Ketogenic Diet for Childhood Epilepsy

Why the Ketogenic Diet Matters

About a quarter to a third of children with epilepsy develop drug-resistant disease, meaning seizures persist despite two suitable medications. For these children, options are limited, and each uncontrolled seizure carries developmental and safety risk. The ketogenic diet is one of the few non-drug treatments with randomised evidence in this group.

The diet is high in fat and low in carbohydrate. It shifts the brain’s main fuel from glucose toward ketone bodies, reproducing part of the metabolic state of fasting. Proposed anti-seizure mechanisms include altered GABA and glutamate signalling, stabilised cellular energy supply, and effects on the gut microbiome.

Ketogenic dietary therapy has been used since 1921 and was long reserved as a last resort. The 2018 International Ketogenic Diet Study Group consensus advises considering it earlier, after roughly two to three medications have failed. Two questions then dominate the clinic visit: which children are most likely to respond, and which of the diet’s forms to choose.

The Options Compared

Four ketogenic dietary therapies are in mainstream use, and they trade off potency against ease of daily life. The classic diet uses a strict ratio of fat to combined protein and carbohydrate, most often 4:1, so that about 90% of calories come from fat. It is the most studied form and is usually started with dietitian supervision.

The medium-chain triglyceride diet supplies part of its fat as MCT oil, which yields more ketones per calorie. This allows more carbohydrate and protein, though the oil can cause gastrointestinal upset. Trial data show no clear efficacy difference between the MCT and classic diets.

The modified Atkins diet is a lighter version. It limits carbohydrate to roughly 10 to 20 g per day but places no restriction on protein, fluid, or calories, and needs no food weighing. The low-glycaemic-index treatment is lighter still, allowing 40 to 60 g of low-glycaemic carbohydrate daily. Both are started as outpatients.

The table below summarises the strongest randomised and pooled evidence in children. Every sample size and effect estimate was confirmed against a primary source, and every listed reference resolves.

Seizure Control: How the Ketogenic Diet Protocols Differ

Strong evidence supports the ketogenic diet reducing seizures in drug-resistant childhood epilepsy. In the first randomised trial, children assigned to the diet had a mean fall of 75% in baseline seizures against a rise in controls, and 38% reached at least a 50% reduction compared with 6% on usual care.

The Cochrane review pooled four childhood trials. Children on a ketogenic diet were about six times as likely to achieve a 50% reduction, at a risk ratio of 5.80 (95% CI 3.48–9.65), and about three times as likely to become seizure-free, at a risk ratio of 3.16 (95% CI 1.20–8.35). A 2025 meta-analysis of nine trials found a similar pattern, including higher odds of complete seizure freedom.

Head-to-head data are scarcer. The clearest comparison randomised 158 children to the classic diet, the modified Atkins diet, or the low-glycaemic-index treatment. Median seizure reduction at 24 weeks was 66%, 45%, and 54% respectively, and the proportions reaching a 50% response were similar across arms.

That trial was designed to test whether the lighter diets were noninferior to the classic diet. Both the modified Atkins and low-glycaemic comparisons crossed the prespecified 15-percentage-point margin, so noninferiority was not proven. The classic diet gave the largest median reduction, but the difference did not reach a clinically defined threshold in either lighter arm.

A separate randomised comparison of the 4:1 classic diet against the modified Atkins diet in 104 children found no significant overall difference. Mean baseline seizures at three months were 39% on the classic diet and 48% on the modified Atkins diet. Responder and seizure-free rates trended higher with the classic diet across every category without reaching significance.

Timing also differs by protocol. Consensus data indicate that most children who respond do so within about two weeks, so a short trial can signal likely benefit. In the three-arm trial, seizures fell quickly over the first four weeks on the classic and modified Atkins diets, but more gradually over 10 to 12 weeks on the low-glycaemic diet.

Tolerability and Safety: How the Options Differ

Most adverse effects are gastrointestinal and appear in the early weeks. Constipation, vomiting, low energy, and hunger are the common complaints, and pooled data place constipation near 38% and reduced appetite near 20%. These effects are usually mild and rarely force the diet to stop when managed well.

The adverse-event burden differs by protocol. In the three-arm trial, treatment-related events occurred in 56% of the classic-diet group, 57% of the modified-Atkins group, and 33% of the low-glycaemic group, a significantly lower rate for the lightest diet. The head-to-head classic-versus-modified-Atkins trial found more frequent excess urinary calcium on the classic diet, at 43% against 23%.

Longer-term concerns need monitoring rather than routine avoidance. Raised blood lipids are reported in 14% to 59% of children on the classic diet, an increase that is usually transient and often normalises by about a year. Kidney stones occur in 3% to 7%, and preventive oral citrate lowered the rate to under 1% in one series.

Growth deserves particular attention in children. Every study following the classic diet beyond six months has shown some slowing of growth, with younger children most affected. Rare but serious events include cardiomyopathy, a prolonged QT interval, and pancreatitis, some linked to selenium deficiency, which supports scheduled nutritional and cardiac surveillance.

Subgroups That Change the Choice

For two metabolic disorders, the ketogenic diet is the treatment of first choice rather than a fallback. In GLUT1 deficiency syndrome, glucose transport into the brain is impaired, and in pyruvate dehydrogenase deficiency, glucose cannot be fully metabolised. Ketone bodies bypass both defects, and the 2020 GLUT1 consensus names dietary therapy the standard of care.

The consensus group considered the diet first-line for GLUT1 deficiency even in a child without seizures, and it advises against the low-glycaemic diet for these disorders because it does not sustain enough ketosis. In both conditions, earlier treatment is associated with better outcomes, which argues for prompt recognition.

Several epilepsy syndromes respond well enough to warrant early use. The consensus lists infantile spasms, Dravet syndrome, myoclonic-atonic (Doose) epilepsy, tuberous sclerosis complex, Angelman syndrome, and febrile infection-related epilepsy syndrome among conditions where response commonly exceeds 70%. In one prospective series of new-onset infantile spasms, 56% became spasm-free within two weeks of starting the classic diet.

Age shapes the protocol choice. In children under two, the classic diet outperformed the modified Atkins diet, with seizure freedom in 53% versus 20% in that subgroup. Most consensus centres therefore offer only the classic diet to this age group, while reserving the lighter diets for older children and adolescents who value easier adherence.

Some children are less suited to dietary therapy. A child with a clearly resectable focal lesion may do better with surgery; in one comparison, prolonged seizure freedom occurred in 48% on the diet against 59% after surgery. Lennox-Gastaut syndrome shows only modest benefit, with about 51% reaching a 50% reduction.

Contraindications are absolute for disorders of fat metabolism. These include primary carnitine deficiency, carnitine palmitoyltransferase and translocase deficiencies, fatty-acid beta-oxidation defects, pyruvate carboxylase deficiency, and porphyria. Screening blood and urine studies before initiation are intended to exclude these conditions when the cause of epilepsy is unclear.

Quality and Consistency of the Evidence

The evidence is stronger for whether the diet works than for which protocol is best or which child benefits most. The strength of each conclusion varies with study design and blinding, summarised below by outcome.

• Overall seizure reduction — moderate. Multiple randomised trials and a Cochrane review agree on benefit, but every included trial was unblinded, creating a high risk of performance and detection bias that may inflate effect sizes.
• Protocol comparison — moderate. Two randomised trials point the same way, with the classic diet numerically ahead, yet noninferiority of the lighter diets was not proven and samples were modest.
• Syndrome-specific response — low. Data for GLUT1 deficiency, Dravet syndrome, and infantile spasms come mainly from case series and cohorts, so the direction is consistent but the certainty is limited.
• Long-term outcomes — low. Few studies follow children beyond two years, so durability of benefit and late safety signals remain incompletely characterised.

What the Evidence Does Not Show

The data stop short of several claims that are easy to assume. The three-arm trial did not show that the modified Atkins or low-glycaemic diets are noninferior to the classic diet; both breached the prespecified margin, so equivalence remains unproven. Nor do the trials show that any protocol produces seizure freedom for most children, since absolute seizure-free rates are modest.

In infants specifically, a 2023 randomised trial found that the classic diet did not differ in efficacy or tolerability from a further antiseizure medication. That result tempers the assumption that the diet is superior to another drug at this age, though it confirmed the diet was reasonably safe.

Two mechanistic questions also remain open. Blood ketone levels do not reliably predict who will respond, so they are an imperfect biomarker. And whether the diet improves cognition or behaviour beyond seizure control, and what its long-term vascular effects are, has not been settled.

Practical Implications

The evidence supports offering a ketogenic diet after about two medications have failed, and considering it sooner for GLUT1 deficiency, pyruvate dehydrogenase deficiency, and the syndromes with high reported response. Screening for fatty-acid oxidation disorders beforehand is a safety step the consensus treats as essential.

Protocol choice can be individualised rather than ranked by potency alone. The classic diet suits infants under two and the two metabolic indications, while the modified Atkins and low-glycaemic diets are reasonable for older children and adolescents where adherence and adverse-event burden matter. The low-glycaemic diet carried the fewest side effects in trials.

A ketogenic diet is delivered by a team, not a prescription pad. A paediatric neurologist and a dietitian share monitoring of growth, lipids, and micronutrients, with supplementation and follow-up at defined intervals. A trial of at least three months, with early signs often visible within two weeks, gives a fair test before deciding to continue or stop.

Article Information and References