Ultra-processed foods and dementia risk: what the 2026 research actually shows

Short answer: In most large studies, people who eat the most ultra-processed food have a modestly higher risk of dementia and cognitive impairment than people who eat the least — most estimates fall roughly 15–45% higher, depending on the study and the comparison. The link looks strongest when the eating happens in midlife, and one of the most useful findings is about substitution: in UK Biobank, replacing a tenth of ultra-processed food with unprocessed or minimally processed food was associated with 19% lower relative risk (in a study where 518 of 72,083 people developed dementia over about ten years). All of this is observational, some good studies find nothing, and no trial has tested dementia directly. The sensible response is a set of swaps, not a purity test.

This is educational content, not medical advice. It describes associations found in population research, not predictions about any individual.

What counts as "ultra-processed"?

Almost every study here uses the NOVA system, which sorts food by how much industrial processing it has been through rather than by its nutrients. Ultra-processed foods are industrial formulations built mostly from extracted or modified ingredients — think soft drinks, packaged snacks and sweets, mass-produced breads and breakfast cereals, reconstituted meats, instant noodles and many ready meals. They are cheap, convenient and engineered to be easy to eat a lot of.

That last point matters. In a tightly controlled 2019 NIH trial, 20 adults lived on a ward and ate either an ultra-processed or an unprocessed diet for two weeks each, with meals matched for calories, sugar, fat, fibre and salt on the plate. On the ultra-processed diet they chose to eat about 508 more calories a day and gained about 0.9 kg; on the unprocessed diet they lost about the same. That trial says nothing about the brain directly — but excess weight, type 2 diabetes and high blood pressure are all on the list of fourteen modifiable dementia risk factors, and this is one plausible route from the plate to them.

How big is the association?

Here are the studies worth knowing, from the largest syntheses down:

Two things jump out. First, the numbers vary a lot, and the 2024 meta-analysis has very high heterogeneity (I² = 97%) — the studies measured diet and outcomes differently, so the pooled figure is a rough summary, not a precise estimate. Second, the "moderate vs low" comparison in that same review was not significant. The signal lives mostly at the high end.

About that "58% higher risk" headline

The June 2026 Harvard-led study in the American Journal of Public Health made headlines with a 58% higher dementia risk for the highest consumers. That number is real, but read the interval: 1.003 to 2.48. The lower end is a hair above "no difference at all," and the test for a steady trend across the five intake groups was not significant (p = 0.50). It also helps to know the base rate: 266 of the 5,370 people in the study (5%) developed dementia during follow-up. The steadier result in the same paper is the one that got less attention: cognitive impairment short of dementia, 1.46 (1.13–1.88) with a significant trend (p = 0.03). If you want to know why the interval matters more than the headline, we wrote a whole guide to reading dementia statistics.

It may matter most in midlife

Three independent lines of evidence point the same way:

• In the Framingham Heart Study (1,375 people, 12.7 years of follow-up), each daily serving of ultra-processed food was associated with 13% higher Alzheimer's risk (HR 1.13, 1.03–1.25) — but only in people younger than 68 at the start. In those 68 and older, there was no significant association.
• The 2026 meta-analysis found the same shape: HR 1.19 (1.06–1.33) in people under 60, and no significant association in those over 60.
• A 2026 Dutch cohort of 1,371 adults aged 55 and over (the Longitudinal Aging Study Amsterdam) found no association at all between ultra-processed food and cognitive decline over ten years.

A midlife-weighted pattern fits the wider dementia picture: midlife is when vascular conditions such as high blood pressure and type 2 diabetes appear to shape later dementia risk. It also means the case for acting is strongest for the people who least often think of dementia as their problem. It does not mean diet stops mattering later — and for a different reason: in older age, eating enough protein and energy often matters more than purity, and unplanned weight loss deserves a conversation with a clinician.

Can changing what you eat change the risk?

No trial has randomised people to cut ultra-processed food and then counted dementia cases. The closest evidence is observational but useful:

• Substitution. In UK Biobank, replacing 10% of ultra-processed food by weight with unprocessed or minimally processed food was associated with 19% lower relative dementia risk (HR 0.81, 0.74–0.89). For scale: 518 of the study's 72,083 participants developed dementia over a median of ten years, so the absolute difference for any one person is small.
• Change over time. A 2026 analysis of 7,555 REGARDS participants measured diet twice, about ten years apart. People whose ultra-processed share rose by 10% had higher subsequent risk of cognitive impairment (HR 1.24, 1.12–1.37 by weight); each one-point rise in MIND-diet score went with lower risk (HR 0.90, 0.85–0.96). The outcome was a short screening test, not a dementia diagnosis, so treat it as supportive rather than decisive.

The honest summary: people who move away from ultra-processed food, and towards the kind of eating described in the MIND diet, tend to do better. Whether the change causes the difference has not been tested.

Why the evidence is harder than it looks

• Diet is measured by questionnaire. People misremember and under-report — and NOVA classification from a food-frequency questionnaire is approximate.
• Confounding. Heavy ultra-processed eating travels with lower income, less education, less time and more stress. Studies adjust for these, but only as well as they measure them.
• Not every outcome agrees. A 2025 meta-analysis of 28 studies found higher intake linked to cognitive impairment (RR 1.17, 1.06–1.30) but not significantly to Alzheimer's disease or dementia on their own. And an October 2026 Canadian study of 8,783 adults aged 35 to 69, a single snapshot rather than a follow-up study, linked the highest intake to more body fat and higher odds of silent vascular brain injury on MRI (OR 1.40, 1.10–1.77), but found no association with scores on two thinking tests.
• Some "dementia risk" measures are circular. A 2026 Australian study of 2,192 adults linked each 10% more ultra-processed food to slightly lower attention scores (−0.05) and a higher score on the CAIDE risk tool (+0.24 points). CAIDE is built partly from weight, blood pressure and cholesterol — things diet affects directly — so that result is closer to "diet affects risk factors" than "diet affects the brain."
• A retracted paper still circulates. A 2023 Framingham Offspring analysis in Alzheimer's & Dementia was retracted because its authors lacked the required data-use approval — a process failure, not a finding. A separate Framingham analysis by a different research team was published in 2025 and is the one cited here. If you see "Framingham" and "ultra-processed" together, check which one it is.

What is actually worth doing

1. Swap, don't purge. The substitution finding is about a tenth of the diet. Pick the two or three ultra-processed items you eat most often and find a less-processed swap for each.

2. Start with drinks. Sugary drinks are ultra-processed, easy to replace, and have their own evidence on dementia risk. Water, plain coffee or tea and sparkling water are the easy defaults.

3. Then processed meats and packaged snacks. These are the categories most consistently linked to worse outcomes in our guide to foods linked with higher dementia risk. Nuts, fruit, yogurt, beans and eggs are practical replacements.

4. Don't panic over every label. Ultra-processed foods are not all nutritionally alike, and the studies above measured the overall share of the diet, not single products. The risk signal sits in that overall pattern, especially at the high end, not in any single item.

5. Watch the downstream numbers. If your diet has been heavy on ultra-processed food, the most useful next step may be checking what it has done to your blood pressure and blood sugar — and discussing the results with your clinician.

Diet is one lever among many. The free Solenna risk profile takes about ten minutes and shows which of the modifiable factors — diet, blood pressure, activity, hearing, sleep and the rest — are most worth your attention, so you can put effort where it counts.

How we research and review our content →

An independent physician audited the Solenna app. We published all nine findings →

Educational content only. It does not diagnose any condition and is not a substitute for professional medical advice. If you have diabetes, kidney disease, a history of disordered eating or unexplained weight loss, discuss dietary changes with your clinician.