Plants and your brain

In 60,298 UK Biobank adults followed for nine years, the third eating closest to a Mediterranean pattern (vegetables, legumes, fruit, nuts, olive oil, fish; little processed or red meat) developed 23% less dementia than the third furthest from it, with 882 cases, and the protection held in people at high genetic risk [1]. A meta-analysis of five earlier cohorts put the top third at 33% less mild cognitive impairment or Alzheimer's [5], and a later UK Biobank analysis of 121,521 adults with repeat diet records found the best Mediterranean scorers at about half the dementia rate of the worst [11]. This is the pattern the dementia What Matters diet pill scores: "Whole foods, lots of vegetables" gets the −23%, and the two middle answers get starred, projected numbers between the thirds.

MIND is Mediterranean plus DASH with leafy greens and berries promoted and butter, cheese, fried food and pastries demoted. In 923 Chicago elders followed 4.5 years, the top third of MIND scores developed 53% less Alzheimer's than the bottom third and the middle third 35% less [2], and in the same cohort the top third declined as if 7.5 years younger [3]. Then came the trial: 604 overweight older adults with a family history of dementia randomised to MIND or a control diet for three years, both with calorie counselling. Global cognition improved in both groups by the same amount, and brain scans matched too [10]. The likeliest readings are that three years is too short, that the control group also ate better, or that the cohort numbers carried confounding. MIND is teaching only here: it is the same foods as the diet pill with two emphases, and it has no separate dementia number that survives the trial.

Among 493,888 UK Biobank adults with 2,896 dementia cases over eight years, every extra 25 g a day of processed meat (about one rasher of bacon) went with 44% more dementia and 52% more Alzheimer's, independent of APOE4. The surprise in the same paper: an extra 50 g a day of unprocessed red meat went with 19% less dementia [4], which is hard to square with a blanket "meat causes dementia" story and easy to square with a processing one. Two US cohorts of 133,771 health professionals followed for decades found a quarter serving a day of processed red meat went with 13% more dementia and 1.6 years of extra cognitive ageing per daily serving, and replacing it with nuts and legumes went with 19% less [9]. Teaching only for dementia: the life model already charges processed meat through mortality, and the dementia chart has no separate pill for it.

The paper behind "meat eaters are twice as likely to get dementia" is a 1993 Adventist Health Study sub-analysis of 272 Californians in matched quartets: meat eaters (fish and poultry included) were 2.2 times as likely to develop dementia, rising to 3 times when past meat eating was counted, but the first figure did not reach significance, and the paper's second sub-study of 2,984 unmatched people found no difference at all [12]. The best modern cohort is the Tzu Chi Vegetarian Study: 5,710 Buddhist volunteers in Taiwan, 121 dementia cases in nine years, a third less dementia in vegetarians after adjustment [8]. The authors list the problems themselves: diet asked once, dementia read from insurance claims, and vegetarians who may differ in ways no adjustment catches. In UK Biobank, an overall plant-based score (more plants, fewer animal foods, no quality split) was unrelated to dementia (hazard ratio 1.03) [6]. Teaching only: no vegetarian pill, because the honest reading is that meat-free is a proxy for the pattern in factor 1, not a lever of its own.

Give plant foods a quality score and the null becomes two opposite findings. In 180,532 UK Biobank adults with 1,428 dementia cases, the top fifth on a healthful plant-based index (whole grains, vegetables, fruit, nuts, legumes, tea and coffee) had 18% less dementia, and the top fifth on an unhealthful one (juices, refined grains, potatoes, sweets) had 29% more [6]. Three more cohorts (Health and Retirement Study, Framingham Offspring, Whitehall II; 891 cases) pooled to 29% less for the top third of the healthful index and 42% more for the unhealthful one, with vegetables, nuts, legumes and tea or coffee doing most of the work; a five-cohort meta-analysis of 207,981 people landed at −21% and +24% [13]. Teaching only, but it is the same axis the diet pill already runs on, from "Mostly processed" to "Whole foods, lots of vegetables".

In 72,083 UK Biobank adults over 55, followed about ten years with 518 dementia cases, every 10% of the diet by weight that came from ultra-processed food (packaged snacks, sweetened drinks, reconstituted meats, ready meals) went with 25% more dementia, and modelling a swap of 10% of that for unprocessed or minimally processed food went with 19% less [7]. It is a cohort with a short-ish follow-up and diet recalled at the start, so the number is an association, not a promise. Teaching only.

In 960 Chicago elders tested yearly for about five years, the fifth eating the most green leafy vegetables (about 1.3 servings a day) declined at a rate equivalent to being 11 years younger than the fifth eating the least, after adjusting for exercise, mental activity, seafood and alcohol [14]. In 921 people from the same project, the fifth with the highest intake of flavonols (kale, beans, tea, spinach, broccoli, apples, berries) developed Alzheimer's at about half the rate of the lowest fifth (hazard ratio 0.52) [15]. A 2025 machine-learning pass over 185,012 UK Biobank diets picked out the same foods, leafy greens, berries and citrus among eight groups, and outperformed the MIND score as a predictor [16]. One cohort, one city, for the first two; teaching only. The chart's vegetable question scores servings, not greens specifically.

Across 21 cohorts and 181,580 people, each weekly serving of fish went with 5% less dementia and 7% less Alzheimer's, and higher DHA intake with markedly less of both, though the dose curve is not linear [17]. In the Framingham Offspring cohort, the quarter of adults with the lowest red-cell DHA had brain volumes about two years older than everyone else [18], which is the omega-3 pill on the cognitive fitness chart. Fish is inside the Mediterranean pattern, so it is part of what the diet pill scores; it has no separate dementia pill. Vegans get no long-chain omega-3 from food unless they take algae oil, and the conversion from flaxseed is poor.

The reason a plant-forward plate might protect the brain is mostly vascular: less saturated fat, more fibre and polyphenols, better blood pressure and insulin sensitivity, the same channels as the blood pressure, cholesterol and diabetes pills on the chart. The direct evidence on saturated fat itself is thin: four cohorts with 633 cases pooled to 39% more Alzheimer's and about twice the dementia in the highest-intake group, with 15% more Alzheimer's per extra 4 g a day [19], and the Alzheimer's interval only just clears no effect. The 2024 Lancet Commission left diet off its 14 factors for exactly this reason: strong through cholesterol, blood pressure, diabetes and weight, not yet strong on its own. Teaching only.

Vitamin B12 comes from animal foods or supplements, and low B12 raises homocysteine, which tracks with brain shrinkage. A review of 18 studies found B12 depletion or deficiency in a large share of vegetarians in every group measured, from 11% to 90% of the elderly and 62% of pregnant women, with vegans worse than vegetarians [20]. The intervention side is real: in a two-year trial in older adults with mild cognitive impairment, high-dose B vitamins cut grey-matter loss in Alzheimer's-vulnerable regions up to seven-fold, but only in those who started with high homocysteine [21]. One more caution for the meat-free: in EPIC-Oxford, vegetarians had 22% less heart disease than meat eaters but 20% more stroke, mostly haemorrhagic, and stroke roughly doubles later dementia risk [22]. Teaching only; B12 has no pill on any chart.

Cohorts can rank diets; only trials can prove them, and the dementia trials are few. PREDIMED randomised older Spaniards to a Mediterranean diet with olive oil or nuts against a low-fat control: after four years the 334 with cognitive tests held steady while controls declined, on secondary outcomes with modest effect sizes [23]. DIRECT PLUS randomised 284 adults to guidelines, Mediterranean, or a green Mediterranean diet (walnuts, green tea, Mankai): over 18 months both Mediterranean arms lost less hippocampal volume, the green one least, and less red and processed meat tracked with less shrinkage [24]. The MIND trial found nothing (factor 2). No trial has run long enough to count dementia diagnoses. That gap is why the Lancet Commission keeps diet off its list, and why this chart scores one pill, the Mediterranean pattern from the largest cohort, rather than one for MIND, one for vegetarian, and one for every food: they would be the same answer counted three times.