduce air-pollution exposure where feasible (filtration in homes near major roads). See Environmental toxins for the full air-filtration story.
Get vision corrected. Cataract surgery, glasses, regular eye exams. The signal scales with the severity of impairment rather than with any particular diagnosis — uncorrected functional vision loss is the modifiable variable. See Vision.
Treat depression. It carries the highest individual risk of the 14 factors. Therapy, medication, exercise.

Family history of Alzheimer's: what to do

Most familial risk runs through APOE ε4 carrier status. Genetic testing is available but not always recommended, because the actionable advice is the same whether you're a carrier or not.
The modifiable factors appear to work equally well in carriers: US POINTER found the cognitive benefit of its structured programme was statistically indistinguishable between APOE ε4 carriers and non-carriers, which is at least consistent with the usual assumption that carriers — starting from higher baseline risk — stand to gain more in absolute terms.
For specific genetic testing decisions, work with a genetic counsellor.