A Multicenter Study of the Efficacy of the Ketogenic Diet

Objective  To determine the efficacy of the ketogenic diet in multiple centers.

Design  A prospective study of the change in frequency of seizures in 51 children with intractable seizures who were treated with the ketogenic diet.

Setting  Patients were enrolled from the clinical practices of 7 sites. The diet was initiated in-hospital and the patients were followed up for at least 6 months.

Patients  Fifty-one children, aged 1 to 8 years, with more than 10 seizures per week, whose electroencephalogram showed generalized epileptiform abnormalities or multifocal spikes, and who had failed results when taking at least 2 appropriate anti-epileptic drugs.

Intervention  The children were hospitalized, fasted, and a 4:1 ketogenic diet was initiated and maintained.

Main Outcome Measures  Frequency of seizures was documented from parental calendars and efficacy was compared with prediet baseline after 3, 6, and 12 months. The children were categorized as free of seizures, greater than 90% reduction, 50% to 90% reduction, or lower than 50% reduction in frequency of seizures.

Results  Eighty-eight percent of all children initiating the diet remained on it at 3 months, 69% remained on it at 6 months, and 47% remained on it at 1 year. Three months after initiating the diet, frequency of seizures was decreased to greater than 50% in 54%. At 6 months, 28 (55%) of the 51 initiating the diet had at least a 50% decrease from baseline, and at 1 year, 40% of those starting the diet had a greater than 50% decrease in seizures. Five patients (10%) were free of seizures at 1 year. Age, sex, principal seizure type, and electroencephalogram were not statistically related to outcome.

Conclusion  The ketogenic diet is effective in substantially decreasing difficult-to-control seizures and can successfully be administered in a wide variety of settings.

THE KETOGENIC diet is an individually calculated and rigidly controlled, high-fat, low-protein, low-carbohydrate diet used for the treatment of difficult-to-control seizures. Originally developed in the 1920s, this diet was designed to mimic the biochemical changes associated with starvation.¹ In that era, when few anticonvulsants were available, 60% to 75% of children placed on the diet had a more than 50% decrease in their seizures and 30% to 40% of those had a greater than 90% decrease in the frequency of seizures.² When the clinical efficacy of diphenylhydantoin was reported in 1938, attention turned to new anticonvulsant development and therapy.

As new anticonvulsant medications became available, the diet was used less frequently and lack of experience led to the widespread opinion that the diet did not work and was difficult to tolerate.³ However, a few centers continued to occasionally use the diet.⁴ In the early 1990s Kinsman et al⁵ reported a series of 57 patients, refractory to the newer medications including valproate, who were treated with the classic diet. Their results were similar to the older studies.

Since 1994, when a flurry of media coverage focused attention on the ketogenic diet,⁶ clinicians have shown a renewed interest in its use.^3,7 With funding from the Charlie Foundation to Cure Pediatric Epilepsy (Santa Monica, Calif), physician-nurse-dietician teams from 8 medical centers were trained to provide the ketogenic diet and jointly developed a standardized protocol to prospectively enroll children on the ketogenic diet, to compare outcomes, and to determine factors such as age or type of seizures associated with success or failure. Adverse events associated with the diet were also evaluated. This article is the results of that study.

Children aged 1 to 8 years, having more than 10 seizures per week of any type, whose electroencephalogram (EEG) demonstrated generalized epileptiform abnormalities or multifocal spikes, and who had failed at least 2 appropriate anti-epileptic drugs were eligible for the study and were consecutively enrolled. Children with partial seizures only, those whose EEG showed a single epileptic focus, and those who had evidence of metabolic or degenerative disease were to be excluded from the study group, as were families with psychosocial issues that might preclude compliance. This protocol was approved by the individual centers' institutional review boards and informed consent was obtained from each family.

Children were admitted to the hospital, fasted for 36 hours, and started on the classic 4:1 diet (ratio of grams of fat to grams of protein plus carbohydrate) according to the Johns Hopkins protocol.^3,8,9

Table 1 summarizes the protocol. For 1 to 2 days before fasting the family is instructed to decrease their child's intake of carbohydrates and starches. Fasting is begun after dinner, the evening before admission. Whenever possible, carbohydrate-free anticonvulsant medications should be used. On day 1 the child is admitted to the hospital. Fluids (free of caffeine and carbohydrate) are limited to 60 to 75 mL/kg of body weight with an upper limit of 1200 mL/d. Blood glucose is measured by a fingerstick method using a reagent strip (Dextrostix) every 6 hours unless the level falls below 2.2 mmol/L (40 mg/dL) in which case it is measured every 2 hours. Symptoms simulating hypoglycemia warrant an immediate blood glucose test. Symptoms, or glucose levels lower than 1.3 mmol/L (<25 mg/dL), warrant giving 30 mL of orange juice and measuring the blood glucose level again. Symptomatic hypoglycemia during this fasting is uncommon, even in small children.

On day 2 lack of energy and lethargy are common during the second 24 hours of fasting. Hunger is uncommon. On the evening of day 2, after 48 hours of fasting, one third of the calculated ketogenic diet is given as an "eggnog." The diet is generally calculated on the basis of a given number of calories per kilogram to be provided in a given day, divided into 3 equal meals. Usually a 4:1 ratio is used. A 4:1 ratio eggnog would contain 60 g of 36% cream, 25 g of egg, vanilla, and saccharine for flavor. This would yield 245 calories, approximately 4 g of protein, 2 g of carbohydrate, and 24 g of fats (24:6 or 4:1 ratio). Therefore, if 120 mL of a 4:1 ratio eggnog would usually serve as a meal for a given child, one third would be 60 mL of the eggnog and two thirds would be 120 mL of the eggnog. Although most children will have reached large ketones on reagent strip for urinalysis (>160 mmol/dL while receiving Ketostix) by this time, we begin feeding even if this degree of ketosis is not reached. Excess ketosis may be manifested by nausea or vomiting and may be relieved with small amounts of orange juice followed by continuation of the protocol.

On day 3 one third of the calculated diet is given as eggnog for breakfast and lunch. Two thirds is given beginning with dinner. As the body is shifting to the use of ketones as the primary energy source, general lack of energy and lethargy persist; they will be regained over the ensuing 2 weeks.

On day 4 the child continues two third of the calculated diet as eggnog, and dinner is the first full meal. Occasionally, the child becomes too ketotic or acidotic as evidenced by failure to drink, Kussmaul breathing, pallor, or limpness. In such an event the child is rehydrated with carbohydrate-free fluids and the diet is continued.

On day 5 the child receives a full ketogenic breakfast and is discharged. Each day while the child is hospitalized, the parents (and older children) are involved in classes to learn the rationale behind the diet, the calculation of meals, the weighing of foods, the reading of labels, and the management of the diet during the usual childhood infections. Every child should receive a sugar-free, fat-soluble vitamin supplement, and additional calcium.

After discharge parents are instructed to measure urinary ketones daily in the morning. The diet is individually adjusted (calories or ratio) by telephone to provide maximum control of seizures, maintain the child in large ketones as measured on reagent strip for urinalysis ketosis, and to avoid both significant weight gain and weight loss. We do not routinely monitor blood glucose or electrolytes after discharge, nor do we follow up on serum lipid levels, except for research purposes. We do not alter or abandon the diet even if the lipid levels are elevated.

After 1 month on the diet, if seizures had decreased or were eliminated, anticonvulsant medication could be decreased. Frequency of seizures was tabulated from the baseline parental calendars and from parental calendars collected at the time of follow-up at 3, 6, and 12 months.

Demographic, baseline, and study data were recorded on customized data collection forms and transferred to a Paradox for Windows (Borland International Inc, Scott's Valley, Calif) database. The percentage of decrease from the initial baseline frequency of seizures is reported for all children entering the protocol. Independent variables (age, sex, type of seizures, or EEG) were analyzed using Yates corrected χ² and all P values are 2-tailed. Kaplan-Meier survival analysis was based on percentage of decrease from baseline and examined the probability of remaining on the diet for a year. Adverse events related to the diet were also noted at the follow-up visits.

Fifty-one patients, 34 males and 17 females, mean age 4.7 years (range, 1.3-8.6 years) were enrolled in the study between January 1, 1994, and December 30, 1995. The children had been treated with an average of 7 medications and were having an average of 230 seizures per month (range, 11-1880 seizures) during the baseline period. Twenty of the 51 children had tonic-clonic seizures, 19 had myoclonic seizures, 14 had atonic and/or drop seizures, 10 had tonic seizures, 8 had absence, 6 had complex partial seizures, 5 atypical absence, 2 partial seizures with secondary generalization, and 1 each had infantile spasms, simple partial seizures, and clonic seizures. Seizures for 8 children were unclassified. Many children had more than 1 type of seizure.

Three months after initiation of the diet, 45 (88%) of 51 remained on the diet (Table 2). Six were free of seizures; an additional 7 children, for a total of 13 (25%) of 51 had a greater than 90% decrease in seizures, and 15 (29%) of 51 had a 50% to 90% decrease. Only 6 children (12%) had discontinued the diet.

Six months after initiation of the diet, 15 (29%) of 51 children had a greater than 90% decrease in seizures and 12 (24%) of 51 had a 50% to 90% decrease. Fourteen children (27%) had discontinued the diet, and 2 were lost to follow-up.

One year after initiation of the diet 24 (47%) of 51 remained on the diet, 11 (22%) had a greater than 90% decrease in seizures (5 were free of seizures) and 9 (18%) of 51 had a 50% to 90% decrease in seizures. Four children remained on the diet despite a less than 50% decrease in seizures, usually because the seizures were of less frequency or severity, or because the child was receiving less medication. Longitudinal observation showed that 20 (71%) of 28 children who had better than 50% seizure control at 3 months remained on the diet at 12 months. Most of the 23 who had discontinued the diet did so because they had a lower than 50% decrease in seizures. Four (8%) of the 51 children were lost to follow-up during the first year.

Figure 1 illustrates the Kaplan-Meier survival curves for the probability of remaining on the diet in relation to the success of the diet in controlling seizures. It shows a statistically significant difference in the probability of remaining on the diet between the children who had a greater than 50% decrease in seizures and those who had a lower than 50% decrease in the frequency of seizures (Wilcoxon, P<.001). There was little difference with respect to the probability of remaining on the diet between those whose seizures were totally controlled, more than 90% controlled, and 50% to 90% controlled.

There was no statistically significant difference (P<.05) in outcome (<50% control of seizures vs >50% control of seizures at 3, 6, or 12 months) using χ² for the following variables: age, sex, primary type of seizures, presence or absence of normal background, focal EEG changes, or generalized spike-wave abnormalities. There was a marginally significant difference (P = .04) for children who had a pattern of multifocal spikes. These children did less well at 3 months. There was no significant difference at 6 or 12 months.

While the number of patients at each center was small, each center entered at least 6 children; each had at least 1 child become free of seizures; and each had at least 1 child on the diet at 12 months. Only one center had no children with greater than 50% improvement at 1 year. There was no substantial difference between any of the 7 centers in the efficacy of the diet or in the percentage of children remaining on the diet at 3, 6, and 12 months.

Adverse events attributed to the diet were similar between centers and included lethargy (2), severe dehydration or acidosis (2), behavioral changes (4), increase in infections (2), severe constipation (4), and vomiting (2). The diet was discontinued in 6 children during the first 3 months, in an additional 8 by 6 months, and in a total of 23 by 1 year. Reasons given for discontinuation were medical intolerance or illness (6), "too restrictive" (4), insufficient control of seizures (12), and other (1).

Seven epilepsy groups scattered across the continent, some academic medical centers, some private practices, each with differences in available support staff and differences in prior experience with the diet, entered children in a joint protocol to evaluate the efficacy of the ketogenic diet. An eighth center initiated the study but withdrew for personnel reasons. This protocol was not used and cannot be generalized to all children with epilepsy. Children with partial seizures, those with a single EEG focus, and those with evidence of a degenerative disease were excluded from the protocol, as were families who were believed to be incapable of carrying out this rigorous diet.

This study was neither randomized nor blinded. While we recognize that the placebo effect can be significant, we believe that when 40% of the children with difficult-to-control seizures have a greater than 50% decrease in seizures 1 year after initiating the diet, this is unlikely to be due to placebo effect. In one potentially comparable study¹⁰ of the efficacy of felbamate in the Lennox-Gastaut syndrome, felbamate-treated patients had a 34% decrease in the frequency of atonic seizures (reported by parents), compared with a 9% decrease in the patients who received placebo. Four (11%) of 37 children receiving felbamate were noted to be free of seizures, vs 1 (3%) of 35 receiving placebo. However, these results were based on only 3 months of therapy. In our study, 10% of the children who started the diet continued to be free of seizures over the year that they were followed up. It is also possible to theorize that these children simply became free of seizures based on the natural history of their disorder. However, given this population, and extrapolating from the work of Huttenlocker and Hapke,¹¹ such a resolution of intractable seizures within such a short time would be highly unlikely.

The rate of success was similar between the different centers and similar to studies published 4 and 5 decades ago.² Although recently new and effective anticonvulsant medications have become available, seizures continue to remain difficult to control for many children. Even today, for these children, the success rate of the ketogenic diet exceeds that of most new anticonvulsant medications.¹² Most children benefiting from the diet showed a substantial decrease in seizures during the first 3 months on the diet, but many continued to have even better control over the second 3 months on the diet.

Treatment of epilepsy should consider both the control of the seizures and the adverse effects of the treatment. Figure 1 shows that if effective, children can and will stay on the diet. Mattson and et al¹³ used retention time on the medication to evaluate the relative efficacy and adverse effects of different anticonvulsant regimens. Kinsman et al⁵ used a similar technique in their study of the ketogenic diet. In this multicenter study of the ketogenic diet, 24 (47%) of the original 51 children remained on the diet for 1 year. Four additional patients were lost to follow-up. A lack of efficacy may be tolerated for a short time (3 months) when 33% of those still on the diet had lower than 50% seizure control. However, at 12 months, only 17% of those still on the diet had a lower than 50% decrease in seizures, suggesting that it was not worth tolerating the diet without considerable control of seizure. As noted by Nordli and DeVivo,¹⁴ "the question is not "Does the ketogenic diet have a role in the treatment of epilepsy?" but rather, "How can we maximize using the ketogenic diet and learn from it to benefit all children with epilepsy?"

We agree. Even though we now know that the diet works, we still do not know how it works, but then, we also do not know how most anticonvulsants work. Perhaps the resurgence of interest in the ketogenic diet will stimulate research to find how this dramatically different approach to control seizures affects difficult-to-control epilepsy, and thus may lead to new insights into this still devastating condition. Many questions remain about the ketogenic diet, including the effects on lipids, cognition, and behavior. We, and others, are continuing to investigate some of the many questions that the renewed interest in the diet is causing to be generated.

In a prospective, multicenter trial of the efficacy of the ketogenic diet, the diet was found to be effective in decreasing frequency of seizures in children with difficult-to-control seizures. The degree of effectiveness was similar between institutions, similar across different types of seizures, and similar to the reports from the last 5 decades. Further evaluation of this promising old modality of therapy is warranted.

Accepted for publication April 3, 1998.

This study was supported by the Charlie Foundation to Cure Pediatric Epilepsy, Santa Monica, Calif, and The Johns Hopkins Children's Center Telethon Funds, Baltimore, Md.

The Johns Hopkins Medical Institutions, Baltimore, Md: Eileen P. G. Vining, MD; John M. Freeman, MD; Jane Casey, RN, LCSW; Sophie Hsieh, RN; Regina Homer; Millicent Kelly, RD, LD; Cathy Park, RN; Diana Pillas; Paula Pyzik; Traci D. Swink, MD.Scottish Rite Children's Medical Center, Atlanta, Ga: Edwin Trevathan, Raymond Cheng, MD; Linda McCarty, RN; MD; Linda Trevathan, RN; Caroline Silzle.University of Kentucky, Lexington: Edwin Trevathan, MD; Linda Trevathan, RN.Boston Children's Hospital, Boston, Mass: Gregory L. Holmes, MD; Joan Anderson; Marilyn Kuehn.Dalhousie University, IWK Grace Health Centre, Halifax, Nova Scotia: Carol Camfield, MD; Peter R. Camfield, MD; Mary Height; Edyth Smith.Private practice, Mishawaka, Ind: Robert M. Shuman, MD; Rhonda Hammond, RN; Elaine Huffman; Cindy Tansek, RD.Montefiore Medical Center, Bronx, NY: Karen Ballaban-Gil, MD; Shlomo Shinnar, MD, PhD; Candace Callahan, RN; Christine O'Dell, RN, MSN; Miriam Pappo, RD.University of Texas Medical School, Houston: James W. Wheless, MD; M. Suzanne Berryman, MS, RD, LD, CS; Gretchen Matuszak, MS, RD, LD; Wendy Weisenfluh, RN, MN.

Corresponding author: Eileen P. G. Vining, MD, The Johns Hopkins Medical Institutions, Meyer 2-147-JHH, 600 N Wolfe St, Baltimore, MD 21287-7247.