They ate 500 fewer calories on unprocessed food. A reanalysis asks what they chose.

The 2019 trial that made ultra-processed food a public health category has been cited for six years as evidence that processing causes overeating. What it did not explain is how.

A post-hoc analysis published in The American Journal of Clinical Nutrition on 29 December returns to the original data and reconstructs what participants actually put on their plates [s1]. The answer is less about processing and more about what people reach for when the options change.

The trial being reanalysed

Twenty weight-stable adults, mean age 31.2 years and mean BMI 27, were admitted to the NIH Clinical Center and randomised to receive either an ultra-processed or an unprocessed diet for two weeks, immediately followed by the alternate diet [s2]. Meals were designed to be matched for presented calories, energy density, macronutrients, sugar, sodium and fibre, and participants were told to eat as much or as little as they wanted [s2].

Energy intake was 508 ± 106 kcal/day higher on the ultra-processed diet (p=0.0001), driven by carbohydrate (280 ± 54 kcal/day, p<0.0001) and fat (230 ± 53 kcal/day, p=0.0004) but not protein (−2 ± 12 kcal/day, p=0.85) [s2]. Participants gained 0.9 ± 0.3 kg during the ultra-processed fortnight and lost 0.9 ± 0.3 kg during the unprocessed one [s2].

The key design fact for what follows: the diets were matched on what was offered. Participants chose freely within them.

What the reanalysis found

Two things about the composition of self-selected meals.

First, on the unprocessed diet, participants preferentially chose low-energy-dense components — under 1.0 kcal/g, mostly fruits and vegetables [s1]. The result was meals that were lower in energy (719.4 ± 11.6 kcal unprocessed versus 829.5 ± 12.51 kcal ultra-processed; F(1,19)=14.9, P<0.001) and simultaneously 57% larger by mass (665.5 ± 10.74 g versus 423.5 ± 8.03 g; F(1,19)=82.9, P<0.001) [s1].

More food, fewer calories.

Second, the unprocessed meals had what the authors call a lower "blend index" — a less equal split of energy between carbohydrate and fat. The difference in the blend index of selected components was 0.22 at lunch (95% CI 0.19–0.26, P<0.0001, d=0.76) and 0.24 at dinner (95% CI 0.19–0.28, P<0.0001, d=0.71) [s1]. The assembled meals were correspondingly less balanced between the two (lunch F(1,19)=18.49, P<0.0004, partial η²=0.493; dinner F(1,19)=24.85, P<0.0001, partial η²=0.57) [s1].

Modelled together, low-energy-dense mass and blend index strongly predicted the energy intakes that were actually observed (r=0.78, df=1676, P<0.001) [s1].

The interpretation, and how much of it is data

The authors propose that unprocessed meals reduce energy intake for two reasons: the carbohydrate–fat blend is less balanced, and the meals promote what they term "micronutrient deleveraging" — a compromise struck between consuming calories and consuming micronutrients [s1]. They describe this as a form of nutritional intelligence [s1].

The measurements support the first half of that. Participants did select bulkier, less energy-dense, less carbohydrate-fat-balanced components when the unprocessed menu was in front of them, and those choices predicted intake [s1].

The second half is a hypothesis about motivation, not an observation. Nobody measured whether participants were seeking micronutrients, consciously or otherwise. What was measured is that when the available foods were fruits and vegetables rather than manufactured composites, people ate more grams and fewer calories. An account in which nutrient-seeking drives the choice and an account in which fruits and vegetables are simply what an unprocessed menu contains both fit these data equally well.

Combinations of carbohydrate and fat in near-equal energy proportions are also relatively rare in whole foods and common in manufactured ones — which means the blend index may be partly a description of the food supply rather than of an appetite mechanism.

The limits, which are the same limits as the original

Twenty people [s1][s2]. A metabolic ward, where every meal is supplied and nothing else is available. Two weeks per arm. This is a post-hoc analysis, meaning the comparisons were specified after the data were in hand, and the paper is explicit about that [s1].

The correlation of 0.78 is calculated across 1,676 degrees of freedom, which reflects meals rather than participants [s1] — the effective sample for generalisation is still twenty people.

Nothing here establishes that processing per se is harmful. It establishes that within a controlled menu, the unprocessed condition led people to assemble different plates, and that the difference in those plates accounts statistically for the calorie gap.

Why it matters

Most public advice about ultra-processed food is framed as avoidance. This analysis points at a different lever: the physical composition of a meal.

If the operative variables are the mass of low-energy-dense food on the plate and the carbohydrate-to-fat balance of the meal [s1], those are things that can be changed inside a diet that still contains processed items — and they are testable independently of the processing category entirely. That is the experiment this analysis implies and does not run.

What to watch

Whether anyone tests the blend index and low-energy-dense mass directly, by holding processing constant and varying only those two properties. Until that trial exists, "micronutrient deleveraging" remains a plausible name for a pattern rather than a demonstrated mechanism.

This article is informational and is not medical advice.

Sources

• [s1] Brunstrom JM, Schatzker M, Rogers PJ, Courville AB, Hall KD, Flynn AN. "Consuming an unprocessed diet reduces energy intake: a post-hoc analysis of a randomized controlled trial reveals a role for human nutritional intelligence." The American Journal of Clinical Nutrition, published online 29 December 2025. https://doi.org/10.1016/j.ajcnut.2025.101183
• [s2] Hall KD, Ayuketah A, Brychta R, et al. "Ultra-Processed Diets Cause Excess Calorie Intake and Weight Gain: An Inpatient Randomized Controlled Trial of Ad Libitum Food Intake." Cell Metabolism, 30(1), published online 16 May 2019. https://doi.org/10.1016/j.cmet.2019.05.008