Science for a Healthier Tomorrow
Keto’s liver-fat advantage held up — and stayed liver-specific
All three diets produced a similar ~50% muscle-insulin gain and no significant LDL differences; keto’s 67% liver-fat drop is a relative figure at matched weight loss.
What this means
The 67% figure is a relative change measured after every participant had lost about 10% of body weight on provided meals. Within this trial, the ketogenic diet produced a larger improvement in liver fat and hepatic insulin sensitivity than the Mediterranean and low-fat comparison diets, while all three diets produced the same gain in muscle insulin sensitivity and showed no separation in LDL cholesterol or apolipoprotein B. An earlier, shorter trial of carbohydrate restriction in fatty liver had already pointed in the same direction. What the trial cannot support is the claim that a ketogenic diet is the better diet for metabolic health in general, that these numbers transfer to self-directed eating outside a research kitchen, or that the liver advantage persists once weight loss stops — the last of these being an unknown the authors state themselves: whether the liver advantage persists during weight-loss maintenance. And with 14 people per arm, small changes in who was counted could move several of these percentages substantially.
The trial behind last month’s viral “keto cut liver fat by 67%” headline does support its number — as a relative change in liver fat, in a small trial of people with obesity, prediabetes, and fatty liver who ate study-provided meals and lost about 10% of their body weight. The one-line summaries drop that premise, and they also leave out the trial’s own co-primary results: muscle insulin sensitivity rose by roughly 50% in all three diet groups, and LDL cholesterol and apolipoprotein B showed no differences among the diets. The ketogenic advantage this trial found was specific to the liver and the metabolic measures around it.
What the trial did
Researchers at Washington University School of Medicine in St. Louis randomized 55 adults to eat one of three diets — very-low-carbohydrate ketogenic, Mediterranean, or very-low-fat plant-forward — with all food provided for about four to five months. Participants had a body mass index of 30 to 50, prediabetes, and at least 5% liver fat on MRI; they were 18 to 55 years old, with no diabetes, prior bariatric surgery, menopause, or weight-affecting medications. Fourteen participants per arm — nine women and five men in each, all the women premenopausal — completed and were analyzed, with weight loss of 10.4%, 10.2%, and 10.1% of body weight across the arms. The trial ran from February 2016 to April 2025 under registration NCT02706262, with the two insulin-sensitivity measures prespecified as co-primary outcomes. Provided meals and adherence above 95% make this a high-precision test of macronutrient composition at matched weight loss — a research setting, and a distinct one from what self-directed diets achieve.
Where the diets separated: the liver
Liver fat, measured by MRI, fell in all three arms — by about 67% in the ketogenic group versus about 45% in each comparison arm (P<0.05). Those figures are relative to each participant's own starting liver fat. The hepatic insulin sensitivity index, a co-primary outcome, rose two to three times more in the ketogenic arm (P<0.001), a result that held in a sensitivity analysis including all 55 randomized participants. Hepatic de novo lipogenesis — the liver's manufacture of new fat — and basal endogenous glucose production fell in the keto and Mediterranean arms, stayed flat in the low-fat arm, and fell more in keto than in either (P<0.001). The change in de novo lipogenesis tracked the change in 24-hour insulin exposure (r=0.725, P<0.001). Several secondary markers moved furthest in the keto arm: HbA1c, fasting triglycerides, and large VLDL particles all decreased most in that group, while HDL cholesterol fell in the Mediterranean and low-fat arms and tended to rise in keto (P=0.005).
Where they didn’t: muscle and cholesterol
Skeletal muscle insulin sensitivity, the other co-primary outcome, rose by about 50% in all three arms with no between-group difference (P=0.617). LDL cholesterol and apolipoprotein B “tended to decrease after weight loss, without differences among diet groups,” in the paper’s wording; 24-hour plasma triglycerides likewise showed no differences among diet groups. On the trial’s primary measures, all three diets performed alike.
“Half” is seven people
The press release’s “about half of the participants” who no longer qualified as having prediabetes refers to 7 of the 14 analyzed participants in the ketogenic arm, versus 4 of 14 on the Mediterranean diet and 1 of 14 on the very-low-fat diet (P<0.05). With 14 people per arm, each participant moves that arm's percentage by about 7 points, and the paper gives no confidence interval for these proportions.
Limits worth carrying
All reported outcomes are surrogate measures — clamp-derived insulin sensitivity, MRI liver fat, isotope-traced liver metabolism; the trial measured no clinical events such as cardiovascular disease or new diabetes diagnoses. The analysis is small — 42 of 55 randomized participants, 14 per arm — and the authors state that limited power may have missed between-group differences. The trial was open-label, with participants knowing which diet they ate, and its results come from a provided-food research setting with adherence above 95%. The population was narrow: ages 18 to 55, BMI 30 to 50, prediabetes with fatty liver, nine women and five men per arm. No serious adverse events occurred; estimated kidney function (eGFR) rose in the keto arm (P=0.002), an increase driven by falling cystatin C rather than creatinine, urinary nitrogen excretion rose in the keto arm (P<0.001), and the oxidative-stress marker 8-isoprostane fell (P=0.005). The authors themselves note they cannot determine whether a partial carbohydrate reduction short of full ketosis would reproduce the liver effects, or whether the benefits persist during weight-loss maintenance. Funding came from NIH grants and the Foundation for Barnes-Jewish Hospital, with no industry funding of the trial itself; co-author Samuel Klein reports scientific advisory-board fees from AbbVie, 89Bio, and Boehringer Ingelheim, an investigator-initiated grant from Merck, and support for an industry-initiated trial from Viking Therapeutics.
Sources
- Max C. Petersen, Gordon I. Smith, Sarah S. Farabi, Hector H. Palacios, Mahalakshmi Shankaran, Marc K. Hellerstein, Bruce W. Patterson, Samuel Klein. “Effect of diet macronutrient content on the cardiometabolic response to weight loss: A randomized clinical trial.” Cell Metabolism, published Aug 27, 2026. DOI: 10.1016/j.cmet.2026.07.020 — publisher full text · PubMed PMID 42660124 · NIH manuscript version PMC13551625
- Trial registration: ClinicalTrials.gov NCT02706262 (Washington University School of Medicine; Feb 2016–Apr 2025)
- Press release: “Keto diet cut liver fat by 67% in a clinical trial” — ScienceDaily (Washington University School of Medicine in St. Louis release)
- Context: Browning et al. “Short-term weight loss and hepatic triglyceride reduction: evidence of a metabolic advantage with dietary carbohydrate restriction.” Am J Clin Nutr, 2011. PMID 21367948
