Photo: the Mayo Clinic skyline in Rochester, Minnesota (not a 1921 ward). Michael Hicks from Saint Paul, MN, USA / Wikimedia Commons, CC BY 2.0
Russell Wilder, a doctor at the Mayo Clinic in Rochester, Minnesota, first used a high-fat, very low-carbohydrate diet to treat epilepsy in 1921, and gave the regime the name it still carries. The clinical reference StatPearls credits Wilder with coining the term "ketogenic diet" and with observing that it reduced the frequency and severity of seizures in some of the patients who followed it.
The distance between that and the diet's modern reputation is wide. Wilder was treating a neurological disease in a hospital, and the phrase he put into print was a clinical label. The low-carbohydrate boom that turned ketosis into a selling point arrived in the 2010s, decades later.
The hypothesis started with fasting
Wilder did not begin with fat. He began with fasting. In a 2019 history of dietary treatment in Epilepsy & Behavior, Höhn, Dozières-Puyravel and Auvin set out how the idea took shape. Wilder published two articles in 1921 on the effects of ketonemia on epilepsy, first reporting on the interest of fasting for patients with epilepsy and suggesting that the benefit to seizures might depend on ketonemia rather than on the absence of food itself.
He then hypothesised that equally good results could be obtained with a diet very rich in fat and very low in carbohydrate, which would provoke ketogenesis, and observed the effects of that diet on three patients for the first time. The two papers appeared in the same year, the first on fasting, the second on the diet meant to replace it.
Wilder's 1921 proposal was framed around epilepsy, not specifically around children. Work on children followed at Mayo, where Peterman reported on the diet in paediatric patients in 1924 and 1925. The idea spread quickly for its era: Höhn and colleagues counted nine papers on the ketogenic diet published during the 1920s, involving more than 400 patients with epilepsy.
Clinical versions are precise about proportions in a way popular versions are not. The traditional ratio-based plan, developed for drug-resistant epilepsy and especially for children, sets fat against combined protein and carbohydrate at about four to one, with roughly 90 per cent of calories from fat, 6 per cent from protein and 4 per cent from carbohydrate. Food has to be weighed, and doctors and dietitians supervise. Protein is held to a set share in that arithmetic, which is one reason arguments about how much a person needs, and about where the 0.8 grams per kilogram figure came from, resurface whenever a therapeutic diet is repackaged for general use.
Insulin, anticonvulsants and the 1990s return
The diet lost ground almost as fast as it gained it. Before insulin was isolated in the 1920s, type 1 diabetes was managed mainly with very low-carbohydrate, low-calorie diets that were essentially ketogenic, aimed at reducing glycosuria and hyperglycaemia and often causing severe malnutrition and impaired growth, especially in children. Once exogenous insulin became available, dietary measures were no longer the main treatment. For epilepsy, anticonvulsant drugs had a similar effect, and clinical use of the ketogenic diet declined.
Its return came in the 1990s, driven by renewed success in treating refractory epilepsy, the cases where drugs do not control seizures. Use then widened to a range of cardiometabolic and neurologic conditions, and the diet is now studied and used in settings including obesity, metabolic syndrome and type 2 diabetes.
A clinical observation moving outward until it becomes general dietary advice is a familiar sequence in twentieth-century nutrition, close to what happened with Denis Burkitt's work on fibre. The difference with the ketogenic diet is that the original indication never went away, and it is still the only one that formal guidelines agree on.
What the guidelines actually cover
StatPearls is direct on the point: although current use covers metabolic and neurologic disorders, epilepsy remains the diet's only universally accepted, guideline-supported indication. It is well established in managing drug-resistant epilepsy, glucose transporter type 1 deficiency syndrome and pyruvate dehydrogenase deficiency, where it can substantially reduce seizure frequency and severity.
Everything else sits in a different category. The same reference says multiple randomised controlled trials and meta-analyses have shown effectiveness for weight loss and blood sugar control in obesity and type 2 diabetes, with reductions in body mass index, haemoglobin A1c and triglycerides, while also stating that applications in obesity, diabetes, metabolic dysfunction-associated steatotic liver disease and neurodegenerative conditions remain investigational and lack guidelines from professional medical organisations. Work on cancer, polycystic ovary syndrome and psychiatric conditions sits further back again, on evidence the authors call limited.
The versions sold to the public are not the version clinicians prescribe either. The very-low-carbohydrate form most often used for obesity and type 2 diabetes limits carbohydrate to 20 to 50 grams a day, with moderate protein and 60 to 75 per cent of calories from fat, rather than a fixed ratio and weighed portions. The modified Atkins version, used for adolescents and adults with epilepsy or when strict adherence is difficult, drops the weighing.
One question has stayed open since 1921. Wilder's starting hypothesis was that the ketones in the blood did the work. Höhn and colleagues concluded that available experimental data do not confirm ketosis as the unique mechanism of the ketogenic diet, and that the diet is still being explored to understand the mechanisms involved.




