Does Keto Improve Insulin Sensitivity Without Weight Loss?

Keto can improve blood-sugar markers, especially in type 2 diabetes. The harder question is whether it improves insulin sensitivity even when body weight stays the same. Two 2024 randomized trials tried to answer that question with euglycemic-hyperinsulinemic clamps, the best available test of how tissues respond to insulin. They reached different conclusions.

One tightly controlled 10-day trial found no improvement in liver or muscle insulin sensitivity despite confirmed nutritional ketosis and strict weight maintenance. A three-week crossover trial found greater insulin-stimulated glucose disposal in skeletal muscle, but participants lost an average of 2.2 kg despite an isocaloric prescription. Keto clearly changes glucose metabolism. The evidence does not yet show that ketosis itself reliably sensitizes the body to insulin independent of weight loss.

Why weight control changes the question

Many ketogenic-diet studies combine several interventions at once: carbohydrate restriction, lower insulin exposure after meals, spontaneous reductions in calorie intake, and weight loss. Weight loss alone can improve insulin sensitivity, so a study cannot credit keto's macronutrient pattern for the change unless it accounts for body-weight change.

That is why clamp trials with controlled calories matter. During a clamp, researchers infuse insulin and adjust glucose infusion to hold blood glucose steady. The more glucose required to maintain that target, the more insulin-sensitive the body is under the conditions tested. Glucose tracers can also separate liver glucose output from glucose disposal in peripheral tissues, chiefly skeletal muscle.

A strict 10-day keto trial found no insulin-sensitivity gain

Merovci et al. (2024) enrolled 29 adults with obesity and type 2 diabetes. Participants had BMI values of 27.5 to 42 and HbA1c values of 7.0% to 10.5%. They ate a ketogenic diet containing 5% to 10% carbohydrate and 70% to 80% fat for 10 days. A metabolic kitchen supplied the food, researchers monitored weight daily, and the intervention aimed to maintain body weight rather than produce a calorie deficit.

The study used a two-step clamp with [3-³H]-glucose tracing. The lower-insulin stage assessed suppression of endogenous glucose production, a measure of hepatic insulin sensitivity. The higher-insulin stage assessed total glucose disposal, a standard measure dominated by skeletal muscle insulin sensitivity. Neither endpoint improved after 10 days.

The investigators also tested whether higher circulating beta-hydroxybutyrate (BHB), the main ketone body measured in nutritional ketosis, could do the job on its own. Participants received 8 g of BHB ester every eight hours, producing plasma BHB concentrations above those reached with diet alone. That arm also showed no improvement in hepatic glucose-production suppression or total glucose disposal. The ester added about 122 kcal per day without a matched control, a minor design limitation, but it does not explain a missing insulin-sensitivity benefit.

The trial did find a significant increase in insulin secretion capacity. That is a separate endpoint, not evidence that muscle or liver became more insulin-sensitive. It does, however, raise the possibility that keto could affect glucose control through beta-cell function even when clamp-measured insulin sensitivity stays unchanged. That secondary finding needs replication.

A three-week trial found a muscle-specific signal, with a catch

Luong et al. (2024) studied a different group: adults with obesity but without diabetes, with BMI values of 28 to 40 and HbA1c below 48 mmol/mol. In a randomized crossover design, each participant completed a three-week ketogenic diet period and a control-diet period. The ketogenic diet prescribed 5% carbohydrate, 20% protein, and 75% fat, with individualized calorie targets intended to be isocaloric.

At the end of each period, the researchers used a single-step hyperinsulinemic-euglycemic clamp at 0.6 mU/kg/min, plus glucose and palmitate tracers. The ketogenic diet increased the M-value, reflecting greater insulin-stimulated glucose disposal. Tracer analysis did not show a significant improvement in hepatic insulin sensitivity, which localized the observed benefit to skeletal muscle rather than the liver.

That result is important, but it does not settle the weight-independent claim. Participants lost a mean 2.2 kg during the keto period. The trial did not use doubly labeled water or an equivalent measure to verify that energy intake and expenditure truly matched. A randomized trial in older adults with obesity found that weight-loss interventions can improve clamp-measured insulin sensitivity, although that study's six-month diet-and-exercise interventions cannot assign a specific effect to a 2.2 kg loss in this population.

The crossover design reduces noise from differences between people. Yet a calorie prescription is not the same as verified weight maintenance. The muscle result may reflect keto-specific adaptation, modest weight loss, or both.

Why the two clamp trials may disagree

It is tempting to call 10 days too short and three weeks long enough. The comparison supports that hypothesis, but it cannot prove it. The trials differed in at least four consequential ways.

• Merovci et al. studied people with established type 2 diabetes and HbA1c up to 10.5%; Luong et al. studied people with obesity without diabetes.
• The first trial maintained weight for 10 days; the second ran for three weeks and included mean weight loss of 2.2 kg.
• The studies used different clamp protocols: a two-step design in the diabetes trial and a single-step design in the obesity trial.
• Their insulin doses sampled different parts of the insulin dose-response curve, so their M-values should not be compared as if they were interchangeable.

A longer weight-maintaining trial, ideally lasting four to 12 weeks and using a comparable tracer clamp protocol in both populations, would answer the question more cleanly. It has not yet been done.

Ketones change fuel use, but that is not the same as insulin sensitivity

The BHB ester result argues against a simple story in which higher ketones directly make muscle more responsive to insulin. Dörner et al. (2024) reached a compatible conclusion in healthy participants: endogenous and exogenous ketones shifted respiratory exchange ratio, indicating a change in substrate oxidation away from glucose, but did not produce clamp-level improvement in insulin sensitivity.

Changing fuel preference is real physiology. It does not automatically improve insulin signaling. This distinction matters for ketone supplements. Current human clamp evidence does not support expecting a BHB drink or ester to reproduce any insulin-sensitivity effect of a ketogenic diet.

There are plausible pathways worth investigating. Kim et al. (2022) found that a short-term isocaloric ketogenic diet modulated the NLRP3 inflammasome through BHB and FGF21 in humans. Because chronic low-grade inflammation accompanies metabolic dysfunction, this could contribute to later changes in insulin action. The study demonstrated an inflammatory-pathway effect; it did not demonstrate improved muscle insulin signaling or clamp-measured insulin sensitivity.

Likewise, London et al. (2024) found that a short-term eucaloric low-carbohydrate diet reduced liver triacylglycerol in males with overweight or obesity compared with an isocaloric high-carbohydrate diet. That shows the liver responds quickly to macronutrient composition even without weight loss. But neither clamp trial found significantly improved hepatic insulin sensitivity at its respective endpoint. Less liver fat and better suppression of liver glucose output are related concepts, not interchangeable outcomes.

What the broader evidence can and cannot tell you

A 2025 meta-analysis of randomized trials found that carbohydrate-restricted diets, including ketogenic diets, improve several metabolic markers across diverse adult populations. That supports keto as a legitimate option for improving glycemic measures such as glucose and HbA1c. Meta-analysis cannot isolate a weight-independent clamp effect when the underlying studies differ in calorie intake, weight change, diet composition, and outcome measures.

Another small human experiment adds biological context without proving treatment benefit. In 19 healthy adults exposed to controlled lipid overload, Mey et al. (2023) found that individual BHB responses were associated differently with susceptibility to lipid-induced hepatic and peripheral insulin resistance. This was an association in healthy people during a lipid challenge, not a test of keto for established insulin resistance.

The practical verdict

If you have type 2 diabetes, keto has evidence as a glycemic-management strategy. Reduced carbohydrate intake can lower the glucose that must be handled after meals, and randomized-trial syntheses report improvements in glycemic markers. Medication changes can become necessary quickly, especially with insulin or sulfonylureas, so this diet belongs in a plan with a clinician who can monitor glucose and adjust treatment.

Do not interpret that outcome evidence as proof that keto bypasses the value of weight loss or directly sensitizes muscle through BHB. The best strictly weight-maintained trial found no liver or muscle insulin-sensitivity improvement after 10 days in adults with obesity and type 2 diabetes. The best positive trial found a skeletal-muscle signal after three weeks in adults with obesity but no diabetes, while average body weight fell 2.2 kg.

Keto may improve glycemia through several routes, but current clamp evidence does not establish ketones as an independent insulin-sensitizing agent.

For now, ketone supplements have the weakest practical case: they can raise BHB and alter fuel oxidation, yet the available human trials did not show improved clamp-measured insulin sensitivity. A ketogenic diet may still fit someone's glucose-management goals, but the measurable outcomes matter more than chasing a ketone number.