Diet & Supplements for Prevention

The supplement aisle has a dementia section now, and the trial record has a verdict on it: null, with two partial exceptions. Ginkgo, vitamin E, DHA, curcumin, and even a branded brain diet have each failed dedicated prevention trials. This page documents the nulls honestly — then walks the two signals that survived, and what diet evidence actually shows.

Why Pills Face a Brutal Trial

Alzheimer's pathology begins fifteen to twenty years before symptoms, which makes prevention trials structurally punishing: enroll thousands of healthy people, treat them for years, and wait for endpoints that may never arrive — while the disease, if it comes, was already brewing before the first capsule. Against that backdrop, the supplement industry enjoys an asymmetry. Observational studies generate hope almost for free: people who take fish oil differ from people who do not in income, education, exercise, and everything else, and some of that washes into "fish oil protects the brain." Randomized trials then test the isolated pill against placebo — and, one after another, the hope evaporates.

The pattern has now repeated often enough to be a law of this literature: cohorts suggest, trials dismiss. The chart below is the scoreboard.

The Wall of Nulls

Four flat grey bars: no effect. One amber bar: a modest signal from a single trial. One accent bar: a real subgroup finding that most people misunderstand. The rest of this page unpacks each row.

The Classic Nulls

• 🌿 Ginkgo biloba. The most studied candidate: the GEM trial randomized 3,069 adults aged 75+ to ginkgo or placebo for a median 6.1 years and found no difference in dementia incidence — 3.3 vs 2.9 cases per 100 person-years (DeKosky et al., JAMA, 2008). A Cochrane review reached the same conclusion (Birks & Grimley Evans, 2009). Two decades of marketing, zero prevention.
• 💊 Vitamin E. The oxidative-stress hypothesis sounded right; the trial did not cooperate. In 769 people with mild cognitive impairment followed for three years, 2,000 IU of vitamin E daily did not delay progression to Alzheimer's (Petersen et al., NEJM, 2005). High-dose vitamin E also carries its own risks, which makes the null doubly decisive.
• 🐟 DHA and omega-3s. Two large tests: DHA at 2 g/day for 18 months in 402 people with established Alzheimer's — no slowing of decline (Quinn et al., JAMA, 2010); and the AREDS2 cognitive substudy, roughly 3,500 older adults followed five years — no effect of omega-3 or lutein/zeaxanthin on cognition (Chew et al., JAMA, 2015). Omega-3s may help the heart; the brain-prevention claim failed.
• 🧪 Curcumin. Epidemiologic lore about Indian curry consumption met pharmacology: a 24-week randomized trial in Alzheimer's patients found no cognitive benefit, with absorption problems muddying the dose (Ringman et al., 2012).
• 🧠 "Nootropic" stacks. The evidence-based cognitive enhancers remain caffeine, exercise, and sleep — the drugs and cognition topic keeps that list honest. Everything else in the category is a thin trial behind thick marketing.

The Scoreboard

One interpretive note on the verdict column: "null for dementia prevention" does not mean worthless for everything. Omega-3s have cardiovascular evidence, and vitamin D has a job to do when levels are actually low. The column answers one question — does this prevent dementia? — and the answer is what it is.

The Two Honest Exceptions

Two signals survived the null wave, and both deserve precise framing — because both are routinely oversold:

• 🧬 B vitamins — a subgroup story about homocysteine. The VITACOG trial gave 266 people with MCI a B-vitamin combination (folate 0.8 mg, B12 0.5 mg, B6 20 mg) or placebo for two years. Whole-brain atrophy slowed by roughly 30% overall — but the effect concentrated almost entirely in participants with elevated homocysteine (above ~11.3 μmol/L), where a follow-up analysis showed markedly reduced atrophy in Alzheimer's-vulnerable brain regions (Smith et al., 2010; Douaud et al., PNAS, 2013). The honest translation: B vitamins are not a prevention supplement for well-nourished people. They are a treatment for a measurable abnormality. If homocysteine runs high, correcting it with a clinician is a different project from sprinkling B12 on a healthy diet.
• 💊 Multivitamin — a single modest signal. COSMOS-Mind randomized 2,262 older adults to a daily multivitamin or placebo for three years. Global cognition improved modestly (about 0.07 standard deviations — roughly equivalent to 1.8 years of typical age-related change), with larger memory effects; cocoa extract did nothing (Baker et al., Alzheimer's & Dementia, 2023). That is a real but small result from one trial, and it needs replication before anyone should plan a life around it. Modest, single, pending — those three words are the frame.

⚠️ Clinician territory, stated once: if you have MCI or a dementia diagnosis, any supplement — including B vitamins at trial doses — belongs in a conversation with your clinician, not an online checkout. Trial doses are not grocery doses, and interactions are real.

Diet vs Pills

The diet evidence splits the same way, and the split is instructive. In PREDIMED, older adults at high cardiovascular risk randomized to a Mediterranean diet with extra-virgin olive oil or nuts maintained better cognition over roughly four years than a low-fat control group (Valls-Pedret et al., JAMA Internal Medicine, 2015) — a dietary pattern beating a default diet, in a population with vascular problems to fix. The MIND diet trial then asked the sharper question: does a branded "brain diet" beat any other decent diet? It did not — 604 at-risk older adults on MIND or a calorie-matched control diet both improved slightly, with no difference between groups (Barnes et al., NEJM, 2023).

The honest synthesis: eating patterns matter — the Mediterranean diet topic owns that evidence — but no branded diet has demonstrated specific superiority for dementia prevention, and the null in the MIND trial says the brand was never the active ingredient. Diet also works best as one member of a stack: in FINGER it was one of four simultaneous levers, and the parent topic documents what the package achieved. Finally, the boundary that keeps this page honest: correcting a measured deficiency — B12, D, iron, homocysteine — is not prevention, it is repair, and the audit's blood panel is where those measurements live.

Related Topics