Avoiding ultra-processed food

• Eat fewer calories without counting them, by choosing foods that are harder to overeat.
• Improve overall diet quality: more fiber, whole plant foods and protein, less added sugar and refined starch.
• Lower long-term risk of obesity, type 2 diabetes, heart disease and depression.
• Reduce exposure to additives such as emulsifiers and non-sugar sweeteners whose long-term effects are uncertain.

NOVA was created by Carlos Monteiro and colleagues at the University of São Paulo. It sorts every food into one of four groups by the extent and purpose of processing, not by fat, sugar or salt content [16].

The marker ingredients are things like high-fructose corn syrup, hydrogenated or interesterified oils, hydrolyzed proteins, and classes of cosmetic additives such as flavors, colors, emulsifiers and non-sugar sweeteners [16].

How much of the diet is ultra-processed?

UPF are now the main source of calories in rich countries. US adults got 57.0% of their calories from UPF in 2017 to 2018, up from 53.5% in 2001 to 2002 [17]. In the UK the figure was about 56% from 2008 to 2019 [18]. The US MAHA Commission states that nearly 70% of an American child's calories now come from UPF [19].

Ultra-processed versus processed

Processing alone is not the target. Freezing peas, canning beans, fermenting yogurt and baking bread are all processing. The protocol targets group 4, the formulations built from refined extracts and additives.

There is no single official threshold. People follow the protocol at different levels of strictness:

• Strict: no group 4 products at all. Everything is cooked from whole foods and simple ingredients.
• Target-based: UPF below a set share of calories, often 10% to 20%, measured with a food log. Typical Western diets sit above 50% [17] [18].
• Swap-based: keep the meal pattern, swap the worst UPF categories first. Large US cohorts found that sugary and artificially sweetened drinks and processed meats carried most of the risk, while packaged whole-grain bread, cold cereals and yogurt did not [3] [4] [9].

The swap-based version matters because UPF are not one thing. In three Harvard cohorts, refined breads, sauces, sweetened drinks, animal-based products and ready meals were linked to higher type 2 diabetes risk, while cereals, dark and whole-grain breads, packaged snacks, fruit-based products and yogurt were linked to lower risk [4].

What a day can look like

• Breakfast: oats with milk, fruit and nuts, or eggs with bread made from flour, water, yeast and salt.
• Lunch: a bean and vegetable salad with olive oil and vinegar, or leftovers.
• Dinner: fish, chicken or tofu with rice or potatoes and vegetables.
• Snacks: fruit, plain yogurt, cheese, nuts.

Official guidance has started to move this way. Brazil built its national dietary guidelines around food processing, and the 2025 to 2030 Dietary Guidelines for Americans tell people to avoid highly processed food (see the examples on this page).

Kevin Hall's team admitted 20 weight-stable adults to the NIH Clinical Center. Each person ate an ultra-processed diet for 2 weeks and an unprocessed diet for 2 weeks, in random order, and could eat as much or as little as they wanted. Meals were matched for presented calories, energy density, macronutrients, sugar, sodium and fiber [1].

Results [1]:

• Energy intake was 508 kcal a day higher on the ultra-processed diet.
• The extra came from carbohydrate (280 kcal) and fat (230 kcal), not protein.
• Weight changes tracked intake closely (r = 0.8): +0.9 kg on UPF, -0.9 kg on unprocessed food.

The newer trial

The 2025 UPDATE trial in England asked a harder question: does processing still matter when both diets follow national guidelines? 55 adults ate two delivered 8-week diets that both met the UK Eatwell Guide, one minimally processed and one ultra-processed. Both led to weight loss, but more on minimally processed food: 2.06% versus 1.05% of body weight [7]. That 1 percentage point gap is small but came from diets that were otherwise nutritionally similar.

Why might people eat more?

Critics argue the 2019 result can be explained by measurable traits such as energy density, fiber and glycemic load rather than processing itself [5]. Hall has said in interviews that UPF tend to be more energy-dense and more often hyper-palatable, and that an interim look at a newer trial suggested people did not overeat when those traits were matched; that trial has not yet been published. The popular idea that UPF act like addictive drugs got little support from a PET study in which an ultra-processed milkshake caused no measurable dopamine release in the striatum [20].

The 2024 BMJ umbrella review pooled 45 meta-analyses covering nearly 9.9 million people. Higher UPF exposure was linked to higher risk for 32 of them [2]. The most credible associations (class I) were cardiovascular death (risk ratio 1.50), type 2 diabetes (12% higher per dose increment), anxiety (odds ratio 1.48) and common mental disorders (odds ratio 1.53) [2].

What the evidence does not show

• Causation for disease. No trial has run long enough to test heart attacks, cancer or death. The trials that exist measure weight and calorie intake over weeks [1] [7].
• That every UPF is equal. In the longest US mortality study, all-cause death was only 4% higher in the top quarter of intake, with no link to cancer or cardiovascular death; processed meats and sweetened drinks drove most of the signal [9]. Within each level of diet quality, UPF intake showed no consistent link with mortality [9].
• That the label is precise. Food experts agree only weakly on NOVA groups, with kappa scores of 0.32 to 0.34 [6].

Proposed mechanisms, from best to least supported:

1. Eating more calories. The two feeding trials show people eat more or lose less weight on UPF diets [1] [7]. Energy density and palatability are the main suspects [5].
2. Displacing whole foods. Diets high in UPF tend to be lower in fiber and protective plant compounds; the Lancet series argues this displacement is a key driver of chronic disease [21].
3. Additives and the gut. In mice, the emulsifiers carboxymethylcellulose and polysorbate-80 caused gut inflammation and metabolic syndrome [14]. In a controlled feeding study of 16 healthy adults, 15 g a day of carboxymethylcellulose for 11 days reduced microbial diversity and short-chain fatty acids [15]. That is a signal, not proof of disease.
4. Addiction. Some researchers argue UPF high in refined carbohydrate and fat can be addictive [22]. A PET study did not find the drug-like dopamine surge this idea predicts [20].

Is it the processing or the nutrients?

This is the core of the scientific debate. Critics say NOVA mostly repackages what nutrient profiling already captures: sugar, salt, saturated fat, low fiber and high energy density [23] [5]. Supporters point to cohort analyses that adjust for nutrients and still find risk, and to the UPDATE trial, where both diets met nutrient guidelines and still differed in weight loss [24] [7].

The main criticisms:

• Too broad. One category spans soda, hot dogs, infant formula, protein bars and packaged whole-grain bread. Gibney and colleagues argued this makes it unusable for food-based guidelines [23].
• Inconsistent. When 159 to 177 French food specialists sorted the same foods, agreement was low (Fleiss kappa 0.32 to 0.34), even with ingredient lists in hand [6].
• Confounded. People who eat more UPF also tend to smoke more, exercise less and have more chronic disease at baseline, and cohorts cannot fully remove this [11].
• Mislabels healthy foods. Some group 4 foods track with lower disease risk, including whole-grain breads, cereals and yogurt [4] [3].

Supporters reply that the evidence is now large and consistent across countries and outcomes, that two randomized trials show a causal effect on intake and weight, and that a simple message to prefer minimally processed food is easier to follow than nutrient targets [24] [21].

A fair reading

Both camps agree that sugary drinks, processed meats and refined snack foods are worth cutting. They disagree on whether processing is the cause or a marker, and on whether foods like packaged whole-grain bread deserve the same label as soda.

A practical order of operations:

1. Drinks first. Replace soda and sweetened drinks with water, tea, coffee or milk. Sweetened beverages showed one of the clearest mortality links [9].
2. Processed meats next. Ready-to-eat meat products had the strongest mortality associations in the longest US study [9].
3. Breakfast. Swap sweetened cereals and pastries for oats, eggs, plain yogurt and fruit.
4. Batch cook. Two or three cooking sessions a week cover most meals.
5. Keep easy whole foods. Fruit, nuts, cheese, canned beans and frozen vegetables are processed but not ultra-processed.
6. Read labels. One industrial ingredient or cosmetic additive means group 4 [16].

How to measure adherence

Log everything you eat for 3 to 7 days and estimate the share of calories from group 4 products. Repeat monthly. A simpler proxy is counting UPF servings a day; in the Nurses' Health Study II, the lowest group ate fewer than 4 a day and the highest more than 8.8 [11].

Common mistakes

• Swapping one UPF for another with a health claim, such as protein bars or plant-based meat.
• Cutting nutritious group 3 and group 4 foods such as whole-grain bread and plain yogurt, which were not linked to harm [4].
• Expecting fast weight loss. The UPDATE trial difference was about 1 percentage point of body weight over 8 weeks [7].

• Swapping one UPF for another with a health claim, such as protein bars, 'keto' snacks or plant-based meat.
• Dropping nutritious foods such as whole-grain bread, plain yogurt or canned beans because they are processed.
• Treating 'organic' or 'natural' on a label as proof a product is not ultra-processed.
• Ignoring calories from whole foods: oils, nuts and cheese are easy to overeat too.
• Making it so strict that it becomes stressful or isolating, which can feed disordered eating.

• People with a current or past eating disorder should make diet changes with their care team, since strict food rules can trigger relapse.
• People who rely on fortified products (such as infant formula, medical nutrition drinks or fortified cereals) should not drop them without advice.
• People with diabetes on insulin or sulfonylureas should check with a clinician, since lower calorie intake can change medication needs.

• More time and money spent on shopping and cooking.
• Short-term digestive changes when fiber intake rises quickly.
• Lower intake of nutrients that some UPF are fortified with, such as iodine in iodized products or B vitamins in cereals, if replacements are not chosen with care.