Articles

Classic ketogenic diet versus further antiseizure medicine in

infants with drug-resistant epilepsy (KIWE): a UK,
multicentre, open-label, randomised clinical trial
Natasha E Schoeler, Louise Marston, Laura Lyons, Sally Halsall, Ruchika Jain, Siobhan Titre-Johnson, Maryam Balogun, Simon J R Heales,
Simon Eaton, Michael Orford, Elizabeth Neal, Colin Reilly, Christin Eltze, Elma Stephen, Andrew A Mallick, Finbar O’Callaghan, Shakti Agrawal,
Alasdair Parker, Martin Kirkpatrick, Andreas Brunklaus, Ailsa McLellan, Helen McCullagh, Rajib Samanta, Rachel Kneen, Hui Jeen Tan,
Anita Devlin, Manish Prasad, Rohini Rattihalli, Helen Basu, Archana Desurkar, Ruth Williams, Penny Fallon, Irwin Nazareth,
Nick Freemantle, J Helen Cross, on behalf of the KIWE study group*

Summary
Background Many infancy-onset epilepsies have poor prognosis for seizure control and neurodevelopmental outcome. Lancet Neurol 2023;
Ketogenic diets can improve seizures in children older than 2 years and adults who are unresponsive to antiseizure 22: 1113–24

medicines. We aimed to establish the efficacy of a classic ketogenic diet at reducing seizure frequency compared with See Comment page 1088
further antiseizure medicine in infants with drug-resistant epilepsy. *The KIWE study group is listed
at the end of the paper

Methods In this phase 4, open-label, multicentre, randomised clinical trial, infants aged 1–24 months with drug- Developmental Neurosciences
Research and Teaching
resistant epilepsy (defined as four or more seizures per week and two or more previous antiseizure medications) were Department (N E Schoeler PhD,
recruited from 19 hospitals in the UK. Following a 1-week or 2-week observation period, participants were randomly Prof J H Cross PhD, L Lyons PhD,
assigned using a computer-generated schedule, without stratification, to either a classic ketogenic diet or a further S Halsall PhD, R Jain BSc,
S Titre-Johnson BSc,
antiseizure medication for 8 weeks. Treatment allocation was masked from research nurses involved in patient care,
M Balogun BSc, E Neal PhD,
but not from participants. The primary outcome was the median number of seizures per day, recorded during Prof F O’Callaghan PhD),
weeks 6–8. All analyses were by modified intention to treat, which included all participants with available data. Genetics and Genomic
Participants were followed for up to 12 months. All serious adverse events were recorded. The trial is registered with Medicine (Prof S J R Heales PhD,
M Orford PhD), and Stem Cells
the European Union Drug Regulating Authorities Clinical Trials Database (2013–002195–40). The trial was terminated
and Regenerative Medicine
early before all participants had reached 12 months of follow-up because of slow recruitment and end of funding. Section (S Eaton PhD),
University College London
Findings Between Jan 1, 2015, and Sept 30, 2021, 155 infants were assessed for eligibility, of whom 136 met inclusion Great Ormond Street Institute
of Child Health, London, UK;
criteria and were randomly assigned; 75 (55%) were male and 61 (45%) were female. 78 infants were assigned to a
Dietetics (N E Schoeler),
ketogenic diet and 58 to antiseizure medication, of whom 61 and 47, respectively, had available data and were included and Paediatric Neurosciences
in the modifified intention-to-treat analysis at week 8. The median number of seizures per day during weeks 6–8, (Prof J H Cross, C Eltze MRCPH),
accounting for baseline rate and randomised group, was similar between the ketogenic diet group (5 [IQR 1–16]) and Great Ormond Street Hospital
for Children, London, UK;
antiseizure medication group (3 [IQR 2–11]; IRR 1·33, 95% CI 0·84–2·11). A similar number of infants with at least
Department of Primary Care
one serious adverse event was reported in both groups (40 [51%] of 78 participants in the ketogenic diet group and and Population
26 [45%] of 58 participants in the antiseizure medication group). The most common serious adverse events were Health (Prof L Marston PhD),
seizures in both groups. Three infants died during the trial, all of whom were randomly assigned a ketogenic diet: one Institute of Clinical Trials
and Methodology
child (who also had dystonic cerebral palsy) was found not breathing at home; one child died suddenly and
(Prof N Freemantle PhD),
unexpectedly at home; and one child went into cardiac arrest during routine surgery under anaesthetic. The deaths and PRIMENT Clinical Trials
were judged unrelated to treatment by local principal investigators and confirmed by the data safety monitoring Unit (Prof L Marston,
committee. Prof I Nazareth PhD), University
College London, London, UK;
Research Department, Young
Interpretation In this phase 4 trial, a ketogenic diet did not differ in efficacy and tolerability to a further antiseizure Epilepsy, Lingfield, Surrey, UK
medication, and it appears to be safe to use in infants with drug-resistant epilepsy. A ketogenic diet could be a (C Reilly PhD); Child Neurology
treatment option in infants whose seizures continue despite previously trying two antiseizure medications. Service, Royal Aberdeen
Children’s Hospital, Aberdeen,
UK (E Stephen FRCPH);
Funding National Institute for Health and Care Research. Department of Paediatric
Neurology, Bristol Royal
Copyright © 2023 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. Hospital for Children, Bristol,
UK (A A Mallick PhD);
Department of Neurology,
Introduction is associated with improved developmental outcome,2 Birmingham Children’s
The incidence of epilepsy is greatest in the first 2 years but many epilepsies presenting in infancy are associated Hospital, Birmingham, UK
of life (95% CI 56–88 cases per 100 000 infants per with poor prognosis for seizure control.3 (S Agrawal MRCPCH); Clinical
Medical School, University of
year).1 These young children with epilepsy remain most A ketogenic diet—ie, a high-fat low-carbohydrate diet Cambridge, Cambridge, UK
at risk for continuing seizures and neurodevelopmental designed to mimic the effects of starvation on the (A Parker MRCPCH); School of
compromise in the long term. Early control of seizures body—is a non-pharmacological treatment option for Medicine, University of

www.thelancet.com/neurology Vol 22 December 2023 1113

 Articles

Dundee, Dundee, UK
(Prof M Kirkpatrick PhD); Research in context
Paediatric Neurosciences Unit,
Royal Hospital for Children, Evidence before this study Added value of this study
Glasgow, UK We searched PubMed, Ovid, the Cochrane Database of This randomised controlled trial is the first to evaluate the
(Prof A Brunklaus, PhD); Systematic Reviews, Cochrane CENTRAL, and the National effectiveness of a ketogenic diet compared with further
Department of Paediatric
Neurosciences, Royal Hospital
Institutes of Health clinical trial registry from database antiseizure medicine in infants (aged 1–24 months) with
for Sick Children, Edinburgh, inception to Oct 16, 2019, with the terms “infant(s)” OR drug-resistant epilepsy. This study was done in infants with
UK (A McLellan FRCPCH); “child(ren)”, AND “ketogenic” OR “medium chain triglyceride”, epilepsy who were having four or more seizures per week and
Department of Paediatric AND “epilepsy” OR “spasm(s)” OR “seizure(s)”. Of 33 studies who had not responded to at least two previous antiseizure
Neurology, Leeds Children’s
Hospital, Leeds, UK
identified, two were randomised controlled trials (one medicines. The results provide a valuable evidence base for the
(H McCullagh MRCPCH); compared a classic ketogenic diet with adrenocorticotropic treatment of drug-resistant epilepsy in infants in whom few
Department of Paediatric hormone in infants with infantile spasms, and one assessed a trials have been undertaken previously.
Neurology, University Hospital classic ketogenic diet with a modified Atkins diet in children,
of Leicester, Leicester, UK Implications of all the available evidence
including 37 infants); the remainder were uncontrolled cohort
(R Samanta FRCPCH); Our trial results showed that a ketogenic diet was not more
Department of Neurology, studies. All studies were categorised as low quality. In meta-
efficacious than a further antiseizure medicine but that the diet
Alder Hey Children’s Hospital, analyses of uncontrolled studies, about 59% (95% CI 53–65) of
Liverpool, UK was safe to use in infants aged 1–24 months. A ketogenic diet
infants achieved 50% or more seizure reduction when following
(R Kneen FRCPCH); Department could be considered a treatment option for infants who
of Paediatric Neurology, Royal
a ketogenic diet, and about 33% (26–43) achieved seizure
continue to have seizures despite having tried two antiseizure
Manchester Children’s freedom. Randomised controlled trials in older children (aged
medications. These results support data from previous low-
Hospital, Manchester, UK >2 years) and adults have only compared a ketogenic diet with
(H J Tan MRCP); Department of powered randomised controlled trials in infants with newly
care as usual. Studies in older children and adults have not
Paediatric Neurology, diagnosed infantile epilepsy and observational studies in
Great North Children’s compared a ketogenic diet with further antiseizure medication.
infants with drug-resistant epilepsy.
Hospital, Newcastle, UK An adequately powered randomised controlled trial is needed
(A Devlin MRCPCH); to assess a ketogenic diet versus standard pharmacological
Department of Paediatric
treatment in infants with epilepsy with various seizure types
Neurology, Queens Medical
Centre, Nottingham, UK who have not responded to first-line treatment.
(M Prasad MRCPCH);
Department of Paediatric
Neurology, Oxford University individuals with drug-resistant epilepsy. There are We designed a phase 4 randomised trial in infants
Hospitals, Oxford, UK
several versions of the diet, but the one that is most aged 1–24 months with drug-resistant epilepsy (defined
(R Rattihalli MRCP); Department
of Paediatric Neurology, Royal commonly used in infants is based on a ratio between as four or more seizures per week and at least two
Preston Hospital, Preston, UK 2:1 to 4:1 of fat (g) and protein and carbohydrate (g), previous antiseizure medicines). We aimed to assess
(H Basu MRCPCH); Neurology respectively.4 A Cochrane review5 of ketogenic diets for the efficacy of a classic ketogenic diet on the number of
Department, Sheffield
Children’s Hospital, Sheffield,
epilepsy incorporated data from four randomised seizures per day, compared with further antiseizure
UK (A Desurkar MRCPCH); controlled trials (RCTs) comparing ketogenic diets with medication.
Children’s Neurosciences usual care in children, including the first RCT in
Centre, Evelina London children aged 2–16 years.6 Children who received a Methods
Children’s Hospital, London,
UK (R Williams FRCPCH);
ketogenic diet were more likely to have 50% or more Study design
Department of Paediatric seizure reduction (risk ratio [RR] 5·80, 95% CI We did a phase 4, open-label, multicentre, randomised
Neurology, St George’s 3·48–9·65; p<0·001) and seizure freedom (RR 3·16, clinical trial at 19 hospitals in the UK (all sites and
Hospital, London, UK 1·20–8·35; p=0·02). However, the magnitude of these principal investigators are listed in the appendix, p 129).
(P Fallon MRCPCH)
effects might not be considered clinically plausible. The The Research Ethics Committee provided full ethics
Correspondence to:
Cochrane review highlighted that evidence for the use approval (14/LO/1230) before the trial start. The
Prof J Helen Cross, University
College London Great Ormond of ketogenic diets in infants with epilepsy is scarce.5 Medicines and Healthcare products Regulatory Agency
Street Institute of Child Health, The efficacy of ketogenic diets cannot be accounted for provided approval with annual review because
London WC1N 1EH, UK solely by the accumulation of brain ketones in the body, medicinal compounds without marketing authori­
[email protected]
and various mechanisms of action have been proposed.7 sations in the target population were used as the
See Online for appendix Medium chain fatty acids, particularly decanoic acid, comparator. The protocol has been published (appendix
might enhance neuronal mitochondrial function by pp 1–51).11
stimulating mitochondrial proliferation.8 Decanoic acid
has also been shown to have an antiseizure effect.9 In Participants
younger children (aged <2 years), there is evidence that a Participants were infants (aged 1–24 months) with a
switch to fatty acid oxidation occurs more readily than in confirmed diagnosis of epilepsy and with four or more
older children (aged ≥2 years).10 It should be investigated seizures per week at baseline, who did not respond to
whether decanoic acid enhances the action of ketogenic two or more pharmacological treatments (antiseizure
diets in infancy, and the biochemical basis for effective­ medication or corticosteroids). Exclusion criteria inclu­
ness should be identified. ded diagnosis of a metabolic disease contraindicating use

1114 www.thelancet.com/neurology Vol 22 December 2023

 Articles

of ketogenic diets, progressive neurological disease, intellectual and developmental disabilities, with scores
severe gastroesophageal reflux, or previous treatment on domains including communication, daily living,
with ketogenic diets. A complete list of eligibility criteria socialisation, and motor skills. Scores ranged from
is provided in the appendix (pp 20–69). 20 to 160 (mean 100 [SD 15]). A low score was classified
Parents or guardians of potential participants were as less than 85. Clinical laboratory assessments were
approached initially by a member of their direct health- done at local laboratories (appendix pp 76–77). Parents
care team. Written informed consent was obtained or guardians returned food diaries required for diet
from each parent or guardian before undergoing calculation at a maximum of 1 week into the observation
baseline assessment, following a face-to-face or tele­ period.
phone consultation with an adequate explanation of the For infants who were assigned a ketogenic diet, the
aims, methods, anticipated benefits, and potential components of the diet were calculated by a paediatric
adverse events of the study. Consent was obtained by dietitian and were specific to the infant, accounting for
the local site principal investigator (paediatric the food diary, daily calorie requirements, adequate
neurologist) or delegate. Sex, as reported in local protein intake for growth, and vitamin and mineral
hospital records, was filled on paper case report forms supple­mentation. All diets were implemented according
by the paediatric neurologist, alongside other study to a classic ketogenic diet protocol, based on a ratio
data. Categories for ethnicity were White, Black, Asian, (usually between 2:1 and 4:1) of fat (g) to carbohydrate
and Other. and protein (g), with non-fasting initiation. Further
adjustments to the ketogenic diet were established
Randomisation and masking through regular growth monitoring, seizure diaries,
Participants were randomly assigned to either a classic and daily home measurement of urine or concentrations
ketogenic diet or a further antiseizure medication using of blood ketones. Parents or guardians of infants
a web-based randomisation service. The randomisation assigned a ketogenic diet underwent a thorough
schedule was computer generated using a simple teaching programme before starting the diet, including
randomisation method with no stratification. Allo­ how to manage potential early side-effects, such as
cations were released via email to centres after the excess ketosis and hypo­glycaemia. Infants who were
research nurse had entered participant information younger than 12 months were admitted for diet
onto the randomisation website. This process concealed initiation. An intervention manual (appendix pp 44–50,
treatment allocation from research nurses involved in 95–99) was provided to sites to ensure consistent
patient care. Success of masking was not measured. ketogenic diet implementation and was discussed with
Although it was not possible to mask participants to local dietitians and the dietetic assistant. All dietitians
treatment allocation, efforts were made to minimise involved in the study were in regular contact with the
expectation bias by emphasising in the patient dietetic assistant, and meetings were organised to
information sheet that evidence supporting ketogenic ensure continued cross-site consistency. Consistency of
diets for seizure control is scarce. Serious adverse ketogenic diet implementation was monitored after the
events were initially assessed by local investigators, but 8-week and 12-month visits by the dietetic assistant.
they were masked from the safety monitoring board for For infants who were assigned a further antiseizure
further review. Treatment procedures started within medication, the clinician who was responsible for
5 days of randomisation. managing the infant’s epilepsy prescribed the most
appropriate drug, which was selected depending on
Procedures presenting seizures, epilepsy syndrome, and previous
Before randomisation, participants were observed drugs used. Paediatric neurologists attended an initial
for 2 weeks (or 1 week if the infant was prone to more meeting to discuss clinical practice, forming the basis of a
than two seizures per day), during which time no consensus protocol to ensure consistent delivery of
changes were made to regular antiseizure medication; antiseizure medication (appendix pp 51, 100). Cross-site
emergency seizure treatments continued as required. consistency of antiseizure medication prescription,
Seizure types or frequency, number of emergency according to the protocol, was monitored by the dietetic
seizure treatments required, and health-care system assistant. A general discussion about infant or toddler
contacts due to exacerbation were recorded by parents nutrition, including details such as promotion of
or guardians in a seizure diary. The paediatric breastfeeding, age-appropriate texture progression for
neurologist or research nurse assessed all infants using weaning, food groups, and the importance of iron-rich
the Infant Toddler Quality of Life Questionnaire foods, was done with families of infants in the antiseizure
(ITQOL-97)12 and Vineland Adaptive Behaviour Scales medication group at the randomisation visit. If the infant
(Vineland-II).13 ITQOL-97 is a questionnaire to measure had local dietetic support, it was ensured that this
physical, mental, and social wellbeing, with scores on a monitoring continued, and a referral was made if required.
scale from 0 (worst health) to 100 (best health). Follow-up visits were arranged at 4 weeks, 8 weeks,
Vineland-II is a scale to support the diagnosis of 6 months, 9 months, and 12 months. Assessments at

www.thelancet.com/neurology Vol 22 December 2023 1115

 Articles

these visits included clinical review, physical exam­ (at 8 weeks, 6 months, and 12 months), completion of
ination, documentation of seizure frequency from the tolerability questionnaire by parents or guardians
seizure diaries, review of adverse events and con­ with research nurses, and completion of ITQOL-97 (at
comitant medication, clinical laboratory assessments 8 weeks and 12 months) and Vineland-II (at 12 months).
Parents or guardians were asked to keep daily seizure
diaries for 8 weeks; thereafter, they were requested to
155 patients assessed for eligibility reduce seizure recording to at least 1–2 days per week,
as clinically indicated, until 28 days before the final
19 excluded
12-month visit, when daily seizure recording recom­
10 did not meet inclusion criteria menced. After the 8-week assessment, according to the
6 declined to participate infant’s clinical response to treatment (seizure outcome
3 unable to complete baseline period
and tolerability), a ketogenic diet or antiseizure
medication was continued or changed. Infants in the
136 enrolled and randomly assigned antiseizure medication group who did not have seizure
control at 8 weeks were offered the chance to switch to a
ketogenic diet outside the context of the trial, depending
on waiting lists for ketogenic diet at the study site.
78 assigned to ketogenic diet 58 assigned to antiseizure treatment
Infants on a ketogenic diet without seizure improvement
at the 8-week assessment continued clinical manage­
5 discontinued treatment ment with antiseizure medication, as per clinician
4 family wanted ketogenic decision.
treatment
1 randomised in error During the COVID-19 pandemic, visits were
conducted remotely by telephone or secure videoc­
onference facility if the parents or guardians did not
78 started ketogenic diet 53 started antiseizure medication
wish to travel or bring the child to hospital (or both), or
if there were concerns about adverse events, as advised
11 discontinued 4 discontinued by the local study team. Remote methods were also
1 non-compliant 1 non-compliant used for issuing and collecting seizure diaries and
4 parent overwhelmed, family issues, 2 patient elected for surgery
or withdrew consent 1 principal investigator or medical
completing questionnaires. Blood tests could be done
6 principal investigator or medical decision locally, and existing laboratory samples used for
decision screening if samples were no older than 6 weeks. Other
protocol amendments during the study included
67 continued ketogenic diet to 8 weeks 49 continued antiseizure medication to 8 weeks reducing the inclusion age from 3 months to 1 month,
giving the option of a minimum of a 1 week baseline
instead of 2 weeks for participants who had a seizure
6 missing or incomplete data for the 2 missing or incomplete data for the frequency of more than two seizures per day, and
primary outcome* primary outcome*
several extensions to the recruitment end date. A
summary of protocol amendments can be found in the
61 included in primary outcome analysis at 47 included in primary outcome analysis at appendix (pp 110–18).
8 weeks* (modified intention to treat) 8 weeks* (modified intention to treat)

Outcomes
24 discontinued treatment 17 discontinued treatment The primary outcome was the number of seizures
5 non-compliant 3 non-compliant recorded during weeks 6–8, accounting for the baseline
3 patient deceased 2 family left country
2 family left country 1 principal investigator or medical
rate and randomised group. Secondary outcomes at
1 principal investigator or medical decision 8 weeks were the number of infants who were seizure-
decision 1 parent overwhelmed, family free during weeks 6–8 of the intervention, responder
2 parent overwhelmed, family issues, or withdrew consent
issues, or withdrew consent 10 study ended before 12-month rate (defined as the number of infants with
11 study ended before 12-month visit >50% improvement from baseline in seizure frequency),
visit
tolerance to a ketogenic diet (as assessed by question­
naire or blood results), and the association between
43 continued ketogenic diet to 12 months† 32 continued antiseizure medication to concentration of medium chain fatty acids and seizure
12 months† control. Secondary outcomes at 12 months were
treatment retention (defined as the number of infants
Figure: Trial profile who remained on a ketogenic diet; intervention group
*Some infants with missing data for the primary outcome analysis had data available for secondary outcomes at
8 weeks. †At 12 months, full data for secondary outcomes were only available for 31 infants (missing or
only), quality of life (ITQOL-97 subscales were also
incomplete data n=12) in the ketogenic diet group and 25 infants (missing or incomplete data n=7) in the analysed separately), and neurodevelopmental outcome
antiseizure medication group. (using Vineland-II, with domains also analysed

1116 www.thelancet.com/neurology Vol 22 December 2023

 Articles

separately). Adverse events were recorded throughout

Antiseizure medication Ketogenic diet group
the trial. Serious adverse events were reported to the group (n=58) (n=78)
study sponsor. The local principal investigator reviewed
Age at randomisation, years 1·10 (0·48) 1·23 (0·54)
the adverse events and decided whether they were
Sex
related to the study intervention. Serious adverse events
Male 36/58 (62%) 39/78 (50%)
were reviewed by the data monitoring committee.
Female 22/58 (38%) 39/78 (50%)
Adverse events were coded according to MedDRA for the
Ethnicity
purpose of reporting.
White 40/55 (73%) 59/75 (79%)

Statistical analysis Other 15/55 (27%) 16/75 (21%)

For the primary outcome, based on data from our previous EEG abnormal 45/51 (88%) 59/68 (87%)

study,6 a mean percentage change of 62% (SD 45) in Epilepsy syndrome diagnosis 29/43 (67%) 45/67 (67%)
seizures from baseline in the ketogenic diet group was Epilepsy syndrome or type
used, assuming a change of 90% (SD 50) from the baseline Early myoclonic encephalopathy 0/38 (0%) 1/46 (2%)
seizure level in the comparison group (100% was no Early infantile epileptic encephalopathy 11/38 (29%) 13/46 (28%)
change in frequency of seizures from baseline) at Migrating focal seizures of infancy 1/38 (3%) 0/46 (0%)
90% power and 5% significance, with a superiority study Infantile epileptic spasms syndrome 19/38 (50%) 23/46 (50%)
design. An inflation factor of 1·35 was used to account for Dravet syndrome 1/38 (3%) 2/46 (4%)
therapist effect (dietitian), assuming nine centres with an Epilepsy with myoclonic atonic seizures 1/38 (3%) 0/46 (0%)
average cluster size of eight and an intraclass correlation (Doose syndrome)

coefficient of 0·05. We also inflated for 10% dropout or Lesional focal epilepsy 5/38 (13%) 7/46 (15%)

other methodological challenges. We calculated a sample Genetic diagnosis 14/54 (26%) 18/68 (26%)
size of 68 participants in the antiseizure medication group Other neurological diagnosis 19/54 (35%) 26/67 (39%)
and 92 participants in the ketogenic diet group Developmental delay (reported by medical 49/55 (89%) 65/73 (89%)
team)
(160 in total). Due to slow recruitment, the sample size
Hemiplegia 3/55 (5%) 8/72 (11%)
was recalculated assuming 25% dropout but keeping all
Seizure type
other parameters the same as in the original sample size.
Focal 22/52 (42%) 30/69 (43%)
With 75 participants in the ketogenic diet group and
Spasms 30/52 (58%) 41/69 (59%)
62 participants in the antiseizure medication group (137 in
total), the primary outcome was powered at 80%. Absence 6/52 (12%) 4/69 (6%)

Type 1 error was two-tailed. Myoclonic 10/52 (19%) 9/70 (13%)

Primary outcome data were analysed using a Poisson Clonic 3/52 (6%) 4/69 (6%)
mixed model, accounting for clustering by centre Tonic 10/52 (19%) 18/69 (26%)
(synonymous with therapist). Random allocation and Tonic clonic 10/52 (19%) 7/69 (10%)
timepoint (baseline or 8 weeks) were entered into the Atonic 4/52 (8%) 3/69 (4%)
model as fixed effects, and centre was entered as a random Seizures per day 9 (3–19) 7 (4–21)
effect. Loge of number of days’ data included in the analysis Systolic blood pressure, mm Hg 94 (13) 98 (16)
from 6 weeks to 8 weeks was included as an offset. Diastolic blood pressure, mm Hg 56 (13) 62 (16)
Secondary outcomes were analysed using random effects Pulse, beats per min 126 (17) 126 (22)
logistic models, with centres as the random effects and Temperature, °C 36·8 (0·3) 36·6 (0·4)
randomised group as a fixed effect. The process outcome Weight, kg 9·9 (2·9) 9·6 (2·7)
relating to tolerability, quality of life, and neurodevelopment Weight, SDS –0·02 (1·76) –0·09 (1·52)
was analysed using random effects linear modelling. Length, m 0·75 (0·11) 0·76 (0·09)
Two stopping rules were established for the data Length, SDS –0·45 (1·90) –0·27 (1·79)
monitoring committee: the O’Brien-Fleming14 type rule Head circumference, cm 44·1 (4·1) 44·1 (3·5)
for efficacy and a power family with an exponent of 1·5 for Head circumference, SDS –1·07 (2·58) 1·08 (2·28)
adverse events. Data were analysed twice for the data (Table 1 continues on next page)
monitoring committee. All analyses were by modified
intention to treat, which included all randomly assigned
participants who had data available for the primary Role of the funding source
endpoint. The funder of the study had no role in study design, data
STATA 17 was used for all analyses except for SAS 9.4, collection, data analysis, data interpretation, or writing of
which was also used for the primary outcome. The the report.
statistical analysis plan is provided in the appendix
(pp 101–09). Results
This trial is registered with ClinicalTrials.gov Between Jan 1, 2015, and Sept 30, 2021, 155 infants were
(NCT02205931) and EudraCT (2013–002195–40). assessed for eligibility, of whom 136 met inclusion

www.thelancet.com/neurology Vol 22 December 2023 1117

 Articles

criteria and were randomly assigned. 78 infants were

Antiseizure medication Ketogenic diet group
assigned to a ketogenic diet and 58 to further antiseizure
group (n=58) (n=78) medication (figure). 75 (55%) infants were male
(Continued from previous page)
and 61 (45%) were female. Baseline clinical and
demographic characteristics, including quality of life
ITQOL-97: child’s current health
(ITQOL-97) and neurodevelopment (Vineland-II), were
Overall health 60 (30–60) 60 (30–60)
similar between the ketogenic diet group and
Physical abilities 25 (10–56) 23 (10–62)
antiseizure medication group (table 1), as were baseline
Parent’s or guardian’s satisfaction with 45 (33–58) 51 (39–70)
child’s overall growth and development
clinical laboratory parameters (appendix pp 119–21).
Pain 58 (42–75) 58 (42–75)
All 78 infants who were assigned to the ketogenic diet
Temperament and mood 56 (45–66) 61 (47–72)
group started the diet. 67 (86%) of 78 infants continued
Overall behaviour 65 (58–79) 69 (56–81)
to 8 weeks, of whom 61 (78%) had primary outcome
data available and were included in the modified
Global behaviour 85 (60–85) 73 (60–100)
intention-to-treat analysis. Some infants with missing
Getting on with others 60 (50–66) 55 (48–70)
data for the primary outcome had available data for
Parent’s or guardian’s perceptions of child’s 36 (25–54) 41 (27–50)
general health secondary outcomes at 8 weeks. Of 58 infants assigned
ITQOL-97: change in child’s health
to further antiseizure medication, 53 (91%) commenced
Much worse than 1 year ago 8/29 (28%) 7/43 (16%)
treat­ment, 49 (84%) continued to 8 weeks, and 47 (81%)
Somewhat worse than 1 year ago 3/29 (10%) 7/43 (16%)
had primary outcome data available.
About the same as 1 year ago 11/29 (38%) 10/43 (23%)
The trial was terminated before all participants could
attain 12 months of follow-up data because of slow
Somewhat better than 1 year ago 3/29 (10%) 11/43 (26%)
recruitment and end of funding. Of 67 infants who
Much better than 1 year ago 4/29 (14%) 8/43 (19%)
were randomly assigned to a ketogenic diet more than
ITQOL-97: effect on parent or guardian
12 months before the study end date (and who therefore
Emotional effect 39 (21–54) 50 (29–64)
had a full 12 months of follow-up), 43 (64%) continued
Effect on their time 52 (33–71) 62 (33–76)
the diet to 12 months and 31 (47%) had available data.
Family cohesion 85 (85–100) 85 (60–100)
Of 48 infants randomised to further antiseizure
Vineland-II: communication skills
medication more than 12 months before the study end
Receptive communication, v-scale score 7 (5–10) 8 (6–10)
date, 32 (67%) continued antiseizure medication to
Expressive communication, v-scale score 6 (4–11) 8 (6–11)
12 months and 25 (52%) had available data.
Overall, v-scale score 12 (9–19) 16 (11–20)
Median follow-up was 11·3 months (IQR 2·7–12·1).
Communication domain, standard score 44 (37–62) 60 (46–66) At 8 weeks, the median number of seizures per day
Vineland-II: daily living skills compared with baseline was similar in the ketogenic
Personal, v-scale score 9 (8–11) 9 (8–12) diet group and antiseizure medication group
Domestic, v-scale score 12 (6–13) 11 (0–12) (5 [IQR 1–16] and 3 [IQR 2–11]; IRR 1·33,
Community, v-scale score 10 (0–10) 10 (0–10) 95% CI 0·84–2·11; table 2). Of 63 infants in the
Overall, v-scale score 29 (10–32) 26 (10–32) ketogenic diet group with available data for the
Daily living domain, standard score 66 (57–72) 68 (61–75) responder rate analysis at 8 weeks, 28 (44%) had more
Vineland-II: socialisation skills than 50% seizure reduction compared with 19 (40%) of
Interpersonal relationships, v-scale score 7 (5–9) 8 (5–10) 47 infants in the antiseizure medication group (OR 1·21,
Play, v-scale score 8 (7–10) 9 (7–10) 95% CI 0·55–2·65). Seven (11%) of 63 infants in the
Coping, v-scale score 0 (0–9) 0 (0–9) ketogenic diet group were seizure-free at 8 weeks
Overall, v-scale score 21 (18–26) 22 (19–26) compared with six (13%) of 48 infants in the antiseizure
Socialisation domain, standard score 59 (53–65) 65 (54–73) medication group (OR 0·88, 0·27–2·80). The tolerability
Vineland-II: motor skills score was similar in both groups at the week 8 outcome
Gross, v-scale score 6 (5–8) 6 (6–8) analysis and at other timepoints (table 3).
Fine, v-scale score 6 (6–8) 7 (6–9) No differences between groups were noted for any
Overall, v-scale score 12 (10–15) 14 (12–19) concept within the ITQOL-97 at 12 months, except for the
Motor skills domain, standard score 50 (49–55) 55 (50–61) child’s temperament and mood (β coefficient –6·09,
Overall Vineland-II score 95% CI –11·63 to –0·54) and the child getting along with
Sum of domain standard scores or adaptive 228 (199–244) 236 (208–276) others (β coefficient –6·79, –12·97 to –0·60), which
behaviour composite favoured the antiseizure medication group (table 2). A
Standardised score 54 (48–58) 56 (50–66) similar proportion of parents or guardians in both groups
Data are n/N (%), mean (SD), or median (IQR). SDS=standard deviation score. ITQOL-97=Infant Toddler Quality of Life perceived their child’s health to be much better than 1 year
questionnaire, 97-item full-length version. Vineland-II=Vineland adaptive behaviour scales, 2nd edn. ago (12 [50%] of 24 in the antiseizure medication group
and 11 [37%] of 30 in the ketogenic diet group) or much
Table 1: Clinical and demographic characteristics at screening or baseline
worse than 1 year ago (none [0%] of 24 and one [3%] of 30).

1118 www.thelancet.com/neurology Vol 22 December 2023

 Articles

At 8 weeks, median scores in the ITQOL-97 were

Antiseizure medicine Ketogenic diet group IRR, OR, or
numerically higher (suggesting better health) in the group (n=58) (n=78) β coefficient (95% CI)
ketogenic diet group than the antiseizure medication
Number of seizures at 8 weeks 3 5 IRR 1·33
group for seven of the 12 concepts (appendix p 128). (primary outcome) (2 to 11) (1 to 16) (0·84 to 2·11)
Child’s pain, child’s global behaviour, effect on parental Seizure free at 8 weeks 6/48 (13%) 7/63 (11%) OR 0·88
time, and family cohesion were equal between the groups, (0·27 to 2·80)
although general perceptions of the child’s health was Treatment responsive at 8 weeks 19/47 (40%) 28/63 (44%) OR 1·21
numerically higher in the antiseizure medication group. (0·55 to 2·65)
At 12 months, a numerically larger proportion of infant’s ITQOL-97: child’s current health at 12 months
parents or guardians in the ketogenic diet group perceived Overall health 30 (30 to 85); n=24 60 (30 to 60); n=28 β coefficient 1·23
their child’s health to be much better than 1 year ago than (–12·70 to 15·17)

in the antiseizure medication group (ten [25%] of Physical abilities 47 (7 to 70); n=24 27 (13 to 58); n=22 β coefficient –0·59
(–14·58 to 13·40)
40 individuals vs three [9%] of 32 individuals). By contrast,
Satisfaction with child’s overall 58 (38 to 78); n=24 45 (38 to 70); n=30 β coefficient –4·14
numerically more parents or guardians in the antiseizure growth and development (–14·22 to 5·94)
medication group perceived their child’s health to be Pain 75 (50 to 83); n=24 67 (33 to 83); n=30 β coefficient –11·14
much worse than 1 year ago than those in the ketogenic (–24·65 to 2·36)
diet group (eight [25%] of 32 vs two [5%] of 40; Temperament and mood 68 (60 to 79); n=23 65 (56 to 71); n=30 β coefficient –6·09
appendix p 128). (–11·63 to –0·54)
No differences between groups were noted in the Overall behaviour 67 (60 to 83); n=14 65 (56 to 77); n=14 β coefficient –7·23
(–15·96 to 1·50)
Vineland-II overall standardised score or in the domain
standard score at 12 months (table 2). The overall daily Global behaviour 85 (60 to 100); n=14 85 (60 to 100); n=14 β coefficient 12·72
(–1·56 to 27·00)
living v-scale score was nominally improved in the
Getting on with others 65 (52 to 72); n=12 58 (50 to 66); n=16 β coefficient –6·79
antiseizure medication group (β coefficient 2·23, 95% CI (–12·97 to –0·60)
–4·22 to –0·25). Perceptions of child’s general 41 (30 to 52); n=24 30 (16 to 52); n=30 β coefficient –6·37
A total of 73 serious adverse events were reported in the health (–14·29 to 1·56)
antiseizure medication group and 161 were reported in the ITQOL-97: effect on parent or guardian at 12 months
ketogenic diet group. The proportion of serious adverse Emotional effect 57 (36 to 79); n=24 54 (39 to 68); n=30 β coefficient –5·00
events classified into each Medical Dictionary for (–15·52 to 5·53)
Regulatory Activities (MedDRA) system organ was similar Effect on their time 62 (33 to 90); n=24 57 (43 to 76); n=30 β coefficient –3·11
in each group (table 4). Three infants died during the trial, (–16·80 to 10·58)

all of whom were randomly assigned to the ketogenic diet Family cohesion 85 (60 to 85); n=24 85 (60 to 100); n=30 β coefficient –1·52
(–9·48 to 6·45)
group; deaths were judged unrelated to the treatment.
Vineland-II: communication skills at 12 months
One child with epilepsy and dystonic cerebral palsy was
Receptive communication, 7 (7 to 8); n=9 7 (5 to 7); n=8 β coefficient 0·09
found not breathing at home; cardiopulmonary v-scale score (–1·22 to 1·39)
resuscitation was attempted without success in the Expressive communication, 5 (3 to 7); n=9 5 (3 to 8); n=14 β coefficient 0·68
emergency department. One child died suddenly and v-scale score (–1·49 to 2·85)
unexpectedly at home, a known risk of complex epilepsy. Overall communication, 10 (8 to 14); n=5 11 (9 to 13); n=5 β coefficient 1·17
The third child was no longer on a ketogenic diet at the v-scale score (–3·42 to 5·77)
time of the event; they became bradycardic and went into Communication domain, 48 (43 to 59); n=5 49 (44 to 55); n=5 β coefficient 2·79
cardiac arrest during routine surgery under anaesthetic. standard score (–8·14 to 13·72)

A higher proportion of infants in the antiseizure Vineland-II: daily living skills at 12 months
medication group had changes to the number or dose of Personal, v-scale score 6 (5 to 7); n=11 5 (4 to 7); n=9 β coefficient –1·53
(–3·38 to 0·32)
concurrent antiseizure medications during the study
Domestic, v-scale score 10 (9 to 11); n=21 11 (9 to 11); n=20 β coefficient 0·01
compared with the ketogenic diet group (24 [50%] of (–0·38 to 0·41)
48 infants vs nine [14%] of 66 infants). These modifications Community, v-scale score 10 (9 to 10); n=17 10 (9 to 10); n=22 β coefficient 0·22
included dose increases of concurrent antiseizure (–0·22 to 0·67)
medications or short courses of new antiseizure Overall, v-scale score 16 (15 to 17); n=10 15 (13 to 16); n=8 β coefficient –2·23
medications due to seizure escalation or prophylaxis for (–4·22 to –0·25)
planned admission, except for one (2%) of 66 infants in Daily living domain, 25 (21 to 34); n=10 25 (21 to 28); n=8 β coefficient –0·69
the ketogenic diet group and two (4%) of 48 infants in the standard score (–7·68 to 6·31)
antiseizure medication group, for whom the dose of a (Table 2 continues on next page)
concurrent antiseizure medication was decreased during
the study. Concomitant (non-antiseizure) medications
were changed in a similar proportion of infants in Mean measurements for laboratory parameters, blood
both groups (25 [53%] of 47 infants in the antiseizure pressure, pulse, and body temperature were mostly similar
medication group and 33 [49%] of 67 infants in ketogenic in both groups at 8 weeks (appendix pp 122–24).
diet group). Differences were noted between groups in concentrations

www.thelancet.com/neurology Vol 22 December 2023 1119

 Articles

Antiseizure medicine Ketogenic diet group IRR, OR, or Antiseizure Ketogenic diet
group (n=58) (n=78) β coefficient (95% CI) medicine group group (n=78)
(n=58)
(Continued from previous page)
At least one serious adverse 26/58 (45%) 40/78 (51%)
Vineland-II: socialisation skills at 12 months
event at any time
Interpersonal relationships, 6 (5 to 9); n=12 6 (3 to 7); n=12 β coefficient –1·30
Number of serious adverse 73 161
v-scale score (–3·17 to 0·57)
events
Play, v-scale score 8 (7 to 9); n=13 8 (7 to 9); n=12 β coefficient 0·77
MedDRA system organ class*
(–1·12 to 2·66)
Cardiac disorders 0/73 (0%) 1/161 (1%)
Coping, v-scale score 9 (8 to 9); n=11 9 (8 to 9); n=16 β coefficient –0·07
(–0·99 to 0·84) Gastrointestinal disorders 7/73 (10%) 8/161 (5%)
Overall, v-scale score 21 (18 to 23); n=5 22 (19 to 24); n=5 β coefficient 1·55 General disorders and 3/73 (4%) 2/161 (1%)
(–4·04 to 7·14) administration site
conditions
Socialisation domain, 56 (47 to 59); n=5 54 (53 to 58); n=5 β coefficient 1·12
standard score (–17·13 to 19·36) General system disorders 1/73 (1%) 0/161 (0%)
Vineland-II: motor skills at 12 months Immune system disorders 1/73 (1%) 0/161 (0%)
Gross, v-scale score 5 (4 to 7); n=18 5 (4 to 6); n=19 β coefficient –0·53 Infections and infestations 11/73 (15%) 64/161 (40%)
(–1·54 to 0·48) Injury, poisoning, and 1/73 (1%) 0/161 (0%)
Fine, v-scale score 5 (4 to 7); n=15 5 (3 to 6); n=18 β coefficient –0·33 procedural complications
(–1·85 to 1·19) Investigations 1/73 (1%) 2/161 (1%)
Overall, v-scale scores 9 (9 to 12); n=14 9 (8 to 10); n=15 β coefficient –0·46 Metabolism and nutrition 1/73 (1%) 9/161 (6%)
(–1·95 to 1·03) disorders
Motor skills domain, 48 (45 to 54); n=14 43 (45 to 50); n=15 β coefficient –1·53 Nervous system disorders 34/73 (47%) 56/161 (35%)
standard score (–5·94 to 2·88)
Respiratory, thoracic, and 10/73 (14%) 23/161 (14%)
Vineland-II: overall scores at 12 months mediastinal disorders
Domain standard score or 165 (160 to 171); n=4 168 (162 to 176); n=2 β coefficient 0·96 Surgical and medical 5/73 (7%) 2/161 (1%)
adaptive behaviour composite (–18·12 to 20·03) procedures
Standardised score 40 (39 to 41); n=4 41 (39 to 43); n=2 β coefficient 0·16 Vascular disorders 0/73 (0%) 1/161 (1%)
(–5·34 to 5·67)
Examples of MedDRA codes included in each system organ class are: cardiac arrest
Data are median (IQR) or n/N (%), unless specified otherwise. The secondary outcomes ITQOL-97 and Vineland-II were (cardiac disorders); vomiting, diarrhoea, and haematemesis (gastrointestinal
analysed at 12 months. The number of infants included in these analyses are small due to missing data or not meeting disorders); pyrexia (general disorders and administration site conditions); chest
the minimum requirement for analysis. ITQOL-97=Infant Toddler Quality of Life questionnaire, 97-item full-length discomfort (general system disorders); allergic dermatitis (immune system
version. Vineland-II=Vineland adaptive behaviour scales, 2nd edn. IRR=incidence rate ratio. OR=odds ratio. disorders); pneumonia, viral bronchitis, and lower or upper respiratory tract
infection (infections and infestations); shunt malfunction (injury, poisoning and
Table 2: Primary and secondary outcomes procedural complications); weight decreased (investigations); dehydration,
hypoglycaemia, and metabolic acidosis (metabolism and nutrition disorders);
seizure, status epilepticus, and increased intracranial pressure (nervous system
Antiseizure medicine Ketogenic diet group disorders); pneumonia aspiration, and abnormal respiration (respiratory, thoracic
group (n=58) (n=78) and mediastinal disorders); gastrostomy and oesophagogastric fundoplasty
(surgical and medical procedures); and oesophageal varices (vascular disorders).
4 weeks 44 (41–44) 40 (36–42)
MedDRA=Medical Dictionary for Regulatory Activities. *MedDRA system class
8 weeks* 41 (39–44) 40 (38–42) totals do not add up to 73 and 161 respectively as some participants had more
6 months 40 (4) 39 (4) than one classification within a single serious adverse event.
9 months 41 (3) 41 (3)
Table 4: Serious adverse events in 12 months by randomised group
12 months 41 (39–43) 40 (36–42)
Data are median (IQR) or mean (SD). A lower score refers to increased symptoms
or increased severity of symptoms (or both). *Prespecified secondary outcome. in storage, so data were available for only 17 samples at
baseline and 15 samples at 8 weeks from infants on a
Table 3: Tolerability questionnaire scores by randomised group
keto­genic diet. There was a wide range of baseline
plasma concentrations of medium chain fatty acids, and
of β-hydroxybutyrate, glucose, bicarbonate, urate, an increase in octanoic acid and decanoic acid concen­
creatinine, free carnitine, urine organic acids, urine-to- trations in samples taken at 8 weeks (n=15; table 5).
creatinine ratio, lipids, and acylcarnitine, but these Dodecanoic acid concentrations were similar in baseline
changes were as would be expected for individuals and post-intervention samples. In view of the small
following a ketogenic diet. No out-of-range laboratory number of samples available for analysis, no attempt
parameters were considered clinically significant in either was made to perform any statistical analysis to establish
group. Mean measurements for laboratory parameters, whether there was an association between seizures and
anthropometry SD scores, blood pressure, pulse, and body fatty acid concentrations.
temperature were similar in both groups at 12 months
(appendix pp 125–27). Discussion
71 plasma samples were sent for medium chain fatty To our knowledge, this study is the first randomised
acid analysis. Stability was compromised in 39 samples trial in infants (aged 1–24 months) with drug-resistant

1120 www.thelancet.com/neurology Vol 22 December 2023

 Articles

during the intervention compared with those randomly

Baseline (n=17) 8 weeks (n=15)
assigned to a ketogenic diet, suggesting that infants on
Octanoic acid (μmol/L) 6·7 (2·4–9·1) 10·1 (4·2–14·8) a ketogenic diet might have been more clinically stable.
Decanoic acid (μmol/L) 4·3 (3·2–6·6) 10·2 (4·2–18·1) We acknowledge that a ketogenic diet might take time
Dodecanoic acid (μmol/L) 11·6 (8·1–21·9) 12·9 (9·2–18·5) to be effective, but this delay could also occur with up-
Data are median (IQR). titrating antiseizure medication. Moreover, as the
protocol stated that baseline antiseizure medications
Table 5: Concentrations of plasma medium chain fatty acids at baseline
and 8 weeks in infants assigned to ketogenic diet should not change during baseline and intervention
periods, any changes to antiseizure medications would
be for emergency rescue or dose adjustment of existing
epilepsy (defined as four or more seizures per week and medications for ongoing seizures. In view of the
at least two previous antiseizure medicines) to assess the random­isation and complexity of the epilepsy in these
efficacy of a classic ketogenic diet versus further children, a similar degree of severity of epilepsy syn­
antiseizure medication. Although it was designed as a drome and need for medication change would have
superiority study, no evidence was found that a ketogenic been expected in both groups.
diet was better than additional antiseizure medication at In this study, we used 8 weeks as the primary outcome
achieving seizure control in infants aged 1–24 months, period rather than the standard 3 months used in other
and the two treatments were similarly tolerated. Although RCTs. In infants, many epilepsy syndromes are
the study was not powered for non-inferiority, a ketogenic characterised by high seizure frequency, and 8 weeks
diet was numerically similar in efficacy and tolerability to was considered the longest tolerable period for
further antiseizure medication and appeared safe to use assessment and justifiable time to leave concurrent
in this age group. A ketogenic diet might be considered antiseizure medications unchanged. In our clinical
as a treatment option with standard antiseizure experience, seizure response to a ketogenic diet or an
medication for infants who continue to have seizures antiseizure medication is generally established within
after having tried two antiseizure medications. 4 weeks in infants (although longer might be needed for
Our responder rate of approximately 40% is consistent some antiseizure medications due to slow titration
with other ketogenic diet studies. RCTs comparing periods), which leaves 2–4 weeks for seizure assessment
a ketogenic diet with usual care in older children (aged following the initial titration period. Previous
2–16 years) report responder rates of 34–50% after antiseizure medication studies in this age group have
3–4 months;5 a meta-analysis conducted by the KIWE used standardised titration periods of 1 day to 4 weeks,
study group of uncontrolled studies of ketogenic diets in with 4-day to 4-week stabilisation periods.17–19
infants with epilepsy estimated a responder rate of 59%.15 Most sections of the ITQOL-97 quality-of-life measure
The seizure freedom rate was at the higher end of the at 8 weeks showed a favourable trend towards the
range reported in previous RCTs of older children (1–10%) ketogenic diet group, and more parents or guardians
yet lower than that reported in uncontrolled studies in perceived their child’s health to be much better at
infants (33%). It should be acknowledged that false- 12 months versus baseline, compared with those in the
negative results can occur if seizures were not seen or antiseizure medication group. The feeling of the
recorded (or both) within the intervention period, but parents or guardians of doing something worthwhile
this would apply to individuals in both groups, as well as when administering a ketogenic diet (despite the
throughout the study (baseline and intervention periods), common perception that dietary treatment is an
as with all clinical trials. One further RCT examined the imposition on parents) and their perception of its
efficacy of a ketogenic diet versus adrenocorticotropic benefit for their child should not be ignored,20
hormone in infantile spasms not previously trialled on independent of its efficacy in terms of seizure reduction.
steroids (a standard treatment for this seizure type). However, the potential placebo effect of any unblinded
Ten (62%) of 16 infants on a ketogenic diet were in treatment should be considered.
electro­clinical remission at 28 days compared with Quality-of-life and neurodevelopmental measures at
11 (69%) of 16 on adrenocorticotropic hormone; relapse 12 months (ITQOL-97 and Vineland-II) should be
rates were similar between groups (40% vs 36%).16 interpreted with caution due to substantial loss to
However, this study was done in infants with new-onset follow-up, which challenges internal validity and
spasms (a single seizure type) and was underpowered statistical precision. Communication and socialisation
with small numbers. This population was different to skills being numerically in favour of the ketogenic diet
that in our study in which infants with different seizure group at 12 months is consistent with the only RCT
types were included and those with spasms had not assessing quality-of-life and cognitive and behavioural
responded to standard steroid and vigabatrin treatment. functioning on ketogenic diet compared with usual care
Despite the similar seizure frequencies in both groups at 4 months and 16 months. A trend was reported
in our study, more infants who were assigned to further towards improved activation, increased productivity,
antiseizure medication required medication changes and decreased anxiety and mood-disturbed behaviour,21

www.thelancet.com/neurology Vol 22 December 2023 1121

 Articles

although no difference was found between quality- required for infants. One design has been proposed in
adjusted life-years when comparing ketogenic diet with which baseline duration is adjusted on the basis of
usual care.22 individual seizure burden, and treatment duration is
Consistent with uncontrolled studies on the use of based on seizure response according to the timing of
ketogenic diets in infants,15 clinical or laboratory seizure occurrence rather than using the number of
parameters were not clinically significantly different seizures over a set time period as the primary outcome.23
between groups, except for those that would be expected Furthermore, we had scant medium chain fatty acid
when following a ketogenic diet. The proportion of data, because the stability of many samples
individuals with results out of normal range differed was compromised due to storage issues, preventing
between groups only for specific clinical or laboratory meaningful analysis of these data. Future studies
parameters at varying timepoints relevant to ketogenic should allow resources to aid with sample logistics to
diet use. Serious adverse events were as expected in prevent such difficulties, particularly when samples are
both groups: most commonly an increase in seizures, being transported from other sites.
followed by infections, which were both thought to be Although we started the study with 12 sites, we invited
unrelated to treatment. all UK ketogenic diet centres to participate and included
Retention rates of approximately 50% at 12 months in 19 sites nationwide, encompassing secondary and
this study are similar to those reported in uncontrolled tertiary centres from south England to north Scotland,
studies of infants on a ketogenic diet (aggregated rate with varying sizes of ketogenic diet service. Including
of 43% at 12 months).15 RCTs of a ketogenic diet versus all centres not only optimised recruitment but also
usual care in children with epilepsy also found similar ensured that the range of sociodemographic and clinical
retention rates between groups (RR 1·08, 95% CI diversity encountered in this population was reflected
0·74–1·57; p=0·71).5 Studies evaluating the efficacy of in our cohort. As expected, infantile epileptic spasms
antiseizure medications in infancy are scarce. Protocols syndrome was the most common presenting epilepsy
that were used in studies that assessed first-line syndrome, as is seen in epidemiological studies,1 but
treatment in people with spasms required a short time there were no obvious differences in characteristics
for intervention and outcome, making longer term between groups. Study visits and assessments were
retention rates irrelevant. Only a few studies have been conducted online during the COVID-19 pandemic when
done of second-line treatments in infants, in which possible, and interpreters were available as required.
efficacy was reviewed over a short period. Therefore, In this first RCT assessing the use of a ketogenic diet
our data are the first to provide information about in infants (aged 1–24 months) with drug-resistant
retention rates in infants. epilepsy, we report that a classic ketogenic diet was not
More infants in the antiseizure medication group more efficacious than antiseizure medication. A keto­
switched to a ketogenic diet compared with infants in genic diet was tolerable and safe to use in this age
the ketogenic diet group who switched to antiseizure group. Ketogenic diets might be a treatment option in
medication at 8 weeks. Although changes to antiseizure infants who continue to have seizures despite having
medications for infants assigned to further antiseizure tried two antiseizure medications. Many parents or
medication were not recommended in the protocol guardians viewed a ketogenic diet as positive, even if it
(dose of concurrent antiseizure medications might be did not stop their child’s seizures. A ketogenic diet
decreased due to drug interactions, especially if a high might also improve some aspects of quality-of-life and
dose was being taken), medications or doses were neurodevelopment, but further trials are needed with
changed more often in the antiseizure medication larger cohorts at 12-month follow-up and thereafter,
group than in the ketogenic diet group. Although perhaps with an alternative study design.
efficacy was not different between groups, parent or KIWE study group
guardian satisfaction or infant stability might have been Suresh Pujar, Aikaterini Vezyroglou, Victor Maduekwe, Zoe Simpson,
higher in the ketogenic diet group than the antiseizure Isobel Hardy, Agnieszka Szmurlo, Sara Viadero-Prieto (Great Ormond
Street Hospital for Children, London, UK); Domenico Serino,
medication group. Catriona Ward, Tracy Cameron (Royal Aberdeen Children’s Hospital,
Recruitment was slower than anticipated. Infants Aberdeen, UK); Shane Roberts, Vanessa Bara, Susan Ovington,
presenting with epilepsy often have a high burden of Georgina Dunning (Bristol Royal Hospital for Children, Bristol, UK);
seizures; physicians, thus, perceived an urgent need to Tara Deshpande, Ruth Howman, Julia Ackrill (Birmingham Children’s
Hospital, Birmingham, UK); Helena Champion, Nicole Mills,
treat or change treatment rather than wait to document Hanna Laming, Jennie Sharp (Cambridge University Hospitals);
the baseline rate of seizures. The study was also Alice Jollands, Susan MacFarlane, Anne MacLeod, Debbie Rice,
introduced to families at a late stage during treatment— Tracy Cameron (Tayside Children’s Hospital, Dundee, UK);
ie, not after two antiseizure medications when further Sameer Zuberi, Barry Milligan, Janette Buttle, Carla Rennie
(Royal Hospital for Children, Glasgow, UK); Linsey Mavor,
treatment options are limited. Therefore, we reduced Cheryl Fisher, Mags Millar (Royal Hospital for Sick Children,
the baseline period to 1 week for infants with frequent Edinburgh, UK); Nicola Balatoni, Anna Marcyniuk, Emily Scriven,
seizures and increased the number of sites to aid Rachel Meskell, Kathryn Lightfoot, Alison Craig (Leeds Children’s
recruitment. An alternative trial design might be Hospital, Leeds, UK); Anisa Ahmed, Ellen Wilford,

1122 www.thelancet.com/neurology Vol 22 December 2023

 Articles

Rachel Fox, Laiwah Tsang (University Hospitals of Leicester); charity representatives were on the trial steering committee. In 2020, a
Emma Gosling, Sophie Chidlow, Rebecca Jennings, Clare McIntyre, mother of a child who participated in the study spoke to parents or
Victoria Horsley (Alder Hey Children’s Hospital, Liverpool, UK); guardians of other participants and provided a document of concerns
Deivasumathy Muthugovindan, Dipak Ram, Victoria Whiteley and suggestions, which was discussed during the trial steering
(Royal Manchester Children’s Hospital, Manchester, UK); Lucy Bellis, committee. It was then communicated to the trial management group
Jennie Taylor, Imogen Clarke, Ruth Ord (Great North Children’s to ensure that patients or families were made aware of all support
Hospital, Newcastle, UK); Helen Fazachorley, Ann Brown, Naomi Cox groups and resources available at that time. This work was funded by
(Queens Medical Centre, Nottingham, UK); Natalie Mattos-Harris, the National Institute for Health and Care Research (NIHR) Efficacy
Paula Sugden, Rachael Strang (Royal Preston Hospital, Lancashire, and Mechanism Evaluation Programme (13/0656). We thank all
UK); Sithara Ramdas, Nicola Howell, Judy Wadsworth, Shannah Hatch, participants and their families, all members of the trial steering
Ruth Fisher (Oxford University Hospitals); Hannah Taylor, committee, and members of the data and safety monitoring board.
Kim Redfern, Rachel Harrison (Sheffield Children’s Hospital); Research at Great Ormond Street Hospital NHS Foundation Trust and
Mary-Anne Leung, Tara Randall (Evelina London Children’s Hospital, UCL Great Ormond Street Institute of Child Health is made possible
London, UK); Elena Stefanova, Youne Ng, Rachel Doody, by the NIHR Great Ormond Street Hospital Biomedical Research
Nicole Dos Santos, Sophie Adams (St George’s Hospital, London, UK). Centre. The views expressed are those of the authors and not
necessarily those of the National Health Service, the NIHR, or the
Contributors
Department of Health, and Social Care UK.
JHC, NF, and IN conceptualised and designed the study. NES was
the trial dietetic assistant and wrote the first draft. EN and SJRH References
contributed to protocol development. The local trial management team 1 Eltze CM, Chong WK, Cox T, et al. A population-based study of
comprised of LL, SH, RJ, ST-J, and MB. SJRH and SE had the idea for newly diagnosed epilepsy in infants. Epilepsia 2013; 54: 437–45.
the medium chain fatty acid component of the study and MO analysed 2 Freitag H, Tuxhorn I. Cognitive function in preschool children
the medium chain fatty acid samples. CE, ES, AAM, FO’C, SA, AP, after epilepsy surgery: rationale for early intervention. Epilepsia
MK, AB, AMcL, HMcC, RS, RK, HJT, AD, MP, RR, HB, AD, RW, and 2005; 46: 561–67.
PF oversaw site recruitment, clinical management of participants, and 3 Chevrie JJ, Aicardi J. Convulsive disorders in the first year of life:
data collection for their respective sites. CR helped with scoring and neurological and mental outcome and mortality. Epilepsia 1978;
19: 67–74.
interpretation of quality of life and neurodevelopmental outcome data.
All members of the KIWE study group contributed to clinical or 4 van der Louw E, van den Hurk D, Neal E, et al. Ketogenic diet
guidelines for infants with refractory epilepsy.
dietetic management of participants and data collection for their
Eur J Paediatr Neurol 2016; 20: 798–809.
respective sites. IN oversaw the study conduct. LM and NF did the
5 Martin-McGill KJ, Bresnahan R, Levy RG, Cooper PN. Ketogenic
statistical analyses and accessed and verified the underlying data.
diets for drug-resistant epilepsy. Cochrane Database Syst Rev 2020;
All authors had full access to all study data and accept responsibility to 6: CD001903.
submit the manuscript for publication.
6 Neal EG, Chaffe H, Schwartz RH, et al. The ketogenic diet for the
Declaration of interests treatment of childhood epilepsy: a randomised controlled trial.
NES was supported for a research post by Vitaflo (International) and Lancet Neurol 2008; 7: 500–06.
received grants from Nutricia Advanced Medical Nutrition, Vitaflo 7 Masino S, Rho J. Mechanisms of Ketogenic Diet Action.
(International), and Matthew’s Friends charity, and honoraria from In: Noebels JL, Avoli M, Rogawski MA, et al, eds. Jasper’s Basic
Nutricia Advanced Medical Nutrition, Vitaflo (International), and Mechanisms of the Epilepsies. [Internet] 4th edn. Bethesda, MD:
Dr Schaer. SE, SJRH, and JHC report receiving grants from Vitaflo National Center for Biotechnology Information (US), 2012.
https://doi.org/10.1093/med/9780199746545.003.0078.
(International) and having a patent nutritional product
8 Hughes SD, Kanabus M, Anderson G, et al. The ketogenic diet
(WO2013186570) and a patent anticonvulsant compound
component decanoic acid increases mitochondrial citrate synthase
(WO2016038379A1) issued. JHC reports receiving honoraria from
and complex I activity in neuronal cells. J Neurochem 2014;
Nutricia and grants from GW Pharmaceuticals, Zogenix, Marinius, 129: 426–33.
and Ovid. SJRH reports receiving consultancy fees and PhD
9 Chang P, Terbach N, Plant N, Chen PE, Walker MC, Williams RS.
studentship funding from Vitaflo (International). CE reports receiving Seizure control by ketogenic diet-associated medium chain fatty
honorarium from GW Pharmaceuticals/JAZZ Pharmaceuticals. acids. Neuropharmacology 2013; 69: 105–14.
HJT reports receiving honoraria from UCB Pharma, Nutricia, and 10 Heales SJ, Thompson GN, Massoud AF, Rahman S, Halliday D,
GW Pharmaceuticals. SA reports receiving honorarium from Nutricia. Leonard JV. Production and disposal of medium-chain fatty acids
AP reports receiving honorarium from Biomarin. EN reports receiving in children with medium-chain acyl-CoA dehydrogenase
honorarium from Vitaflo (International). AD reports receiving deficiency. J Inherit Metab Dis 1994; 17: 74–80.
consultancy fees from Nutricia and honoraria from Nutricia, 11 Titre-Johnson S, Schoeler N, Eltze C, et al. Ketogenic diet in the
GW Pharmaceuticals, and Zogenix. AAM reports receiving honoraria treatment of epilepsy in children under the age of 2 years: study
from LivaNova and Danone. NF reports receiving grants from the protocol for a randomised controlled trial. Trials 2017; 18: 195.
National Institute for Health and Care Research, the Medical Research 12 Landgraf J. The Infant/Toddler Child Health Questionnaire:
Council, Cure Parkinson’s Trust, and the European Union, consultancy conceptual framework, logic content, and preliminary
fees from ALK, Sanofi Aventis, Gedeon Richter, Abbott, Galderma, psychometric results. Boston, MA, USA: Health Act, 1994.
AstraZeneca, Ipsen, Vertex, Thea, Novo Nordisk, Aimmune, and Ipsen, 13 American Psychological Association. APA PsychNet: Vineland
and honorarium from Abbott Singapore. All other authors declare no Adaptive Behavior Scales, Second Edition. https://doi.org/10.1037/
competing interests. t15164-000.
14 O’Brien PC, Fleming TR. A multiple testing procedure for clinical
Data sharing trials. Biometrics 1979; 35: 549–56.
On publication of this manuscript, de-identified participant data, the 15 Lyons L, Schoeler NE, Langan D, Cross JH. Use of ketogenic diet
study protocol, and the statistical analysis plan will be available on therapy in infants with epilepsy: A systematic review and
request from academic institutions on receipt of a credible research meta-analysis. Epilepsia 2020; 61: 1261–81.
proposal, approved by the corresponding author. Documents will be 16 Dressler A, Benninger F, Trimmel-Schwahofer P, et al. Efficacy
available in English only, for a prespecified time (typically 12 months) and tolerability of the ketogenic diet versus high-dose
on a password-protected portal. adrenocorticotropic hormone for infantile spasms: a single-center
parallel-cohort randomized controlled trial. Epilepsia 2019;
Acknowledgments
60: 441–51.
Matthew’s Friends (a charitable organisation providing support and
17 Piña-Garza JE, Espinoza R, Nordli D, et al. Oxcarbazepine
education for families using a ketogenic diet) was involved in protocol
adjunctive therapy in infants and young children with partial
coproduction and reviewing study progress. Parent and epilepsy seizures. Neurology 2005; 65: 1370–75.

www.thelancet.com/neurology Vol 22 December 2023 1123

 Articles

18 Piña-Garza JE, Levisohn P, Gucuyener K, et al. Adjunctive 22 Lambrechts DAJE, de Kinderen RJA, Vles JSH, de Louw AJ,
lamotrigine for partial seizures in patients aged 1 to 24 months. Aldenkamp AP, Majoie HJM. A randomized controlled trial of the
Neurology 2008; 70: 2099–108. ketogenic diet in refractory childhood epilepsy. Acta Neurol Scand
19 Piña-Garza JE, Nordli DR Jr, Rating D, Yang H, 2017; 135: 678.
Schiemann-Delgado J, Duncan B. Adjunctive levetiracetam in 23 Auvin S, French J, Dlugos D, et al. Novel study design to assess
infants and young children with refractory partial-onset seizures. the efficacy and tolerability of antiseizure medications for focal-
Epilepsia 2009; 50: 1141–49. onset seizures in infants and young children: a consensus
20 Schoeler NE, MacDonald L, Champion H, et al. Assessing parents’ document from the regulatory task force and the pediatric
attitudes towards ketogenic dietary therapies. Epilepsy Behav 2014; commission of the International League against Epilepsy (ILAE),
39: 1–5. in collaboration with the Pediatric Epilepsy Research Consortium
21 IJff DM, Postulart D, Lambrechts DAJE, et al. Cognitive and (PERC). Epilepsia Open 2019; 4: 537–43.
behavioral impact of the ketogenic diet in children and
adolescents with refractory epilepsy: a randomized controlled trial.
Epilepsy Behav 2016; 60: 153–57.

1124 www.thelancet.com/neurology Vol 22 December 2023