Testing matched meals in a metabolic chamber and brain scanner
The study involved 57 healthy-weight adults aged 18 to 45, with an average body mass index of 22.8. In a randomised crossover design, 32 participants completed both test meals on separate mornings. Each meal provided about 300 calories and was matched on a dozen nutritional measures, including total calories, sugars, fats, sodium, fibre and glycaemic index, to within 1.6 percent.
Participants consumed each meal in a whole-room indirect calorimeter to measure metabolic rate and fuel use, and blood samples were taken at several intervals to monitor glucose and insulin. The two meals were designed to be visually similar and weighed the same (about 156 grams each). One was composed of minimally processed items such as hard-boiled eggs, banana, dried cranberries, cheese and sea salt; the other used ultraprocessed components including deli-style turkey, instant mashed potatoes, frosted wheat cereal, veggie chips, a chocolate chip cookie and white bread with peanut butter and jelly, plus water.
In a separate set of sessions, participants evaluated photographs of 28 familiar foods (14 ultraprocessed and 14 non-ultraprocessed) that had been matched for visual features and nutrients in prior work. On a separate day, after fasting, they underwent functional magnetic resonance imaging while viewing each image, then bid in a Becker-DeGroot-Marschak auction to indicate how much they would pay for each item. The researchers also asked them to estimate typical prices and calories for the foods.
Only a subset had complete data for both metabolic and brain measurements: usable brain scans were obtained from 52 participants, including 29 who also had full metabolic data.
For the baseline metrics, including resting energy expenditure and fasting glucose and insulin, there were no significant differences between the two meal conditions at the start of the visits. Eating the ultraprocessed meal took slightly longer on average (about 1.5 minutes more) than the non-ultraprocessed meal, but post-meal ratings of hunger, fullness or thirst did not differ significantly between meals.
Insulin spikes, fat burning and brain signals
Both meals produced similar overall glucose responses over three hours, with peaks around 40 minutes for each. However, glucose rose faster after the non-ultraprocessed meal, reaching higher levels at 20 minutes, while after the ultraprocessed meal, glucose tended to stay elevated at two to three hours rather than returning to baseline as quickly.
By contrast, the ultraprocessed meal triggered a substantially larger cumulative insulin surge. Insulin levels were similar during the first 20 minutes, but rose higher after 40, 60, 90 and 120 minutes compared with the non-ultraprocessed meal, before both conditions returned toward baseline around three hours.
In the calorimeter, total calories burned over the three hours were higher after the ultraprocessed meal, though the pattern showed a slower fall after peaking around 40 minutes. The respiratory exchange ratio was lower following the ultraprocessed meal, indicating reduced carbohydrate burning and greater fat burning, a sign that the meal altered how the body switches fuel sources post-meal.
DiFeliceantonio highlighted the insulin finding as particularly surprising: “The insulin result was the most surprising.”
In the brain analyses, the strongest relationships emerged in regions tied to reward valuation and sensory processing. In the subset with both metabolic and brain data, carbohydrate burning correlated with activity in the left superior temporal gyrus during exposure to food cues. Within the striatum, a region linked to reward and motivation, carbohydrate burning aligned with activity in the right caudate and left ventral striatum, and the same pattern appeared at two points in the right caudate for the fuel-burning balance.
Across these regions, participants whose carbohydrate burning was most blunted by the ultraprocessed meal showed stronger activity when viewing ultraprocessed food images compared with non-ultraprocessed images. This inverse relationship contrasts with earlier studies that scanned people while tasting sugary drinks rather than viewing pictures of foods.
Importantly, the researchers cautioned that, despite the clearer insulin response, differences in peak insulin between meals did not predict brain responses to the food images. Nor did differences in peak metabolic rate or blood glucose achieve statistical significance predictors once other variables were considered.
For the broader food-valuation analyses, which drew on all 52 brain scans, subjective ratings of healthiness or how often foods were eaten did not fully explain bidding differences. After accounting for health perceptions and consumption frequency, bids for ultraprocessed items did not differ significantly from those for non-ultraprocessed items, although activity in classic valuation regions like the orbitofrontal cortex and anterior cingulate reflected overall bid levels.
When researchers examined how brain activity scaled with bid amounts, they found distinct patterns in the visual processing areas rather than the primary reward centres. In the left fusiform and lingual gyri, activity rose with higher bids for non-ultraprocessed foods but declined for ultraprocessed ones; similar effects appeared in the right putamen and right caudate within the striatum. No brain region showed stronger responses to ultraprocessed bids than to non-ultraprocessed bids in the analyses conducted.
Implications for metabolism, policy and future work
Commenting on the broader significance, a researcher unaffiliated with the study said there is now convincing evidence that ultra-processing matters beyond caloric and nutrient content. “There is a lot of criticism of the concept of ultra-processed foods… what this paper shows quite convincingly is that even if the foods are extremely well-matched for all of those factors, there are still differences in metabolic and brain responses compared to non-UPFs,” said Filip Morys of McGill University. “So, the mere fact that those foods are ultra-processed – might contain some additives or undergo processing in extreme temperatures, etc. – changes how our bodies react to them.”
The Virginia Tech team notes that industrial methods used to enhance texture and mouthfeel may alter digestion kinetics and how nutrients are delivered to the upper gut. They emphasise that the present study was not designed to evaluate long-term health outcomes, but to expose immediate metabolic and neural responses to meals differing only in processing level.
DiFeliceantonio also stressed that the work is a step toward understanding mechanism rather than prescribing policy. “The study was not designed to assess chronic outcomes… This gives us some clues. The insulin response is high, the shift to carbohydrate oxidation is slower, and that might help us understand why they lead to poor metabolic health outcomes.”
On policy and definitions, the researchers noted that a working definition of ultra-processed foods exists in science, but public policy and labeling definitions remain unsettled. “The idea that we don’t have a working research definition of ultra-processed foods. We do. What we don’t have is a policy and labeling definition,” said Morys.
DiFeliceantonio urged caution in interpreting the brain findings, given the smaller brain-imaging sample and the indirect nature of comparing metabolic responses to brain activity recorded on a separate day. She suggested future work could bring the two measures together in more naturalistic settings, such as imaging during actual ultra-processed food consumption rather than photo viewing, and across broader populations and meal sizes.
The study, titled Metabolic and neural responses to ultraprocessed foods: a randomized, controlled, crossover study, lists Zach Hutelin, Monica Ahrens, Mary Elizabeth Baugh, Emmanuel Nartey, Delbert L. Herald III, Alexandra L. Hanlon and Alexandra G. DiFeliceantonio as authors.
