— June 26, 2026 · Ultra-Processed Food and an Estimated 3.4 Million Deaths a Year: A 45-Country Model

Table of Contents

1. The Study at a Glance
2. What Was Done
3. What Was Found, Number by Number
4. How the Model Turns Diets Into Deaths
5. How Ultra-Processed Food May Raise Risk
6. How It Fits With Earlier Research
7. What It Cannot Show
8. What the Authors and Agencies Said
9. What Remains Unknown
10. Sources and Primary Documents
11. Key Research Papers
12. Connections
13. Featured Videos

The Study at a Glance

• Paper — “Ultraprocessed food consumption and the global burden of noncommunicable disease: a 45-country comparative risk assessment,” American Journal of Clinical Nutrition 124(3):101417. Online June 26, 2026; September 2026 print issue.
• Who — seven researchers from universities in Ho Chi Minh City, Vietnam, and Taipei, Taiwan. The authors report no conflicts of interest.
• Design — comparative risk assessment: a mathematical model, not a trial and not a new cohort of patients.
• Inputs — ultra-processed food intake as a share of calories from 47 national or regional diet surveys; relative risks per 10% of calories taken from published meta-analyses; the WHO Global Health Estimates for 2021.
• Outcomes — eight noncommunicable diseases, with type 2 diabetes chosen in advance as the main outcome, plus all-cause deaths.
• Threshold — intake above 10% of calories was counted as excess.
• Range of intake — 4.0% of calories (Ethiopia) to 57.5% (United States).
• Main results — 41.4% of type 2 diabetes in the United States attributed to ultra-processed food; an estimated 3.4 million deaths a year (1.3 to 6.0 million) and 5.67 million years of healthy life lost to type 2 diabetes each year across the countries modelled.
• Pattern — the estimated burden was concentrated in high-income countries.

Back to Table of Contents

What Was Done

The study rests on the NOVA classification, devised by Brazilian researchers led by Carlos Monteiro. NOVA sorts foods not by their nutrients but by how much, and for what purpose, they have been industrially processed. It has four groups: unprocessed or minimally processed foods, processed culinary ingredients, processed foods, and ultra-processed foods (NOVA group 4). A 2019 guide by the NOVA authors describes ultra-processed foods as industrial formulations made by breaking whole foods into substances, chemically modifying them and reassembling them, frequently with “cosmetic additives.” Its practical test is the ingredient list: a product counts as ultra-processed if it contains at least one item “never or rarely used in kitchens (such as high-fructose corn syrup, hydrogenated or interesterified oils, and hydrolysed proteins),” or additive classes such as flavours, flavour enhancers, colours, emulsifiers, sweeteners and thickeners.

The team gathered the share of daily calories that came from NOVA group 4 foods in 47 dietary surveys. Together these made 47 country or regional estimates covering high-, upper-middle-, lower-middle- and low-income economies. For 45 of them, the WHO’s 2021 Global Health Estimates held matching cause-of-death and disease data, so the full burden calculation was done for those 45. That is where the “45-country” in the title comes from.

For each disease, the authors took a dose-response relative risk from published meta-analyses: how much the risk of that disease rises for every additional 10% of calories from ultra-processed food. They then set a theoretical minimum risk exposure level of 10% of calories. Intake at or below that level was treated as carrying no extra risk; only intake above it was counted. With those pieces they calculated a population attributable fraction for each disease in each country, the share of cases the model assigns to ultra-processed food intake above the threshold, and applied it to the WHO’s numbers of deaths and of disability-adjusted life years (DALYs). To express uncertainty, the whole calculation was repeated 10,000 times in a Monte Carlo simulation.

Type 2 diabetes was named in advance as the primary outcome. The abstract states that eight noncommunicable disease outcomes were assessed but names only type 2 diabetes and all-cause deaths. The paper’s reference list cites earlier research on type 2 diabetes, cardiovascular disease, mental health, cancer and deaths from all causes. Which eight outcomes were finally included, and the risk figure used for each, are set out in the full text, which was not available for this report.

Back to Table of Contents

What Was Found, Number by Number

Intake. The share of calories from ultra-processed food ranged from 4.0% in Ethiopia to 57.5% in the United States. In other words, by these surveys more than half of what the average American ate came from NOVA group 4 foods.

Type 2 diabetes, by income group. The average share of type 2 diabetes attributed to ultra-processed food fell steeply with national income:

• High-income countries (32 estimates): 20.2%
• Upper-middle-income countries (6 estimates, Lebanon excluded from this average): 14.2%
• Lower-middle-income countries (6 estimates): 0.9%
• Low-income countries (2 estimates): 2.7%

The low-income average rests on only two estimates, which is why it sits slightly above the lower-middle-income figure.

Type 2 diabetes, highest countries. The four highest country figures, each with its 95% confidence interval:

• United States: 41.4% (29.2% to 52.2%)
• United Kingdom: 40.8% (28.4% to 51.6%)
• Netherlands: 36.2% (24.6% to 46.6%)
• Norway: 35.5% (24.2% to 45.5%)

Deaths and DALYs. Across the countries modelled, intake above the 10% threshold “was associated with an estimated 3.4 million all-cause deaths (95% CI: 1.3, 6.0 million)” each year. The precise central figure was 3,362 thousand. For type 2 diabetes alone, the estimate was 5,670 thousand DALYs a year (3,266 thousand to 8,348 thousand), or about 5.7 million years of healthy life lost to early death and to living with the disease.

Moving the threshold. When the authors moved the threshold anywhere between 0% and 15% of calories, the total of attributable deaths ranged from 2,469 thousand to 6,590 thousand, roughly 2.5 to 6.6 million. The abstract describes the results as “stable” across these checks. The range shows that the headline count depends heavily on where the line is drawn, while every version of it stays in the millions.

The authors’ conclusion, in their words: “UPF consumption above a 10% energy threshold is associated with a sizable NCD burden that varies markedly across countries and is concentrated in high-income settings.”

Back to Table of Contents

How the Model Turns Diets Into Deaths

A comparative risk assessment is the same method the WHO and the Global Burden of Disease project have used since the early 2000s to rank tobacco, high blood pressure, alcohol and other risks against one another. It has three moving parts.

1. Exposure — how much of the risk a population has. Here, the average share of calories from ultra-processed food in each country’s survey.
2. Relative risk — how much each step of exposure raises the chance of a disease, taken from meta-analyses of cohort studies. Here, the rise in risk per extra 10% of calories.
3. Counterfactual — the comparison world. Here, a world in which nobody’s intake goes above 10% of calories.

From these comes the population attributable fraction (PAF): the share of a disease’s cases or deaths that would not occur in the comparison world, if the relative risk reflects cause and effect. The PAF is then multiplied by the actual number of deaths or DALYs from that disease, taken here from the WHO’s 2021 estimates, to give an “attributable” count.

How the pieces combine. The paper runs this calculation country by country and simulates the uncertainty in every input. Its abstract does not list each country’s inputs, so the description below uses only figures that published papers state.

• A 2025 meta-analysis found that each extra 10% of calories from ultra-processed food was linked to a 3% higher risk of death from any cause (relative risk 1.03). A 2023 meta-analysis found a 12% higher risk of type 2 diabetes per extra 10% of calories.
• The further a country’s average intake sits above the 10% comparison level, the more of these ten-point steps it has, and the larger its attributable share. The survey averages ranged from 4.0% of calories (Ethiopia), below the comparison level, to 57.5% (the United States).
• For the United States, the paper attributes 41.4% of type 2 diabetes (95% CI 29.2–52.2%) to ultra-processed food. The 2025 meta-analysis, using the same kind of method, put the share of premature deaths attributable to ultra-processed food at 14% in the United States and the United Kingdom.

This is why the United States, the United Kingdom and the Netherlands rise to the top: with the same risk per step applied everywhere, the countries that eat the most ultra-processed food carry the largest attributable share. It also shows why the threshold matters: lowering it adds steps, and therefore deaths, for every country.

DALYs add together the years of life lost to early death and the years lived with illness, weighted for how disabling the illness is. The Monte Carlo simulation reruns the calculation many times, each time drawing the uncertain inputs from within their plausible ranges. The spread of the 10,000 results gives the 95% interval: 1.3 to 6.0 million deaths.

Back to Table of Contents

How Ultra-Processed Food May Raise Risk

The burden model does not test any mechanism; it takes the size of the risk as given. Other research proposes several ways a diet built on ultra-processed food could raise the risk of chronic disease:

• Eating more without noticing. In a 2019 US National Institutes of Health inpatient trial, 20 adults ate as much as they wanted from an ultra-processed diet for two weeks and an unprocessed diet for two weeks. The meals offered were matched for calories, energy density, macronutrients, sugar, sodium and fibre. People ate 508 more calories a day on the ultra-processed diet, gained 0.9 kg, and lost 0.9 kg on the unprocessed diet. The trial measured intake and weight over four weeks, not disease; a lasting calorie surplus is the usual route to obesity, itself a major risk factor for type 2 diabetes. The site’s page on ultra-processed foods and weight gain describes the trial in detail.
• Poorer nutrient quality. In US national survey data from 2009–2010 (9,317 people), diets with a larger share of ultra-processed food contained less protein, fibre, vitamins A, C, D and E, zinc, potassium, phosphorus, magnesium and calcium, and more carbohydrate, added sugar and saturated fat.
• Food structure and additives. The first paper of the 2025 Lancet series on ultra-processed food lists overeating “driven by high energy density, hyper-palatability, soft texture, and disrupted food matrices,” lower intake of protective plant compounds, and higher intake of “toxic compounds, endocrine disruptors, and potentially harmful classes and mixtures of food additives.” Some of these are covered on the site’s pages on food additives and industrial seed oils.
• Displacement of whole foods. Each calorie from a NOVA group 4 product is a calorie not taken from unprocessed or minimally processed foods such as vegetables, fruit, beans, whole grains, eggs, meat and fish prepared in the kitchen. The same Lancet paper argues that this displacement of “long-established diets centred on whole foods and their culinary preparation” is the core of the problem.

Which of these matters most, and whether they act together, is not settled. The burden estimate treats ultra-processed food as a single exposure, so it cannot tell them apart.

Back to Table of Contents

How It Fits With Earlier Research

The risk figures. A 2024 umbrella review in The BMJ gathered 45 pooled analyses covering 9,888,373 people and graded how credible each link was. It rated as “convincing” the link between ultra-processed food and cardiovascular disease deaths (risk ratio 1.50, 95% CI 1.37 to 1.63, though the GRADE quality was very low), and the dose-response link with type 2 diabetes (risk ratio 1.12 per step, 1.11 to 1.13, moderate quality). It rated as “highly suggestive” the links with all-cause death (1.21), heart-disease death (1.66), depression (1.22) and obesity (odds ratio 1.55). Most of the 45 analyses were rated low or very low quality; only four were moderate.

Type 2 diabetes in particular. A 2023 study of three large US cohorts (198,636 women and men, 19,503 diabetes cases) found a 46% higher risk in the top fifth of ultra-processed food intake than in the bottom fifth. Its meta-analysis of 415,554 people found a 12% higher risk per 10% step, rated high quality. The same study also found that ultra-processed foods are not all alike. Refined breads, sauces and spreads, sweetened drinks, processed animal-based products and ready-to-eat mixed dishes were linked to higher risk. Some other subgroups, including cereals and dark and whole-grain breads, were linked to lower risk.

Earlier burden models. The June 2026 paper extends a line of single-country and several-country models:

• Brazil, 2023. With ultra-processed food at 13% to 21% of calories, about 57,000 deaths among adults aged 30 to 69 in 2019 (95% uncertainty interval 33,493 to 82,570) were attributed to it. That is 10.5% of premature deaths at those ages.
• Eight countries, 2025. A new meta-analysis gave a relative risk of 1.03 per 10% step for death from any cause. The attributable share of premature deaths ranged from 4% in Colombia to 14% in the United Kingdom and the United States.
• Belgium, 2025. 23% of type 2 diabetes cases, 24% of diabetes DALYs and 21% of obesity cases were attributed to ultra-processed food.

The new paper’s diabetes shares for high-income countries (20% on average, around 40% at the top) sit in the same range as the Belgian estimate. Its contribution is scale: 45 countries in one framework, with low- and middle-income countries that earlier models left out.

The wider debate. In November 2025 The Lancet began a three-part series whose first paper concluded that the displacement of traditional diets by ultra-processed food “is a key driver of the escalating global burden of multiple diet-related chronic diseases,” resting on “more than 100 prospective studies, meta-analyses, randomised controlled trials, and mechanistic studies.” The June 2026 model starts from that body of evidence and asks how large the burden would be if those associations are causal.

An older thread runs through the site as well. In 1942 the German hygienist Werner Kollath sorted foods into six “value grades,” from fresh and unaltered to isolated preparations, in his book Die Ordnung unserer Nahrung, decades before NOVA brought a processing-based grouping into survey research.

Back to Table of Contents

What It Cannot Show

The authors’ own list of limitations is in the full text, which could not be read for this report. The points below are general limits of this kind of model, which apply whatever the full text says.

• These are modelled deaths, not counted ones. The abstract says intake was “associated with an estimated” 3.4 million deaths. No death record lists ultra-processed food as a cause. The figure is a calculation of how many deaths the published risk figures would account for.
• Cause and effect is assumed, not tested. The relative risks come largely from observational cohort studies. People who eat a lot of ultra-processed food differ from those who eat little in income, smoking, activity and much else. Cohort studies adjust for what they measure, but leftover confounding can remain. A burden model inherits whatever bias sits in its risk figures. The one controlled feeding trial cited above lasted four weeks in 20 people and measured calories and weight, not disease.
• One risk figure for everyone. The model applies the same risk per 10% step in every country. Yet what counts as ultra-processed in one country’s diet, and what it replaces, can differ greatly from another’s. The 2023 diabetes study found that some ultra-processed subgroups were linked to lower risk, not higher.
• A straight-line dose-response. Risk is assumed to rise by the same proportion with every 10% step, from just above the threshold to more than half of all calories.
• The threshold is a choice. Setting “minimum risk” at 10% of calories is a judgement. The authors’ own checks show that moving it between 0% and 15% moves the death total between about 2.5 and 6.6 million.
• Survey quality varies. The 47 surveys were carried out in different years with different methods. Classing reported foods into NOVA groups involves judgement, and dietary recall is known to be imperfect.
• Wide uncertainty. The 95% interval for deaths runs from 1.3 million to 6.0 million, a nearly five-fold range.
• Little outside scrutiny so far. No press release, independent commentary or news coverage of this paper was found as of the date of this report.

Back to Table of Contents

What the Authors and Agencies Said

The authors. The abstract closes: “The income gradient points to an opportunity for early preventive food policy in countries undergoing the nutrition transition.” The “nutrition transition” is the shift from home-cooked traditional diets towards packaged industrial food that typically follows rising incomes. On this reading, low- and middle-income countries, where intake is still low, have the most to gain from early policy. The authors report no conflicts of interest.

United States regulators. No agency statement about this particular paper was found. The paper appeared while US agencies were working on the question. On July 23, 2025, the FDA and the US Department of Agriculture asked for public input on a uniform federal definition of ultra-processed foods (“Ultra-Processed Foods; Request for Information,” 90 FR 35305, published July 25, 2025, Docket No. FDA-2025-N-1793). A second notice (90 FR 45229, September 19, 2025) extended the comment period to October 23, 2025. On January 7, 2026, the Department of Health and Human Services and the Department of Agriculture released the 2025–2030 Dietary Guidelines for Americans, which take direct aim at ultra-processed food. The June 2026 paper cites those guidelines.

No international verdict yet. No statement from the WHO or another international body on this paper was found.

Back to Table of Contents

What Remains Unknown

• Which outcomes, which risks. The full list of the eight diseases, and the risk figure behind each, sit in the paywalled full text. So do the death counts by disease.
• Which part of ultra-processed food matters. Whether the risk comes from additives, from refined starch, sugar and fat, from soft, energy-dense textures that speed eating, from what these foods push out of the diet, or from all of these, is still being studied.
• Whether cutting intake lowers disease. The causal step in the model rests on cohort studies and short feeding trials. None of the research cited here is a long-term randomised trial testing whether eating less ultra-processed food lowers rates of diabetes, heart disease or death.
• Where the true minimum lies. Whether there is a level of intake with no extra risk, and whether 10% of calories is that level, is not known.
• The countries without data. The 45 countries modelled are a minority of the world’s countries, so the 3.4 million is a total for those countries rather than a full global count.
• The direction of travel. If intake in lower-income countries rises toward high-income levels, the model implies their burden would grow too. That forecast has not been tested.

Back to Table of Contents

Sources and Primary Documents

Every figure attributed to the June 2026 study comes from its PubMed abstract, its conflict-of-interest declaration and its Crossref publisher record. The full text is subscription-only and was not read.

1. Nguyen TA, Nguyen NN, Nguyen HS, Linh TTD, Nguyen NTK, Hoang YN, Ho DKN (2026). Ultraprocessed food consumption and the global burden of noncommunicable disease: a 45-country comparative risk assessment. American Journal of Clinical Nutrition 124(3):101417. Epub June 26, 2026. — doi:10.1016/j.ajcnut.2026.101417 · PubMed PMID: 42362116
2. Food and Drug Administration and Department of Agriculture (July 25, 2025). Ultra-Processed Foods; Request for Information, 90 FR 35305, FR Doc. 2025-14089, Docket No. FDA-2025-N-1793. — Federal Register notice
3. Monteiro CA, Cannon G, Levy RB, Moubarac JC, Louzada ML, Rauber F, et al. (2019). Ultra-processed foods: what they are and how to identify them. Public Health Nutrition 22(5):936-941. The NOVA definition used by the study. — PubMed PMID: 30744710

Back to Table of Contents

Key Research Papers

Background papers on the risk figures, the method and earlier burden estimates. Each was checked against PubMed by its DOI, and its abstract was read before it was cited.

1. Hall KD, Ayuketah A, Brychta R, Cai H, Cassimatis T, Chen KY, et al. (2019). Ultra-processed diets cause excess calorie intake and weight gain: an inpatient randomized controlled trial of ad libitum food intake. Cell Metabolism 30(1):67-77.e3. — PubMed PMID: 31105044
2. Lane MM, Gamage E, Du S, Ashtree DN, McGuinness AJ, Gauci S, et al. (2024). Ultra-processed food exposure and adverse health outcomes: umbrella review of epidemiological meta-analyses. BMJ 384:e077310. — PubMed PMID: 38418082
3. Chen Z, Khandpur N, Desjardins C, Wang L, Monteiro CA, Rossato SL, et al. (2023). Ultra-processed food consumption and risk of type 2 diabetes: three large prospective U.S. cohort studies. Diabetes Care 46(7):1335-1344. — PubMed PMID: 36854188
4. Nilson EAF, Ferrari G, Louzada MLC, Levy RB, Monteiro CA, Rezende LFM (2023). Premature deaths attributable to the consumption of ultraprocessed foods in Brazil. American Journal of Preventive Medicine 64(1):129-136. — PubMed PMID: 36528353
5. Nilson EAF, Delpino FM, Batis C, Machado PP, Moubarac JC, Cediel G, et al. (2025). Premature mortality attributable to ultraprocessed food consumption in 8 countries. American Journal of Preventive Medicine 68(6):1091-1099. — PubMed PMID: 40293384
6. Gutierrez-Ortiz C, Guariguata L, Dénos C, Peñalvo JL, Vandevijvere S (2025). Impact of ultra-processed foods consumption on the burden of obesity and type 2 diabetes in Belgium: a comparative risk assessment. BMC Public Health 25(1):1097. — PubMed PMID: 40121412
7. Monteiro CA, Louzada ML, Steele-Martinez E, Cannon G, Andrade GC, Baker P, et al. (2025). Ultra-processed foods and human health: the main thesis and the evidence. Lancet 406(10520):2667-2684. — PubMed PMID: 41270766
8. Martínez Steele E, Popkin BM, Swinburn B, Monteiro CA (2017). The share of ultra-processed foods and the overall nutritional quality of diets in the US: evidence from a nationally representative cross-sectional study. Population Health Metrics 15(1):6. — PubMed PMID: 28193285
9. Ezzati M, Lopez AD, Rodgers A, Vander Hoorn S, Murray CJ; Comparative Risk Assessment Collaborating Group (2002). Selected major risk factors and global and regional burden of disease. Lancet 360(9343):1347-1360. — PubMed PMID: 12423980

PubMed Topic Searches

1. PubMed: Ultra-processed food and comparative risk assessment
2. PubMed: Ultra-processed food, attributable fraction and mortality
3. PubMed: Ultra-processed food and type 2 diabetes, dose-response
4. PubMed: NOVA classification in dietary surveys
5. PubMed: Nutrition transition in low- and middle-income countries

Back to Table of Contents

Connections

• News Archive
• Ultra-Processed Foods (UPF) — the NOVA classification and the research on NOVA group 4
• Ultra-Processed Foods and Weight Gain — the 2019 inpatient trial and why these foods are easy to overeat
• FDA and USDA Seek a Definition of Ultra-Processed Foods — the 2025 federal request for information
• New Dietary Guidelines Take Aim at Ultra-Processed Food — the 2025–2030 guidelines, released January 2026
• Food Additives — emulsifiers, colours, sweeteners and other additive classes
• Processed Seed Oils — industrial oils in the food supply
• Type 2 Diabetes — the study’s primary outcome
• Obesity — causes, health effects and research
• Metabolic Syndrome — the cluster of risks that precedes diabetes and heart disease
• Cardiovascular Disease — heart and blood-vessel disease
• Coronary Artery Disease — the most common form of heart disease
• Depression — one of the outcomes linked to ultra-processed food in the 2024 umbrella review
• Kollath’s Order of Our Food — an early classification of food by degree of processing
• Food Preservatives and High Blood Pressure — a later news report on additives and health
• Beneficial Foods — the site’s section on whole foods

Back to Table of Contents

Featured Videos