In nutrition research, most trials converge on the same conclusion: calorie deficit decides the outcome, and the specific diet is just packaging. But a Washington University School of Medicine trial published August 27 in *Cell Metabolism* breaks that pattern — three diet groups lost the same amount of weight, yet liver outcomes favored only one. The ketogenic group cut intrahepatic fat by an average of 67%, while the Mediterranean and plant-based low-fat groups both landed at 45%.
Trial design
- Population: metabolically unhealthy adults with obesity, prediabetes, and fatty liver disease
- Three arms: ketogenic (4% carbs / 73% fat), Mediterranean (50% / 35%), plant-based low-fat (70% / 15%)
- Full feeding: 100% of food provided by the lab for five months, with individualized calorie allocation
- Intrahepatic fat reduction: keto -67%, Mediterranean -45%, plant-based low-fat -45%
- 24-hour insulin: -74% on keto (vs -44% and -27% in the other arms)
- The largest reduction in hepatic DNL (de novo lipogenesis — the liver's on-site conversion of carbohydrate into fat), plus the largest glucagon rise
- Hepatic insulin sensitivity improved 2- to 3-fold (p<0.001); HbA1c fell the most
- No differences in LDL, apoB, or 24-hour triglycerides across groups — don't celebrate the cardiovascular angle yet; this trial gives it no award
- WashU Medicine press release (2026-08-27)
- ScienceDaily republication
- EurekAlert
- Paper: *Cell Metabolism*, DOI 10.1016/j.cmet.2026.07.020
That last detail is the entire cleverness of the design: calories were calibrated so weight loss was deliberately equalized at ~10% in all three groups. Once the scale reads the same, any difference in liver outcomes can only come from *what* participants ate, decoupled from *eating less*. Most diet trials can't do this, because weight loss never syncs up — changes in liver fat get mixed with contributions from weight loss, and you can't untangle the two.
Results
Caveats, ranked by importance
1. The population was metabolically unhealthy adults with obesity (obesity plus prediabetes plus fatty liver) — not healthy people. That's both a limitation and a selling point: it matches the fatty liver demographic, but there's no direct extrapolation to lean, healthy individuals. 2. The sample was small, and the two press releases disagree: the WashU release says 55 randomized, the Cell Press release says 42. Likely 55 enrolled and 42 completed, but neither states it explicitly. Unverified — the only number I couldn't fully confirm. 3. Five months only; no long-term outcomes. 4. The imaging modality for liver fat measurement isn't specified in public materials (presumably MRI-based). Unknown — waiting on the full text. 5. Corresponding author Samuel Klein has consultancy relationships with multiple pharma companies (AbbVie, 89Bio, Boehringer Ingelheim, Merck, Viking) — disclosed per convention, but readers should know.
The authors acknowledge their own limits: the underlying mechanisms aren't fully understood (does GLP-1 play a role?); whether carbohydrate restriction must actually induce ketosis to work, and whether the benefit must occur during active weight loss, remain unanswered.
Why it matters
Fatty liver affects a striking share of the population — first author Max C. Petersen (assistant professor) said in interviews that roughly 75% of adults with obesity have fatty liver (as quoted; independently unverified). In the GLP-1 era, the default assumption is that weight loss equals the entire therapeutic benefit. This trial says: hold on — with identical weight on the scale, organ-level outcomes can differ. For people with existing fatty liver, the carbohydrate side of the diet may be the real lever.
I plan to wait for replication before changing my own diet — nutrition science has fooled us with single-arm studies too many times. That said, a fully controlled feeding, randomized, three-arm trial with mechanistic biomarkers is the ceiling of evidence for a dinner-table question. The remaining question is whether that ceiling is high enough.