In most nutrition research, the conclusion is the same: the calorie deficit decides the outcome, and the specific diet is just packaging. A trial from Washington University School of Medicine, published August 27 in *Cell Metabolism*, refuses to cooperate — three groups lost the same amount of weight, but the liver only rewarded one of them. The ketogenic group saw intrahepatic fat fall by an average of 67%, while the Mediterranean and plant-based low-fat groups both achieved 45%.
Study design
The cleverness lies entirely in the design:
- Participants: metabolically unhealthy adults with obesity — prediabetes plus fatty liver disease
- Three arms:
- Ketogenic: 4% carbohydrate / 73% fat
- Mediterranean: 50% / 35%
- Plant-based low-fat: 70% / 15%
- Protocol: 100% of food provided by the lab for five months, with individualized calorie rationing that deliberately equalized weight loss at ~10% in all groups
- WashU Medicine press release
- ScienceDaily
- EurekAlert
- Paper: *Cell Metabolism*, DOI 10.1016/j.cmet.2026.07.020
This equalization is the key trick: the scale showed the same numbers, but liver outcomes diverged — so the difference must come from *what* was eaten, decoupled from *eating less*. Most diet trials can't do this, because weight loss is never synchronized, and liver fat changes end up confounded with weight loss effects.
Liver fat results
| Diet | Mean intrahepatic fat reduction | |---|---| | Ketogenic | 67% | | Mediterranean | 45% | | Plant-based low-fat | 45% |
Proposed mechanism
The paper offers a fairly complete mechanistic chain:
1. Carbohydrate cut to 4% → 24-hour insulin fell 74% (vs -44% Mediterranean, -27% low-fat); glucagon rose the most 2. Greatest suppression of hepatic de novo lipogenesis (DNL — the liver's process of synthesizing fat from carbohydrate) 3. Hepatic insulin sensitivity improved 2–3 fold (p<0.001); HbA1c also fell the most 4. Result: hepatic fat -67%
Notably on the flip side: LDL, apoB, and 24-hour triglycerides showed no differences across the three groups — don't get excited about the cardiovascular line, this trial didn't award it any prize.
The first author is Max C. Petersen (assistant professor); the corresponding author is Samuel Klein (Danforth Professor, a veteran in this field).
Caveats, ranked by importance
1. Population: metabolically unhealthy adults with obesity (obesity + prediabetes + fatty liver), not healthy people. This is both a limitation and a selling point — it precisely matches the fatty liver population — but it doesn't extrapolate to lean or healthy individuals. 2. Small sample, conflicting press releases: the WashU release says 55 randomized, the Cell Press release says 42. Likely 55 enrolled, 42 completed, but neither states it clearly. This is the only figure not fully verified. 3. Five months only, no long-term outcomes. 4. Imaging method for liver fat (presumably an MRI-based protocol) is not specified in the public materials. Unknown — awaiting the full text. 5. Klein has consultant relationships with multiple pharma companies (AbbVie, 89Bio, Boehringer Ingelheim, Merck, Viking). Disclosed per convention, but readers should know.
The authors' own stated limitations are candid: the underlying "basic mechanisms" aren't fully understood (does GLP-1 play a role?); whether carbohydrate restriction must induce "ketosis" to work, and whether the benefit must occur during active weight loss, both remain unanswered.
Why it matters
The fatty liver population is enormous — Petersen said in an interview that roughly 75% of adults with obesity have fatty liver disease (as quoted; not independently verified). In the GLP-1 era, the default assumption is "weight loss is the entire benefit." This trial says: wait — with identical scale numbers, organ-level outcomes can differ. For people with fatty liver, the carbohydrate side of the equation may be the real lever.
I plan to wait for replication before changing my own diet. Nutrition science has been burned by single-arm studies too many times. That said: provided meals, randomized, three arms, full mechanistic markers — for questions about the dinner table, this is the ceiling of evidence quality. The remaining question is whether the ceiling is high enough.