SEARCH

SEARCH BY CITATION

Keywords:

  • Ketogenic diet;
  • Epilepsy;
  • Fasting;
  • Children;
  • Randomized crossover trial.

Summary

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Despite over 80 years of use, the ketogenic diet (KD) has never been tested in a blinded manner. Twenty children with intractable Lennox-Gastaut syndrome (LGS) were fasted 36  h and then randomized to receive the classic KD in conjunction with a solution containing either 60 g/day of glucose or saccharin. Parents and physicians were blinded to both the solution composition and level of ketosis. A crossover to the KD with the alternate solution occurred following the sixth day and a repeat fast. A 24-h electroencephalography (EEG) was obtained at baseline and after each arm. After administration of the solution, there was moderate evidence of a reduction in parent-reported seizures between the glucose and saccharin arms, with a median difference of 1.5 seizures per day (p = 0.07). There was no reduction in the number of EEG-identified events, with a median reduction of 7 events per day (p = 0.33). Ketosis was not completely eliminated in the glucose-added arm.

Despite widespread evidence that the ketogenic diet (KD) is an effective therapy for many children with epilepsy, the absence of blinded, randomized, and controlled (class 1) data has been an impediment to its full acceptance (Henderson et al., 2006; Wiznitzer, 2008). A recently completed controlled study demonstrated highly significant results in favor of the KD, but was not blinded (Neal et al., 2008). Our study was designed to assess, in a blinded, controlled fashion, the short-term efficacy of the KD in children with Lennox-Gastaut syndrome (LGS). We believed that a study solution containing 60 g of glucose, when given over a day, would negate both urinary and serum ketosis and therefore create a placebo arm (Huttenlocher, 1976;  Freeman & Vining, 1999), whereas an artificial sweetener (saccharin) similar in taste would represent a treatment arm.

Methods

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Children with intractable atonic-myoclonic or “drop” seizures associated with LGS were selected for the study because of the daily frequency of their seizures and the historical response to the KD ( Markand, 1977;  Vining et al., 1998). Affected children have many of these drop seizures per day, characterized by either mild slumping or more active falls to the floor, often associated with injury. The  efficacy of the KD in decreasing the frequency of these atonic-myoclonic seizures was the subject of this study therefore.

Inclusion criteria included ages 1–10 years, prior exposure to at least two anticonvulsants, electroencephalography (EEG) evidence within 6 months of the typical LGS pattern of 2–2.5 Hz spike and slow wave discharges, and an average of at least 15 atonic-myoclonic seizures per day by parental records over the prior month. Children were excluded if there was evidence for a metabolic disorder, treatment with steroids or adrenocorticotropic hormone (ACTH) in the prior month (which would theoretically negate ketosis), or previous KD treatment. This study was approved by both the Johns Hopkins Pediatric General Clinical Research Center and Institutional Review Boards.

Children meeting recruitment criteria were admitted, and a 24-h Digitrace (16-channel ambulatory) EEG (SleepMed, Inc., Columbia, SC, U.S.A.) was obtained. Parents were required to room-in with their children and were asked to push a Digitrace button whenever they saw their child have a drop or atonic-myoclonic seizure and to also record these push-button events on a calendar. Other seizures, including absence, tonic, and generalized tonic–clonic seizures, were not recorded. Although parents in general observed their children during waking hours, the duration, sampling, and timing of the observation period for these parent-identified seizures was not standardized over the 24-h period.

The EEG was then reviewed by the electroencephalographer (E.P.G.V.) to identify the electrographic signature (often electrodecremental changes) that accompanied these push-button events. The entire record was also reviewed for additional identical “electrographic” events. Utilizing the EEG-identified events as one end point was intended to minimize any bias from the parents and any loss of data that may have occurred when a child was not being continuously observed by a parent.

One hundred thirteen parents originally inquired about the study, with 33  children (29%) qualifying by historical records. Thirteen patients were subsequently disqualified either because they failed to meet sufficient seizure frequencies on the day of the baseline EEG or the reader was unable to correlate push-button with EEG-identified events because of a grossly distorted EEG background. Twenty children who had at least 15 distinct EEG-identified events were enrolled.

The study design is shown in  Fig.  1. All children were initially fasted for 36 h. Serum glucose was measured every 6 h, with small amounts of oral glucose given for glucose <30 mg/dL and clinical symptoms (emesis, fatigue). Twenty-four-hour EEGs were obtained on days 1, 6, and 11. The reader and caretakers were blinded to the treatment arm.

image

Figure 1.  Study design and implementation.

Download figure to PowerPoint

During each day of the study, the child was given a  solution of sweetened, flavored water that replaced an equivalent portion of the typical carbohydrate-free fluid allotment. Children were randomized in a double-blinded manner to receive first either a solution containing 60 g of glucose (in three separate 20-g aliquots per day) or a similar saccharin solution. The order of presentation was determined by the study nurse who also created the solutions and checked urinary ketosis and serum β-hydroxyburate (BOH) daily, using ketone strips and a portable meter, respectively. Parents, children, and physicians were blinded to serum BOH and urine ketone results as well. Medications were not changed during the study period. All children were seen subsequently as outpatients for efficacy after discharge.

Statistical analyses included the two-tailed Wilcoxon rank-sum test for the comparison of changes in outcome variable in response to treatment for the two-period crossover, assessing the treatment effect and the difference in carry-over effects. The correlation between the two methods of measurement (push-button versus EEG-identified) was assessed using Spearman correlation coefficients. The type I error rates for all the statistical comparisons were set at a level of 0.05. All analyses used SAS statistical software 8.2 or JMP version 6.0 (SAS Institute, Cary, NC, U.S.A.).

Results

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

Twenty children were enrolled from November 1997 to November 2002, with 11 given glucose first. Mean age was 3.9 years (range: 1.0–7.4 years), and eleven (55%) were male. A 4:1 ketogenic ratio (fat:carbohydrate and protein) was used in 13 patients (65%), with the others started on a 3:1 ketogenic ratio. In general, a 3:1 ketogenic ratio was used for children 3 years of age or younger to increase protein allowance. All children completed the 12-day protocol.

There was a trend towards statistical significance based on parent-reported clinical events between the ends of the saccharin and glucose arms; median change, −1.5 seizures per day (range, −48 to 21 seizures per day, p = 0.07). Dissociation existed between clinical and EEG-identified events. There was no statistical difference in EEG-identified events between the ends of the two arms; median change, −7 events per day (range, −141 to 357 events per day, p = 0.33). Six children showed both >50% seizure reduction with saccharin in addition to <50% improvement with glucose, compared to three children who demonstrated the opposite response (p = 0.50). The sequence of treatment arms did not make a difference in EEG-identified events (p = 0.32), but had borderline statistical significance for a difference in the push-button events (p = 0.07), with the saccharin arm given first being more efficacious.

At the end of the first evaluation period (day 6), the total EEG-identified events decreased by a median of −22.5 seizures per day (range, −517 to 112 seizures per day, p = 0.03) and the push-button events by a median of −14.5 seizures per day (range, −52 to 14 seizures per day, p = 0.001). Comparing baseline seizure frequency to day 12, there was a significant reduction in seizures with a median decrease of −34 seizures per day (p  = 0.003). Sixty-five percent of patients experienced >50% reduction in seizures over the study period. Six months after discharge, 80% had a >50% decrease in reported seizures, and at 12 months 65% still had a >50% decrease.

On the final days of each saccharin arm of the study, urinary ketones were uniformly large (80–160 mg/dL);  however even during the glucose arm, ketones were still typically trace to moderate (15–60 mg/dL). There was a significant difference between the serum BOH of children during the glucose arm compared to the saccharin arm (2.7 versus 6.0 mmol/L, p < 0.001). Four children (22%) had absent urinary ketosis on the final day of the glucose arm, although all had measurable levels of serum BOH.

The protocol was well-tolerated despite the two 36-h fasting periods over 12 days. Six children had emesis, typically during the fasting period or immediately afterwards, one of whom required 1 day of glucose-free intravenous fluids. Three additional children were fatigued during the fasting period. Hypoglycemia (serum glucose <30 mg/dL) occurred in six children (25%) during a fast.

Discussion

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

We were successful in designing and completing a randomized, crossover-design, double-blind study of the KD during which children, parents, and physicians involved were unaware of treatment arm assignment, ketosis level, and EEG results. Unexpectedly, only when the blind was broken and results analyzed, did we become aware that the study design resulted in an active control. Despite preliminary evidence that 60 g glucose per day would be sufficient to negate the ketosis of the ketogenic diet (Huttenlocher, 1976); ketosis was not eliminated in the planned placebo group. No child in the glucose-added arm of this study had absent serum BOH, and even the lower levels of BOH in this group may have been sufficient to reduce seizures, as has been demonstrated in studies of both a modified Atkins diet (Kossoff et al., 2006;  Kang et al., 2007) and a low glycemic index treatment (Pfeifer & Thiele, 2005). Alternatively, the two 36-h fasting periods may have reduced seizures so effectively that even theoretically higher glucose amounts, were they given, may not have led to a worsening of seizures in these short-duration treatment arms. Lastly, fasting itself can have a dramatic and immediate effect on seizures as seen in this study and in others (Freeman & Vining, 1999;  Than et al., 2005;  Kossoff et al., 2008). Nevertheless, there was a highly significant decrease in the number of clinically observed seizures in both groups over the 12-day study (p < 0.001).

Future studies of the KD using this novel glucose and saccharin design may be warranted and could avoid this study's weaknesses. We suspect that by both providing additional quantities of glucose (>100 g/day) and not fasting (Bergqvist et al., 2005), ketosis could be rapidly eliminated successfully. In addition, allowing enough time for children to return to a normal, nonketogenic diet (and baseline seizure frequency) might result in an effective crossover study design, but would be difficult to implement. Evaluating children as outpatients over months rather than days might also be more reflective of the actual usage of the KD, as designed in a recently completed trial (Neal et al., 2008). Lastly, while EEG identification of seizures may seem a more objective end point, they are clearly different than drop seizures and may not represent clinically important events in children with LGS.

Unfortunately, the dramatic effects of fasting and probably insufficient glucose quantities in this study prevented statistically significant differences between the two arms of the study. We believe future studies can learn from and improve upon this study design.

Acknowledgments

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References

This study was supported by the National Institutes of Health (NIH) grant no. RO1NS35980-01A1 and by the Pediatric Clinical Research Unit, NIH/National Center for Research Resources grant no. MO1-RR00052. We wish to express our appreciation to our colleagues, James E. Rubenstein, MD, for his care of patients in the study, Diana Pillas for her recruitment efforts, our dietician, Jane R. McGrogan, RD, and our study nurse, Heather Hladky, RN. We also express our appreciation to Digitrace for providing us with their equipment for this study.

Conflict of interest: 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. None of the authors have any conflicts of interest to disclose.

References

  1. Top of page
  2. Summary
  3. Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. References