In this guide
Keto has been on trial in the court of public opinion for years, and the headlines swing wildly. One week it's a miracle for blood sugar. The next it's a cancer risk, a cholesterol catastrophe or a fad that fades by February. So when a well-controlled randomised trial from Washington University in St. Louis reported at the end of August that a ketogenic diet cut liver fat by two-thirds, the internet did what it always does and picked a side.
We read the university's news release, the coverage and the trial's design, and tried to answer the questions that matter if you're deciding what to eat: how good is the evidence, who is it relevant to, and what doesn't it show?
The short version: in a small, tightly controlled trial of adults with obesity and prediabetes, all three diets (keto, Mediterranean and low-fat plant-forward) produced about 10% weight loss because the researchers set the calories that way. The ketogenic diet cut liver fat by about 67%, versus about 45% on each of the other two, produced the biggest fall in insulin, and returned roughly half of its participants to normal blood sugar (versus about 29% and 7%). LDL cholesterol, ApoB and triglycerides didn't differ between diets. It's encouraging for liver and glucose health. It's also a short, small, supervised trial where all food was provided, so it can't tell us how keto works in real life over years.
Why This Study Spread So Fast
Three things made it irresistible. First, fatty liver is common. Metabolic dysfunction-associated steatotic liver disease (MASLD, formerly NAFLD) affects roughly a third of adults worldwide by many estimates, and most people don't know they have it. Second, the study pitted keto against the Mediterranean diet, the diet most doctors prefer, which sets up an easy "winner" narrative. Third, it's a proper randomised trial with provided food, which is rarer in nutrition research than people assume.
That last point is also why it deserves a careful read. Good design gives it credibility, and small size and short length limit what it can tell us.
The Washington University Trial at a Glance
| Item | Details |
|---|---|
| Where and who | Washington University School of Medicine in St. Louis; first author Max Petersen, senior author Samuel Klein; published in Cell Metabolism, 27 August 2026 |
| Participants | 55 adults with obesity and prediabetes enrolled; 42 completed (about 14 per diet) |
| Design | Randomised trial of three diets, with all meals provided and weekly dietitian visits |
| Calories | Adjusted so that everyone lost about 10% of body weight over roughly 4–5 months |
| Main outcomes | Liver fat, insulin and glucose measures, cardiometabolic risk markers |
| Funding | US National Institutes of Health and the Foundation for Barnes-Jewish Hospital |
| Authors' caution | Small, highly controlled, relatively short; long-term effects unknown |
The Three Diets That Were Compared
| Diet | Carbohydrate | Fat | Character |
|---|---|---|---|
| Ketogenic | About 4% of calories | About 73% | Very low carb; relatively high in saturated fat (reported at about 52 g per 2,000 kcal) |
| Mediterranean | About 50% | About 35% | Olive oil, fish, nuts, legumes, vegetables, whole grains |
| Plant-forward low-fat | About 70% | About 15% | Predominantly plant foods, very low fat |
These aren't loose labels. They're specific macronutrient targets, and every meal was prepared to hit them, which is far more precise than a typical diet study where people are told to "follow" a plan and left to improvise.
The Design Strength: Food Provided, Weight Loss Matched
Most diet comparisons are muddied by a simple problem: people on different diets lose different amounts of weight, so you can't tell whether benefits come from the diet or from the weight loss. Here, the team tailored calorie intake so that everyone lost around 10%. That makes the comparison cleaner, because if outcomes differ, they're less likely to be explained by one group simply losing more.
Providing all food also removes the usual self-report fog. Researchers know what people ate. That's a major strength, and also a reason the results may not transfer directly to someone cooking and shopping alone.
The Results: Liver Fat, Insulin, Glucose and Prediabetes
| Outcome | Ketogenic | Mediterranean | Low-fat plant-forward |
|---|---|---|---|
| Weight loss | About 10% | About 10% | About 10% |
| Liver fat reduction | About 67% | About 45% | About 45% |
| Insulin | Largest fall | Smaller | Smaller |
| Glucagon | Largest rise | Smaller | Smaller |
| Fasting glucose (as reported in coverage) | Roughly twice the fall seen on the other diets | — | — |
| Prediabetes remission | About half (roughly 7 of 14) | About 29% (roughly 4 of 14) | About 7% (roughly 1 of 14) |
In plain English: all three diets worked on weight and liver fat, but keto worked more on the liver and on insulin biology. Prediabetes remission followed the same direction, with keto ahead. The glucagon finding is interesting because glucagon is the hormone that mobilises stored fuel, and its rise alongside falling insulin is exactly the hormonal profile you'd expect to help the liver shed fat.
What Didn't Differ: LDL, ApoB and Triglycerides
This is the line that cholesterol-watchers jumped on. Despite the ketogenic diet's higher saturated fat, the study found no significant difference between diets in LDL cholesterol, ApoB (a measure of atherogenic particles) or 24-hour triglycerides. That's reassuring as an average over the group.
The word "average" matters, though. Individual responses to a high-fat diet vary enormously, and some lean people see LDL rise sharply, as we explain in our guide to the lean mass hyper-responder pattern. With about 14 people per diet, a few strong responders wouldn't move the group mean much. So read this as "no group-level problem in a short trial", not "no one's LDL ever rises on keto".
How to Read "67% Versus 45%"
Three cautions help you interpret headline percentages in a small trial:
- These are relative changes. A 67% fall from a high starting liver fat is a bigger absolute reduction than a 67% fall from a low one. Look for baselines in the paper before judging clinical importance.
- Small groups, wide uncertainty. With around 14 people per diet, a handful of unusually strong responders can shift an average. Confidence intervals, which we haven't seen reproduced in coverage, would show how precise the estimates are.
- "Remission" counts are tiny. "About half" is roughly seven people; "7%" is roughly one. They point in a direction, but they're not rates you can plan your life around.
It's also worth noting that 13 of the 55 enrolled participants didn't finish. That's about a quarter, a reminder that even with food provided and weekly support, strict diets are hard to stick to.
Why Keto Might Lower Liver Fat More
The study offers a plausible mechanism, but the paper doesn't prove it. Very low carbohydrate intake drives insulin down. Lower insulin and higher glucagon shift the body toward releasing and burning stored fat, and reduce the liver's own fat production from carbohydrate (de novo lipogenesis). Secondary coverage reported that this liver pathway was reduced more on keto. Mediterranean and plant-forward diets also lowered liver fat substantially, probably through weight loss, better food quality and more fibre.
The most important message may be the one in the middle: 10% weight loss by any of these diets cut liver fat by nearly half. Keto looked better still, but weight loss itself was doing most of the work in all three arms.
What the Trial Can't Tell You
- Real-life adherence. Everyone received food. Shopping, cooking, eating out and cravings are where most diets succeed or fail.
- Long-term outcomes. A 4–5 month trial can't show effects on heart attacks, cancer, bone health or how long benefits last.
- The weight-loss gap in the real world. Because calories were set to equalise weight loss, the study can't say which diet would lead to more loss when people eat freely.
- Diet quality within keto. A ketogenic diet built from olive oil, fish and vegetables isn't the same as one built from processed meat and butter. The study's menu details matter.
- People without prediabetes. Everyone who finished had obesity and prediabetes. Results may differ in lean, healthy people or in those with established diabetes.
- Individual cholesterol responses. Group averages hide outliers.
The Case for the Mediterranean Plate
Don't write off the Mediterranean diet because keto won one round. It matched the low-fat diet on liver fat, improved glucose control and, most importantly, is the diet with the best long-term evidence behind it. The PREDIMED randomised trial found that a Mediterranean diet supplemented with olive oil or nuts reduced major cardiovascular events by about 30% in higher-risk adults over several years. It's also easier to live with for most people, because nothing is banned.
A fair summary is that keto may offer an extra metabolic edge in the short term under supervision, while the Mediterranean pattern has the deeper long-term record and better sustainability. They aren't mutually exclusive. Many people do well on a lower-carbohydrate Mediterranean-style diet that sits in between.
Who Should Check With a Doctor First
| If you… | Take care because… |
|---|---|
| Take insulin or sulfonylureas | Very low carbohydrate intake can cause low blood sugar; doses often need adjusting |
| Take an SGLT2 inhibitor (such as dapagliflozin or empagliflozin) | Combining these with very low-carbohydrate diets raises the risk of ketoacidosis, a medical emergency; regulators have issued warnings about this |
| Are pregnant, breastfeeding or have type 1 diabetes | Keto isn't generally appropriate without specialist supervision |
| Have gallbladder disease or a history of pancreatitis | A high-fat diet may worsen symptoms |
| Have kidney disease or kidney stones | Dietary changes can affect kidney health; individual advice is needed |
| Have a history of disordered eating | Rigid rules about food groups can feed unhelpful patterns |
| Have very high LDL on a high-fat diet before | You may be a hyper-responder; track your lipids |
If You Decide to Try Keto: A Quality-First Approach
- Get a baseline. Ask your GP about liver enzymes, HbA1c, lipids and, if you have risk factors, liver health. Our body-fat guide and body fat calculator can help you track composition between visits.
- Set your targets deliberately. Use the keto macro calculator and carb calculator, and see our guides to calculating net carbs, fat and protein and the keto macro calculator.
- Choose fats thoughtfully. Lean towards olive oil, nuts, seeds, avocado and oily fish, with plenty of non-starchy vegetables, rather than relying on processed meats.
- Keep a moderate deficit. Use the calorie deficit calculator. The study's benefits came with about 10% weight loss over months, not crash dieting.
- Re-check at 8–12 weeks. Repeat lipids and glucose markers, and discuss the results with your clinician.
- Plan the exit. Long-term data are thin, so decide in advance how you'll move toward a more varied pattern, such as a lower-carbohydrate Mediterranean plate, once your goals are met.
We covered a very different keto study recently, MIT's mouse research on fat metabolism and intestinal tumours. Together, the two are a good reminder that one dietary pattern can look beneficial for one organ in one setting and raise questions in another, and that "keto" is a family of diets, not a single thing.
The Bottom Line
This is a well-designed, honestly reported trial with encouraging news for liver fat and insulin on a ketogenic diet, and with no sign of worse LDL, ApoB or triglycerides on average. It's also a trial of 42 people, over a few months, with all food provided and weight loss deliberately matched. It tells us that keto can be metabolically potent. It doesn't tell us that it's the best diet for you, or that it's safe over years.
For most people with obesity and prediabetes, the first big win is the same on every diet in this study: a steady 10% weight loss. If keto helps you get there and you can keep it up safely, it's a reasonable option with medical input. If a Mediterranean plate suits your life, it did nearly as well on the liver and has far more long-term evidence behind it. The best diet remains the one you can sustain, supervised where it matters, and built from good food.
Frequently Asked Questions
Does the keto diet reduce liver fat more than other diets?
In a 2026 Washington University randomised trial, a ketogenic diet reduced liver fat by about 67%, compared with about 45% on a Mediterranean diet and about 45% on a plant-forward low-fat diet, with all groups losing about 10% of body weight. The trial was small and short.
Did keto raise cholesterol in the trial?
Not on average. LDL cholesterol, ApoB and 24-hour triglycerides didn't differ significantly between the three diets. Individual responses to high-fat diets vary, and with about 14 people per diet, a few strong responders wouldn't shift the average.
Did keto reverse prediabetes?
About half of the keto group (roughly 7 of 14) returned to normal blood sugar, compared with about 29% on the Mediterranean diet and about 7% on the low-fat diet. Those are small numbers, and they apply to a short, supervised trial.
Why did the researchers make everyone lose the same amount of weight?
To isolate the effect of diet composition. If one diet had caused more weight loss, any benefit might simply reflect the extra weight lost. Equalising it makes the comparison cleaner.
Can I do keto if I take diabetes medication?
Only with medical supervision. Insulin and sulfonylureas can cause low blood sugar on very low-carbohydrate diets, and SGLT2 inhibitors combined with such diets carry a risk of ketoacidosis. Never change medication on your own.
Is the Mediterranean diet still better long term?
It has stronger long-term evidence, including the PREDIMED trial's roughly 30% reduction in major cardiovascular events in higher-risk adults. In the Washington University trial it matched the low-fat diet on liver fat. Many people find it easier to sustain.
Is keto safe long term?
Long-term human data are limited, so no one can say with confidence. Many practitioners suggest periodic reassessment, a quality-first approach to fats and an exit plan. A separate mouse study also raised questions about intestinal effects, though it has no human data yet.
Does this mean keto is best for fatty liver?
Not necessarily. It suggests keto can reduce liver fat quickly under supervision, but weight loss by any of the diets cut liver fat substantially. Talk to your clinician about the right approach for your liver health.
How do I work out keto macros safely?
Start with your calories, set protein first, keep net carbohydrates low and fill the remainder with fats from quality sources. Our keto macro calculator and net-carbs guide walk through the steps.
References & Sources
- Petersen, M. C., Klein, S. et al. (2026). Randomized clinical trial of diet macronutrient content and the cardiometabolic response to weight loss (ketogenic, Mediterranean and plant-forward low-fat diets). Cell Metabolism, published 27 August 2026.
- Washington University School of Medicine in St. Louis (27 August 2026). Keto diet has greater health benefits than other popular weight-loss diets, according to clinical trial. EurekAlert news release.
- Secondary coverage: M3 Global Research, Blavity and independent clinician commentary on the trial (September 2026).
- Estruch, R. et al. (2018). Primary prevention of cardiovascular disease with a Mediterranean diet supplemented with extra-virgin olive oil or nuts (PREDIMED). New England Journal of Medicine, 378(25), e34.
- Younossi, Z. M. et al. (2023). The global epidemiology of nonalcoholic fatty liver disease (NAFLD) and nonalcoholic steatohepatitis (NASH): a systematic review. Hepatology, 77(4), 1335–1347.
- US Food and Drug Administration. Drug safety communications on SGLT2 inhibitors and ketoacidosis.
- MacrosMeasure (2026). Free Keto Macro Calculator.
- MacrosMeasure (2026). How to Calculate Net Carbs, Fat, and Protein for a Ketogenic Diet.
Calculate your own numbers — free.
18 science-backed calculators covering everything in this article and more.