Ultra-Processed Foods May Raise Brain Risk Beyond Diet Quality

A high Mediterranean, DASH, or MIND diet score may not tell the whole neurological story. In a large U.S. prospective cohort, a greater share of calories from ultra-processed foods was associated with higher risks of cognitive impairment and stroke even after researchers adjusted for all three dietary-pattern scores. That finding does not prove that processing causes brain disease. It does mean that a food can help a diet score while still carrying a processing-related risk signal.

The practical implication is straightforward: assess food quality at two levels. Ask whether the diet contains vegetables, legumes, whole grains, fish, and other foods rewarded by established diet patterns. Then ask how much of the diet comes from industrial formulations rather than recognizable foods. The second question has received far less attention, despite its relevance to the way many Americans eat.

Processing predicted brain outcomes after diet-quality adjustment

The strongest direct evidence comes from a 2024 Neurology analysis by Bhave et al. using the prospective REGARDS cohort of Black and White U.S. adults enrolled from 2003 to 2007. The investigators used a baseline food-frequency questionnaire, classified foods with the NOVA system, and modeled ultra-processed food intake as a percentage of the total diet.

For incident cognitive impairment, the analysis included 14,175 participants. Each 10-percentage-point increase in ultra-processed food intake was associated with a 16% higher hazard of cognitive impairment (hazard ratio 1.16, 95% confidence interval 1.09 to 1.24; p=0.0000101). For incident stroke, the analysis included 20,243 participants. The equivalent increase in ultra-processed food intake was associated with an 8% higher stroke hazard (hazard ratio 1.08, 95% confidence interval 1.02 to 1.14; p=0.0112).

The important design feature is the adjustment strategy. The models included Mediterranean, DASH, and MIND diet-adherence scores at the same time, along with age, sex, race, region, body mass index, physical activity, smoking, and cardiovascular comorbidities. In other words, the association did not disappear when researchers accounted for conventional measures of dietary quality. That is more informative than a simple comparison of people who ate more or less packaged food, although it still cannot remove every difference between those groups.

The inverse association pointed in the same direction. Each 10-percentage-point increase in unprocessed or minimally processed foods was associated with a 12% lower hazard of cognitive impairment (hazard ratio 0.88, 95% confidence interval 0.83 to 0.94; p=0.000183) and a 9% lower hazard of stroke (hazard ratio 0.91, 95% confidence interval 0.86 to 0.95; p=0.000213).

These are relative hazard estimates, not an individual forecast. The study does not provide an absolute risk change for a person who swaps 10% of calories, and it does not establish that one specific substitution will prevent a stroke or cognitive decline. It also did not report formal tests of whether risk rises linearly across intake levels, so no defensible threshold can be drawn from these results.

This is a common exposure, not a niche dietary issue

A 2024 BMJ umbrella review by Lane et al. synthesized 45 pooled epidemiological analyses across health outcomes. It reported that ultra-processed foods supply about 58% of dietary energy in the United States and 42% in Australia. At those levels, a 10-percentage-point shift is realistic enough to matter in population terms.

The review found adverse associations across cardiometabolic, mental-health, and mortality outcomes. It also described several plausible explanations: ultra-processed products often have high energy density, sodium, added sugar, and saturated fat, while providing less fiber and fewer micronutrients. Additives, altered food structure, microbiome changes, and inflammation are often proposed as additional pathways.

Plausible is not proven. The umbrella review found no poolable randomized trials of ultra-processed foods and health outcomes during its pilot screening. No human or animal experiment in this evidence base shows that a named additive, such as an emulsifier, directly causes neuroinflammation, blood-brain barrier damage, or stroke independently of the rest of the diet. The mechanism remains a hypothesis, not a verdict in a lab coat.

A healthy pattern may not offset high ultra-processing

The same basic concern appears outside neurological outcomes. In an Italian prospective cohort, Bonaccio et al.'s 2025 Moli-sani study found that high ultra-processed food consumption was associated with higher all-cause and cause-specific mortality even among people with high Mediterranean-diet adherence. Its endpoint was mortality, not cognitive impairment or stroke, so it cannot confirm the REGARDS brain findings. It does support the narrower point that a favorable pattern score may not fully capture the consequences of processing.

Diet patterns still matter. A 2025 prospective multi-cohort study by Shi et al. examined 10 dietary patterns and major brain disorders, identifying the MIND diet as the most favorable pattern across dementia, stroke, depression, and anxiety. That study did not test NOVA-classified ultra-processed foods as an independent exposure. It therefore reinforces the value of pattern quality without answering whether processing adds risk beyond it.

Likewise, a 2025 systematic review and meta-analysis of plant-based diets reported cognitive findings consistent with diets built around less processed plant foods. But plant-based studies assess dietary patterns; they cannot isolate processing as the reason for an association.

What the evidence cannot settle

Every central estimate here comes from observational research. People who eat more ultra-processed food may differ in income, food access, work schedules, health literacy, healthcare use, and other factors that statistical adjustment cannot measure perfectly. Reverse causation is less of a concern in a prospective design than in a cross-sectional survey, but residual confounding remains possible.

Exposure measurement also limits confidence. REGARDS relied on a single baseline food-frequency questionnaire. Diets change over time, and assigning questionnaire items to NOVA categories requires judgment. Researchers have not formally validated consistent NOVA classification across studies. The cohort also included only Black and White U.S. adults, in a food environment with unusually high ultra-processed food exposure, so the hazard ratios may not transfer to populations with much lower exposure.

REGARDS did report a statistically significant interaction by race for stroke: the ultra-processed-food association was larger in Black than White participants (interaction hazard ratio 1.15, 95% confidence interval 1.03 to 1.29; p=0.015). The study did not establish why, and the result needs replication before it should guide individualized risk estimates.

There is another reason to avoid overconfidence: published evidence available here lacks a well-powered prospective null finding for cognitive decline or stroke after comparable adjustment. That gap could reflect a real, consistent association. It could also reflect publication bias, limited research in lower-exposure settings, or incomplete replication. Absence of a null result is not proof that null results do not exist.

What to do with an association

The evidence supports treating ultra-processed-food reduction as a sensible precaution, not as a proven neurological treatment. Keep the foundations of a Mediterranean-, DASH-, or MIND-style diet, but inspect the products doing the work. Foods that meet a nutrient target yet arrive as industrial formulations deserve scrutiny.

Aim to shift some calories from NOVA Group 4 products toward unprocessed or minimally processed foods: vegetables, fruit, beans, intact grains, nuts, eggs, plain dairy, fish, and simply prepared meats. The REGARDS results quantify an association with such a shift; they do not guarantee an outcome for any one person.

There is no evidence-based cutoff for a safe share of ultra-processed calories, and no additive blacklist proven to protect the brain. The defensible move is less glamorous: make more of the diet recognizable before it becomes a label-reading contest. For neurological health, the available cohort evidence says processing belongs on the checklist alongside nutrients and dietary patterns.