A ketogenic diet reduced liver fat more effectively than a Mediterranean or a low-fat, plant-forward diet — even when all three produced the same weight loss.
Researchers at Washington University School of Medicine in St. Louis studied 42 participants with obesity, prediabetes and elevated liver fat, randomly assigning them to one of three diets and supplying all their food for four to five months. The findings appeared in Cell Metabolism.
Why controlled feeding makes this trial unusual
Nutrition research is notoriously unreliable, and the reason is almost always the same: people do not eat what they say they eat. Most diet studies depend on self-reported intake, which is systematically inaccurate, and on participants adhering to instructions over months without supervision.
Providing all food for four to five months eliminates that. Researchers know precisely what each participant consumed, which is what allows the central comparison — matched weight loss across diets — to mean anything.
The three diets were ketogenic (4% carbohydrate, 73% fat), Mediterranean (50% carbohydrate) and plant-forward/low-fat (70% carbohydrate). All produced approximately 10% body-weight loss.
The results
The keto group cut liver fat by 67%, against about 45% in each of the other groups — despite equivalent weight loss.
About half of keto participants no longer met prediabetes criteria, versus 29% on the Mediterranean diet and 7% on the low-fat diet.
And the keto group did not show increased blood fats or cholesterol despite consuming 73% of calories as fat — which the researchers called the most surprising result.
Why matched weight loss is the whole point
Liver fat falls when people lose weight, reliably and substantially. That has been established for years, and it is why weight loss is the standard recommendation for fatty liver disease.
The question this trial answers is different: does composition matter beyond the weight change? Because all three groups lost the same amount, the extra 22 percentage points of liver fat reduction in the keto arm cannot be attributed to weight loss. Something about carbohydrate restriction specifically was doing additional work.
“Reducing carbohydrate intake can have important therapeutic effects on metabolism beyond just weight loss alone,” said senior author Samuel Klein.
A plausible mechanism
The likely explanation is straightforward. The liver manufactures fat from excess carbohydrate through a process called de novo lipogenesis, and that pathway is a meaningful contributor to fat accumulation in the liver — particularly with high intakes of certain sugars.
Cutting carbohydrate to 4% of calories largely shuts that production line down. The liver stops making new fat while continuing to export and oxidise what it has, so stores deplete faster than weight loss alone would achieve.
It also explains the counterintuitive lipid result. Dietary fat and blood cholesterol are connected less directly than intuition suggests, and in a setting of substantial weight loss and near-zero carbohydrate, the metabolic effects appear to have offset what the fat intake alone might have produced.
Why liver fat matters
Fatty liver disease has become extremely common alongside rising obesity and type 2 diabetes, and it is not benign. A subset of patients progress to inflammation and scarring, and from there to cirrhosis and liver failure — a leading and growing indication for liver transplantation.
Until recently there was no approved drug, and management meant weight loss and treating associated conditions. A drug has since reached the market for the inflammatory stage, but diet remains foundational, which is why evidence about which diet has practical weight.
The prediabetes result
The spread on prediabetes reversal — roughly 50% versus 29% versus 7% — is arguably more striking than the liver fat figures, and mechanistically coherent: liver fat is closely tied to hepatic insulin resistance, and reducing it improves the liver’s response to insulin.
Whether this represents durable metabolic improvement or a physiological consequence of near-zero carbohydrate intake at the moment of testing is a genuine question. Glucose tolerance can behave unusually on ketogenic diets, and the finding is best interpreted alongside the liver fat data rather than alone.
The caveats
42 people is small. The population was specific — obesity with prediabetes and elevated liver fat — and results may not extend to people without those features.
The controlled feeding that makes the trial rigorous also makes it unrepresentative. Participants received all their food; nobody shopped, cooked, ate at restaurants or navigated a family meal. Free-living adherence to a 4%-carbohydrate diet is considerably harder, and long-term adherence to ketogenic diets is poor in practice.
Cholesterol responses also vary between individuals. Group averages showing no increase do not mean no participant’s lipids rose, and some people do respond to high saturated fat intake with substantial LDL increases.
How to read it
This is a mechanistic result establishing that carbohydrate restriction affects liver fat beyond its effect on weight — not a recommendation that everyone adopt a ketogenic diet. It does not mean keto is right for everyone, and the trial cannot speak to what happens after five months or in people managing their own food. This summarises research and is not medical advice.