Adults who reported eating more ultra-processed food were more likely to have certain blood-vessel-related marks on brain scans in a 2026 study. Their scores on two thinking tests, however, were not significantly different. That combination is the heart of the finding. A scan can detect a change that a short test does not, but this study cannot show that food caused the scan findings or predict who will have a stroke.
The research raises a useful everyday question: what can a person do with an uncertain finding without treating every packaged food as dangerous? The sensible answer is to look at the overall pattern of meals, make practical swaps when convenient, and keep the limits of the evidence in view.
What the researchers measured
The 2026 JAMA Network Open study (https://pubmed.ncbi.nlm.nih.gov/42879154/) analyzed 8,783 adults in a Canadian research cohort. Their average age was about 58. Researchers estimated how much of each person's diet came from ultra-processed foods, measured body fat, reviewed brain magnetic resonance imaging (MRI) scans, and administered two cognitive tests. They then compared people with different reported levels of intake while accounting for several other characteristics that might influence the results.
Ultra-processed foods are industrial formulations in the NOVA classification. The category can include sweet drinks, confectionery, packaged snacks, and ready meals. It also includes products with quite different nutritional profiles. A label based on processing is therefore a starting point for describing a diet, not a complete score of a food's health value.
The brain outcome combined covert infarcts and a high burden of white matter hyperintensities. A covert infarct is an area that looks like an older small interruption of blood flow, even though the person may never have noticed a stroke. White matter hyperintensities are bright spots on a particular MRI image. They can reflect changes in the tiny blood vessels that supply the brain. Neither mark means that a person currently has dementia, and the study did not follow participants to see whether a future stroke occurred.
The investigators also used the Montreal Cognitive Assessment and the Digital Symbol Substitution Test. These are structured tasks that sample memory, attention, and mental speed. They cannot capture every part of a person's day-to-day thinking. In this sample, greater ultra-processed food intake was associated with higher body fat and greater odds of the combined MRI outcome. It was not significantly associated with the cognitive test scores. The paper's imaging finding and its cognitive finding should be reported together.
Why one snapshot cannot establish cause
This was a cross-sectional analysis: diet information, scans, and tests were considered at roughly the same stage of participants' lives. It does not show what their brains looked like before they ate the reported diet. It cannot tell whether an individual scan mark developed before or after a dietary pattern began. Nor can it rule out the possibility that other differences between people helped produce the association.
For example, blood pressure, diabetes risk, smoking, activity, income, and access to food may relate both to what people eat and to blood-vessel health. Statistical adjustment can reduce some of that mixing, but it cannot measure every relevant influence perfectly. Food questionnaires also depend on memory and on how accurately foods can be placed in the NOVA category. These are reasons to avoid a headline that says ultra-processed food damages the brain.
The scan result deserves attention because the researchers measured an actual brain marker rather than relying only on symptoms. Yet the marker is still different from an outcome that readers can feel, such as a change in memory, a diagnosed stroke, or loss of independence. The study did not establish that replacing a product would reverse the MRI marks. A prospective study with repeated diet measurements and scans, or a controlled intervention that measures brain outcomes, would answer a stronger question.
How does this fit with other human research?
A Brazilian cohort study (https://pubmed.ncbi.nlm.nih.gov/36469335/) followed 10,775 adults for a median of eight years. People with more ultra-processed food in their diets had a slightly faster decline on tests of global thinking and executive function. Executive function includes abilities such as organizing and switching between tasks. That study looked at change over time, which the Canadian snapshot could not do. It still observed people's chosen diets rather than assigning them to eat particular foods, so other differences could have contributed. Its measured effect was small, and it did not include MRI scans to show what happened in the brain.
A UK Biobank analysis (https://pubmed.ncbi.nlm.nih.gov/40752041/) studied 58,423 adults for later neurological diagnoses and a smaller MRI group of 5,462 people. Higher reported ultra-processed food intake was associated with dementia diagnoses and some differences in brain structure. The MRI measures in that paper were mainly gray matter features, not the same vascular marks used in the Canadian study. The dietary measure was also based on the weight of foods rather than the share of calories, which makes a simple comparison of exposure levels misleading. As another observational study, it cannot prove that processing caused the outcomes.
The studies therefore point in a similar direction for some brain measures, but they do not form a single experiment. The 2026 Canadian study found no significant link with its cognitive tests, while the Brazilian study saw a small difference in the rate of change over years. Those results can coexist: a one-time test and a repeated test answer different questions, and the populations and measurements differ. They also warn us against saying that all studies found impaired thinking.
What does ultra-processed mean for an actual meal?
Processing alone does not tell the whole nutritional story. Washing, freezing, cooking, canning, and fermenting can make food easier to use without making it a poor choice. Even within the ultra-processed category, foods differ in fiber, salt, added sugar, and how often they replace more nourishing options. A packaged product can make a balanced meal possible for someone with limited money, time, energy, or kitchen access.
For a practical change, start with the foods you eat most often. Water or unsweetened tea can replace some sugary drinks. Oats with berries, nuts, or seeds can replace a sweet breakfast snack when that fits the day. Beans or lentils, frozen vegetables, and a whole grain can become a quick meal. Canned beans are convenient; rinsing them can reduce some added salt. These examples focus on adding useful foods rather than treating a single ingredient as a threat.
Someone who relies on packaged plant-based foods need not discard all of them. Compare products on ingredients and nutrition, and consider the whole meal. A fortified food can also have a useful role in a plant-based diet. The MRI study did not test any of these swaps, so they should be seen as reasonable ways to improve overall diet quality, not proven methods for preventing brain lesions.
Questions the study still leaves open
One question is whether all foods in the ultra-processed group carry the same association. The researchers estimated total intake. That does not establish that a sweet drink, a fortified plant-based staple, and a packaged meal have equal effects. A more detailed analysis would need to separate food groups, look at what each food replaced, and consider fiber and other nutrients. It would also need to avoid assuming that a single ingredient explains a broad diet pattern.
Another question is timing. The MRI marks may have accumulated over many years. A food questionnaire captures an estimate of a person's usual intake, not a full history from early adulthood. People also change their diets after learning that they have high blood pressure or diabetes. Such changes can blur the relationship between current eating and an older scan finding. Repeated measurements would help researchers see which came first.
The most important unanswered question is whether a change in meals changes a person's future health. A study could follow people who make a specific, feasible dietary change and compare later scans and real events such as stroke. That would take time and careful measurement. Until such evidence exists, a reader should avoid turning an association into a personal prediction. A person who eats more packaged food is not destined for a brain injury, and a person who eats very little is not guaranteed protection.
What should a reader take from the MRI finding?
The finding is a reason to study diet and vascular brain health more carefully. It is not a diagnosis for anyone who eats packaged food. It does not mean that a person with a bright spot on a scan will develop symptoms, or that a person with a normal scan can ignore blood pressure and other established health concerns.
If you have a sudden drooping face, weakness on one side, trouble speaking, or a sudden severe new neurological symptom, seek emergency care. Food changes are a long-term choice and do not replace urgent assessment. For ordinary meal planning, the useful direction is modest: make room for vegetables, fruit, legumes, whole grains, nuts, and other satisfying foods you can access and enjoy. Future research may tell us whether reducing particular ultra-processed foods changes MRI findings or clinical outcomes. This study alone cannot.
