A majority of the calories Americans eat now come from ultra-processed foods. In its August 2025 data brief, the National Center for Health Statistics put the figure at 55.0% of total calories for everyone aged 1 and older during August 2021–August 2023, and 61.9% for children and adolescents aged 1–18, against 53.0% for adults 19 and older. Sandwiches and burgers, sweet bakery products, savory snacks and sweetened beverages were among the top calorie sources.

Those numbers explain why “how bad is this?” has become one of the loudest questions in nutrition. The past few weeks have produced two very different kinds of answer, and it is worth keeping them apart: a very large tally of observational data, and the first serious attempt to test the question experimentally.

The biggest count yet — with the researchers’ own caveat

On 22 September, a dose-response meta-analysis in Family Medicine and Community Health pooled 51 prospective cohort studies covering 8,819,894 adults. Higher ultra-processed food (UPF) intake was associated with higher risk of cardiovascular events (hazard ratio 1.24, 95% CI 1.14–1.36), cancer (1.12, 1.04–1.21), overweight and obesity (1.23, 1.17–1.30), metabolic syndrome and diabetes (1.24, 1.08–1.42), depression and anxiety (1.27, 1.15–1.40), digestive diseases (1.31, 1.16–1.48) and all-cause mortality (1.18, 1.10–1.26). Each additional 100 g per day carried small additional risk, the authors reported.

They graded that body of evidence “low or moderate” — their words, not a critic’s. Every included study was observational. UPF intake was estimated from food-frequency questionnaires, 24-hour recalls or diet histories, and people who eat differently differ in other ways that statistics can only partly strip out. A hazard ratio of 1.24 describes a population pattern; it is not a personal prediction.

Where the pattern bends

Two of the newest studies complicate the headline. In a prospective analysis from the Alpha Omega Cohort, published in Current Developments in Nutrition, researchers followed 4,365 people who had already had a heart attack for a median of 14 years, during which 1,012 died of cardiovascular causes. A high-UPF diet was not consistently or significantly linked to cardiovascular or coronary heart disease mortality — but each 10-percentage-point increase in the share of energy from UPF was associated with higher stroke mortality (1.21, 1.02–1.42). The Dutch team noted that the high-UPF diets in their cohort were also lower in whole, cardioprotective foods, which is one way a “processed food” effect could really be a “not enough vegetables” effect — the ground our archive piece on signs of low vegetable intake covers.

A much smaller study, in Nutrients, tracked 114 US infants across complementary feeding. Ultra-processed foods rose from 9 to 20 kcal per kilogram per day between 9 and 12 months of age, reaching 30% of total energy at 9 months and 38% at 12. Higher UPF intake tracked with higher added sugar and sodium intake. The cohort is small and the findings are associations, not causes, but they show the category arriving early — and the authors point to savory snacks and toddler bars as frequent contributors.

The experiments are only now being designed

The more telling news is methodological. The protocol for ULTRA-CGM, a randomised controlled trial in 50 healthy Korean adults aged 20–39 who already get at least 25% of their energy from UPF, was published in BMJ Open on 6 October and registered as NCT07175701. Participants are randomised to a single 40-minute session on reducing UPF intake, or to standard dietary counselling based on national guidelines, with 10 days of continuous glucose monitoring before and after. The primary outcome is postprandial glucose response.

That is a protocol, not a result, and 50 people monitored for ten days cannot say anything about cancer, stroke or longevity — only about short-term glucose. But it points at the gap that has defined this field: after thousands of papers, almost no randomised trials have tested whether changing UPF intake itself changes health. Part of the difficulty is structural — you cannot blind people to what they eat, and diets drift over the years a disease endpoint would need. Still, the evidence hierarchy has a hole where experiments should be, and the cohort studies above call for trials in their own conclusions.

“Ultra-processed” is a category still under construction

Even the definition is unsettled. Harvard’s Nutrition Source notes that the US Department of Agriculture treats washing, cutting, freezing and packaging as processing, which means almost everything in a supermarket qualifies to some degree, while the Nova classification used in this research sorts foods into four groups by purpose and degree of industrial processing.

A January 2026 review in BMJ Nutrition, Prevention & Health examined 14 studies that separated UPF into food groups rather than treating it as one block. Processed meats, sugar-sweetened and artificially sweetened drinks, and fats, spreads and sauces showed higher risk; breads, biscuits, breakfast cereals and whole grains were associated with lower risk. The review’s authors disclose compensation from the Physicians Committee for Responsible Medicine, a group that promotes plant-based eating — worth knowing when reading their argument that Nova should distinguish animal from plant foods.

A separate NHANES analysis of 47,999 adults, published in the American Journal of Public Health in July, tried to separate processing from nutritional quality. Each 10% of energy from UPF was associated with all-cause mortality (hazard ratio 1.04, 95% CI 1.02–1.07) and prevalent metabolic syndrome (odds ratio 1.07, 1.05–1.09). Adjusting for overall diet quality attenuated those associations only partly; adjusting for saturated fat, added sugar or sodium alone barely moved them. The signal is not purely about a handful of nutrients — but it is not yet a clean verdict on processing either.

What guidance actually says

The World Health Organization’s healthy-diet fact sheet does not ban a category. It holds that a variety of minimally processed and unprocessed foods low in unhealthy fats, free sugars and sodium are the foundation of a healthy diet, and it sets limits: salt under 5 g a day (2 g sodium), saturated fat under 10% of energy, trans fat under 1%, and at least 400 g of fruit and vegetables a day for anyone over 10.

A narrower takeaway than the internet version

That is the honest shape of it. As currently defined, the UPF category swallows soda, ham, packaged pastries, protein bars and sliced whole-grain bread, and the research does not implicate them equally. The associations are strongest and most consistent for the items nutrition guidance long ago flagged — sweetened drinks, processed meats, salty and sugary packaged snacks. If anything has been tested and repeated, it is that.

None of this is medical or dietary advice, and it is not a reason to overhaul a diet on the strength of an observational hazard ratio. Anyone managing a health condition, or considering a significant change, is better served by a conversation with a clinician or a registered dietitian. The trials that would sharpen these answers have only just been designed.