What Is Ultra-Processed Food — and Does It Really Matter?

The framework is called NOVA, developed by Carlos Monteiro’s group at the University of São Paulo. It sorts food not by nutrients but by what has been done to it industrially:¹

• Group 1 — unprocessed or minimally processed: vegetables, meat, milk, eggs, oats, frozen berries, plain yogurt.
• Group 2 — culinary ingredients: oil, butter, sugar, salt.
• Group 3 — processed foods: group 1 + group 2, simply combined: cheese, canned fish, sourdough bread, salted nuts.
• Group 4 — ultra-processed: industrial formulations built largely from substances extracted or synthesized from food — protein isolates, modified starches, glucose syrup, hydrogenated fats — plus additives that make the result look, taste and feel like food: emulsifiers, colorants, flavorings, sweeteners.

The unifying idea: group 4 products are engineered products designed for convenience, shelf life and maximal appeal, rather than foods in the traditional sense.

The practical shortcut is disarmingly simple: read the ingredient list and ask whether you could make this in a home kitchen. Flour, water, salt, yeast — that is bread, group 3, even from a factory. Add emulsifiers, dough conditioners and a sweetener to keep it soft for three weeks, and it has crossed into group 4. If the list contains things you cannot buy as a home cook — maltodextrin, soy protein isolate, mono- and diglycerides — you are holding an ultra-processed product. No single additive is the villain; the pattern is the marker.

Yes — and the signal is unusually consistent. In the French NutriNet-Santé cohort of over 100,000 adults, every 10-percentage-point increase in the ultra-processed share of the diet came with a measurably higher rate of cardiovascular disease over the following years.³ In 2024, an umbrella review in The BMJ pooled the meta-analyses across 45 outcomes and graded the evidence: the associations rated convincing or highly suggestive included cardiovascular-related death, type 2 diabetes, anxiety and common mental disorders.²

Two honest caveats belong next to that. These are observational data — they show that heavy ultra-processed eaters fare worse, not by itself why. And the risk increases are meaningful but not dramatic for any single person: this is population-level arithmetic, where a daily habit multiplied across millions of people becomes a public-health problem.

The study that moved ultra-processed food from correlation toward causation was run by Kevin Hall at the US National Institutes of Health in 2019. Twenty adults lived at the clinic for a month and ate two weeks of each diet: one ultra-processed, one minimally processed — matched for calories, sugar, fat, fiber and salt in what was offered, with participants free to eat as much or little as they liked. On the ultra-processed diet they spontaneously ate about 500 kcal more per day and gained roughly a kilo in two weeks; on the unprocessed diet the same people lost weight.⁴

Because the menus were nutritionally matched, the usual suspects — sugar, fat, salt — cannot explain the gap. Something about the products themselves (energy density, softness and eating speed, hyper-palatable combinations) drives passive overeating. It is a small, short study, but it is the cleanest causal evidence in the field.

Here is the sharpest question you can ask about this research: a person who eats a processed-meat sandwich every day is eating salami and white factory bread. If they get sicker over the years, how does anyone know which one did it? Epidemiologists attack this from three directions:

• Statistical adjustment: the cohort studies control for total energy, sugar, salt, fat, fiber and overall diet quality. The ultra-processed association survives — so it is not merely a marker of a nutritionally poor diet.²³
• Category-level analyses: when a large European study followed each ultra-processed sub-category separately, the risk was driven mainly by animal-based products (processed meats) and sweetened beverages — while ultra-processed breads and cereals showed no such risk, even trending mildly the other way.⁵ The bread on that sandwich is, on current evidence, largely acquitted; the salami is not.
• Independent mechanisms: for processed meat specifically, the case does not rest on questionnaires. Nitrite curing produces nitrosamines, and the IARC working group judged the colorectal-cancer evidence strong enough to classify processed meat as a group 1 carcinogen.⁶

Full disentanglement is impossible in observational data — daily salami eaters differ from non-eaters in a thousand ways. But when adjustment, sub-category analysis and mechanism all point at the same foods, the verdict stops being a coin flip.

The strongest objection to NOVA is that group 4 is enormous. A whey protein powder, a fortified plant milk, a supermarket wholegrain loaf and a bag of cheese puffs all land in the same box, yet their health effects are clearly not equivalent — the category-level data above proves exactly that. “Ultra-processed” is best read as a warning label on a shelf, not a verdict on every product standing on it. That is also why our companion piece ranks the specific products with the worst evidence instead of condemning the whole group.

• Count the share, not the sins. The risk data scales with the proportion of your diet that is ultra-processed. An occasional ready meal in a diet built on real food moves nothing.
• Prioritize the proven offenders: daily processed meat and daily sweetened drinks are where the evidence is concrete. Start there, not with the protein powder.
• Use the kitchen test when shopping — short ingredient lists of recognizable things beat any front-of-pack health claim.
• Cook simple. The realistic goal is shifting the ratio toward group 1 and 3 — which is precisely what our recipes are built around.