Cutting down on ultra-processed foods

What counts as ultra-processed

The common definition comes from the NOVA classification of the group around Carlos Monteiro at the University of São Paulo. It sorts foods into four groups by the extent and purpose of industrial processing: unprocessed or minimally processed foods, processed culinary ingredients such as oil or sugar, processed foods and ultra-processed foods. The fourth group is created by breaking foods down into individual substances, chemically modifying them and recombining them, often with cosmetic additives and elaborate packaging.

The practical test that the NOVA authors themselves propose: does the ingredient list contain at least one ingredient that is rarely used in the kitchen, such as glucose-fructose syrup, hydrogenated or interesterified fats or hydrolyzed protein? Or an additive that shapes taste and appearance, that is, flavorings, flavor enhancers, colorings, emulsifiers, sweeteners, thickeners? Then the product counts as ultra-processed. On sweeteners and sugar substitutes, see also Erythritol and Xylitol.

How it might work

The obvious suspicion is that ultra-processed foods simply contain more sugar, fat and salt. The 2019 NIH study eliminated exactly that: the meals offered were matched for calories, energy density, macronutrients, sugar, sodium and fiber, and yet participants ate markedly more on the ultra-processed diet, mainly carbohydrates and fat, but not protein.

Several pathways are discussed as explanations. The 2025 Lancet series names high energy density, very high palatability, soft texture and a destroyed food structure as drivers of overeating, plus fewer protective plant compounds and more contaminants and additive mixtures. In a meta-analysis of ten randomized trials, eating rate was the clearest difference: ultra-processed food was eaten faster. And the extra intake depended on whether the ultra-processed diet was more energy-dense. When both were equally energy-dense, it largely disappeared.

What the studies show

NIH metabolic study, Cell Metabolism 2019

20 weight-stable adults lived on the research ward of the US health agency NIH and received, in random order, two weeks each of an ultra-processed and an unprocessed diet, as much or as little as they wanted. On the ultra-processed diet, they ate 508 kilocalories more per day, 280 of them from carbohydrates and 230 from fat. They gained 0.9 kilos, and lost 0.9 kilos on the unprocessed diet. Weight closely followed energy intake.

UPDATE study, Nature Medicine 2025

55 adults in England with overweight who had previously obtained at least half of their energy from ultra-processed foods received eight weeks each of minimally processed and ultra-processed food deliveries, both put together according to the official British dietary guidelines; 50 were included in the analysis. Both diets lowered weight, the minimally processed one by 2.1 percent, the ultra-processed one by 1.1 percent. Fat mass, triglycerides and cravings fell more with the minimally processed diet, whereas LDL cholesterol fell more with the ultra-processed one. The study was funded by a medical foundation and a non-governmental organization, without industry involvement.

Feeding study in men, Cell Metabolism 2025

Men ate ultra-processed and unprocessed diets in a crossover design, each with a normal and with an increased amount of calories. On the ultra-processed diet, body weight and the ratio of LDL to HDL rose, regardless of the amount of calories. Follicle-stimulating hormone fell, sperm motility tended to decrease, and in the blood there was a trend toward more of a phthalate breakdown product.

Meta-analysis of randomized trials, 2026

Ten randomized trials, ultra-processed versus less processed. Weight gain and eating rate were significantly higher on the ultra-processed diet, energy intake tended to be higher, but depending on the statistical model not always significantly. When both diets had the same energy density, practically no difference in energy intake was seen any longer. The authors rate the certainty of the evidence as low and point out that the diets in the previous studies were not matched for nutrients.

Umbrella review, BMJ 2024

45 pooled analyses from observational studies with a combined almost 9.9 million people. The authors rated as convincing the associations with cardiovascular death, with a 50 percent higher risk, with type 2 diabetes, with a 12 percent higher risk in the dose-response analysis, and with anxiety disorders and common mental disorders. Highly suggestive was a 21 percent higher risk of death overall. According to GRADE, however, only 4 of the 45 analyses were of moderate quality, 22 of low and 19 of very low quality.

Lancet series, 2025

The first paper of a three-part Lancet series summarizes more than 100 prospective studies, meta-analyses, randomized trials and mechanistic studies and concludes that the displacement of traditional dietary patterns by ultra-processed foods is a major driver of diet-related chronic diseases. The first author is Carlos Monteiro, who also wrote the 2019 NOVA guide.

How consistently NOVA is applied, European Journal of Clinical Nutrition 2022

French nutrition experts were asked to assign 120 branded products with ingredient lists and 111 generic foods to the NOVA groups. Agreement between the raters was low, with a kappa of 0.32 and 0.34 on a scale up to 1, even when the ingredient list was available. For some of the products, the assignment was unambiguous; for others, it was spread across all four groups.

Where the data stop

For heart attack, cancer or mortality, there are only observational data, and according to GRADE these are predominantly of low or very low quality. People who eat a lot of ultra-processed food also differ in many other ways; this can only partly be adjusted for statistically. The randomized trials are small, run for two to eight weeks and measure weight and risk factors, not diseases.

What exactly has the effect is also open. Part of the overeating can be explained by energy density and eating rate, and in the 2026 meta-analysis it largely disappeared when both diets were equally energy-dense. Whether processing itself is harmful or is mainly a feature that coincides with unfavorable nutrients has not been settled. Added to this is the blurred classification: if experts classify the same products alike only with a kappa of 0.32, the measurement of intake in large studies is also uncertain. And UPDATE shows that an ultra-processed diet following dietary recommendations does not automatically perform badly: it also lowered weight and LDL cholesterol, only the weight less strongly.

How to do it

Read the ingredient list. According to the NOVA guide, a single typical ingredient is enough to classify a product as ultra-processed: glucose-fructose syrup, hydrogenated fats, hydrolyzed protein or additives such as flavorings, flavor enhancers, colorings, emulsifiers, sweeteners and thickeners. A short list of ingredients that you also have in your kitchen is the opposite.

Start with what you eat most often and replace it with dishes made from basic ingredients. In UPDATE, the advantage of minimally processed food was about one percent of body weight in eight weeks, in people who had previously obtained at least half of their energy from ultra-processed foods. The lever is therefore greatest where the share is high.

Eat more slowly. Eating rate was the clearest difference between the diets in the randomized trials. And do not take the classification too precisely: if even experts disagree on whether a product belongs in group 3 or 4, arguing about it is not worth it. What matters is the overall share of fresh and minimally processed foods. Anyone also interested in carbohydrates will find the data on this under ketogenic diet.

Safety

Eating fewer ultra-processed foods usually means more food made from basic ingredients; no serious risks have been described for this. In UPDATE, mild gastrointestinal complaints were common under both diets. Anyone who sorts foods only by degree of processing may, however, discard products that perform well by other standards: in UPDATE, LDL cholesterol fell more under the ultra-processed diet following dietary recommendations than under the minimally processed one.

Sources

• NOVA guide “Ultra-processed foods: what they are and how to identify them” — Monteiro et al., Public Health Nutrition 2019
• NIH metabolic study — Hall et al., Cell Metabolism 2019, 20 adults, 2 weeks each
• Umbrella review on ultra-processed foods — Lane et al., BMJ 2024, 45 pooled analyses, 9.9 million participants
• UPDATE study — Dicken et al., Nature Medicine 2025, 55 adults, 8 weeks each
• Feeding study on metabolism and fertility in men — Preston et al., Cell Metabolism 2025
• Meta-analysis of randomized trials on energy intake and weight — Robinson et al., Critical Reviews in Food Science and Nutrition 2026
• Lancet series, part 1 — Monteiro et al., The Lancet 2025
• Agreement in NOVA classification — Braesco et al., European Journal of Clinical Nutrition 2022