Efficacy, safety, and tolerability of the low glycemic index treatment in pediatric epilepsy


Address correspondence to Dr. Elizabeth A. Thiele, Pediatric Epilepsy Program, 175 Cambridge Street, Suite 340, Boston, MA 02114, U.S.A. E-mail: ethiele@partners.org


Purpose: To report the efficacy, safety, and tolerability of the low glycemic index treatment (LGIT) in pediatric epilepsy.

Methods: A retrospective chart review was performed on patients initiating the LGIT at the Massachusetts General Hospital between January 2002 and June 2008. Demographic and clinical information including seizure type, baseline seizure frequency, medications, blood chemistries, side effects, and anthropometrics were collected. Initiation of the LGIT was done in an outpatient setting. Patients were educated by a dietitian to restrict foods with high glycemic index and to limit total daily carbohydrates to 40–60 g. Change in seizure frequency was assessed at 1-, 3-, 6-, 9-, and 12-month follow-up intervals.

Results: Seventy-six children were included in the study. Eighty-nine percent had intractable epilepsy (≥3 antiepileptic drugs). A greater than 50% reduction from baseline seizure frequency was observed in 42%, 50%, 54%, 64%, and 66% of the population with follow-up available at 1, 3, 6, 9, and 12 months, respectively. Increased efficacy was correlated with lower serum glucose levels at some time points, but not with β-hydroxybutyrate (BOHB) changes or ketosis status at any time point. Only three patients reported side effects (transient lethargy). Blood urea nitrogen (BUN) was elevated in approximately one-third of follow-up laboratory studies. No significant changes were seen in body mass index (BMI) or BMI z-score at any follow-up interval. The most cited reason for treatment discontinuation was the restrictiveness of the diet, in 18 patients (24%).

Conclusion: The LGIT was associated with reduced seizure frequency in a large fraction of patients, with limited side effects.

Dietary therapy has long been recognized as an effective treatment for a broad range of seizure disorders (Lennox, 1960; Bailey et al., 2005). The high fat, low carbohydrate ketogenic diet (KD), which was developed to mimic the biochemical changes associated with fasting (i.e., acidosis, dehydration, and ketosis) (Wilder, 1921; Greene et al., 2003) is effective for pharmacoresistant epilepsies (>90% reduction in seizure frequency in about one-third of patients) and has gained widespread acceptance (Freeman et al., 1998; Bailey et al., 2005). Nevertheless, many patients find it difficult to comply with the KD or find it unpalatable. Effective alternative dietary therapies for epilepsy are needed.

The antiepileptic effects of both fasting and the KD have been associated with decreased blood glucose and increased blood ketone levels (Owen et al., 1967); however, recent research indicates that ketosis alone cannot account for the anticonvulsant effects of the KD (Greene et al., 2003) and suggests that regulation of blood glucose may be at least partly responsible for these effects. Glycemic index (GI) describes the tendency of foods to elevate blood glucose (Jenkins et al., 1981). GI is calculated from the incremental area under the blood glucose curve after feeding, indexed to ingested glucose = 100. Foods with high GI (e.g., most refined carbohydrates) produce substantial increases in blood glucose and insulin levels, whereas foods with a low GI (e.g., meat, dairy, some fruits, some vegetables, and some unprocessed whole-grain foods) induce lower postprandial plasma glucose and insulin profiles (Bell & Sears, 2003). By limiting the quantity of carbohydrates consumed and restricting sources of carbohydrates to low-GI foods, the low glycemic index treatment (LGIT) is designed to prevent dramatic postprandial increases in blood glucose (Pfeifer & Thiele, 2005). The LGIT, compared to the KD, allows for a more liberal total carbohydrate intake (Fig. 1) but restricts foods to those that produce relatively little elevation in blood glucose (GI <50). Diets based on GI have been extensively studied in diabetics and patients with cardiovascular disease, and have been shown to have beneficial effects on blood glucose control (Brand et al., 1991; Fontvieille et al., 1992; Wolever et al., 1992; Jarvi et al., 1999; Rizkalla et al., 2004).

Figure 1.

 Percentage of carbohydrate, fat, and protein intake on the typical (Western) diet, KD, and LGIT. KD, ketogenic diet. LGIT, low glycemic index treatment.

Based on the observations that carbohydrate intake can rapidly terminate the protective effects of fasting and the KD (Lennox & Cobb, 1928) and that glucose levels are positively correlated with seizure susceptibility in animal models (Schwechter et al., 2003), we used the LGIT in children as a treatment for intractable epilepsy. The purpose of this review was to assess the efficacy, safety, and tolerability of the LGIT in a substantial cohort of pediatric epilepsy patients.


A retrospective chart review was performed on 131 pediatric patients who were educated to the LGIT at the Pediatric Epilepsy Clinic of the Massachusetts General Hospital between January 2002 and June 2008. Patients with epilepsy were placed on the LGIT either because: (1) families believed their child could not comply with the KD, (2) the wait for admission to initiate the KD was greater than 2 weeks, or (3) they were unable to tolerate the constraints of the KD and, therefore, transitioned from the KD to the LGIT. Patients were excluded from the data set if: (1) seizure frequency could be assessed neither at 1 nor at 3 months because of loss to follow-up, (2) they had been treated with the KD within 5 months prior to initiating the LGIT (due to concerns of residual benefits), or (3) they had been treated with the LGIT for a period of time shorter than that expected to show three seizures at baseline seizure frequency (e.g., a patient with one seizure every 2 weeks was excluded if on the diet <6 weeks).

The LGIT was initiated in an outpatient setting with an experienced dietitian and neurologist, with follow-up visits every 1–3 months. Before initiation, each patient’s dietary intake was assessed with a 3-day food record using household instruments to measure portion sizes. These records were analyzed with Nutritionist Pro software (Version 2.4, 2005, First Database Division; The Hearst Corporation, San Bruno, CA, U.S.A.) to determine daily caloric intake; based on this intake and the recommended daily allowance (RDA) for age, the dietitian determined the caloric requirement for the LGIT. At initiation, patients and their families received education on the LGIT and were instructed to eliminate high-GI carbohydrates (GI >50) from the diet and to limit total carbohydrates to 40–60 g/day (∼10% of daily calories). They were provided with a reference list of foods based on the International Tables of GI Values (Foster-Powell et al., 2002; Brand-Miller & Foster-Powell, 2007). Recommended goals for daily fats and proteins were set to ensure that the patients’ caloric requirements were met.

At initiation and at all follow-up visits, weight and height were measured using a hospital-grade digital scale and a wall-mounted stadiometer. Seizure type, baseline seizure frequency, and prior and current antiepileptic drugs (AEDs), as reported by parents and patients, were documented by the neurologist and the dietitian. Patients and parents were specifically asked about symptoms suggestive of possible side effects (e.g., constipation, abdominal pain, or symptoms of acidosis). Baseline and follow-up blood glucose, CO2, β-hydroxybutyrate (BOHB), blood urea nitrogen (BUN), creatinine, and urine specific gravity were also collected. Calorie and carbohydrate intake were adjusted during the treatment period to optimize seizure control, growth, and tolerability. If patients continued to have seizures, the LGIT was fine-tuned by ensuring that foods were low GI or by further reducing total carbohydrates. Total caloric intake was adjusted if patients experienced unwanted weight changes. Follow-up food records were obtained as necessary for fine-tuning, but not on a regular basis. Carbohydrate intake was determined by parental report at follow-up visits. Medications were adjusted as necessary for optimal seizure control.

For this retrospective study, we examined patient records generated during clinic visits, notes from telephone calls, e-mail communications, and laboratory results. Seizure frequency at baseline was compared with that at 1-, 3-, 6-, 9-, and 12-month follow-up. Baseline blood glucose, serum CO2, serum BOHB, BUN, serum creatinine, urine specific gravity, and BMI were also compared with follow-up at 1, 3, 6, 9, and 12 months. Growth charts were generated using EpiInfo (version 3.3, October 5, 2004) to determine BMI z-scores. If follow-up data at a given time interval was not available, individuals were excluded from analysis of the variable(s) for which data were missing. Adverse events, reasons for LGIT discontinuation, and medication changes were tabulated.

Statistical tests were performed using SPSS (v. 11.5). Paired-sample t-test was used to assess changes in BMI, BMI z-score, serum glucose, serum CO2, serum BOHB, BUN, serum creatinine, and urine specific gravity. Spearman’s correlation was used to assess the relationships between efficacy and AED changes, between efficacy and changes in anthropometrics, between efficacy and changes in serum BOHB, and between efficacy and serum glucose levels. The Mann-Whitney test was used to assess the relationship between efficacy and ketosis status. The Kruskal-Wallis test was used to assess the relationship between efficacy and seizure type (partial, generalized, or mixed). Tests of significance were two-tailed, and alpha was set at 0.05.


Of the 131 patients initially identified, 30 were excluded because they transitioned from the KD, and 25 were excluded either because they were on the diet too briefly to determine efficacy or because despite their education on the diet they never started dietary therapy. Of the 76 children (ages 1.5–22 years) included in the analysis (Table 1), 13 had previously been on the KD but had been off the KD 1.5–7 years before initiating the LGIT. Sixty-eight (89%) had previously been on three or more AEDs without effective seizure control (and were thus considered to have intractable epilepsy), seven had previously been on two AEDs, and one had previously been on one. Length of time on the LGIT was 2–199 weeks (median 22) with 27 (36%) continuing at the time of this study.

Table 1.   Baseline patient characteristics at time of LGIT initiation
 Patients treated with LGIT (N = 76)
  1. AED, antiepileptic drug; LGIT, low glycemic index treatment.

Age, mean ± SD, years9.65 ± 4.96
Current AEDs, mean ± SD, years2.3 ± 1.1
Total AEDs, mean ± SD, years5.9 ± 2.8
Male/female, n (%)34 (45)/42 (55)
Cognitive functioning, n (%)
 On track33 (43)
 Off track43 (57)
Seizure frequency, n (%)
 ≥1 per day47 (62)
 ≥1 per week but <1 per day16 (21)
 <1 per week13 (17)
Seizure type, n (%)
 Partial onset26 (34)
 Primary generalized, single type21 (28)
 Mixed29 (38)

Based on parental reports, carbohydrate intake while on the LGIT ranged from 15–150 grams/day. At 3 months, mean carbohydrate intake was estimated at 53 ± 18 (based upon the 33 patients with available carbohydrate intake data at that follow-up point). Ranges differed from initial goals for several reasons: (1) reduced carbohydrates were necessary for improved seizure control in some patients; (2) carbohydrate restrictions were relaxed after good seizure control had been achieved and maintained in some patients; or (3) compliance with the LGIT was not always ideal.


Seizure reductions from baseline at 1, 3, 6, 9, and 12 months are shown in Fig. 2, whereas the percentages of patients remaining on the diet at each follow-up interval are shown in Fig. 3. With increasing duration on the diet, those with a reduction of more than 50% from baseline rose, measuring 42%, 50%, 54%, 64%, and 66%, respectively, at the five follow-up intervals. Efficacy did not differ significantly with regard to seizure type (partial, generalized, or mixed) at any of the follow-up intervals (Kruskal-Wallis test, p > 0.05).

Figure 2.

 Percent seizure reduction, compared with baseline, at 1, 3, 6, 9, and 12 months. NC/increase, no change or increased seizure frequency.

Figure 3.

 Percent of patients remaining on LGIT, transitioned to KD, discontinued, or lost to follow-up at each follow-up interval. LGIT, low glycemic index treatment. KD, ketogenic diet.

Reasons for discontinuation

Forty-nine patients (64%) discontinued the LGIT after 2–108 weeks (median 12); reasons for discontinuation are given in Fig. 4. Fifteen were transitioned to the KD (duration on LGIT 2–43 weeks, median 8). Within this group, most used the LGIT only as a transitional treatment while waiting to be admitted for KD initiation. Another small portion of this group was transitioned following partial seizure control on LGIT, with the hope that the KD would provide complete seizure control. Among the remaining 61 patients who were not transitioned to KD, length of time on the LGIT was longer (3–199 weeks, median 32), including 27 (44%) continuing at the time of this study.

Figure 4.

 Reasons for discontinuation. LGIT, low glycemic index treatment. KD, ketogenic diet.

For the 18 patients who ultimately found the LGIT too restrictive, duration of treatment was 3–108 weeks (median 12). Among these patients, efficacy at 1 month was as follows: three had >90% reduction in seizure frequency, one had 50–90% reduction, two had <50% reduction, five had no change, one had increased seizure frequency, and 1-month efficacy data were unavailable for six patients. Of these 18 patients, 10 had problems with restrictiveness from the time of diet initiation, three of whom specifically reported such problems despite a perceived benefit on the diet. These patients commonly reported that it was too easy for them to “cheat” (i.e., eat more carbohydrates than what was allowed, and eat carbohydrates with a higher GI). One of these 10 patients experienced extreme agitation and tantrums upon the limiting of her dietary options. Another 6 of the 18 patients reported problems with restrictiveness only after they perceived the diet to not be sufficiently efficacious. One of these six reported difficulty with compliance specifically due to the social restrictions of the diet. One of the two remaining patients was in a residential facility that reportedly struggled to maintain the diet’s restrictions for the patient, while the family of the other remaining patient found the LGIT to be too expensive and was unable to comply for financial reasons.

Of the nine patients who stopped the LGIT because of ineffectiveness, duration of treatment was 6–47 weeks (median 29). At the time of diet termination, two of these patients had experienced a <50% seizure reduction, four had experienced no change, and three had experienced an increase in seizure frequency. Of the seven patients with either no change or increase, three had relatively long treatment durations (29, 40, and 47 months, respectively) despite the lack of efficacy. The first remained invested because of the intractability of his seizures (with >100 daily at baseline, and no seizure reduction at 3 months) and also because he lost 4.2 kg during his first 3 months on the LGIT, improving his BMI from 27.1 to 25.2 kg/m2. The second remained invested because of early seizure reduction that only later deteriorated, as well as a loss of 6.8 kg over the 9 months of treatment, reducing her BMI from 21.8 to 18.8 kg/m2. The third remained invested due to early seizure reduction that would later deteriorate.

Two patients discontinued the LGIT for reasons other than those described earlier. One had become seizure free on the combination of AEDs and the LGIT. She was tapered off the LGIT after 74 weeks. To date, this patient had been tapered off all medications but reinitiated one AED because of behavioral changes and epileptiform features on EEG; however, she remains seizure free. Another patient developed streptococcal pharyngitis while on the LGIT, and discontinued the LGIT during the illness. Five patients were lost to follow-up after 23, 21, 20, 9, and 7 weeks with reported seizure reductions at those times of <50%, >50%, increase, >50%, and >50%, respectively; it is unclear whether these five patients are still on the LGIT.

Interestingly, one patient continued on the LGIT despite an extended period of increased seizure frequency. At the time of LGIT initiation, she was also being treated with levetiracetam (LEV). Three months after initiation, LEV was increased in an effort to optimize seizure control. She experienced increased seizure frequency over the following 4 months, during which time LEV dosage was decreased and lamotrigine (LTG) was added. The patient’s parents felt the LGIT to be beneficial, and the combination of LGIT, LEV, and LTG ultimately resulted in seizure freedom.

Medication changes

AED regimens were changed during the course of the LGIT in most patients. To assess whether changes in medications were a confounding factor in LGIT efficacy, we compared 1-, 3-, 6-, 9-, and 12-month efficacy among patients with reduced AEDs, unchanged AEDs, and increased AEDs. No significant correlation was found between efficacy and AED changes at any of these follow-up intervals (Spearman’s correlation, p > 0.05 all five tests). Of note, the percentage of patients with an increase in AEDs increased with increasing duration on the diet, measuring 31%, 42%, 57%, 68%, and 83%, respectively, at the five follow-up intervals. In addition, we performed a subanalysis of 1-, 3-, and 6-month efficacy in only the patients with no AED changes over these intervals (Fig. 5). The 9- and 12-month follow-ups were excluded here because of insufficient numbers.

Figure 5.

 Percent seizure reduction, compared with baseline, at 1, 3, and 6 months, in only subset of patients with no change in antiepileptic drugs (AEDs) at given end point.

Side effects

Patients reported minimal side effects while on the LGIT. One patient who discontinued the LGIT after 13 weeks reported fatigue occurring in the middle of the day, which she attributed to the carbohydrate restrictions of the LGIT. Another patient reported lethargy and vomiting upon initiating the LGIT. These symptoms were suspected to be secondary to acidosis; however, laboratory values were not available. A third patient, identified as one of the 131 educated on LGIT but not included in this report’s analysis due to insufficient follow-up data, also reported a lack of energy upon initiating the diet.


Baseline and follow-up anthropometrics were available for 19, 28, 19, 17, and 9 patients, respectively, at the 1-, 3-, 6-, 9-, and 12-month follow-up intervals. Neither BMI nor BMI z-score changes from baseline were significant over any of these intervals (pair-samples t-test, p > 0.05). Nor did changes in BMI or in BMI z-score correlate with degree of seizure reduction at any of these time intervals (Spearman’s correlation, p > 0.05).

Blood chemistries

Serum glucose changes from baseline were available for 23, 26, 24, 14, and 8 patients, respectively, at 1-, 3-, 6-, 9-, and 12-month follow-up. These changes were not significant at any time point (paired-samples t-test, p > 0.05). Follow-up serum glucose levels were available for 26, 30, 25, 19, and 12 patients who also had efficacy data at the 1-, 3-, 6-, 9-, and 12-month follow-up intervals, respectively. A lower serum glucose correlated with a higher treatment efficacy at both the 1-month (Spearman’s correlation coefficient = −0.396, p = 0.045, Fig. 6A) and the 12-month (Spearman’s correlation coefficient = −0.662, p = 0.019, Fig. 6E) follow-up intervals. No correlation was found between serum glucose and efficacy at the 3-, 6-, or 9-month intervals (Fig. 6B–D).

Figure 6.

 Correlation between blood glucose level and LGIT efficacy at (A) 1-month, (B) 3-month, (C) 6-month, (D) 9-month, and (E) 12-month follow-ups. LGIT, low glycemic index treatment.

Baseline and follow-up serum BOHB levels were available for 16, 20, 16, 10, and 5 patients, respectively, at the five follow-up intervals. Change from baseline was significant at the 1-month [mean increase = 1.0 mM; 95% confidence interval (CI) = 0.3–1.7; p = 0.010), 6-month (mean increase = 1.3 mM; 95% CI = 0.4–2.3; p = 0.009), and 9-month (mean increase = 1.4 mM; 95% CI = 0.1–2.7; p = 0.039) follow-ups. Changes were not significant at the 3-month (mean increase = 0.4 mM; 95% CI = −0.1–0.8; p = 0.087) and 12-month (mean increase = 1.1 mM; 95% CI = −0.7–2.8; p = 0.167) follow-ups. Thirteen, 15, 16, 9, and 5 of these patients also had efficacy data at these follow-up intervals, respectively. Changes from baseline in serum BOHB did not correlate to efficacy at any of these points (Spearman’s correlation, p > 0.05). Specifically, efficacies were compared in patients with BOHB ≤0.5 mM and those with BOHB >0.5 mM, indicating ketosis, at each follow-up interval (Table 2), and again no difference was found in efficacy between these two groups at any of the five follow-up labs (Mann–Whitney test, p > 0.05).

Table 2.   Relationship between ketosis status (by serum BOHB level) and seizure reduction efficacy at 1-, 3-, 6-, 9-, and 12-month follow-ups
 Ketosis = BOHB >0.5Relationship with efficacy (p-value)a
n (%)Mean ± SD, mMRange, mM
  1. aMann–Whitney test.

  2. BOHB, β-hydroxybutyrate.

Baseline (n = 62)3 (5)4.3 ± 3.81.4–8.6
1-month follow-up (n = 22)10 (45)2.1 ± 1.10.7–4.00.283
3-month follow-up (n = 26)9 (35)1.7 ± 1.10.6–3.40.874
6-month follow-up (n = 21)10 (48)2.3 ± 1.80.7–5.50.918
9-month follow-up (n = 16)5 (31)2.9 ± 1.51.2–4.70.661
1-year follow-up (n = 11)4 (36)1.8 ± 1.10.7–2.70.648

Serum BUN, CO2, and urine specific gravity findings are shown in Table 3. No patient had a serum creatinine level exceeding 1 mg/dl. Changes from baseline serum BUN were available for 22, 24, 23, 12, and 7 patients, respectively, at 1-, 3-, 6-, 9-, and 12-month follow-up. A significant increase was seen at 1 month (mean increase = 3.5; 95% CI = 0.9–6.0; p = 0.011), 3 months (mean increase = 2.0; 95% CI = 0.3–3.6; p = 0.023), 6 months (mean increase = 3.3; 95% CI = 1.1–5.5; p = 0.005), and 9 months (mean increase = 3.2; 95% CI = 0.2–6.2; p = 0.040). The increase at 12 months did not reach statistical significance (mean increase = 1.9; 95% CI = −2.3–6.0; p = 0.319). Of note, many patients demonstrated transient increases in BUN that were not consistent from one follow-up lab to the next. Changes from baseline to 1-, 3-, 6-, 9-, and 12-months in serum CO2, serum creatinine, and urine specific gravity were not significant (paired-samples t-test, p > 0.05).

Table 3.   Serum BUN, serum CO2, and urine specific gravity findings at baseline, 1, 3, 6, 9, and 12 months in patients with elevated BUN
 Elevated BUNa
n (%) BUN range, mg/dl Decreased serum CO2bElevated urine specific gravityc
  1. a>18 mg/dl in patients under 18 years of age; >21 mg/dl in patients 18 years of age or older.

  2. b<21 mM.

  3. c>1.015.

  4. BUN, blood urea nitrogen.

Baseline (n = 50)5 (10)19–2213 of 3
1-month follow-up (n = 31)10 (32)19–3014 of 7
3-month follow-up (n = 34)9 (26)19–2524 of 4
6-month follow-up (n = 27)9 (33)19–2406 of 6
9-month follow-up (n = 18)6 (33)19–2615 of 5
1-year follow-up (n = 13)4 (31)19–2501 of 2


Although the exact therapeutic mechanisms of the fasting state and the KD remain unknown, many of the metabolic changes that accompany such diets have been investigated. The most obvious change is the elevation of serum ketone bodies and their partial substitution for glucose as a major metabolic substrate of the brain (Owen et al., 1967; Appleton & DeVivo, 1974; DeVivo et al., 1978; Schwartzkroin, 1999). Although many have proposed that ketosis is the primary mediator of the antiepileptic effects of the KD, a correlation between the degree of ketosis (concentration of ketone bodies) and the level of seizure control has not been demonstrated consistently (Appleton & DeVivo, 1974; Bough et al., 1999). Furthermore, animal studies have shown that ketone bodies do not directly alter excitatory or inhibitory synaptic transmission in hippocampal slices (Thio et al., 2000). These findings suggest that ketosis alone may not account for the anticonvulsant effects of the KD (Greene et al., 2003). Fasting and the KD also have significant effects on glucose levels (Schwartzkroin, 1999; Greene et al., 2001). Glucose production and utilization are significantly decreased in the fasting state or while on the KD, resulting in a net decrease in blood glucose levels (Haymond et al., 1983). Blood glucose levels also stabilize in the absence of additional energy substrates, even for prolonged periods (Valencia et al., 2002). More recent studies of flurothyl-induced seizures in rats have shown circulating glucose levels to be positively correlated with seizure susceptibility (Schwechter et al., 2003). Furthermore, with food or glucose intake, seizures can be induced rapidly in association with rising blood glucose and falling ketone levels (Lennox & Cobb, 1928; Huttenlocher, 1976; Greene et al., 2003). These findings suggest that glucose and perhaps related metabolic factors play a significant role in seizure control. The LGIT is designed to prevent dramatic fluctuations in blood glucose.

With substantially higher sample size and follow-up evaluation than has previously been explored in the literature (Pfeifer & Thiele, 2005), this pediatric epilepsy population demonstrates the efficacy of the LGIT in a population with an average of approximately six former or current AEDs. Observed efficacy in this population (Figs. 2 and 5) approached that of the KD (Henderson et al., 2006). Not unlike the KD, however, the exact mechanism of the LGIT remains unclear. Although not extensively studied in epilepsy, low-GI diets are known to lower postprandial plasma glucose and insulin profiles among diabetics and patients with cardiovascular disease (Brand et al., 1991; Fontvieille et al., 1992; Wolever et al., 1992; Jarvi et al., 1999; Rizkalla et al., 2004). In a recent study on the metabolic effects of a low-GI diet in a healthy population, investigators used the MiniMed continuous glucose monitor, which measures blood glucose every 5 min over a 24-h period; they found that individuals on a low-GI diet for 1 week had improved glucose profiles (Brynes et al., 2005). In the present study, we report a correlation between lower blood glucose levels and improved efficacy of the diet, but not at all time points (Fig. 6). Therefore, although it is possible that the seizure protection afforded by the LGIT is due to lower, more stable blood glucose and insulin levels, higher-powered studies with more strictly regulated postprandial blood glucose tests (the present study used random blood glucose tests) are needed to confirm such a theory.

We did not observe a relationship between BOHB levels and efficacy in this study. Among the patients with documented ketosis on the LGIT, efficacy ranged from >90% reduction to increased seizure frequency. Furthermore, some patients achieved dramatic improvements in seizure frequency, even in the absence of ketosis. Of 12 patients with >90% reduction at 3 months, four had normal BOHB (0.0–0.2 mM), three had elevated BOHB (0.7–1.7 mM), and five had no follow-up BOHB levels available at that time point; ketosis is, therefore, not required for excellent seizure control. Similarly, effective seizure control has been documented in the absence of elevated ketones in an Atkins-type diet (Kossoff et al., 2006). On the KD, comparatively, evidence for a correlation between seizure control and serum BOHB is inconsistent (Bough et al., 1999; Gilbert et al., 2000) and thus the role of elevated blood ketone bodies in the treatment of epilepsy remains unclear. On the LGIT, however, it remains possible that the elevation of BOHB seen in some patients may have some neuronal effect.

In our experience, the LGIT is better tolerated and more palatable than the KD. The KD, although efficacious, continues to be difficult for most patients and families in terms of compliance and tolerability. In addition to the rigid and complicated meal plans, the food is viewed as unpalatable. Furthermore, the dietary restrictions of the KD may be accompanied by psychosocial issues: some patients experience social isolation because they eat obviously distinct foods from their peers. In contrast, only 18 (24%) of the patients in the present study found the LGIT too restrictive. Anecdotal evidence suggests that the LGIT is easier to tolerate for several reasons. First, meals are easier to prepare so that detailed meal plans are not required and portions need not be weighed with a gram scale. Second, the LGIT is more palatable because of the liberalized carbohydrate content and decreased fat content. Third, there are fewer psychosocial issues because the food is viewed as more “normal” and patients are able to eat outside the home without having to prepare special meals in advance. Finally, the LGIT, unlike the KD, can be initiated in an outpatient setting. However, despite the relative ease of management compared with the KD, the LGIT remains too restrictive for some families. In addition, some patients note that the liberal nature of the LGIT, specifically in comparison to the KD, actually make “cheating” easier and thus compliance more difficult. Accordingly, the determination of which patients will better tolerate the LGIT over the KD must be made on a case-by-case basis.

Our study also suggests that the LGIT may be associated with fewer side effects than the KD. Common side effects reported with the KD include metabolic acidosis, renal calculi, acute pancreatitis, constipation, and hyperlipidemia (Sinha & Kossoff, 2005). None of these adverse events were noted on the LGIT. Three patients had transient lethargy on the LGIT, one of which was suspected to be caused by acidosis and spontaneously resolved without further sequelae, whereas the other two only maintained the diet for a short duration as a result. The approximately one-third of patients with elevated BUN at each follow-up set of labs demonstrates the expected result of any diet high in protein such as the LGIT. However, although a relationship has been demonstrated between high protein intake and declining renal function in patients with established kidney disease, no such relationship has been observed in patients with normal renal function (Martin et al., 2005). Therefore, preexisting renal disease may be a contraindication for this diet, but compromised renal function would not otherwise be expected. Relative dehydration may have also contributed to the increased BUN levels in this population, as we did observe a high percentage of patients with elevated urine specific gravity among those with elevated BUN (Table 3). A third potential cause for increased BUN is metabolic acidosis, which we did not observe at an appreciably high rate in this population. It is, therefore, recommended that patients on the LGIT be closely followed by labs for modification of protein intake, fluid intake, and potential potassium citrate supplementation, as necessary. In addition, the BMI and BMI z-scores indicate that with appropriate supervision by an experienced dietician, it is possible to avoid undesired weight changes while instituting the LGIT.

This retrospective study had several limitations. Duration of therapy was limited, in part because some patients in our population used the LGIT only as a transitional therapy. This limited our ability to assess longer-term efficacy. Most patients also altered AED regimens over the course of their treatment, which could be partly responsible for changes in seizure control. However, subgroup analyses suggest that AED changes did not significantly affect changes in seizure frequency. The fact that LGIT efficacy appeared to increase at successive follow-up appointments was likely secondary to selection bias: patients with better seizure control tended to remain on the diet longer. Finally, the small sample size (especially with regard to laboratory results) limited our statistical power; future studies with larger sample sizes may show more consistently significant relationships than did this study.

Low-carbohydrate diets should be used with caution as they do alter the body’s normal metabolism (Bravata et al., 2003). It is, therefore, recommended that patients be followed by an experienced dietitian and neurologist. Initial and follow-up monitoring of weight and blood chemistries is also recommended.


We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Conflicts of interest: None of the authors has any conflict of interest to report.