Ultra-processed food

What it is

Ultra-processed food is a category in the NOVA classification, which sorts foods by the extent and purpose of industrial processing [1]. The category covers foods such as breakfast cereals, sweets and mass-produced bread [2]. In the UK, more than half of average energy intake is reported to come from ultra-processed food [2], and participants in one recent trial started at 67.4% of calories a day [2]. Several alternative classification systems have been proposed, and their definitions differ [1].

How it might work

Proposed explanations for the overeating seen in trials include high energy density, soft textures and fast eating rates [3]. In a trial of 69 adults eating ultra-processed sandwich lunches, a slow-eating, low-energy-density meal led to 573 fewer calories than a fast, high-energy-density one [3]. A Japanese crossover trial found lower chewing frequency per calorie on the ultra-processed diet, alongside 813.5 extra kcal a day [4]. Researchers have not settled which features of processing matter most [1].

What the research says

The strongest direct evidence comes from controlled feeding trials, where everyone eats only study meals [5]. In the best-known, 20 adults stayed at the NIH Clinical Center and ate an ultra-processed or unprocessed diet for two weeks each, in random order; meals were matched for presented calories, energy density, macronutrients, sugar, sodium and fibre, and people ate as much as they wanted [6]. They ate 508 kcal a day more on the ultra-processed diet and gained 0.9 kg, losing 0.9 kg on the unprocessed diet [6]. A 2024 crossover trial in nine overweight Japanese men found 813.5 extra kcal a day and 1.1 kg more weight gain over one-week periods [4]. In the 2025 UPDATE trial, 55 adults in England ate ultra-processed or minimally processed diets built on the UK Eatwell Guide for eight weeks each: weight fell on both, but 1.01 percentage points less on ultra-processed, about 0.96 kg, with roughly 1 kg less fat-mass loss and higher triglycerides on the ultra-processed diet [2]. A 2024 review of randomised trials (four studies, 455 participants) found most outcomes unchanged, with lower energy intake and body weight when ultra-processed food was reduced in the one controlled feeding trial, graded moderate certainty [5]. In a 54-hour whole-room calorimeter trial of 21 young adults, a protein-enriched ultra-processed diet cut intake by about 196 kcal a day versus a standard ultra-processed diet but did not prevent overeating [7]. A 2025 crossover trial found higher body weight and a higher LDL:HDL ratio on ultra-processed food regardless of calorie load [8].

For long-term disease the evidence is observational [9]. A 2024 BMJ umbrella review pooled 45 meta-analyses covering 9,888,373 people; higher intake was linked to 32 of 45 health outcomes [9]. The strongest links, classed as convincing by the authors, were 50% higher cardiovascular disease mortality and 12% higher type 2 diabetes risk per 10% increase in dietary share, the latter rated moderate quality [9]. Suggestive links included 21% higher all-cause mortality, 66% higher heart-disease mortality, 55% higher obesity and 22% higher depression [9]. A second umbrella review updated 39 meta-analyses with 122 studies: kidney-function decline and wheezing in children and adolescents were convincing, and diabetes, overweight, obesity, depression and common mental disorders were highly suggestive [10]. It found no study reporting a beneficial association [10]. A third umbrella review of 16 meta-analyses found moderate-certainty evidence linking each 50 g a day of ultra-processed food to higher death and cardiovascular risks [11].

Individual meta-analyses put numbers on several of those links: across 18 cohorts with 1,148,387 people and 173,107 deaths, the highest intake was linked to 15% higher all-cause mortality, while each 10% step up in intake was linked to about 10% higher risk [12]. Across 12 cohorts, the highest intake was linked to 48% higher type 2 diabetes risk, with a steeper rise above about 300 g a day [13]; a pooled analysis of three Korean cohorts and a meta-analysis of 17 prospective studies found 10-11% higher diabetes risk per 10% increment [14]. For depression, a meta-analysis of six prospective cohorts found 32% higher risk with high intake [15], an earlier review found 22% higher prospective risk plus higher cross-sectional odds of mental disorder symptoms [16], and a target trial emulation in 11,192 older adults linked four or more servings a day to 10% higher risk of depressive symptoms [17]. For cardiovascular events, an 11-paper dose-response analysis found 35% higher risk at the highest intake [18].

Doses used in studies

In the feeding trials, the ultra-processed and comparison diets each lasted one to eight weeks [6][4][2]. The NIH trial used two weeks per diet [6]; the Japanese trial one-week periods with a two-week washout [4]; the UPDATE trial eight weeks per diet [2]. In the NIH and Japanese trials the meals offered were matched for calories and macronutrients [6][4]. In UPDATE, both diets were built to follow the UK Eatwell Guide [2]. The cohort evidence concerns usual diets, with participants typically grouped by the share of daily calories from ultra-processed food [9]. These are descriptions of study diets, not recommendations.

Safety and interactions

Side effects: in the UPDATE trial, mild gastrointestinal side effects were common on both an ultra-processed and a minimally processed diet, and the trials reviewed here did not report serious harms [2][5]. Upper limit: no upper limit or toxic dose is defined for ultra-processed food as a category; the concern in the evidence is long-term disease risk with high intake [9]. Medicine interactions: the studies read did not report interactions between ultra-processed food and medicines. Pregnancy and breastfeeding: no randomised trial data in pregnancy were found; one review reported that an educational programme in overweight pregnant women did not reduce ultra-processed food intake, so any benefits could not be attributed to that change [5]. Breastfeeding was not covered by the studies read. People with conditions such as diabetes, heart disease or reduced kidney function: the higher risks in the cohort studies cover these outcomes, so it is worth discussing diet changes with a doctor or dietitian [10].

Caveats and open questions

The main open question is whether the processing itself causes disease or whether ultra-processed intake is a marker of an overall poor diet; the 2025 umbrella review said this remains unclear [11]. The NOVA category mixes foods with different links to health: in one meta-analysis, processed meats were linked to higher diabetes risk while ultra-processed cereals, breads and packaged savoury snacks were not [13]. Definitions are debated - a review of six classification systems found none that clearly specifies which processing techniques matter [1]. Trials have measured weight, calorie intake and blood markers over weeks, not disease over years [5]. Only four of the 45 pooled analyses in the BMJ review were rated moderate quality, with the rest low or very low [9]. Registered trials are still testing the short-term metabolic effects of ultra-processed versus minimally processed diets.