Nearly every diet comparison has the same flaw: the winning group loses more weight, so you cannot tell whether the diet did anything beyond causing the weight loss. A trial published in Cell Metabolism on 27 August 2026 removed that problem — and the results are more interesting because of it.
What they did
Researchers at WashU Medicine randomly assigned 55 adults with metabolically unhealthy obesity to one of three diets for about five months:
- A low-carbohydrate, high-fat ketogenic diet
- A high-carbohydrate, low-fat, plant-forward diet
- A Mediterranean diet, sitting between the two
The critical design feature: all three groups lost roughly 10% of their starting weight. The weight loss was equalised. Any difference that remained had to come from the composition of the diet rather than the amount of weight lost.
What they found
| Outcome | Ketogenic | Other two diets |
|---|---|---|
| Liver fat reduction | 67% | 45% |
| Liver insulin sensitivity | 2–3× greater improvement | baseline improvement |
| 24-hour blood glucose | −20% | −8% |
| Prediabetes remission | 50% | 29% Mediterranean · 7% plant-forward |
One result cuts against a simple “keto wins” reading: muscle insulin sensitivity improved by about 50% in all three groups equally. The diets were interchangeable there. The advantage was specific to the liver.
Why the liver, specifically
The finding fits what is already understood about hepatic metabolism. The liver converts excess carbohydrate into fat through de novo lipogenesis. Cut carbohydrate sharply and you cut the substrate for that process directly, while ketosis increases fat oxidation.
This matters because fatty liver disease — now called MASLD — affects roughly a quarter of adults worldwide, is closely tied to insulin resistance, and has very few effective treatments. Hepatic insulin resistance is also what allows the overnight glucose release behind high morning blood sugar.
The limits, stated plainly
Fifty-five people. This is a mechanistic trial, not a population study. It shows what happens under controlled conditions, not what happens across a country.
Five months. Long enough to move liver fat, nowhere near long enough to say anything about heart attacks, cancer or mortality — the outcomes on which the long-running debate about ketogenic diets actually turns.
Adherence in a trial is not adherence in life. Participants in feeding studies receive support most people never get. Ketogenic diets have a well-documented attrition problem outside research settings, and a diet followed for five months and abandoned in the sixth is not the diet that was tested.
It says nothing about LDL cholesterol. High-saturated-fat ketogenic diets raise LDL substantially in a subset of people. A trial reporting liver outcomes does not address cardiovascular risk, and those two things can move in opposite directions in the same person (why LDL still matters).
The comparison diets also worked. A 45% reduction in liver fat is a substantial result. The headline is a difference between three effective interventions, not between one that works and two that fail.
What it reasonably supports
For someone with fatty liver disease or prediabetes specifically, this is real evidence that carbohydrate restriction adds something beyond the weight loss — which was the honest open question, and is now better answered.
It does not establish that ketogenic diets are superior for general health, for weight loss itself (all groups lost the same), or for anyone without a metabolic problem. And the Mediterranean diet retains something no ketogenic diet has: long-term randomised evidence for cardiovascular events.
If you want the non-dietary levers, the evidence is strongest for resistance training, walking after meals and reducing liquid sugar (the full protocol, and why carb quality matters).
When to see a doctor
Ask about liver enzymes and a fatty liver assessment if you have prediabetes, type 2 diabetes, obesity, or elevated waist-to-height ratio — MASLD is usually silent until it is advanced. Talk to a doctor before starting a ketogenic diet if you take insulin or a sulfonylurea, where the risk of hypoglycaemia requires dose adjustment; if you have type 1 diabetes; if you are pregnant; or if you have kidney or established liver disease.
Anyone starting a ketogenic diet should have a lipid panel before and about three months after, because the LDL response varies enormously between individuals.
The honest bottom line
This trial did the hard thing: it matched weight loss across three diets, so the differences that remained belong to the diets themselves. The ketogenic arm cut liver fat by 67% against 45%, tripled the improvement in liver insulin sensitivity, and reversed prediabetes in half of participants. That is a genuine finding about the liver. It is 55 people over five months, it says nothing about cholesterol or long-term outcomes, and the other two diets worked too.
Frequently asked questions
Does a keto diet reverse fatty liver disease?
This randomised trial found a 67% reduction in liver fat on a ketogenic diet compared with 45% on Mediterranean and plant-forward diets, with weight loss matched across all three groups at about 10%. That supports carbohydrate restriction having an effect on the liver beyond weight loss alone. It involved 55 people over five months, so it is mechanistic evidence rather than proof of long-term reversal, and the comparison diets produced substantial reductions too.
Is keto better than the Mediterranean diet?
For liver fat and hepatic insulin sensitivity in this trial, yes. For muscle insulin sensitivity, no — all three diets improved it by about 50% equally. And for long-term cardiovascular outcomes the Mediterranean diet has randomised evidence that ketogenic diets do not. Better depends entirely on which outcome you are asking about, and this trial answered a narrow question well.
Does keto raise cholesterol?
It can, substantially, in a subset of people, particularly when the diet is high in saturated fat. This trial reported liver and glucose outcomes and did not address LDL cholesterol, so it cannot reassure you on that point. Anyone starting a ketogenic diet should have a lipid panel before beginning and again after about three months, because individual LDL responses vary widely.