A trial published August 27, 2026 in Cell Metabolism found that a ketogenic diet cut liver fat by 67% on average, versus 45% on a Mediterranean diet and 45% on a low-fat, plant-forward diet, even though all three groups lost the same 10% of body weight. Prediabetes reversed in about half the keto group, versus 29% on Mediterranean and 7% on the plant-forward diet.
What did the new keto vs. Mediterranean vs. plant-forward study actually find?
Researchers at Washington University School of Medicine in St. Louis, led by first author Max Petersen and senior author Samuel Klein, enrolled 55 adults with obesity, prediabetes, and fatty liver disease. Forty-two completed the full trial. Participants were randomly assigned to one of three diets for four to five months: a ketogenic diet (about 4% of calories from carbohydrate, 73% from fat), a Mediterranean diet (about 50% carbohydrate, 35% fat), or a very-low-fat, plant-forward diet (about 70% carbohydrate, 15% fat).
The detail that makes this trial worth reading past the headline: researchers individually adjusted each person’s calorie intake alongside their macronutrient targets, so every group lost approximately the same 10% of starting body weight. That design choice is what let them ask a sharper question than “which diet helps people lose more weight.” They asked what happens when weight loss is held equal and only the diet’s composition differs. If you want to see what a comparable deficit looks like against your own numbers instead of a study average, the calorie deficit calculator will build that from your actual stats.
This design choice is also why the study is getting attention beyond the usual diet-comparison coverage. Most head-to-head diet trials report which group lost more weight and stop there, which makes it nearly impossible to tell whether any downstream health improvement came from the diet itself or simply from carrying less body weight. Holding weight loss equal across all three arms removes that confound and isolates a narrower, more useful question: given the same amount of weight lost, does what someone ate to get there still matter for their liver and blood sugar. Funded by the National Institutes of Health and the Foundation for Barnes-Jewish Hospital, the trial had participants meet regularly with dietitians and receive prepared meals built to each diet’s macronutrient targets, an approach designed to keep adherence high during the four-to-five-month study window.
| Diet | Carbohydrate | Fat | Liver fat change | Prediabetes remission |
|---|---|---|---|---|
| Ketogenic | ~4% | ~73% | −67% | ~50% |
| Mediterranean | ~50% | ~35% | −45% | 29% |
| Plant-forward (low-fat) | ~70% | ~15% | −45% | 7% |
Figures as reported across the trial’s press coverage; weight loss was matched at roughly 10% of starting body weight across all three groups. Study: Petersen, Klein, et al., Cell Metabolism, August 27, 2026, Washington University School of Medicine.
Is keto really better for fatty liver disease than other diets?
On the liver-specific measures this trial tracked, the ketogenic group came out ahead by a wide margin. Beyond the headline liver fat number, researchers monitored participants’ metabolism over a full 24-hour window and found the keto group had the largest rise in glucagon, a hormone that promotes fat breakdown in the liver, alongside a roughly 20% drop in 24-hour blood glucose readings compared with about 8% in the other two groups.
That’s a coherent, mechanistically plausible result: a hormone known to mobilize liver fat rose more on the diet that also showed the biggest liver fat drop. It’s still evidence from one trial in a specific population: people who already had obesity, prediabetes, and diagnosed fatty liver disease (sometimes called steatotic liver disease in current medical terminology). It doesn’t yet speak to anyone simply curious about trying keto without those underlying conditions. A result this specific is worth taking seriously without treating it as the final word on which diet is “better” in some universal sense.
It’s also worth being precise about what “better” means here. This trial measured liver fat, prediabetes status, and metabolic markers over four to five months. It did not measure long-term liver disease progression, cardiovascular events, or whether the improvement holds up once a structured, dietitian-supported program ends. A diet that outperforms on one well-measured outcome in one trial is meaningful evidence, not a complete verdict on every dimension that might matter for someone managing fatty liver disease over years of follow-up well beyond this trial’s four-to-five-month window.
How does liver fat connect to prediabetes and insulin sensitivity?
Excess fat stored in the liver interferes with the organ’s response to insulin, part of why fatty liver and prediabetes so often show up in the same person. This trial measured insulin sensitivity in two separate tissues, and the pattern is worth sitting with: muscle insulin sensitivity improved by a similar amount, roughly 50% from baseline, across all three diets. Liver insulin sensitivity told a different story, improving two to three times more in the keto group than in the other two.
That split matters for interpreting the prediabetes numbers. Half the keto group no longer met the criteria for prediabetes after four to five months, compared with 29% on Mediterranean and just 7% on the plant-forward diet. Since prediabetes remission depends heavily on how well the liver, and not only the muscles, responds to insulin, the diet that moved liver insulin sensitivity the most also produced the largest share of people reversing their diagnosis. The three numbers, liver fat, liver insulin sensitivity, and prediabetes remission, move together in this dataset instead of telling three unrelated stories.
| Measure | Ketogenic | Mediterranean | Plant-forward |
|---|---|---|---|
| Muscle insulin sensitivity | ~50% improvement | ~50% improvement | ~50% improvement |
| Liver insulin sensitivity | 2-3x greater improvement | Baseline comparison | Baseline comparison |
| 24-hour blood glucose | ~20% reduction | ~8% reduction | ~8% reduction |
| Prediabetes remission | ~50% | 29% | 7% |
Muscle insulin sensitivity moved by a similar amount on all three diets; liver insulin sensitivity and prediabetes remission did not. That split is the clearest evidence in this trial that diet composition, not just weight lost, shaped the outcome.
What are the limits of this study?
A few limits deserve more than a passing mention, because they change how much weight this single trial should carry.
Small sample, real attrition. Fifty-five adults enrolled; 42 completed the trial and are reflected in the outcome numbers above. Split three ways, that puts each diet arm at roughly 14 people who finished. That’s a workable size for the kind of detailed metabolic monitoring this study ran, tracking hormones and glucose over full 24-hour windows in a controlled setting is expensive and labor-intensive research, but it’s still a small group for drawing broad conclusions, and the researchers behind the study said as much themselves: Scientific American’s coverage quotes them pointing to the trial’s “short timeline and small sample size” as reasons more research is needed.
Short duration. Four to five months shows what happens during active, closely monitored weight loss. It says nothing about what happens at month twelve, month twenty-four, or after participants stop following a structured diet with weekly dietitian support, an element the trial itself provided and that most people trying to replicate this at home won’t have in the same form.
A narrow population. Every participant already had obesity, prediabetes, and diagnosed fatty liver disease going in. That’s precisely the group most likely to show measurable liver improvement from a diet change, and it’s also precisely why the result can’t be assumed to generalize to someone without those conditions, or to someone with a different liver condition entirely.
Still just one trial. This is the first study to compare these three specific diets head to head while deliberately matching weight loss, which is exactly why it’s newsworthy. It also means no independent research group has yet reproduced the finding. A single well-designed trial is genuinely useful evidence. It is not the same thing as a settled scientific consensus, and framing it that way would overstate what one study, however carefully controlled, can establish on its own.
Support most people won’t have at home. Participants met regularly with dietitians and had prepared meals available during the trial. That level of structure tends to raise adherence to any diet, keto included, well above what someone following the same macronutrient targets alone at home is likely to sustain. A result built on closely supported adherence doesn’t automatically transfer to someone building the same eating pattern without that scaffolding.
None of this makes the finding uninteresting. It’s a rare example of a nutrition trial holding weight loss constant on purpose, which is exactly what lets a diet-composition claim like this one be tested at all instead of just asserted. It just isn’t, on its own, the diet debate’s final answer.
Does this mean you should switch to keto?
That’s an individual medical decision, not one this article is positioned to make. What the evidence here supports is narrower and more useful than a blanket recommendation: for adults with obesity, prediabetes, and existing fatty liver disease, a structured ketogenic diet produced a larger reduction in liver fat and a higher rate of prediabetes remission than a Mediterranean or plant-forward diet over four to five months, on top of matched weight loss.
What it doesn’t support is a general claim that keto is the healthiest diet, that it’s the right choice for someone without these conditions, or that anyone with an existing liver disease, prediabetes, or diabetes should change their eating pattern without medical guidance. A very-low-carbohydrate diet also carries its own considerations around sustainability, nutrient variety, and interaction with certain medications, none of which this trial was designed to evaluate. Anyone weighing a major diet change for an existing liver condition or blood sugar issue should bring this study to a doctor or registered dietitian, who can weigh it against personal labs, medications, and history in a way a single trial’s averages cannot.
It’s also worth remembering that this trial compared three specific, clearly defined diets against each other. It didn’t test any of them against “doing nothing,” or against every possible variation of low-carbohydrate eating. A loosely followed low-carbohydrate diet at home, without the structured calorie adjustment and dietitian support the trial provided, is a different intervention than the one these numbers describe, and there’s no data here to say how closely a home version would track the trial’s results.
Building a Mediterranean, low-carb, or plant-forward plan on Calorilator
Whichever direction this research leaves you curious about, building the pattern itself is the more useful next step than debating diet labels in the abstract. Calorilator’s Mediterranean meal plan and low-carb meal plan tools generate a real 7-day plan around each pattern’s macro targets, and the macro calculator linked above will translate any of these percentage splits into grams and calories for your own numbers, not a generic study average.
The plant-forward arm in this trial leaned heavily on carbohydrate from whole foods over refined ones, a distinction our whole vs. processed foods explainer covers in more depth: a plant-forward diet built around vegetables, legumes, and whole grains behaves differently in the body than one built around packaged, ultra-processed carbohydrate sources, even at similar macronutrient percentages. That same distinction applies underneath all three diet arms in this study. A ketogenic or Mediterranean plan built around whole ingredients is a different proposition than the same macro split filled in with convenience food, and it’s worth keeping that in mind whichever pattern you build toward next.