The ketogenic diet
what the studies say

What it is: the carbs it takes

• Weak evidence · experts or few studies Between 20 and 50 g of carbs a day, not counting fiber, is the figure the American Diabetes Association consensus gives to reach ketosis (ADA 2019). It is an expert consensus: no trial has set the exact number.
• Weak evidence · experts or few studies The average split is 70-80% of calories from fat, 5-10% from carbs and 10-20% from protein (Harvard T.H. Chan School of Public Health, general-audience review).
• Weak evidence · experts or few studies Below those 20-50 g insulin drops and the body uses more fat (StatPearls, NCBI).

In a single meal: how many carbs at once?

The consensus gives a figure for the day, but it does not say how to split it. And having 50 g over five meals is not the same as all in one sitting. No study has set a maximum per meal, or compared the same total split in different ways. What has been measured:

• Measured in people · small studies A large amount at once takes you out of ketosis for hours. With 75 g of glucose, in women who had been in ketosis for almost four years, blood ketones dropped 46% within an hour, stayed below the ketosis level for about two or three hours and recovered after five (Cooper 2023, 10 women).
• Measured in people · small studies For beginners it is harder to get back. With 72 g at breakfast, in people who had been on a ketogenic diet for a week, ketosis took two to five days to recover (Kempf and Martin 2024, 10 people, no control group).
• Measured in people · small studies Even a normal ketogenic meal lowers ketones for up to three hours: their lowest point of the day was after breakfast (Urbain and Bertz 2016, 12 people).
• Weak evidence · experts or few studies The guidelines for the ketogenic diet in epilepsy, where this is watched most closely, ask for carbs to be spread evenly throughout the day (Schoeler 2025). It is a clinical practice consensus, not a measured figure.
• Measured in people · small studies And the daily total has some leeway: in a trial with controlled food intake, people left ketosis at around 90-105 g of carbs a day (Li 2026, 22 adults). Each person responds differently.

Do you lose more weight than with other diets?

A little more at first, and the advantage shrinks over time.

• Strong evidence · meta-analyses At six months or more, about 2 kg more than with a low-fat diet: 11 trials with 1,369 people (Mansoor 2016). It includes low-carb diets that were not all ketogenic.
• Strong evidence · meta-analyses At a year or more, less than 1 kg more: 0.91 kg in 13 trials with 1,415 people (Bueno 2013), and 1.15 kg in a review of 53 trials with 68,128 people whose authors see no reason to prefer one or the other (Tobias 2015).
• Strong evidence · meta-analyses With the same calories, the difference disappears: «little or none» compared with a balanced diet, at 3-6 months and at 1-2 years, in 19 trials (Naude 2014).
• Strong evidence · meta-analyses Between named diets, the differences are small. Two network meta-analyses, of 48 and 121 trials, conclude that at 12 months almost everything evens out and that what counts is the diet each person sticks to (Johnston 2014; Ge 2020).
• Moderate evidence · trials In the largest single trial, with 609 people, at 12 months there was no significant difference between low-carb and low-fat. Note: its low-carb diet ended up at 30%, it was not ketogenic (Gardner 2018).
• Strong evidence · meta-analyses Where it does help is with hunger: while in ketosis people feel slightly less hungry, mainly because it avoids the rise in hunger that losing weight usually brings. The changes are small (Gibson 2015).

Does it speed up your metabolism or burn more fat?

• Measured in people · small studies No. With food controlled in a metabolic ward and the same calories, after switching to a ketogenic diet energy expenditure barely rose, fat loss slowed down and more lean mass was lost: 17 men (Hall 2016).
• Measured in people · small studies In an analysis of 32 controlled feeding studies, with the same calories, the low-fat diet even gave slightly higher expenditure (26 kcal a day) and fat loss (Hall and Guo 2017). Very small differences: calories are what matter.

Cholesterol and blood fats

• Strong evidence · meta-analyses LDL, the «bad» cholesterol, usually goes up: 0.35 mmol/L on average, about 13 mg/dL, in 27 trials with 1,278 people. Triglycerides go down and HDL goes up (Wang 2024). The same compared with low-fat diets (Bueno 2013; Mansoor 2016).
• Weak evidence · experts or few studies In people of normal weight it rose considerably more, about 42 mg/dL of LDL, although that comes from only three trials (Joo 2023). If you do it without being overweight, get your lipids checked.
• Weak evidence · experts or few studies The National Lipid Association sums it up like this: mixed effects on LDL, lower triglycerides, not better for weight loss and little data beyond two years (NLA 2019).

Type 2 diabetes

• Strong evidence · meta-analyses It improves blood sugar in the first months, with no clear advantage after a year. At six months, more people reached an HbA1c below 6.5% (57% versus 31%), but when it was required without medication the effect was small and not significant, and at 12 months it fades: 23 trials (Goldenberg 2021).
• Strong evidence · meta-analyses The fewer the carbs, the lower the HbA1c in the first year; after a year, the same as with more carbs (Snorgaard 2017; Sainsbury 2018).
• Strong evidence · meta-analyses On a strict ketogenic diet, no advantage in glucose or weight compared with the control diet; only slightly better HDL and triglycerides (Choy 2023).
• Weak evidence · experts or few studies The American Diabetes Association accepts it as an option, but with supervision from the start, to adjust insulin and medication and avoid low blood sugar (ADA 2019).

Does too much protein kick you out of ketosis?

It is a very common saying, and Harvard repeats it. But what has been measured, which is little and in very few people, does not confirm it.

• Moderate evidence · trials With 30% protein people entered ketosis just the same, and they ate less and were less hungry: a trial with 17 men and the food provided (Johnstone 2008). Note: that trial used very few carbs, 4% of calories.
• Measured in people · small studies Protein hardly turns into glucose: of 23 g of egg protein, only 3.9 g ended up as glucose over eight hours (Fromentin 2013, 8 people). With 30% protein and no carbs, the body made less glucose in total, not more (Veldhorst 2009, 10 people).

Side effects

• Weak evidence · experts or few studies The «keto flu» — headache, tiredness, nausea, dizziness, digestive discomfort — is most common in the first week. It has only been described from online forums, not measured in trials (Bostock 2020, 300 users).
• Weak evidence · experts or few studies Kidney stones in around 6%, almost half of them uric acid, in mostly observational studies with many children with epilepsy (Acharya 2021).
• Moderate evidence · trials In a trial with 120 people, mild adverse effects were more frequent with the low-carb diet (Yancy 2004).
• Weak evidence · experts or few studies Constipation and bad breath are also described and, in the long term, high cholesterol, stones and lower bone density (StatPearls; Harvard).

Who should NOT do it, or only with their doctor

• People who take SGLT2 inhibitors for diabetes (dapagliflozin, empagliflozin, canagliflozin): the European Medicines Agency warned of rare cases of ketoacidosis, some fatal, and lists not eating enough as a risk factor (EMA 2016). There are cases linked to combining them with a ketogenic diet, one a week after starting (Mistry 2021). Don’t start it without talking to your doctor.
• Pancreatitis, liver failure and fat metabolism disorders (carnitine deficiency, porphyrias…): absolute contraindication (StatPearls).
• Diabetes treated with insulin: only with supervision, because of low blood sugar while the dose is adjusted (ADA 2019).
• Pregnancy, breastfeeding, kidney disease and anyone who has had an eating disorder: there is not enough research to recommend it (ADA 2019). During breastfeeding there is a case of severe ketoacidosis ten days after starting (von Geijer 2015).

How many people stick with it?

• Moderate evidence · trials After a year about half of those who started were still on it, and what was most related to the weight lost was sticking to the diet, not which diet it was: 160 people (Dansinger 2005).
• Moderate evidence · trials The weight advantage at six months was no longer significant after a year, with high dropout in both groups (Foster 2003).
• Strong evidence · meta-analyses Much of the lower effectiveness of very low-carb diets was explained by worse adherence (Goldenberg 2021).

What is NOT proven

• That it takes off more weight than any other diet in the long term. After a year the advantage is less than 1 kg or not significant, and with the same calories there is none.
• That it speeds up your metabolism or burns more fat on its own. Studies with controlled food intake say the opposite.
• That it cures or reverses diabetes. The improvement is seen at six months, depends on medication and does not hold up after a year.
• That it improves cholesterol. Triglycerides and HDL improve, but LDL goes up.
• That it is safe in the long term. There is very little data beyond two years.

How our app builds it, and which part is our decision

We explain it because almost nobody does. Our diet app has a ketogenic option, and this is how it builds it:

• At most 10 g of carbs per meal, and for the day 10 per meal up to 50. With one meal, 10 g; with two, 20; with three, 30; with four, 40; with five or more, 50. The daily cap, up to 50 g, comes from the studies; the 10 per meal are our decision. No study sets a maximum per meal, but it has been measured that a large amount at once takes you out of ketosis for hours (we explain it above), so spreading them out is the prudent choice. It is a convention, not a measured threshold.
• We count net carbs: fiber is subtracted, the same way the ADA range counts them, so the number in the app and the cap speak the same language.
• Our protein is higher than the classic one: around 30% of calories, compared with the 10-20% of the reference. That is on purpose: we calculate it on your lean mass to protect muscle while you lose weight (Helms 2014; we explain it in the protein guide). What has been measured says that with 30% protein you stay in ketosis, but that trial used fewer carbs than we do: the exact combination has not been measured.
• Fat takes up the rest of the calories, which is what defines the diet.

Frequently asked questions