Vagus nerve stimulation in children with intractable epilepsy: a randomized controlled trial


Drs Sylvia Klinkenberg at Department of Neurology, Maastricht University Medical Center, PB 5800, 6202 AZ Maastricht, the Netherlands. E-mail:


Aim  The aim of this study was to evaluate the effects of vagus nerve stimulation (VNS) in children with intractable epilepsy on seizure frequency and severity and in terms of tolerability and safety.

Method  In this study, the first randomized active controlled trial of its kind in children, 41 children (23 males; 18 females; mean age at implantation 11y 2mo, SD 4y 2mo, range 3y 10mo–17y 8mo) were included. Thirty-five participants had localization-related epilepsy (25 symptomatic; 10 cryptogenic), while six participants had generalized epilepsy (four symptomatic; two idiopathic). During a baseline period of 12 weeks, seizure frequency and severity were recorded using seizure diaries and the adapted Chalfont Seizure Severity Scale (NHS3), after which the participants entered a blinded active controlled phase of 20 weeks. During this phase, half of the participants received high-output VNS (maximally 1.75mA) and the other half received low-output stimulation (0.25mA). Finally, all participants received high-output stimulation for 19 weeks. For both phases, seizure frequency and severity were assessed as during the baseline period. Overall satisfaction and adverse events were assessed by semi-structured interviews.

Results  At the end of the randomized controlled blinded phase, seizure frequency reduction of 50% or more occurred in 16% of the high-output stimulation group and in 21% of the low-output stimulation group (p=1.00). There was no significant difference in the decrease in seizure severity between participants in the stimulation groups. Overall, VNS reduced seizure frequency by 50% or more in 26% of participants at the end of the add-on phase The overall seizure severity also improved (p<0.001).

Interpretation  VNS is a safe and well-tolerated adjunctive treatment of epilepsy in children. Our results suggest that the effect of VNS on seizure frequency in children is limited. However, the possible reduction in seizure severity and improvement in well-being makes this treatment worth considering in individual children with intractable epilepsy.


Antiepileptic drug


Vagus nerve stimulation

What this paper adds

  •  This study is the first randomized controlled trial of the efficacy of vagus nerve stimulation in children with intractable epilepsy.
  •  This study suggests that the effect on seizure frequency is moderate in this highly refractory population.
  •  Possible reductions in seizure severity, improvements in well-being, and a favourable side-effect profile are additional benefits.

Vagus nerve stimulation (VNS) is a neuromodulatory treatment that is used as an adjunctive therapy for individuals with medically refractory epilepsy who are not eligible for epilepsy surgery or in whom surgery has failed, and in whom non-epileptic events are excluded. VNS consists of chronic intermittent electrical stimulation of the vagus nerve, delivered by a programmable pulse generator (neurocybernetic prosthesis [NCP]; Cyberonics Inc, Webster, TX, USA).

Randomized active controlled trials, which have predominantly included adults, have demonstrated the safety and efficacy of VNS:1,2 seizure frequency decreased by more than 50% in 23 to 31% of individuals in the treatment groups compared with 13 to 15% in the placebo group. These trials led to the U.S. Food and Drug Administration approval in 1997 of the use of VNS as adjunctive therapy in individuals older than 12 years with partial epilepsy refractory to treatment with available antiepileptic drugs (AEDs).

The effectiveness of VNS might be more variable in children than in adults. Numerous prospective and retrospective studies at various centres worldwide describing more than 650 children, aged 0 to 19 years, have reported a reduction in seizure frequency of more than 50% in 0 to 90%.3–28 However, these studies were uncontrolled, and there was a large variation in study groups, for example regarding age, epilepsy syndromes, and follow-up duration, which varied from 3 months to 10 years.

No randomized active controlled paediatric trial that unequivocally demonstrates the efficacy of VNS in children has yet been conducted. This study was carried out with the aim of evaluating the tolerability and effectiveness of VNS in children with intractable epilepsy. Moreover, we sought to identify responder characteristics that may improve future participant selection.


Study design

This study was a randomized active controlled double-blinded add-on study. The study was divided into two phases: baseline (12wks) and blinded (20wks) treatment phases. During the treatment phase, participants received either high (therapeutic) or low (active control) stimulation. This active control group was incorporated to protect the blinding, because participants can detect stimulation. Additionally, after the blinded phase all participants underwent a non-controlled follow-up, in which they received high stimulation (19wks; add-on phase).

All parents, guardians, and participants aged 12 years or above gave informed consent, and all procedures were approved by the ethics committee of Maastricht University Medical Center.


Forty-one children with medically refractory epilepsy participated in the study (See Fig. SI enrolment flowchart, supporting information published online). Children were included if they had medically refractory epilepsy despite adequate and stable AED concentrations, were aged between 4 and 18 years at implantation, and were not eligible for epilepsy surgery, and if written and signed informed consent was obtained from parents or guardians. Exclusion criteria included the following: non-epileptic seizures; a documented history of generalized status epilepticus in the previous 3 months; evidence of a progressive cerebral lesion, degenerative disorder, or malignancy in the previous 5 years; the presence of unstable medical disease (i.e. cardiovascular, hepatic, renal, musculoskeletal, gastrointestinal, metabolic, endocrine) in the previous 2 years; schizophrenia or any psychotic symptomatology; a high risk of complications (obstructive respiratory disease, gastric disorders, cardiac rhythm disorders); a history of alcohol or drug abuse, or of psychiatric disorder requiring electroconvulsive therapy or chronic use of major tranquillizers (neuroleptics, antidepressants) in the previous 6 months; regular treatment with antihistamines, metoclopramide, or central nervous system-active compounds; and treatment with an experimental drug during the previous 30 days.

Device and implantation

All surgical procedures were performed at the Maastricht University Medical Center, the Netherlands. Bipolar electrodes were placed around the left vagus nerve and connected to the programmable pulse generator, which was implanted subcutaneously or underneath the pectoral muscle below the clavicle.

Randomization, blinding, and device settings

The treating neurologist (MM), participants, parents, and guardians were blinded to treatment conditions. Participants were allocated to a treatment condition by one trial nurse (LL) using a computer program. The trial nurse also adjusted device settings according to a fixed protocol: 2 weeks after surgery, the device was set to the parameters depicted in Table I. Thereafter, in the treatment group the current was increased stepwise at 2-week intervals to the maximally tolerated output current (maximum 1.75mA) on the basis of the clinical evaluation of the treating neurologist. The stepwise increase was stopped when a participant experienced a reduction in seizure frequency of 50% or more or was delayed or set back to the highest tolerable level in the event of side effects occurring.

Table I.   Initial vagal nerve stimulation device settings
SettingTreatment groupActive control group
Output current (mA)0.250.25
Pulse width (ms)0.50.1
Frequency (Hz)301
Duty cycle: on (s)/off (min)30/514/60

In the active control group, the output current was increased temporarily during each visit and switched back to the initial output current at the end of the visit. At the end of the blinded phase, the output parameters of the active control group were adjusted according to the schedule and parameters of the treatment group. The maximally applied current at the end of this phase was set at 2.25mA for both groups.

Outcome parameters

Seizure frequency was recorded by participants’ parents or guardians using a diary. Several seizure types were scored separately. Seizures were classified according to the International League Against Epilepsy classification.29 Seizure frequency was recorded in the 12 weeks before implantation to serve as a baseline.

Seizure severity was estimated by administering the adapted Chalfont Seizure Severity Scale (NHS3) at baseline and at the end of the blinded and add-on phase to the caregiver who witnessed the seizures.30 This scale includes seven seizure-related factors and generates a score from 1 to 27: the higher the score, the more severe the seizures.

Overall satisfaction was evaluated at the end of the non-blinded add-on phase by standardized questioning of the parents or guardians.

Adverse events were recorded at each visit (12 in total) by questioning parents or guardians using a semi-structured interview.

IQ was assessed using the Peabody Picture Vocabulary test31 and the Beery–Buktenica Developmental Test of Visual-Motor Integration, 5th edition,32 a developmental test of visual motor integration. To calculate IQ, raw scores were converted to age equivalents in months, which were divided by age in months at the date of testing and multiplied by 100. Finally, the mean of both tests was calculated.

Statistical analysis

Power analysis was performed based on the assumption that 40% and 5% of participants in treatment and control groups respectively, had a clinically relevant response, which was defined as a reduction in seizure frequency of 50% or more. Assuming a level of significance of 0.05 (two-tailed) and a power of 80%, the size of our study population was calculated at 2 × 18.

Statistical analysis was performed using SPSS version 19 for MacOS (SPSS Inc, Chicago, IL, USA).

Seizure frequency during the blinded phase was expressed as a percentage change compared with seizure frequency during baseline for each participant. Seizure frequency changes were classified into different categories (≥50% increase, no response [<50% increase or decrease], ≥50% decrease). The number of participants experiencing a 50% or more reduction in seizure frequency was compared between the high- and low-output groups using Fisher’s exact test. Furthermore, we compared the percentage change in seizure frequency between the low- and high-output groups. First, the percentage change was transformed by calculating the natural logarithm of each value as these data were not normally distributed. The resulting values were normally distributed in both groups and, therefore, were compared using the Student’s t-test. The same analysis was repeated for the final 30 days of the blinded phase in order to assess any lag in efficacy or increased effectiveness over time. Overall improvement of seizure frequency was calculated by comparing the seizure frequency at the end of the add-on phase with baseline by a related-samples Wilcoxon signed-rank test. Finally, seizure frequency changes were classified into different categories (≥50% increase, no response, ≥50% decrease).

To assess seizure severity, first a mean NHS3 score was calculated for each participant by summing up the scores from each time point and dividing them by the number of seizure types. The median differences between groups were compared using a Mann–Whitney U test. Overall improvement of seizure severity was calculated by comparing the NHS3 score at the end of the add-on phase with baseline using a paired t-test.

We analysed the relationship between the final outcome at the end of the add-on phase and the following parameters by correlation analysis (Kendall’s tau): age at implantation, age at onset, and number of AEDs that had been used before implantation. The Mann–Whitney U test was used for the variables sex and presence of bilateral interictal discharges, and the Kruskal–Wallis test was used for aetiology.

A p-value <0.05 was considered statistically significant.



Forty-one children (mean age 11y 2mo; range 3y 10mo–17y 8mo; median 12y 4mo; 23 males; 18 females) underwent implantation of a VNS device after appropriate regulatory approval. Recruitment took place over a 2-year period. Seizure onset occurred at a mean age of 2 years 4 months (range 0–12y, median 1y 2mo).

Two participants had undergone unsuccessful epilepsy surgery several years before VNS. Participants had been exposed to an mean of 7.2 AEDs and were receiving a mean of 2.4 AEDs at implantation. Of the 41 children, 15 had been on a ketogenic diet.

The majority of participants had localization-related epilepsy, which in 10 participants was cryptogenic and in 25 was symptomatic. Four participants had symptomatic generalized epilepsy and only two participants had idiopathic generalized epilepsy. All but three children had learning disabilities.* Twenty-five participants had multiple seizure types, and 16 participants had a single seizure type.

An overview of the baseline characteristics of both groups is provided in Table II. There were no significant differences at baseline between the high- and low-output stimulation groups with regard to sex, age at implantation, age at onset, seizure frequency, International League Against Epilepsy classification (localization related: symptomatic, cryptogenic; generalized: idiopathic or symptomatic), or number of AEDs used before surgery.

Table II.   Baseline characteristics of study groups
 High-output stimulationLow-output stimulation
  1. ILAE, International League Against Epilepsy; AEDs, antiepileptic drugs.

Number of participants (male/female)21 (11/10)20 (12/8)
Mean age at implantation, y:mo (range)10:11 (3:10–17:8)11:6 (4:2–17:2)
Mean age at onset, y:mo (range)2:10 (0–12y)1:8 (0–5y)
Median age at onset, y:mo (range)1:2 (0–12y)1:2 (0–5y)
Mean interval onset-implantation, y:mo (range)7:8 (2–16y)9:5 (3–15y)
Median seizure frequency at baseline (seizures/day)2.1 (0.1–53.7)0.9 (0.1–31.7)
ILAE classification
 Localization related19 (90%)16 (80%)
 Symptomatic15 (71%)10 (50%)
 Cryptogenic4 (19%)6 (30%)
 Generalized2 (10%)4 (20%)
 Idiopathic02 (10%)
 Symptomatic2 (10%)2 (10%)
 Mean total exposure of AEDs7.0 (5–10)7.3 (4–14)

Three out of 41 participants who entered the study were excluded for seizure frequency calculations because of unreliable/incomplete seizure diaries. Two of the excluded participants were in the high-output group and one was in the low-output stimulation group. Additionally, information for four participants regarding the last 30 days of the add-on phase was missing.

Effects of low versus high stimulation – end of randomized controlled blinded phase

Seizure frequency

Three participants in the high-stimulation group and four in the low-stimulation group experienced a reduction in seizure frequency of 50% or more (p=1.000; Fig. 1a). Among participants receiving high stimulation, the median seizure frequency increased by 23.4%, compared with baseline, whereas among participants receiving low stimulation, seizure frequency fell by a median of 8.8% (comparison log-transformed values, p=0.61). Focusing on the last 30 days of the blinded phase, the median seizure frequency decrease was 3.1% and 5.1% in the high- and low-stimulation group respectively (comparison log-transformed values, p=0.47).

Figure 1.

 Effect of vagal nerve stimulation on seizure frequency and severity. (a) Response distribution for high- and low-stimulation group during double-blind phase ≥50% increase, no response, ≥50% decrease; (b) Distribution overall response at the end of the add-on phase 50% decrease, no response, ≥50% increase; (c) Mean seizure severity score according to the Chalfont Seizure Severity Scale, baseline compared to end of add-on phase, *p<0.05.

Seizure severity

The mean decrease in NHS3 score was 0.3 in the high-stimulation group and 0.6 in the low-stimulation group (p=0.71).

Overall effects of stimulation – end of add-on phase

Seizure frequency

Nine out of 34 participants experienced a 50% or more seizure frequency reduction, five experienced a 50% or more increase, and 20 did not respond at all (Fig. 1b). Seizure frequency decreased from a median of 1.61 seizures per day during the baseline phase to a median of 1.12 seizures per day at the end of the add-on phase (p=0.02).

Seizure severity

Seizure severity also improved, with a decrease in mean score from 9.5 at baseline to 8.3 at the end of the add-on phase (p<0.001; Fig. 1c).

Overall satisfaction

Overall, parents or guardians of 32 of the participants perceived some kind of positive effect at the end of the add-on phase as assessed by standardized questioning.

Responder characteristics

There was weak evidence of a correlation between seizure frequency reduction at the end of the add-on phase and age at onset (τ=0.213; p=0.08): children with a lower age at onset tended to have a better response. Age at implantation and the number of AEDs that had been used before implantation were not correlated with seizure frequency reduction. No statistically significant differences in seizure frequency reduction were observed between males and females, between different aetiologies, and between the children with or without bilateral interictal activity.

Adverse events

The most frequently reported adverse events were voice alterations, coughing, and throat pain (Table III). The majority of side effects were transient and most of them were stimulus related – some did not appear to be directly related to VNS. Reported behavioural changes consisted of agitation, crying, or frequent startles. Wound infection occurred in two participants. There was no need for device removal – both infections were successfully treated with short-term antibiotics. There were no other surgery-related side effects. Discontinuation during the study because of side effects did not occur.

Table III.   Adverse effects reported by participants, parents or guardians
Adverse eventNumber of participants
Voice alterations8
Throat pain3
Tingling sensations in throat2
Behavioural changes3
Spontaneous swelling around stimulator1
Pain around stimulator during exercise1


This study is the first randomized active controlled study on the effectiveness of VNS in children. We observed no statistically significant difference in seizure frequency reduction and seizure severity when comparing high- and low-stimulation groups. However, both seizure frequency and seizure severity at the end of the add-on phase were significantly decreased compared with baseline. A reduction in seizure frequency of 50% or more was reached in nine out of 34 of the participants at the end of the add-on phase. There was a trend towards a significant correlation between age at onset and a favourable response to VNS. Strikingly, the completion rate and overall satisfaction were very high, as stimulation was discontinued in only one out of 41 participants, because of the lack of effect, and more than three-quarters of parents and guardians reported some kind of improvement.

Our results confirm previous reports demonstrating that VNS in children is safe when performed by experienced neurosurgeons: no major adverse events occurred and side effects of VNS were mild, mostly transient, and related to stimulation. Infection occurred in only two participants and was completely resolved with short-term antibiotics.

We were not able to demonstrate a favourable effect of high stimulation versus low stimulation, as has been observed in adult randomized trials.1,2 Several factors may account for this. First of all, the size of our population might have been too small. Owing to the lack of other randomized studies in children, the power analysis was based on open-label studies, which suggested a larger effect than in adults. According to the current results, future studies should include a larger study population. Second, there are differences in vagus nerve electrophysiology between adults and children: threshold currents are higher and conduction velocities are lower in younger children than in older children and adults, indicating that maturation of the vagus nerve is not yet complete in young children.33 Moreover, the developing brain may respond differently to VNS. De Herdt et al.34 additionally observed a lower efficacy of VNS in children than in adults.

Our results particularly contrast with the findings of Murphy,5 who reported a median seizure frequency reduction of 23% after 3 months of stimulation. This difference may be explained by differences in experimental design: Murphy5 included 41 participants from the Compassionate Protocol as well. As this protocol is uncontrolled, VNS effectiveness may have been overestimated. Indeed, several other studies observed no seizure frequency reduction in some of their study population7,8,11,12,22,24 or in the entire study population.26,27

In our study, seizure frequency reduction in the lower output group was higher than expected. A decrease in seizure frequency may have resulted from the natural fluctuation of the disease, which is probably higher in children than in adults.7(addendum) Indeed, children were twice as likely as adults to respond with a greater than 50% seizure frequency reduction during placebo treatment.35 Although active controlled treatment is not equivalent to placebo treatment, it is unlikely that low stimulation may yield a true effect, as previous studies have demonstrated that the chosen combination of current intensity and pulse width does not evoke an action potential.33,36

Overall, VNS reduced seizure frequency by 50% or more in nine out of 34 participants. It has been proposed that the effectiveness of VNS is influenced by several factors. First, it has been suggested that learning disability may be a negative predictor.20,26,37,38 As all but three of our participants had learning disabilities, this may have contributed to the modest effect of VNS in our study. The chosen stimulation parameters may also have affected VNS efficacy. In our study, the output current was no higher than 1.75mA during the blinded phase and 2.25mA during the add-on phase in order to prevent demyelination of the nerve, which can occur even at normal output currents (1.5mA, 250μs, 20Hz).39 It is unlikely that a further increase in output current would have resulted in a larger effect of VNS, as our results demonstrated that a higher output current was not correlated with a more favourable response. Furthermore, even a low output current (<1mA) can reduce seizure frequency in a substantial portion of participants.40 Moreover, we did not adjust the duty cycle to rapid cycling, that is, a mode of stimulation with a faster intermittent pulse stimulation. We presume that this did not influence seizure reduction significantly because several studies have shown that rapid cycling does not provide any additional persistent seizure control over normal cycling.10,41,42 We cannot exclude the possibility that a longer follow-up would have yielded different results. Some studies have suggested that a larger percentage of participants respond at longer follow-up,23,43,44 while others do not demonstrate an effect of longer stimulation.12,26 However, in our study the number of responders at the end of the add-on phase was not significantly different from that in the blinded period.

An increase in seizure severity in response to VNS was observed as well, just as described in the adult trials2,45 and in one retrospective and two prospective paediatric trials.4,7,10 This might be explained by the natural course of the disease, which is variable over time: in both of the participants described by Parker et al. (addendum)7 who experienced an initial increase in seizure frequency, seizure frequency returned to baseline level when follow-up duration was prolonged. In one out of the six participants described by Helmers et al.10 who experienced a more than 50% increase in seizure frequency after 3 months, the increased seizure frequency persisted after 6 months.

Several other studies have tried to identify a profile of responders. In line with the results of Patwardhan et al.,8 we found a trend towards a correlation between age at onset and response to VNS. According to Janzsky et al.,46 the absence of bilateral interictal epileptic discharges and the presence of malformation of cortical development were factors predicting a favourable outcome. In our responding group, seven out of nine participants in whom 50% or more seizure frequency reduction was achieved had bilateral interictal epileptic dischargers compared with 18 out of 25 non-responders. Only one of the participants with malformation of cortical development (n=8) had a 50% or more seizure frequency reduction. Callosotomy before VNS treatment was reported to be associated with a positive response,10,46 but this was not the case in our participant who had undergone a callosotomy.

Finally, it is important to realize that a moderate effect on seizure frequency does not signify low utility of VNS. After all, this population of participants is highly refractory. For comparison, only 3% of children with refractory epilepsy become seizure-free after the addition of a third AED.47 Furthermore, in contrast to the effect of AEDs, when children respond positively to VNS, this response is long-lasting.5,10,25,48 The benefits therefore might be of great value for the individual child with therapy-resistant epilepsy. This is especially true when the safety of the implantation procedure and favourable side-effect profile of VNS are taken into account. Moreover, irrespective of the (lack of) effect on seizure frequency, the possible reduction in seizure severity and improvement in well-being make this treatment worth considering.49,50


  • *

    North American usage: mental retardation.


Sylvia Klinkenberg and Marlien W Aalbers contributed equally to this study.

Online Material/Supporting Information

Additional material and supporting information for this paper may be found online.