Safety and Tolerability of the Ketogenic Diet in Pediatric Epilepsy: Effects of Valproate Combination Therapy


  • David A. Lyczkowski,

    1. Pediatric Epilepsy Program, Departments of Neurology and Pediatrics, Massachusetts General Hospital, Boston, Massachusetts, U.S.A.
    Search for more papers by this author
  • Heidi H. Pfeifer,

    1. Pediatric Epilepsy Program, Departments of Neurology and Pediatrics, Massachusetts General Hospital, Boston, Massachusetts, U.S.A.
    Search for more papers by this author
  • Soumit Ghosh,

    1. Pediatric Epilepsy Program, Departments of Neurology and Pediatrics, Massachusetts General Hospital, Boston, Massachusetts, U.S.A.
    Search for more papers by this author
  • Elizabeth A. Thiele

    1. Pediatric Epilepsy Program, Departments of Neurology and Pediatrics, Massachusetts General Hospital, Boston, Massachusetts, U.S.A.
    Search for more papers by this author

Address correspondence and reprint requests to Dr. E.A. Thiele at Pediatric Epilepsy Program, Massachusetts General Hospital, 175 Cambridge St., Suite 340, Boston, MA 02114. E-mail:


Summary: Purpose: To evaluate safety and tolerability of ketogenic diet (KGD) and valproate (VPA) cotherapy in the treatment of intractable seizures.

Methods: The patient records of children who underwent KGD initiation at the Massachusetts General Hospital for Children from February 2002 to September 2004 were retrospectively assessed. Efficacy was measured by comparing reported seizure frequency at baseline and at 3-month intervals thereafter. Adverse events and reasons for terminating the diet were tabulated.

Results: Of 71 patients who underwent KGD initiation, 24 were concomitantly using VPA at the time of initiation. The most serious adverse events were two cases of acute pancreatitis (2.8%), both of which occurred in patients not taking VPA. The most common complications in both groups were acidosis (39.4%), nausea and vomiting (23.9%), hypertriglyceridemia (21.1%), lethargy (18.3%), and behavioral changes and irritability (15.5%). No significant difference in adverse-event profiles was found between the VPA group and the non-VPA group. At 1 year, 32 patients remained on the diet, including 11 in the VPA group. Efficacy was nearly identical in these two groups.

Conclusions: KGD and VPA combination therapy is relatively safe and effective in refractory pediatric epilepsy. Adverse-event profiles of patients on KGD and VPA cotherapy are similar to those of patients on the KGD without VPA. In considering possible treatment options for intractable seizures, cotherapy with these two modalities should not be excluded for safety or tolerability concerns. In some patients, this combination may provide optimal seizure control.

The ketogenic diet (KGD) can effectively treat otherwise intractable epilepsy in children (1–4). Because of the refractory nature of their seizures, candidates for the KGD are commonly taking antiepileptic drugs (AEDs) at the time of diet initiation; polytherapy is common. Valproate (VPA), a common broad-spectrum AED, may be a component of this polytherapy.

Adverse events have been associated with the KGD and also with VPA treatment. In addition to relatively common complications such as dehydration, lethargy, and gastrointestinal disturbances, the KGD may predispose patients to renal stones, chronic metabolic acidosis, hyperlipidemia, cardiomyopathy, and pancreatitis (5–8). VPA commonly causes nausea, emesis, tremor, and weight gain. It also is believed to increase risk of acute pancreatitis, thrombocytopenia, and polycystic ovarian syndrome. Furthermore, in rare cases, VPA toxicity has been associated with hyperammonemic encephalopathy and acute hepatic failure (9,10).

Because of these complications, it has often been suggested that KGD and VPA cotherapy may be contraindicated (7). Because both VPA and KGD can induce carnitine deficiency, it also has been suggested that their cumulative effects may cause hepatotoxicity by a carnitine-related mechanism (11). Clinical evidence for the possible contraindications of KGD and VPA cotherapy remains largely anecdotal, however.


To evaluate the relative side effects and tolerability of the KGD with and without VPA cotherapy, we undertook a retrospective study of the records of children initiated on the KGD through the Massachusetts General Hospital for Children Pediatric Epilepsy Program during a 31-month period from February 2002 to September 2004. Patients who underwent initiation at other institutions but attended this Ketogenic Diet Program were excluded. After the inpatient diet initiation, clinic visits were scheduled at 2 weeks, 1 month, 3 months, and every 3 months thereafter. We examined hospital admission notes and discharge summaries, outpatient clinic notes, notes from telephone conversations, e-mails, and other documents generated before March 2005.

Diet efficacy was evaluated based on patient or family reports or both of seizure frequency documented at each visit. (Although some patients experienced decreased seizure intensity or improved cognitive ability without exhibiting significantly reduced seizure frequency, these variables were too difficult to quantify retrospectively and were therefore omitted from this study.) Complications associated with the diet were tabulated at diet initiation and at all follow-up visits. When a patient discontinued the diet, the reason was noted. AED regimens and serum carnitine levels were tracked. When on occasion KGD clinic visits were missed, data were estimated based on trends in existing data, interim phone contact with the families, clinicians' reports from later visits, and reports from other practitioners at the Massachusetts General Hospital. χ2 and analysis of variance were performed by using SPSS v. 11.5.


From February 2002 to September 2004, 71 children underwent KGD initiation at the Massachusetts General Hospital (Table 1). During an inpatient stay of ≥3 days, each patient successfully initiated the KGD and had documented ketosis. In nearly all patients, moderate acidosis (blood CO2 <23.0 mM) developed, and nearly all had a decrease in blood glucose during the course of the admission.

Table 1. Patient profiles at time of diet initiation
group (n = 47)
VPA group
(n = 24)
(n = 71)
Age (yr;mo) (mean yr ± SD)0;7–20;11 (6.52 ± 4.43) 1;5–20;6 (8.34 ± 5.23)    0;7–20;11 (7.13 ± 4.76)
Number of patients on polypharmacotherapy (≥2 AEDs) 37 (78.7%)22 (91.7%)    59 (83.1%)
Number of concomitant AEDs, including VPA (mean ± SD)     0–5 (3.23 ± 1.11)    1–4 (2.75 ± 0.85)        0–5 (2.46 ± 1.04)
Seizure type
 Partial-onset (focal) seizures 26 (55.3%)16 (66.7%)    42 (59.2%)
 Generalized seizures 19 (40.4%) 8 (33.3%)    27 (38.0%)
 Uncertain 2 (4.3%)0 (0.0)      2 (2.8%)

From February 2002 to early May 2003, 33 patients were initiated on the diet by consuming 1/3 of their estimated daily caloric requirement on day 1, 2/3 on day 2, and full caloric intake on day 3. Beginning in May 2003, 38 patients were initiated on a modified protocol, receiving the full-strength diet from day 1. Patients did not fast before or on admission. At initiation, each diet was calculated for a ratio of fats to carbohydrate + protein. Two patients had KGD ratios <3.0:1; 12 had ratios of 3–3.4:1; 17 had ratios of 3.5–3.9:1; and 40 had ratios of 4:1.

Patients continued their usual AED regimens throughout initiation of the diet, although AEDs were commonly adjusted or eliminated in the months that followed. Diet ratios and caloric content were adjusted during these clinic visits to optimize seizure control, to control weight gain/loss, and for diet tolerability.

Adverse events are listed in Table 2. Pancreatitis, observed in two patients, was the most serious adverse event; of note, neither of these patients was receiving combination therapy with VPA. The most common adverse event was symptomatic acidosis, which resolved in all cases by administration of polycitra-K or bicarbonate or both. No significant relation was observed between KGD ratio and adverse effects.

Table 2. Adverse events
 Non-VPA group
(n = 47)
VPA group
(n = 24)
(n = 71)
  1. aSymptomatic acidosis refers to plasma CO2 below normal (<23.0 mM) with nausea, vomiting, and/or lethargy.

  2. bOther adverse events: non-VPA group includes one case each of alopecia, amenorrhea, edema, hyperammonemia, premature adrenarche, and symptomatic hypoglycemia. VPA group includes one case each of anemia, brittle hair, rash, and vigabatrin-related retinopathy.

Gastrointestinal18 (38.3%)10 (41.7%)28 (39.4%)
 Constipation4 (8.5%) 3 (12.5%)7 (9.9%)
 Diarrhea1 (2.1%)1 (4.2%)2 (2.8%)
 Nausea/vomiting10 (21.3%) 7 (29.2%)17 (23.9%)
 Reflux1 (2.1%)2 (8.3%)3 (4.2%)
 PO refusal1 (2.1%)1 (4.2%)2 (2.8%)
 Duodenal ulcers0  1 (4.2%)1 (1.4%)
Infectious disease10 (21.3%) 5 (20.8%)15 (21.1%)
 Otitis media 6 (12.8%) 4 (16.7%)10 (14.1%)
 Upper respiratory tract infection4 (8.5%)2 (8.3%)6 (8.5%)
 Recurrent viral illness1 (2.1%)1 (4.2%)2 (2.8%)
 Gastroenteritis1 (2.1%)0  1 (1.4%)
Renal 9 (19.1%) 3 (12.5%)12 (16.9%)
 Hematuria 6 (12.8%)1 (4.2%)7 (9.9%)
 Renal stone3 (6.4%)0  3 (4.2%)
 Urinary tract infection2 (4.3%)0  2 (2.8%)
 Dysuria, urinary incontinence0   2 (8.3%)2 (2.8%)
 Symptomatic acidosisa19 (40.4%) 9 (37.5%)28 (39.4%)
 Lethargy11 (23.4%)2 (8.3%)13 (18.3%)
 Behavioral changes, irritability, hyperactivity, depression 9 (19.1%) 4 (16.7%)11 (15.5%)
 Severe hypercholesterolemia (>300)2 (4.3%)2 (8.3%)4 (5.6%)
 Severe hypertriglyceridemia (>500) 9 (19.1%) 6 (25.0%)15 (21.1%)
 Pancreatitis2 (4.3%)0  2 (2.8%)
 Other adverse eventsb 6 (12.8%) 4 (16.7%)10 (14.1%)

In addition to the complications listed in Table 2, we observed substantial elevations in liver function tests (LFTs) in two patients, both of whom were taking VPA cotherapy. In patient A, a 9-year-old boy with idiopathic partial complex seizure disorder and electrical status epilepticus of sleep, LFT elevation was noted ∼3 weeks after diet initiation, with a concomitant increase in serum triglycerides (SGPT, 201 U/L; normal, 10–55; SGOT, 161 U/L; normal, 10–40; TRIG, 383 mg/dl; normal, 40–150). However, the patient reported no symptoms, and the abnormal values spontaneously resolved by the time of the next blood tests, 6 months later. In patient B, a 17-month-old girl with neurofibromatosis-1, LFT elevation occurred 2 weeks after diet initiation (SGPT, 229 U/L; SGOT, 283 U/L). At this time, the patient showed no symptoms other than lethargy. VPA was tapered 6 months after diet initiation because of improved seizure control. Owing to phlebotomy difficulties, blood samples were not obtained until 1 year later, at which time SGPT had returned to the normal range (29 U/L), and SGOT remained moderately elevated at 73 U/L. Unfortunately, LFT values are not available from the time of VPA taper to distinguish between VPA and KGD as possible causes of the elevated LFTs.

Although no patients reported symptoms suggestive of carnitine deficiency (cardiomyopathy, skeletal muscle weakness, or signs of acute encephalopathy), 19 received l-carnitine supplementation while on the diet; of these, presupplement post–diet-initiation carnitine levels were available for 10. Postsupplement levels were excluded from our analysis. Mean serum total carnitine of 16 patients in the VPA group and 39 in the non-VPA group was within normal limits (mean ± SD for VPA, 41.3 ± 14.3 μM; non-VPA, 47.0 ± 12.8; p = 0.154; reference range, 27–71). In both groups, mean esterified carnitine was well above the normal range, and the mean free/total carnitine ratio was well below normal. The VPA group had significantly lower esterified carnitine (VPA, 21.5 ± 8.3 μM; non-VPA, 28.0 ± 9.5; p = 0.021; reference range, 4–16) and significantly higher free/total carnitine ratios (VPA, 49.1 ± 7.3%; non-VPA, 41.6 ± 8.1%; p = 0.002; reference range, 68–92%) than did the non-VPA group.

Reductions in seizure frequency at the 3-, 6-, and 12-month follow-up visits are shown in Fig. 1. At 3 months, 40% of those remaining on the diet had >90% reduction in seizure frequency, 36% had 50–90% reduction, and 24% had <50% reduction. At 6 months, the percentages of patients were 43%, 28%, and 28%, respectively. At 12 months, the percentages were 47%, 25%, and 29%, respectively. No significant differences were found between the patients receiving VPA cotherapy and those not taking VPA (p > 0.05). No significant relation was found between KGD ratio and effectiveness.

Figure 1.

Reduction in seizure frequency, valproate cotherapy vs. no valproate cotherapy

Thirty-five patients are known to have discontinued the diet within 2 years of initiation. (Because many initiated the diet <2 years before the time of this study, the actual number of patients who quit within 2 years will be greater than 35 [49.3%] of 71. Of the 71 patients in our study, nine have completed 2 years on the diet, 35 quit before reaching 2 years, 24 continue on the diet but have not yet reached 2 years, and the remaining three have moved or lost contact. Fig. 1 shows the actual rate of discontinuance at 3, 6, and 12 months.) Because of the severity of its restrictions, the diet requires a major commitment from patients and their parents, and the most common reason for discontinuing the diet was inability to maintain compliance (15 patients). The next most common reason was ineffectiveness (14 patients). Other than one patient who discontinued because of depression, only two patients discontinued because of illness. Both of these patients had pancreatitis, and as noted earlier, neither was taking VPA cotherapy. No significant relation was found between KGD ratio and rate of discontinuance or tolerability.


Adverse events

Gastrointestinal complaints and symptomatic acidosis were the most common adverse events in this study. Infectious diseases, most commonly otitis media and upper respiratory tract infections, also were common among patients on the diet. More serious complications were rare. The two cases of acute pancreatitis were of greatest concern, and three cases of nephrolithiasis also were diagnosed. Of these, none was cotreated with VPA. Although elevated LFTs were noted in two patients in the VPA group, at least one (and possibly both) spontaneously resolved without discontinuation of either KGD or VPA therapy, and no clinical symptoms were noted. Although hepatotoxicity is a concern in children treated with VPA, and although it has been argued that the KGD may increase this risk (7), no cases of severe liver toxicity were observed. Consistent with other studies, hypertriglyceridemia was noted in 15 (21.1%) of our patients, and hypercholesterolemia in four (5.6%) patients (12,13). Subsequent to completion of the diet, lipid levels may be expected to return to normal with few significant long-term effects (14).

No complication occurred with significantly greater frequency in the VPA group than in the non-VPA group, and the overall rate of complications also was not significantly different between the two groups (p > 0.05). It is important to note, however, that the most serious idiosyncratic reactions to VPA (including pancreatitis, renal tubular dysfunction, and hepatotoxicity) occur so rarely that their absence in our patient population cannot be considered statistically significant. For example, fatal VPA-induced hepatotoxicity is estimated at 1:15,000 among those taking VPA overall. Incidence in the highest-risk category, children younger than two years receiving polytherapy, is estimated at 1:600 (15).

Our findings corroborate those of Kang et al. (6), who reported complications in a cohort of 129 patients on the KGD. Rates of most of the adverse events reported therein are similar to those in our study. Of particular note, the Kang et al. cohort included 96 patients taking VPA while on the KGD. As in our study, complication rates did not differ significantly between the VPA and non-VPA groups in their study.

Ballaban-Gil et al. (7) studied a smaller cohort of 52 patients on the KGD, of whom 29 (56%) received VPA cotherapy. Five (9.6%) patients in that study had serious adverse events, including one with severe hypoproteinemia; one with hypoproteinemia, lipemia, and hemolytic anemia; one with Fanconi's renal tubular acidosis; and two with increased LFTs. The AED regimens of four of these five patients included VPA. (The only patient with complications who did not take VPA was the patient in whom hemolytic anemia developed.) Thus in that study, four (13.8%) of the 29 in the VPA–combination therapy group had serious complications, as did one (4.3%) of the 23 not receiving VPA combination therapy.


Pancreatitis, although not a common complication of the KGD, occurred in two of the patients in our study. Case reports have suggested that the diet triggered hypertriglyceridemia-induced pancreatitis in one patient already treated for hypertriglyceridemia and in another with glucose transport protein deficiency (16,17). In rare cases, acute pancreatitis also has been associated with VPA therapy (18,19). Although reported incidences of asymptomatic elevated serum amylase with VPA range ≤24%, incidence of acute pancreatitis is relatively small with only two cases of 2,416 patients (1,044 patient-years) reported during clinical trials (9,19). The two patients in our study in whom acute pancreatitis developed were not receiving VPA cotherapy. In both cases, the pancreatitis developed 4 months after diet initiation. The KGD may have played a role in the occurrence of pancreatitis in these cases, although other etiologies (e.g., viral) cannot be excluded.

Renal complications

In three of our patients, renal stones developed, but none of these was taking VPA. Of these three, one was taking clobazam and topiramate (TPM); one was taking TPM, zonisamide, and levetiracetam (LEV); and the third was taking clonazepam (CZP) and LEV. No other renal complications were observed. Ballaban-Gil et al. (7) reported one patient taking VPA for 3 years in whom Fanconi's renal tubular acidosis developed 3 months after initiation of the KGD. Tubular function returned to normal with discontinuation of VPA, IV rehydration, electrolyte supplementation, polycitra-K, and bicarbonate supplementation. In the 11 cases reported in the literature, proximal tubular dysfunction secondary to VPA occurred most commonly after years of VPA therapy (20–22). We have found no evidence, however, to suggest that KGD increases that risk.

Possible hepatotoxicity

As described earlier, we observed elevated LFT values in two patients (patients A and B) with VPA and the KGD, but we saw no clinical signs of hepatotoxicity. These resolved spontaneously in patient A without discontinuation of VPA or KGD. In patient B, VPA was discontinued for reasons unrelated to the LFTs, and unfortunately we lack laboratory results that would show whether the spike in LFTs resolved before or after the VPA taper. The subsequent moderate elevation of patient B's SGOT level is likely unrelated to VPA, because the drug had been discontinued 6 months before the time of the blood test and is more likely related to the acute infection or complications thereof for which the patient was admitted.

In the study by Ballaban-Gil et al. (7), two patients also showed elevated LFTs. Taken together, therefore, that study and ours included a total of four patients with substantially elevated LFTs, all of whom were taking VPA cotherapy. Only one of those cases seems to reflect a serious VPA-related hepatotoxicity with clinical correlation (23). The transient increase in liver enzymes in patients taking VPA, apparently observed in two of our patients and one of Ballaban-Gil, is normally self-resolving and without clinical manifestation, and absent other findings is little cause for concern (11,24–26). According to one study, transient increases in SGOT and SGPT that resolved without adjusting VPA dosage occurred in 44% of patients taking the drug (23,27). In sum, the data in our study and in others to date do not support the claim that cotherapy with KGD and VPA is associated with a higher risk of hepatotoxicity than is either treatment alone.

Carnitine deficiency

Carnitine is an amino acid derivative involved in fatty acid metabolism. Studies have shown that KGD increases serum acylcarnitine levels, reduces the free-to-total carnitine ratio, and reduces total carnitine (11,28). Although not unanimously, some studies suggest that VPA also affects serum carnitine levels and that carnitine deficiency may be implicated in VPA-induced hepatotoxicity (29–31). We found no evidence of any clinical effects of hypocarnitinemia in our patient population. In fact, those on VPA/KGD cotherapy had free-to-total carnitine ratios that were significantly higher, and therefore closer to normal, than did those on KGD without VPA. Like other practitioners, we have provided l-carnitine supplementation to some KGD patients with hypocarnitinemia because of the low risk of adverse effects associated with the supplement (28). It remains unclear whether such supplementation has any real effect. Although it is prudent to monitor serum carnitine levels of patients on the KGD, the additional risk of hypocarnitinemia associated with VPA should not generally preclude cotherapy.


Overall efficacy of the KGD in reducing seizure frequency was similar to that typically reported in the literature. One recent summary of nine efficacy studies published between 1925 and 1998 shows 37% with >90% seizure control, 30% with 90–50% control, and 33% with <50% control (1). At 12 months after initiation, our figures are similar (46.9%, 25.0%, and 28.1%, respectively). An important, novel finding is that efficacy did not differ significantly between the VPA-cotreated group and the non–VPA-cotreated group (p > 0.05).

In several patients, cotherapy with VPA and KGD was necessary for optimal seizure control. For example, patient C was a 10-year-old girl at the time of diet initiation with a history of generalized epilepsy since age 3 years with mild developmental delay, attention-deficit/hyperactivity disorder, and behavioral difficulties. Her seizures were of mixed phenotype, but at the time of initiation, they manifested as behavioral outbursts. At the time of diet initiation, she was taking VPA, lamotrigine (LTG), and LEV. Her behavioral spells improved immediately and dramatically on the KGD, and only occasional seizure activity was noted in the first few months of the diet. LEV and LTG were stopped within 5 months after initiation. When a VPA taper was subsequently attempted, seizures worsened and were characterized by unresponsiveness and incontinence of urine; EEG showed increased frequency of generalized spike-and-wave discharges. VPA was increased again, with dramatic improvement in seizure control, cognitive function, and behavior. Three years after initiation of the diet, still taking VPA, the patient tapered off the KGD because of difficulty with compliance. Seizure activity and behavioral problems subsequently increased, suggesting that the combination of KGD and VPA was more effective for this patient than either treatment alone.

Patient D was a 7-year-old girl at diet initiation, with epilepsy and global developmental delay secondary to neonatal hypoxic ischemic encephalopathy, status post ventriculoperitoneal shunt placement at age 6 months for hydrocephalus. At the time of initiation, she was taking LEV and CZP, and she had around four partial seizures per week. After initiation of the diet, she had very good seizure control. CZP was tapered. An increase in seizure activity occurred 1 year after initiation, at which time VPA was added, and seizure frequency decreased. The patient continued on KGD, VPA, and LEV for 3 months before discontinuing the diet because of difficulty with compliance. Subsequently, the patient had up to five seizures per day, a dramatic increase from baseline. In patient D, as in patient C, concomitant use of the KGD and VPA seemed to provide better seizure control than did either agent alone.


The KGD and VPA are relatively safe and effective therapies for epilepsy. Adverse-event profiles of patients on KGD and VPA cotherapy are similar to those of patients on KGD without VPA. With proper monitoring, this combination of therapies can be safely administered. In considering possible treatment options for intractable seizures, simultaneous use of these two modalities should not be excluded because of safety concerns. Furthermore, in some patients, optimal seizure control has been achieved only with KGD and VPA cotherapy. For patients who have responded well to one or both of these therapies, the combination may provide better control than either agent alone.


Acknowledgment:  We thank Elisabeth Winterkorn, M.D., and Jacqueline Winterkorn, Ph.D., M.D., for their careful readings of this manuscript.