Losing the same weight, keto dieters shed more liver fat

In a trial of 42 adults with fatty liver and prediabetes, each group lost about 10% of their weight. Liver fat fell by two-thirds on a keto diet and by under half on Mediterranean or very-low-fat diets. But the groups were small, liver fat was a secondary measure taken while people were still eating keto, and the trial can't say whether the edge lasts or what months of keto do to cholesterol.

9 October 2026 · How we rate evidence

For four months, every bite the 42 volunteers ate came out of a freezer bag. The kitchen at Washington University in St. Louis packed 500-calorie entrées to order: omelets with bacon and sausage for one group, minestrone and quinoa for another, red beans and rice for the third. The goal was to make three very different diets produce the same weight loss, so any difference left over would belong to the food.

The claim going around. "Keto diet cut liver fat by 67% in a clinical trial," ScienceDaily reported on 29 August, reposting a Cell Press release. Labroots, and Fox News and Blavity via MSN, led with the same figure. The trial is careful and the 67% is real, but it isn't the size of keto's advantage.

What they did. Max Petersen, Samuel Klein and colleagues recruited adults aged 18 to 55 with obesity, prediabetes (blood sugar above normal but below diabetes) and fatty liver, meaning at least 5% of the liver was fat on an MRI scan. Fifty-five were randomly assigned to one of three diets. The keto diet got 4% of its calories from carbohydrate and 73% from fat, with 23% from saturated fat. The Mediterranean diet was 50% carbohydrate and 35% fat. A plant-forward very-low-fat diet was 70% carbohydrate and 15% fat.

Calories were cut until each person had lost about 10% of their weight, which took four to five and a half months. The groups landed at 10.4%, 10.2% and 10.1%, 11 to 12 kg each. After three or four weeks at steady weight, everyone was retested, while still eating their assigned diet. Thirteen people dropped out, and 42 were analysed, 14 per diet.

What they found. Losing 10% of body weight helped everyone. Liver fat fell by about 45% on the Mediterranean and very-low-fat diets, and muscle insulin sensitivity rose about 50% on all three, with no difference found between diets. The keto group's liver fat fell further, by 67% on average: from 19.4% of the liver to 6.2%, against about 18% to 10% on the other two. In absolute terms, keto removed roughly 3 to 5 more percentage points of liver fat than either alternative (our arithmetic from the paper's table; the keto group started a little higher).

That difference, not the 67%, is the measure of keto's edge.

Keto also led on several blood-sugar measures. Long-term blood sugar (HbA1c) fell 0.6 points on keto, against 0.1 and 0.4 on the others. Half the keto group no longer met the definition of prediabetes, against 29% and 7%. And a liver insulin sensitivity index, one of the trial's two main outcomes, rose two to three times as much on keto.

On the side of possible harms, LDL cholesterol, the "bad" kind, fell on average on all three diets (by 10 mg/dL on keto, 22 on the Mediterranean diet and 8 on the very-low-fat diet), though clearly only on the Mediterranean diet. The authors report no difference among groups, and first author Max Petersen called the absence of a cholesterol rise the "most surprising finding." That holds for fasting cholesterol and for triglycerides across the whole day, but on keto, fatty acids in the blood rose 63% over 24 hours and triglycerides ran higher after meals. LDL was measured just after weight loss, which lowers it, and with 14 people per group the trial was too small to rule out a difference of around 12 mg/dL, the gap between keto and Mediterranean (our calculation from the paper's ranges). No serious side effects were reported; the table of lesser ones is in a supplement we couldn't open.

What it doesn't show. First, liver fat was a secondary outcome, one of dozens of things measured, with no correction for testing so many. The trial's registered main outcome was insulin sensitivity. Second, the groups were small, and the analysis included only the 42 who finished; nearly a third of the very-low-fat group dropped out, against under a fifth on keto, and the reasons aren't given in the parts of the paper we could open.

Third, and most important, everyone was retested while still on their diet. Liver fat and fasting insulin respond to carbohydrate within days: the same lab found in 2009 that liver fat fell about 30% in the first 48 hours of a calorie-cut low-carbohydrate diet, against 9% on a high-carbohydrate one, and a 2024 Danish study of 11 men saw a 35% fall after four days of low-carb eating without weight loss. Part of keto's edge may be the state of eating almost no carbohydrate, not a lasting change. The liver insulin index has the same problem, since it is calculated from fasting insulin, which a near carb-free diet lowers directly. As run, the trial can't tell the two apart.

The same 2009 study cuts the other way, though. After 7% weight loss and at least four weeks at steady weight, with the low-carbohydrate group still in ketosis, it found no clear difference in liver fat between the diets. So being on keto at retest doesn't guarantee an edge. That trial had 11 people per group with a wide spread, allowed more carbohydrate (under 50 g a day, against 20 g), supplied food only for the first two days, and only about half its volunteers had fatty liver.

What it could mean for people. For someone with fatty liver and prediabetes, the question is whether the kind of diet matters at equal weight loss. Here it did, on the day of testing. Whether it matters months later, and whether the liver benefit carries a cholesterol cost, this trial wasn't built to answer. A 2023 trial in Denmark of people with type 2 diabetes, not designed to match weight loss, found LDL came out about 14 mg/dL worse on a low-carbohydrate diet over six months. And the Helsinki-led study found signs of a starvation-like state in keto livers that its authors link to liver injury.

Why it's still interesting. Many diet trials fail to match weight loss or keep people on the diet, as the authors note. By cooking every meal and titrating calories, this team got three groups to within a third of a percentage point of each other. With weight off the table, keto still moved liver fat further. That makes it a clean test, if a small one.

Open questions.

Does the extra liver-fat loss last once carbohydrates come back? Rescan everyone after all three groups switch to the same mixed diet at steady weight; the registry lists follow-up visits for up to two years. If the edge persists, a keto phase during weight loss could be a one-off tool; if it fades, the benefit lasts only as long as the diet.

What happens to LDL cholesterol after months of keto at steady weight? Run a longer trial with enough people to detect a 10 mg/dL difference in LDL and in apolipoprotein B, a count of the particles that carry it. The answer decides whether the liver benefit comes with a heart-health cost.

Does it take full ketosis? The authors say they can't tell whether a milder carbohydrate cut would work. A trial with several carbohydrate levels at matched weight loss would show whether something easier to stick to would do.

A note on timing. The paper was published online on 27 August 2026 and ScienceDaily ran the release two days later; more coverage followed into mid-September. The trial ran from 2016 to 2025.