The people in the vegan arm of a 16-week trial did not eat less food. They ate more of it. Total daily food weight rose from about 1,825 grams to 1,930 grams, an increase the analysis treated as statistically flat. Over the same period, their daily calorie intake fell by 491.
The food-group table published alongside the JAMA Network Open analysis on August 6 shows exactly where those calories went, and the arithmetic is more surprising than the headline result.
The Four Numbers That Did Most of the Work
Meat, poultry, fish and shellfish dropped from about 102 grams a day to 1.4. Dairy went from 129 grams to 1.5. Eggs went from 30 grams to 0.1. Those three groups had been supplying roughly 402 calories a day between them, and they were gone.
The fourth number is the one nobody talks about. Nuts and added fats fell from 39 grams a day to 12.5, a cut of about two-thirds by weight. Because that group carries roughly 6.5 calories per gram, more than any other food category in the study, that shift of 27 grams removed about 145 calories a day. Gram for gram, nothing else on the plate came close.
What replaced them was bulk. Vegetables rose by 143 grams a day. Fruit rose by 80 grams. Legumes went from 36 grams to 110, roughly a tripling, and nondairy alternatives more than doubled. Grains barely moved in the vegan group, though they fell sharply in the control group.
Net across the two arms, the vegan group's calorie intake dropped 357 calories a day more than controls, with a confidence interval running from 192 to 522.
Two Plant Foods Got More Calorie Dense, Not Less
Here is the counterintuitive part. Overall energy density in the vegan group fell about 30%, from roughly 1.1 to 0.7 calories per gram. But the contribution from legumes rose, from 0.7 to 1.1 calories per gram. The contribution from nuts and fats rose too, from 6.5 to 7.2.
That is not a contradiction. Those figures track what each group contributed to the diet, weighted by how much of it people ate. Legumes went up because participants ate three times as many of them, not because beans became more fattening. The total still fell because entire high-density categories were removed from the plate.
The practical translation is that the 30% drop came from subtraction at the top of the density scale, not from every food getting lighter. Vegetables barely budged, at 0.5 calories per gram before and 0.4 after. They were already low.
Reductions in energy density tracked with greater weight loss, at r = 0.28, and the relationship held after adjusting for calorie intake, at r = 0.21. The authors read that as energy density doing work beyond simply lowering intake. It remains an association within one trial population.
Who Was in the Room
The parent trial randomized 244 adults with overweight in Washington, D.C., between January 2017 and February 2019, and 223 completed it after a later correction to the original count. In the originally published trial, 87% of participants were women and the mean age was 54. The vegan arm's weight advantage over controls came to a treatment effect of 5.9 kilograms.
The energy density question was examined after the fact, using diet records collected for a different purpose, and the statistical analysis was run between April and June 2026. Food intake came from three-day self-reported diaries at week 0 and week 16, which is the weakest link in the chain. People underreport, and they underreport unevenly.
The control group made no dietary changes at all, a design registered in advance on ClinicalTrials.gov. That comparison shows the vegan pattern beat doing nothing. It does not show it beats a Mediterranean pattern, a higher-protein pattern, or simply cooking with less oil.
Funding came from the Physicians Committee for Responsible Medicine, which advocates for plant-based diets and employs the lead author and the senior author. The journal states the funder had no role in design, analysis, or the decision to publish. Readers can weigh the measurements and the framing separately.
The Lever That Is Not About Being Vegan
Strip out the label and what the table describes is a mechanical relationship. Fat carries about nine calories per gram against four for protein or carbohydrate. Water and fiber carry weight without calories. And controlled feeding work has repeatedly found that people tend to eat a fairly consistent weight of food at a meal, which is why lowering the calories packed into each gram lowers intake without anyone deciding to eat less.
Any diet that moves weight toward water-rich, fibrous foods and away from concentrated fats will move in the same direction. Adding a serving of beans and cutting the cooking oil in half is the same lever, and the trial's own numbers suggest the added-fat side of that equation is doing more per gram than anything else.
None of this changes clinical guidance, and a secondary analysis of a 16-week trial would not be expected to. Anyone considering a large dietary shift, particularly people who are pregnant, managing diabetes, taking medications affected by diet, or with a history of disordered eating, should work it through with a clinician or registered dietitian first. Vegan patterns also require attention to vitamin B12 and, depending on choices, iron, zinc, calcium, iodine, and omega-3 intake.
Key Questions Answered
Which food change removed the most calories per gram?
Nuts and added fats. Cutting that group by about 27 grams a day removed roughly 145 calories, because it carries about 6.5 calories per gram.
Did participants eat less food?
No. Total food weight rose slightly and the change was not statistically significant. Calorie intake fell because the food was less concentrated.
Why did legumes appear to get more calorie dense?
The figure tracks each group's weighted contribution to the diet. Legume intake roughly tripled, so their share of the total rose even though beans themselves did not change.
How big was the effect against the control group?
A drop of 357 calories a day, with a confidence interval from 192 to 522.
What are the main limitations?
Self-reported diet records, a post hoc analysis, 16 weeks of follow-up, a control group that changed nothing, and a trial population that was 87% women with a mean age of 54.
Do you have to go vegan to get this effect?
No. The mechanism is compositional. Adding water-rich, high-fiber foods and cutting added fats moves energy density in the same direction on any eating pattern.
