1. The rule this page applies
Almost everything sold for brain health is sold on a mechanism: this compound crosses the blood–brain barrier, that one is a precursor to a neurotransmitter, the other one is an antioxidant and oxidative stress drives neurodegeneration. Every one of those statements can be true while the capsule does nothing you would notice.
This site has watched that gap open repeatedly and in public. Inflammation drives coronary disease, and suppressing IL-6 exactly as designed changed nothing. Carotid wall thickness tracks atherosclerosis, and ENHANCE nearly killed a drug that turned out to work. And in 2026 the brain got its own worked example, which is the most important thing on this page.
EVOKE and EVOKE+, The Lancet, March 2026. 3,808 people aged 55–85 with amyloid-confirmed early Alzheimer's disease, randomised to oral semaglutide or placebo across 566 sites in 40 countries. The observational literature had been encouraging for years: people taking GLP-1 drugs for diabetes appeared to develop dementia less often.
Biomarkers of neuroinflammation improved by up to 10%, statistically significantly. There was no difference in clinical progression at two years — not on the primary endpoint, not on daily function, not on any cognitive secondary, and not in any prespecified subgroup. A pooled analysis found no delay in progression from mild cognitive impairment to Alzheimer's disease.
The point is not that GLP-1 drugs are useless. It is that a large, well-conducted, biologically plausible trial moved the marker and did not move the person, in exactly the field where every supplement on this page makes its case from markers.
2. Multivitamins — the one with real randomised evidence, and it is small
This is the claim that deserves the most care, because it is the only one here backed by large randomised trials and it is routinely reported at several times its actual size.
COSMOS randomised 21,442 US adults aged 60 and over to a daily multivitamin (Centrum Silver), cocoa extract, both or neither. Three cognitive sub-studies ran inside it:
| Sub-study | n | Finding |
|---|---|---|
| COSMOS-Mind | 2,262 | Global cognition improved; difference 0.07 z-score points at 3 years. Cocoa did nothing |
| COSMOS-Web | 3,562 | Immediate recall improved. Memory retention, executive function and object recognition did not |
| COSMOS-Clinic | 573 | In-person neuropsychological testing |
| Pooled | ~5,200 | Global cognition 0.07 SD, episodic memory 0.06 SD |
Where “two to three years of cognitive aging” comes from, and why it is not a measurement. The investigators took the observed difference and asked how many years of ordinary age-related decline it corresponds to, using a cross-sectional regression of performance against age in the same cohort. That is a legitimate way to make a small number intelligible. It is not a finding that anyone's brain was two years younger. An effect of 0.07 standard deviations is small by any conventional reading, and the translation into years is a rhetorical device applied to it afterwards.
And the trial that points the other way is rarely mentioned. The Physicians' Health Study II randomised 5,947 men, mean age 71.6, and followed cognition for an average of ten years. It found no multivitamin benefit on cognitive change. COSMOS is more recent, better instrumented and more consistent across its sub-studies; PHS II is larger in follow-up time. Both are real, and the field has not reconciled them.
One caveat worth stating before the grade. COSMOS was a US population, and part of what a multivitamin does in any population is correct marginal deficiencies. How much of the effect is that — rather than something a well-fed person would also get — is not established, and it is a reason to expect less rather than more from it in a country with a different diet.
Graded Moderate for a small effect on memory in older adults, and explicitly ungraded for anything a person would notice day to day, because no trial has tested that.
Consistency note, because this site says something different elsewhere. The cholesterol calculator places a multivitamin in its lowest tier and describes the evidence for cardiovascular events as weak to null. That remains correct, and it is not in conflict with this page: COSMOS tested cognition, not cardiovascular outcomes, and the evidence is genuinely better for the former. A multivitamin is a reasonable thing to take and a poor thing to rely on.
3. Choline — a real requirement, an oversold dementia claim, and an argument this site has already had
The nutrient is real and under-consumed. Choline is the precursor to acetylcholine and a structural component of cell membranes. There is no RDA, but an Adequate Intake has been set at 550 mg/day for men and 425 mg/day for women, rising to 450 mg in pregnancy and 550 mg in lactation. The tolerable upper limit is about 3,500 mg/day, and the ways people exceed it are supplements rather than food: high doses cause a fishy body odour, sweating and low blood pressure.
The dementia claim needs reading carefully. The Finnish cohort most often quoted (Kuopio Ischaemic Heart Disease Risk Factor Study, ~22 years, middle-aged and older men) found a 28% lower dementia risk in the highest intake group for phosphatidylcholine specifically — and total choline intake showed no statistically significant reduction on its own. That distinction is dropped in almost every retelling, including the ones that cite the study by name. It is observational, in men, in one country, and phosphatidylcholine intake tracks eggs and meat, which track a great many other things about a person's life.
And here it collides with the rest of this site. Phosphatidylcholine is also the substrate gut bacteria convert to TMAO, which the cardiovascular literature associates with higher risk. So the same dietary pattern is being recommended for the brain and questioned for the heart.
This page does not resolve that, because it is not resolved. What it will not do is quote the brain half and stay quiet about the other. The practical position both halves support: get choline from food rather than a supplement, and do not go hunting for more of it than the Adequate Intake on the strength of an observational subgroup. Two eggs and an otherwise ordinary diet reaches the target.
And for the reader who wants a decision rather than a tension: both halves of this argument are consistent with eggs in ordinary amounts. The brain half does not ask you to eat more than that; the heart half does not ask you to eat fewer. What neither supports is a choline supplement taken to push the number up.
The 2023 Japanese trial of 300 mg/day of egg yolk choline over 12 weeks reported improved verbal memory against placebo. Small, short, single-country, one supplement form. Graded Emerging.
4. Blueberries, matcha and the antioxidant argument
Blueberries. The best trial is Wood et al., American Journal of Clinical Nutrition, 2023 — King's College London — a 12-week randomised, double-blind study in 61 healthy adults aged 65–80. Participants took 26 g of freeze-dried wild blueberry powder containing 302 mg of anthocyanins, or a placebo matched for appearance, taste, fibre and vitamin C. Results: flow-mediated dilation improved by 0.86%, 24-hour ambulatory systolic blood pressure fell by 3.59 mmHg, and on the cognitive side immediate word recall and accuracy on a task-switching test improved. A 2025 meta-analysis in people who already had mild cognitive impairment found significant improvements in episodic and language memory.
What 26 g actually means, because it is the easiest number on this page to get wrong.
26 g is the freeze-dried powder, not the fruit. Freeze-drying removes almost all the water, so that powder is the equivalent of 178 g of fresh wild blueberries — which the investigators put at roughly 75 to 80 berries, or a generous cupful. That count assumes the large cultivated kind, which weigh around 2 g each. Wild lowbush berries weigh roughly a third of that, so 178 g of wild berries is a far larger number of them — the weight is the thing to go by, not the count.
Weigh out 26 g of actual blueberries and you get about ten. That is somewhere near an eighth of the dose that was tested, and it is a very easy mistake to make from the abstract, because both numbers appear in it and only one of them is food.
And they do not have to be wild. Wild blueberries carry more anthocyanin per gram than the large cultivated kind, which is why the trials use them — but the study's senior investigator has said plainly that other blueberry types have shown benefits in this literature too. Frozen are fine, and frozen wild blueberries are cheap, sold in most UK supermarkets, and keep. A handful on breakfast is a reasonable approximation of the trial. A teaspoon of powder stirred into a smoothie is not.
That is better evidence than most foods have. It is also 61 people for 12 weeks — and note which result was the strong one. The vascular endpoints moved with far more confidence than the cognitive ones, and 3.59 mmHg off 24-hour systolic pressure is a real number by the standards of the blood pressure chapter. That is a hint the mechanism is blood flow rather than anything specific to the brain — and blood flow is what the whole dementia chapter is about.
Graded Emerging for cognition and Moderate for the vascular effect, which is the honest split. Either way a bowl of blueberries is a good thing to eat, and unlike most of this page it is food rather than a capsule.
Matcha and EGCG. The study cited for neurogenesis — EGCG increasing neuronal survival in the adult hippocampus and producing more elaborate dendritic trees — is a rodent study. It is presented in popular coverage as though it described human brains. There is reasonable human evidence that L-theanine with caffeine improves attention acutely, which is a much smaller and much more defensible claim than reversing brain aging.
One safety point that rarely accompanies the recommendation: concentrated green tea extract has a documented hepatotoxicity signal at supplement doses, which is why several regulators have set limits on EGCG in food supplements. Drinking matcha is not the concern; buying high-dose EGCG capsules is a different act. Graded Emerging for cognition, and drink it because you like it.
5. Omega-3, which this site covers elsewhere
Low DHA is associated with smaller total brain volume and more white matter hyperintensity on MRI in the Framingham cohort. That is observational and cross-sectional, and the same caveat applies as everywhere else on this page.
The cholesterol calculator and its measurement chapter carry the trial evidence for omega-3 in full, including the comparator problem that makes REDUCE-IT contested. Rather than repeat it here and let the two accounts drift apart, this section points there. Nothing in the brain literature changes the dose or the argument.
6. What this page is not saying
None of the above is a reason to stop eating berries, drink less tea, or throw out a multivitamin. They are cheap, safe at food doses, and a multivitamin has more randomised support for cognition than most things in a health shop.
What it is saying is that the order matters. The interventions with the largest effects on how your brain works in twenty years are on the essentials page, and every one of them is unglamorous: blood pressure, hearing, movement, sleep, not smoking, staying connected to other people. A capsule is what people reach for because it is the easiest thing to do, not because it is the thing most likely to work.