Ultra-Processed Foods and Health: The NOVA System Explained

NOVA divides all foods into four groups based on the degree of processing they undergo.

**Group 1 — Unprocessed or minimally processed foods:** These are natural foods that have undergone minimal processing for preservation or basic preparation — peeling, cutting, drying, freezing, pasteurising, fermentation. Examples include fresh fruit and vegetables, dried legumes, plain yoghurt, unsalted nuts, eggs, plain meat and fish, and plain milk. These foods retain their original nutritional structure.

**Group 2 — Processed culinary ingredients:** Substances derived from Group 1 foods or from nature, used in cooking to prepare Group 1 foods. Examples include pressed oils, butter, flours, starches, sugar, salt, and vinegar. These are not typically eaten alone but are used to make meals.

**Group 3 — Processed foods:** Products made by adding Group 2 ingredients to Group 1 foods, primarily to extend shelf life or improve flavour. Examples include canned fish in olive oil, salted nuts, cured meats like prosciutto or pancetta, hard cheeses, freshly baked bread made with only flour, salt, water, and yeast. Processing is recognisable and ingredients are few.

**Group 4 — Ultra-processed foods:** Industrial formulations made primarily from refined substances derived from foods — isolated protein, modified starches, interesterified fats — combined with cosmetic additives including emulsifiers, colours, flavour enhancers, artificial sweeteners, and glazing agents. The defining characteristic is that ultra-processed foods contain ingredients or combinations of ingredients that you would not find in a domestic kitchen. Examples include carbonated soft drinks, packaged snacks and crisps, mass-produced bread with more than 5–6 ingredients, most breakfast cereals, instant noodles, reconstituted meat products, and most commercially produced desserts.

The mechanisms by which ultra-processed foods drive adverse health outcomes are an active research question, with two dominant — and not mutually exclusive — hypotheses.

**The Additives Hypothesis** proposes that specific chemical additives in ultra-processed foods cause harm independently of nutritional composition. The most studied examples are emulsifiers. A landmark 2015 study in Nature (Chassaing et al.) showed that dietary emulsifiers carboxymethylcellulose and polysorbate-80, at concentrations approximating typical human dietary exposure, disrupted the mucus layer protecting the intestinal epithelium in mice, increased intestinal permeability, caused low-grade inflammation, and promoted metabolic syndrome features. Subsequent observational data in humans have found associations between emulsifier intake and inflammatory bowel disease and type 2 diabetes. Artificial sweeteners have independently been associated with gut microbiome disruption and altered glucose regulation in randomised trials, though the evidence is less consistent.

**The Matrix Hypothesis** argues that the problem is not specific additives but the destruction of the food matrix itself — the complex physical and chemical structure of whole foods that governs how nutrients are digested, absorbed, and interact with gut physiology. Ultra-processing systematically breaks down cell walls, denatures proteins, disrupts fat-water interfaces, and reduces particle size. This alters transit time, satiety signalling, glycaemic response, and the physical availability of nutrients to gut bacteria. A whole almond and almond butter made from only almonds are nutritionally similar on paper but behave very differently in the gut because the matrix is intact in the whole nut and disrupted in the butter.

The NutriNet-Santé cohort is the largest and most frequently cited source of evidence on ultra-processed food consumption and health outcomes. This French prospective cohort has followed over 100,000 adults since 2009, with detailed dietary assessments via 24-hour recalls. A series of publications from this cohort have found that each 10 % increase in the proportion of ultra-processed foods in the diet is associated with a 12 % higher risk of overall cancer, a 14 % higher risk of cardiovascular disease, a 10 % higher risk of type 2 diabetes, and elevated all-cause mortality.

Crucially, these associations hold after adjustment for standard nutrient-based metrics — total energy intake, saturated fat, fibre, sodium, and sugar. This suggests that processing degree captures health-relevant information that nutrient composition does not. The Spanish SUN cohort (Rico-Campà et al., BMJ, 2019) found that consuming 4+ servings of ultra-processed foods daily was associated with a 62 % higher all-cause mortality risk compared to fewer than 2 servings daily, again after comprehensive nutritional adjustment.

The most compelling experimental evidence comes from a 2019 NIH randomised controlled trial by Kevin Hall and colleagues, published in Cell Metabolism. Twenty adults were admitted to an inpatient unit for four weeks and randomised to either an ultra-processed or unprocessed diet matched for total energy, sugar, fat, fibre, and macronutrients. Those on the ultra-processed diet consumed an average of 508 kcal more per day and gained weight; those on the unprocessed diet lost weight. This is one of the few RCTs in nutritional science that directly tests processing degree — and its results are striking.

Practical application of NOVA requires answering some genuinely difficult questions. Is Greek yoghurt ultra-processed? No — plain Greek yoghurt is Group 1. Flavoured Greek yoghurt with added colours and thickeners is probably Group 4. Is wholegrain bread ultra-processed? A sourdough made from flour, water, and salt is Group 3. A mass-produced 'wholegrain' loaf containing mono- and diglycerides, dough conditioners, and added vitamins is Group 4 — even if the front-of-pack label says 'high in fibre.' Is smoked salmon ultra-processed? No — salting and smoking are ancient preservation techniques placing it in Group 3 at most.

Some common foods are surprisingly classified as ultra-processed: infant formula (contains industrial food-grade emulsifiers), most plant-based meat alternatives (contain methylcellulose, soy protein isolate, modified starches), many 'healthy' snack bars (contain concentrated fruit paste, added vitamins, flavourings), and mass-produced oat drinks (often contain added oil, emulsifiers, and gums). This does not mean these foods are uniquely toxic — it means they fall within the NOVA Group 4 definition and should be contextualised alongside the epidemiological evidence.

Moving down the NOVA scale does not require radical dietary overhaul. Most high-impact swaps involve replacing frequently consumed ultra-processed items with minimally processed equivalents.

**Mass-produced bread → sourdough or home-baked bread:** A true sourdough contains flour, water, salt, and a starter culture. Three ingredients, no emulsifiers, and the fermentation process reduces phytates and improves nutrient bioavailability. **Packaged breakfast cereal → rolled oats with fruit and nuts:** Plain oats are a Group 1 food. Topped with banana and walnuts, you have a breakfast with more fibre, less sugar, and no additives. **Flavoured yoghurt → plain yoghurt with fresh fruit:** Plain whole-milk yoghurt is Group 1. Add fresh berries and a small spoon of honey and the result is nutritionally superior to any fruit-flavoured yoghurt. **Commercial salad dressing → olive oil and lemon:** Most bottled dressings contain emulsifiers, modified starch, and flavour enhancers. A drizzle of extra virgin olive oil with lemon juice takes fifteen seconds and contains none of them. **Fruit juice drinks → whole fruit or water:** The combination of low-fibre, high-fructose liquid with emulsifiers and flavourings in commercial 'juice drinks' is a textbook Group 4 product. A whole orange is Group 1.