Vagus Nerve Stimulation for the Treatment

of Medically Intractable Seizures

Results of a 1-Year Open-Extension Trial
Martin C. Salinsky, MD; Basim M. Uthman, MD; Ruzica K. Ristanovic, MD; J. F. Wernicke, MD;
W. Brent Tarver, MS; and the Vagus Nerve Stimulation Study Group

Background: Chronic vagus nerve stimulation (VNS)
continues to be evaluated as an adjunctive treatment for
medically intractable seizures. A previous randomized
controlled trial of 114 patients demonstrated a signifi-
cant decrease in seizure frequency during 3 months of
VNS at effective stimulation levels.
Objective: To evaluate the efficacy of 1 year of VNS
therapy for the treatment of medically refractory partial
seizures and the relationship between initial and long-
term response.
Patients and Methods: All patients exiting the ran-
domized controlled study of VNS for treatment of
medically refractory partial seizures were offered
indefinite treatment extension as part of an open-label
trial. One hundred (88%) of 114 patients completed
12 months of VNS treatment at effective stimulation
levels. Fourteen patients discontinued VNS treatment
prior to 1 year, principally because of the treatment's
lack of efficacy. These 14 patients were retained in the
present analysis using an intent-to-treat approach.
Antiepileptic drug use was monitored throughout the
trial. Seizure frequency was analyzed in 4 sequential
3-month treatment periods.
Results: Compared with pretreatment baseline, there was
a significant decrease in seizure frequency during each
of the 3-month treatment periods. Seizure frequency was
reduced by a median of 20% during the first 3 months
of VNS treatment and by 32% during stimulation months
10 through 12. Response during the first 3 months of VNS
treatment was a statistically significant predictor of re-
sponse at months 10 through 12. The observed reduc-
tion in seizure frequency was not explained by overall
changes in antiepileptic drug use.
Conclusions: The results indicate that VNS remains an
effective adjunctive therapy for medically refractory par-
tial seizures over a period of at least 1 year. Response dur-
ing the first 3 months of treatment is predictive of long-
term response.
Arch Neurol. 1996;53:1176-1180

From the Oregon Health
Sciences University Epilepsy
Center, Portland
(Dr Salinsky); Department
of Neurology, University of
Florida, Gainesville
(Dr Uthman); Department of
Neurology,
Rush-Presbyterian-St Luke's
Medical Center, Chicago, Ill
(Dr Ristanovic); and
Cyberonics Inc, Webster, Tex
(Dr Wernicke and
Mr Tarver). A complete list
of members of the Vagus Nerve
Stimulation Study Group
appears on page 1180.
Chronic
vagus nerve
stimulation (VNS) has
been investigated as a
novel adjunctive treat¬
ment for medically intrac¬
table seizures. Fourteen patients with
poorly controlled partial seizures re¬
ceived VNS treatment in 2 single-blind
crossover pilot studies.u A fully implant-
able pacemakerlike signal generator de¬
livered trains of electrical pulses to the left
vagus nerve via bipolar stimulating elec¬
trodes.3 The pilot studies suggested that
VNS resulted in both acute and long-
term reductions in seizure frequency and
a randomized controlled trial of VNS fol¬
lowed.4 During the double-blind treat¬
ment phase, 114 patients with medically
refractory partial seizures received either
high-level stimulation (presumed effec¬
tive dose) or low-level stimulation (pre¬
sumed subtherapeutic dose). Twelve
weeks of VNS treatment followed a 12-
week pretreatment baseline evaluation.
Mean reduction in seizure frequency was
24.5% for the high-level stimulation group
vs 6.1% for the low-level stimulation
group, a statistically significant differ¬
ence indicating that VNS therapy can re¬
duce the frequency of medically intrac¬
table partial seizures.5 Treatment emergent
side effects were largely limited to a tran¬
sient hoarseness occurring during the
stimulus delivery.
Epilepsy is a chronic disorder usu¬
ally requiring long-term treatment; this
is particularly true for difficult to control
seizures. However, VNS and novel
See Methods on next page

Downloaded From: http://archneur.jamanetwork.com/ by a University of Iowa User on 06/10/2015

 METHODS
Chronic vagus nerve stimulation was accomplished using
an implanted pacemakerlike generator (model NCP100,
Cyberonics Ine, Webster, Tex) connected to 2 stimulating elec¬
trodes wrapped around the left vagus nerve. The generator
was implanted in the left subclavicular fossa.7 Stimulator set¬
tings were programmed via a telemetry wand placed on the
skin overlying the generator. Constant current output was
provided in 2 operating modes. In standard mode a train of
pulses with preset characteristics was delivered at fixed in¬
tervals throughout the day (typically 30 seconds of stimula¬
tion every 5 minutes). In manual activation mode patients could
trigger additional pulse trains by placing and removing a pocket
magnet over the generator. Patients and caregivers were in¬
structed to use this mode during auras or seizures.
RANDOMIZED CONTROLLED TRIAL
(PRIOR STUDY)
The randomized controlled trial of VNS (Study E03) for
the treatment of medically intractable seizures included 114
patients. The design and results of study E03 have been pub¬
lished.3 All patients had medically intractable seizures domi¬
nated by partial seizures, with a seizure frequency of at least
6 per month (median, 0.79 seizures per day) confirmed dur¬
ing a 3-month baseline observation period and with no more
than 2 weeks between seizures. Patients were excluded from
the study if they had any progressive or unstable medical
condition or were receiving more than 3 AEDs or under¬
going any experimental therapies. Mean patient age was 33
years (range, 13-52 years).
Following the baseline evaluation, all patients under¬
went implantation with identical stimulators. Two weeks
later the generator was activated and patients were ran¬
domized to receive high-level stimulation (presumed ef¬
fective dose) or low-level stimulation (presumed ineffec¬
tive dose). The low-level dose was necessary to preserve
the blind by providing patients with a sensation of stimu¬
lation. This blinded, parallel treatment phase continued for
12 weeks, during which patients continued to receive base¬
line AEDs at stable doses (Figure 1 ).
OPEN-EXTENSION TRIAL (CURRENT STUDY)
At the conclusion of the blinded, parallel phase of the study,
all patients were invited to continue VNS treatment indefi¬
nitely in an open-extension trial. The blind was broken.
Stimulator settings for patients initially randomized to low-
level stimulation were adjusted to the high-level stimula¬
tion range. Patients initially randomized to receive high-
level stimulation continued to receive this treatment. During
the open extension, investigators were free to adjust the stimu¬
lator settings (current, frequency, pulse width, on time, and
off time) according to clinical response. Changes in AED or
AED dosage were allowed during the open-extension phase.
Monthly seizure counts and reports of potential side effects
were obtained on standard collection forms.
ANALYSIS
This analysis includes the first 12 months of high-level
stimulation for each patient, regardless of assignment
Downloaded From: http://archneur.jamanetwork.com/ by a University of Iowa User on 06/10/2015
during the randomized controlled trial. The study design
is illustrated in Figure 1. For patients randomized to high-
level stimulation during the randomized controlled trial,
this analysis includes the 3 months of the randomized trial
and the first 9 months of the open extension (total of 12
of months high-level stimulation). For those patients who
received low-level stimulation during the randomized trial,
this analysis includes the first 12 months of the open ex¬
tension (ie, the first 12 months of high-level stimulation).
Combining patients initially randomized to the high- and
low-level treatment groups was justified based on a pre¬
liminary analysis that revealed no significant differences in
response during long-term high-level stimulation.6 For each
patient, monthly seizure counts were compiled into 3-month
analysis periods. Period 1 included the first 3 months of
high-level stimulation, period 2 included months 4 through
6 of high-level stimulation, and periods 3 and 4 included
months 7 through 9 and 10 through 12, respectively.
Monthly seizure counts were tabulated for each pa¬
tient using a last visit, carried-forward, intent-to-treat ap¬
proach. For those patients who discontinued VNS treat¬
ment prior to completing 12 months of high-level
stimulation, the last monthly seizure count was used (car¬
ried forward) to estimate the frequency of seizures over the
remainder of the 12-month assessment period. Similarly,
missing data points were filled in using the previous monthly
seizure count for that patient. This conservative analysis
method minimized the potential bias introduced by ex¬
cluding study dropouts, many of whom had a relatively poor
initial response to VNS treatment.
Each 3-month analysis period was compared with the
3-month pre-VNS baseline assessment period for that pa¬
tient. Percentage of difference in seizure frequency (com¬
pared with baseline) was calculated for each patient at each
of the 4 analysis periods. The distribution of percentage
change values at each analysis period was nonnormal (Sha-
piro-Wilk statistic). Therefore, for each analysis period a
Wilcoxon signed rank test evaluated the hypothesis that
the mean percentage change in seizure frequency for the
entire group equaled 0. Differences between the 4 analysis
periods were evaluated using the nonparametric Fried¬
man analysis of variance (ANOVA) test.
Absolute seizure frequency distributions were also non-
normally distributed for each analysis period. Therefore,
group differences in median seizure frequency from base¬
line to each analysis period were evaluated using the Wil¬
coxon signed rank test. Differences in seizure frequency be¬
tween the 4 analysis periods were evaluated by the Friedman
ANOVA test. Differences in the distribution of responders
(defined by a a50% decrease in seizure frequency) at
months 1 through 3 and 10 through 12 were analyzed us¬
ing the McNemar test for no association.
All analyses were performed on an intent-to-treat ba¬
sis, including the entire group of 114 patients with im¬
planted devices. A supplementary analysis was performed ex¬
cluding 24 patients who had protocol violations during the
controlled portion of the study (pure pool), including 18 pa¬
tients with a 6- to 12-month interval between baseline and
randomized treatment and 6 patients who were improperly
randomized.5 Statistical analyses were performed at the Or¬
egon Health Sciences University, Portland, using monthly sei¬
zure counts collected by Cyberonics Ine, the study sponsor.
All statistical tests were performed using SAS software (ver¬
sion 6.04, SAS Institute Ine, Cary, NC) and were 2-tailed.
 High
12 wk
12 wk /
Baseline Low
—
Randomized Phase
/
/
Patients Originally Randomized High
s.-;__'-;-1
Mo 0/
¿_
1 4 5 6
Patients Originally Randomized
Figure 1. Diagram of the study design including the randomized controlled trial and the open-extension trial. The present analysis includes the first 12
months of high-level stimulation (months O through 12) regardless of assignment during the randomized phase. Patients originally randomized to high- and
low-level stimulation were combined by offsetting the start dates by 3 months.
Figure 2. Percentage of baseline seizure frequency during the first 12
months of high-level vagus nerve stimulation for all patients (N= 114) and
separately for patients initially randomized to the high- or low-level
stimulation groups during the controlled portion of the study. Each bar
represents the median value for the group during the specified stimulation
period. Baseline is defined as 100%.
antiepileptic drugs (AEDs) are usually tested over rela¬
tively short periods of 2 to 4 months. Treatments may
be effective for the first several months and then lose ef¬
ficacy. Alternatively, some treatments could be more ef¬
fective when used over longer periods. Demonstration
of long-term efficacy is therefore an important factor to
be considered prior to acceptance of a new treatment of
epilepsy.
exiting the randomized controlled trial of
Patients
VNS were offered indefinite treatment extension as part
of an open-label trial. A preliminary analysis of the first
67 patients entering long-term treatment suggested that
the reduction in seizure frequency observed during the
controlled study phase was maintained through up to 18
months of follow-up.6 A subgroup of patients had a sta¬
tistically significant decrease in seizure frequency after
16 to 18 months of VNS treatment compared with the
first 3 months of treatment. However, the preliminary
analysis included only those patients who elected to con¬
tinue therapy, introducing a significant bias toward im¬
proved efficacy over time.
Downloaded From: http://archneur.jamanetwork.com/ by a University of Iowa User on 06/10/2015
Extension
7 8 io 11
_?
yí
Low
We report the results of a 1-year follow-up intent-
to-treat analysis of all 114 patients who underwent sig¬
nal generator implantation and received stimulation dur¬
ing the randomized controlled trial of VNS. We examined
the efficacy of long-term stimulation in the treatment of
refractory seizures and the relationship between initial
response and long-term response in individual patients.
RESULTS
One hundred (88%) of the 114 patients entering the ran¬
domized controlled study completed at least 12 months of
high-level stimulation. Fourteen patients discontinued treat¬
ment prior to completing 12 months. The most common
reason for discontinuation was patient request because of
perceived lack of efficacy (9 patients, 3 to 11 months of high-
level stimulation). Two patients died during the 12-
month observation. One patient drowned after 3 months
of high-level stimulation. The other patient died of throm¬
botic thrombocytopenic purpura after 6 months of high-
level stimulation. Two patients discontinued treatment be¬
cause of technical problems (1 electrode lead breakage at
7 months and 1 generator failure at 1.5 months). One pa¬
tient discontinued treatment after suffering a nonfatal myo¬
cardial infarction after 2 months of high-level stimula¬
tion. Of the 14 patients discontinuing treatment, 8 were
initially randomized to high-level stimulation and 6 to low-
level stimulation. All 14 patients were included in the pres¬
ent analysis by carrying forward the last monthly seizure
count prior to discontinuing VNS treatment.
Figure 2 illustrates the percentage of baseline sei¬
zure frequency for all seizures during each 3-month stimu¬
lation period. Each bar represents the median value for
the 114 patients in the intent-to-treat group. The reduc¬
tion in seizure frequency during the first 3 months of high-
level stimulation (period 1) was maintained over the
following 9 months. At each analysis period the median
percentage decrease in seizure frequency was signifi¬
cantly greater than 0 (Table). There was a trend sug¬
gesting improved seizure control with longer use of VNS.
However, intraindividual differences in seizure fre¬
quency between analysis periods 1 and 4 were not
significantly different from 0 (Wilcoxon signed rank test,
P=.19) and the ANOVA including all 4 treatment
periods did not reach statistical significance (P=.33).
 Seizure Frequency for All Patients During Baseline Evaluation and the First Year of Stimulation*
Frequency Baseline Period 1
Seizures per day (P) 0.79 0.50 (<.01)
Percent of baseline seizures (P) 100 79.8 (<.01)
==50% Reduction in seizures, % .. .t 28
*Periods 1 through 4 are sequential 3-month analysis periods. For seizures per
the 114 patients. values are from the Wilcoxon signed rank test.
1Ellipses indicates baseline serves as control period.
Results for the intent-to-treat group (n= 114) and the pure
pool group (n=90) were similar, with a median 31.9%
decrease in seizure frequency during stimulation months
10 through 12 (analysis period 4) in the intent-to-treat
group vs a 36.2% decrease in the pure pool. The 14 pa¬
tients who did not complete 1 year of high-level stimu¬
lation had a median 11% decrease in seizure frequency
at stimulation months 10 through 12 based on the last
visit, carried-forward analysis.
Patients initially randomized to the high-level stimu¬
lation group and those initially in the low-level stimula¬
tion group had similar results during the first 12 months
of high-level stimulation (Figure 2). Median percentage de¬
crease in seizure frequency during period 4 was 31.6% for
patients initially randomized to low-level stimulation and
32.1% for patients initially randomized to high-level stimu¬
lation. There were no statistically significant differences be¬
tween the 2 subgroups at any of the 4 analysis periods.
Absolute seizure frequency at each analysis period
was significantly reduced compared with baseline (Table
1). A 5-group Friedman ANOVA test, including the pre¬
treatment baseline and all 4 analysis periods, was sig¬
nificant at P<.01. However, there were no significant dif¬
ferences between the 4 analysis periods. Wilcoxon signed
rank tests indicated a significant decrease in absolute sei¬
zure frequency at each 3-month stimulation period com¬
pared with the pretreatment baseline (Table 1). The per¬
centage of patients who achieved at least a 50% reduction
in seizure frequency also remained stable across the 4
analysis periods (Table 1).
Thirty-two (28%) of 114 patients achieved at least
a 50% reduction in seizure frequency during the first 3
months of high-level stimulation (period 1). Of these 32
patients, 21 (66%) maintained a 50% or greater reduc¬
tion in seizures during stimulation months 10 through
12 (period 4). Patients who failed to achieve a 50% re¬
duction in seizures during period 1 were relatively un¬
likely to do so during period 4 (14 [17%] of 82 pa¬
tients) The response to VNS treatment at period 4 was
not significantly different from the response at period 1
.
(McNemar test for no association, P=.55).
Complete AED use data at 1-year follow-up were
available for 99 of the 100 continuing patients. Com¬
pared with baseline, the total number of AEDs used in¬
creased in 7 patients, decreased in 17 patients, and was
unchanged in 75 patients. There was no change in AEDs
or dosage in 35 patients. The remaining 64 patients re¬
corded 33 dosage increases of more than 10%, 30 dos¬
age decreases of more than 10%, and 7 AED substitu¬
tions. To provide a summary of overall change in AED
Downloaded From: http://archneur.jamanetwork.com/ by a University of Iowa User on 06/10/2015
Period 2 Period 3 Period 4
0.60 (<.01) 0.58 (<.01) 0.53 (<.01)
75.3 (<.01) 74.3 (<.01) 68.1 (<.01)
24 25 31
day and percent of baseline seizures, each value represents the median for
load for each patient, we considered an increase or de¬
crease in the number of AEDs used as the primary factor
(an AED substitution was defined as a neutral change).
Patients with no change in the number of AEDs used at
1-year follow-up were considered to have an increased
AED load if any drug had a dosage increase of greater than
10% and a decreased load if any drug dosage decreased
more than 10%. Offsetting changes of 2 AEDs in any 1
patient were considered neutral. Using these defini¬
tions, 21 patients had an overall increase in medication
load, 31 patients had an overall decrease, and 47 pa¬
tients were unchanged or neutral. Among the 35 pa¬
tients who had at least a 50% decrease in seizure fre¬
quency at months 10 through 12 of VNS treatment, 7 had
an overall increase and 9 an overall decrease in medica¬
tion load at 1-year follow-up.
Two deaths occurred during the 12-month open-
extension trial: 1 related to thrombotic thrombocytope¬
nic purpura and 1 to drowning. The relationship of these
deaths to VNS treatment is questionable. One patient with
no previous history of cardiac disease experienced a non-
fatal myocardial infarction after 8 weeks of stimulation.
No other cardiac events were observed in the 114 pa¬
tients during the 12-month follow-up. There were no gas¬
trointestinal complications. Two devices malfunc¬
tioned. One generator locked into a continuous DC
stimulation leading to a unilateral vocal cord paralysis
with subsequent partial recovery. One electrode wire frac¬
tured leading to discontinuation of therapy. Chronic side
effects of VNS treatment (in &5% of patients) were iden¬
tical to those observed during the randomized con¬
trolled trial and consisted mainly of mild hoarseness dur¬
ing stimulus delivery.5
COMMENT
The results indicate that the decrease in seizure fre¬
quency demonstrated during a 3-month randomized con¬
trolled trial of VNS treatment was sustained throughout
12 months of continued treatment. There was a statisti¬
cally significant decrease in seizure frequency during each
of 4 sequential 3-month treatment periods compared
with
baseline. There was no evidence of a loss of efficacy over
time. We observed a trend toward improved seizure con¬
trol with longer use of VNS. However, differences be¬
tween the first and last 3 months of high-level stimula¬
tion did not reach statistical significance based on this
analysis of 114 patients.
Several potentially important uncontrolled vari¬
ables could have influenced the results of this open-
 The Vagus Nerve Stimulation Study Group

Participating Institutions and Principal Investigators

Baylor College of Medicine, Houston, Tex: R. George, MD. Hans Berger Kliniek, Breda, Holland: A. Sonnen, MD.
McMaster University, Hamilton, Ontario: A. Upton, MD. Oregon Health Sciences University, Portland:
M. Salinsky, MD. Rush-Presbyterian-St Luke's Medical Center, Chicago, 111: R. Ristanovic, MD; D. Bergen, MD. St Louis
University, St Louis, Mo: W. Mirza, MD. St Luke's Hospital, St Louis: W. Rosenfeld, MD. Southern Illinois University,
Springfield: D. Naritoku, MD. Temple University, Philadelphia, Pa: R. Manon-Espaillat, MD. Thomas Jefferson Univer¬
sity, Philadelphia: G. Barolat, MD. Tulane University, New Orleans, La: J. Willis, MD. Universität Erlangen-Nurnberg,
Erlangen, Germany: H. Stefan, MD; T. Treig, MD. Universitätsklinic Bonn, Bonn, Germany: A. Hufnagel, MD. University
of Alabama, Birmingham: R. Kuzniecky, MD. University of Florida, Gainesville: B. Uthman, MD; B. J. Wilder, MD.
University of Göteborg, Sweden: L. Augustinsson, MD, PhD; E. Ben-Menachem, MD, PhD. University of Miami, Miami,
Fla: E. Ramsay, MD. Cyberonics Ine, Webster, Tex: J. F. Wernicke, MD, PhD; W. B. Tarver, MS; J. Allen, PhD.

label follow-up study. Placebo effects may have oc¬
curred once the patients and investigators were unblinded,
and these effects may have contributed to the further re¬
ductions in seizure frequency following the blinded phase
of the trial. Antiepileptic drugs were adjusted according
to clinical response and new AEDs were allowed, poten¬
tially improving seizure control. However, the analysis
of total AED load at 1 year of treatment compared with
baseline revealed an overall decrease in the number of
AEDs used and in AED dosages, suggesting that AED
changes did not contribute to the observed long-term ef¬
the first 12 months of high-level stimulation. Analysis of
the 2 groups served as an internal replication of the results
since the original assignments to the high- and low-level
stimulation groups were random. A 12-month follow-up
analysis of 13 patients exiting the 2 VNS pilot studies had
similar results, with an average 38% decrease in seizure fre¬
quency after 9 months of stimulation.4 In the present study,
response at 1 year was significantly related to response dur¬
ing the first 3 months of high-level stimulation. Therefore,
the group response to VNS treatment at analysis period 4
was largely owing to the same patients continuing to ben¬

ficacy of VNS treatment.
Given the open-label design of this study we used a
conservative intent-to-treat analysis, carrying forward the
last visit data on patients who discontinued VNS treat¬
ment before completing 12 months of high-level stimu¬
lation. The majority of these patients had a relatively poor
response to VNS treatment with a median 11% decrease
in seizure frequency at the time of discontinuation (as
opposed to a 20%-32% decrease for all 114 patients). In¬
cluding the last visit data on these patients in place of
subsequent visits eliminated the possibility of skewing
the overall results by removing poor responders from the
analysis and also assumed that poor responders would
not improve with a longer duration of treatment.
Our use of an intent-to-treat analysis is the likely
explanation for differences between the current results
and those previously reported in a preliminary analysis
of the first 67 patients exiting the 3-month controlled trial.6
The previous study reported a 41.4% median decrease
in seizure frequency at months 10 through 12 of high-
level stimulation, as opposed to 31.9% in the present
analysis. However, the preliminary analysis included only
those patients who continued VNS treatment through¬
out the follow-up period and did not account for 12% of
patients who discontinued VNS treatment or were not
adequately followed up. Conclusions regarding a fur¬
ther decrease in seizure frequency with continued VNS
treatment up to 18 months should be interpreted cau¬
tiously, given that 25% of patients were excluded from
the prelimary analysis at the 16- to 18-month interval.
Additional observations suggest that the beneficial ef¬
fects of VNS treatment are sustained during long-term treat¬
ment. In the present study, separate analyses of patients origi¬
nally randomized to the high- and low-level stimulation
groups revealed nearly identical response profiles during
efit rather than transient or inconsistent responses in dif¬
ferent subsets of patients. Taken together, these observa¬
tions indicate that patients who benefit from VNS treatment
are likely to continue to benefit over a period of at least 1
year. Longer observation periods with larger numbers of
patients will be needed to determine whether the response
to VNS treatment changes beyond 1 year of stimulation and
whether any long-term safety problems exist.
Accepted for publication August 5,1996.
The authors thank J. Annegers, PhD, J. Cereghino,
MD, and J. Buchhalter, MD, PhD, for helpful comments
on the manuscript and Jay e Thompson, MS, for statistical
assistance.
Reprints: Martin C. Salinsky, MD, Oregon Health Sci¬
ences L/niversity Epilepsy Center (CDW3), 3181 Samfack-
son Park Rd, Portland, OR 97201.
REFERENCES
Penry JK, Dean JC. Prevention of intractable partial seizures by intermittent vagal
stimulation in humans: preliminary results. Epilepsia. 1990;31(suppl):40S-43S.
Uthman BM, Wilder BJ, Hammond EJ, Reid SA. Efficacy and safety of vagus
nerve stimulation in patients with complex partial seizures. Epilepsia. 1990;31
(suppl):44S-50S.
Terry R, Tarver B, Zabara J. An implantable neurocybernetic prosthesis sys-
tem. Epilepsia. 1990;31(suppl):33S-37S.
Uthman BM, Wilder BJ, Penry JK, et al. Treatment of epilepsy by stimulation
of the vagus nerve. Neurology. 1993;43:1338-1345.
Vagus Nerve Stimulation Study Group. A randomized controlled trial of chronic
vagus nerve stimulation for treatment of medically intractable seizures. Neu-
rology. 1995;45:224-230.
George R, Salinsky M, Kuzniecky R, et al. Vagus nerve stimulation for treat-
ment of partial seizures: long-term follow-up on first 67 patients exiting a con-
trolled study. Epilepsia. 1994;35:637-643.
Reid SA. Surgical technique for implantation of the neurocybernetic prosthe-
sis. Epilepsia. 1990;31 (suppl):38S-39S.

Downloaded From: http://archneur.jamanetwork.com/ by a University of Iowa User on 06/10/2015