Diet, Your Metabolome, and APOE4: What the 2025 Mediterranean-Diet Study Actually Found

If you carry two copies of APOE4, closely following a Mediterranean diet was associated with about 35% lower dementia risk, the biggest benefit of any genetic group. That result landed on the exact people who have spent years being told their genes had already settled the question.

It comes from an August 2025 study in Nature Medicine led by Yuxi Liu and colleagues at Mass General Brigham, the Harvard Chan School, and the Broad Institute. What lifts it above another “eat your vegetables” headline is the design. The team did not just track who ate well and who got dementia. They measured what the diet did to people’s blood chemistry, and the protection was strongest in the carriers with the most to lose.

What they actually found

The 35% is best understood as a slope, not a one-time discount. In the genotyped sample of about 16,500 women, each one-point rise in a person’s Mediterranean-diet score came with roughly 35% lower dementia risk for two-copy carriers, against only about 4 to 5% for people with one copy or none. That gap is the real story: the same food moved the needle far more for the highest-risk genotype.

It holds when you slice it the other way. Comparing the closest adherers to the loosest, two-copy carriers had about 28% lower risk (hazard ratio 0.72, 95% confidence interval 0.58 to 0.89), again the strongest of any genotype, and it replicated in a separate group of about 1,490 men. The study followed these groups from 1989 to 2023, with 485 dementia cases among the 4,215 women and 121 among the men, so this is a long-term signal, not a snapshot.

Now the honest part, because a relative cut means nothing without the baseline it applies to. Someone with two APOE4 copies carries a lifetime Alzheimer’s risk in the range of 40 to 55% by age 85 (see APOE4 genotypes explained and is Alzheimer’s inevitable for homozygotes). Trimming a number that high by a quarter to a third is meaningful help, but it is not a cure. It is also the opposite of what a “genes are destiny” model predicts: the diet did the most for the people who needed it most.

Why your blood chemistry is the missing link

Most diet-and-dementia studies stop at “people who ate this way did better.” This one asked why, by reading each person’s metabolome, the thousands of small molecules circulating in the blood that show what your body is actually doing with your food and your genes.

The pattern pointed straight at fat handling, which is exactly where APOE does its job. APOE is, at its core, a protein that ferries cholesterol and other lipids around the body and brain, and the E4 version does that poorly. In carriers, two families of cholesterol-linked lipids (cholesteryl esters and sphingomyelins) tracked with higher dementia risk, while a third family (glycerides) tracked with lower risk. Those links were sharpest in two-copy carriers and largely absent in everyone else, which fits what we already know about E4: clumsy lipid handling feeds the inflammation and the sluggish amyloid clearance behind the disease. (More on that thread in APOE4, cholesterol and the heart and APOE4 and saturated fat.)

Here is the finding that earns the study its place. About 40% of the diet’s benefit in carriers appeared to run through these blood-chemistry shifts, and that link showed up only in carriers, not in non-carriers or the group as a whole. In plain terms, the diet does not look like magic. It looks like it is correcting the exact lipid signals E4 gets wrong. The authors treat this causal layer as exploratory rather than proven, which is the right call for an observational study.

One more honest note: the metabolome was not even the best predictor here. A validated blood marker for Alzheimer’s, p-tau217, flagged roughly threefold higher dementia risk comparing the top and bottom quarters, outperforming the broad metabolite panel (see blood-based biomarkers for Alzheimer’s).

The numbers in one place

What	Figure	Context
Diet effect, two copies, per 1-point score rise	~35% lower risk	vs ~4-5% for one copy or none
Diet effect, two copies, closest vs loosest adherence	HR 0.72 (CI 0.58-0.89)	strongest of any genotype; replicated in men
Lifetime Alzheimer’s risk, two copies (background)	~40-55% by age 85	the baseline a relative cut applies to
Benefit running through blood-chemistry shifts	~40%	carriers only; absent in non-carriers; exploratory
p-tau217, top vs bottom quarter	~3x risk	outpredicted the metabolite panel
Study scale	485 cases / 4,215 women; 121 / 1,490 men	followed 1989 to 2023, replicated across cohorts

What this doesn’t prove

Keep your skeptic’s hat on, because three limits matter. This is observational, built from food questionnaires rather than a randomized trial, and people who eat this way differ in other ways the analysis cannot fully erase (a good habit to bring to any headline: reading a study like a skeptic). The two-copy group is also small by nature, since only about 2 to 3% of people carry two copies, so its confidence intervals are wide. The reassuring counterweight is that the effect replicated across two independent cohorts. And adding the metabolite data improved actual risk prediction only modestly, mostly in the early years of follow-up, so this is a window into mechanism, not a crystal-ball blood test.

What to actually do

“Eat better” is not a protocol, so here is the concrete version of the pattern the study rewarded: heavy on vegetables, fruit, nuts, whole grains, legumes, fish, and olive oil, and light on red and processed meat and alcohol.

A defensible weekly template if you carry APOE4:

• Make olive oil your default fat. Extra virgin, used generously. This is your main monounsaturated-fat lever.
• Eat fish about twice a week, oily fish included. The DHA angle has its own carrier wrinkle in omega-3, DHA and APOE4.
• Lean on plants daily: legumes most days, a small handful of nuts, and vegetables as the bulk of the plate.
• Choose whole grains over refined, and push red and processed meat to occasional.
• Rethink the wine. Older diet scores gave points for moderate alcohol, but the pattern here was low in it, and for an APOE4 brain the math is different. Read alcohol and the APOE4 brain before treating wine as health food, and do not start drinking because an index rewards it.

For tracking, a standard lipid panel is the obvious start, and ApoB rather than LDL gives a cleaner read of the particle burden E4 mishandles. If you do not know your genotype yet, how to get tested for APOE4 is step zero. And keep this in proportion: the strongest prevention evidence still comes from doing several things at once, as in the FINGER trial, which pairs diet with exercise, sleep, and blood pressure control.

Common questions

Does this mean I can out-eat my genes? Not exactly. It is strong observational evidence that diet substantially shifts risk through a believable lipid mechanism, in a small but replicated group of high-risk carriers. That is real reason for optimism, not a guarantee, and a third off a high baseline still leaves elevated risk.

Should I get metabolomic testing? Not yet for routine use. The metabolite panel barely improved prediction over what clinicians already have, and p-tau217 outdid it. Put your energy into the diet, your ApoB, and the rest of the toolkit you can act on today.

Sources

• Liu Y, et al. Interplay of genetic predisposition, plasma metabolome and Mediterranean diet in dementia risk and cognitive function. Nature Medicine, 2025. 10.1038/s41591-025-03891-5
• Alzforum: Metabolites Spurred by Mediterranean Diet May Fend Off Dementia
• Medscape: Can a Mediterranean Diet Offset Genetic Alzheimer’s Risk?
• News-Medical: Mediterranean diet lowers dementia risk by altering key metabolites
• CNN: Mediterranean diet lowers risk of dementia by 35% in people at most risk
• National Institute on Aging: Study reveals how APOE4 gene may increase risk for dementia