Peanuts and Brain Health: What That New Trial Actually Found

The headline, and the thing underneath it

At the end of October 2026 a tidy little story went round: eat two servings of peanuts a day, improve insulin signalling in the memory centres of your aging brain. Lovely headline — cheap food, a brain scan, and the word hippocampus, which is the neuroscience equivalent of a puppy in an advert.

So we opened the paper.^[1] It is competent work by a serious group: Lucia Kerkhof and colleagues at Maastricht University Medical Center, who have ground away at nuts and vascular function for years. The problem is not the science. It is the gap between what they measured and what you were told they measured.

What the new peanut brain study actually did

Thirty-one healthy adults, average age 67 ± 4, ate 60 g a day of unsalted, skin-roasted peanuts for 16 weeks, then an 8-week washout, then 16 weeks of no peanuts as the control. Crossover, so everyone did both arms. Median compliance was 100%, which for a four-month food intervention suggests people quite like being told to eat peanuts.^[1]

The outcome was not memory. It was brain insulin responsiveness, measured in a wonderfully strange way: insulin goes up the nose, reaching the brain without flooding the bloodstream, and you watch how cerebral blood flow responds on arterial spin labeling MRI. A brain that responds still has working insulin signalling; one that does not is, roughly, insulin resistant. Since brain insulin resistance keeps turning up alongside cognitive decline, the logic is sound. Liver fat, fasting glucose, lipids and inflammation were measured too.

The result, with the error bars left in

A whole-brain analysis — meaning you search the entire brain rather than a region you picked in advance — found one region, in the right hemisphere, where the insulin-driven blood flow response differed after peanuts.

Three things jump out. First, p = 0.044 from a whole-brain search in 31 people is significance by the skin of its teeth — the statistical equivalent of a coin landing on its edge. Second, the effect size is −5.4 with a standard deviation of ±5.4: when the spread matches the signal, individual people were going in opposite directions.

Third, and this is the one that bothers us, nobody actually knows where that blob was. The Harvard-Oxford atlas assigns probabilities, and the highest was 20%, for the lingual gyrus — primarily visual cortex. The hippocampus, star of every headline, came in at 14%. Calling a 20%-confident cluster a “memory region” is a choice. The authors are far more careful than their coverage: their conclusion says “potentially beneficial” and “may contribute.” Those hedges were load-bearing, and they fell off in translation.

One last wrinkle: the finding was a reduced blood-flow response to intranasal insulin, which the authors read as improved responsiveness. Defensible — this literature is full of blunted and exaggerated responses meaning different things in different tissues — but it is an interpretation, not a reading off a dial. A smaller number in a half-cubic-centimetre cluster still has to be argued into meaning something good.

Wait — this is the same 31 people as last year

Thirty seconds on the trial registry reframes everything. This is not a new study. Both papers come from one trial, NCT05724654, which ran February 2023 to July 2024 with an actual enrolment of 31.^[3] The first, in Clinical Nutrition in late 2025, reported the blood flow and memory outcomes.^[2] This one reports the brain-insulin mechanism in the same participants.

That sequence is completely normal science: publish the primary outcome, then the mechanistic sub-study. What is not fine is the impression the coverage leaves — that evidence is accumulating, that a second study has confirmed the first. It has not. One trial of 31 people has been described to you twice. We have written before about how a single study becomes a wave of headlines; this is the quieter version of the same trick. For what it is worth, the original paper also drew a published critique in Clinical Nutrition this year, with a reply from the Maastricht group^[4][5] — a sign specialists did not treat it as settled either.

The memory result, counted honestly

So what did the first paper find? Real things, modestly sized. Global cerebral blood flow rose 3.6% (1.5 mL/100 g/min, 95% CI 0.3 to 2.6, p = 0.014) and grey matter blood flow 4.5% (p = 0.002) — solid numbers on the pre-specified primary outcome, and the strongest part of the story.^[2]

Memory is where it thins out. Verbal memory improved 5.8% on the delayed-recall condition of one task, which in plain units means 1.4 more words remembered (95% CI 0.0 to 2.7, p = 0.043). A lower bound of zero is the statistics telling you “no effect at all” is still on the table. Executive function and psychomotor speed showed nothing. One task, one condition, 1.4 words, 31 people: a hint, not a reason to restructure your diet.

The finding nobody tweeted

Our actual editorial point is buried in the back half of the first paper. During the peanut arm, systolic blood pressure fell by 5 mmHg (95% CI −8 to −2, p = 0.004) and pulse pressure by 4 mmHg (95% CI −7 to −1, p = 0.006).^[2]

Now compare the p-values. Viral brain-insulin cluster: 0.044. Memory: 0.043. Blood pressure: 0.004. By an order of magnitude, the most robust thing that happened here was a blood pressure change — and it is the one result with a large, independent, well-replicated literature tying it to brain aging, midlife hypertension being among the better-established modifiable risk factors for later cognitive decline. A sustained 5 mmHg drop is not trivial.

Which makes the honest version of this story almost the inverse of the headline. Peanuts probably did help these brains slightly, most plausibly by doing something boring and vascular — better flow, lower pressure — which is also the likeliest explanation for the 1.4 words. The exotic brain-insulin mechanism is the weakest link in the chain, and it is the one that got the press release.

If a mechanism must be named: peanuts are unusually rich in L-arginine (C₆H₁₄N₄O₂), the substrate your endothelium turns into nitric oxide (NO) to relax blood vessels, and their fat is largely oleic acid (C₁₈H₃₄O₂). Both fit a vascular story. Neither was tested here — so treat that as a hypothesis with good manners, not a finding.

Three caveats that matter more than the p-values

1. The control arm ate nothing

This was peanuts versus no peanuts, not peanuts versus a calorie-matched snack. So the trial cannot separate peanuts from 350 extra calories of fat, protein, magnesium, niacin and fibre — or from whatever those calories displaced. For a vascular outcome, that is a real gap.

2. Single-blind means the participants knew

You cannot blind someone to eating 60 g of peanuts a day. Everyone in the peanut arm knew they were in the healthy condition for four months, and both blood pressure and cognitive test scores are susceptible to that. The crossover design helps; it does not erase it.

3. Who paid

The Peanut Institute Foundation is listed as a collaborator on the trial registry.^[3] The authors state the funder had no role in design, conduct, analysis or writing, and we have no reason to doubt it — but industry money reliably shifts which questions get asked and which results get amplified. In fairness, the famous Harvard nut-mortality analysis was part-funded by the International Tree Nut Council.^[6] Almost all nut research has a nut body in the acknowledgements. That is the field, not a scandal; it is just why we read methods instead of summaries.

Zoom out: the case for peanuts is actually strong

None of this means skip the peanuts. It means the brain scan is the wrong reason to eat them, because a much better one has existed for a decade.

In two Harvard cohorts totalling 118,962 people followed up to 30 years — over three million person-years, 27,429 deaths — nut intake tracked inversely with total mortality in a clean dose-response. Against people eating no nuts, the pooled hazard ratio was 0.93 for less than weekly, falling to 0.80 (95% CI 0.73 to 0.86) at seven or more times a week, p < 0.001 for trend.^[6]

And for peanuts specifically — which matters, because peanuts are cheap legumes rather than expensive tree nuts — a Vanderbilt-led analysis across three cohorts and 14,440 deaths, deliberately including low-income American and Chinese populations where most nut intake is peanuts, found hazard ratios of 0.79 and 0.83 for the highest versus lowest quintiles, driven mainly by cardiovascular death.^[7] Their closing line is the useful one: peanuts, given their affordability, may be a cost-effective way to improve cardiovascular health.

Observational data, with all the usual baggage — nut eaters tend to do other sensible things. But it is enormous, consistent across ethnic groups and income levels, and it points the same way as the trial's blood pressure finding. That is the shape evidence takes when something is probably true, and it is the pattern we keep hitting in food before pills: the boring whole food outscores the extracted compound.

What we would actually do

Eat peanuts. Unsalted, skins on if you can get them, because that is what was tested. Treat them as a swap for something worse rather than an addition to everything else: 60 g is about 350 calories and roughly 30 g of fat, so bolting it onto an unchanged diet buys you weight gain, a reliably bad trade for your brain.

Do not expect to notice anything — a 1.4-word memory gain is invisible from the inside. If you want to know whether it is doing something, watch your blood pressure cuff rather than your recall. It is the best-evidenced outcome here and the one you can measure at home for twenty euros.

And keep the ranking straight. Peanuts sit in the same drawer as the other cheap, unglamorous, well-evidenced things — the free levers of sleep, blood pressure, fitness and strength. Not the drawer with the short list of supplements that clear the bar, and certainly not one needing an MRI to justify it. The people who keep their memories into their 80s, as far as the SuperAgers work can tell, are not optimising one food. They are maintaining a system.

The clean verdict

A good group ran a careful small trial and published its mechanism honestly, hedges intact. Then a press cycle sanded off the hedges, promoted a 20%-probability location to “the hippocampus,” and sold the second paper from one 31-person study as confirmation of the first.

Underneath it, the signal looks real but modest and probably vascular: more blood flow, lower pressure, maybe 1.4 extra words. Add two decades of cohort data showing nut eaters die less, and peanuts come out as exactly what they are — a cheap, boring, genuinely good food. Fine conclusion. It just did not need an MRI scanner. For more of this sort of thing, our hype detox collection is where the louder claims go to be measured.