August 30, 2026 · Science · Nutrition · Biomarkers

What the Keto Diet Did to Liver Fat

A ketogenic diet cut liver fat by 67% in a randomized trial. Liver fat fell by an average of 45% in the Mediterranean arm and 45% in the plant-forward arm over the same period. Participants in all three groups lost about 10% of their starting body weight, so the gap in liver fat did not come from one group losing more.

The trial randomly assigned 55 adults with obesity, prediabetes and fatty liver to one of the three diets for around five months, according to Washington University School of Medicine in St. Louis. Participants received all of their food for the duration and met a study dietitian weekly. The first author is Max C. Petersen, an assistant professor of medicine there, and the results were published on August 27, 2026 in Cell Metabolism.

The arms differed sharply in composition. ScienceDaily reports the ketogenic diet at 4% carbohydrate and 73% fat, the Mediterranean diet at 50% carbohydrate and 35% fat, and the plant-forward diet at 70% carbohydrate and 15% fat.

The short version

Supplying every meal for five months to 55 people is an expensive way to answer a narrow question, and it bought a cleaner answer than a food-diary study can. With weight loss held at about 10% in every arm, the liver fat difference has to come from what the diets were made of. That is worth knowing. Two questions stayed out of scope. The trial did not test whether anyone eats this way once the food stops arriving. It did not test whether a liver fat number that falls further means fewer people get sick years later. Neither is answered by a larger effect on a scan.

Why Equal Weight Loss Makes the Liver Fat Result Cleaner

Matching weight loss across the arms is what gives this trial its value. Most comparisons between diets confound two things at once: how much weight a person lost, and what they ate to lose it. Weight loss alone reduces liver fat, so a diet that produces more of it will look better on a scan for reasons that have nothing to do with its composition. Holding the loss at about 10% in all three groups removes that explanation.

Several of the reported outcomes split along the same line. Washington University reports that insulin sensitivity in muscle rose roughly 50% from baseline across every group, which tracks the shared weight loss rather than the diet. Insulin sensitivity in the liver improved two to three times more on the ketogenic diet. Twenty-four-hour blood glucose fell 20% in the ketogenic group against 8% in the others. Blood insulin fell 74% on the ketogenic diet, 44% on the Mediterranean diet and 27% on the high-carbohydrate diet.

Prediabetes status moved in the same direction. Half the ketogenic participants reversed their prediabetes, against 29% on the Mediterranean diet and 7% on the high-carbohydrate diet. ScienceDaily adds that the ketogenic group showed the largest rise in glucagon and did not show an increase in blood fat or cholesterol levels. No lipid figures were published in either account, so the size of that lipid result is not established.

Why a Liver Panel Does Not Report Liver Fat

The endpoint this trial moved is one the standard liver blood panel cannot see. Our guide to the hepatic function panel lists what those seven tests report. They are ALT, AST, alkaline phosphatase, bilirubin, albumin and total protein. Every one is an enzyme or a synthetic product, and none is a measure of fat content. The same guide records the American College of Gastroenterology position that a normal ALT does not exclude liver disease, and that the panel contains no imaging and no fibrosis measure. Reading this trial through a subscription blood panel would therefore miss the result entirely.

That gap matters commercially, because the population in this trial is the population these testing platforms sell to. The ACG guidance our page cites recommends ultrasound screening for fatty liver in one specific group. It is people with a raised BMI, features of metabolic syndrome, and a mildly raised ALT. That covers most adults who buy a broad panel and find one liver value out of range. The panel raises the question, and an imaging study a clinician orders answers it. No amount of extra blood markers substitutes.

Who Should Not Read This as a Recommendation

Anyone with type 1 diabetes, a history of pancreatitis, or an inherited disorder of fat metabolism sits outside this trial's population and should treat the result as background. So does anyone taking an SGLT2 inhibitor, where very low carbohydrate intake interacts with a known ketoacidosis risk that a prescriber has to weigh. The trial enrolled adults with obesity, prediabetes and fatty liver, and it says nothing about people who have none of the three.

People already at a stable weight with normal liver values are also outside it. The mechanism on display here operates through weight loss plus composition, and there was no arm testing what happens to a lean person who removes carbohydrate. Anyone in either group who wants to act on this should raise it with the clinician managing their care rather than with a testing subscription.

What Would Change This Read

A trial with hard outcomes would change it immediately. Liver fat on a scan is a marker. Nobody yet knows whether people who reduce it further go on to develop less cirrhosis, fewer cardiovascular events and less type 2 diabetes. Nothing in a five-month feeding study can answer that. A longer trial powered for those outcomes would settle whether the 67% figure describes a health difference or a measurement difference.

A free-living replication would change it in a different way. Every meal in this trial arrived from the study kitchen, and the ketogenic arm at 4% carbohydrate is the hardest of the three to reproduce from a supermarket. A version that gave people the same instructions and no food would test the diet a reader can actually adopt. Published lipid data would settle the third question, because the claim that cholesterol did not rise currently rests on a sentence with no numbers behind it.

Related Coverage

An earlier briefing on this site covered a Stanford analysis tying LDL increases on low-carb diets to a genetic score, and noted that the analysis did not test a ketogenic diet. This trial did, and reports no rise in blood fat or cholesterol without publishing the figures. Both lead to the same practical step for anyone cutting carbohydrate. Measure the lipid response instead of predicting it. The guides on this site covering ApoB and fasting insulin deal with the two markers that move most on this kind of dietary change. The prediabetes reversal guide covers what the remission threshold in this trial means in laboratory terms. Anyone tracking liver fat on a keto diet should ask a clinician about an imaging measure rather than a broader liver enzyme panel, and get the baseline before the diet changes.

Sources

  • Washington University School of Medicine, "Low-carb diet delivers added liver benefits beyond weight loss," medicine.washu.edu (accessed August 30, 2026). Source of the 55-participant randomization, the five-month duration and the provided-food design. Also the source of the 67% and 45% liver fat figures, the 10% weight loss, the 50% muscle insulin sensitivity change, and the two-to-three-fold hepatic insulin sensitivity difference. The 20% and 8% glucose figures, the 74%, 44% and 27% insulin figures, and the 50%, 29% and 7% prediabetes reversal figures come from the same release.
  • ScienceDaily, "Keto diet cut liver fat by 67% in a clinical trial," August 29, 2026, sciencedaily.com (accessed August 30, 2026). Source of the macronutrient composition of the three arms, the glucagon observation, the statement that blood fat and cholesterol did not increase, and the 42-participant count noted above.
  • Max C. Petersen et al., "Effect of diet macronutrient content on the cardiometabolic response to weight loss: A randomized clinical trial," Cell Metabolism, 2026, doi.org/10.1016/j.cmet.2026.07.020 (citation as published by ScienceDaily. The full text was not retrievable at the time of writing).

Frequently Asked Questions

Was this trial randomized?

Yes. Washington University School of Medicine reports that 55 adults with obesity, prediabetes and fatty liver were randomly assigned to one of three diets for about five months. Random assignment is what separates this from the observational diet research that dominates the field, because it removes the self-selection that makes healthy-user effects impossible to untangle. The trial was small, and a 55-person randomized comparison across three arms leaves fewer than 20 people in each group.

Did the keto group lose more weight than the others?

No, and that is the point of the design. Washington University reports that participants lost about 10% of their starting body weight across all three groups. Because the weight loss was matched, the remaining differences in liver fat, blood glucose and insulin trace back to what the diets were made of.

Did the ketogenic diet raise cholesterol?

The ScienceDaily write-up reports that the ketogenic group did not show an increase in blood fat or cholesterol levels. No figures for LDL cholesterol, ApoB or triglycerides appear in either the university release or that write-up. The size of the effect on lipids is therefore not established. Individual lipid responses to a very low carbohydrate diet vary widely, which is a separate finding covered in an earlier briefing on this site.

How was liver fat measured?

Neither the Washington University release nor the ScienceDaily summary names the measurement method. What both make clear is that liver fat is not a value a blood panel returns. A hepatic function panel reports enzymes and proteins such as ALT, AST, alkaline phosphatase, bilirubin and albumin, and it contains no imaging and no measure of liver fat or fibrosis. Anyone wanting to track the endpoint this trial moved needs an imaging study ordered by a clinician rather than a broader blood panel.

Does the trial mean a ketogenic diet is the best diet for fatty liver?

No. In 55 people, with all food supplied for about five months, the ketogenic arm reduced liver fat more than two other diets at the same weight loss. It did not measure heart attacks, strokes, liver failure or death, and it did not test whether anyone could sustain the diet once the food stopped arriving. Diet choice for fatty liver disease is a clinical decision that belongs with a physician who knows the rest of the history.

What does prediabetes remission mean in this trial?

Washington University reports that half the participants on the low-carbohydrate diet reversed their prediabetes, against 29% on the Mediterranean diet and 7% on the high-carbohydrate diet. The release does not state which laboratory criteria defined reversal or how long the participants held that status after the trial ended. Both details matter for reading the number, because prediabetes is defined by thresholds that a person can cross back and forth across in either direction.