"Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain." That is the actual title of the 2019 Cell Metabolism study that built the modern case against ultra-processed food (UPF), written in the flattest, most confident register science allows: cause, not correlate. I have spent a career watching that word get stretched past what a study can hold, so I went back to the trial. Twenty adults were admitted to the NIH Clinical Center for 28 days and fed two weeks of an ultra-processed diet and two weeks of a minimally processed one, in random order, with every meal matched for calories, energy density, macronutrients, sugar, sodium, and fiber, according to the NIH's own account of the trial. On the ultra-processed diet, people ate 508 calories a day more than on the minimally processed one, almost entirely as extra carbohydrate and fat, a gap far too consistent to be chance (P = 0.0001). That is a real number, from a tightly controlled study, and it deserves to be taken seriously.
It is also, according to a growing pile of 2026 commentary, not proof that processing itself did it. The most recent entry is a perspective by Jimmy Chun Yu Louie, a dietetics researcher and associate professor at Swinburne University of Technology in Melbourne, published in the American Journal of Clinical Nutrition this June. The full paper sits behind a paywall I could not clear, but its abstract, readable on PubMed, confirms the core claims: Louie reviews four trials that have directly pitted ultra-processed diets against minimally processed ones, and his complaint, in the abstract's own words, is that "reporting consistently emphasized findings aligning with expectations of harm while downplaying neutral or contradictory evidence." The general problem he is describing, that a comparison diet can differ from the other in more than processing level at once, shows up in at least one of those four trials on its own published methods: the NIH trial matched calories, macronutrients, sugar, sodium, and fiber between its two diets, but it did not set out to match how those diets tasted or felt in the mouth, so palatability and texture, separate from processing, cannot be ruled out as part of what drove the extra eating. His point is not that the 2019 finding is wrong. It is that trial coverage since 2019 has treated an unresolved mechanism as a settled one.
The UK trial that found a one-kilogram gap, and the confounds riding along with it
The trial that triggered the current round of scrutiny is the UPDATE study, led by Samuel Dicken and colleagues at University College London, with Kevin Hall, the same NIH scientist behind the 2019 trial, among the senior authors, published in Nature Medicine in October 2025. The trial's own disclosures are worth noting alongside the findings: Dicken reports book royalties and consultancy income of his own, and senior author Rachel Batterham has been an employee and shareholder of Eli Lilly since 2023 and discloses honoraria and consultancy income from Novo Nordisk and Eli Lilly, both makers of GLP-1 weight-loss drugs with a commercial stake in a drugs-not-diet framing of obesity. Fifty-five adults with overweight or obesity, all habitually eating at least half their calories as UPF, were fed two eight-week diets in random order, both built to follow the UK's Eatwell Guide, Britain's official healthy-eating template, similar in role to the "MyPlate" chart in the United States. One diet ran on minimally processed ingredients; the other ran on reformulated, "healthier" ultra-processed products carrying real nutrition claims, the kind of thing already sitting on supermarket shelves marketed as the responsible choice. Participants lost more weight on the minimally processed diet, 2.06 percent of body weight against 1.05 percent on the ultra-processed one, a gap of about one kilogram over eight weeks (P = 0.024). Read plainly, that looks like proof that ultra-processed food makes people gain weight even when a government's own healthy-eating guide is followed to the letter.
The rest of the trial's own data complicates that reading. Triglycerides, a blood fat tied to heart disease, improved on the minimally processed diet and got worse on the ultra-processed one; LDL cholesterol, the "bad" cholesterol, actually improved more on the ultra-processed diet than on the minimally processed one; and blood pressure showed no real difference between the two arms at all. The trial's own authors wrote that the greater weight loss on the minimally processed diet "did not translate into significant improvements in cardiometabolic risk factors [the blood pressure and cholesterol markers that shape heart disease risk] over the ultra-processed diet, except triglycerides." They also disclosed, in their own limitations section, that the ultra-processed diet was more energy dense, rated as tastier, and delivered in branded packaging carrying health claims that may have shaped how much people thought was an appropriate portion. None of that is processing. It is calories per bite, flavor, and marketing, riding along with a label.
Three letters land in Nature Medicine, each with a different objection
Three separate letters attacked the trial's conclusions. David Ludwig, a Harvard Medical School professor of pediatrics and nutrition, working with Walter Willett, a longtime Harvard epidemiology and nutrition professor, and Mary Putt, a biostatistician at the University of Pennsylvania, argue that the four-week washout between the two eight-week diet periods may not have been long enough to clear carryover effects: the body can take weeks to adjust to a major change in diet, and weeks more to un-adjust, a problem well known in drug trials and, as the same three authors argue at length in a 2025 companion paper devoted to that bias alone, routinely overlooked in nutrition studies. Eric Robinson, a psychologist at the University of Liverpool who studies eating behavior, and Ciaran Forde, who chairs a sensory-science and eating-behavior research group at Wageningen University, argue the diets were simply nutritionally mismatched in ways sufficient to explain the weight gap without invoking processing as a mechanism at all. A third letter, from Zixuan Wang, CEO of ZHT Lab, an artificial-intelligence research firm in Beijing, and Can Peng, an intern there, raised a more granular objection: the trial's own numbers produce two different, separately significant estimates of how much less people ate on the minimally processed diet. The body-composition data implied a gap of 170.4 fewer calories a day, and the trial's self-reported intake data put the same gap at roughly 327 calories a day, both statistically significant (P = 0.005 for each), nearly double one another, two measurements of the same effect inside one dataset that should broadly agree and instead do not. Wang and Peng's own disclosure states their company "has no business or financial interests related to diet or nutrition." Dicken's team published a joint reply to all three; I was not able to read its full text to confirm exactly how they answered each point, but the trial's own limitations section, quoted above, had already conceded several of the same confounds before any of the three letters ran.
Millions of person-years, and a consistent, if imprecise, signal
Robinson widened the argument in June, together with Faidon Magkos, a nutrition researcher at the University of Copenhagen, and Forde, in a Science perspective reviewing five obesity trials on UPF from the US, UK, Denmark, and Japan. Their verdict, as Robinson put it in comments distributed to reporters through the UK's Science Media Centre (a press-briefing service that circulates named scientists' on-record reactions to new research): the available trials give only weak support for an effect that comes specifically from processing, as opposed to the food's basic makeup. Their fix is not to abandon the concern but to redirect it, toward foods that are soft, calorie dense, high in saturated fat and salt, and low in fiber and protein, properties that plausibly drive overeating on their own, whether or not a plant is involved in making them. Responding through that same briefing, Seamus Higgins, an associate professor of food process engineering at the University of Nottingham, agreed the trial evidence for a processing-specific effect is weak, but added that large cohort studies, including recent Lancet analyses, keep finding strong associations between UPF intake and worse outcomes that are "increasingly difficult to dispute." The disagreement is not about whether something is going on. It is about whether the something is processing, or the nutrition profile that keeps showing up wearing processing's clothes.
That cohort evidence is worth naming plainly, because it is the reason this is not an academic argument. A 2024 umbrella review in Clinical Nutrition gave its strongest rating, convincing, to just two outcomes: about 25 percent higher risk of kidney function decline and 42 percent higher risk of childhood wheezing. One tier down, rated highly suggestive rather than convincing, it also found a meaningfully higher risk of type 2 diabetes, obesity, and depression (roughly 23, 26, and 40 percent higher, respectively). A separate umbrella review published via BMJ rated as convincing, its own top tier, roughly 50 percent higher cardiovascular mortality (risk ratio 1.50) tied to high UPF intake, plus a type 2 diabetes risk that climbed step by step with each added increment of UPF in the diet (risk ratio 1.12 per unit). Most individual outcomes still carry that review's own lower-confidence ratings, its authors said so themselves, but the direction is consistent across dozens of cohorts and millions of person-years. Meanwhile, per the CDC's own 2025 analysis of national nutrition data, American adults get 53 percent of their daily calories from ultra-processed food, and American children get nearly 62 percent. This is not a fringe exposure being investigated for theoretical harm. It is most of what most people eat.
Act on the population-level signal the way you would act on any other large, repeated public-health finding: eat more fiber, more whole foods, less of the soft, sweet, effortless stuff, advice that needed no crossover trial to justify it in the first place. What the newest trials cannot yet do is tell you that "ultra-processed" is itself the poison, separate from the calories, the softness, and the marketing that happen to travel with it. Louie's proposed fix, matching future trial arms on energy density, fiber, and macronutrients so processing can be tested on its own, is the kind of unglamorous, expensive study nobody will fund with a press release. Until someone runs it, the honest sentence is not "ultra-processed food causes disease." It is that food built to be eaten fast, in large amounts, past the point of fullness, correlates with disease, and a great deal of ultra-processed food happens to be built exactly that way. That is a real problem. It is just not the same claim.
