Epilepsy Research 64 (2005) 115125

The ketogenic diet influences the levels of excitatory and inhibitory

amino acids in the CSF in children with refractory epilepsy
a
Elfving b , Urban Ungerstedt b , Per Amark

Maria Dahlin a, , Ase

a

Department of Pediatrics, Astrid Lindgren Childrens Hospital, Karolinska Hospital, SE-171 76 Stockholm, Sweden
b Department of Physiology and Pharmacology, Karolinska Institute, 171 77 Stockholm, Sweden
Received 6 July 2004; accepted 11 March 2005

Abstract
The ketogenic diet (KD) is an established treatment for medically refractory pediatric epilepsy. Its anticonvulsant mechanism
is still unclear. We examined the influence of the KD on the CSF levels of excitatory and inhibitory amino acids in 26 children
(mean age 6.1 years) with refractory epilepsy. Seventeen amino acids were determined before and at a mean of 4 months after
the start of the KD. Seizures were quantified. Highly significant changes were found in eight amino acids: increases in GABA,
taurine, serine, and glycine and decreases in asparagine, alanine, tyrosine and phenylalanine. However, aspartate, glutamate,
arginine, threonine, citrulline, leucine, isoleucine and valine/methionine remained unchanged. A significant correlation with
seizure response was found for threonine (P = 0.016). The GABA levels were higher in responders (>50% seizure reduction)
than in nonresponders during the diet (P = 0.041). In the very good responders (>90% seizure reduction), the GABA levels were
significantly higher at baseline as well as during the diet. Age differences were found with significantly larger decreases in
glutamate and increases in GABA in connection with the diet in younger children. Our results indicate that the KD significantly
alters the levels of several CSF amino acids that may be involved in its mechanism of action and the increase in GABA is of
particular interest.
2005 Elsevier B.V. All rights reserved.
Keywords: Amino acids; GABA; Ketogenic diet; Children; Cerebrospinal fluid; CSF

1. Introduction
The ketogenic diet (KD) is a high-fat, low-protein
and very low-carbohydrate diet that has been used in the

Corresponding author. Tel.: +46 8 5177 7024;

fax: +46 8 5177 7457.
E-mail address: maria.dahlin@karolinska.se (M. Dahlin).
0920-1211/$ see front matter 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.eplepsyres.2005.03.008

treatment of pediatric epilepsy since the 1920s (Wilder,

1921). The total caloric needs of the particular child
and the caloric contents of the meals are calculated
by a dietician. Currently, the KD is used primarily to
treat pharmacologically refractory childhood epilepsy
(Freeman et al., 1998; Vining, 1999).
The efficacy of the diet in the treatment of pediatric epilepsy is well established. Studies have shown

116

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

that one-half to two-thirds of the participating children achieve a >50% seizure reduction (Vining, 1999).
Despite its long use and satisfying efficacy, the mechanism of action by which the KD suppress seizures
remains unknown. Several hypotheses have been
proposed. The high-fat and low-carbohydrate intake
induces nonglycolytic pathways with subsequent ketosis. Animal studies have shown that ketosis is necessary for the anticonvulsant effect of the KD (Appleton
and DeVivo, 1974; Nakazawa et al., 1983). However,
further studies have suggested that the ketone betahydroxybuturic acid was not directly involved in the
anticonvulsant effect (Bough et al., 2000), indicating that ketosis may not be a sufficient prerequisite.
Changes in the brain energy reserve, alterations of the
brain acidbase balance and effects of changes in the
lipid profile have also been proposed as modes of action
(Stafstrom, 1999).
Alterations of the amino acid neurotransmitters,
including GABA, have been suggested as a mechanism of action. In a study of adult rats fed a ketogenic
diet, no changes were found in the total brain GABA
levels (Al-Mudallal et al., 1996). In animal studies,
ketones induced increases in GABA and decreases in
aspartate levels (Erecinska et al., 1996; Daikhin and
Yudkoff, 1998), but in mice fed a KD, reduced aspartate but unchanged GABA was found (Yudkoff et al.,
2001). These authors hypothesize that ketosis induces
changes in the brains handling of amino acids with
reduced transamination of glutamate to aspartate and
hence increased GABA. In a pilot study using magnetic
resonance spectroscopy (MRS), two out of three investigated children showed increased GABA levels (Wang
et al., 2003). However, studies in humans evaluating the
levels of amino acid neurotransmitters in relation to the
KD have not been conducted.
The aim of this study was to compare the levels of
excitatory and inhibitory amino acids in the CSF before
and after institution of the KD in a cohort of children
with medically refractory epilepsy.

2. Material and methods

2.1. Ethics
The study was approved by the Ethics Committee
of the Karolinska Hospital and the informed consent

of the parents and, when possible, of the patients, was

obtained.
2.2. Patients
The present study was conducted at the Neuropediatric Department, Karolinska Hospital, during
19982002. It was part of a larger prospective open
trial on the efficacy and safety of the KD in children
with epilepsy. The patients were enrolled consecutively
as they attended the Epilepsy Outpatient Clinic and the
decision to start the KD was made.
The inclusion criteria were an age of 118 years,
pharmacologically refractory epilepsy with a prior trial
of at least three AEDs with appropriate doses and
plasma levels, epilepsy surgery having been considered, no medical contraindications for a trial with the
KD, the family and patient are considered to be able
to cooperate in a trial with the KD, and approval
for performing two lumbar punctures (LPs). Thirtyfive children satisfied the criteria. No patient was lost
to follow-up. However, nine children were excluded
because the second LP could not be performed due
to the fact that the diet was discontinued or the parents did not agree to the LP (seven and two children,
respectively). Thus, 26 children were included in the
final cohort.
Fifteen boys and 11 girls participated in the study.
Their mean age was 6.1 years (3.0 S.D.; median, 5.5;
range, 1.315.8); for demographic data, see Table 1.
The onset of epilepsy occurred at a mean age of 0.8
years ((0.8 S.D.; range, 0.03.0). The type of epilepsy
and the types of seizures as well as epileptic syndromes
were classified according to the International League
Against Epilepsy (ILAE) classification, 1981 and 1989.
Generalized epilepsy was found in the vast majority,
22 children, whereas 4 had partial epilepsy. The most
common seizure types were myoclonic and tonic. At
the time of the study, the mean number of seizure types
was 2.5 (0.8 S.D.). As can be seen from the table, an
epileptic syndrome or etiology could be identified in
22 children. All but one child were mentally retarded
and the majority had other associated neurological disorders, such as autism spectrum disorder and/or motor
handicaps. Prior to entering the study, the patients had
been on long-term therapy with a mean of 7.0 AEDs
(range, 410). At the time of the study, they were on
treatment with a mean of 2.1 AEDs (range, 04).

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

117

Table 1
Demographic data on the study population
Patient
no.

Age
(years)

Sex Epilepsy
onset
(year)

Epilepsy
typea

Seizure
type(s)b

Etiology or epileptic
syndromec

Ton, myocl, Ab
Spasms, GTC
Ton, myocl, GTC
Ton, myocl
Ton, myocl, GTC
Ton, myocl, Ab,
GTC
Ton, myocl,
spasms
Ton, spasms
Myocl, GTC
Ton, myocl, Ab
Ton, myocl
Myocl, GTC
Aton
Myocl, GTC
Spasms, CP, GTC
Myocl, aton, Ab
Myocl, Ab, GTC
GTC
Ton, myocl, CP

Ohtahara syndrome
IS, LGS
Unknown
LGS
HIE, IS
Aicardi syndrome
Inverted duplicated
chrom15, LGS
Aicardi syndrome
Doose syndrome
LGS
Mb down, IS
Postinfectious
Unknown
Cortical dysplasia
LGS
Doose syndrome
IS, LGS
Unknown
Mitochondrial
disorder (complex 1)
Holoprosencephali
HC, corpus callosum
agenesis, IS
LGS
Hereditary
Unknown
Cortical dysplasia
Post-encephalitis

1
2
3
4
5
6

1.3
2.5
2.7
3.5
3.7
3.9

M
F
M
M
F
F

0.0
0.8
0.7
1.7
0.5
0.1

G
G
G
G
G
G

4.0

0.6

8
9
10
11
12
13
14
15
16
17
18
19

4.0
4.5
4.8
5.1
5.1
5.3
5.8
6.0
6.1
6.3
6.7
6.8

F
M
F
M
F
F
F
M
F
M
F
M

0.1
2.5
0.7
0.4
0.4
1.3
1.5
0.2
0.8
0.3
0.3
1.0

G
G
G
G
G
G
G
G
G
G
G
G

20
21

8.2
10.1

M
M

0.1
0.3

G
G

22
23
24
25
26

15.8
8.6
8.7
9.0
9.8

M
M
M
F
M

0.5
0.6
2.5
3.0
0.3

G
P
P
P
P

Ton, myocl, Ab
Myocl, aton,
spasms, CP
Ton, myocl, Ab
Myocl, GTC
Ton, CP
SP
CP, GTC

No.
previous
AEDs

Concomitant
AED(s)d

Seizure
reduction
(%)

7
4
6
5
6
7

VPA, PHT, CZP

VPA, TPM
VPA
LTG, VPA, CZP
VPA, LTG, LOR
OXC, VPA, CZP

>50
>50
>90
>50
>50
>50

LTG, TPM

<50

5
6
8
7
5
7
9
7
9
9
6
6

TPM, CZP
VPA, CZP
CLB
VPA, LTG
CLB
TPM
0
VPA, LTG
TPM, PRM
VPA, CLB
LTG, PB, TPM
VPA, LTG, CZP

>90
100
100
100
<50
>50
100
<50
>50
<50
<50
<50

4
8

LTG, PHT, CZP

VPA, CZP

>50
<50

PHT, OXC, VPA, PB

CZP
PHT, PB, CZP
PRM, VPA
VPA

>50
<50
>50
100
<50

7
8
10
8
10

Epilepsy types: G, generalized; P, partial.

Seizure types: myocl, myoclonic; tonic, tonic; aton, atonic; spasms, atypical infantile spasms; Ab, atypical absences; SP, simple partial; CP,
complex partial; GTC, generalized tonic-clonic. IS, Infantile spasms.
c Etiology: LGS, LennoxGastaut syndrome; HIE, hypoxic-ischemic encephalopathy; HC, hydrocephalus.
d AEDs: VPA, valproic acid; PHT, phenytoin; CZP, clonazepam; TPM, topiramate; LTG, lamotrigine; LOR, lorazepam; OXC, oxcarbazepine;
CLB, clobazam; PRM, primidone; PB, phenobarbital.
b

2.3. Study design

The parents and other caregivers made daily notes of
the number and type(s) of seizures on the seizure calendars used routinely in our clinic. Calculations of these
notes were used to determine the seizure response. The
mean seizure frequency the month before the KD (and
LP1) was compared with the mean frequency the month
before the second admission for LP2. The parental
reports were compared with the observations of seizure

frequency made by the epilepsy nurses during hospitalizations. Children with a <50% seizure reduction were
classified as nonresponders and those with a >50%
seizure reduction as responders. The responders were
further subclassified as good responders if they had a
5090% seizure reduction and as very good responders
if they had a >90% seizure reduction.
To initiate and maintain the KD, we followed a
standardized protocol for the classic KD, which is a
slightly modified version of the protocol of the Johns

118

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

Hopkins Hospital (Swink et al., 1997). Most children

were started at a 4:1 (fat:proteins and carbohydrates)
ratio. A minimum of 1 g/kg body weight per day of protein was used. Supplements of vitamins and minerals,
including selenium, as well as carnitine, were given to
all children.
All patients were admitted to the neuropediatric
ward for a 45 day stay for initiation of the KD.
Before starting the diet, blood samples were drawn in
the morning before breakfast to obtain baseline serum
levels of protein and albumin, and the first LP was performed shortly after breakfast. The second LP (LP2)
and blood sampling were performed according to the
same routine during a 2-day follow-up hospitalization.
The LP2 was done 36 months after starting the KD
provided that the major fine-tuning steps to maximize
its efficacy had been performed and adjusted metabolically. Local anaesthesia was induced by application of
a cream (Emla) and the spinal tap was performed in the
L3L4 interstitium. On each occasion, a total of 4 ml
of CSF was collected and the first milliliter of the sample was sent off for protein analysis. Aliquots of 1.5 ml
were sampled in each of two Eppendorf tubes. These
were immediately cold-centrifuged and the supernatant
was collected and frozen at 70 C. The CSF from the
4th ml obtained was used later for analyses of amino
acids.
2.4. Amino acid analyses
The CSF samples were diluted five time with distilled water (50 l CSF + 200 l distilled water). The
amino acids were analyzed using high performance liquid chromatography (HPLC) (Lindroth and Mopper,
1979) with fluorescence detection (CMA/280, CMA
Microdialysis AB, Stockholm, Sweden). Seventeen
primary amino acids were analyzed by precolumn ophthaldialdehyde (OPA) derivatization in a CMA200
refrigerated Microsampler, (CMA Microdialysis AB,
Stockhom, Sweden) and separated on a Nucleosil C18
reverse phase column (5 m; 60 mm 4 mm, Knauer,
Berlin). The column was eluted with a 0.05 M sodium
acetate buffer (pH 6.72) containing 5% methanol and
2.5% tetrahydrofuran and separation was performed
using a gradient with methanol (Dabrosin et al., 1997).
The amino acid concentrations were measured using
a chromatography data system (EZChrom; Scientific
Software, Inc., San Ramon, CA, USA).

2.5. Statistical evaluation

To determine the statistical significance of differences, the CSF levels of the amino acids before and
during the KD were compared in all children using
either Students t-test for paired samples or a Wilcoxon
matched-pairs test depending on whether or not the differences showed a normal distribution. The amino acid
levels in responders were compared to those of nonresponders before as well as during the diet using Students t-test for unpaired samples or a MannWhitney
U-test. Equivalent comparisons were made between the
very good responders and the nonresponders. The levels of amino acids in relation to age were analyzed by
means of the MannWhitney U-test and the Wilcoxon
matched-pairs test. The level of significance in all statistical tests was set at P < 0.05.

3. Results
3.1. Amino acid levelsgeneral changes
At the time of the CSF sampling while on the KD,
19 children were at a KD ratio of 4:1, six at a ratio of
3:1 and one child had a ratio of 2.5:1.
The levels of protein in the CSF before and during
the diet were compared (n = 22). The mean protein level
before the KD was 0.21 0.1 and 0.22 0.1 S.D. g/l
during the KD and the comparison did not show any
significant changes (P = 0.43).
The mean serum concentration of protein (n = 21)
before and during the KD was 70.3 4.4 and
69.2 3.9 S.D. g/l, respectively, and of albumin
(n = 24), 40.8 3.5 and 40.6 3.8 S.D. g/l. The serum
protein or albumin levels were not significantly
changed in connection with the diet.
The CSF samples obtained at baseline immediately
before starting as well as during the KD were taken
with a mean interval between samples of 4.5 (S.D.
3.2) months. The concentrations of the 17 determined
amino acids from samples obtained before and during
the diet in all 26 children are given in Table 2 with
means and SDs. For each amino acid, the levels at
baseline and during the diet were compared statistically. As can be seen, highly significant changes were
found for asparagine, serine, taurine, GABA, glycine,
alanine, phenylalanine and tyrosine. Decreased levels

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

Table 2
Comparison of the CSF concentrations of amino acids obtained
before and during the ketogenic diet in all children (n = 26)
Amino acid

Before KD

Asparagine
Serine
Taurine
GABA
Glycine
Alanine
Phenylalanine
Tyrosine

11.6
39.7
3.98
3.33
6.9
25.3
8.9
12.6

3.1
10.5
1.16
1.57
1.4
6.7
2.1
4.5

Aspartate
Glutamate
Threonine
Citrulline
Arginine
Leucine
Isoleucine
Valine/methionine

0.39
0.84
31.2
5.6
25.4
12.8
5.2
14.2

0.47
0.97
10.9
2.9
5.8
2.2
0.9
3.1

During KD

P-value

10.4
44.6
5.50
3.71
8.9
20.4
6.9
6.9

2.5
9.4
1.67
1.66
2.8
4.1
1.5
1.4

0.0015
0.0038
0.0001
0.0050
0.0002
0.0003
0.0000
0.0000

0.28
0.63
34.1
5.0
24.9
12.2
5.6
13.3

0.15
0.25
11.0
1.9
5.9
2.0
1.2
3.1

0.68 (ns)
0.17 (ns)
0.08 (ns)
0.11 (ns)
0.61 (ns)
0.19 (ns)
0.15 (ns)
0.08 (ns)

The means S.D.s are presented. All amino acid levels are given in
mol/l.

in tyrosine were seen in all 26 children, in phenylalanine in 25 and in alanine in 22, and increased levels
were seen in taurine in 25 children, in glycine in 23
and in GABA in 21. On the other hand, no significant
changes were seen in the levels of aspartate, glutamate, threonine, citrulline, arginine, leucine, isoleucine
or valine/methionine.
3.2. Amino acid levels and seizure response
The seizure response to treatment with the KD is
given for the individual patient in Table 1. A comparison of the seizure frequency before and during the
diet showed that 17 children (65%) were responders
and nine were nonresponders. Seven of the responders
were very good responders (five being seizure-free) and
10 were good responders. The response groups did not
differ significantly in age.
The CSF levels of amino acids in responders and
in nonresponders before as well as during the KD are
given with means and S.D.s in Table 3. The table also
shows the means of the absolute differences in the
amino acid levels, i.e. the magnitude of the changes
from the baseline level before starting the diet, for the
respective response groups. The amino acid levels were
compared statistically in responders and nonresponders

119

before starting the diet as well as during the diet and also
the absolute differences. Threonine was significantly
different with increases in nonresponders compared to
responders during the KD (P = 0.016), but no differences at baseline and the absolute differences were significantly different (P = 0.036). The GABA levels were
significantly higher in responders than in nonresponders during the diet (P = 0.041), but before starting the
diet the levels were not significantly different although
there was a trend towards significance (P = 0.075). The
absolute differences in GABA levels obtained before
and during the diet were compared in the response
groups and were not significantly different.
In the light of these latter findings, the GABA levels were compared according to the subclassification of
seizure response as is presented in Fig. 1. The means,
S.D.s and ranges of the GABA levels are shown for the
respective response groups. In the very good responders, the mean GABA level before and during KD
was 4.47 and 4.80 mol/l, respectively, in the good
responders, it was 3.24 and 3.77 mol/l, and in the
nonresponders it was 2.55 and 2.79 mol/l. As can be
seen, the GABA levels increased in connection with
the diet in all response groups and higher GABA levels were seen in those with a better seizure response
before as well as during the diet. A comparison between
the GABA levels in very good responders and nonresponders showed a significant difference at baseline (P = 0.023) as well as during the diet (P = 0.023),
but a comparison of the absolute differences showed
no difference between the response groups. Similar
comparisons between very good responders and good
responders as well as between good responders and
nonresponders showed no differences.
Equivalent comparisons of all other amino acid levels in relation to the three response groups were made
and changes were found in asparagine. Fig. 2 shows the
mean, S.D. and range of asparagine levels in relation to
seizure response before and during the diet. The mean
asparagine level before and during KD was 9.5 and
8.7 mol/l, respectively, in the very good responders,
12.3 and 11.2 mol/l in the good responders, and 12.4
and 10.9 mol/l in the nonresponders. The asparagine
levels in very good responders, as compared to nonresponders showed a significant difference (P = 0.038)
before the KD and an increase, although not significant (P = 0.053), during the diet. In the comparison
between very good responders and good responders,

120

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

Table 3
Comparison of the CSF levels of determined amino acids in responders and nonresponders before as well as during the KD with means S.D.s
(n = 26). The absolute differences between levels before and during the diet in responders and nonresponders are also given. Values reaching
statistical significance are shown as bold values
Amino acid

Before KD
Responders

P-value During KD
Nonresponders

Responders

P-value Absolute differences

Nonrespondersa

0.48
0.47
0.19
0.041
0.69
0.19
0.17
0.52
0.016

0.98
5.2
1.6
0.45
1.9
6.0
2.0
5.7
0.5

1.42
4.4
1.3
0.25
2.1
2.7
2.0
5.5
7.5

0.52
0.82
0.49
0.58
0.83
0.18
0.93
0.90
0.036

0.17
0.28
0.13
0.98
0.73
0.77
0.59

0.09
0.09
0.84
0.47
0.33
0.38
0.80

0.14
0.44
0.15
0.68
1.2
0.36
1.1

0.80
0.38
0.37
0.93
0.36
0.97
0.79

Nonresponders

Asparagine
Serine
Taurine
GABA
Glycine
Alanine
Phenylalanine
Tyrosine
Threonine

11.2
40.4
4.2
3.75
6.9
25.6
8.6
12.8
29.9

3.3
12.5
1.3
1.6
1.6
7.5
2.4
5.2
13.0

12.4
38.3
3.6
2.55
6.9
24.6
9.4
11.4
33.6

2.5
5.4
0.8
1.2
1.0
5.3
1.4
4.1
5.2

0.35
0.63
0.25
0.075
0.81
0.72
0.37
0.76
0.43

10.2
45.6
5.8
4.20
8.8
19.6
6.6
7.1
30.5

2.8
10.2
1.7
1.7
3.0
3.8
1.6
1.6
9.7

10.9
42.7
4.9
2.79
9.0
21.9
7.4
7.1
41.1

Aspartate
Glutamate
Citrulline
Arginine
Leucine
Isoleucine
Valine/methionine

0.35
0.66
5.4
25.3
12.4
5.2
13.8

0.4
0.2
3.6
6.9
2.5
1.0
3.8

0.45
1.18
5.9
25.5
13.6
5.3
14.8

0.5
1.6
1.1
2.8
1.4
0.6
1

0.29
0.57
0.68
0.95
0.21
0.66
0.46

0.26
0.58
4.6
24.9
12.1
5.5
13.0

0.2
0.2
2.2
6.4
2.2
1.2
3.5

0.31 0.1
0.74 0.3
5.8 1.10
24.8 5.1
12.4 1.4
5.7 1.1
13.7 2.2

2.1
7.8
1.5
1.3
2.7
4.6
1.1
1.7
10.4

P-value

Respondersa

The amino acid levels are in mol/l.

a Mean.

Fig. 1. The CSF GABA concentrations related to seizure response before (open boxes) as well as during (striped boxes) the ketogenic diet. The
children with a seizure reduction of >90% are referred to in the text as very good responders (n = 7), a reduction of 5090% as good responders
(n = 10) and those with a reduction of <50% as nonresponders (n = 9). The box and whisker plots show the mean, S.D. and complete range of
the GABA concentrations for all 26 children in the study.

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

121

Fig. 2. The CSF asparagine concentrations related to seizure response before (open boxes) as well as during (striped boxes) the ketogenic
diet. The children with a seizure reduction of >90% are referred to in the text as very good responders (n = 7), a reduction of 5090% as good
responders (n = 10) and those with a reduction of <50% as nonresponders (n = 9). The box and whisker plots show the mean, S.D. and complete
range of the asparagine concentrations for all 26 children in the study.

a trend towards significance was seen during the diet

(P = 0.052), but not before it. Comparisons between
good responders and nonresponders showed no differences. Thus, asparagine levels seemed to differ, at least
before starting the diet, between those with a very good
seizure response and those who did not respond.
3.3. Amino acid levels and age
To examine whether age could influence the levels of amino acids, the younger (<5.5 years; n = 13)
and the older (>5.5 years; n = 13) patients were compared concerning all amino acids before as well as
during the KD. Changes were found in aspartate, glutamate and GABA, for which the means and S.D.s are
given in Table 4. As can be seen, significant differences were found between the age groups for aspartate
and glutamate during the diet but not before it and for
GABA both before and during the diet. Also, the absolute differences in the levels before and during the diet
were compared between the age groups and significant
differences were found in glutamate (P = 0.041) and
GABA (P = 0.026) with larger decreases in glutamate
as well as larger increases in GABA in the younger

children than in the older ones although no differences

were found in aspartate.
In the comparison of all 26 children before and during the KD, the levels of aspartate and glutamate did
Table 4
Comparison of the CSF levels of aspartate, glutamate and GABA in
relation to the age of the child before as well as during the KD
Younger

Older

P-value

Aspartate
Before
KD

0.40 0.51a
0.20 0.09

0.37 0.46a
0.35 0.17

ns
0.007

Glutamate
Before
KD

0.66 0.25b
0.52 0.15

1.03 1.36a
0.75 0.28

ns
0.045

GABA
Before
KD

4.00 1.48b
4.58 1.33

2.66 1.40a
2.84 1.53

0.046
0.006

The means S.D.s are given. The younger children were <5.5 years
old (n = 13) and the older >5.5 years old (n = 13). The amino acid
levels are given in mol/l.
a Levels before KD compared to during KD were not significantly
different (P = ns).
b Levels before KD compared to during KD were significantly
different (P < 0.05).

122

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

not show significant changes (Section 3.1), which is

why an equivalent comparison was made also within
the respective age groups. In the younger age group,
glutamate levels before, as compared to during, the
diet were significantly different (P = 0.005), with lower
levels during the diet, but no difference could be found
within the older age group. No significant changes were
observed for aspartate in either age group. A comparison was also made of GABA levels before and during
the diet and showed significant increases during the
diet in the younger age group (P = 0.023), but not in
the older one.
As described above, the GABA levels at baseline
before diet were already higher in younger children
than in older ones and the GABA levels were plotted
against age and a correlation was found (r = 0.53).
However, seizure response was not equally distributed
within the age groups. In the younger age group, 11
children were responders and 2 were nonresponders
and, in the older age group, 6 were responders and 7
were nonresponders.
3.4. Amino acid levels and seizure type
The amino acid levels were also examined in relation to the two most common seizure types, myoclonic
and tonic. The children with myoclonic seizures were
compared to those without such seizures regarding
amino acid levels before as well as during the KD. A
significant difference was noted for glycine in which
larger increases were found in those without myoclonic
seizures compared to those with such seizures during
the diet, but no changes were seen before starting the
diet. Equivalent comparisons regarding tonic seizures
were made but no significant differences were found.
3.5. Amino acid levels and antiepileptic drugs
(AEDs)
The levels of amino acids were analyzed in relation
to the three most frequently used AEDs in this study,
valproate, lamotrigine and clonazepam. No significant
differences were found on comparing the amino acid
levels in children on comedication with, compared to
those without, comedication with one of these drugs. In
children on valproate (n = 14), compared to those not
on valproate (n = 12), the baseline levels of aspartate
were 0.41 and 0.36 mol/l, respectively, and those of

GABA were 2.9 and 3.8 mol/l, and the differences

were not significant.

4. Discussion
Studies of CNS amino acid levels in humans can be
made by analyzing material from brain biopsies, microdialyses and CSF or by MRS. The method available for
us was to obtain CSF samples by lumbar puncture. To
be able to draw conclusions, this would assume the
existence of a relationship between the amino acid levels in the CSF and in the brain. GABA levels in cisternal
CSF and whole-brain tissue were correlated in an animal study (Bohlen et al., 1979). In a human study on
GABA levels comparing suboccipital and lumbar sampling of the CSF, a rostro-caudal concentration gradient
of about 1%/ml CSF was found and the conclusion was
that a valid comparison of CSF amino acid levels is
restricted to similar fractions in the sampling procedure
(Grove et al., 1982). Since equivalent CSF fractions
were analyzed in the present study, we believe that, in
this respect, our amino acid levels should be reliable.
AEDs may induce changes in CSF amino acid levels. This has been reported regarding GABA during
vigabatrin treatment in humans (Ben-Menachem et al.,
1988), but no child in our study was on vigabatrin.
In epilepsy patients, valproate has been reported to
decrease aspartate and to increase GABA CSF levels
(Araki et al., 1988), but no such changes were found in
our cohort.
In the present study, a >50% seizure reduction was
seen in 65% of the children completing the study. If
the initally starting but later excluded children were
counted as nonresponders (i.e. on an intention-to-treat
basis), 49% of the children would have had such a
seizure reduction. These figures are in agreement with
those of earlier studies in which a >50% seizure reduction was found in 54% and 5355% of the children at 3
and 6 months, respectively, on KD in prospective studies (Freeman et al., 1998; Vining et al., 1998) and in
67% in a retrospective study (Hassan et al., 1999).
In our study, eight brain amino acids were significantly altered in connection with the KD. GABA, the
predominant inhibitory neurotransmitter in the CNS,
increased significantly. There is only one report in
humans on GABA levels during the KD in which
an increased GABA level was found, using MR

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

spectroscopy, in two out of a total of three investigated children (Wang et al., 2003). Animal studies on
adult rats fed the KD did not find significant changes in
brain levels of GABA, aspartate or glutamate (Appleton
and DeVivo, 1974), or in total brain GABA concentrations (Al-Mudallal et al., 1996). However, an in vitro
study on rat synaptosomes stimulated with ketones by
Erecinska et al. (1996) showed an increase in GABA
and a decrease in aspartate. In further studies by the
same group, the results were verified in mice injected
with ketones but, in mice fed the KD, aspartate was
changed but not GABA (Yudkoff et al., 2001) and
the authors hypothesize that ketosis induces changes
in brain amino acid handling with a decrease in the
rate of the transamination of glutamate to aspartate and
enhanced conversion of glutamate to GABA. Regarding glutamate, we found a significant decrease in the
subgroup of younger children (<5.5 years) and no
changes were found in aspartate. However, asparagine,
a precursor to aspartate, was decreased. Differences in
species, age, duration of time on the diet and methodology between our study and that of Yudkoff (2001)
may explain the differences in results.
On comparing the amino acid levels with the seizure
response, we found significant changes in threonine,
but the clinical significance of this finding as regards
seizure propensity still seems unclear. The responders
had significantly higher GABA levels than nonresponders during the diet and there was a trend towards
significance also at baseline, but the absolute differences were not significantly different. When the very
good responders were compared to nonresponders significantly higher GABA levels were found at baseline
as well as during the KD, but the absolute differences
were not significantly different. Similar findings were
seen regarding asparagine, but with lower levels in very
good responders. Thus, the magnitude of the changes
in these amino acid levels in connection with the diet
between response groups was not different and may not
directly explain the efficacy. However, one might speculate on the possibility that the mechanism for seizure
efficacy of the KD could involve a threshold level of
GABA or that, at a certain level, GABA influences other
factors that mediate the effect.
Our findings of significantly larger decreases in glutamate and increases in GABA in connection with
the diet in younger children may imply a more efficient metabolic handling of the GABA shunt at this

123

age when forced to utilize nonglycolytic substrates as

ketone bodies, which may result in a lower tendency
to hyperexcitability in these children. One may speculate that the modes of action of the KD might be
age-dependent, a possible explanation of the observation in our study of a better seizure response in the
younger children. Some human studies have observed
a better efficacy of the KD in younger children than
older ones or adults (Huttenlocher, 1976; Schwartz et
al., 1989). However, in the large prospective study by
Freeman et al. (1998), no significant difference in outcome in relation to age was found, although children
>8 years old were slightly less likely to achieve seizure
improvement.
We found GABA levels at baseline before starting the diet to be related to age, with higher levels in
the younger children than in the older ones. GABA
levels have been reported to involve an age dependency; however, a correlation has been found between
increasing GABA and increasing age within the age
span investigated in our study (Goldsmith et al., 1987;
Schmiegelow et al., 1990). Thus, if this holds true also
for our patient cohort, one would have expected lower
GABA levels in the younger age group, which is contrary to our findings. The unexpectedly high baseline
GABA levels in younger children in our study may
be explained by the findings that seizure response was
not equally distributed between the age groups as there
were more responders in the younger age group and
that significantly higher baseline GABA levels were
found in the very good responders. Our patient group
is too small to compare statistically the GABA levels according to seizure response in the younger age
group or to compare the number of responders and
nonresponders according to age group. Therefore, safe
conclusions cannot be drawn as to whether the high
GABA levels observed at baseline in those with a
very good response are due to an inherent GABA
metabolism that predisposes to a good seizure response
or to an age dependency or unknown factors. However, it may be that in the children in whom the best
seizure efficacy occurred, the metabolic handling of
the GABA shunt was different even before starting
the diet. These findings may imply the possibility of
using CSF analyses of GABA and asparagine as predictors of good efficacy of the diet, but further studies
with larger study populations are needed to elucidate
this.

124

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

Prominent decreases in phenylalanine and tyrosine

were found in our study. Appleton and DeVivo (1974)
found significant decreases in the serum levels of these
amino acids in their animal model of the KD. These
amino acids are precursors of norepinephrine (NE), an
inhibitory neurotransmitter in the CNS. Recently, in a
study of knock-out mice lacking NE and fed a ketogenic
diet, NE was found to be required for the anticonvulsant effect of the KD (Szot et al., 2001). The levels
of NE in our cohort are not known but one may speculate that a reduction in precursors might predispose
to a larger pool of NE, which may promote increased
inhibition.
The significant elevations of taurine and glycine
observed in our study have not been described before
in connection with the KD in humans. Taurine is an
inhibitory neurotransmitter and it has been demonstrated in an animal study that induced epileptiform
discharges were suppressed by taurine and, after an
initial proconvulsive effect, also by glycine (Kirchner
et al., 2003).
We found serine levels to be significantly elevated
after introducing the diet. In the animal study by
Appleton and DeVivo (1974) the high-fat-diet animals
had significantly higher serum serine levels. Deficiencies in D-serine have been reported to be involved in
the seizures observed in patients with the 3-PGDH
enzyme deficiency (de Koning et al., 2003). To date,
the significance of our findings pertaining to serine are
unclear.
In conclusion, we found several CSF amino acids
to be changed after initiation of the KD and of particular interest are the alterations of GABA levels. The
altered amino acids include those that increase inhibition as well as decrease excitation and some have
as yet unclear actions. The increases and decreases
in these brain amino acid levels are a clear indication that the diet could influence excitability in the
CNS. It is probable that, in some children, some of
these changes could be involved in the mechanism
of action of the KD to achieve seizure control. However, it is also probable that the diet exerts its action
through several different mechanisms, among which
lipid and hormonal alterations may be included. Different individuals may respond to different modes of
action, which is why the seizure response is not likely
to be explained exclusively by changes in amino acid
levels.

Acknowledgments
This study was supported financially by the Margaretahemmet Society and the Sunnerdahl Handicap
Fund.

References
Al-Mudallal, A., LaManna, J.C., Lust, W.D., Harik, S.I., 1996. Dietinduced ketosis does not cause cerebral acidosis. Epilepsia 37
(3), 258261.
Appleton, D.B., DeVivo, D.C., 1974. An animal model for the ketogenic diet. Epilepsia 15, 211227.
Araki, K., Harada, M., Ueda, Y., Takino, T., Kuriyama, K.,
1988. Alteration of amino acid content of cerebrospinal fluid
from patients with epilepsy. Acta Neurol. Scand. 78, 473
479.
Ben-Menachem, E., Persson, L.I., Schechter, P.J., Haegele, K.D.,
Hardenberg, J., Dahlgren, L., Mumford, J.P., 1988. Effects of
single doses of vigabatrin on CSF concentrations of GABA,
homocarnosine, homovanillic acid and 5-hydroxyindoleacetic
acid in patients with complex partial epilepsy. Epilepsy Res. 2,
96101.
Bough, K.J., Yao, S.G., Eagles, D.A., 2000. Higher ketogenic ratios
confer protection from seizures without neurotoxicity. Epilepsy
Res. 38 (1), 1525.
Bohlen, P., Huot, S., Palfreyman, M.G., 1979. The relationship
between GABA concentrations in brain and cerebrospinal fluid.
Brain Res. 167, 297305.
Ungerstedt, U., Hammar, M., 1997.
Dabrosin, C., Hallstrom, A.,
Microdialysis of human breast tissue during the menstrual cycle.
Clin. Sci. 92, 493496.
de Koning, T.J., Snell, K., Duran, M., Berger, R., Poll-The, B., Surtees, R., 2003. l-Serine in disease and development. Biochem. J.
371, 653661.
Daikhin, Y., Yudkoff, M., 1998. Ketone bodies and brain glutamate
and GABA metabolism. Dev. Neurosci. 20, 358364.
Erecinska, M., Nelson, D., Daikhin, Y., Yudkoff, M., 1996. Regulation of GABA level in rat brain synaptosomes: fluxes through
enzymes of the GABA shunt and effects of glutamate, calcium,
and ketone bodies. J. Neurochem. 67, 23252334.
Freeman, J.M., Vining, E.P.G., Pillas, D.J., Pyzic, P.L., Casey, J.C.,
Kelly, M.T., 1998. The efficacy of the ketogenic diet-1998: a
prospective evaluation of intervention in 150 children. Pediatrics
102, 13581363.
Goldsmith, R.F., Earl, J.W., Cunningham, A.M., 1987. Determination of -aminobutyric acid and other amino acids in cerebrospinal fluid of pediatric patients by reversed-phase liquid
chromatography. Clin. Chem. 33/10, 17361740.
Grove, J., Schechter, P.J., Hanke, N.F.J., de Smet, Y., Agid, A., Tell,
G., Koch-Weser, J., 1982. Concentration gradients of free and
total -aminobutyric acid and homocarnosine in human CSF:
comparison of suboccipital and lumbar sampling. J. Neurochem.
39, 16181622.

M. Dahlin et al. / Epilepsy Research 64 (2005) 115125

Hassan, A.M., Keene, D.L., Whiting, S.E., Jacob, P.J., Champagne,
J.R., Humphreys, P., 1999. Ketogenic diet in the treatment
of refractory epilepsy in childhood. Pediatr. Neurol. 21 (2),
548552.
Huttenlocher, P.R., 1976. Ketonemia and seizures: metabolic and
anticonvulsant effects of two ketogenic diets in childhood
epilepsy. Pediatr. Res. 10, 536540.
Kirchner, A., Breustedt, J., Rosche, B., Heinemann, U.F., Schmieden,
V., 2003. Effects of taurine and glycine on epileptiform activity
induced by removal of Mg2+ in combined rat entorhinal cortexhippocampal slices. Epilepsia 44 (9), 11451152.
Lindroth, P., Mopper, K., 1979. High performance liquid chromatograhic determination of subpicomole amounts of amino
acids by precolumn fluorescence detervatization with ophthaldialdehyde. Anal. Chem. 51 (11), 16671674.
Nakazawa, M., Kodama, S., Matsuo, T., 1983. Effects of ketogenic
diet on electroconvulsive threshold and brain contents of adenosine nucleotides. Brain Dev. 5, 375380.
Schmiegelow, K., Johnsen, A.H., Ebbesen, F., Mortensen, T.,
Mourier Berg, A., Thorn, I., Skov, L., stergaard, J.R., Srensen,
O., 1990. Gamma-aminobutyric acid concentration in lumbar
cerebrospinal fluid from patients with febrile convulsions and
controls. Acta Paediatr. Scand. 79, 10921098.
Schwartz, R.H., Eaton, J., Bower, B.D., Aynsley-Green, A., 1989.
Ketogenic diets in the treatment of epilepsy: short-term clinical
effects. Dev. Med. Child Neurol. 31, 145151.

125

Stafstrom, C.E., 1999. Animal models of the ketogenic diet: what

have we learned, what can we learn? Epilepsy Res. 37, 241
259.
Swink, T.D., Vining, E.P.G., Freeman, J.M., 1997. The ketogenic
diet. Advances in Pediatrics, vol. 44. Mosby-Year Book, pp.
297329.
Szot, P., Weinshenker, D., Rho, J.M., Storey, T.W., Schwartzkroin,
P.A., 2001. Norepinephrine is required for the anticonvulsant
effect of the ketogenic diet. Dev. Brain Res. 129, 211214.
Vining, E.P.G., Freeman, J.M., Ballaban-Gil, K., Camfield, C.S.,
Camfield, P.R., Holmes, G.L., Shinnar, S., Shuman, R., Trevathan, E., Wheless, J.W., the ketogenic diet multi-center study
group, 1998. A multicenter study of the efficacy of the ketogenic
diet. Arch. Neurol. 55, 14331437.
Vining, E.P.G., 1999. Clinical efficacy of the ketogenic diet. Epilepsy
Res. 37, 181190.
Wang, Z.J., Bergqvist, C., Hunter, J.V., Jin, D., Wang, D., Wehrli, S.,
Zimmerman, R.A., 2003. In vivo measurement of brain metabolites using two-dimensional double-quantum MR spectroscopyexploration of GABA levels in a ketogenic diet. Magn. Reson.
Med. 49, 615619.
Wilder, R.M., 1921. The effect of ketonemia on the course of
epilepsy. Mayo Clin. Bull. 2, 307.
Yudkoff, M., Daikhin, Y., Nissim, I., Lazarow, A., Nissim, I., 2001.
Brain amino acid metabolism and ketosis. J. Neurosci. Res. 66,
272281.