On a low-fat vegan diet, plates stayed as heavy but calories fell
In a 16-week trial of 244 adults with overweight, people told to go vegan lost about 6 kg more than people who changed nothing. By their own food diaries they ate roughly the same weight of food as before, but cut about 360 calories a day more than the other group did. A new reanalysis says fewer calories per gram explains the weight loss. The data fit that idea but can't single it out.
10 October 2026 · How we rate evidence
- Claim
- Weight loss (Solid): In the trial's main report (2020), the low-fat vegan group lost 6.4 kg over 16 weeks and the no-change group 0.5 kg, a difference of 5.9 kg (range 5.0 to 6.7). Weight was one of the trial's registered main outcomes.
Calories per gram (Early, close to Speculative): The energy density of the vegan group's diet, meaning calories per gram of food and drink, fell from 1.1 to 0.7; the control group's went from 1.0 to 0.9. The difference between groups was 0.28 calories per gram (range 0.19 to 0.38). Taken from food diaries, in an analysis that wasn't planned in advance.
Food weight (Early, close to Speculative): The total weight of food and drink people wrote down showed no clear difference between groups: the vegan group's rose by about 107 grams a day more than the control group's (range 99 grams less to 312 grams more). Same diaries and analysis.
Link to weight (Early, close to Speculative): Across everyone in the trial, bigger drops in energy density went with more weight loss: a correlation of 0.28 (range 0.15 to 0.39), or 0.21 after allowing for calorie intake. Also not planned in advance.
Popular claim (Overstated): "Low-fat vegan diet helps people lose weight without eating less" (ScienceDaily, 17 Aug 2026, reprinting the Physicians Committee for Responsible Medicine release). The lead author, quoted in the release: "The answer is energy density" and "You don't have to eat less." By their diaries, the vegan group cut about 357 calories a day more than the control group. - Studied in
- People: 244 adults in Washington, DC, with a BMI of 28 to 40 and no diabetes; 87% women, average age 54. 122 per group; 223 of the 244 finished and were analysed
- Design
- Randomized trial, 2017 to 2019, not blinded. The vegan group had weekly classes and no calorie limit; the control group was asked to make no changes and had no classes. Weight was measured by staff who didn't know people's group. Energy density was not a registered outcome: it was calculated in 2026 from 3-day food diaries the participants kept at the start and at 16 weeks. The correlation compares people across both groups, not randomized groups
- Status
- Peer-reviewed research letter, JAMA Network Open 9(8):e2627483, published 6 Aug 2026. Main results: JAMA Network Open 2020. Registered as NCT02939638. The 2020 report has three published corrections (the funder's role, wording on calorie burning, and the number who finished); none changed the weight results
- Replicated
- Weight loss: partly; other trials, some independent, and a review that includes this trial report weight loss on vegan and whole-food plant diets (we read their summaries and the review's list of trials, not the full papers). Energy density and its link to weight: no independent test found
- Funding
- Physicians Committee for Responsible Medicine (PCRM), a nonprofit that promotes plant-based diets, funded this analysis; the 2020 trial also had grants through the Yale Diabetes Center. Four of the five authors work at PCRM. The letter says the funder had no role in the study. For the 2020 report of the same trial, the journal published a correction replacing that statement with one saying the PCRM employees among the authors were involved in its design, analysis and writing. Neal Barnard is PCRM's unpaid president and earns royalties from books on nutrition; a statistician, Richard Holubkov, received fees from PCRM
- We read
- Full letter (free via PubMed Central, CC BY-NC-ND), its table and figure, the trial protocol and data statement, the registry record, and the 2020 main report in full with its three published corrections. Not the participant flow chart, an image we couldn't open. The PCRM release as reprinted by ScienceDaily; headlines from News-Medical, MSN and the Jamaica Gleaner
- Would change our mind
- Up: a feeding study that lowers calories per gram while keeping fat and food types the same, and finds people eat less and lose weight. Down: an objective measure of what people ate, rather than diaries, showing a much smaller calorie gap
Before the trial, the people later put on a vegan diet wrote down about 1.8 kg of food and drink a day. Sixteen weeks later they wrote down about 1.9 kg. Their calories went the other way, from about 1,830 a day to about 1,340. They were eating as much by weight, but the food itself had changed.
The claim going around. "Low-fat vegan diet helps people lose weight without eating less," ScienceDaily reported on 17 August, reprinting a press release from the Physicians Committee for Responsible Medicine (PCRM), which ran the study. News-Medical's version read "without cutting back on food," and the Jamaica Gleaner ran the story in September. In the release, lead author Hana Kahleova says: "The answer is energy density."
What they did. Between 2017 and 2019, PCRM researchers in Washington, DC, working with colleagues at Yale, randomly split 244 adults with overweight into two groups. One was asked to eat a low-fat vegan diet: vegetables, grains, legumes and fruit, no animal foods, no added oils, no calorie limit. That group also went to weekly classes with cooking demonstrations. The other group was asked to change nothing and got no classes. Everyone knew which group they were in.
The main results came out in 2020. The vegan group lost 6.4 kg on average; the control group lost 0.5 kg. The new paper, a short research letter published on 6 August, goes back to the food diaries people kept for three days at the start and at 16 weeks. It asks why the vegan group lost weight when nobody told them to eat less. The analysis wasn't planned in advance; the team ran it this year.
What they found. The vegan diet cut calories per gram sharply. Counting food and drinks together, energy density fell from 1.1 to 0.7 calories per gram, a drop the paper puts at about 30%. In the control group it barely moved, from 1.0 to 0.9. Calories fell in both groups, but by about 490 a day in the vegan group and 130 in the control group, a gap of 357 calories a day. The total weight of what people ate didn't clearly change in either group.
| Per day, week 0 to week 16 | Vegan group | Control group |
|---|---|---|
| Food and drink, grams | 1,825 to 1,931 | 1,907 to 1,924 |
| Calories | 1,834 to 1,343 | 1,790 to 1,656 |
| Calories per gram | 1.1 to 0.7 | 1.0 to 0.9 |
| Weight lost (2020 report) | 6.4 kg | 0.5 kg |
The table shows where the change came from. The vegan group dropped about 260 grams a day of meat, dairy and eggs, and cut the calories they got from nuts and fats by nearly two-thirds. In their place came vegetables, fruit, beans and plant milks, close to 390 grams a day of them. Across all participants, people whose energy density fell most tended to lose the most weight.
The weight result is the sturdy part: a randomized trial, weight a registered main outcome measured on scales by staff who didn't know people's group, 223 of 244 finishing, and a narrow range around the 5.9 kg gap (5.0 to 6.7 kg).
What it doesn't show. First, "without eating less" holds only by weight. In calories, the vegan group ate about a quarter less than before (our arithmetic). The release's own summary says so. They weren't told to count calories or cut portions, and that part of the message holds.
Second, the study can't show that energy density is why they lost weight. The diet changed several things at once: less fat, more fibre, no animal foods, plus weekly classes the control group didn't get. Any of those could have lowered intake. The link to weight comes from a correlation across everyone in the trial. The vegan group had both the bigger drop in energy density and the bigger weight loss, so a correlation across both groups mostly restates that group difference: by our rough arithmetic from the paper's group averages, group membership alone would produce most of it. It isn't an independent test. And a correlation of 0.28 means energy density lined up with about 8% of the variation in weight change (our arithmetic). How the two line up within the vegan group alone isn't reported.
Third, the calorie figures come from diaries kept by people who knew their group and were being taught what to eat. The control group reported eating 134 fewer calories a day yet lost only half a kilo. In the 2020 report, a regression linked the reported calorie gap to only about half a kilo of weight loss, against the 5.9 kg measured on the scales, a sign that diaries are noisy guides for individuals. That noise also weakens correlations, so the true link could be stronger or weaker than 0.28.
The participants were volunteers, mostly women, average age 54, followed for 16 weeks. PCRM, which promotes plant-based diets, ran the trial and paid for this analysis.
What it means for people. The question readers care about is whether you can lose weight without going hungry. This study can't answer that, because it didn't measure hunger. What it does show is that over four months this group wrote down as much food by weight, cut calories without being told to, and lost about 6 kg more than people who changed nothing. Other trials have found weight loss on vegan diets too. A 2022 review of 11 trials by Danish researchers, in Obesity Reviews, put the average at about 4 kg compared with control diets. Five of those trials, including this one, were led by PCRM researchers. What this letter adds is a likely route, fewer calories per gram. It doesn't show that route is what did the work, or that it works without the rest of the package.
Why it's still interesting. "Eat less" is advice most people fail to follow for long. "Eat differently, as much as you like" is easier to live with, and this study shows the arithmetic of how it can work: take out the densest foods, add bulky low-calorie ones, and calories fall while plates stay full. None of that requires going fully vegan. Water-rich vegetables, fruit and beans are available on any diet. That turns the finding into a question anyone could test.
Open questions.
Is it the calories per gram, or the fat, or leaving out animal foods? A feeding study could serve two diets with the same fat and the same kinds of foods but different energy density, and measure what people eat. If energy density drives intake, the advice works in any kitchen. If fat or animal foods are the key, the advice is narrower.
Does it last? The trial ran 16 weeks. Following energy density, calories and weight for a year or more, with an objective measure of intake rather than diaries, would show whether people's appetites adjust. A strategy that fades after six months isn't a weight-management strategy.
How much did the classes matter? The control group got no classes. A trial where both groups get the same support, but only one changes diet, would show whether the food alone produces the same result. That decides whether a leaflet is enough, or whether people need the course.
A note on timing. The trial ran from 2017 to 2019, and its weight results were published in 2020. The energy density letter came out on 6 August 2026. ScienceDaily reposted the release 11 days later, and coverage was still appearing in mid-September.