Ketogenic diet

What a ketogenic diet is

Ketogenic is the name for a diet that contains so few carbohydrates that the liver produces ketone bodies from fat, above all beta-hydroxybutyrate. In the studies, the limit is usually at most 50 grams of carbohydrates per day. In a Swedish feeding study, only 4 percent of energy came from carbohydrates, 77 percent from fat and 19 percent from protein.

There are several variants. The classic ketogenic diet of epilepsy treatment is the strictest; alongside it, there are variants with MCT fats and the milder modified Atkins diet. What distinguishes the studies is often less the fat than what else is on the plate: in Keto-Med, the ketogenic diet included plenty of non-starchy vegetables, while added sugar and white flour were dropped, as were legumes, fruit and whole grains. Ketone bodies can also be drunk; what such drinks do is on the page on exogenous ketones.

Why you lose weight with it

The popular explanation is: less insulin, more fat burning, a metabolic advantage beyond calories. The research group around Kevin Hall at the US health agency NIH tested this on a metabolic ward. 17 overweight men ate a high-carbohydrate diet for four weeks and then a ketogenic diet for four weeks, with the same amounts of calories and protein. Energy expenditure in the metabolic chamber rose by 57 kilocalories per day, close to the limit of what can be measured. Fat loss did not become faster in the ketogenic phase, but slower.

That people nevertheless lose weight well on a ketogenic diet is shown by the randomized trials. According to these data, however, the route runs through what and how much is eaten, not through a metabolic effect of its own.

What the studies show

Randomized epilepsy trial, Lancet Neurology 2008

145 children aged 2 to 16 with at least daily seizures, in whom two medications had failed, started the diet either immediately or three months later. After three months, the number of seizures in the diet group was 62 percent of baseline, in the control group 137 percent. More than a halving of seizures was seen in 38 percent of the children on the diet and 6 percent without. The trial was not blinded, and 103 of the 145 children could be analyzed.

Cochrane review on epilepsy 2020

13 studies with 932 participants, 711 of them children and 221 adults. In children, seizure freedom and an at least halved number of seizures were markedly more frequent with the diet than with usual treatment. In adults, no one became seizure-free. The authors rate the certainty of the evidence for all endpoints as low to very low, mainly because no study was blinded.

Keto-Med, American Journal of Clinical Nutrition 2022

40 people with prediabetes or type 2 diabetes each ate a ketogenic and a Mediterranean diet for twelve weeks, in random order; 33 were included in the main analysis. Long-term blood sugar, HbA1c, did not differ between the phases and improved with both diets. Under the ketogenic diet, triglycerides fell by 16 instead of 5 percent, whereas LDL cholesterol rose by 10 percent, while it fell by 5 percent on the Mediterranean diet. On the ketogenic diet, participants consumed less fiber and three other nutrients, and the follow-up suggests that the Mediterranean diet was easier to stick to.

Meta-analysis ketogenic versus Mediterranean, 2026

Randomized trials in adults up to March 2026. The ketogenic diet lowered weight by 3.6 kilos more, waist circumference by 4.3 centimeters, body fat percentage by 2.1 percentage points, and insulin more strongly. For fasting blood sugar, HbA1c, LDL, HDL, triglycerides, blood pressure and CRP, no difference was found. For triglycerides and LDL, the authors found signs of publication bias.

Long-term meta-analysis, British Journal of Nutrition 2013

13 randomized trials lasting at least twelve months, at most 50 grams of carbohydrates per day versus a low-fat diet. The advantage in weight was 0.9 kilos in 1,415 patients. Triglycerides and diastolic blood pressure fell more, HDL rose, and so did LDL, by 0.12 millimoles per liter.

Metabolic ward, American Journal of Clinical Nutrition 2016

17 overweight men, first four weeks high-carbohydrate, then four weeks ketogenic with the same amount of calories, without random order. Energy expenditure rose by 57 kilocalories per day in the metabolic chamber and by 151 with the doubly labeled water method. Fat loss slowed under the ketogenic diet, while at the same time more fat-free mass was lost.

LDL in young, lean women, Nutrients 2021

Randomized feeding study from Sweden: 24 healthy women of normal weight, four weeks each on a ketogenic diet and on a diet following the recommendation of the Swedish Food Agency; 17 completed both phases. LDL rose in every single woman, on average by 1.82 millimoles per liter, and ApoB rose along with it.

What users report: survey of 548 low-carb eaters, 2022

The research group around Nicholas Norwitz analyzed an online survey of 548 adults on a low-carbohydrate diet. The lower the body mass index, the more LDL rose, and a low ratio of triglycerides to HDL, actually a sign of good metabolic health, predicted the largest increases. The authors called 100 respondents with LDL of 200 or more, HDL of 80 or more and triglycerides of at most 70 milligrams per deciliter “lean mass hyper-responders”. In an accompanying case series of five patients, even a moderate return to carbohydrates lowered LDL markedly. It is self-reporting by volunteers, not a representative sample.

KETO study on coronary arteries, JACC Advances 2024

80 people with an LDL of at least 190 milligrams per deciliter triggered by carbohydrate restriction, on average 272 and at most 591, on a ketogenic diet for an average of 4.7 years. They were compared with 80 matched individuals from a cohort in Miami with a mean LDL of 123. In calcium score and plaque burden on cardiac CT, no difference was found. It is a cross-sectional study with selected participants, not a follow-up study.

Retracted one-year analysis, 2026

The follow-up analysis of the same group, titled “Plaque Predicts Plaque, ApoB Does Not”, appeared in 2025. In May 2026, it was retracted at the request of the authors and editors: after publication, concerns about the methodology had arisen that affect the reliability of the data, and the errors were too large for a correction.

Where the data stop

The weight advantage shrinks over time. After at least twelve months, just under one kilo remains compared with a low-fat diet, and in Keto-Med the follow-up suggests that the Mediterranean diet was easier to stick to. For adults with epilepsy, the benefit is uncertain; in the Cochrane review, no one became seizure-free.

The biggest gap is LDL cholesterol. That it rises in many people, and strongly in some, is shown by feeding studies. What this means for heart attack and stroke over years has not been studied in any randomized trial. The thesis that a very high LDL is less dangerous in lean, metabolically healthy “hyper-responders” rests on an online survey and a cross-sectional study of selected participants. The follow-up analysis that was meant to back this up has been retracted, and a peer-reviewed reanalysis is not available as of November 2026.

How to do it

In the studies, at most 50 grams of carbohydrates per day count as ketogenic. Build the diet around plenty of non-starchy vegetables, avoid added sugar and white flour, and pay attention to fiber, because it was lower on the ketogenic diet than on the Mediterranean diet in Keto-Med. If you only want to eat fewer carbohydrates in the evening, that is a different topic with its own body of evidence, see Cutting carbohydrates in the evening; anyone who wants to work with meal timing rather than nutrients will find that under intermittent fasting.

Have your blood lipids measured before starting and after about three months, at least LDL, HDL and triglycerides, ideally also ApoB. In the Swedish study, LDL rose in every participant after just four weeks. If it rises sharply, that is a finding to discuss with your doctor, not a sign of particularly good ketosis; in the Norwitz case series, even a moderate return to carbohydrates lowered LDL markedly. Another blood lipid value that can hardly be influenced by diet is described on the page on lipoprotein(a).

Do not count on a metabolic advantage. Anyone who loses weight does so through the amount of calories, and in the long run what counts is whether you stick to the diet. Anyone taking insulin or other blood-sugar-lowering medication switches only with medical supervision, because otherwise the dose no longer matches the diet. In epilepsy, the diet is a therapy in specialized centers with dietitians, not for self-experimentation.

Safety

The most common side effects in the epilepsy studies were constipation, vomiting, diarrhea, lack of energy and hunger; in the 2008 trial, 10 children dropped out of treatment before the three-month analysis, 6 of them because of intolerance. In Keto-Med, participants consumed less fiber and three other nutrients. The most important issue for adults is LDL cholesterol, which rises in randomized trials, by 10 percent in Keto-Med and by 1.82 millimoles per liter in young, lean women. Long-term data on the heart and blood vessels from randomized trials do not exist. The epilepsy diet is carried out under medical supervision.

Sources

• Randomized epilepsy trial — Neal et al., Lancet Neurology 2008, 145 children
• Cochrane review of ketogenic diets in drug-resistant epilepsy — Martin-McGill et al., 2020, 13 studies, 932 participants
• Keto-Med, ketogenic versus Mediterranean in prediabetes and type 2 diabetes — Gardner et al., American Journal of Clinical Nutrition 2022
• Meta-analysis ketogenic versus Mediterranean — Kun et al., Reviews in Endocrine and Metabolic Disorders 2026
• Long-term meta-analysis ketogenic versus low-fat, at least 12 months — Bueno et al., British Journal of Nutrition 2013
• Metabolic ward, isocaloric ketogenic diet — Hall et al., American Journal of Clinical Nutrition 2016, 17 men
• LDL under a ketogenic diet in young women of normal weight — Burén et al., Nutrients 2021
• “Lean mass hyper-responder” phenotype, survey of 548 adults — Norwitz et al., Current Developments in Nutrition 2022
• KETO study, plaque on cardiac CT with high LDL on keto — Budoff et al., JACC Advances 2024
• Retraction of the one-year analysis “Plaque Predicts Plaque, ApoB Does Not” — JACC Advances 2026