The Science of Ultra-Processed Food: How Food Is Made Matters More Than What It Contains

Nutrition science has long evaluated food by breaking it into components: grams of fat, grams of sugar, milligrams of sodium. Since the 2010s, however, findings have accumulated that this method cannot explain. Prepare two diets matched for calories, macronutrients, sugar, sodium and fiber, let people eat as much as they want, and one of them leads to more than 500 extra calories a day. The difference lies not in composition but in how the food was made.

NOVA: Classifying Food by Process, Not Nutrients

This perspective originated with the NOVA classification proposed by Professor Carlos Monteiro and colleagues at the University of Sao Paulo, formalized in Public Health Nutrition in 2019. It sorts food into four groups according to the nature and purpose of processing rather than nutrient content.

Group 1 covers unprocessed and minimally processed foods (vegetables, meat, fish, rice, eggs, milk). Group 2 covers processed culinary ingredients (oil, sugar, salt). Group 3 covers processed foods (canned goods, cheese, bread). Group 4 is ultra-processed food (UPF): ready-to-consume industrial formulations built from substances not found in a domestic kitchen -- high-fructose corn syrup, hydrogenated oils, protein isolates, emulsifiers, colorings, flavorings, thickeners. Soft drinks, sweet pastries, snacks, instant noodles, processed meats, and many seasoned frozen meals fall here.

The premise of NOVA is not that these products contain harmful ingredients, but that being manufactured this way may itself constitute an independent risk factor.

The Decisive Experiment: Inside an NIH Metabolic Ward

The hypothesis was tested in 2019 by Dr. Kevin Hall and colleagues at the US National Institutes of Health. Twenty adults were admitted to a metabolic ward and given an ultra-processed diet and a minimally processed diet for two weeks each. The two diets were matched for total calories presented, energy density, macronutrients, sugar, sodium and fiber. Participants ate as much as they wanted.

The result was unambiguous. On the ultra-processed diet, participants consumed an average of 508 additional calories per day and gained about 0.9 kilograms over two weeks. On the minimally processed diet, they lost about 0.9 kilograms. The same individuals, the same nutrient profile -- only processing differed.

In nutrition science, a randomized crossover trial conducted in a controlled inpatient setting with matched nutrients is among the strongest available forms of evidence. This single study elevated "degree of processing" from an observational correlation to a candidate cause.

What 9.9 Million People Show

The epidemiological evidence has grown in parallel. An umbrella review published in the BMJ in 2024 synthesized 45 meta-analyses drawn from 14 review articles, covering roughly 9.9 million participants. Associations between ultra-processed food and cardiovascular-related mortality, type 2 diabetes, and common mental disorders such as anxiety and depression were graded as "convincing" -- the highest evidence class applied.

Individual cohorts point the same direction. In France's NutriNet-Sante study, each 10-percentage-point increase in the share of ultra-processed food in the diet was associated with a 12% higher risk of cardiovascular disease (BMJ, 2019), a 14% higher risk of all-cause mortality (JAMA Internal Medicine, 2019), and a 12% higher risk of overall cancer (BMJ, 2018).

Why People Overeat It

Several mechanisms appear to operate together.

First, eating rate. Heat and mechanical processing destroy the food matrix -- the physical structure of food -- leaving products that require almost no chewing. In the NIH study, participants consumed significantly more calories per minute on the ultra-processed diet. Satiety signals take 15 to 20 minutes to reach the brain, and food that can be eaten quickly exploits that lag.

Second, energy density and nutrient dilution. Removing water and fiber while adding fats and refined carbohydrates raises the calories contained in a given volume. Stretch receptors in the stomach respond to weight and volume, so fullness registers later.

Third, additives and the gut. A 2015 Nature paper showed that administering the common emulsifiers carboxymethylcellulose and polysorbate-80 to mice altered the composition of the gut microbiota, thinned the intestinal mucus layer, and induced low-grade inflammation and metabolic dysfunction. Human replication is still under way, but the assumption that additives are metabolically inert no longer holds securely.

Japan's Position: The World's Most Favorable Starting Point

Internationally, Japan occupies a notably advantageous position. In the United States, roughly 58% of total energy intake comes from ultra-processed foods (BMJ Open, 2016). Analyses of Japanese household food purchases put the figure at around 28% (Public Health Nutrition, 2019). A meal structure of rice, miso soup, grilled fish and pickles resolves almost entirely into NOVA Groups 1 and 2.

That advantage is not fixed. Growth in single-person households, expansion of the prepared-meal market, and the ubiquity of convenience stores are all pushing the processing level of the Japanese diet upward. The long-term stagnation in vegetable intake documented by the National Health and Nutrition Survey belongs to the same trend. Starting from the world's lowest baseline is precisely why the direction of travel matters.

A Caution Against Oversimplification

Researchers have also warned against reading the evidence as "ultra-processed equals harmful." In 2023, an analysis of the large European EPIC cohort (approximately 267,000 participants) found that while ultra-processed food overall was associated with multimorbidity of cancer and cardiometabolic disease, the strength of association varied substantially by subgroup. Sugar-sweetened and artificially sweetened beverages and processed meats showed clear risk elevations, whereas certain breads and breakfast cereals showed no significant association.

Because NOVA classifies by process, it groups together products with radically different nutrient profiles. Whether unsweetened wholegrain bread belongs in the same category as carbonated soft drinks remains an active academic dispute. In practice, the sensible response is not to choose between "processing" and "nutrient profile" as metrics, but to use both.

What Changes: How to Read a Label

The most immediate practical consequence concerns food labels. Conventional advice directs attention to the nutrition panel; the processing lens moves it to the ingredient list. The more entries that name substances absent from a home kitchen -- syrups, modified oils, emulsifiers, acidity regulators, flavorings, antioxidants -- the closer the product sits to Group 4.

The realistic goal is not zero ultra-processed food. Complete elimination is unachievable in most working lives, and unachievable targets do not produce behavior change. What works better is substitution at the points of highest frequency: water or tea instead of sweetened drinks, rice and eggs instead of a sweet pastry at breakfast, nuts or yogurt instead of packaged snacks. The more often a meal repeats, the more the substitution compounds.

Moving from "what to eat" to "how what you eat was made" adds one dimension to nutrition -- but for consumers it simplifies the decision rather than complicating it. No calculation is required. For now, the most practical indicator is whether the ingredient list is short and written in words you recognize.

Back to STORIES