What an Additive Ban Does Not Tell You About Ultra-Processed Food

Tesco gave a difficult food debate a number shoppers could hold on to: 114. The neat number makes the rule feel easy to grasp. That was how many additives the supermarket said it had banned or restricted in its own-brand foods. The policy cuts selected additives from own-brand recipes. Alice Ritchie, Tesco's head of healthy and sustainable diets, said customers wanted to eat fewer processed foods but found the subject hard to navigate.¹

The rule changes real recipes. A colour may disappear. A sweetener may be restricted. A supplier may have to return to the formulation bench. Yet the finished food can still be ultra-processed under NOVA, the research system that looks at the whole formulation and the purpose of processing. The supermarket list also cannot tell whether a food is high in salt, low in fibre, useful for an allergy, fortified with a needed nutrient or affordable enough to eat regularly. The label on the pack still asks a wider question.

That gap matters because the controlled trial at the heart of this article did not test those 114 ingredients one at a time. The key contrast sat on the plate, not in one part. At the US National Institutes of Health, Kevin Hall and his colleagues brought 20 adults into a clinical ward for 28 days. Each person spent two weeks eating from an ultra-processed menu and two weeks eating from an unprocessed menu, in random order. They could eat as much or as little as they wanted.⁴

On the ultra-processed menu, the participants ate an average of 508 more calories each day and gained about 0.9 kilograms. That boundary keeps the result both useful, clear and fair. During the unprocessed period, they lost about the same amount. Those two whole diets caused different short-term intake and weight outcomes under the conditions of the trial. The experiment did not identify a guilty additive, prove long-term disease or make every packaged loaf, ready meal and tin of beans equivalent.⁴

A classification built for a different question

“Ultra-processed” did not begin as a warning to print on a packet. In 2010, Carlos Monteiro and colleagues at the University of São Paulo proposed a new way to sort foods. Instead of looking mainly at nutrients or familiar food groups, they grouped food by the extent and purpose of its processing. Their first version had three groups, which they applied to food-availability data from 48,470 Brazilian households.³

The framework later became the four-group system called NOVA. Its lens is the make-up of the food as a whole. Group one covers unprocessed or minimally processed foods. Group two covers cooking ingredients such as oil, butter and sugar. Group three includes processed foods made by combining those first groups. Group four contains ultra-processed formulations: products assembled largely from industrial ingredients, often with additives and little intact whole food.^3,4

This gave researchers a way to describe a change across an entire diet. They could measure what happened as foods and ingredients used in ordinary cooking were displaced by ready-to-eat industrial formulations. An additive blacklist performs another job. It works ingredient by ingredient inside one retailer's supply chain. One rule looks at parts; the other looks at form.

So check a shorter ingredient list and a NOVA group separately. Removing one colour from a soft drink genuinely changes the recipe. Removing an ingredient used to place a product in NOVA's fourth group may also change its category. Check the category again using the recipe that remains. An ingredient rule and NOVA can therefore point to different information without cancelling each other.

Processing also has useful purposes. Hall's paper notes that ultra-processed foods can be inexpensive, convenient, shelf-stable and relatively safe from microbes.⁴

The useful question is therefore not whether processing exists. It is what the formulation does when foods like it become the repeated base of a diet. Daily habits give the choice its real weight.

Twenty people and two menus

Before Hall's experiment, researchers could see that people who ate more ultra-processed food often had worse health outcomes. Those studies could not easily separate the food from income, activity, other dietary habits and existing health. Hall's team wanted to test one causal link directly: would people eat different amounts when the diets were controlled? Inside the ward, that question became a daily rhythm of trays, meals and leftovers.

Ten men and ten women lived continuously in the National Institutes of Health clinical centre. They crossed from one two-week menu to the other. The researchers designed the meals to present similar calories, carbohydrate, fat, sugar, sodium and fibre. Counting the drinks made the two planned menus look similar in energy density. The non-drink foods were not matched this closely, and the diets also differed in the proportions of protein, added sugar, types of fibre and types of fat. Those differences mean the diets were substantially matched, not identical, so the trial cannot assign the outcome to processing alone.

Each day brought three meals and snacks. The participants decided how much to eat.⁴

The turn came at the table. Each cleared tray turned an abstract label into a count of food actually eaten. The meals felt much alike, but the intake did not. People ate more energy from the ultra-processed menu, mainly through extra carbohydrate and fat. They rated the two diets similarly for pleasantness and familiarity, but ate the ultra-processed meals faster. Weight rose during one period and fell during the other.⁴

The supported concern comes before the uncertainty: under these controlled conditions, the ultra-processed dietary pattern caused people to eat more and gain weight over two weeks. The trial measured that short-term link; it did not establish what happens over years. That is a firm point with a narrow frame.

The trial did not measure a separate share for every cause because several features changed together. It did, however, measure one behavioural mechanism: participants ate faster on the ultra-processed menu, and faster eating was moderately correlated with extra intake. Non-drink foods were about 85% more energy dense, which the authors said likely contributed. Protein leverage is the idea that people may keep eating until they reach a protein target. Under a perfect-leverage assumption, it could explain no more than about half the difference. Softer foods delaying fullness remained a proposed, unmeasured path.⁴ This leaves a firm boundary: pace was measured, while the design cannot apportion the full effect among pace, energy density and protein leverage. The diets were substantially matched, not identical. A 20-person inpatient study cannot establish the long-term effect of every product in a broad and changing category.

Those limits are not an all-clear. They define the next research question rather than erasing the result already measured. Evidence that separated the different features could show which ones matter most, which subgroups carry the concern and which foods do not. Until then, placing the whole effect on one additive goes beyond the trial. The unknown driver does not wipe out the clear effect.

The practical conditions differed as well. Hall's team estimated that ingredients for 2,000 calories a day cost $106 a week for the ultra-processed menu and $151 for the unprocessed one at the local supermarket used in its calculation.⁴ That is not a current British shopping budget. It does show why advice to “replace everything” can fail before dinner is made. A practical choice must make room for price, time, storage, cooking facilities and access. A sound choice must also fit real life.

Concern without a blanket verdict

Years later, the question left the ward and entered the messier world of ordinary diets. A 2025 rapid evidence update by the United Kingdom's Scientific Advisory Committee on Nutrition provides the longer-term boundary used here. It found consistent and concerning associations between higher ultra-processed food consumption and adverse health outcomes. The committee also found that the long-term evidence remained almost entirely observational, meaning researchers observed people's diets rather than assigning them for years.⁵ At its latest check for new evidence, SACN recognised that the research had improved but kept its existing advice and did not start another full review.⁸

Income, body weight, total energy intake and other parts of the diet were handled inconsistently across studies. This leaves confounding: another factor may be related both to what people eat and to the health outcome. The committee could not determine how much of the association came from processing itself and how much came from established risks such as excess calories, saturated fat, salt, free sugars or processed meat.⁵

The category did not move as one block. The broad name hides a wide range of foods. Meat products and sweetened drinks were among the subgroups more consistently associated with poor outcomes. Results for breads, cereals, dairy foods, snacks, sauces and ready-to-eat foods were mixed. Vegetarian alternatives were not associated with adverse outcomes in the subgroup analyses SACN reviewed.⁵

This is why “every ultra-processed food is harmless” and “every ultra-processed food is equally harmful” both outrun the evidence. The controlled trial establishes a reason to take the dietary pattern seriously. The observational evidence adds concern across longer periods. Neither turns a category into a product-by-product verdict. The evidence now spans a controlled trial and longer-term associations, but it still does not give shoppers a product-by-product answer.

Why a visible rule still leaves people unsure

Tesco's policy responds to a real wish for something easier to use. The retailer said more than 80% of its customers were trying to eat fewer processed foods but found the issue hard to navigate. A numbered list offers a visible answer: selected additives will disappear or face restrictions in own-brand recipes.¹

But the list leaves a gap when it is asked to answer a question it was not designed for. Familiar words alone do not settle the meal. Instead, compare salt, saturated fat and sugar directly. An unfamiliar ingredient may support microbial safety, shelf life, fortification or an allergy-safe formulation. That is why a shopper still needs to check fibre, protein, fat, salt and sugar on their own terms.

The confusion reaches beyond one supermarket. In a public dialogue published in 2026, UK Research and Innovation brought 132 people through six workshops and more than 20 hours of discussion. Organisers first showed the group earlier research on how much of UK diets comes from ultra-processed food. Many participants reacted with alarm. They were surprised both by its reach and by the lack of a clear scientific definition. As the workshops continued, some called food advertising “crafty” or “insidious”. Others said “choice is an illusion” where healthier options cost more or are hard to find.⁶

Those discussions are evidence about knowledge, trust and access, not proof of biological harm. They help explain why a simple rule feels welcome and why it cannot finish the decision. A useful rule must work with a household's money, time, health needs, taste and access. The best advice has to work inside those limits.

Three questions for the packaged food that remains

A shopper can now act on the whole-diet concern without treating every packaged food alike. Here, lower regret means acting on the clearest concern without ignoring what a household can afford, store or use. The lower-regret choice begins where the strongest evidence and existing nutrition advice already agree. In a British supermarket or at home, reduce sugar-sweetened drinks, confectionery and frequent foods high in energy, saturated fat, salt or sugar and low in fibre. Make fruit, vegetables, pulses and wholegrain or higher-fibre staples the routine base where price, time, medical needs and access allow.⁵

For a packaged food that still serves a useful purpose, ask three questions together.

First, what role does it play? A daily staple deserves a different decision from an occasional sweet or salty extra. Second, what does the nutrition panel show about calories, saturated fat, salt, sugar and fibre? Third, what does the formulation make possible—or make easier to overconsume—including preservation, fortification, allergy safety, texture and convenience? Together, those checks put the food back in context and give the next shop a clear place to start.

For food, this approach points towards recognisable, minimally processed foods as the dependable base of a diet.

One separate check for the pack

The food decision and the pack are two parts of the same shop, but they need different evidence. The food evidence above cannot verify the material that carries it. After judging the food, ask what carries it. For packaging, ask whether the material is home-compostable and can return to earth safely. Treat that as a question, not a claim. Ask the brand or maker for a named material specification or independent certificate that covers composition and end of life. A vague “natural” label is not proof. A suitable fortified drink or allergy-safe loaf may remain the better choice. Food safety, nutrition, price and packaging each need their own evidence. No single badge can do all of that work.

Back at the shelf, the number 114 can tell a shopper that Tesco changed a supplier rule. The number alone does not say whether the recipe has left NOVA's fourth group, how often the food will be eaten or what a realistic replacement costs. The trial supplies the clear learning: whole dietary patterns can change how much people eat even when no single additive has been named as the cause. The practical response is to change the repeated discretionary foods most aligned with that concern, then judge the necessary packaged ones by role, nutrition and formulation together. That keeps the action tied to what the evidence can support.