IPM Take
Obesity policy still tends to treat weight loss as the principal therapeutic endpoint, but this study suggests that two patients can lose the same proportion of body weight and experience meaningfully different metabolic responses depending on how that weight loss is achieved.
In adults selected specifically because they had obesity, prediabetes and hepatic steatosis, all three diets produced roughly 10% weight loss and approximately 50% improvement in skeletal-muscle insulin sensitivity. However, the very-low-carbohydrate ketogenic diet produced a two- to three-fold greater improvement in hepatic insulin sensitivity, reduced liver fat by 67% compared with about 45% on the other diets, and produced larger improvements in several measures of glycaemic control.
The personalised-medicine implication is not that everyone with obesity should be prescribed keto. It is that nutrition may need to become more phenotype-specific. A patient with prominent hepatic steatosis and insulin resistance may have different therapeutic priorities from someone whose dominant problem is hypertension, dyslipidaemia, sarcopenia or cardiovascular disease.
The policy challenge is that clinical nutrition is rarely funded or organised with that degree of personalisation. Health systems increasingly pay for sophisticated obesity medicines while dietetic care often remains standardised, time-limited or difficult to access. If diet composition can produce benefits beyond weight loss itself, nutrition becomes part of precision treatment rather than an instruction to “eat better.”
Executive Summary
Researchers at Washington University School of Medicine compared three markedly different dietary strategies in adults with metabolically unhealthy obesity, defined in this study by the presence of prediabetes and hepatic steatosis. Fifty-five participants were randomized, with 42 completing the controlled weight-loss intervention and contributing to the main comparative analysis. All food was supplied by the research team, allowing the investigators to produce closely matched weight loss of approximately 10% while varying dietary macronutrient composition.
The ketogenic diet provided roughly 4% of energy from carbohydrates and 73% from fat. The Mediterranean diet provided approximately 50% of energy from carbohydrate and 35% from fat, while the plant-forward diet provided around 70% from carbohydrate and 15% from fat.
Muscle insulin sensitivity improved by about 50% across all three groups, indicating that this benefit was largely associated with weight loss itself. Liver metabolism behaved differently. Hepatic insulin sensitivity improved two to three times more with the ketogenic diet, while intrahepatic triglyceride content fell 67%, compared with approximately 45% with the Mediterranean and plant-forward diets. Twenty-four-hour glucose fell about 20% with keto versus 8% with the other diets.
At the study endpoint, 50% of participants in the ketogenic group no longer met prediabetes criteria, compared with 29% in the Mediterranean group and 7% in the plant-forward group. Importantly, LDL cholesterol, apolipoprotein B and 24-hour triglyceride concentrations did not differ significantly among diets during the study.
The trial was small, highly controlled and short term, so it cannot establish that ketogenic diets are superior for long-term cardiovascular outcomes or suitable for all patients.
Why it matters
- HTA bodies: Lifestyle interventions are often treated as homogeneous comparators in obesity and metabolic-disease assessments. These data suggest that dietary composition can influence intermediate metabolic outcomes independently of the amount of weight lost, which may become relevant when evaluating integrated obesity and MASLD care pathways.
- Payers: Obesity treatment increasingly includes high-cost pharmacotherapy, yet access to intensive dietetic support remains inconsistent. If different dietary strategies provide different benefits according to metabolic phenotype, reimbursement models may need to support personalised nutritional management alongside medicines rather than treating diet as a generic prerequisite for drug coverage.
- Industry / innovation partners: The next phase of obesity care may combine pharmacotherapy with phenotype-guided nutrition. The opportunity lies not in declaring one universal “best diet,” but in identifying which metabolic characteristics predict additional benefit from different dietary strategies and integrating those signals into personalised treatment pathways.
The most interesting finding from this trial is not that a ketogenic diet produced more weight loss. It did not.
Researchers deliberately designed the study so that adults following three very different diets lost approximately the same amount of body weight, around 10% of their starting weight. That allowed them to ask a question that is difficult to answer in ordinary diet studies: when weight loss is held approximately constant, does the composition of the diet still matter metabolically?
The answer appears to be yes, at least for the specific population studied.
Participants had obesity together with prediabetes and excess liver fat, placing them at increased risk of type 2 diabetes, metabolic liver disease and cardiovascular complications. They were randomized to a very-low-carbohydrate ketogenic diet, a Mediterranean diet or a very-low-fat plant-forward diet, with the research team supplying all meals and participants meeting regularly with dietitians.
All three approaches worked in one important respect. With similar weight loss, skeletal-muscle insulin sensitivity improved by roughly 50% across the groups, supporting the familiar clinical message that moderate weight reduction itself can substantially improve metabolic health.
The liver told a different story.
Participants following the ketogenic diet experienced a two- to three-fold greater improvement in hepatic insulin sensitivity than those on the Mediterranean or plant-forward diets. Liver fat declined by 67%, compared with approximately 45% in the other two groups, while 24-hour glucose and insulin profiles improved more strongly with carbohydrate restriction.
That distinction matters because cardiometabolic disease is heterogeneous. Two people with the same BMI may carry very different patterns of visceral adiposity, hepatic fat, insulin resistance, dyslipidaemia and cardiovascular risk. A treatment strategy based exclusively on the number of kilograms lost may therefore miss biologically important differences in how individual patients respond.
From weight-loss diets to metabolic phenotypes
The findings fit naturally with the direction of personalised medicine, although the evidence is not yet strong enough to justify phenotype-based dietary guidelines.
For a patient with marked hepatic steatosis, hyperinsulinaemia and prediabetes, improving liver metabolism may be a particularly important therapeutic objective. For another patient, LDL cholesterol, blood pressure, kidney disease, frailty or long-term dietary sustainability may matter more. The “best” dietary strategy could therefore depend partly on which disease process clinicians are trying to modify.
This becomes especially relevant as GLP-1-based therapies transform obesity care. The study’s investigators explicitly note that many of the patients represented in the trial would now also be candidates for anti-obesity pharmacotherapy, but argue that dietary composition can still influence metabolic outcomes beyond the weight loss itself.
That challenges a simplistic future in which pharmacotherapy produces the weight loss and nutrition becomes secondary. Instead, medication and diet may eventually be used together, with the drug helping achieve or maintain weight reduction while nutrition is selected according to the patient’s dominant metabolic abnormalities.
The trial also provides an important counterpoint to concerns that the high fat content of ketogenic diets must automatically worsen conventional lipid risk. During this short controlled intervention, researchers found no significant between-group differences in LDL cholesterol, apolipoprotein B or 24-hour triglyceride concentrations.
That finding should not be interpreted as proof of long-term cardiovascular safety. The study was not designed or powered to measure myocardial infarction, stroke, cardiovascular mortality or long-term atherosclerotic risk, and its duration was measured in months rather than years.
Precision nutrition will need more than a diet label
There are equally important practical limitations. Only 42 participants completed the comparative intervention, the food was supplied by investigators, adherence was unusually intensive and the population was deliberately selected for obesity, prediabetes and fatty liver. These conditions are useful for understanding biological effects, but they are very different from everyday care, where cost, culture, appetite, family circumstances and long-term adherence determine whether a nutritional strategy survives beyond the consultation room.
Nor does the study prove that ketosis itself was necessary. It remains possible that less severe carbohydrate restriction could deliver some of the hepatic benefit while being easier to sustain, a question that will require further trials.
This is therefore not a mandate to replace Mediterranean diets with ketogenic diets. It is evidence against another form of one-size-fits-all thinking.
Weight loss remains important, but the route to weight loss may influence which metabolic abnormalities improve most. For personalised cardiometabolic medicine, that means the relevant treatment endpoint may eventually become broader than body weight alone.
The policy consequence follows naturally. If healthcare systems want precision obesity care, they cannot personalise only the prescription pad while leaving nutrition generic. They will need the dietetic workforce, monitoring tools and reimbursement structures that allow dietary treatment to be adapted to the metabolic phenotype and circumstances of the individual patient.
