Journal of Psychiatric Research 36 (2002) 219–227

www.elsevier.com/locate/jpsychires

Vagus nerve stimulation (VNS) synchronized BOLD fMRI suggests

that VNS in depressed adults has frequency/dose dependent effects
Mikhail Lomareva,e,*, Stewart Denslowa, Ziad Nahasb, Jeong-Ho Chaea,b,f,
Mark S. Georgea,b,c,d, Daryl E. Bohninga
a
Department of Radiology, Medical University of South Carolina, 171 Ashley Avenue, Charleston, SC 29425, USA
b
Department of Psychiatry, Medical University of South Carolina, 171 Ashley Avenue, Charleston, SC 29425, USA
c
Department of Neurology, Medical University of South Carolina, 171 Ashley Avenue, Charleston, SC 29425, USA
d
The Ralph H. Johnson Veterans Hospital, Charleston, South Carolina, USA
e
Institute of the Human Brain, St. Petersburg, Russia
f
The Catholic University of Korea, Seoul, South Korea

Received 27 August 2001; received in revised form 8 February 2002; accepted 22 February 2002

Abstract
Stimulation of the vagus nerve in the neck can reduce seizures in epilepsy patients, and may be helpful in treating depression.
PET studies have shown that vagus nerve stimulation (VNS) in epilepsy patients causes acute dose (intensity) dependent changes in
regional cerebral blood flow. We sought to use the newly developed VNS synchronized fMRI technique to examine whether VNS
BOLD signal changes depend on the frequency of stimulation. Six adults with recurrent depression were scanned inside a 1.5 T MR
scanner. Data were acquired at rest, with the VNS device on for 7 s, and also, for comparison, while the patient listened to a tone for 7 s.
In two separate back-to-back sessions, the VNS stimulation frequency was set to either 5 or 20 Hz. Data were transformed into Talair-
ach space and then compared by condition. Compared to 5 Hz, 20 Hz VNS produced more acute activity changes from rest in regions
similar to our initial VNS synchronized fMRI feasibility study in depression. Brain regions activated by hearing a tone were also
greater when VNS was intermittently being applied at 20 Hz than at 5 Hz. In depressed adults, left cervical VNS causes regional
brain activity changes that depend on the frequency of stimulation or total dose, or both. In addition to the acute immediate effects
of VNS on regional brain activity, this study suggests further that VNS at different frequencies likely has frequency or dose
dependent modulatory effects on other brain activities (e.g. hearing a tone). # 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Vagus nerve stimulation; Depression; FMRI

1. Introduction generator that can deliver intermittent electrical current

for variable on and off times at different intensities, fre-
Vagus nerve stimulation (VNS; Zabara, 1985, 1992) quencies, or pulsewidths. Thus, VNS can be administered
with the Neurocybernetic Prosthesis (NCP1) Generator with a range of at least five different use parameters
(Cyberonics, Inc., Houston, TX) has shown beneficial (intensity, frequency, pulsewidth, on-time, off-time). Sci-
clinical effects in treating epilepsy (Ben-Menachem et al., entists do not fully understand the neurobiological
1994; Handforth et al., 1998), and has recently shown effects of these different use parameters, either alone or
promise in treating patients with major depression (George in combination, although VNS at different intensities
et al., 2000a,b; Rush et al., 2000; Sackeim et al., 2001). has different effects on emotional memory (Clarke et al.,
Vagus nerve stimulation is applied through an elec- 1999) and pain perception (Ness et al., 2000; Kirschner
trode wrapped around the left vagus nerve in the neck. et al., 2000).
The electrode is connected to a subcutaneous pulse Following the discovery by Zabara in 1985 that VNS
could stop seizures, work was done to better understand
* Corresponding author. Tel.: +1-843-792-5571; fax: +1-843-792-
the use parameters. Researchers used animal studies
5067. with EEG and EMG to determine the use parameters
E-mail address: [email protected] (M. Lomarev). most likely to be effective for epilepsy (Zabara, 1992),
0022-3956/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S0022-3956(02)00013-4
220 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227

which were then employed in initial clinical trials in exploration of the effects of different stimulation para-
epilepsy (Ben-Menachem et al., 1994; Handforth et al., meters and regimen, and, possibly, extend its applica-
1998) and then adopted for the current clinical studies in tion to other disorders.
patients with depression (George et al., 2000a,b; Rush et In this current preliminary study, we wished to exam-
al., 2000; Sackeim et al., 2001). Still, there is incomplete ine whether different frequencies of VNS have differing
understanding of the regional neurobiologic effects of dif- brain effects. We therefore used interleaved VNS and
ferent VNS use parameters. More information is needed Blood Oxygen Level Dependent (BOLD) fMRI to
about the effects of different VNS parameters on regional measure the regional cerebral blood flow related chan-
brain activity to help set optimal dosing in clinical use. ges resulting from intermittent applications of 7 s of
Several groups have used positron emission tomo- either 20 Hz or 5 Hz VNS in patients with major
graphy (PET) to investigate the chronic effects of VNS depression.
(Ko et al., 1996; Garnett et al., 1992). Another group
has studied acute effects of VNS by comparing the dif-
ferences between PET images acquired before and dur- 2. Subjects and method
ing VNS stimulation (Henry et al., 1998, 1999).
Unfortunately, the low temporal resolution of PET 2.1. Subjects
limits its observations to integrated effects over at least
a minute (longer than most VNS trains), and its depen- Nine subjects, who had participated in a recent VNS
dence on radioactive tracers makes it unsuitable for clinical study for treatment-resistant depression (Rush
repeated use. We recently demonstrated the feasibility et al., 2000), were initially recruited for this study and
of performing functional magnetic resonance imaging brought in for scanning. Four of them participated in
(fMRI), with its relatively high spatial and temporal our previous VNS fMRI study (Bohning et al., 2000).
resolution, during VNS (Bohning et al., 2001). This Of these initial nine, three had to be excluded from the
technique can reveal the exact location and level of the final data analysis because the device would not restart
brain’s immediate response to VNS, and allows for the while in the MR scanner, leaving six subjects for the
examination of the regional brain effects of different final analysis. There were three men and three women;
VNS settings, possibly leading to more effective treat- their ages ranged from 43 to 59 years (mean 49.5  6.1
ment regimens. S.D. years; Table 1). These subjects had their NCP pulse
Intermittently stimulating a single nerve cell elec- generators implanted in an MRI-compatible fashion,
trically at different frequencies produces drastically dif- i.e. with the lead pins oriented along the long axis of the
ferent changes in neuronal behavior. Low frequency body (Maniker et al., 2000, and unpublished indepen-
stimulation induces long term depression (LTD); while dent determination by DEB). They varied markedly in
intermittent high frequency stimulation produces long their mood state at the time of scanning, and their clin-
term potentiation (LTP). Similarly transcranial mag- ical response to VNS. The time since the start of their
netic stimulation (TMS) of the cortex at different fre- therapeutic VNS varied from 8 to 19 months (mean
quencies can temporarily inhibit (Wassermann et al., 14.2  4.1 S.D. months). They were on diverse chronic
1998), excite (Sawaki et al., 1999) or even arrest func- VNS settings as well as taking a variety of anti-
tion (Epstein, 1998). Knowing whether VNS at different depressant medications. The subjects signed a written
frequencies or doses evokes similar spatial patterns of informed consent that was approved by the Medical
brain activation would greatly advance knowledge of University of South Carolina’s Institutional Review
the basic mechanisms of action of VNS, accelerate the Board for Human Research.

Table 1
Subject demographics

Patients Age Months from implantation HDRSa28

Before NCPb implantation On day of fMRI

Pt1 43 16 42 13
Pt2 56 8 39 23
Pt3 50 11 26 15
Pt4 44 17 35 26
Pt5 45 14 44 6
Pt6 59 19 31 21
a
HDRS28, 28 item Hamilton Depression Rating Scale.
b
NCP, the Neurocybernetic Prosthesis.
 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227 221

2.2. General experimental design placed on the subjects’ neck. Electrodes were positioned
over the route of the implanted VNS leads, typically
Subjects were scanned in a 1.5T clinical MR scanner about 1.5 cm apart, just above and below the scar of the
(Signa Horizon LX, SR 77 gradients, software Rel. 8.3 incision made during surgery for attaching the implan-
M5, GE Medical Systems, Inc. Milwaukee, WI) with a ted VNS electrodes to the vagus nerve. Subjects were
send/receive RF head coil. Prototype software provided given earplugs and headphones, and instructed to lie
by GE was used for the multi-slice single-shot gradient- quietly with their eyes closed and listen for the tone.
echo EPI-fMRI acquisitions (6464 matrix, FOV=270 While lying on the gantry outside of the scanner bore,
mm,
=88 , TE=40.0 ms, slice thickness=8.0 mm, the subjects inserted their head into the scanner head
gap=0.0 mm, with fat saturation). Fifteen contiguous 8 coil and adjusted their position until they were centered
mm thick axial slices were acquired parallel to the and comfortable. Their head was then stabilized with
Anterior Commissure-Posterior Commissure (AC-PC) foam-padded Velcro restraints. Subjects were then
Line, and centered so that the AC-PC line was in slice moved into scanning position in the scanner bore.
number 5. A set of T1-weighted structural images (TE Bypassing the scanner’s ECG processing hardware/
20 ms, TR 600 ms) was acquired with the same slice software, the NCP System’s pulse generator signal
coverage for anatomical reference. Details of the recorded from the patient’s neck was fed to an isolated
method were published earlier (Bohning et al., 2001). bioamplifier and high performance band pass filter
Two fMRI scanning sessions were performed on each (LabLink V System Model V75–05 and Model V75–48,
subject in a randomized counterbalanced method across respectively, Coulbourn Instruments, Allentown, PA),
subjects, with subjects having to exit the MR scanner 2.0 kHz to 4.0 kHz. This system allows for synchroni-
between the runs in order to have the VNS device reset. zation of the MRI scanning cycle with the VNS gen-
Three of six subjects received 5 Hz during the first ses- erator cycle and has been previously described (Bohning
sion and three during the second session. In each ses- et al., 2001).
sion, the VNS generator was programmed to deliver a 7
s on–108 s off (nominal) stimulation cycle. Because of 2.4. Data processing
the device’s clock cycle, and the fact that the VNS
device ramps up and down, this created an effective sti- The reconstructed structural images and fMRI
mulation cycle of 13 s on and 103 s off, 116 s for the k-space raw data were transferred to a Sun workstation
entire cycle. The pulse width was 500 ms; the current (Sun Microsystems, Inc., Mountain View, CA, USA),
(intensity) settings, were left at the patient’s treatment and the structural images were stored for anatomical
level setting, and ranged from 0.25 mA to 1.25 mA reference. The raw fMRI data was first reconstructed
(mean 0.79  0.27; see Table 3). After completing the into images using research software from GE (‘‘epir-
entire MRI scan (lasting about 90 min), their device was econ’’) and then subjected to a standard set of fMRI
reset to the individual’s original parameters. data processing steps, described below.
As an internal reference for comparing VNS response Motion Correction: Using MEDx 3.2 (Sensor Sys-
at different settings, a control auditory stimulus was tems, Inc., Sterling, VA), a check was performed on
used. During the MRI scan procedure, a 440 Hz tone each image set to determine if subject movement was
was played through the scanner’s sound system to less than 2 mm along all of the three major axes (x, y, z).
headphones on the subject and interleaved in 7 s trains All of the data sets met this test. In three data sets,
of 100 ms long pulses, 100 ms apart, on alternate 58.0 s movement in at least one of the three directions was
epochs of the VNS epoch-TONE epoch stimulation between 1 and 2 mm, so the images were motion cor-
cycle. During each cycle, 40 15-slice volumes of BOLD- rected and co-registered to the images in the acquisition
EPI images were acquired with the TR=2.9 s. The
VNS-REST1-TONE-REST2 cycle was repeated 10
times for a total of 580 s or 9 min, 40 s (see Fig. 1). In
summary, the experiment has a repeating block design
with four epochs: VNS, Rest 1, Tone, Rest 2; each
epoch is 10 images per slice/29 s; the block of four
epochs is 40 images per slice/116 s. The VNS and TONE
stimuli occur for 7 s at the start of the respective epochs.

2.3. VNS synchronization and fMRI scan acquisition

Fig. 1. Experimental paradigm of synchronized interleaving of TON
and fMRI acquisitions. Each fMRI cycle is composed of two epochs:
Before subjects entered the scanner, standard adhesive (1) 13 s (7 s peak) of VNS followed by 45.0 s of rest to measure the
backed MRI-compatible electrodes (re-usable silver associated hemodynamic response and (2) 7 s on a 440 Hz tone (TON)
chloride, Coulbourn Instruments, Allentown, PA), were followed by 51.0 s of rest.
222 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227

midway through the first rest (REST1) epoch using the 3. Results
motion correction algorithm in MEDx, which is based
on AIR (Woods et al., 1998). There were no adverse effects and none of the subjects
Generating Group Data in Talairach Space: Using noticed any aberrant VNS stimuli while in the scanner.
the SPM96 module (Frackowiak et al., 1997) in MEDx
3.2, each person’s functional data set was spatially nor- 3.1. Group data, within run: 20 Hz (140 stimuli)
malized into Talairach space (Talairach and Tournoux,
1988) by deforming the image volume to match the Fig. 2a,b show the 3D orthogonal view maximum
standard SPM96 MR brain template with an affine intensity projection (MIP) of the results of the group
transformation; input and output voxel dimensions VNS-REST2 and TONE-REST2 cluster analyses,
were 4.224.228 mm and 444 mm, respectively. respectively, in the VNS 20 Hz (140 stimuli) experiment.
Each person’s data was then high-pass filtered to In the VNS-REST2 comparison (Fig. 2a), areas of sig-
remove slow signal drift, and spatially smoothed with nificant activation (voxel P < 0.001, extent P < 0.05) can
an 8 mm kernel size. High pass filtration of individual be seen bilaterally in the posterior part of the orbito-
data was done with the cutoff period of 232 s. frontal cortex (mainly in gyrus rectus), more in the right
Identification of voxels with statistically significant hemisphere, and also in both frontal poles (mainly
levels of activation: using SPM96 in MEDx 3.2, a pixel- frontal superior gyrus). In addition, activity can be seen
by-pixel non-paired bi-directional t-test was used to in the hypothalamus and the left globus pallidus. With
compare the individual fMRI time series data to a reduction of statistical threshold (pixel P < 0.05, extent
delayed boxcar model (VNS, REST1, TONE, REST2; P < 0.05) bilateral activation in the thalamus was
REPEAT X 10). In each cycle, acquisitions 1–5 were revealed. In the TONE-REST2 comparison (Fig. 2b)
taken as the VNS activation period and acquisitions 31– large areas of significant (pixel P < 0.001, extent P < 0.05)
39 were taken as the resting period (REST2) at the end activation can be seen bilaterally in the auditory cortex.
of the cycle. Given the BOLD time series data from our See also Table 2 for the exact coordinates.
first VNS study, we were concerned about the lingering
effects of VNS on blood flow immediately following 3.2. Group data, within run: 5 Hz (35 stimuli)
device termination. Since a preliminary analysis revealed
higher Z-values when the VNS was compared with For the VNS 5 Hz (35 stimuli; Fig. 2c), at the same
REST2 than with REST1 (immediately following the level of statistical significance as above, there were no
VNS), or when VNS was compared with both REST1 areas of significant activation of VNS-REST2. The
and REST2, both VNS and the TONE were compared same areas of activation were found during tone pre-
to REST2. sentation in the VNS 5 Hz (35 stimuli) run (Fig. 2d) as
Identification of Activated Areas in Group Data: was found in the VNS 20 Hz run, although the total
Nonthresholded individual Z-maps were averaged amount of activation was reduced (see comparison
according to Bosch (2000). Using the SPM statistics below, Table 3, 324.7 vs. 366.3 voxels and 745.3 vs.
functions in MEDx 3.2, a cluster analysis was also per- 1291.2 voxels).
formed on the group Z-map data to identify areas of the
brain showing significantly more or less BOLD-fMRI 3.3. Individual data, between run comparison
signal during the VNS and TONE conditions than dur-
ing the REST2 condition (Friston et al., 1994). A cluster We generated individual VNS 20 Hz Z-maps for each
was defined as a group of spatially connected voxels subject, on spatially normalized data. The numbers of
with individual Z-score > 3.09 (one-tailed P < 0.001) or statistically significant voxels are listed in Table 3, for
> 2.326 (P < 0.05), and a spatial extent threshold (Fris- both activations and deactivations. Individual response
ton et al., 1997), based on the number and Z-scores of was heterogeneous, with differences across individuals
voxels in it, greater than P < 0.05. These Z-maps were in the amount and location of activated areas. In five of
generated for the contrasts VNS-REST2 and TONE- six patients with exactly the same fMRI data acquisition
REST2 both within individuals and for the group and processing, the degree of activation (number of
within and over the two frequencies (5, 20 Hz). voxels) was higher under the influence of 20 Hz VNS
Quantitative analysis of overall brain reaction (both than 5 Hz VNS [Interestingly and perhaps of some
increases and decreases) to VNS and TONE was done importance, the only subject (Pt4) who demonstrated
based on the numbers of voxels with Z > 3.09 and the opposite trend, with a decrease of the number of
Z < 3.09 statistical level (P < 0.001). A nonparametric activation voxels (from 7 with 5 Hz to 1 with 20 Hz),
paired randomization test and a two by two Chi square was also the worst VNS treatment responder of the six
test of frequency of significant voxels in the volume of subjects. This subject also demonstrated the highest
the brain were used to compare overall activation across degree of relative deactivation in comparison in the
the different VNS and TONE conditions. group (Table 3). As a result this subject’s total reaction
 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227 223

Fig. 2. SPM 3D maximum intensity projection views of clusters of activated and deactivated voxels (group data). Activated areas at: (a) VNS-
REST2, 20 Hz experiment; (b). TONE-REST2 at 20 Hz VNS experiment (c) VNS-REST2, 5 Hz experiment; (d) TONE-REST2 at 5 Hz VNS
experiment. Deactivated areas: (e) VNS-REST2, 20 Hz experiment; (f) TONE-REST2, 20 Hz experiment; (g) TONE-REST2, 5 Hz VNS experiment.
P-values correspond for Z-value of individual voxel and cluster extent.
224 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227

Table 2
Brain areas affected by the Vagus Nerve Stimulation (VNS) and tone

Condition Activation clusters Deactivation clusters

Talairach cluster Size Brain structures Talairach cluster Size Brain

center, x, y, z (mm) (voxels) center, x, y, z (mm) (voxels) structures

VNS at 20 Hz 12.2; 56.8; 5.4 251 Right Frontal pole 5.5; 87.7; 30.1 126 Occipital pole
(P<0.001; P<0.05)
0.3; 5.9; 17.2 183 Posterior orbitofrontal cortex, 6.3; 35.1; 64.7 94 G. temporalis sup.,
extends to hypothalamus (0, 0, 0), medial surface
left Gl. Pallidus ( 16, 0, 4)
Tone at 20 Hz 55.0; 22.9; 1.1 569 Right G. temporalis sup. 4.1; 80.3; 4.5 702 Occipital pole
(P<0.001; P<0.05) and medius
52.4; 29.5; 7.6 431 Left G. temporalis sup,
Brodman 42
Tone at 5 Hz 56.0; 25.2; 6.8 475 Right G. temporalis sup., 25.0; 17.2; 24.1 76 Medial surface of the
(P<0.001; P<0.05) Brodman 42 left temporal lobe
55.7; 31.7; 8.6 474 Left G. temporalis sup., 25.0; 17.2; 24.0 76 Medial surface of the
Brodman 42 left temporal lobe
VNS, 20 Hz-5 Hz 9.9; 58.7; 6.3 177 Right Frontal Pole
(P<0.01; P<0.05)
Tone, 20 Hz-5 Hz 3.4; 25.0; 24.3 429 Pons
(P<0.01; P<0.05)

Table 3
Numbers of different pixels activated by condition

Intensity (mA) VNS 5 Hz VNS 20 Hz Tone at 5 Hz Tone at 20 Hz

>3.09 >3.09 >3.09 >3.09+ > 3.09/ >3.09 >3.09+ >3.09 >3.09+
< 3.09 < 3.09 <3.09 < 3.09 < 3.09

Pt1 1.25 0 0 31 198 5.39 169 169 37 2753

Pt2 1.00 0 0 5016 6995 0.39 0 7 118 118
Pt3 0.75 4 112 4421 10197 1.31 349 518 498 519
Pt4 0.75 7 7 1 117 116.00 983 983 533 933
Pt5 0.50 0 33 98 98 0.00 58 58 72 203
Pt6 0.50 0 0 4 4 0.00 389 463 3214 3221
Mean 0.79 1.8 25.3 1595.2 2934.8 0.84 324.7 366.3 745.3 1291.2
RTa 0.006c 0.016c 0.17d 0.047d
CHIb 0.00001c 0.00001c 0.00001d 0.00001d
a
RT P-value of randomization test comparing number of voxels (paired test).
b
Chi2, two by two Chi square test of frequency of significant voxels (P<0.001) in the volume of the brain.
c
0.006; 0.016; 0.00001; 0.00001, VNS 20 Hz VNS 5 Hz comparisons.
d
0.174; 0.0474; 0.00001;.00001, TONE 20 Hz TONE 5 Hz comparisons.

to VNS (number of activated and deactivated voxels) from randomization and CHI square tests respectively,
was higher at 20 Hz similar to the other patients]. Table 3). And again, only one of the patients (Pt4)
Tone presentation in the 20 Hz VNS run was also demonstrated the opposite trend.
accompanied with greater activation than during the 5
Hz run, but this trend reached statistical significance 3.4. Individual data, examination of response
only in the Chi square test. Activation was slightly heterogeneity and comparison with clinical variables
lateralized to the right hemisphere in the 20 Hz experi-
ment (569 vs. 431 voxels in acoustical clusters, Table 2) We performed scatter plots of these pixel activations
in comparison to the 5 Hz experiment (475 vs. 474 vox- on the 20Hz-REST2 comparison and ordinally ranked
els). Total reaction of the brain (activation and deacti- individual Z-maps with respect to the following clinical
vation) was significantly higher during the 20 Hz VNS variables: time from implant, VNS clinical response dur-
run than during the 5 Hz run (P=0.047 and 0.00001 ing the clinical trial, and Degree of Depression (Hamilton
 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227 225

Score) at the time of scanning. In these few subjects no Laere et al., 2000; Ring et al., 2000). The different time
consistent trend could be detected for these variables. domains over which the different imaging modalities
sample brain information likely accounts for much of the
inconsistencies across the studies. However, other likely
4. Discussion important differences include study sample and diagnosis,
time from implant, and concomitant medications.
The present study in depressed adults using the VNS/ (2) Although this is a small study, there appears to be
fMRI technique has three preliminary findings. First, this a frequency/dose effect of VNS on acute blood flow
study confirms the previous VNS/fMRI study (Bohning et changes. High frequency/dose VNS (20 Hz, at least 186
al., 2001) in terms of the immediate neuroanatomy affec- stimuli at 0.25 mA) increased BOLD-fMRI in numerous
ted by VNS as seen with fMRI, including the hypothala- regions. At the same level of statistical significance,
mus and orbitofrontal cortex, both implicated in mood there was no brain activation at 5 Hz. The 5 and 20 Hz
disorders. Second, this study is consistent with a fre- stimulation runs were identical and balanced by inten-
quency/dose effect of acute VNS on bilateral regional sity and duration (train length), but because of the
brain changes. Finally, there is an interesting suggestion block design, the higher frequency run (20 Hz) also
of a dose dependent modulating effect of VNS on other delivered more stimuli (140 stimuli) than did the lower
brain activity during the off time between acute VNS. frequency (5 Hz, at least 46 stimuli at 0.25 mA). Thus,
Before discussing these results, it is important to point differences in brain activity during the two runs may be
out some limitations of the study. These include the small due to frequency, or total stimuli, or both. Unfortunately,
sample size and the confound of intermingling frequency this design cannot distinguish these two parameters. Fur-
and dose. The subjects are a mixed cohort with treatment ther studies are needed in order to tease apart the brain
resistant depression who are taking multiple medications. effects of these two different parameters (frequency,
They have divergent clinical responses and differing times total number of stimuli).
from stimulation and variable VNS treatment settings. Because of the small amount of activation seen in he 5
One cannot readily generalize these findings to other Hz VNS data at a reasonable level of statistical sig-
diseases without caution. We were able to detect local nificance, we were not able to perform a formal analysis
statistically significant rCBF changes associated with of whether VNS at different frequencies or doses affects
the VNS despite the fact that the patients were clinically different brain regions. This is a key question in the
heterogeneous. The paired study design with 5 and 20 development of VNS as a potential therapy in other
Hz tested in two adjacent runs likely minimizes some of neuropsychiatric disorders. In an exploratory attempt at
the impact of this clinical heterogeneity. Nevertheless, addressing this question, we generated group Z-maps
these results should be viewed with caution until they for 5 Hz at P < 0.1 level, and overlaid them on the 20 Hz
are replicated. It is also conceivable that these changes maps at the levels noted above. Although there were dif-
are not directly neuronal. We are measuring a surrogate ferences in regional activation, we are not convinced they
of blood flow, and theoretically vagus effects on vascular are more than noise, given the need for the low sig-
flow could be a confound. This is unlikely however given nificance level at 5 Hz. Future studies are needed using
the anatomical distribution and the timing of the changes. different VNS settings that each produce robust responses
Turning to the results of the paper, (1) in general, the in order to address this very important question of whe-
VNS maps during the high frequency (20 Hz) run con- ther VNS at different frequencies alters different brain
firm our previous fMRI study and the known neuroa- regions.
natomy of vagus nerve stimulation in depression. In the (3) Finally, this study hints that the brain reacts dif-
present study, 20 Hz VNS increased BOLD-fMRI ferently in the off time between on times of VNS at dif-
response in the orbitofrontal cortex, frontal pole, hypo- ferent settings. That is, 20 Hz VNS (with more stimuli)
thalamus, left pallidum, and, less significantly, the tha- had effects on the brain’s response to hearing a tone
lamus. Our previous study included the 20 Hz data of (more activity overall, and more activity in the right
four of these subjects, so this should not be considered a auditory cortex), compared to during the 5 Hz stimula-
truly independent replication. It should also be noted tion. Specifically, although the VNS was not stimulat-
that the data analysis of our previous study differed in ing, and had not stimulated for almost 50 s, there were
that the individual data sets were intensity normalized differences in the brain’s response to the tone across the
and averaged time point by time point without temporal two runs. There was a trend for more activation elicited
smoothing to obtain a group data set. In general, how- by the tone presentation in the 20 Hz experiment. Sum-
ever, the regional activation found in the current study ming both activations and deactivations, there was a
is consistent with our previous fMRI findings (Bohning significant difference in reaction to the tone across the
et al., 2001) and brain areas reactive to VNS when two runs. The results in this study are thus the first
studied by [15O]H20 PET (Ko et al., 1996; Henry et al., brain imaging results to confirm behavioral data show-
1999) or perfusion SPECT (Vonck et al., 2000; Van ing lingering effects of VNS at different settings. For
226 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227

example, using just behavioral measures without ima- Clark KB, Naritoku DK, Smith DC, Browning RA, Jensen RA.
ging, Clark and colleagues (1999) showed that epilepsy Enhanced recognition memory following vagus nerve stimulation in
human subjects. Nature Neuroscience 1999;2:94–8.
patients performed better at an emotional memory task
Epstein CM. Transcranial magnetic stimulation: language function.
at certain VNS intensity settings than at others. In Journal of Clinical Neurophysiology 1998;15:325–32.
similar work, Ness and colleagues found that the thresh- Frackowiak R, Friston KJ, Chris D, et al. Human brain function. San
old to thermal pain decreased in patients with epilepsy Diego: Academic Press; 1997.
across a wide range of VNS intensities (Ness et al., 2000). Friston KJ, Worsley KJ, Frackowiak RJ, Mazziotta JC, Evans AC.
More behavioral and imaging studies, alone and in com- Assessing the significance of focal activations using their spatial
extent. Human Brain Mapping 1994;1:214–20.
bination, are needed to elucidate this modulatory effect Friston KJ. Cluster analysis. Human Brain Mapping 1997;5:133–6.
of VNS. A natural extension of our imaging results would Garnett ES, Nahmias C, Scheffel A, Firnau G, Upton AR. Regional
be to formally test whether auditory perception is beha- cerebral blood flow in man manipulated by direct vagus stimulation.
viorally better during VNS at 20 Hs than at 5 Hz. Pacing Clinical Electrophysiology 1992;15:1579–80.
Another important observation was the heterogeneity George MS, Sackeim HA, Marangell LB, Husain MM, Nahas Z,
Lisanby SH, et al. Vagus nerve stimulation. A potential therapy for
of individual response to VNS. Because of the large resistant depression? Psychiatric Clinic of North America 2000a;23:
amount of information acquired, BOLD fMRI allows 757–83.
for the formal statistical analysis of brain activity within George MS, Sackeim HA, Rush AJ, Marangell LB, Nahas Z, Husain
individuals. In this small sample single study, it is MM, et al. Vagus nerve stimulation: a new tool for brain research
impossible to determine fully the sources of this hetero- and therapy. Biological Psychiatry 2000b;47:287–95.
Handforth A, DeGiorgio CM, Schachter SC, Uthman BM, Naritoku
geneity. Potential causes of the heterogeneity range DK, Tecoma ES, et al. Vagus nerve stimulation therapy for partial-
from the inherent variability of the VNS/fMRI scanning onset seizures: a randomized active-control trial Neurology 1998;
system, to the differences in our subjects with respect to 51:48–55.
age, gender, intensity, clinical condition at the time of Henry TR, Bakey RE, Votaw JR, Pennell PB, Epstein CM, Faber TL, et
scanning, time from implant, baseline VNS settings before al. Brain blood flow alterations induced by therapeutic vagus nerve
stimulation in partial epilepsy: I. acute effects at high and low levels of
and after the MRI study, concomitant medications, and stimulation. Epilepsia 1998;39:983–90.
co-morbid diagnoses. With a small sample of 6, pre- Henry TR, Votaw JR, Pennell PB, Epstein CM, Bakay RA, Faber TL,
liminary ranking of individual VNS/fMRI responses et al. Acute blood changes and efficacy of vagus nerve stimulation in
with respect to these variables failed to reveal any trends. partial epilepsy. Neurology 1999;52:1166–73.
Kirschner A, Birklein F, Stefan H, Handwerker HO. Left vagus nerve
Future studies with larger sample sizes and covariate
stimulation suppresses experimentally induced pain. Neurology
analyses are needed to examine these important issues and 2000;55:1167–71.
their influence on the individual VNS responses. Ko D, Heck C, Grafton S, Apuzzo ML, Couldwell WT, Chen T, et al.
Future studies combining VNS and fMRI are needed Vagus nerve stimulation activates central nervous system structures
to extend these important new findings of different in epileptic patients during PET H215O blood flow imaging. Neu-
regional brain activity when the vagus nerve is stimulated rosurgery 1996;39:426–30.
Maniker A, Liu W-C, Marks D, Moser K, Kalnin A. Positioning of
at different frequencies. vagal nerve stimulators: technical note. Surgical Neurology 2000;53:
178–81.
Ness TJ, Fillingim RB, Randich A, Backensto EM, Faught E. Low
Acknowledgements intensity vagal nerve stimulation lowers human thermal pain
thresholds. Pain 2000;86:81–5.
Ring HA, White S, Costa DC, Pottinger R, Dick JP, Koeze T, et al. A
Funded in part by grants from the Dana Foundation SPECT study of the effect of vagal nerve stimulation on thalamic
and Cyberonics. The authors acknowledge helpful com- activity in patients with epilepsy. Seizure 2000;9:380–4.
ments from Burke Barrett of Cyberonics. The authors are Rush AJ, George MS, Sackeim HA, Marangell LB, Husain MM,
also grateful for administrative help with scanner access Giller C, et al. Vagus Nerve Stimulation (VNS) for treatment-resis-
tant depressions: a multicenter study. Biological psychiatry 2000;47:
from Ken Roozen, PhD and Sue Hallchurch of GE.
276–86.
Sackeim HA, Keilp JG, Rush AJ, George MS, Marangell LB, Dormer JS,
et al. The effects of vagus nerve stimulation on cognitive performance in
References patients with treatment-resistant depression. Neuropsychiatry Neuro-
psychology and Behavioral Neurology 2001;14:53–62.
Ben-Menachem E, Manon-Espaillat R, Ristanovic R, Wilder BJ, Ste- Sawaki L, Okita T, Fujiwara M, Mizuno K. Effects of subthreshold
fan H, Mirza W, et al. Vagus nerve stimulation for treatment of transcranial magnetic stimulation on choice reaction time and
partial seizures: 1. A controlled study of effect on seizures. First correlation with motor cortical activation. Kobe Journal of Medical
International Vagus Nerve Stimulation Study Group. Epilepsia Science 1999;45:165–79.
1994;35:16–26. Talairach J, Tournoux P. Co-planar stereotactic atlas of the human
Bohning DE, Lomarev MP, Denslow S, Nahas Z, Shastri A, George brain. New York: G. Thieme Stuttgart; 1988.
MS. Feasibility of Vagus Nerve Stimulation-Synchronized blood Van Laere K, Vonck K, Boon P, Brans B, Vandekerckhove T,
oxygen level-dependant functional MRI. Investigative Radiology Dierckx R. Vagus nerve stimulation in refractory epilepsy: SPECT
2001;36:470–9. activation study. Journal of Nuclear Medicine 2000;41:1145–54.
Bosch V. Statistical analysis of multi-subject fMRI data: assessment of Vonck K, Boon P, Van Laere K, D’Have M, Vandekerckhove T,
focal activations. Journal of Magnetic Resonance Imaging 2000;11:61–4. O’Connor S, et al. Acute single photon emission computed
 M. Lomarev et al. / Journal of Psychiatric Research 36 (2002) 219–227 227

tomographic study of vagus nerve stimulation in refractory epilepsy. intramodality validation. Journal of Computer Assisted Tomo-
Epilepsia 2000;41:601–9. graphy 1998;221:139–52.
Wassermann EM, Wedegaertner FR, Ziemann U, George MS, Chen R. Zabara J. Peripheral control of hypersynchronous discharge in epi-
Crossed reduction of human motor cortex excitability by 1-Hz trans- lepsy. Electroencephalography and Clinical Neurophysiology 1985;
cranial magnetic stimulation. Neuroscience Letter 1998;250:141–4. 61:162.
Woods RP, Grafton ST, Holmes CJ, Cherry SR, Mazziotta JC. Zabara J. Inhibition of experimental seizures in canines by repetitive
Automated image registration: I. General methods and intrasubject, vagus stimulation. Epilepsia 1992;33:1005–12.