Twenty-five adults with type 2 diabetes went low-fat vegan for 16 weeks with no calorie limit. Liver fat fell by a third and HbA1c by a full point. They also lost 6 kg, and nobody was there to compare them with

Fat stored in the wrong places, inside the liver and inside muscle cells, is one of the better-supported explanations for why insulin stops working properly in type 2 diabetes. Strict low-calorie diets clear it out, and blood sugar improves when they do. The catch is that they mean counting, weighing and going hungry. A team from the Physicians Committee for Responsible Medicine in Washington, DC, working with Yale School of Medicine, asked whether a change in what people eat, with no limit on how much, would do something similar. Their paper came out in BMJ Open Diabetes Research & Care on 8 October 2026.

They phoned 1,244 people to find 25 who fitted and were willing: adults with type 2 diabetes managed by diet or by tablets other than sulfonylureas, an HbA1c between 6.5% and 10.5% and a BMI between 26 and 40. People on insulin or on weight-loss drugs such as the GLP-1 medicines were not taken. For 16 weeks the 25 were asked to eat a low-fat vegan diet of vegetables, grains, beans and fruit, with no animal products and no added fats, aiming for roughly 75% of energy from carbohydrate, 15% from protein and 10% from fat. No food was provided. They cooked at home or ate out, met weekly as a group, saw a dietitian, took a vitamin B12 tablet and were told to eat until full.

Twenty-four finished. Their weight was down by an average of 6.0 kg, 6.6% of where they started. Fat inside the liver, measured by magnetic resonance spectroscopy at Yale, was down by 33%. Visceral fat, the kind packed around the organs, fell by 290 cubic centimetres. HbA1c dropped by a full percentage point, LDL cholesterol by 22 mg/dL, and the fasting measure of insulin resistance fell. From blood samples taken over three hours after a standard liquid breakfast, the team also calculated that the insulin-producing cells of the pancreas were responding better to rising glucose.

The medication detail is the one that will make people sit up. Seventeen people were on metformin at the start. One dropped out. Of the other sixteen, nine stayed on the same dose, five had it reduced and two came off it. Nobody started or increased a glucose-lowering drug. So the one-point fall in HbA1c happened while the tablets were, on balance, going down. Those changes were made by the participants' own doctors, not by the study.

Now for what the study cannot say, and the authors say it first. There was no comparison group. Everyone got the diet, the weekly meetings and the dietitian, and everyone knew what they were eating. And although nobody was told to eat less, they did: their food diaries showed about 346 kcal a day less than before. A third less liver fat alongside a 6.6% weight loss is roughly what weight loss by any route tends to bring. The paper puts it plainly: the changes 'cannot be attributed specifically to dietary composition independent of the accompanying reduction in energy intake and weight loss'. Whether there is anything particular about vegan food here, beyond the fact that a plate of beans and rice is hard to overeat, is a question this design cannot answer.

Two more things are worth knowing. The number of participants was worked out from the expected fall in fat inside muscle cells. That fell by 13%, with a margin of error running from a 30% fall to a 3% rise, so it could be chance. The paper's discussion nonetheless speaks of 'comparable reductions' in liver and muscle fat. And of the 6 kg lost, about 2.4 kg was lean tissue, around 40% of the total, on a diet where protein intake went down and in a group asked not to change their exercise habits. The authors flag this themselves and suggest future trials add resistance training or pay attention to protein. For people in middle age and beyond, lost muscle is not a small footnote.

Who ran the study matters too. The Physicians Committee is a non-profit that campaigns for plant-based eating; it part-funded the work and paid five of the ten authors for it, and its president, a co-author, earns royalties from books on nutrition. US National Institutes of Health grants supported the Yale side, and the statistician was not told what the diet was. None of that makes the scans wrong. It does mean the natural next step, a randomised trial against another diet that produces the same weight loss, ought to be run by people with no stake in the answer. The 25 volunteers were also, in the authors' words, 'health conscious and willing to make substantial changes to their diet', picked from more than a thousand callers.

Our reading: this is a small, careful before-and-after study showing that people with type 2 diabetes who switch to a low-fat vegan diet and lose 6 kg see their liver fat and blood sugar improve, without being told to count anything. That is useful. For some people, a rule about what is on the plate is easier to live with than a rule about how much. It does not show that a vegan diet beats any other way of losing the same weight, and it ran for only 16 weeks. If you take diabetes medication, a big change in diet can bring your blood sugar down faster than you expect, so plan it with your doctor rather than adjusting tablets yourself. And a fully vegan diet needs a reliable source of vitamin B12, which is why the trial handed one out.