Objectives. To systematically review and synthesize the available evidence on the efficacy of the ketogenic diet in reducing seizure frequency for children with refractory epilepsy.
Data Sources. Medline searches were performed using the keywords epilepsy/therapy, dietary therapy, and epilepsy, and the text word ketogenic diet. The Cochrane Library of clinical trials was searched using the term ketogenic diet. Bibliographies of recent review articles and relevant primary research reports, as well asCurrent Contents were reviewed for additional relevant citations.
Study Selection. Studies were selected for inclusion in the review that reported the reduction of seizure frequency following treatment with the ketogenic diet in children with refractory epilepsy. The outcome measures used were the percentage of patients with: 1) complete elimination of seizures, 2) >90% reduction in seizures, and 3) >50% reduction in seizures.
Results. The evidence consists entirely of uncontrolled studies. Of 11 studies identified for this review, 9 are retrospective series of patients from a single institution. Two studies are prospective, 1 of which is a multicenter trial. The results of these studies are consistent in showing that some children benefit from the ketogenic diet, demonstrated by a significant reduction in seizure frequency. Estimates of the rates of improvement by combined analysis (confidence profile method) are complete cessation of all seizures in 16% of children (95% confidence interval [CI]: 11.0–21.7); a greater than 90% reduction in seizures in 32% (95% CI: 25.3–39.8); and a greater than 50% reduction in seizures in 56% (95% CI: 41.2–69.7). It is unlikely that this degree of benefit can result from a placebo response and/or spontaneous remission.
Conclusions. Although controlled trials are lacking, the evidence is sufficient to determine that the ketogenic diet is efficacious in reducing seizure frequency in children with refractory epilepsy.
Intractable or refractory epilepsy is defined by inadequate control of seizures despite optimal treatment with conventional medications. Of the 2.5 million patients with epilepsy in the United States, 25% to 30% can be considered to have intractable epilepsy.1,,2 Poorly controlled epilepsy has been associated with higher rates of mortality, unemployment, and cognitive impairments.3
A number of new antiepileptic drugs (AEDs) have been approved for use in refractory epilepsy (eg, felbamate, lamotrigine, gabapentin, and vigabatrin)and numerous trials of these agents as add-on therapy for refractory seizures in adult patients have demonstrated a modest benefit. A recent meta-analysis included an analysis of 10 placebo-controlled trials of gabapentin and 11 placebo-controlled trials of lamotrigine.4 Response rates, defined as the percentage of patients with a greater than 50% reduction in seizures, were 20% for gabapentin versus 9.3% for placebo, and 21% for lamotrigine versus 8.9% for placebo. Although primarily labeled for adults, these agents are being used frequently by clinicians for children with refractory seizures.5
A second option for patients with refractory epilepsy is surgery. A resurgence of interest in surgical treatments for refractory seizures has occurred over the last decade. Definite indications for surgery include the presence of a epileptiform focus that is amenable to surgical resection. Approximately 10% of patients with refractory epilepsy meet these criteria.2 In such patients, the success rate of surgery in eliminating or substantially improving seizures is up to 80%,3,,6 and this improvement seems to be stable for at least 4 years.7 Controversy exists in the literature concerning indications for other types of antiepileptiform surgery (temporal lobectomy, callosotomy, and hemispherectomy).7 The morbidity of neurosurgery for intractable seizures is not well-reported, with small retrospective series reporting on results from 1 type of surgery.3,8–10Serious complications have been reported, such as postoperative motor deficits, recurrent central nervous system bleeding, hydrocephalus, and wound infections. Unfortunately, the rates at which these complications occur remain ill defined.
Over the last decade, the ketogenic diet has gained popularity as another treatment option for this group of patients. Dietary measures have been described for the treatment of epilepsy since ancient times. Anecdotal reports documenting the success of fasting or starvation in the treatment of seizures exist as far back as the 5th century B.C.11 Interest in this form of therapy was rekindled in the early 20th century after reports by physicians of dramatic improvements in seizure frequency after a period of fasting.
In the 1920s, pediatricians at Johns Hopkins University postulated that the antiepileptic effect of starvation resulted from ketosis, ie, the presence of ketone bodies in the circulation. These physicians demonstrated that it was possible to maintain a state of ketosis without prolonged starvation, by severely limiting the intake of carbohydrates and proteins, and thereby forcing the body to use ketone bodies as the predominant fuel source. The classic ketogenic diet, developed at Johns Hopkins, contains fats in a 4:1 ratio to carbohydrates. The amount of protein is regulated also so that ∼90% of calories are derived from fat. This diet was used as a treatment for epilepsy fairly commonly in the 1920s and 1930s. In the late 1930s and 1940s, as effective antiepileptic drugs, such as phenytoin and phenobarbital, were introduced into clinical practice, the ketogenic diet was largely replaced by drug therapy.12
The mechanism of effect of ketosis on seizures is not understood. Various theories11 have postulated that: 1) there is a direct stabilizing effect of ketone bodies on the central nervous system; 2) resulting acidosis accompanying ketosis modifies the seizure threshold; 3) changes in fluid and electrolyte balance result in reduced seizures; and 4) change in lipid concentration induced by the diet has an antiseizure effect.
Despite the lack of a well-defined mechanism of action, numerous reports have appeared in the literature that have suggested benefit of this diet in reducing the frequency of seizures. The objective of this present study is to systematically review and synthesize the literature evidence reporting on the efficacy of the ketogenic diet in reducing seizure frequency in children with refractory epilepsy.
Medline searches were performed using the keywords epilepsy/therapy cross-referenced with the textwords ketogenic diet, and then using the keywords dietary therapy crossreferenced with the keyword epilepsy.
A World Wide Web Search was performed using the term ketogenic diet on the Alta Vista search engine. The Cochrane Library of clinical trials was searched using the term ketogenic diet. Bibliographies of recent review articles and relevant primary research reports, as well asCurrent Contents were reviewed for additional relevant citations.
Review of bibliographies revealed references dating back to the 1920s and 1930s. Because of the uncertainty of comparing studies of that era to more current research, article retrieval and review was limited to studies published in 1970 or later.
Studies were selected for inclusion in the review that met the following inclusion criteria: 1) reported relevant health outcomes after treatment with the ketogenic diet in children with refractory epilepsy (refractory was defined as suboptimal control of seizures despite multiple medication trials or intolerance to any effective medications); and 2) treatment given was either the classic ketogenic diet or a modification of this diet (eg, medium chain triglyceride diet).
The main outcome measure evaluated is a reduction in seizure frequency. The optimal outcome is complete elimination of seizures. A 50% or greater reduction in seizures is considered clinically significant. Many studies also reported the percentage of patients achieving near complete elimination of seizures, usually as a percent reduction of 90% or more. These 3 outcome measures (complete elimination of seizure, >90% reduction in seizures, and >50% reduction in seizures) are the outcomes that are reported in this review. Studies were excluded if they contained only subjective outcomes (eg, subjectively improved or not improved). A single study was excluded for this reason.13
Combined analysis of the percentage of patients achieving the relevant outcome measure was performed using the confidence profile method (Fastpro software, Academic Press, Inc, New York, NY). For each outcome measure, a test of homogeneity was first run among the included studies with the following results: complete elimination of seizures (χ2 = 15.5; P = .05); 90% reduction (χ2 = 10.3; P = .07); and 50% reduction (χ2 = 42.8;P < .001). Because each of these tests of homogeneity reached or approached statistical significance, a random effects combined model was used.
An overall assessment of study quality was not performed because of the uncertain validity of overall quality measures.14 Rather, specific aspects of the study design that were believed to be most likely to contribute to variability in outcomes were abstracted. The first of these elements was prospective versus retrospective study design. A sensitivity analysis was performed using only the prospective studies. The second set of elements related to the patient population. Although the treatment given and the outcome measurements were relatively uniform, the patient populations and the degree to which these populations were described varied considerably. The following data elements were abstracted: mean age of patients, type of seizures, mean pretreatment seizure frequency, mean number of pretreatment AEDs, development stage/IQ, and completeness of reporting inclusion/exclusion criteria. For each of these factors, a score of 0 (no information on that parameter), 1 (incomplete information on that parameter), or 2 (complete information on that parameter) was given. Sensitivity analysis was performed on the studies that were rated the highest on these specific elements.
A total of 11 studies were identified that were published since 1970 and met the inclusion criteria for detailed review15–24 (also J. M. Freeman et al, unpublished data, July 1998). This body of evidence consists of uncontrolled studies of children treated with the ketogenic diet. There are no controlled studies that directly compared the ketogenic diet to drug therapy or surgery. Patients treated in these studies were all refractory to treatment with AEDs, although the definition of refractory varied among studies. In general, all patients had failed or were intolerant of treatments with multiple drug regimens.
Data from the studies reporting efficacy of the ketogenic diet are summarized in Tables 1 and2. Ten studies were full-length articles published in peer-reviewed journals.15–24 The final study was an unpublished manuscript obtained with permission from the authors (J. M. Freeman et al, unpublished data). Nine of the 10 published studies were retrospective clinical series of patients treated with the ketogenic diet at a single institution.15–23 Vining et al24 is a prospective, multicenter, uncontrolled trial enrolling 51 patients from 7 clinical centers. Finally, Freeman et al (unpublished data) is a prospective study of 150 consecutive patients treated at 1 institution.
Nine studies reported the percentage of children who became seizure-free,15–21,24 (also J. M. Freeman et al, unpublished data) with a range of 7% to 33%. The percentage of patients with a greater than 90% reduction in seizures was reported in 6 studies18,21–24 (also J. M. Freeman et al, unpublished data) and ranged from 22% to 56%. Nine studies reported the percentage of patients who had a greater than 50% reduction in seizures, ranging from 29% to 100%17–24 (also J. M. Freeman et al, unpublished data).
Combined analysis of these outcome data were next performed. Three separate analyses were performed by the confidence profile method (FastPro) for the outcomes of percentage of patients with complete elimination of seizures, percentage of patients with a greater than 90% reduction in seizures, and the percentage of patients with a greater than 50% reduction in seizures. The combined point estimate for the outcome of percentage of patients who became seizure-free was 15.8% with a 95% confidence interval (CI) of 11.0 to 21.7. For the outcome of greater than 90% reduction in seizures, the point estimate was 32.2% with a 95% CI of 25.3 to 39.8. For the outcome of greater than 50% reduction in seizures, the point estimate was 55.8% with a 95% CI of 41.2 to 69.7 (Table 4).
Two sensitivity analyses were performed, the first limited to prospective studies (n = 2)24 (also J. M. Freeman et al, unpublished data), and the second limited to studies with a quality score for the description of patient population of greater than 4 (n = 5)15,,16,20,24 (also J. M. Freeman et al, unpublished data). As shown in Table 4, there is a slight diminution of the treatment effect for each of the sensitivity analyses. However, the results are not substantially different from the percentages of responders using all the studies.
Compliance with the diet was inconsistently reported. The definition of compliance varied. Some studies restricted compliance to those patients who had clinical improvement and did not consider patients without improvement who stopped the diet to be noncompliant. The range of reported noncompliance was 3% to 32%.
Adverse effects were not consistently reported in the 9 clinical series reviewed. Mild gastrointestinal symptoms were common when reported,16,,21,22 occurring in one third to one half of treated children. Other adverse effects, such as kidney stones and metabolic abnormalities, occurred in <5% of children when reported.15,,23 Vining et al24 systematically reported adverse events in 51 patients. Complications consisted of lethargy, acidosis, constipation, vomiting, and an increased number of infections. These adverse events occurred with a frequency of 4% to 8% (Table 1). One retrospective study,25 which did not meet the inclusion criteria for this review, specifically examined the incidence of kidney stones in 120 patients treated with the ketogenic diet over a 10-year period. This study reported kidney stones in 5% of patients (6/120), as ascertained by review of patient records.
The results of this analysis suggest that approximately half of children with refractory epilepsy will have a clinically meaningful improvement after treatment with the ketogenic diet. Because these results are based on uncontrolled studies, it is possible that the results could be explained by the placebo effect, spontaneous remission, and/or random variation. However, it is unlikely that these factors could account for the degree of seizure reduction seen in these trials.
To estimate the magnitude of the placebo effect in the treatment of refractory epilepsy, the placebo response in recent add-on trials of newer AEDs for patients with refractory epilepsy was examined. These data are on treatment of adults, unfortunately corresponding data on children is lacking. A recent meta-analysis summarizes the relevant data.4 This study analyzed 51 placebo-controlled trials of 6 newer antiepileptic agents (some not available in the United States). In all trials, response rates were defined by the percentage of patients with greater than 50% reduction in seizures. The percentage of patients receiving placebo who responded ranged between 6.2% and 13.8%. Combining results from all patients treated with placebo, 137 of 1396 patients or 9.8% responded to placebo. This is contrasted with the estimate of 55.8% (95% CI: 41.2–69.7) of patients who had a greater than 50% response to the ketogenic diet, making the placebo effect an extremely unlikely explanation for these findings.
Spontaneous remission occurs in pediatric epilepsy. For patients with intractable epilepsy, the natural history of the disorder is not well defined. A single retrospective study was identified that reported on spontaneous remission in patients with refractory epilepsy.1 In this study, spontaneous remission was related to intelligence level and was observed to occur at a rate of 4% per year in patients with normal intelligence per year, and at 1.5% per year in patients with mental retardation. This rate of 1.5% to 4% per year is compared with the estimated 15.8% (95% CI: 11.0–21.7) of patients who became seizure-free on the diet, most within a much shorter time period than 1 year. The lower limit of this CI is more than twice as high as the upper limit of the incidence of spontaneous remission. Also, a comparison of the patient population in this study with the populations for the included studies (Table 4) reveals that the patients treated in the ketogenic diet studies have more severe epiliepsy. Thus, spontaneous remission might account for a small number of patients who improve on the ketogenic diet but is unlikely to explain the majority of the improvement.
This analysis is limited by several factors. The overall quality of this body of literature is not high. There are no controlled trials of the ketogenic diet compared with drugs or surgery, and the majority of the clinical series reported in the literature are retrospective in nature. However, sensitivity analysis by quality of the studies resulted in only a slight dimunition of the treatment effect. A second limitation is the lack of a biological explanation for the treatment effect. Although several theories on mechanism of action have been proposed, none are supported by empirical data.
There are only limited data on adverse effects of the ketogenic diet. Effects on growth and development over the intended course of the diet (2–3 years) are not known. Some authors commented that children may show growth retardation,15 but this has not been systematically examined. It is expected that cholesterol levels and triglyceride levels will rise on the diet. Delgado et al26reported in abstract form that 5 of 17 children (29%) developed significant hypercholesterolemia (greater than 250 mg/dL) while on the diet, and 4 of 17 (24%) developed hypertriglyceridemia. The mean cholesterol for the 5 children with hypercholesterolemia was 367 mg/dL with a range of 253 to 512. The significance of this degree of hypercholesterolemia for a 2- to 3-year period is not known.
In summary, the ketogenic diet seems to be efficacious in reducing the frequency of seizures in children with refractory epilepsy. Some children will benefit, as demonstrated by a reduction in seizures that is unlikely to be attributable to a placebo effect or to spontaneous improvement. Compared with alternatives, this improvement is in the range or greater than that reported with the addition of newer AEDs. For a properly selected subset of patients, surgery may achieve as good or better seizure control but involves the potential for greater morbidity associated with neurosurgery. Therefore, treatment with the ketogenic diet should be considered a valid therapeutic option for children with refractory epilepsy.
The work for this project was supported by the Blue Cross and Blue Shield Association Technology Evaluation Center.
We would like to acknowledge the input of the Blue Cross and Blue Shield Association Technology Evaluation Center Medical Advisory Panel for their assistance in the formulation, review, and revisions to the enclosed manuscript. The members of this panel are James Armitage, MD; Wade Aubry, MD; J. Lawrence Colley, ME; Helen Darling, MA; Richard D. Dent, ME; David Eddy, MD; PhD, Alan Garber, MD, PhD; Kathleen Goonan, MD; Michael Hattwick, MD; I. Craig Henderson, MD; Albert Jonsen, PhD; Peter Margolis, MD, PhD; Barbara McNeil, MD, PhD; Richard Roberts, MD, JD; Earl Steinberg, MD, MPP; Clifford Waldman, MD; Paul Wallace, MD; and Les Zendle, MD.
We would also like to thank Kathleen Zegler, PharmD, for her assistance with manuscript preparation and editing.
- Received November 18, 1999.
- Accepted November 18, 1999.
Reprint requests to (F.L.) Division of General Internal Medicine, Northwestern Medical Faculty Foundation, 676 N St Clair, 18-200, Chicago, IL 60611. E-mail:
- AED =
- antiepileptic drug •
- CI =
- confidence interval
- Barron TF,
- Hunt SL
- ↵Swink TD, Vining EPG, Freeman JM. The ketogenic diet: 1997. In: Advances in Pediatrics: Mosby-Year Book. St Louis, MO: Mosby; 1997:297–328
- Huttenlocher PR,
- Wilbourn AJ,
- Signore JM
- Trauner DA
- Sills MA,
- Forsythe WI,
- Haidukewych D,
- et al.
- Herzberg GZ,
- Fivush BA,
- Kinsman SL,
- et al.
- Copyright © 2000 American Academy of Pediatrics