Background: In 2019, Hall et al. reported a randomized clinical trial showing that an ultraprocessed diet increases energy intake by ∼500 kcal/d compared with an unprocessed diet.
Objective: This post-hoc analysis assessed whether participants selected meal components with specific nutritional characteristics and how this affected energy intake.
Methods: Twenty weight-stable adults received an ad libitum ultraprocessed or unprocessed diet for 2 wk, followed by the alternate diet. ANOVA and t tests assessed diet effects; a linear mixed model assessed predictors of meal size.
Results: With the unprocessed diet, participants selected components with a less-equal blend of energy from carbohydrate and fat ["blend index" difference; lunch = 0.22 (95% CI: 0.19, 0.26), P< 0.0001, d = 0.76; dinner = 0.24 (95% CI: 0.19, 0.28), P< 0.0001, d = 0.71]. These components formed meals that had a lower blend index (less balanced) than ultraprocessed meals [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]. With the unprocessed diet, participants preferentially chose low-energy-dense components (<1.0 kcal/g, mostly fruits and vegetables), creating meals lower in energy (unprocessed = 719.4 ± 11.6 kcal compared with ultraprocessed = 829.5 ± 12.51 kcal), [F(1,19) = 14.9, P < 0.001, η2G = 0.0457], yet significantly larger (57%) by mass (unprocessed = 665.5 ± 10.74 g compared with ultraprocessed = 423.5 ± 8.03 g), [F(1,19) = 82.9, P < 0.001, η2G = 0.274]. Modeled together, low-energy-dense mass and blend index strongly predict observed energy intakes (r = 0.78, df = 1676, P < 0.001).
Conclusions: Unprocessed meals may reduce energy intake because: 1) they have a less balanced carbohydrate-fat blend; and 2) they promote a form of nutritional intelligence whereby a compromise is struck between consuming calories and consuming micronutrients, which we refer to as "micronutrient deleveraging." This trial was registered at clinicaltrials.gov as NCT03407053.

