Are ultra-processed foods dangerous for your health?

1. The concern is primarily a dietary pattern and repeated exposure, not one package

Nova describes a formulation. It does not measure a toxic dose or define a threshold above which one food becomes dangerous. An occasional serving therefore does not predict disease, just as a product outside Nova 4 does not guarantee a healthy diet.

Most health studies examine the daily share of ultra-processed products by calories, weight, or servings. People with the highest intake may also differ in smoking, physical activity, income, sleep, access to cooking, and many other habits. Analyses adjust for some of those factors but can never guarantee that all are removed.

ANSES concluded that associations exist but that the overall weight of evidence remains low for several relationships and that no consensus definition perfectly characterizes ultra-processed foods. “Low” does not mean no signal or proof of safety; it limits the strength of any causal conclusion. ANSES — opinion on intake of so-called ultra-processed foods

• Learn how to identify ultra-processed foods without a misleading list →

2. Trials and cohorts do not answer the same question

A randomized trial assigns diets under controlled conditions. It reduces certain differences between participants and can show a causal effect of the tested menu on a short-term outcome such as energy intake or weight. Keeping people in a metabolic unit for years, however, would be unrealistic and unethical for many outcomes.

A cohort study observes what thousands of people report eating, then records diagnoses or deaths. It can examine cancer or cardiovascular disease over time but cannot capture every difference between people with high and low intake. An association that persists after adjustment remains an association.

A systematic review or meta-analysis combines studies under a stated method. It can improve precision and examine consistency without removing biases shared by the included studies. Ten imperfect cohorts do not become a randomized trial simply because researchers pool them.

3. Hall’s trial found greater energy intake on one specific diet

In 2019, 20 adults ate two ad libitum diets in a metabolic unit for two weeks each in randomized order. The offered menus were presented as matched for calories, macronutrients, sugar, sodium, and fiber. On the ultra-processed diet, participants consumed about 508 kcal more per day and gained approximately 0.9 kg; on the unprocessed diet, they lost approximately 0.9 kg. Hall et al., 2019 — randomized metabolic-unit trial

This is important causal evidence for those two menu sets over that period. It shows that when food is available without restriction, properties of that ultra-processed diet can lead people to eat more. It does not identify which ingredient, additive, or process caused the effect or show that every Nova 4 product behaves the same way.

The trial was very small and short. Energy density excluding beverages, texture, eating rate, and the way fiber was provided differed. It measured no cancer, diabetes, or cardiovascular events. The firm conclusion concerns excess energy intake and short-term weight change, not universal chronic danger.

4. The UPDATE trial found weight loss on both diets, but more on the minimally processed diet

The UPDATE crossover trial published in 2025 randomized 55 adults; 50 provided data for at least one diet, and 43 completed both eight-week periods. Both diets followed the United Kingdom’s Eatwell Guide. The minimally processed diet produced a mean weight reduction of 2.06%, compared with 1.05% on the ultra-processed diet. The difference was −1.01 percentage points, with a 95% confidence interval of −1.87 to −0.14. Dicken et al., 2025 — UPDATE trial in Nature Medicine

The result qualifies the claim that a diet high in Nova 4 foods prevents all weight loss: both periods produced weight loss. Under comparable overall dietary guidance, however, the minimally processed version produced approximately twice the relative reduction across eight weeks.

The sample was 90.9% women. Participants lived freely, an order effect was possible, and the ultra-processed diet had greater energy density. Several secondary outcomes did not differ; LDL cholesterol fell more during the ultra-processed period. The trial compared complete diets, not the individual danger of every product.

5. A 12-month trial in 2026 found a limited weight benefit from advice to reduce ultra-processed foods

In a 2026 trial of 148 people with obesity, advice to reduce ultra-processed foods was added to an energy-restricted diet. The weight difference was statistically detectable, but the authors considered it not clinically important, with no clear differences in the other assessed outcomes. 12-month randomized trial, 2026 — reducing ultra-processed foods and body weight

This study tested a strategy closer to everyday life and over a longer period than metabolic-unit trials. It also shows that “reducing Nova 4” is not one uniform intervention: participants choose which foods to replace, creating differences in calories, adherence, and substitution quality.

The result neither justifies abandoning reduction efforts nor supports promises of major weight loss. A sustainable energy deficit remains the primary driver of fat loss. Processing may make that deficit easier or harder through satiety, energy density, eating rate, and convenience.

• Understand why energy balance remains the basis of weight change →

6. Large cohorts find associations with cancer, cardiovascular disease, and diabetes

In NutriNet-Santé, a French cohort of 104,980 participants, each 10-percentage-point increase in the proportion by weight of ultra-processed foods was associated with higher relative risks of overall cancer, with a hazard ratio of 1.12, and breast cancer, with an HR of 1.11. Another analysis of 105,159 participants reported an HR of 1.12 for cardiovascular disease per 10-point increase. Fiolet et al., 2018 — ultra-processed foods and cancer, Srour et al., 2019 — ultra-processed foods and cardiovascular diseases

An analysis of 104,707 participants associated each 10-point increase with the incidence of type 2 diabetes, with an HR of 1.15 and then 1.13 after additional nutritional adjustment. Attenuation without disappearance of the signal suggests that the measured nutrients explain part of the association but cannot identify what explains the remainder. Srour et al., 2020 — ultra-processed foods and type 2 diabetes

These risk ratios compare average trends, not an individual’s fate. Questionnaires, recipes, and classifications may be imperfect, and consumers with the highest intake differ in other behaviors. Consistency across several results supports caution and further research without turning 1.12 into a personal probability or proof of one mechanism.

7. A BMJ umbrella review found many signals, often with low certainty

A BMJ umbrella review published in 2024 included 45 pooled analyses representing nearly 9.9 million participants. Thirty-two outcomes showed a direct association with exposure to ultra-processed foods. Four analyses reached the statistical category described as “convincing.” Lane et al., 2024 — umbrella review of health outcomes

A statistical evidence hierarchy should not be confused with causal certainty. Under GRADE, only four analyses had moderate certainty, 22 had low certainty, and 19 had very low certainty. Most data came from observational studies vulnerable to residual confounding and measurement error.

Multiple meta-analyses may also share primary studies. Adding participant counts illustrates the scale of the evidence base, not a fully independent population for every outcome. The review supports cautious public-health recommendations, not the claim that “every ultra-processed food causes 32 diseases.”

8. Not all Nova 4 subgroups show the same signal

In the European EPIC cohort, 266,666 participants were followed for multimorbidity combining cancer, cardiovascular disease, and diabetes. Higher consumption of approximately 260 g per day was associated with an HR of 1.09. The signal mainly involved processed animal products and sugary drinks; it was not clear for breads and cereals or plant-based alternatives. Cordova et al., 2023 — subgroups and multimorbidity in EPIC

Across three US cohorts published in 2024, total intake was associated with cardiovascular disease, HR 1.11, and coronary heart disease, HR 1.16, while the association with stroke, HR 1.04, was not significant. Sugary drinks and processed meats showed unfavorable associations; some breads, cereals, dairy products, and snacks showed inverse associations. Mendoza et al., 2024 — ultra-processed, subgroups and cardiovascular risk

These differences may reflect the foods, nutrients, substitutions, and measurement methods. They neither automatically clear one subgroup nor prove that another acts alone. Above all, they show that useful guidance should prioritize the best-documented and most frequent sources instead of treating Nova 4 as one homogeneous block.

9. Several plausible mechanisms may act together

High energy density, soft texture, limited chewing, and faster eating can increase energy intake before satiety signals slow the meal. A highly palatable formulation that is widely available and served in large portions may reinforce this pattern without requiring any additive to be toxic.

Dietary displacement also matters: more sugary drinks, snacks, or formulated meals can mean fewer fruits, vegetables, legumes, whole grains, and home-cooked meals. The average diet may then contain less fiber or more sodium, sugar, and saturated fat. Nutritional adjustments do not always erase the association, but they confirm that these factors explain part of it.

ANSES also examines disruption of the food matrix, newly formed substances, certain additives, mixtures, and packaging-related exposures. It recommends separating these dimensions instead of attributing every effect to the Nova number alone. Several mechanisms remain hypotheses or have specific evidence that cannot be extended to the entire category. ANSES — analysis of possible mechanisms and research needs

10. Reduce the main sources without sacrificing nutrition fundamentals

Start with seven days of observation. Identify sugary drinks, snacks, desserts, formulated meat products, and ready-to-eat meals that recur. Record servings, calories, protein, fiber, and what each product replaces. This hierarchy prevents you from spending all your effort on an occasional condiment.

Choose a substitute with the same function: sparkling water or an unsweetened drink for soda, a simple meal cooked in extra portions for a daily prepared dish, fruit and plain yogurt for a formulated snack, or bread with a shorter ingredient list for a comparable product. The substitute must remain affordable, satisfying, compatible with available time, and enjoyable enough to repeat.

For a physique or performance goal, the priorities remain total energy, sufficient protein, essential fats, carbohydrates suited to performance, then food quality, fiber, and micronutrients. Reducing Nova 4 foods may simplify that structure but does not replace tracking. A handful of minimally processed nuts can provide substantial energy; a formulated protein option may remain useful on occasion.

• Build a complete, flexible strength-training diet →

11. Track for two weeks, then adjust one lever

After the swap, monitor hunger, digestive comfort, training energy, preparation burden, and budget for 1–2 weeks. If body weight is a goal, use an average across several days rather than one weigh-in. If the alternative is not sustainable, change its format or frequency instead of concluding that any reduction is impossible.

Nalko does not assess whether a product is dangerous. The app distributes usable calories across the groups used in the analysis, maintains an undetermined category, and displays an internal ranking of Nova 2–4 products. This lets you observe recorded changes without inventing an individual risk estimate or official health score.

The proportionate conclusion in 2026 is twofold: reducing high exposure is a precaution consistent with short trials, cohorts, and dietary guidance; banning every product or promising that one rule prevents disease exceeds the evidence. A sustainable diet should be judged as a whole.

• Week 1: observe recurring products, portions and contexts.
• Week 2: replace a dominant source without changing everything else.
• Measure: usable calories and macros, hunger, digestion, performance, cost, and time.
• Decide: keep, adapt the substitution or choose another frequent source.
• Avoid: compensating, assigning blame, or removing a medically necessary product.