A tightly matched meal experiment tested more than calories and macros

The Nature Metabolism randomized, controlled crossover study, published online on October 5, 2026, enrolled 57 healthy-weight adults. In a crossover design, participants could be studied under both meal conditions rather than being assigned to only one.

The researchers created approximately 300-kcal ultra-processed and non-ultra-processed meals using the NOVA processing framework. Importantly, the meals were matched within less than 1.6% for weight, energy, energy density, total carbohydrate, fat, protein, available carbohydrate, glycemic index, glycemic load, fiber, sodium and water.

That level of matching makes the study particularly relevant to a common question: if two meals have similar calories and macros, must the body respond in exactly the same way? This experiment suggests the answer is no, at least for certain short-term physiological measurements.

It does not mean the meals were identical in every respect. As the study design cannot isolate every difference between foods, ingredient identity, food matrix, texture, particle size, additives, digestion and other unmeasured properties could have differed. The trial therefore cannot identify the specific feature responsible for its findings.

Insulin area under the curve was greater after the ultra-processed meal

The detailed metabolic analysis involved a 32-person subset. Participants completed a 50-minute baseline measurement, ate one of the test meals, then underwent three hours of post-meal monitoring. Blood samples were collected at baseline and at 5, 20, 40, 60, 90, 120 and 180 minutes.

Across that monitoring window, the ultra-processed meal produced a significantly greater postprandial insulin area under the curve. Insulin concentrations were also higher in the ultra-processed condition at 40, 60, 90 and 120 minutes.

Area under the curve is a way to summarize the total response over a defined time period. Here, the finding means insulin exposure across the three-hour measurement period was greater after the ultra-processed meal. It should not be described as proof of chronic hyperinsulinemia, insulin resistance or diabetes. The study measured an acute response to one prescribed meal.

Overall glucose exposure was similar, but glucose timing differed

The insulin result does not mean the ultra-processed meal produced a higher overall glucose response. The study reported no significant difference in glucose area under the curve between conditions.

However, the glucose curves were not identical. Glucose rose faster and was higher at 20 minutes after the non-ultra-processed meal. Later, glucose remained relatively more elevated after the ultra-processed meal. Both meals peaked at approximately 40 minutes, and the researchers reported a significant time-by-condition interaction.

This distinction matters. Similar total glucose exposure over three hours can coexist with different timing and shapes of glucose responses. Likewise, the greater insulin area under the curve is an observation that requires explanation, not evidence that the study found a greater overall glucose area under the curve.

The trial also measured post-meal fuel use

Participants in the metabolic subset completed measurements in a whole-room indirect calorimeter before and for three hours after their meals. The ultra-processed condition had greater metabolic-rate area under the curve and greater fat oxidation. The non-ultra-processed condition had greater respiratory exchange ratio and carbohydrate oxidation.

These are short-term differences in measured energy expenditure and substrate oxidation after the test meals. They do not show that ultra-processed foods permanently raise metabolic rate, create a lasting fuel-use pattern or produce a particular long-term body-composition outcome.

Brain-response findings are exploratory, not a causal explanation

A separate fMRI session was completed by 52 participants. Participants viewed images of 28 foods and completed a Becker-DeGroot-Marschak monetary-auction task intended to assess food valuation.

The researchers reported exploratory associations between differences in peak carbohydrate oxidation and striatal activation to food cues. They also reported processing-category differences in valuation-related responses in visual cortex and striatum. These findings are interesting signals for future work, but the imaging session was separate from the meal-metabolism sessions.

The results do not establish that brain activity caused the insulin difference, nor that it causes cravings, increased eating, obesity or metabolic disease. Subjective liking and wanting did not differ significantly between the meal conditions.

Why equal calories may not guarantee identical physiology

Calories, macronutrients and glycemic characteristics remain relevant variables in nutrition and metabolism. What this trial adds is a controlled example in which closely matched values for those variables did not yield identical post-meal insulin, substrate-oxidation and other metabolic measurements.

A reasonable interpretation is narrow: food processing or food form may influence acute post-ingestive physiology beyond the conventional nutrition-label variables matched in this protocol. It is not reasonable to conclude that calories are irrelevant, that macronutrients no longer matter, or that every ultra-processed food produces the same response.

NOVA is a classification system based on the extent and purpose of industrial processing. It can be useful for describing dietary patterns, but it is not a complete measure of nutritional quality or a direct biological mechanism. A NOVA category alone does not specify a food's calories, macronutrients, matrix, additives, glycemic response or health effect.

How this study differs from the earlier NIH weight-gain trial

This acute matched-meal study complements, but does not replicate, the earlier NIH inpatient randomized crossover trial. In that 2019 study, adults ate ultra-processed and unprocessed diets freely over two-week diet periods. Participants consumed more calories and gained weight during the ultra-processed diet phase.

The questions were different. The newer study prescribed a single closely matched meal and measured acute metabolic responses. The earlier study tested ad libitum intake and weight change over longer diet periods. Neither study should be used to erase the distinction between a short-term insulin measurement and long-term changes in voluntary eating or body weight.

What the new trial can and cannot tell us

The strength of this study is the unusually close matching of calories, meal weight, energy density, macronutrients, available carbohydrate, glycemic measures, fiber, sodium and water. That design makes the greater insulin response after the ultra-processed meal a useful challenge to the assumption that matching calories and macros guarantees the same immediate physiology.

Its limitations are equally important. This was an acute study in a relatively small group of healthy-weight adults, with metabolic analyses in 32 people. It cannot establish how the findings apply to people with obesity, prediabetes, diabetes, insulin resistance or other metabolic conditions. It did not test long-term insulin sensitivity, body-weight change, diabetes incidence or chronic disease outcomes.

For now, the most accurate takeaway is specific: closely matched ultra-processed and non-ultra-processed meals produced different acute insulin responses in this trial, even though overall glucose area under the curve was not significantly different. The study raises important questions about food processing metabolism, but it does not yet identify the causal food property or settle the long-term health implications.