DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY

ORIGINAL ARTICLE

Favourable response to ketogenic dietary therapies: undiagnosed

glucose 1 transporter deficiency syndrome is only one factor
NATASHA E SCHOELER 1,2

| JUDITH HELEN CROSS 2,3,4 | SUZANNE DRURY 5 | NICHOLAS LENCH 5,6 |

JACINTA M MCMAHON 7 | MARK T MACKAY 8,9 | INGRID E SCHEFFER 8,10 | JOSEMIR W SANDER 1,11,12 |
SANJAY M SISODIYA 1,12
1 NIHR University College London Hospitals Biomedical Research Centre, UCL Institute of Neurology, London; 2 UCL Institute of Child Health, London; 3 Young
Epilepsy, Lingfield; 4 Great Ormond Street Hospital for Children, London; 5 NE Thames Regional Genetics Service, Great Ormond Street Hospital for Children, London;
6 Congenica Ltd, Cambridge, UK. 7 Epilepsy Research Centre, The University of Melbourne, Austin Health, Melbourne, Vic.; 8 Royal Childrens Hospital, Melbourne,
Vic.; 9 Murdoch Childrens Research Institute, Melbourne, Vic.; 10 Departments of Medicine and Paediatrics, The University of Melbourne, Austin Health, Melbourne,
Vic., Australia; 11 Sichting Epilepsie Instellingen Nederland (SEIN), Heemstede, the Netherlands; 12 Epilepsy Society, Chalfont, St. Peter, UK.
Correspondence to Sanjay M Sisodiya at Department of Clinical and Experimental Epilepsy, UCL Institute of Neurology, Queen Square, London WC1N 3BG, UK. E-mail: s.sisodiya@ucl.ac.uk
This article is commented on by Klepper on pages 896897 of this issue.

PUBLICATION DATA

Accepted for publication 16th March 2015.

Published online 23rd April 2015.
ABBREVIATIONS

GLUT1-DS Glucose transporter type 1 deficiency syndrome

KDT
Ketogenic dietary therapies

AIM We aimed to determine whether response to ketogenic dietary therapies (KDT) was due
to undiagnosed glucose transporter type 1 deficiency syndrome (GLUT1-DS).
METHOD Targeted resequencing of the SLC2A1 gene was completed in individuals without
previously known GLUT1-DS who received KDT for their epilepsy. Hospital records were used
to obtain demographic and clinical data. Response to KDT at various follow-up points was
defined as seizure reduction of at least 50%. Seizure freedom achieved at any follow-up point
was also documented. Fishers exact and gene-burden association tests were conducted
using the PLINK/SEQ open-source genetics library.
RESULTS Of the 246 participants, one was shown to have a novel variant in SLC2A1 that was
predicted to be deleterious. This individual was seizure-free on KDT. Rates of seizure freedom
in cases without GLUT1-DS were below 8% at each follow-up point. Two cases without
SLC2A1 mutations were seizure-free at every follow-up point recorded. No significant results
were obtained from Fishers exact or gene-burden association tests.
INTERPRETATION A favourable response to KDT is not solely explained by mutations in
SLC2A1. Other genetic factors should be sought to identify those who are most likely to
benefit from dietary treatment for epilepsy, particularly those who may achieve seizure
freedom.

Ketogenic dietary therapies (KDT) can be an effective

treatment option for some people with drug-resistant epilepsy;1 they are the treatment of choice for glucose transporter type 1 deficiency syndrome (GLUT1-DS).2 The
only known cause of GLUT1-DS is mutation in the
SLC2A1 gene, which encodes a glucose transporter that is
widely expressed, including in endothelial cells of the
bloodbrain barrier.3 The energy deficit due to impaired
glucose transport across the bloodbrain barrier is overcome by the use of ketone bodies (which do not rely on
the glucose transporter system to enter the brain) as a
cerebral energy substrate. Not all clinically diagnosed
cases of GLUT1-DS have a mutation detected in
SLC2A1.4,5
People without mutations in SLC2A1 may achieve at
least 50% seizure reduction when following KDT,
although very few achieve seizure freedom.6 In the absence
of GLUT1-DS or other metabolic disorders, such as pyruvate dehydrogenase complex deficiency, evidence for
2015 Mac Keith Press

factors that predict response to KDT is inconsistent,7 preventing a mechanistic guide to their effective use.
The phenotypic spectrum of GLUT1-DS has expanded
considerably since the first description of the disorder8
and, thus, the diagnosis may be missed. A diagnosis, however, will alter the course of clinical management and suggest the need for lifelong adherence to KDT. As part of an
ongoing study evaluating predictors of response to KDT,
we sequenced SLC2A1 in individuals who followed KDT
for their epilepsy to determine whether response was, in
fact, the result of undiagnosed GLUT1-DS. There may be
SLC2A1 variants that are not disease-causing but may still
affect glucose transport.

METHOD
Participants
The project gained ethical approval through relevant ethics
committees or institutional review boards. Informed consent was obtained from all study participants or from their
DOI: 10.1111/dmcn.12781

969

parents/guardians in the case of minors or adults with

intellectual disability.
From April 2011 to December 2012, participants were
recruited from the following sites in the UK: Great
Ormond Street Hospital for Children, London; National
Hospital for Neurology and Neurosurgery, London;
Evelina London Childrens Hospital, London; St
Georges Hospital, London; Young Epilepsy (including
Matthews Friends clinics for Ketogenic Dietary Therapies), Surrey; Birmingham Childrens Hospital, Birmingham; Addenbrookes Hospital, Cambridge; Alder Hey
Childrens Hospital, Liverpool; Bristol Royal Hospital
for Sick Children, Bristol; and from The Royal Childrens Hospital, Melbourne, and Austin Health, Melbourne, Australia. The number of participants recruited
from each site with KDT response and SLC2A1
sequencing data available is given in Appendix S1
(online supporting information).
Criteria for study inclusion were individuals aged at
least 3 months who were either following KDT or who
had followed KDT in the past for epilepsy. Exclusion criteria were individuals who discontinued KDT before the
3-month point because of lack of tolerability (those who
discontinued KDT before the 3-month point owing to
lack of response or seizure increase were not excluded);
individuals with known GLUT1-DS, pyruvate dehydrogenase complex deficiency, or known to have other metabolic disorders; and individuals with progressive myoclonic
epilepsies.
In the UK clinics, all those who were eligible for
recruitment were invited to participate, bearing in mind
the limited numbers of eligible participants (KDT are only
accessible through specialist centres or clinics). Most Australian participants were recruited prospectively.

Phenotypic data collection

All participants underwent electroclinical phenotyping to
establish seizure type and epilepsy syndrome. This involved
medical history, examination, and review of electroencephalography and imaging studies. Demographic data were
obtained from medical records.
Ketogenic dietary therapies response was defined in
terms of seizure frequency. Seizure frequency was
estimated in 28-day epochs before starting the diet
(baseline), and before 3-month, 6-month, 12-month, 18month, and 24-month follow-up after the start of KDT,
where applicable. Not all patients were treated for
24 months. Clinic letters and seizure diaries, where
already used as part of clinical monitoring, were used to
estimate seizure frequency at each time point. Those with
at least 50% seizure reduction were classified as responders; those with <50% seizure reduction were nonresponders. A case/control study design was adopted:
responders were classified as cases and non-responders as
controls. Response at 3-month follow-up was used as the
primary outcome.

970 Developmental Medicine & Child Neurology 2015, 57: 969976

What this paper adds

Of 246 participants, one (0.4%) had a putatively deleterious variant in
SLC2A1.
Fewer than 8% of participants without glucose transporter type 1 deficiency
syndrome (GLUT1-DS) were seizure-free at each follow-up point.
Single-variant and gene-burden association tests gave no significant results.
Ketogenic dietary therapies (KDT) response is not fully explained by SLC2A1
mutation/common variation.

Genotypic data collection

DNA was extracted from venous blood using Autogen
(AutoGen Inc, Hollister, MA, USA), Fujifilm (Fujifilm
Corporation, Tokyo, Japan), or QiaAmp Blood Maxi kits
(Qiagen GmbH, Hilden, Germany).
Libraries were prepared using the Illumina TruSeq
Custom Amplicon Library Preparation protocol, according to the manufacturers instructions (TruSeq Custom
Amplicon Library Preparation Guide, 2013 Illumina,
Inc., San Diego, CA, USA). Paired-end sequencing reads
were mapped to the February 2009 assembly of the
human genome (Genome Reference Consortium human
genome, build 37) using Burrows-Wheeler Aligner software9 with default settings. Sequence variants were called
using VarScan2 (mpileup2cns)10 with a minimum total
read depth requirement of 30 and default settings. Variants were limited to coding regions and 14 and 6 base
pairs 50 and 30 respectively. Variants were annotated
using Ensembl Variant Effect Predictor Perl Script.
Fishers exact test and the gene-burden association test
were conducted using PLINK/SEQ software version 0.08
(https://atgu.mgh.harvard.edu/plinkseq). The burden test
is the default gene-based test in PLINK/SEQ, which
compares the burden of case-to-control non-reference
alleles by collapsing variants within a gene into a single
value. By default, p values for the burden test are
based on adaptive permutation, giving the empirical significance: the test is stopped early if it appears that the
ultimate empirical p value will not be highly significant
(https://atgu.mgh.harvard.edu/plinkseq/assoc.shtml#perm).
A Bonferroni-corrected significance threshold was set,
based on a=0.05 and four tests (corrected threshold=0.01;
two phenotypes and two association tests).
Power calculations
Power calculations were conducted using PGA Power Calculator.11 An alpha level was set according to the effective
number of tests (two phenotypes and two association tests).
As there is no naturally existing ketogenic diet, disease (or
trait) prevalence is hard to gauge. The disease prevalence
of epilepsy (0.5%) was therefore used, as in other studies.12
The genetic architecture of response is unknown, so power
calculations were performed using co-dominant, dominant,
and recessive penetrance models.
As shown in Figure 1, with a sample size of 245 (as in
the 3mo follow-up association analysis), assuming a codominant penetrance model, variants with a minor allele
frequency of approximately 0.15 and a relative risk of two

Detectable relative risk (RR1)

7
Recessive model
6

4
Dominant model

2
Co-dominant model
1

0.1

0.2
0.3
Marker allele frequency

0.4

0.5

Figure 1: Minimum detectable relative risk and disease allele frequency curves for a cohort of 245 people, with 80% power, assuming a disease prevalence of 0.5%, a=0.01, 127 cases, and control-to-case ratio of 0.9.

253 eligible cases recruited

SLC2A1 sequencing failed in seven individuals

246 cases with SLC2A1 sequencing

data and KDT response data
(at any follow-up point)

245/246a with 3mo KDT follow-up

data
163/246 remained on KDT at the 6mo point.
6mo follow-up data available for 153 cases
111/246 remained on KDT at the 12mo point.
12mo follow-up data available for 106 cases
73/246 remained on KDT at the 18mo point.
18mo follow-up data available for 68 cases
46/246 remained on KDT at the 24mo point.
24mo follow-up data available for 46 cases

Figure 2: Flow chart of inclusion process and number of participants with ketogenic dietary therapies (KDT) response data at each follow-up point.
a
Three-month follow-up data were not available for one patient, for whom data at other time points were available.

or more could be detected with 80% power; assuming a

dominant penetrance model, variants with a minor allele
frequency of approximately 0.15 with a relative risk of
approximately 2.5 or more could be detected with 80%

power; assuming a recessive penetrance model, variants

with a minor allele frequency of approximately 0.35 with a
relative risk of approximately three or more could be
detected with 80% power.
GLUT1 and Ketogenic Diet Response Natasha E Schoeler et al.

971

Table I: Clinical characteristics of cohort (n=246)

Characteristic

Number

Sex
Age at seizure onset, mean (SD)
Age at diet onset, mean (SD)
Cause of epilepsy

n=129 males, n=117 females

2y 1mo (3y 9mo)
7y 11mo (8y 3mo)
Genetic n=28

Structuralmetabolic n=70

Epilepsy syndrome

Number of failed antiepileptic

drugs, mean (SD)
Diet type

Oral/tube fed

Mutation in SCN1A n=14

Mutation in CDKL5 n=3
Mutation in PCDH19 n=1
Mutation in MECP2 n=3
Mutation in STXBP1 n=1
Chromosome translocation n=2
Chromosome q22 deletion n=1
17p13.3 deletion n=1
3p25 deletion n=1
Cardio-facio cutaneous syndrome (unknown which gene) n=1
Hypoxicischemic encephalopathy n=20
Perinatal stroke (without hypoxicischemic encephalopathy) n=1
Haemorrhage n=3
Neonatal hypoglycaemia n=2
Polymicrogyria n=9
Pachygyria n=1
Infection n=6
Periventricular leukomalacia n=1
Cystic encephalomalacia n=1
Occipital lobe abnormality n=1
Temporal lobe abnormality n=3
Hypoplasia n=2 (one with cerebellar hypoplasia and one with
hypoplasia of the corpus callosum)
Cortical dysplasia n=2
Porencephaly n=1
Lesions n=1
Neurofibromatosis type 1 n=1
Epidermal naevus syndrome n=1
Tuberous sclerosis n=5
Hippocampal sclerosis n=2
Lissencephaly n=2
Ohtahara syndrome n=1
Unknown n=4

Unknown cause n=148

Dravet syndrome/severe myoclonic epilepsy of infancy n=15
LennoxGastaut syndrome/spectrum n=11
Childhood absence epilepsy n=2
Juvenile myoclonic epilepsy n=2
Juvenile absence epilepsy n=3
Epilepsy with myoclonic atonic (astatic) seizures (Doose syndrome) n=14
Epilepsy with myoclonic absences n=1
Epilepsy with myoclonic atonic seizures and myoclonic absences n=1
Myoclonic epilepsy (unspecified) n=7
Epilepsy of infancy with migrating focal seizures n=3
Ohtahara syndrome n=1
West syndrome n=15
Undiagnosed n=171
6.61 (6.276.94) (unknown for three cases)
Classical ketogenic diet n=162
Medium-chain triglyceride ketogenic diet n=44
Modified Atkins diet n=39
Unknown n=1
Oral n=167
Tube-fed n=64
Oral and tube-fed n=15

RESULTS
Ketogenic dietary therapies response data were available for
253 individuals. SLC2A1 sequencing failed in seven individuals
owing to low quantity or quality DNA. SLC2A1 sequencing
data and 3-month KDT response data were available for 246
individuals. The inclusion process is illustrated in Figure 2.
972 Developmental Medicine & Child Neurology 2015, 57: 969976

Of the 246 (0.4%) participants, one had an SLC2A1

mutation that was predicted to be deleterious.

Phenotypic data
The clinical characteristics of the cohort are summarized
in Table I.

The following numbers of individuals were responders

at the follow-up points after the start of KDT: 127 out of
245 (52%) at 3 months, of whom nine (4%) were seizurefree; 106 out of 163 (65%) at 6 months, of whom 10 (6%)
were seizure-free; 89 out of 111 (80%) at 12 months, of
whom seven (6%) were seizure-free; 57 out of 73 (78%) at
18 months, of whom three (4%) were seizure-free; 37 out
of 46 (80%) at 24 months, of whom three (7%) were seizure-free.
Three cases were seizure-free at every follow-up point
recorded (all followed the diet, or have currently been following the diet, for approximately 2y) and all were weaned
from all antiepileptic drugs. Four cases were seizure-free
for approximately 12 months or longer but, subsequently,
they either had occasional breakthrough seizures associated
with illness or seizure control deteriorated (although seizure frequency was still dramatically less than before KDT
was commenced).
Details of these individuals response to KDT and their
clinical data are found in Table SI (online supporting
information).

Gene sequence data

A frequency breakdown of the SLC2A1 variants detected in
responders and non-responders at 3-month follow-up is
given in Table II.
One (c.482A>G p.Gln161Arg at chr1: 43396331) of
the three SLC2A1 missense variants was predicted to be
of functional consequence, with a SIFT score of 0 and
PolyPhen score of 0.995. This variant was present in
one individual (genotype T/C), with early-onset epilepsy
(syndrome unspecified) and mild global developmental
delay. The individual remained seizure-free on KDT
for 2 years. The individual was diagnosed with GLUT1DS.
The variant at position chr1:43392783 (c.1408G>C
p.Gly470Arg) was not predicted to be deleterious, with a
SIFT score of 0.4 and a PolyPhen score of 0.103. This
variant was present in one individual who had a marked

increase in seizure frequency when KDT started, so the

diet was stopped immediately. The variant at position
chr1:43424313 (c.10A>G p.Ser4Gly) was not predicted to
be deleterious, with a SIFT score of 0.18 and a PolyPhen
score of 0. This variant was present in one individual who
followed KDT for 7 years, with approximately 50% to
75% seizure reduction maintained throughout this period.
This person was weaned off all antiepileptic drugs
8 months after starting KDT.
None of these three variants were found in the 1000 Genomes Project browser (http://browser.1000genomes.org),
Exome Variant Server (http://evs.gs.washington.edu/EVS),
or ClinVar (http://www.ncbi.nlm.nih.gov/clinvar).
All synonymous and non-coding variants detected in our
cohort with a minor allele frequency <2% were analysed
with Alamut (Interactive Biosoftware, Rouen, France). No
variants were predicted to affect splicing.
No significant results (p<0.01) were obtained from allelic, dominant, or recessive Fishers exact tests: the lowest p
value obtained was 0.15 for chr1:43408966 (empirical p
value under the recessive model). No significant results
were obtained from the gene burden test (number of nonreference alleles in responders=172; number of non-reference alleles in non-responders=130): p=1.

DISCUSSION
We have shown that a favourable response to KDT cannot
be explained solely by mutations in SLC2A1 and this supports findings from a previous study.6 With a sample size
of 245, we had 80% power to detect associations of variants with a minor allele frequency of at least 0.15 and a
relative risk of at least 2, depending on the genetic model
assumed. Variants in SLC2A1 with a smaller effect on
KDT response, or less common variants, may exist, but we
did not have power to detect them, although such effect
sizes are unlikely to be clinically relevant when attempting
to predict response. Patients who become seizure-free on
KDT should still be tested for SLC2A1 mutations; our
results show that there are other, as yet unknown factors

Table II: SLC2A1 variant non-reference allele counts in responders and non-responders to ketogenic dietary therapies at 3 months
Variant location
(build 37/hg19)

SLC2A1 (NM_006516)
variant

Protein
change

chr1:43392754
chr1:43392783
chr1:43392819
chr1:43393384
chr1:43394612
chr1:43394666
chr1:43394887
chr1:43395354
chr1:43395635
chr1:43396331
chr1:43396414
chr1:43408966
chr1:43408984
chr1:43424313

c.1437C>T
c.1408G>C
c.1372C>A
c.1170C>T
c.1065A>G
c.1011C>T
c.966C>T
c.777C>T
c.588G>A
c.482A>G
c.399C>T
c.45C>T
c.27G>A
c.10A>G

p.Pro479=
p.Gly470Arg
p.Arg458=
p.Ile390=
p.Leu355=
p.His337=
p.Val322=
p.Ile259=
p.Pro196=
p.Gln161Arg
p.Cys133=
p.Ala15=
p.Thr9=
p.Ser4Gly

Variant class

Non-reference
allele frequency in
non-responders

Non-reference
allele frequency in
responders

Non-reference
allele frequency in people
seizure-free at 3mo

Synonymous
Missense
Synonymous
Synonymous
Synonymous
Synonymous
Synonymous
Synonymous
Synonymous
Missense
Synonymous
Synonymous
Synonymous
Missense

0
0.004
0
0
0.004
0.013
0.013
0
0.178
0
0.182
0.218
0
0

0.008
0
0.004
0.004
0.02
0.024
0.02
0.004
0.186
0.004
0.184
0.248
0.008
0.004

0
0
0.06
0.06
0.06
0
0
0
0.22
0.06
0.22
0.39
0
0

GLUT1 and Ketogenic Diet Response Natasha E Schoeler et al.

973

that may predict favourable response. Only one participant

in our cohort was shown to have a previously unknown
variant in SLC2A1 that was predicted to be deleterious.
The marked response to KDT in this individual was compatible with a genetic diagnosis of GLUT1-DS, and the
phenotype of the individual was along the widening spectrum of GLUT1-DS,5,8,13 with infantile-onset generalized
epilepsy and mild developmental concerns.
The functional consequences of the other two missense
variants detected, although predicted to be benign, are
unknown. It is unlikely that these variants cause GLUT1DS. Neither of the individuals harbouring these variants
became seizure-free, although one had a very good
response to KDT.
None of the individuals who failed SLC2A1 sequencing
had achieved seizure freedom from KDT. One had been
following KDT for longer than 8 years with more than
90% seizure reduction, two individuals had no change in
seizure frequency, and four had variable seizure frequency
when following KDT. Previous results from Sanger
sequencing and targeted resequencing of SLC2A1 in the
individual with more than 90% seizure reduction, performed in other laboratories, revealed no SLC2A1 mutations.
Synonymous variation can influence protein function
and phenotypic traits in several ways, including through
messenger RNA splicing, stability, or secondary structure.14,15 Variation in the 50 and 30 untranslated region
may also cause loss of function and influence human disease.16 It is thus conceivable that such variation may give
rise to GLUT1-DS with different grades of severity,
depending on the residual activity of GLUT1, and different grades of response to KDT. We did not find a significant difference in the frequency of non-reference alleles in
SLC2A1 variants between responders and non-responders,
independent of variant class. Variation outside the 50 and
30 untranslated region was not sequenced and may also be
relevant for response to KDT.
Individual single nucleotide polymorphisms may have a
greater phenotypic effect in combination with other variants within the same gene or pathway. No significant
results, however, were obtained from the gene burden test.
With the ever-expanding genotypic and phenotypic spectrum of GLUT1-DS,5,8,17 it may be the case that some
variants have a small effect on KDT response such that a
much larger sample size would be needed to detect a population association.
Seizure-freedom rates in people with GLUT1-DS following KDT are highly variable: one study reported that
two out of six (33%) individuals became seizure-free,17
while another reported that nine out of 10 (90%) became
seizure-free.18 Both these seizure-freedom rates are higher
than those commonly reported in those without GLUT1DS following KDT, as reflected in our cohort and noted
in a previous study6. Even people with GLUT1-DS who
become seizure-free on KDT may not, however, necessarily maintain this in the long term. For example, two
974 Developmental Medicine & Child Neurology 2015, 57: 969976

individuals with GLUT1-DS who became seizure-free

both had a partial relapse after a few months, despite good
compliance.17 In another study, 12 out of 15 (80%) people
became seizure-free with the onset of ketosis, but only 10
of these 12 remained seizure-free on KDT as monotherapy
for the whole period of follow-up (range 25y 6mo).19 As
shown in our cohort, a small proportion of people without
SLC2A1 mutations can achieve seizure freedom when following KDT. Type of KDT does not appear to be a contributing factor to seizure-freedom rate in our cohort. The
two cases without SLC2A1 mutations that remained seizure-free on KDT both had epilepsy with myoclonic atonic seizures. Other factors, including variation in other
genes, are likely to contribute to the variability in response.
It is of interest that two of the four individuals who
achieved seizure freedom but subsequently relapsed had
mutations in PCDH19.
A limitation of this study was that response to KDT was
only defined in terms of seizure frequency. KDT can influence other seizure-related factors, such as reduced seizure
severity,20,21 and other parameters, such as improved cognition.22 These benefits may not be related to seizure
reduction20,2325 or decreases in concomitant antiepileptic
drug usage.24 SLC2A1 variation might have been associated
with response to KDT if response had been defined in
terms besides seizure reduction alone. These factors would
have to be consistently studied to redefine KDT efficacy.
Our cohort included patients with known genetic or
structuralmetabolic causes of epilepsy, although the cause
of epilepsy was unknown for most of our cohort. Although
it may be unlikely that people with, for example, Dravet
syndrome or polymicrogyria, also have GLUT1-DS, we do
not know this for a fact. Therefore, we believe there is a
need to establish whether there are unsuspected SLC2A1
disease-causing mutations in KDT responders with whatever type of epilepsy. It is also important to consider that
we are not looking only for disease-causing mutations in
SLC2A1. There may be variants that are not damaging
enough to cause a GLUT1-DS type phenotype, but may
still affect glucose transport. If only people with no known
genetic or structural-metabolic cause of epilepsy and who
had successful sequencing are considered, one out of 148
cases had an unsuspected SLC2A1 mutation. Although it is
important to know the cause in this one patient, it remains
clear that SLC2A1 mutation does not account for either all
responses or all responders.
Lack of detected mutations in SLC2A1 does not necessarily exclude GLUT1-DS in all patients. At each of the
study sites (with the exception of The Royal Childrens
Hospital, Melbourne), it is standard practice to test children with early-onset epileptic encephalopathy for
GLUT1-DS deficiency (particularly if they have a very
favourable response to KDT) by measuring cerebrospinal
fluid glucose or cerebrospinal fluid/blood glucose ratio.
This was not done in all our cases, so the possibility cannot be excluded that some GLUT1-DS cases remained
undiagnosed. There is no known evidence of other genetic

causes of GLUT1-DS, although not all those with cerebrospinal fluid parameters consistent with GLUT1-DS
have SLC2A1 mutations.26
Despite its limitations, this study provides further evidence that a favourable response to KDT is not fully
explained by mutations in SLC2A1. Seizure-freedom rates,
particularly long-term rates, in people without SLC2A1
mutations are low but still present. Our sample size was
larger than a previously reported cohort6 and we confirmed
the findings of that report. This is the first study to investigate genotypicphenotypic correlations between SLC2A1
variation and response to KDT using single-variant and
gene-burden association tests. A larger sample size may be
needed to detect associations with variants of smaller effect
sizes, but such variants are unlikely to be of clinical significance. Other genetic factors should be sought to identify
all those who are most likely to obtain benefit from KDT.
A CK N O W L E D G E M E N T S
This work was undertaken at University College London Hospitals NHS Foundation Trust and University College London, and
a proportion of funding was received from the UK Department of
Healths NIHR Biomedical Research Centres funding scheme.
NES was supported by a UCL Impact Studentship in conjunction
with The Katy Baggott Foundation and Epilepsy Society. JWS
received research support from Epilepsy Society, the Dr Marvin
Weil Epilepsy Research Fund, Eisai, GlaxoSmithKline, World
Health Organization, EU FP7, and the US National Institutes of
Health (NIH), and was consulted by and received fees for lectures
from GSK, Eisai, and UCB Pharma. JHC received funds to the
department for research into the ketogenic diet from Vitaflo.
Honoraria for speaking were also made to the department on her
behalf from Nuntranslated regionicia. JHC and IES have written
a cookery book, Ketocooking, funds from the sale of which are
donated to their respective departments. SMS received research

support from the Epilepsy Society, The Wellcome Trust, The

European Commission, Dravet Syndrome UK, Epilepsy Action,
MRC, NIH, and The Katy Baggott Foundation, and received
research support/fees from lectures from Eisai, GlaxoSmithKline,
and UCB Pharma. The remaining authors have no conflicts of
interest in relation to this work. We thank the patients and their
families for participating in the research. We are grateful to the
following clinicians and staff who assisted with recruitment and
provided phenotypic details: C Eltze, M Sewell, C Fitzachary, and
G Fitzsimmons (Great Ormond Street Hospital, London); R Williams, M-A Leung, T Randall, L Alford, A Tomalin, and E
Hughes (Evelina London Childrens Hospital, London); H Chan,
C Ellerton, C Maritz, and R Lachmann (National Hospital for
Neurology and Neurosurgery, London); B Concannon, S Philip,
and V Hopkins (Birmingham Childrens Hospital, Birmingham);
P Fallon, O Stone, and N Dos Santos (St Georges Hospital,
London); V Aldridge, E Neal, T Stein, and A Desurkar (Matthews Friends Clinics for Ketogenic Dietary Therapies, Surrey);
H Champion, A Parker, and A Maw (Addenbrookes Hospital,
Cambridge); R Kneen and B Evans (Alder Hey Childrens Hospital, Liverpool); M Dunlop and H Edwards (Bristol Royal Hospital
for Sick Children, Bristol); J Bicknell-Royle (The Royal Childrens Hospital, Melbourne, Australia); C Leu and A Tostevin
(UCL Institute of Neurology, London). We also thank J White
(UCL Genetics Institute) for discussions.

SUPPORTING INFORMATION
The following additional material may be found online:
Appendix S1. Number of participants recruited from each
study site, with ketogenic dietary therapy response data and
SLC2A1 sequencing data.
Table SI. Clinical characteristics of individuals who were seizure-free at every follow-up point recorded (n=3), or who were
seizure-free for approximately 12 months with subsequent relapse
(n=4).

REFERENCES
1. Neal EG, Chaffe H, Schwartz RH, et al. The ketogenic
diet for the treatment of childhood epilepsy: a
randomised controlled trial. Lancet Neurol 2008; 7: 500

syndrome in patients treated with ketogenic diet. Epilepsy Behav 2014; 32: 768.

Common genetic variation and susceptibility to partial

7. Schoeler NE, Cross JH, Sander JW, Sisodiya SM. Can

we predict a favourable response to Ketogenic Diet

6.
2. Klepper J, Leiendecker B. GLUT1 deficiency syndrome
2007 update. Dev Med Child Neurol 2007; 49: 70716.

12. Kasperaviciute D, Catarino CB, Heinzen EL, et al.

Therapies for drug-resistant epilepsy? Epilepsy Res 2013;

106: 116.

epilepsies: a genome-wide association study. Brain 2010;

133: 213647.
13. Arsov T, Mullen SA, Rogers S, et al. Glucose transporter 1 deficiency in the idiopathic generalized epilep-

3. Young CD, Lewis AS, Rudolph MC, et al. Modulation

8. Pearson TS, Akman C, Hinton VJ, Engelstad K, De

of glucose transporter 1 (GLUT1) expression levels

Vivo DC. Phenotypic spectrum of glucose transporter

14. Chamary JV, Parmley JL, Hurst LD. Hearing silence:

alters mouse mammary tumor cell growth in vitro and

type 1 deficiency syndrome (Glut1 DS). Curr Neurol

non-neutral evolution at synonymous sites in mammals.

in vivo. PLoS ONE 2011; 6: e23205.

Neurosci Rep 2013; 13: 342.

sies. Ann Neurol 2012; 72: 80715.

Nat Rev Genet 2006; 7: 98108.

4. Hashimoto N, Kagitani-Shimono K, Sakai N, et al.

9. Li H, Durbin R. Fast and accurate short read alignment

15. Hunt R, Sauna Z, Ambudkar S, Gottesman M,

SLC2A1 gene analysis of Japanese patients with glucose

with Burrows-Wheeler transform. Bioinformatics 2009;

Kimchi-Sarfaty C. Silent (synonymous) SNPs: should

transporter 1 deficiency syndrome. J Hum Genet 2011;

25: 175460.

we care about them? Methods Mol Biol 2009; 578: 23

10. Koboldt DC, Chen K, Wylie T, et al. VarScan: vari-

56: 84651.

39.

5. Leen WG, Klepper J, Verbeek MM, et al. Glucose

ant detection in massively parallel sequencing of indi-

16. Chatterjee S, Pal JK. Role of 50 - and 30 -untranslated

transporter-1 deficiency syndrome: the expanding clini-

vidual and pooled samples. Bioinformatics 2009; 25:

regions of mRNAs in human diseases. Biol Cell 2009;

cal and genetic spectrum of a treatable disorder. Brain

22835.

2010; 133: 65570.

6. Ramm-Pettersen A, Nakken KO, Haavardsholm KC,
Selmer

KK.

Occurrence

of

GLUT1

deficiency

101: 25162.

11. Menashe I, Rosenberg PS, Chen BE. PGA: power calcu-

17. Tzadok M, Nissenkorn A, Porper K, et al. The many

lator for case-control genetic association analyses. BMC

faces of glut1 deficiency syndrome. J Child Neurol 2014;

Genet 2008; 9: 36.

29: 34959.

GLUT1 and Ketogenic Diet Response Natasha E Schoeler et al.

975

18. Ramm-Pettersen A, Nakken KO, Skogseid IM, et al.

21. Kossoff EH, Mcgrogan JR, Bluml RM, et al. A modified

Good outcome in patients with early dietary treatment

Atkins diet is effective for the treatment of intractable

of GLUT-1 deficiency syndrome: results from a retro-

pediatric epilepsy. Epilepsia 2006; 47: 4214.

spective Norwegian study. Dev Med Child Neurol 2013;

55: 4407.
19. Klepper J, Scheffer H, Leiendecker B, et al. Seizure control and acceptance of the ketogenic diet in GLUT1 deficiency syndrome: a 2- to 5-year follow-up of 15 children
enrolled prospectively. Neuropediatrics 2005; 36: 3028.
20. Hallbook T, Lundgren J, Ingmar R. Ketogenic diet
improves sleep quality in children with therapy-resistant
epilepsy. Epilepsia 2007; 48: 5965.

22. Farasat S, Kossoff EH, Pillas DJ, et al. The importance

behavior: preliminary report of a prospective study. Dev

Med Child Neurol 2001; 43: 3016.
25. Lambrechts DA, Bovens MJ, De La Parra NM, et al.
Ketogenic diet effects on cognition, mood, and psycho-

of parental expectations of cognitive improvement for

social adjustment in children. Acta Neurol Scand 2013;

their children with epilepsy prior to starting the keto-

127: 1038.

genic diet. Epilepsy Behav 2006; 8: 40610.

23. Nordli DR, Kuroda MM, Carroll J, et al. Experience
with the ketogenic diet in infants. Pediatrics 2001; 108:
12933.
24. Pulsifer MB, Gordon JM, Brandt J, Vining EP, Freeman
JM. Effects of ketogenic diet on development and

976 Developmental Medicine & Child Neurology 2015, 57: 969976

26. Klepper J. Absence of SLC2A1 mutations does not

exclude Glut1 deficiency syndrome. Neuropediatrics 2013;
44: 2356.

Copyright of Developmental Medicine & Child Neurology is the property of Wiley-Blackwell

and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.