Ultra-processed food is the dietary concept of the decade, and it arrived with an unusual amount of evidence behind it. It is also genuinely contested among nutrition scientists — not the association with poor outcomes, which is well replicated, but what exactly is doing the damage.
Both parts are worth understanding, because the practical advice changes depending on which is right.
The definition, and why it is awkward
The term comes from the NOVA classification, developed by Carlos Monteiro’s group. It sorts food into four groups by degree of industrial processing rather than by nutrients.
Group 1 is unprocessed or minimally processed: vegetables, meat, eggs, milk, plain oats. Group 2 is culinary ingredients: oil, butter, salt, sugar. Group 3 is processed food, made by combining groups 1 and 2 — bread, cheese, tinned fish, salted nuts.
Group 4 is ultra-processed, defined as industrial formulations typically composed of sugars, oils, fats, salt and additives not used in home kitchens.
The practical test that follows from this is the one worth remembering: look for ingredients you would not find in a domestic kitchen. Emulsifiers, protein isolates, modified starches, hydrolysed proteins, colourings, flavourings, non-sugar sweeteners, thickeners with numbers.
The awkwardness is that the category is broad and internally inconsistent. Wholemeal supermarket bread with an emulsifier is ultra-processed. So is a can of energy drink. Nutrition scientist Kevin Klatt has pointed out that NOVA does not always track actual industrial transformation — edamame pasta, made by pulverising and extruding beans, can count as minimally processed because edamame is its only listed ingredient.
Keep that in mind. It matters later.
What the evidence shows
The observational data is extensive and consistent.
An umbrella review in the BMJ in 2024 evaluated the existing meta-analytic evidence. It found moderate certainty of evidence for associations with all-cause mortality, cardiovascular disease incidence and mortality, type 2 diabetes and colorectal cancer.
The effect sizes: per 50 g/day increment, a summary risk ratio of 1.02 for all-cause mortality and 1.04 for cardiovascular disease incidence. For type 2 diabetes, 1.12 per 10% increase in dietary share.
A 2025 dose-response meta-analysis pooling 18 prospective cohorts — 1,148,387 participants, 173,107 deaths — found that the highest consumers had a 15% higher risk of all-cause mortality than the lowest, with each 10% rise in the ultra-processed share of the diet adding roughly 10% to mortality risk.
And, unusually for nutrition, there is a randomised trial.
Kevin Hall’s team at the NIH admitted participants to a metabolic ward and fed them either an ultra-processed or an unprocessed diet for two weeks each, matched for calories presented, macronutrients, sugar, sodium and fibre. Participants ate freely.
On the ultra-processed diet they consumed roughly 500 kcal more per day and gained weight. On the unprocessed diet they lost it.
That is the strongest single piece of evidence in this field, because it removes the confounding that plagues everything else. Same nutrients on paper, different processing, different outcome.
The part that is genuinely unresolved
Here is where the honest version diverges from the confident version.
All the mortality evidence is observational. People eating the most ultra-processed food differ systematically from those eating the least — income, education, smoking, physical activity, time pressure. Studies adjust for what they measure. Residual confounding is likely in a category this closely tied to socioeconomic status.
The category behaves inconsistently when you break it apart. A US cohort of over 206,000 people found sugary drinks and processed meat associated with clear cardiovascular risk, while cereals, baked goods and yogurt showed neutral or even protective associations. An analysis of the EPIC cohort found the same pattern for diabetes.
That does not prove some ultra-processed foods are protective — yogurt eaters differ from soda drinkers in other ways. But if subcategories within group 4 diverge that much, processing alone may not be what explains the pattern.
The candidate mechanisms are several and not mutually exclusive: energy density, how quickly the food can be eaten, low fibre and protein, effects of specific additives such as emulsifiers on the gut, and simple displacement of other foods. Hall’s trial shows something real happens. It does not identify which of these is responsible.
What to actually do
The practical advice survives the uncertainty, because it points the same way under every explanation.
Read ingredients, not nutrition panels. The front of a package is marketing and the nutrition table can be gamed. The ingredient list is where the answer is. Long lists containing substances you do not recognise are the signal.
Focus on the categories with the clearest evidence. Sugary drinks and processed meat carry the most consistent risk associations across studies. If you change nothing else, change those.
Do not treat the label as binary. Tinned beans, frozen vegetables, plain yogurt, wholemeal bread and tinned fish are processed to varying degrees and are among the most useful foods available — cheap, high in fibre or protein, and easy. Rejecting the whole category on principle removes those and makes a diet worse, not better.
Watch the share, not the item. The dose-response data is about the proportion of your diet, not about whether any single item is permitted. Ultra-processed food accounts for roughly 57% of calories in the UK and 58% in the US. Moving from 60% to 40% is a substantial change achieved without banning anything.
Cook the base, buy the rest. The realistic version for most households is not cooking everything. It is that the foundation of most meals — the vegetables, grains, pulses, meat or fish — comes from group 1, and the convenience sits around it rather than replacing it.
What this means over decades
The individual effect sizes here are small. A risk ratio of 1.02 per 50 g increment is not the kind of number that should alarm anyone about a single meal.
What makes it matter is the exposure. Ultra-processed food is not an occasional item; for most people in the UK and US it is the majority of what they eat, every day, for their whole adult life. Small effects multiplied by that kind of exposure are how population-level patterns get made.
This is also why the honest uncertainty about mechanism does not change the recommendation much. Whether the problem is emulsifiers, or energy density, or eating rate, or the fibre displaced — the intervention is the same. Shift the share downwards. You do not need to know which mechanism you are avoiding.
This article is for general information and is not medical advice.
Sources
- Lane MM, Gamage E, Du S, et al. Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses. BMJ, 2024;384:e077310. https://pmc.ncbi.nlm.nih.gov/articles/PMC10899807
- Ultra-processed foods and risk of all-cause mortality: an updated systematic review and dose-response meta-analysis of prospective cohort studies. Systematic Reviews, 2025. https://link.springer.com/article/10.1186/s13643-025-02800-8
- Hall KD, Ayuketah A, Brychta R, et al. Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metabolism, 2019;30:67–77.
- Monteiro CA, Cannon G, Levy RB, et al. Ultra-processed foods: what they are and how to identify them. Public Health Nutrition, 2019;22:936–941.
- American Council on Science and Health. Beyond NOVA: how to interpret the evidence on ultra-processed foods, 2026. https://www.acsh.org/news/2026/07/29/beyond-nova-how-interpret-evidence-ultra-processed-foods-50248
This article is for general information and is not medical advice. If a health problem is affecting your daily life, speak to your GP.
Comments
We read every comment. Be kind, stay on topic, and please don't ask for medical advice — we can't give it.