Does the MIND diet prevent Alzheimer's disease?

The only randomised trial says no over three years. 604 older adults with a family history of dementia were assigned to MIND or to a control diet, both with mild caloric restriction and weight loss support. Global cognition improved in both, with a difference of 0.035 standardized units and a confidence interval that crosses zero, p = 0.23, and no difference in white matter hyperintensities, hippocampal volume or brain volume. The cohorts that made the pattern famous associate it with 39 to 53 percent lower Alzheimer's risk, and those are observational.

So the cohort findings were wrong?

Not necessarily, and this is the honest answer rather than a hedge. Three years is short for a cognitive endpoint, the participants were cognitively unimpaired at entry so there was little decline to slow, and the control arm was itself a supported, calorie-restricted diet rather than a bad one. Those are reasons a real effect could have been missed. They are not evidence that one was there. What the trial does settle is that the association is not reproducible by assignment over three years in that population.

Is it better than just following a Mediterranean diet?

In the largest independent cohort, no. Among 121,521 UK Biobank participants followed a median 10.9 years, the top quartile of a Mediterranean score carried a dementia hazard ratio of 0.53 against 0.61 for MIND, and Mediterranean was also the stronger association for Alzheimer's disease and for vascular dementia. In the autopsy data the Mediterranean association with beta amyloid was stronger too. MIND was built to beat the pattern it came from and the independent data does not show that it does.

Which part of it matters most?

Green leafy vegetables are the only component with a standalone finding worth quoting: in 581 autopsied brains, people in the highest tertile of intake had less global Alzheimer's pathology than those in the lowest, p = 0.0038. That is one component association inside one observational analysis, which is a reason to eat them rather than a demonstration that they are the active ingredient.

Is there anyone it is a bad fit for?

Anyone on warfarin who suddenly increases green leafy vegetables without telling their prescriber, because vitamin K is what warfarin dosing is set against. Anyone with a nut or fish allergy, since two of the ten positive components are those. And anyone who should not drink, since wine is one of the fifteen scored components, although it is not required and no study here has isolated it.

Whole-food9 human studies cited

MIND

Written by Aaron CuhaReviewed Sep 2026

Also called: Mediterranean-DASH Intervention for Neurodegenerative Delay, MIND diet

The cohort data is strong and consistent across independent populations. The one randomised trial ran three years in 604 people and was null against an active control diet on every cognitive measure and every brain MRI measure. That null is the most important thing on this page.

What it actually is

A hybrid of the Mediterranean and DASH patterns, scored rather than prescribed. Fifteen components: ten to eat, being green leafy vegetables, other vegetables, berries, nuts, olive oil, whole grains, fish, beans, poultry and wine, and five to limit, being red meat, butter and stick margarine, cheese, pastries and sweets, and fried or fast food. The distinction that matters for reading the evidence is that it began life as a scoring instrument applied to food frequency questionnaires in an existing cohort, not as a diet anyone was asked to follow. It was turned into a prescription later, for one trial, and that trial is the null.

The mechanism its advocates propose

Not a mechanism so much as an assembly, and its designers described it that way. Individual foods and nutrients had been associated with slower cognitive decline in earlier work, notably green leafy vegetables, berries and long-chain omega-3 from fish, while saturated fat, trans fat and fried food had been associated with faster decline. The MIND score selects the components from the Mediterranean and DASH patterns that had that specific dementia literature behind them and drops the ones that did not, adding berries and green leafy vegetables as separate scored items. The honest description of that construction is that it is hypothesis generating, and it was first tested in the same cohort whose data was used to build it.

What happened when calories were controlled

Strongest design available: prescribed, self-reported

Participants were told what to eat and reported back. Under-reporting is the norm rather than the exception, so the actual intake in each arm is unknown.

No calorie-controlled feeding trial of this pattern exists and none is likely, because cognitive endpoints need years and feeding studies run for weeks. What does exist is a three-year randomised trial in which both arms received mild caloric restriction and weight loss support, so calories were matched by design across arms, and it found no difference in cognition or brain imaging.

Barnes 2023 is the trial and it is the centre of this page. Two sites, 604 older adults without cognitive impairment but with a family history of dementia, a body mass index above 25 and a suboptimal diet at baseline. 301 to MIND and 303 to a control diet, both with mild caloric restriction, both with dietary counselling and support to promote weight loss, three years. 93.4 percent completed. Global cognition improved in both groups, by 0.205 standardized units on MIND and 0.170 on control, mean difference 0.035 (95% CI minus 0.022 to 0.092), p = 0.23. Changes in white matter hyperintensities, hippocampal volume and total gray and white matter volume were similar in the two groups.
What that trial does not show is worth stating as carefully as what it does. The control arm was not a bad diet, both arms were calorie restricted, both received counselling and weight loss support, and the participants were cognitively unimpaired at entry with three years of follow-up. Those are all reasons a real effect could have been missed. It is a null against an active comparator in a motivated population over three years, not a demonstration that diet is irrelevant to cognition.
It also removes a confound that the cohort data cannot remove. Because both arms lost weight under the same caloric prescription, the trial rules out the possibility that an apparent MIND benefit is a weight loss benefit. Nothing separated, which means there is nothing to attribute.
The nearest randomised comparator is PREDIMED-NAVARRA, which tested a Mediterranean diet rather than MIND. 522 participants at high vascular risk, cognition assessed after 6.5 years of intervention against a low fat control. Mediterranean with extra virgin olive oil scored 0.62 points higher on the Mini-Mental State Examination (95% CI 0.18 to 1.05, p = 0.005) and 0.51 higher on the Clock Drawing Test (p = 0.001); with nuts, 0.57 (p = 0.015) and 0.33 (p = 0.048). Three caveats travel with it: those are fractions of a point on a 30-point instrument, cognition was assessed only at the end rather than tracked from baseline, and PREDIMED had a documented randomisation failure that is reported in full on the Mediterranean entry.

What happened when people just ate it

The observational layer is large, consistent and the reason anyone has heard of this pattern. Morris 2015 related the score to incident Alzheimer's disease in 923 participants aged 58 to 98 followed a mean 4.5 years: the top tertile of MIND scores carried a hazard ratio of 0.47 (95% CI 0.26 to 0.76) and the second tertile 0.65 (0.44 to 0.98), while for DASH and Mediterranean only the top tertile reached significance, at 0.61 and 0.46. That detail, that moderate adherence appeared to work for MIND and not for the other two, is what made the score popular. The companion paper related it to cognitive decline in 960 participants over 4.7 years and found the score associated with slower decline in global cognition and in each of five domains, with the top against bottom tertile difference described as equivalent to being 7.5 years younger in age. That 7.5 years figure is the number that travelled and it is a translation of a slope difference in an observational cohort into an age equivalent, not a measured outcome. Both papers come from the Rush Memory and Aging Project, the cohort the score was built in. Independent populations broadly agree. A French cohort of 1,412 people aged 65 and over, followed a median 9.7 years with 356 incident dementias, found a hazard ratio of 0.89 (0.83 to 0.95) per point of a French-adapted MIND score for dementia and 0.88 for Alzheimer's disease, with better white matter microstructure in an MRI substudy of 175 and no association with gray matter volume. UK Biobank is the largest and it contains the finding advocates skip: among 121,521 participants followed a median 10.9 years with 621 dementias, the top quartile of MIND scores carried a hazard ratio of 0.61 (0.48 to 0.78), and the plain Mediterranean score in the same analysis carried 0.53 (0.45 to 0.63). The Mediterranean score was stronger than the MIND score for all-cause dementia, Alzheimer's disease and vascular dementia alike. A 106-person analysis in stroke survivors found slower global decline in the top tertile over a mean 5.9 years, with 106 people in it.

Protein, and whether it confounds the result

MIND sets no protein target. Its protein comes from fish, poultry and beans, with red meat scored as a component to limit, and no trial or cohort of this pattern has measured protein intake as an exposure or lean mass as an outcome. The randomised trial applied mild caloric restriction and weight loss support to both arms and reported no body composition data at all, which matters because weight loss in older adults is where lean mass is most worth watching. Anyone combining this pattern with a weight loss drug or a training programme gets nothing from it on protein and will need to solve that elsewhere.

The measured intakes behind the protein figures are on the protein page.

What reliably moves

Marker	Direction	From
Global cognition over three years, randomised	No difference against an active control diet	Both groups improved, 0.205 against 0.170 standardized units, mean difference 0.035 with a confidence interval from minus 0.022 to 0.092, p = 0.23. 604 participants, 93.4 percent completion. This is the headline result of the pattern's only trial.
Brain MRI over three years, randomised	No difference	White matter hyperintensities, hippocampal volume and total gray and white matter volume were similar between the groups in the imaging subsample.
Incident Alzheimer's disease, observationally	Down	Hazard ratio 0.47 for the top tertile in the cohort the score was built in, and 0.61 for the top quartile in UK Biobank. Association, in people who chose the diet.
Incident dementia in independent cohorts	Down, and not by more than a plain Mediterranean score	0.89 per point in a French cohort. In UK Biobank, MIND 0.61 against Mediterranean 0.53 for all-cause dementia, with Mediterranean also stronger for Alzheimer's disease and vascular dementia. The hybrid did not beat the pattern it was built from.
Postmortem beta amyloid and global Alzheimer's pathology	Lower	In 581 autopsied participants, MIND was associated with lower global pathology (p = 0.034) and less beta amyloid (p = 0.050); Mediterranean scored p = 0.039 and p = 0.004 on the same measures. Green leafy vegetables carried the strongest single-component association, top against bottom tertile p = 0.0038.
Cognition independent of pathology	Better, which complicates the amyloid story	In 569 decedents from the same cohort, the association between MIND score and cognition near death barely changed when Alzheimer's and other pathologies were added to the model, from 0.119 to 0.111. The authors read that as cognitive resilience rather than less disease, which sits awkwardly alongside the amyloid finding from the same cohort.
Body weight	Down in both arms of the trial, by design	Both arms received mild caloric restriction and weight loss support. That is why the trial cannot be read as a diet-versus-weight comparison, and it is also why the null cannot be explained away as a weight difference.

Long term, and hard outcomes

Three years randomised and null. Up to sixteen years observational and positive. The gap between those two is the entire story of this pattern and both halves have to be reported. The observational finding is unusually robust for nutrition: it replicates in a French population cohort and in UK Biobank, it survives adjustment for physical activity, smoking and vascular burden in the pathology work, and it holds in analyses that exclude people who already had cognitive impairment at dietary assessment. The randomised finding is also unusually clean for nutrition: 604 participants, three years, 93.4 percent completion, prespecified global cognition endpoint, MRI secondary endpoint, and nothing separated. Neither result cancels the other. The most defensible reading is that people who eat this way have less dementia and slower decline, that this has been demonstrated in several independent populations, and that assigning the pattern to motivated older adults with a family history of dementia for three years does not reproduce it. Whether that is because three years is too short, because the participants were too well to decline measurably, because the control diet was itself good, or because the association was never causal, is not resolved by any evidence that exists.

Citations

Human2023
Trial of the MIND Diet for Prevention of Cognitive Decline in Older Persons.
New England Journal of Medicine
604 older adults without cognitive impairment but with a family history of dementia, three years, both arms on mild caloric restriction with weight loss support, 93.4 percent completion. Global cognition improved 0.205 against 0.170 standardized units, mean difference 0.035 (95% CI minus 0.022 to 0.092), p = 0.23. White matter hyperintensities, hippocampal volume and total gray and white matter volume all similar. Null.
Human2015
MIND diet associated with reduced incidence of Alzheimer's disease.
Alzheimer's and Dementia
923 participants aged 58 to 98, mean 4.5 years, food frequency questionnaire. Alzheimer's disease hazard ratio 0.47 (0.26 to 0.76) for the top MIND tertile and 0.65 (0.44 to 0.98) for the second, against 0.61 for the top DASH tertile and 0.46 for the top Mediterranean tertile. Observational, in the cohort the score was built in.
Human2015
MIND diet slows cognitive decline with aging.
Alzheimer's and Dementia
960 participants of the Memory and Aging Project over a mean 4.7 years. MIND score associated with slower decline in global cognition (beta 0.0092, p < 0.0001) and in each of five domains, with the top against bottom tertile difference described as equivalent to being 7.5 years younger. The authors write that replication in a dietary intervention trial would be required.
Human2023
Association of Mediterranean-DASH Intervention for Neurodegenerative Delay and Mediterranean Diets With Alzheimer Disease Pathology.
Neurology
581 autopsied older adults, mean age at death 91. MIND associated with lower global Alzheimer's pathology (p = 0.034) and less beta amyloid (p = 0.050); Mediterranean with p = 0.039 and p = 0.004 on the same outcomes. Green leafy vegetables, top against bottom tertile, p = 0.0038. Findings persisted after adjustment for physical activity, smoking and vascular burden.
Human2021
MIND Diet, Common Brain Pathologies, and Cognition in Community-Dwelling Older Adults.
Journal of Alzheimer's Disease
569 decedents with dietary data, cognitive testing near death and complete autopsy. Higher MIND score associated with better cognition near death (beta 0.119, p = 0.003), essentially unchanged when Alzheimer's and other brain pathologies were added to the model (0.111, p = 0.003). Read by the authors as cognitive resilience rather than less pathology.
Human2022
Association of a MIND Diet with Brain Structure and Dementia in a French Population.
Journal of Prevention of Alzheimer's Disease
1,412 participants aged 65 and over from the Three-City Bordeaux cohort, median 9.7 years, 356 incident dementias. Hazard ratio 0.89 (0.83 to 0.95) per point for dementia and 0.88 (0.81 to 0.96) for Alzheimer's disease. In an MRI substudy of 175, lower diffusivity in the splenium of the corpus callosum and no association with gray matter volume.
Human2025
Adherence to four dietary indices and the risk of all-cause and cause-specific dementia: Findings from the UK Biobank study.
Diabetes, Obesity and Metabolism
121,521 participants, median 10.9 years, 621 dementias. Top against bottom quartile all-cause dementia: Mediterranean 0.53 (0.45 to 0.63), MIND 0.61 (0.48 to 0.78), Recommended Food Score 0.66. Mediterranean was the stronger association for all-cause dementia, Alzheimer's disease and vascular dementia. The Healthy Diet Indicator showed no association with any outcome.
Human2019
Mediterranean-Dash Intervention for Neurodegenerative Delay (MIND) Diet Slows Cognitive Decline After Stroke.
Journal of Prevention of Alzheimer's Disease
106 community cohort participants with a clinical history of stroke, mean 5.9 years of follow-up. Top against lowest tertile of MIND score showed slower global cognitive decline (beta 0.08, 95% CI 0.0074 to 0.156) after adjustment. 106 people, observational.
Human2013
Mediterranean diet improves cognition: the PREDIMED-NAVARRA randomised trial.
Journal of Neurology, Neurosurgery and Psychiatry
The nearest randomised comparator, testing Mediterranean rather than MIND. 522 participants at high vascular risk, cognition assessed after 6.5 years against a low fat control. Mini-Mental State Examination higher by 0.62 points with olive oil (p = 0.005) and 0.57 with nuts (p = 0.015); Clock Drawing higher by 0.51 and 0.33. Fractions of a point, assessed only at the end, and within a trial whose randomisation failure is documented on the Mediterranean entry.

What people report

These are uncontrolled self-reports, not evidence. They are here because they tell you what to expect and what to watch for, which the trial literature does not. They cannot tell you whether anything works.

Described as the easiest pattern here to adopt, because it adds foods rather than banning categories and has no elimination phase.
No perceptible change, reported very commonly, which is the expected experience of a pattern whose endpoint is measured in cognitive test scores over years rather than in anything a person can feel.
Confusion about scoring, since the pattern is defined by servings per week of fifteen items and most people adopt an approximation rather than the instrument.
Berries and green leafy vegetables described as the hardest components to hit year round, and the most expensive.
People adopting it for a parent with a family history of dementia, which is exactly the population the randomised trial recruited and exactly the population in which it found no difference.
Disappointment and pushback after the 2023 trial was published, with a common and fair objection that three years is short for a cognitive endpoint, and a less fair one that the trial was flawed because it did not find what the cohorts found.
Weight loss reported by people who adopted it as a general healthy-eating pattern, which is what both arms of the trial were doing.

Sources: General brain-health and caregiver communities, dietitian and neurology practice write-ups, and the adherence data inside the trial, which is unusually good at 93.4 percent completion across three years. Uncontrolled self-report otherwise, and in this pattern self-report cannot speak to the endpoint at all, because nobody can feel a 0.035 standardized unit difference in a global cognition score.

Who this is wrong for

Anyone treating it as proven dementia prevention. Its only randomised trial found a difference of 0.035 standardized units in global cognition over three years, with a confidence interval spanning zero, and no difference on any brain imaging measure.
Anyone on warfarin who starts loading green leafy vegetables without telling their prescriber. Those are the single component with the strongest association in the pathology data and the main dietary source of vitamin K, which is exactly what warfarin dosing is titrated against.
Anyone with a tree nut or fish allergy who cannot substitute. Two of the ten positive components are nuts and fish, and the score has no substitution rule.
Anyone who should not drink alcohol. Wine is one of the fifteen scored components, it is not a requirement of the pattern, and no trial or cohort here has isolated its contribution from the other fourteen.
Anyone who expects it to outperform a plain Mediterranean pattern. In the largest independent cohort the Mediterranean score was associated with lower dementia risk than the MIND score for all-cause dementia, Alzheimer's disease and vascular dementia, and in the autopsy data the Mediterranean association with beta amyloid was the stronger of the two.
Anyone reading a score as a prescription. This was designed as a scoring instrument for food frequency questionnaires, and the one attempt to convert it into an assigned diet is the study that came back null.

Questions

Does the MIND diet prevent Alzheimer's disease?: The only randomised trial says no over three years. 604 older adults with a family history of dementia were assigned to MIND or to a control diet, both with mild caloric restriction and weight loss support. Global cognition improved in both, with a difference of 0.035 standardized units and a confidence interval that crosses zero, p = 0.23, and no difference in white matter hyperintensities, hippocampal volume or brain volume. The cohorts that made the pattern famous associate it with 39 to 53 percent lower Alzheimer's risk, and those are observational.
So the cohort findings were wrong?: Not necessarily, and this is the honest answer rather than a hedge. Three years is short for a cognitive endpoint, the participants were cognitively unimpaired at entry so there was little decline to slow, and the control arm was itself a supported, calorie-restricted diet rather than a bad one. Those are reasons a real effect could have been missed. They are not evidence that one was there. What the trial does settle is that the association is not reproducible by assignment over three years in that population.
Is it better than just following a Mediterranean diet?: In the largest independent cohort, no. Among 121,521 UK Biobank participants followed a median 10.9 years, the top quartile of a Mediterranean score carried a dementia hazard ratio of 0.53 against 0.61 for MIND, and Mediterranean was also the stronger association for Alzheimer's disease and for vascular dementia. In the autopsy data the Mediterranean association with beta amyloid was stronger too. MIND was built to beat the pattern it came from and the independent data does not show that it does.
Which part of it matters most?: Green leafy vegetables are the only component with a standalone finding worth quoting: in 581 autopsied brains, people in the highest tertile of intake had less global Alzheimer's pathology than those in the lowest, p = 0.0038. That is one component association inside one observational analysis, which is a reason to eat them rather than a demonstration that they are the active ingredient.
Is there anyone it is a bad fit for?: Anyone on warfarin who suddenly increases green leafy vegetables without telling their prescriber, because vitamin K is what warfarin dosing is set against. Anyone with a nut or fish allergy, since two of the ten positive components are those. And anyone who should not drink, since wine is one of the fifteen scored components, although it is not required and no study here has isolated it.