Why Ultraprocessed Foods Drive Weight Gain — and Why Kevin Hall Left the NIH
Kevin Hall ran the first controlled trial showing that people eat about 500 extra calories a day on an ultraprocessed diet — even when the diets are matched for sugar, fat, salt and fiber, and even when both diets are rated equally pleasant. His follow-up work points to two properties as the likely drivers: calorie density and hyperpalatable foods. He also explains why he took early retirement from the NIH after his research communication was repeatedly blocked.
Overview
Kevin Hall spent two decades at the NIH running metabolic ward studies — participants living on-site for a month, every gram of food weighed, every leftover measured. His landmark ultraprocessed food trial gave people double the calories they needed and simple instructions: eat as much or as little as you like. On the ultraprocessed menu they spontaneously ate roughly 500 extra calories a day and gained weight and body fat; on the minimally processed menu they lost both.
Strikingly, they rated the meals equally pleasant and reported similar hunger and fullness — so taste alone does not explain the effect. Follow-up analysis pointed to energy density and the number of hyperpalatable foods on the plate as the strongest correlates of how much people ate, and a new four-arm trial was designed to separate the two. Hall also unpacks the NOVA classification: where it came from, why it is more about formulation than processing in practice, and why he thinks healthy ultraprocessed foods deserve serious attention rather than being treated as an oxymoron.
On the carbs-versus-fat question, his chamber studies found that a calorie is very nearly a calorie — low-carb diets showed no metabolic advantage, which does not mean they cannot work for individuals. He closes with the story of blocked talks, a killed press release, edited answers attributed to him, and the decision to retire rather than accept interference in how science is communicated.
Key quotes
5Are we really listening to the science, or are we only interested in purveying the preconceived narrative?
They ate about 500 calories per day on average more than the same people when they were exposed to the minimally processed food environment.
Either people are lying, or you have to take the data at face value.
We actually have to start seriously thinking about what do healthy ultraprocessed foods look like.
Science gets in the way of a good narrative.
Key ideas
9The effect was real before the mechanism was known
The first trial showed a large difference in calorie intake without isolating any single cause. Everyone read their favourite theory into the result, but only a new study varying one property at a time can settle it.
Where the ultraprocessed concept came from
Carlos Monteiro noticed Brazilians were buying less sugar and less oil while obesity kept rising. The sugar and oil had simply moved inside pre-prepared industrial foods, which suggested a new lens on the food supply.
NOVA in practice describes formulation, not processing
Manufacturers are not required to disclose how a food is made, so in practice the ingredient list becomes the marker. That is why the label is a little squishy and why edge cases are unavoidable.
Ultraprocessed is not automatically unhealthy
Whether a given product harms you is an empirical question, not a definitional one. A jarred sauce that makes a whole-grain, plant-rich meal happen more often can be a good addition to a real life.
Not all subcategories behave the same way
In observational data, sugar-sweetened beverages and processed meats sit consistently on the harmful side. Sweets and desserts land near neutral and ultraprocessed breads and cereals even lean favourable, which Hall treats as intriguing rather than settled.
The metabolic ward is the only way to measure intake honestly
Self-reported intake underestimates calories, and even take-home study food leaks by hundreds of calories a day. Housing people for a month and weighing every leftover to a tenth of a gram is the price of a clean number.
Liking and wanting are different systems
Both menus were rated equally pleasant, so conscious pleasure is not the driver. A gut-brain nutrient-sensing route operating below awareness appears to shape how much gets eaten in the hours around a meal.
Energy density and hyperpalatability stand out
Across every diet pattern tested, meals that were more calorie dense and contained more hyperpalatable items were eaten in larger amounts. The same coefficients held on minimally processed menus, suggesting general properties rather than processing magic.
A calorie is very nearly a calorie
Cutting carbs or cutting fat by the same number of calories produced almost identical fat loss in chamber studies, with no metabolic advantage for low carb. Where low-carb diets work, appetite is the more likely explanation than fat trapping.
Practical takeaways
7- 1
Judge the product, not the category 36:10
Instead of banning a whole label from your kitchen, look at what a convenience product actually enables. If it makes a vegetable-rich, whole-grain meal happen on a Tuesday night, it is earning its place.
- 2
Start with the two consistent offenders 43:00
Sugar-sweetened beverages and processed meat products show up at the top of the harm list again and again. Reducing those two is the least controversial move available.
- 3
Watch calorie density before you watch labels 1:22:00
How many calories sit in the same volume of food was one of the strongest predictors of how much people ate. Building meals around water-rich, bulky foods lowers intake without any willpower story.
- 4
Notice combinations, not single nutrients 1:24:00
Hyperpalatable items are defined by pairs of nutrients crossing thresholds together — fat plus sugar, fat plus salt, salt plus carbs. Spotting those pairings on your plate is more useful than counting one nutrient.
- 5
Eating speed matters 1:10:00
The faster people ate a meal, the more calories they took in from it. Slowing the meal down is a lever available at every table, with no shopping list attached.
- 6
Life circumstances may beat macronutrients 1:42:00
The people who struggled most often had a job loss, a separation or another disruption rather than the wrong macros. Stability, support and time to plan deserve as much attention as the diet name.
- 7
Stop hunting for the one magic diet 1:35:00
Low-carb and low-fat approaches produced similar fat loss on average, and the predictors of who does better did not replicate. Pick the pattern you can actually live with and stop treating it as a moral position.
Topics & chapters
16A researcher walks away
Kevin Hall announces early retirement from the NIH and describes blocked interviews and a quashed press release.
The effect without a mechanism
Why the first trial showed a strong result that nobody could confidently explain.
How the concept was born in Brazil
Carlos Monteiro's observation about falling sugar and oil purchases alongside rising obesity.
The four NOVA categories
From minimally processed foods and culinary ingredients to processed and ultraprocessed.
What the definition gets wrong
Marketing and profit motives inside a food definition, and why that irritates a reductionist scientist.
Can an ultraprocessed food be healthy?
Why the original framing treated the idea as an oxymoron, and why Hall disagrees.
Which subcategories actually track with harm
Beverages and processed meat versus the puzzling neutrality of sweets and desserts.
Inside the metabolic ward
A month of residency, double the calories offered, and every leftover weighed in the basement.
Five hundred extra calories a day
Spontaneous weight and fat gain on one menu, loss on the other, with matched nutrients.
Equally pleasant, unequally eaten
The pleasantness ratings that surprised everyone, and the liking-versus-wanting distinction.
Why these foods win in the marketplace
Cheap commodity inputs, constant reformulation and a selection process that rewards repeat purchase.
Energy density and hyperpalatability
The post-hoc analysis that identified two candidate drivers and the four-arm trial built to test them.
The carbs versus fat wars
Chamber studies, computer models, and the finding that low carb offers no metabolic advantage.
Why the same diet works for one person and not another
Life disruption, social support and stability as underrated predictors of success.
The censorship story
A denied talk, a killed press release, edited answers attributed to him, and an unanswered letter.
What happens next
A handshake, a plan for better nutrition facilities, and the call that ended it.