Ultra-Processed Food: What the NOVA Classification Actually Tells Us

The term “ultra-processed food” has become one of the most cited concepts in nutritional epidemiology over the past decade. It appears in government dietary guidelines, academic journals, newspaper headlines, and social media debates with roughly equal frequency. The classification system behind it — NOVA, developed by Carlos Monteiro and colleagues at the University of Sao Paulo — has fundamentally changed how researchers categorize food. What it has not done is resolve the question of why some processed foods appear to harm health while others do not.

The Four NOVA Groups

NOVA divides all foods and beverages into four categories based on the nature, extent, and purpose of processing they undergo.

Group 1: Unprocessed or minimally processed foods. Fresh fruits, vegetables, grains, legumes, nuts, eggs, meat, milk. Minimal processing includes removal of inedible parts, drying, pasteurization, freezing, and fermentation without added substances. The food remains recognizable as its original biological source.

Group 2: Processed culinary ingredients. Substances extracted from Group 1 foods and used in cooking: oils, butter, sugar, salt, flour, starch. These are rarely consumed on their own. They exist to prepare and season Group 1 foods.

Group 3: Processed foods. Group 1 foods modified by Group 2 ingredients using methods like canning, bottling, and non-alcoholic fermentation. Canned vegetables in brine, artisanal bread made from flour-water-salt-yeast, cheese, cured meats. These foods are recognizable versions of the original food with a limited number of added ingredients.

Group 4: Ultra-processed foods. Industrial formulations typically containing five or more ingredients, including substances not commonly used in domestic kitchens — high-fructose corn syrup, hydrogenated oils, modified starches, protein isolates, emulsifiers, humectants, flavor enhancers, and colorants. The group includes soft drinks, packaged snacks, reconstituted meat products, instant noodles, most commercial breads, sweetened cereals, and flavored yogurts.

The Epidemiological Signal

The epidemiological evidence linking ultra-processed food consumption to adverse health outcomes is substantial and growing. Large prospective cohort studies — including the French NutriNet-Sante study, the Spanish SUN cohort, and the UK Biobank — have consistently associated higher ultra-processed food intake with increased risk of obesity, type 2 diabetes, cardiovascular disease, depression, and certain cancers, even after adjusting for total energy intake, body mass index, and socioeconomic factors.

A 2024 umbrella review published in The Lancet synthesized the evidence from multiple systematic reviews and meta-analyses, concluding that higher consumption of ultra-processed foods was associated with adverse health outcomes across nearly every category examined. The magnitude of the associations was not trivial: the most-exposed groups showed 20-50% higher risk for several chronic conditions compared to the least-exposed groups.

These findings have driven policy action. Brazil, France, Israel, and several other countries have incorporated NOVA-based guidance into their national dietary recommendations. The classification has moved from academic curiosity to regulatory relevance in under a decade.

The Classification Problem

The epidemiological signal is strong. The mechanistic explanation is less clear, and this is where NOVA’s utility as a scientific tool comes under scrutiny.

The core issue is that NOVA groups foods by how they are made, not by what they contain nutritionally. This produces some classifications that are difficult to defend from a health perspective. Commercially produced whole-grain bread that contains an emulsifier falls into Group 4 alongside candy bars and sugar-sweetened beverages. A flavored yogurt with a small amount of added pectin is ultra-processed; the same yogurt without pectin is merely processed. Industrial baby formula — engineered to approximate the nutritional profile of breast milk as closely as current science allows — sits in Group 4 alongside energy drinks.

Critics, including Michael Gibney of University College Dublin, have argued that NOVA conflates genuinely harmful products with nutritionally adequate or even beneficial ones under a single label. The classification captures something real about the food supply — the proliferation of industrially formulated, hyper-palatable, nutrient-poor products — but its boundaries are drawn by processing method rather than nutritional outcome, which creates categorical noise.

What Is Actually Driving the Health Effects?

If ultra-processed food consumption correlates with poor health outcomes, several mechanisms could explain the association, and they are not mutually exclusive.

Energy density and palatability engineering. Ultra-processed foods are designed to be consumed quickly and in large quantities. They tend to be energy-dense, low in fiber, and optimized for combinations of sugar, fat, and salt that override satiety signaling. A landmark 2019 randomized controlled trial by Kevin Hall and colleagues at the NIH found that participants offered an ultra-processed diet consumed approximately 500 more calories per day than those offered an unprocessed diet, and gained weight accordingly — despite both diets being matched for available calories, macronutrients, sugar, sodium, and fiber. The ultra-processed foods were simply eaten faster and in greater quantity.

Displacement of whole foods. Every calorie consumed from ultra-processed sources is a calorie not consumed from minimally processed foods. In populations where ultra-processed foods constitute 50-60% of total energy intake — as they do in the United States, United Kingdom, Canada, and Australia — the displacement effect is enormous. Diets high in ultra-processed foods are consistently lower in fiber, potassium, magnesium, and various phytonutrients, not because ultra-processed foods are toxic, but because they crowd out the foods that provide these nutrients.

Additives and industrial substances. Some food additives commonly found in ultra-processed products — certain emulsifiers, artificial sweeteners, and nanoparticle-based colorants — have shown adverse effects on the gut microbiome in animal models and small human studies. Carboxymethylcellulose and polysorbate 80, for example, have been shown to erode the intestinal mucus layer and promote inflammatory changes in mouse models. Whether these effects translate to clinically meaningful outcomes in humans at typical dietary exposures remains an active area of research.

Food matrix disruption. Processing that disrupts the physical structure of food — grinding, extruding, hydrolyzing — alters how nutrients are released and absorbed during digestion. The same amount of starch delivered in a whole grain versus a puffed cereal produces different glycemic responses, different insulin responses, and different satiety profiles. NOVA captures this dimension imperfectly, but it captures it.

A More Useful Framework

NOVA has been extraordinarily valuable as a research tool. It gave epidemiologists a vocabulary and a classification system that revealed patterns invisible under traditional nutrient-based analysis. The consistent association between ultra-processed food consumption and disease risk is not an artifact of poor study design — it has been replicated across diverse populations, study methods, and dietary assessment tools.

Where NOVA falls short is as a consumer-facing guide. Telling someone to “avoid ultra-processed food” without context leads to confusion when they discover that their whole-wheat sandwich bread, protein bar, and soy milk all qualify. The actionable insight is not that industrial processing is inherently dangerous, but that a dietary pattern dominated by formulated, hyper-palatable, nutrient-poor products is associated with overconsumption and poor health outcomes — and that shifting toward whole, minimally processed foods consistently improves both diet quality and metabolic markers.

The distinction between these two framings — “processing is bad” versus “your overall dietary pattern matters” — may sound subtle. In practice, it determines whether NOVA becomes a useful heuristic or a source of unnecessary anxiety about ingredients lists.

Michael Torres is the Food Science Editor at Daily Bite Lab. He holds a PhD in Food Systems and Public Health from the University of Michigan.