For years, the phrase “ultra-processed” has functioned as a one-word verdict on a bowl of cereal, a can of beans, or a glass of plant-based milk — a shorthand for dietary danger that lumped them alongside potato chips and soda. A sweeping Harvard analysis presented yesterday at the largest annual nutrition science conference in the world suggests that verdict has been doing real harm to real diets
Researchers from Harvard T.H. Chan School of Public Health analyzed data from more than 200,000 U.S. adults tracked across four decades — one of the largest dietary investigations ever conducted — and found that overall nutritional quality, not how industrially processed a food happens to be, was the dominant driver of heart disease, type 2 diabetes, and mortality risk, according to findings presented at 8:54 a.m. ET on Tuesday, July 28 at NUTRITION 2026 in National Harbor, Maryland. Those findings pose a pointed challenge to the public health establishment’s increasingly vocal anti-processing consensus — and to the classification tool that consensus is built on.
The practical implication is specific and directly actionable: when a reader sets down a package of fortified whole-grain cereal and picks up a bag of white rice instead because the cereal’s ingredient list triggered an app’s “ultra-processed” warning, that trade may be making their diet worse, not better. The study’s lead researcher, Xiaowen Wang, PhD, a postdoctoral research fellow in the Department of Nutrition at Harvard T.H. Chan School of Public Health, put it plainly: “Advising people to universally ‘avoid ultra-processed foods’ can lead to lower diet quality since healthy ultra-processed foods are beneficial and unhealthy minimally-processed foods are detrimental,” Wang told the American Society for Nutrition.
NOVA System Classifies by Process, Not Nutrition
To understand the study, it helps to understand the tool it challenged. The NOVA classification system — developed in 2009 at the University of São Paulo by epidemiologist Carlos Augusto Monteiro — sorts foods into four groups based entirely on the extent and purpose of industrial processing applied to them, not on their nutritional content. Group 1 is unprocessed or minimally processed (fresh fruits, plain yogurt, milk). Group 2 is processed culinary ingredients (olive oil, salt, sugar). Group 3 is processed foods (cheese, canned vegetables, salted nuts). Group 4 — “ultra-processed” — covers industrially manufactured products containing food substances rarely used in kitchen cooking: high-fructose corn syrup, hydrogenated oils, protein isolates, cosmetic additives like emulsifiers and artificial colors.
NOVA’s own documentation is explicit on one critical point: the classification does not comment on the nutritional value of foods and is not intended for nutrient profiling. That is a principled design choice — Monteiro’s argument is that the industrial process itself signals something about food, independent of what comes out of it. But it creates a structural problem: nutritious foods can be classified as ultra-processed while nutritionally empty foods can be classified as minimally processed. Canned lentil soup lands in Group 4 alongside sugary soda. Fruit juice and white rice, which deliver little beyond rapidly absorbed carbohydrates, stay in Groups 1 and 3. When dietary guidance built on NOVA tells people to “avoid ultra-processed foods,” it cannot distinguish between these cases.
The Wang study was specifically designed to exploit that gap. Rather than applying NOVA alone — which would confound processing level with diet quality, since the two variables naturally travel together in real-world eating — the research team applied NOVA alongside the plant-based diet index (PDI), a scoring tool that measures the overall nutritional quality and healthfulness of a diet independently of how foods were made. The PDI assigns positive scores to healthful plant foods (whole grains, vegetables, legumes, nuts) and negative scores to unhealthful foods regardless of their processing tier, allowing researchers to hold nutritional quality constant and then ask what processing level alone contributes. Using data pooled from three major longitudinal cohort studies tracking more than 200,000 U.S. adults from the 1980s through the 2020s, the team sorted every participant into one of four diet profiles: healthy/ultra-processed, healthy/minimally-processed, unhealthy/ultra-processed, and unhealthy/minimally-processed.
The results were clear. About half of all participants eventually died, developed heart disease, or were diagnosed with type 2 diabetes over the study period. Diet quality predicted risk powerfully — healthy diets were associated with substantially lower risk; unhealthy diets raised it. But within each diet-quality tier, risk was nearly identical regardless of whether foods were ultra-processed or minimally processed. A bowl of fortified whole-grain cereal with plant-based milk was as protective as a bowl of hand-cooked oats. A meal of white rice and fruit juice was as risky as a fast-food meal. The NOVA tier, held constant for quality, contributed essentially nothing to the risk picture, according to the study’s findings.
NOVA’s Critics Have a Point
The study’s findings land in the middle of a live academic debate about NOVA itself. A 2025 review published in the Proceedings of the Nutrition Society found substantial heterogeneity in health outcomes across ultra-processed food subtypes, noting that the NOVA classification’s binary approach fails to account for nutritional composition, fortification benefits, and cultural food traditions. Products like sugar-sweetened beverages were consistently associated with adverse outcomes, the review found, while fortified cereals and certain dairy products showed neutral or protective effects — precisely the distinction NOVA cannot make.
A multi-institution effort to restructure NOVA has been running since January 2025 under Prof. Susanne Bügel of the University of Copenhagen, aimed at incorporating nutritional content and food matrix effects — the chemical and physical properties of food that affect how it is digested and metabolized — into a next-generation classification system. The project carries a significant conflict of interest that has not received wide coverage: it is funded by the Novo Nordisk Foundation, the charitable arm of the Danish pharmaceutical company that makes Ozempic and Wegovy. NOVA’s creator, Carlos Monteiro, publicly repudiated the initiative, telling its lead researcher not to use the NOVA name or claim any connection with his system, and characterizing the project’s framing as “an effort to delegitimise independent research and promote an alternative classification system more favourable to industry.” Bügel has maintained that the project’s scientific goals have not changed.
NOVA’s defenders have a substantive response. Monteiro’s core claim is that the industrial food manufacturing process — specifically, the use of substances like hydrogenated oils, high-fructose corn syrup, and cosmetic additives that are not cooking ingredients — produces foods engineered to be hyperpalatable and to override satiety signaling in ways that nutritional profiling cannot capture. The most rigorous evidence in support of this mechanism comes from a 2019 randomized controlled trial at the NIH Clinical Center led by Kevin Hall, PhD, a senior investigator at the National Institute of Diabetes and Digestive and Kidney Diseases. In that study, 20 healthy adults were admitted to a metabolic unit for 28 days and assigned to either an ultra-processed diet or an unprocessed diet for two weeks at a time, in random order. Crucially, the two diets were matched for total calories, energy density, macronutrients, sugar, sodium, and fiber — eliminating the nutritional variables that NOVA critics point to. Participants were told to eat as much or as little as they wished. On the ultra-processed diet, they spontaneously consumed a mean of 508 more kilocalories per day, gained an average of 0.9 kilograms (about 2 pounds), and lost weight during the unprocessed phase.
That finding — excess calorie intake that persisted even when the nutritional profiles were held constant — is not something the Wang observational study can explain away. If processing affects appetite regulation independently of nutritional quality, then two diets with identical quality scores but different processing levels may not be equally healthy in the long run, even if their disease associations look similar in cohort data
Read more:AI Diet Tools Have a Multicultural Blind Spot: NUTRITION 2026 Convenes Saturday in National Harbor
Recent Research Shows a More Complex Picture
The Harvard findings do not stand alone, and they do not settle the debate. A separate Harvard analysis published in The BMJ in May 2024, drawing on more than 114,000 participants across the Nurses’ Health Study and the Health Professionals Follow-up Study — followed for a median of more than 30 years — found that certain ultra-processed food subcategories, particularly processed meats, sugary breakfast foods, and sugar- or artificially sweetened beverages, were associated with meaningfully elevated mortality risk. That study found a 4% higher all-cause mortality risk for the highest quartile of total ultra-processed food consumption, and an 8% higher risk of neurodegenerative mortality — but it also acknowledged that ultra-processed foods is a diverse category, with many items that researchers consider healthy.
A June 2026 study from the Friedman School of Nutrition Science and Policy at Tufts University, published in the American Journal of Public Health and analyzing nationally representative dietary data from nearly 48,000 U.S. adults across two decades, similarly found that higher ultra-processed food intake was associated with worse cardiometabolic markers and a 4% higher risk of all-cause mortality. Critically, those associations persisted even after adjusting for overall diet quality using the Food Compass nutritional scoring system — a finding that suggests processing may carry some independent health signal beyond what nutritional profiling captures. A parallel AJPH analysis by researchers at Harvard and the University of California San Francisco found that diet quality was “more consistently associated with cardiometabolic health than UPF intake” — lending direct corroboration to the Wang finding — but also confirmed that the combination of high ultra-processed food intake and low diet quality produced the most adverse metabolic profiles.
What emerges from these studies, taken together, is that “ultra-processed food is bad” is a significant oversimplification, and “only nutritional quality matters” may be an oversimplification in the other direction. The appetite-regulation mechanism that Hall’s feeding trial documented appears real, even if it does not show up cleanly in long-term observational data. The heterogeneity within the ultra-processed category — the gap between a can of soda and a can of chickpeas, both classified as Group 4 — is a genuine methodological problem for dietary guidance built on NOVA alone.
What to Actually Read on the Label
Wang offered two practical takeaways from the research. For individuals, the guidance is a reframe: rather than scanning packages for signs of industrial processing, focus on whether a food is nutrient-dense — high in fiber, vitamins, minerals, and beneficial plant compounds, and low in added sugar, sodium, and saturated fat. A bowl of fortified oat cereal with plant-based milk passes that test regardless of its NOVA classification. A glass of fruit juice or a bowl of white rice does not, regardless of how minimally processed they are.
“Focus on what you’re eating overall, not just whether it’s processed,” Wang said. “A healthy plant-based diet rich in whole grains, fruits, vegetables, nuts, legumes and healthy plant oils is beneficial, even if it includes some processed foods.”
For policymakers and dietary guideline bodies, Wang suggested that frameworks built around the Mediterranean diet or high-quality plant-based diets — which specify what to eat rather than how it was made — are better positioned to translate into improved health outcomes than blanket guidance to avoid ultra-processed foods. She also noted that pushing food manufacturers to improve the nutritional content of packaged foods could help close the access gap, making healthy eating more practical for people who rely on convenience foods.
That last point carries real weight for the roughly 60% of daily calories in American diets that already come from ultra-processed foods. A policy that effectively tells most Americans their diets are wrong without offering a realistic path to a different diet is not the same as a policy that guides manufacturers to make the convenience foods people already buy more nutritious
Significant Caveats Apply
The study was presented as a conference abstract at NUTRITION 2026 and has not yet been published in a peer-reviewed journal, meaning the full methodology and statistical details have not gone through formal external review. Conference abstracts should be treated as preliminary findings. The analysis also focused on plant-derived components of participants’ diets to enable consistent cross-group comparisons — a methodological choice that simplifies the picture and may limit generalizability to the full diversity of American dietary patterns. And as all cohort studies, it establishes associations, not causation: the researchers cannot rule out residual confounding from variables they did not measure.
The 2025–2030 Dietary Guidelines for Americans, released in January 2026 under significant scientific controversy, were already criticized by Harvard researchers for contradicting the guidelines’ own retained 10% saturated fat limit. That controversy was a central topic at NUTRITION 2026. How this new evidence factors into the next revision — whether the guidelines move toward nutrient-based language and away from processing-level language — is a policy question that NUTRITION 2026 put squarely on the table.
Why This Matters Beyond the Label
The study’s broadest implication is a call to shift the terms of nutritional conversation: away from process labels and toward what the evidence consistently shows matters — the overall nutritional profile of what people eat
The “ultra-processed” framing drew public attention to a real problem in modern diets, and that attention was warranted: diets dominated by fast food, processed meat, and sugary beverages genuinely predict worse health outcomes, and Hall’s feeding trial established that the food manufacturing process itself may affect appetite regulation in ways nutritional labels do not fully capture. But if that framing is now guiding people to distrust fortified cereals and canned beans while leaving fruit juice and white rice unchallenged — if it is making diets worse in the name of making them healthier — then the label has overshot its usefulness.
What the evidence collectively supports is a more precise target: specific ultra-processed food categories (sugar-sweetened beverages, processed meats, sugary snacks) show consistent harm signals driven by both their nutritional profiles and, possibly, their processing-related effects on appetite. Nutritious ultra-processed foods (fortified whole grains, canned legumes, plant-based dairy alternatives) show no such signal. The two groups should not share a category that treats them as equivalents — and four decades of data from 200,000 Americans suggests that, for chronic disease risk, they effectively do not.
Frequently Asked Questions
Is “ultra-processed” a reliable guide to whether food is healthy?
According to the Wang study presented at NUTRITION 2026 — the largest analysis to directly separate diet quality from processing level in a single framework — processing level alone does not reliably predict health outcomes. A healthy diet is protective whether it includes some ultra-processed foods or not; an unhealthy diet is risky whether it is built from fast food or from minimally processed white rice and fruit juice. The NOVA classification system explicitly does not account for nutritional content, which creates situations where nutritious fortified cereals and canned legumes (ultra-processed) appear more dangerous than nutritionally depleted whole foods. When in doubt about a specific food, check added sugar, sodium, fiber, and protein content rather than NOVA tier, per the Wang study’s findings.
What specific ultra-processed foods have the strongest documented harm signals?
Not all ultra-processed foods carry equal risk. The most consistent harm signals in the published literature — including a 2024 Harvard analysis in The BMJ and a 2025 critical review in the Proceedings of the Nutrition Society — are concentrated in sugar-sweetened beverages, processed meats, sugary breakfast foods, and dairy-based desserts. Fortified cereals and certain dairy products have shown neutral or even protective associations in the same studies. The Wang research suggests these should not be treated as a single category for guidance purposes.
Why does the NOVA food classification system put fortified cereal in the same category as soda?
NOVA was designed to classify foods by the extent and purpose of industrial processing, not by their nutritional content. Its creator, Carlos Monteiro, argues that the industrial manufacturing process — not nutrients alone — explains why ultra-processed diets promote overeating, citing a 2019 NIH feeding trial in which matched nutritional profiles did not eliminate the caloric excess of an ultra-processed diet. Critics, including the 2025 Louie review and now the Wang analysis, argue that this process-based grouping creates guidance failures by treating nutritious and harmful foods as equivalent. An ongoing international project to develop a next-generation classification that incorporates nutritional content is running through December 2026, though it has drawn criticism from NOVA’s creator for being funded by the Novo Nordisk Foundation — the pharmaceutical company that makes Ozempic.
What should I look at on a food label instead of checking whether a food is ultra-processed?
Based on the Wang study’s practical guidance: focus on nutrient density. Check added sugar (aim low), sodium (aim low), dietary fiber (aim high), and protein. Look for whole grains as a first ingredient in grain products. Fortified foods that contribute meaningful amounts of calcium, vitamins, and iron can be part of a healthy diet regardless of their NOVA classification. Foods to prioritize reducing regardless of processing level: sugar-sweetened beverages, processed meats, refined-grain products with high added sugar — all of which show consistent harm signals whether assessed by NOVA or by nutritional profiling, according to the Wang study.
ⓒ 2026 TECHTIMES.com All rights reserved. Do not reproduce without permission
Tags:Heart Disease
