top of page

Ultra-Processed Food and the NOVA Classification: What the Research Shows

Writer: Kim Berglin
Kim Berglin
Oct 13
5 min read

Grab a package of bread off a grocery shelf and you've touched one of nutrition science's most contentious debates without realizing it. A four-tier sorting system called NOVA now shapes headlines and government dietary guidance around the world. It has also reshaped how ordinary shoppers judge what counts as real food. Researchers who study ultra-processing disagree sharply, though, on what the category measures. One of the strongest pieces of evidence comes from a 2019 government-funded trial that locked volunteers in a hospital ward for a month and tracked every calorie they ate. Enormous population reviews, pooling data from millions of people, point in a similar direction. A growing group of food scientists argues the whole framework rests on shakier ground than headlines suggest. Below is a look at what the research supports, and where it stops short.

A Classification Built on Processing, Not Nutrients

Brazilian epidemiologist Carlos Monteiro and colleagues at the University of São Paulo introduced NOVA in 2009, sorting every food into four groups by degree and purpose of industrial processing rather than by nutrient content. Group 1 covers unprocessed or minimally processed foods such as fruit, vegetables, milk, and eggs. Group 4, ultra-processed food, covers industrial formulations assembled from substances rarely found in a home kitchen, among them hydrogenated oils, protein isolates, flavor compounds, and emulsifiers. Soda and packaged snack cakes fall into this last group. So do most breakfast cereals, some wholegrain breads, flavored yogurts, and plant-based milk alternatives that look nutritionally reasonable on a label.

Kevin Hall's Controlled Feeding Experiment

NIH metabolism researcher Kevin Hall ran the most direct test yet in 2019, admitting 20 adults to the NIH Clinical Center for a full month and feeding each person two weeks of an ultra-processed diet and two weeks of a minimally processed diet, order randomized. Both diets matched for calories, sugar, fiber, fat, and carbohydrate content, and participants could eat as much or as little as they wanted. People eating the ultra-processed menu consumed roughly 500 more calories per day than on the minimally processed menu, ate faster, and gained about 0.9 kilograms over two weeks while losing a similar amount on the other diet. Hall called the result "the first study to demonstrate causality" and noted that the two diets "showed a clear and consistent difference."

Hall and his co-authors flagged an important caveat: a slight difference in protein content between the two menus could account for as much as half the calorie gap, and they called for follow-up trials using a reformulated ultra-processed diet to isolate which specific factor drives overeating. Hall pointed to a practical barrier distinct from biology as well. Preparing less-processed meals, he said, "takes more time and more money." Budget and time pressure, more than willpower, explain why many households default to convenience.

A Massive Review Raises the Stakes

Melissa Lane and colleagues published the biggest synthesis of this evidence in BMJ in February 2024, an umbrella review pooling 45 meta-analyses drawn from 14 systematic reviews and covering close to 10 million participants. Lane's team graded results as convincing, highly suggestive, suggestive, or weak, borrowing the standard grading scale used across epidemiology. Convincing evidence linked higher ultra-processed food intake to roughly 50% greater risk of dying from cardiovascular disease and a 48-53% greater risk of anxiety or other common mental disorders. Highly suggestive evidence pointed to a 21% higher risk of death from any cause among the heaviest consumers, plus elevated risk of obesity and sleep problems. Independent public and academic funding backed every one of the 14 reviews, the authors verified, addressing a common criticism of industry-funded nutrition research.

Experts Divide Over What the Review Proves

Reaction split immediately along methodological lines. Clare Collins, a laureate professor at the University of Newcastle, pointed to a limitation baked into the review's design: "The studies are observational, which means cause and effect cannot be proven." Every piece of evidence in the umbrella review traces back to food-frequency questionnaires and dietary recall surveys, methods suited to spotting association rather than mechanism. Alan Barclay, a consultant dietitian and honorary associate at the University of Sydney, pushed further, calling NOVA itself crude and arguing the review's findings "do not and cannot provide evidence of causality." Daisy Coyle, a research fellow at The George Institute for Global Health, read the same data as reason for immediate action, calling for advertising restrictions, warning labels, school bans, and subsidies for fresh food because, in her words, "existing nutrition policies in Australia aren't enough to tackle this problem."

Food Scientists Challenge NOVA's Scientific Basis

Not every critique comes from outside nutrition science. Thomas Henle, chair of food chemistry at Technische Universität Dresden, co-signed a 2025 open letter arguing that NOVA's fourth category lacks objective metrics and depends mainly on counting ingredients and additives rather than measuring anything about health. Counting ingredients and additives this way, the letter argues, lumps wholegrain bread and infant formula into the same bucket as sausages and sugary soda, obscuring real nutritional differences between products that share a processing label but little else. Henle's letter also traces NOVA's origin to a deliberate health-policy goal, a framing Monteiro and his University of São Paulo research group have long disputed, arguing instead that NOVA fills a gap nutrient-based classification leaves open.

Bottom Line

Controlled feeding data and enormous population reviews point in the same rough direction: diets heavy in ultra-processed food associate with worse outcomes, and Kevin Hall's locked-ward trial offers the closest thing nutrition science has to causal proof that these products drive overeating, at least in a two-week window. Scaling that finding up to lifetime chronic disease risk requires leaning on correlational data, though, and researchers like Clare Collins and Alan Barclay are right that food-frequency questionnaires cannot prove cause and effect on their own. NOVA's harshest critics raise a fair point too: a classification system built around ingredient counts and additive lists captures something real about industrial food, but it also catches nutritious products such as wholegrain bread and plant-based milk in the same net as soda. A useful middle read treats UPF status as a flag worth noticing rather than a verdict on any single food, and weighs overall diet pattern more heavily than processing category alone when predicting health outcomes.

Diet quality debates like this one rarely stay confined to nutrient counts and processing labels; how a person eats matters as much as what sits in the cart. Readers curious about that behavioral side can find more in our piece on mindful eating and the psychology of satiety, which covers why slowing down at a meal changes how satisfied a person feels regardless of a food's processing tier.

Sources

 
 
 

Comments


bottom of page