Does a preference for fatty foods prior to commencing treatment with the ketogenic diet predict the efficacy of this diet?

Article Outline

• Abstract
• 1. Introduction
• 2. Methods
  • 2.1. Particpants
  • 2.2. Procedure
  • 2.3. Statistical analysis
• 3. Results
• 4. Discussion and conclusion
• Appendix A. Supplementary data
• References
• Copyright

Abstract 

The ketogenic diet can be effective in children who suffer from drug-resistant epilepsy. However, it is still hard to predict how large an effect this diet will have for an individual child. Previous data suggests a high-fat food preference is more likely in those with epilepsy, as assessed by in-person forced-choice design. The aim of this study is to examine whether a partiality to fatty foods prior to commencing the ketogenic diet can be used as a predictive factor for the efficacy of this diet in children with drug-resistant epilepsy.

Data from 43 children aged between 2 and 19 years was used in this retrospective, non-controlled, non-randomised, open study. All children had followed the ketogenic diet for a period of 3 months or more. Before commencing the diet, a food record was collected for each child to determine the percentage of daily energy-intake accounted for by fats. Parents of the participants completed a questionnaire to measure fat-preference in the pre-diet period and received a score to objectify the efficacy of the treatment.

The raw scores on the food record and on the questionnaire were divided into subgroups. Subsequently Kendall's tau-b was calculated for the correlation between each combination of variables. A non-significant correlation was found for the relationship between the food record and the questionnaire (p=.939), the relationship between the food record and the efficacy of the treatment (p=.827) and the relationship between the questionnaire and the efficacy of the treatment (p=.539). This means treatment efficacy cannot be predicted by the child's food preference.

Keywords: Epilepsy, Ketogenic diet, Food preferences

Back to Article Outline

1. Introduction 

Epilepsy is acknowledged as a disease worldwide. With an estimated prevalence of 1%, it is the most widespread major neurological disorder.¹ According to statistics published by the RIVM (Dutch National Institute for Public Health and the Environment),² there are upwards of 7600 new cases identified in the Netherlands each year. Twenty-five percent of these cases are not treatable with medication, resulting in annually more than 1900 people hearing that they could possibly suffer from uncontrollable seizures for the rest of their lives. Fortunately for these people a non-medicinal treatment also exists, namely a treatment with the ketogenic diet.

The ketogenic diet is a diet that consists predominantly of fats, with a fat to carbohydrates and protein ratio becoming as much as 3:1 or 4:1.³ Although the exact response mechanism still has not been clearly defined, it has long been known that a high concentration of ketones in the body can cause major reductions in seizure frequency. Multiple studies have shown that the ketogenic diet can accomplish a more than 90% reduction in seizures in one-third of all patients.⁴

However, although one-third of all patients is promising, this treatment is still by no means 100% effective. Researchers have often asked why this ketogenic diet works for some patients but not for others, and multiple studies have attempted to answer this question. In recent studies, children with a comorbid GLUT1 deficiency syndrome or a pyruvate–dehydrogenase-complex-deficiency appear to do very well on the ketogenic diet in comparison with children without such deficiencies.⁵ Studies on mice have shown that genetic factors also play a role.⁶ Factors that do not seem to be of any influence are a hypo- or hyperglycaemic status (with the exception of evaluating for Glut-1 deficiency),⁷ a fluctuating or lasting increased or reduced body-mass-index,⁸ age, seizure type (except for myoclonic-astatic epilepsy and infantile spasms, these children also appear to do very well on the diet) or etiology.⁹ A disorder in fat metabolism is indicated as negatively influencing the effectiveness of the ketogenic diet.⁴

Despite all these studies mentioned above and other studies that are not described here, there are still no clear results that can help clearly define which factors predispose a patient to a positive reaction on the ketogenic diet. It is vital not to underestimate the importance of resolving this problem: not only could this spare patients a potentially disappointing treatment, but the economic savings resulting from more effective allocation of this treatment are substantial. In 2005, Beghi et al. estimated the costs of the ketogenic diet to be $US 36.234, per child for a treatment period ranging from 6 to 12 months.¹⁰ This is a substantial amount of money that would be better spent on a child that shows a good response to the diet than on someone who shows no response at all.

It therefore remains crucial to try to better understand what factors have a predictive value in the efficacy of the ketogenic diet. The hypothesis arose that the search for these factors should continue on a more individual level with respect to the patient. It is known that patients suffering from many physical disorders automatically show compensating behaviour. This applies for simple cases, such as a hypotensive patient who gets a thirst signal and starts drinking, as well as for more complex disorders like a patient suffering from depression who grabs for a chocolate bar.¹¹ Could it be that a similar response mechanism exists for an epileptic patient? Amari et al. asked themselves the same question.¹² They studied the preference for high-fat foods in children with seizures in comparison with high-carbohydrate foods and a control group. Their results show that children with seizures indeed prefer high-fat foods more than high-carbohydrates foods and differ significantly with the control group. They also asked the parents of these children to predict the preference of their child. It appeared that parents in both groups (children with seizures and controls) displayed low accuracy in these predictions.

Nor Amari et al., nor someone else in the literature, studied the relationship between a preference for high-fat foods in children with epilepsy and the efficacy of the ketogenic diet. When various parents of children with epilepsy being treated with the ketogenic diet indicated to the researchers that their children seemed to have a preference for fatty foods prior to this diet, our research question was born. This study aims to analyse the question if a preference for fatty foods prior to the start of the ketogenic diet is a predictive factor for the efficacy of this diet in children with drug-resistant epilepsy.

Back to Article Outline

2. Methods 

2.1. Particpants 

For this study data from 43 children in the age group 2–19 years were included. All of these children were treated with the ketogenic diet at the Wilhelmina's Children Hospital in Utrecht (the Netherlands) during or prior to this study. Children were eligible for inclusion if they followed the ketogenic diet for at least 3 months.

Those children who were younger than 2 years at start of the treatment, who had a comorbide metabolic disorder, who prior to the diet were completely fed by enteral tube feeding or those who were resident in an institution at the time, were excluded from the study. When selecting which children to send the questionnaire, two more exclusion criteria were applied: no questionnaire was sent to those children whose diet had started prior to 2003, nor to the families of children who had passed away.

2.2. Procedure 

Every child treated with the ketogenic diet had to provide a food record written up for the period prior to commencing the diet. This record registers everything the child eats during 3 days, 2 weekdays and 1 day in the weekend. Of every item the precise quantity is measured and written up in the record. A completed record is analysed with a specialized calculation program that uses the Dutch Composition Table 2003.¹³ From these documents the child's eating pattern could be discerned and the percentage of nutrition comprising of fats could be determined.

To identify a potential preference for fatty foods, the researchers searched for a validated questionnaire aimed at children, which could demonstrate this preference. Unfortunately, such a questionnaire did not yet exist, so a new questionnaire was designed in cooperation with a dietician. This questionnaire consisted of 20 questions which all had to be answered by the parents of the child.

The questionnaire was sent to 34 children. An accompanying letter explained the purpose of the study and asked the children's parents, not the children themselves, to complete the questionnaire. It was emphasised before each question that the answer had to be consistent with the situation before their child started the ketogenic diet.

Out of the 34 questionnaires sent, 28 (82%) were returned. Twenty-three (68%) of these were usable for analysis and 4 (12%) were incomplete and could not be taken into account for the analysis. One questionnaire (3%) was returned stating ‘not participating’.

Every parent who filled in and returned a complete or incomplete questionnaire later received a telephone call from one researcher to go through the questions once again together, enabling any ambiguities to be clarified for both parties. Following these telephone calls, the answers on the questionnaires were converted to scores and a total score was calculated for every child.

Finally, the efficacy of the ketogenic diet needed to be determined for each child that participated in the study. This was accomplished by means of a 4-points scale; a score of 1 equated to a reduction in seizure frequency of 100%, a score of 2, 3 and 4 equated to a seizure reduction of 90–100%, 50–90% and <50% respectively. A fifth group has been created by combining the scores 1 and 2 into one group (seizure reduction ≥90%). All participating children were assessed by the same neurologist.

2.3. Statistical analysis 

The design of this study was a retrospective, non-controlled, non-randomised, open study. In order to answer the research question, four analyses were run on the data. All analyses were done with the statistical program ‘SPSS for Windows’, version 15.0.

Before beginning the analyses both the score for the fat-intake prior to the ketogenic diet, as well as the total score on the questionnaire were divided into groups. These transformations had to be done because of the floor- and ceiling effects and skewed distributions, which were seen after data acquisition.

Participants were divided on the basis of fat-intake prior to treatment in the following manner: when less than 25% of the daily energy-intake consisted of fat, this was classified as group 1. When the fat-intake fell between 25% and 35%, this was labelled group 2. Lastly, a total fat-intake of more than 35% was designated as group 3.

Participants were grouped using the total score of the questionnaire in much the same way: a total score of less than 45 points (low preference for fats) became class 1, a score between 45 and 75 points (neutral preference for fats) class 2, and a score of 75 points or more (high preference for fats) class 3.

Three variables originated from these transformations, each with several subgroups. The first variable is the percentage of fats in the daily energy-intake prior to treatment with the ketogenic diet, divided into three groups. The second variable is the total questionnaire score, and was also divided into three groups. The third and final variable is the efficacy of the treatment. This variable consists of five subgroups. During the first analysis the correlations between all these subgroups are computed: Kendall's tau-b was calculated to show a possible relationship between one subgroup from one variable with a subgroup from another variable.

Subsequently, all three variables were compared to each other as whole groups. Initially the association between the scores on the questionnaire and the efficacy of the treatment is determined. This was achieved by calculating Kendall's tau-b for the correlation between these two variables. This analysis will demonstrate if children who have a preference for fatty foods (as indicated by their parents) show a different result of treatment with the ketogenic diet than children without this assumed preference. Because only 23 data pairs could be used in this analysis, the researchers have searched for a way to include more data.

To accomplish this, Kendall's tau-b is computed for the relationship between the fat-intake prior to the treatment and the efficacy of the treatment. This answers the question whether children who were already used to a large portion of fat in their nutrition, showed a different treatment outcome to children who did not eat such large amounts of fat. However, this still does not tell us anything about a possible preference for fatty food.

To determine if there is a connection between the amount of fat eaten prior to start of the diet and a preference for fats, predicted by the parents of the children, these two variables (variables 1 and 2) are compared to each other. Using Kendall's tau-b, the correlation between the fat-intake prior to the treatment and the score on the questionnaire is determined. When this correlation appears to be sufficiently high, we can assume that there is indeed a relationship. This relationship then can be used to interpret the analysis described in the previous paragraph: it is now plausible to assume that a demonstrated correlation between the amount of fat in the nutrition before the ketogenic diet and the efficacy of the treatment, is the result of a child's preference for fats.

When there fails to be a correlation between the score on the questionnaire and the score on the food record, we can assume that parents are no good in predicting the food preference of their child, a result that needs to be expected after Amari's study,¹² and cannot use the results of the questionnaire to say something about preferences of the child.

Back to Article Outline

3. Results 

Before starting the analyses, the unprocessed scores of some variables were divided into subgroups. Subsequently, frequency distributions were drawn up for each variable subgroup.

Table 1 displays the frequency distribution for the fat-intake prior to commencing the ketogenic diet. This table shows that in the case of 5 children (11.6%), their daily energy-intake consists of less than 25% fat. The fat percentage of the daily energy-intake lay between 25% and 35% for 21 children (48.8%), and 35% or higher for 17 children (39.5%).

Table 1. Frequency distribution for the percentage of energy-intake that is accounted for by fats prior to start of the ketogenic diet.

The frequency distribution of the questionnaire scores that measured a preference for fatty foods are presented in Table 2. Using this questionnaire, 4 children (17.4%) scored lower than 45 points, 8 children (34.8%) scored between 45 and 75 points and 11 children (47.8%) scored 75 points or more.

Table 2. Frequency distribution for fat-preference.

Table 3 shows the frequency distribution for the efficacy of the treatment. As can be seen here, 8 children (18.6%) presented with a seizure reduction of 100%. An additional 8 children (18.6%) showed a reduction of 90–100%. Six children (14.0%) showed reduction in seizures of 50–90% and 21 children (48.8%) displayed a reduction of less than 50%.

Table 3. Frequency distribution for efficacy of the treatment.

After the formation of these subgroups, the inter-correlation for every combination of subgroups could be computed by means of Kendall's tau-b. Table 4 shows the result of this analysis.

Table 4. Inter-correlation coefficients (Kendall's tau-b) per subgroup.

	Percentage of fats in food prior to ketogenic diet			Preverence for fats
	<25% (n=5)	25–35% (n=21)	>35% (n=17)	Low (n=4)	Neutral (n=8)	High (n=11)
Preverence for fats
Low (n=4)	−.178	.250	−.133
Neutral (n=8)	.259	−.334	.163
High (n=11)	−.112	.129	−.054
Efficacy of the treatment
≥90% seizure reduction (n=16)	.166	−.098	−.010	.387	−.342	.032
100% seizure reduction (n=8)	.009	−.243	.244	.395	−.335	.020
90–100% seizure reduction (n=8)	.196	.121	−.256	.092	−.094	.020
50–90% seizure reduction (n=6)	.060	−.136	.100	−.178	.259	−.112
<50% seizure reduction (n=21)	−.203	.191	−.061	−.250	.151	.045

When comparing the subgroups of the fat-intake prior to the ketogenic diet and the subgroups of the efficacy of the treatment, there does not appear any significant correlation.

None of the correlations between the fat-intake prior to the ketogenic diet and a preference for fatty food is significant.

Furthermore, following the comparison between a preference for fat and the efficacy of the treatment, no significant correlation is found. There does however seem to be a tendency towards a positive correlation between a low preference for fats and a good efficacy of the treatment (100% seizure reduction) and a negative correlation between a low preference for fats and a poor efficacy of the treatment (less than 50% seizure reduction). However, as mentioned none of these correlations showed a significant value.

The ensuing analyses were completed to determine the correlations between the three covering variables. Kendall's tau-b is calculated for the correlation between the score on the questionnaire and the efficacy of the treatment. This analysis provides a correlation coefficient of .131 (p=.539). Kendall's tau-b was also calculated for the relationship between the efficacy of the treatment and the daily fat-intake prior to the treatment. This analysis showed a correlation of −.028 (p=.827). Finally a correlation of .012 (p=.939) was found when Kendall's tau-b was computed for the relationship between the fat-intake prior to treatment and the score on the questionnaire. None of these last three analyses can be considered to show a significant correlation.

Back to Article Outline

4. Discussion and conclusion 

The aim of this study was to determine if a preference for fatty foods prior to start of a treatment with the ketogenic diet could be applied as a predictive factor for the efficacy of this diet for children with drug-resistant epilepsy. To answer this question, a food record, a questionnaire to indicate a fat-preference and a score for the efficacy of the treatment were used.

After data analysis, there did not appear to be any relationship between the percentage of fat in the nutrition prior to commencing the ketogenic diet and an actual preference for fatty food, as measured by the questionnaire. As concluded by Amari et al., parents are typically not accurate in predicting the food preference of their child.¹² With the food record, a reliable measurement of the child's eating habits, we controlled for this. Amari's conclusion is proved by our results once again and makes it hard to draw any conclusions from the absence of the relationship between the scores on the food record and on the questionnaire.

The fat percentage of the nutrition prior to the treatment showed no significant correlation with the efficacy of the treatment when these two whole variable groups were compared. Also, no significant correlation was found when comparing the different subgroups. To make our analysis more specific, the group of children who showed an efficacy of the treatment ≥90% seizure reduction, was divided in two groups. The first group consisted of children who were free of seizures after the treatment, the second group were those who showed a ≥90% reduction in seizures, but were not free of seizures. After this division, there still failed to be a significant correlation, but the correlation between children who had a contribution of more than 35% of fats in their daily energy-intake prior to the ketogenic diet and the children who presented with a seizure reduction of 100% after the treatment showed an almost significant positive correlation. One could argue now that children who ate a large amount of fat prior to start of the treatment, have a better chance at a seizure reduction of 100% in comparison with children who are used to smaller amounts of fat. However, because the overall-correlation is low and not considered significant, this statement cannot actually be justified. Furthermore a high fat percentage in the nutrition prior to the treatment shows a negative, albeit insignificant correlation with a seizure reduction of 90–100%. A reduction of 90–100% in seizure frequency is an equally substantial result, however children who are used to large amounts of fats in their diet supposedly have a smaller chance of achieving this reduction. In conclusion, it would seem that it makes no difference to the efficacy of the ketogenic diet whether a child is used to more or less fat intake prior treatment with the diet.

The final correlation that was determined, was the one that would show a relationship between a preference for fatty foods and the efficacy of the treatment. This relationship could not be found. When parents cannot accurately predict their child's preference, no significant correlation can be expected. The fact that there indeed was no significant correlation found, supports the conclusion of Amari et al. about the predictive ability of parents. On closer inspection at our figures, there even seems to be a tendency of a positive relationship between a low preference for fat and a good efficacy of the treatment (100% seizure reduction) and a negative relationship between a low preference for fat and a lesser efficacy of the treatment. Although none of these correlations were significant, it is a further reason to doubt the validity of the hypothesis or the ability of parents to predict the food preference of their child. To conclude, these findings do not support the hypothesis that children who have a higher preference for fatty foods show a different efficacy of the treatment with the ketogenic diet than children without such a preference, but does not invalidate it either because of the non-validated questionnaire.

The methodological design of this study remains debatable. Firstly, the choice of a retrospective study automatically has some disadvantages; the parents who were finally sent a questionnaire were sometimes required to answer questions about a time more than 5 years in the past, and they can therefore not realistically be expected to recollect every detail of the situation. This obviously prevents issues for the reliability of this study.

A further concern for this study is that it made use of an invalidated questionnaire for measuring food-preferences and did not compare the resulting scores with a control group. Therefore, results that originate from the questionnaire must be interpreted cautiously.

In addition to the above issues, the low power of the test remains problematic. Due to the nature of the research, only a small effect-size was expected. Consequently, a large N was needed to reach the desired power. Unfortunately, it was impossible to realize the required number of participants, which resulted in a lower than desired power of the test. This also decreased the chances of finding a possible significant result.

Nevertheless, despite these methodological shortcomings and the insignificant results, an important conclusion can still be drawn from this study. The research could not demonstrate a better or different efficacy of the ketogenic diet for children who ate more fatty foods prior to this treatment in comparison with children who did not ate as much fats. This means that the fact that parents sometimes come to the researcher stating they expected a good efficacy of the treatment because their child ate a fat-rich diet before starting the ketogenic diet, seems just a coincidence. In practice this means that physicians do not need to take this fact into account in deciding whether to start a treatment with the ketogenic diet or not (independent of tolerability of the diet). The fact that the child likes a diet with a lot of fat does not predict success of the ketogenic diet. Likewise, parents who think their child dislikes fatty foods do not have to worry about their child not responding at all or to a lesser extent to the ketogenic diet (again, independent of tolerability).

To strengthen these conclusions, further research is still required on this subject. Firstly, a validated questionnaire must be designed to objectify a fat-preference more reliably. The most reliable questionnaire would be the one that correlates with in-person, forced-choice paradigm, as in Amari's study.¹² Secondly, this study should be undertaken following a prospective, rather than a retrospective, design thus avoiding the disadvantages mentioned above. One possibility would be to perform a prospective controlled preference test using a in-person forced-choice design, in which samples of foods of equal number of high fat and low fat foods are presented at several meal times to children with epilepsy. Finally, including more participants in a future study would increase the power of the test. By doing so, small effect-sizes will have a better chance of being detected.

Back to Article Outline

Appendix A. Supplementary data 

Back to Article Outline

References 

1. Zupec-Kania BA, Spellman E. An overview of the ketogenic diet for pediatric epilepsy. Nutrition in Clinical Practice. 2008;23:589–596
   • View In Article
2. (RIVM) Carpay JA, Gijsen R, Poos MJJC. How many times does epilepsy occur and how many people die because of it? In: Public Health Discovering the Future, National Pointer Public Health. Bilthoven, The Netherlands: RIVM, http://www.nationaalkompas.nl/gezondheid-en-ziekte/ziekten-en-aandoeningen/zenuwstelsel-en-zintuigen/epilepsie/omvang/, Published online 20.02.2004.
   • View In Article
3. Neal EG, Chaffe H, Schwartz RH, Lawson MS, Edwarts N, Fitzsimmons G, Whitney A, Cross JH. The ketogenic diet for the treatment of childhood epilepsy: a randomised controlled trial. http://neurology.thelancet.com Published online 3.05.2008. DOI: 10.1016/S1474-4422(08)70092-9.
   • View In Article
4. Kossoff EH, Zupec-Kania BA, Amark PE, et al. Optimal clinical management of children receiving the ketogenic diet: recommendations of the Internal Ketogenic Diet Study Group. Epilepsia. 2008;1–14:DOI: 10.1111/j.1528-1167.2008.01765.x
   • View In Article
5. Klepper J, Leiendecker B, Riemann F, Baumeister FA. The ketogenic diet in German-speaking countries: update 2003. Klinikal Padiatrie. 2004;216:277–285
   • View In Article
6. Dutton SB, Escayg A. Genetic influences on ketogenic diet efficacy. Epilepsia. 2008;49:67–69
   • View In Article
   • CrossRef
7. Bergqvist AGC, Schall JI, Richard EL, Gallagher PR, Stallings VA. Predictive power of first morning glucose and the ketogenic diet. Neuropediatrics. 2007;38:193–196
   • View In Article
   • CrossRef
8. Hamdy RF, Turner Z, Pyzik PL, Kossoff EH. Lack of influence of body mass index on the efficacy of the ketogenic diet. Journal of Child Neurology. 2007;22:1167–1171
   • View In Article
   • CrossRef
9. Henderson CB, Filloux FM, Alder SC, Lyon JL, Caplin DA. Efficacy of the ketogenic diet as a treatment option for epilepsy: meta-analysis. Journal of Child Neurology. 2006;21:193–198
   • View In Article
   • MEDLINE
10. Beghi E, Frigeni B, Beghi M, De Compadri P, Garattini L. A review of the costs of managing childhood epilepsy. Pharmacoeconomics. 2005;23:27–45
    • View In Article
    • MEDLINE
    • CrossRef
11. Willner P, Benton D, Brown F, Cheeta S, Davies G, Morgan J, et al. “Depression” increases “craving” for sweet rewards in animal and human models of depression and craving. Psychopharmacology (Berl). 1998;136:272–283
    • View In Article
    • MEDLINE
    • CrossRef
12. Amari A, Dahlquist L, Kossoff EH, Vining EPG, Trescher WH, Slifer KJ. Children with seizures exhibit preferences for foods compatible with the ketogenic diet. Epilepsy & Behavior. 2007;11:98–104
    • View In Article
13. Dutch composition table. The Hague, The Netherlands.: Dutch Nutrition Centre; 2003;ISBN 905151770383
    • View In Article