Ultra-processed food
The more ultra-processed food people eat, the more disease and earlier death they tend to have, and in short feeding trials people overeat on it without noticing. Whether processing itself does harm beyond the calories, sugar, salt and fat these foods carry is plausible but still unproven — and the headline risk figures are softer than they sound.
Cut back where the signal is strongest first: sugary and diet drinks, and processed meats. Then use the kitchen test on labels and swap toward whole foods at a pace you can keep. Getting ultra-processed food from the 50–60% of calories most Western adults eat down toward 10–20% is a realistic destination; perfection is not the goal.
What "ultra-processed" means — and why experts argue about it
Ultra-processed food (UPF) supplies about 53% of adults' calories in the United States, about 54% in the United Kingdom, and 62% in American children.[1] [2] The US adult figure was 56% in the 2017–2018 national survey; the drop to 53% by 2021–2023 is the first significant fall recorded.
The standard definition is the NOVA classification, which sorts food by how it was made rather than what nutrients it contains:
| Group | What it is | Examples |
|---|---|---|
| 1 — Unprocessed or minimally processed | Whole foods, at most dried, frozen, fermented or pasteurised | Vegetables, fruit, legumes, plain oats, milk, eggs, raw meat, plain yogurt |
| 2 — Culinary ingredients | Pressed, refined or milled from group 1 | Olive oil, butter, sugar, salt, honey |
| 3 — Processed foods | Group 1 plus group 2, to keep longer | Tinned fish, artisan bread, salted nuts, hard cheese |
| 4 — Ultra-processed | Industrial formulations with at least one ingredient rarely used at home, or a cosmetic additive | Soft drinks, packaged bread, reconstituted meat, instant noodles, flavoured yogurt, most protein bars |
One qualifying ingredient is enough to put a food in group 4: something you could not buy for home cooking, or one of the cosmetic additives clustered at the end of the list, such as emulsifiers, thickeners, colours, flavour enhancers and non-caloric sweeteners.[3]
Serious critics argue that NOVA mixes up which ingredients a food contains with how it was processed, and that much of the harm may simply be the nutrient profile.[4] It is also hard to apply consistently: when 159 food and nutrition specialists sorted 120 supermarket products with the ingredient lists in front of them, their agreement was only fair (a kappa of 0.32, where 1 is perfect agreement and 0 is chance).[5] Against that, in many cohorts, though not all, the harm signal persists after adjusting for nutrient quality.[6]
The headline numbers are real, but softer than they sound
The link is consistent and dose-dependent, but observational and graded low certainty. The biggest overview, pooling 45 meta-analyses covering roughly 10 million people, found high UPF intake linked to 32 different bad outcomes.[7] Each was graded both for how credible the association looks and for how much confidence the evidence deserves. Four reached the top credibility tier — cardiovascular death (about 50% higher risk, but very low certainty), type 2 diabetes (about 12% higher per step up in intake, moderate certainty), anxiety (about 48% higher) and common mental disorders (about 53% higher), both low certainty. Across all 45, only four earned moderate certainty and none earned high.
For death from any cause, the largest dose-response analysis — 18 cohorts, 1.15 million adults — found the heaviest UPF eaters had about 15% higher mortality than the lightest, with a plausible range of roughly 9% to 22%.[8] The often-quoted 62% comes from a single Spanish cohort of university graduates with just 335 deaths, and its plausible range ran from 13% to 133% higher.[9]
The risk per 10% of calories is unsettled: about 10% higher mortality in that same analysis, with the cohorts disagreeing wildly, and about 3% in a second analysis built for an eight-country model.[10] Read it as roughly 3–10% per 10% of calories. The model's estimate that ultra-processed food causes between about 4% (Colombia) and 14% (UK and US) of premature deaths depends entirely on treating these associations as cause and effect.
A 2025 three-part Lancet series concluded that UPFs harm health across the major organ systems and called for policy action.[11] It was led by the researchers who created NOVA, and the journal later published five critical letters with the authors' reply.[12]
15% or 4%? How much is the company these foods keep
Much of the association may be smoking, income and overall diet rather than processing; the clearest test comes from two US cohorts of nurses and doctors: among 114,000 of them, followed for over three decades, the heaviest quarter of UPF eaters had only 4% higher mortality than the lightest — 1,536 against 1,472 deaths per 100,000 person-years.[13] None of that excess came from cancer or cardiovascular disease. Within any band of overall diet quality, the UPF link largely vanished, while diet quality predicted death at every level of UPF intake.
The two figures are compatible; three things pull the health-professional figure down. The gap between heavy and light eaters was narrow (about 7.4 against 3.0 servings a day). Removing smoking from the statistical model made the association much stronger, a direct measure of how much work confounding does. And diet was re-measured every four years in an unusually health-conscious group. So the 15% describes how much worse heavy UPF eaters fare in the population at large; the 4% is closer to what a non-smoker who already eats reasonably well should expect from cutting back further. The diet-quality result is not the last word: a 2026 analysis of 47,999 US adults found the mortality link only partly weakened by a broader diet-quality adjustment.[14]
Three more reasons for caution:
- Hidden confounders need not be strong. For the link with weight gain, an unmeasured factor — depression, a tendency to overeat, poor access to good food — would need a risk ratio of only about 1.55 with both UPF intake and weight, beyond the confounders already adjusted for, to explain it away.[15]
- UPF predicts deaths it cannot cause. In a 450,000-person European cohort, higher intake also went with about 12% more accidental deaths per step up in intake — which the researchers read as residual confounding.[16]
- Illness may come first. People who are already unwell may drift toward cheaper, more convenient food.
What the feeding trials show
In short-term, tightly controlled feeding trials, people who eat freely from an ultra-processed diet eat more without trying to, but no trial has yet pinned that on processing rather than on how energy-dense and quick to eat the food is.
The foundational trial. In a 2019 US government metabolic-ward study, people on a UPF diet ate about 500 kcal a day more than on an unprocessed one.[17] A Japanese pilot pointed the same way — nine men ate about 800 kcal a day more over one week, and chewed less per calorie — but with so few people the true figure could be anywhere from 340 to 1,280 kcal.[18] A re-analysis of 2,733 meals from the US trial found energy density, eating rate and hyper-palatable foods tracked how much people ate; additives were never tested.[19]
The 2025 London trial — important but contested. In a University College London (UCL) trial, 55 adults, nine in ten of them women, spent eight weeks on each of two diets built to UK healthy-eating guidelines, one ultra-processed and one minimally processed, eating as much as they liked.[20] Both diets produced weight loss, the minimally processed one about twice as much — a difference of about 1 kg over eight weeks.
Three formal critiques followed. The critics argue the diets were mismatched in a way that could explain the result without invoking processing:[21] on the trial's own corrected figures, the ultra-processed diet was about a quarter more energy-dense.[22] A 1-kg difference is of limited public-health importance, and the handling of missing data and of carry-over between the two diet periods was questionable.[23] And body fat was measured by bioimpedance, never by a DXA (X-ray body-composition) scan, so the fat-loss signal may partly be a water shift;[24] total body water fell about half a kilogram more on the minimally processed diet.[25] Several authors have obesity-drug industry ties, one wrote a popular anti-UPF book, and the 2019 trial's lead author is a co-author, so the two trials are not fully independent.
Texture alone moves intake. A two-week trial in 41 healthy-weight young adults matched the calories served, held palatability and UPF content constant, and changed only how fast the food could be eaten: people ate 369 kcal a day less on the slower version, and the gap held.[26] Its lead author also co-wrote the first critique above and declares food-industry speaking fees. Still, it is strong evidence that eating speed matters, and that texture may be doing more of the work than "processing" itself.
Two results that show less than they seem. Inside a weight-loss trial, a statistical model of 356 people's diet records linked swapping 10% of calories from ultra-processed to minimally processed food with modestly more weight and fat loss over six months — an association, not a tested swap.[27] And a 2026 trial gave 27 young adults two fully provided 14-day diets, 81% against 0% ultra-processed, matched even for energy density. Because the diets were provided, it could not test overeating; at a buffet afterwards intake was the same, except that the 18- to 21-year-olds ate more after the UPF period.[28]
What it might be doing in the body
These pathways are plausible, but mostly built on animal, cell and observational data, with very few human trials.
Speed and palatability. Processed food needs less chewing and is emptied from the stomach faster, so calories can arrive before fullness hormones catch up. UPFs also cluster around combinations rare in nature — fat with salt, fat with sugar, carbohydrate with salt. A quantitative definition of these "hyper-palatable" foods flagged 62% of the 7,757 items in the US food-composition database.[29] Long-term consumption is proposed to blunt the brain's dopamine reward signalling, a tolerance pattern seen in animals and hinted at in human brain scans; "food addiction" is still a questionnaire-measured research idea, not a diagnosis.[30]
The gut. UPF-heavy diets are linked to lower microbial diversity and fewer of the bacteria that maintain the gut lining (from a narrative review, not a systematic one).[31] The proposed chain — additives thin the mucus layer, bacterial fragments leak into the blood, low-grade inflammation follows — is well shown in animals but thin in people: human trials of emulsifiers have so far found microbiome changes but no inflammation.
A fibre clue. Imidazole propionate, a gut-bacterial product that interferes with insulin signalling, is higher in type 2 diabetes.[32] Among 7,432 US health professionals, the top fifth for blood levels had 82% higher coronary heart disease risk than the bottom fifth. The dietary factor that mattered was fibre, especially pectin, not the amino acid it is made from: more fibre went with lower levels, and the amino acid and gut bacteria pushed levels up only in people eating little fibre.[33] No study has measured UPF intake against it directly; the link runs through the low fibre typical of these diets.
Missing plant compounds. Minimally processed foods carry roughly ten times the antioxidant capacity of ultra-processed ones — but ordinary processed foods scored no better than ultra-processed, and the study measured foods, not what people ate.[34]
The additives: what is actually known
The human trials of additives are few, though no longer negligible.[35]
| Additive | Where you find it | What the evidence shows |
|---|---|---|
| Emulsifiers (carboxymethylcellulose, polysorbate 80) | Ice cream, baked goods, sauces, plant milks | Erode the gut mucus layer and promote inflammation in animals; linked to inflammatory bowel disease and metabolic syndrome, mostly in animal and observational studies. In people, two short trials (seven adults for 11 days; 60 adults for four weeks) found microbiome changes but no inflammatory signal[36] [37] |
| Carrageenan | Plant milks, processed cheese, deli meats | Colitis models use a degraded form, not the food additive — but in the four-week trial, food-grade carrageenan modestly raised gut permeability against people's own starting level, though not against placebo.[38] Contested |
| Non-caloric sweeteners (sucralose, saccharin, aspartame) | Diet drinks, low-sugar yogurts, protein bars | Shift the microbiome; only saccharin and sucralose worsened glucose responses in the best trial, very person-specific — see Sugar substitutes |
| Maltodextrin | Sports drinks, sauces, "fat-free" foods | Thins the mucus layer in rats; human trials show shifts in gut bacteria but did not measure inflammation |
| Azo dyes (Red 40, Yellow 5) | Sweets, drinks, cereals | Cause colitis in mice. In a trial of 297 UK children, dye mixtures slightly increased hyperactivity, but a preservative was also in the drinks, so the dyes cannot be isolated[39] |
| Titanium dioxide | Whitener in sweets, sauces, supplements | Builds up in the gut lining; oxidative stress and DNA damage in lab studies |
Titanium dioxide is the one additive a major regulator has removed: the EU withdrew it in August 2022, after its food-safety authority concluded it "can no longer be considered safe" because damage to DNA could not be ruled out. It is still permitted in the United States, where a petition to remove it has been pending since 2023.
Brain, aging and autoimmunity: early signals
Mental health and cognition. Anxiety and common mental disorders are among the strongest associations in the umbrella review (see above), with no long-term trials. On cognition there are only five studies, three of which found worse outcomes; the reviewers say future work must account for overall diet quality to tell whether processing matters at all.[40]
Biological aging. In a snapshot of 16,055 US adults, each extra 10% of calories from UPF went with about 2.4 months (0.21 years) of extra biological age on a blood-marker score, and the top fifth of eaters (68–100% of calories) were 0.90 years older biologically than the bottom fifth.[41] Adjusting for diet quality weakened this only modestly, but a single snapshot cannot show cause. A DNA-methylation pilot in 15 women found differences between heavy and light eaters — a lead, nothing more.[42]
Frailty and muscle. Across 246,261 people, high UPF intake went with about 40% higher risk of frailty in studies that followed people over time. The link to weak muscle strength was much smaller, appeared only in snapshot studies, and was absent in the two that followed people forward.[43] One study of 886 older Spanish adults linked more than three servings a day to shorter telomeres, the protective caps on chromosomes, but it too was a snapshot.[44]
Autoimmune disease. Additives have been proposed to disturb the gut barrier and tilt the immune system toward inflammation.[45] Among 207,000 UK adults followed for twelve years, the top fifth of UPF eaters had 17% more new rheumatoid arthritis — an association whose plausible range only barely clears no effect.[46]
Not all ultra-processed foods are equal
The category runs from diet soda to fortified wholegrain bread, and the harm is uneven. In the health-professional cohorts, the remaining signal came mainly from:[47]
- Processed meats and ready-to-eat meat, poultry and seafood products — the strongest signal
- Sugar-sweetened and artificially sweetened drinks
- Dairy-based desserts
- Ultra-processed breakfast foods
Some group-4 foods show no detectable harm. Among 267,000 European adults followed for a median of 11 years, industrial breads and cereals and plant-based alternatives were not linked to developing two or more of cancer, cardiovascular disease and type 2 diabetes.[48] "No harm detected" is not "protective", and it does not hold everywhere: in the US cohorts, ultra-processed breads and breakfast foods carried a small but clear excess mortality. A fortified plant milk or packaged wholegrain bread is a reasonable stepping stone for a busy adult, not a destination.
What to actually do
- Run the kitchen test. If a label lists something you could not buy for home cooking — soy protein isolate, high-fructose corn syrup, hydrogenated or interesterified oils, dough conditioners, hydrolysed proteins — it is ultra-processed.
- Scan the end of the ingredient list. Emulsifiers (soy lecithin, carrageenan, polysorbate 80), thickeners (xanthan gum, maltodextrin), colours, flavour enhancers and sweeteners mark a group-4 food, whatever the front of the pack says.
- Ignore the marketing. "All-natural", "fortified", "plant-based", "high-protein" and "gluten-free" say nothing about processing. Added protein does not fix overeating: in 21 adults spending 54 hours in a sealed metabolic chamber, a high-protein ultra-processed diet improved energy balance, but people still overate.[49]
- Make the easy swaps first.
| If you eat | Try instead |
|---|---|
| Sugary or diet soda | Water, sparkling water, tea, black coffee |
| Flavoured yogurt | Plain yogurt or kefir with fruit and honey |
| Boxed breakfast cereal | Rolled or steel-cut oats; plain yogurt with nuts |
| Mass-produced sliced bread | Fresh bread (flour, water, salt, yeast) or wholegrain sourdough |
| Margarine and spreads | Olive oil, butter, avocado, nut butter |
| Processed meat (deli, sausage, bacon) | Roasted whole cuts, tinned sardines, eggs, legumes |
| Packaged snacks | Nuts, fruit, hard cheese, dark chocolate |
| Frozen ready meals | Batch-cooked one-pot meals from whole ingredients |
- Aim for a share, not perfection. No study has found a safe threshold. The big dose-response analyses assumed a straight line rather than testing for one, and the one cohort that looked for a threshold found the link with cardiovascular death disappeared below roughly 2.3 to 2.4 servings a day — but it never examined death overall.[50] So the 10–20% target is about practicality, not biology; the strongest signals, sugary drinks and processed meat, come first, and going lower is optional.
What governments and expert panels say
Governments are moving faster than the evidence graders.
- United States. The 2025–2030 Dietary Guidelines say to "limit highly processed foods, added sugars, and refined carbohydrates" ("highly processed", not NOVA's term).[51] The advisory committee behind them had graded the evidence linking UPF to obesity as only "limited". The FDA revoked approval of the dye Red No. 3 in January 2025, but manufacturers have until January 2027 to reformulate.[52] A proposed "Low / Med / High" front-of-pack label for saturated fat, salt and added sugar is not yet final.[53] A first federal definition of ultra-processed food went for final review in August 2026.
- United Kingdom. The government's nutrition advisers looked again in April 2025 and kept their 2023 verdict: the associations are concerning, but it is unclear whether processing itself or the calories, saturated fat, salt and sugar are to blame, so advice stays nutrient-based.[54]
- Elsewhere. Brazil has told people to avoid ultra-processed food since 2014, and six other countries name ultra-processing in their guidelines.[55] The World Health Organization is drafting a guideline that does not yet exist.
- American Heart Association. Its 2025 advisory says mechanisms "remain unclear" and that most UPFs overlap with foods already targeted for saturated fat, sugar and salt — so it does not endorse an independent processing effect — yet concludes that "evidence supports food policies that limit UPF intake while avoiding unintended consequences".[56]
Further reading
- Lane MM et al. Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses. BMJ 2024.[57]
- Fang Z et al. Association of ultra-processed food consumption with all cause and cause specific mortality — population based cohort study. BMJ 2024.[58]
- Liang S et al. Ultra-processed foods and risk of all-cause mortality: an updated systematic review and dose-response meta-analysis of prospective cohort studies. Syst Rev 2025.[59]
- Hall KD et al. Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake. Cell Metab 2019.[60]
- Dicken SJ et al. Ultraprocessed or minimally processed diets following healthy dietary guidelines on weight and cardiometabolic health: a randomized, crossover trial (with three Matters Arising critiques and reply). Nat Med 2025.[61]
- Forde CG et al. Eating rate has sustained effects on energy intake from ultraprocessed diets: a 2-week ad libitum dietary randomized controlled crossover trial. Am J Clin Nutr 2026.[62]
- Monteiro CA et al. Ultra-processed foods and human health: the main thesis and the evidence (Lancet Series, paper 1). Lancet 2025.[63]
- Cardoso BR et al. Association between ultra-processed food intake and biological ageing in US adults: findings from NHANES 2003–2010. Age and Ageing 2024.[64]
- Rondinella D et al. The Detrimental Impact of Ultra-Processed Foods on the Human Gut Microbiome and Gut Barrier. Nutrients 2025.[65]
- Vadiveloo MK et al. Ultraprocessed Foods and Their Association With Cardiometabolic Health: Evidence, Gaps, and Opportunities. Circulation 2025 (AHA science advisory).[66]
- UK SACN — statement on processed foods and health, 2023, and rapid evidence update, April 2025.[67]
- Harvard T.H. Chan — Processed Foods and Health (general-audience overview, last updated 2023).[68]