Efficacy and Safety of Transcranial Magnetic
Stimulation in the Acute Treatment of Major
Depression: A Multisite Randomized Controlled Trial
John P. O’Reardon, H. Brent Solvason, Philip G. Janicak, Shirlene Sampson, Keith E. Isenberg,
Ziad Nahas, William M. McDonald, David Avery, Paul B. Fitzgerald, Colleen Loo, Mark A. Demitrack,
Mark S. George, and Harold A. Sackeim
Background: We tested whether transcranial magnetic stimulation (TMS) over the left dorsolateral prefrontal cortex (DLPFC) is effective
and safe in the acute treatment of major depression.
Methods: In a double-blind, multisite study, 301 medication-free patients with major depression who had not benefited from prior
treatment were randomized to active (n ? 155) or sham TMS (n ? 146) conditions. Sessions were conducted five times per week with TMS
Hamilton Depression Rating Scale (HAMD) and response and remission rates with the MADRS and HAMD.
severity between groups at baseline), as well as on the HAMD17 and HAMD24 scales at weeks 4 and 6. Response rates were significantly
events (4.5%) that were generally mild and limited to transient scalp discomfort or pain.
Conclusions: Transcranial magnetic stimulation was effective in treating major depression with minimal side effects reported. It offers
clinicians a novel alternative for the treatment of this disorder.
Key Words: Clinical trial, efficacy, major depression, safety, TMS
and Lopez 1996). Treatment is often challenging; an estimated
20%–40% of patients do not benefit sufficiently from existing
antidepressant interventions including trials of medication and
psychotherapy (Greden 2001). A substantial proportion of pa-
tients manifest a chronic, treatment-resistant course of illness,
resulting in a need for additional treatment options (Rush et al.
ajor depressive disorder (MDD) is a common, recurrent,
and frequently chronic disorder that is a leading con-
tributor to functional impairment and disability (Murray
2006; Trivedi et al. 2006). Transcranial magnetic stimulation
(TMS) has been proposed as one such alternative (George et al.
1997, 1995b; George and Wassermann 1994).
During TMS, a time-varying current is discharged in an
insulated coil placed on the scalp surface. This generates a brief
dynamic magnetic field that is orthogonal in orientation to
current flow in the coil (Amassian et al. 1992; Roth et al. 1991a,
1991b). The scalp and skull are transparent to the magnetic field,
which induces current flow when it reaches a conductive me-
dium such as neural tissue and with it the potential to modulate
neural circuitry in a therapeutic fashion.
Several single-center, controlled studies of TMS have been
conducted that, in most cases, have supported the hypothesis
that TMS manifests antidepressant properties when delivered to
the left or right dorsolateral prefrontal cortex (DLPFC; Avery et al.
2005; Fitzgerald et al. 2003; George et al. 1997, 2000; Klein et al.
1999; Loo et al. 1999, 2001; Pascual-Leone et al. 1996). This target
was initially selected (George et al. 1995b) based on imaging
findings implicating this region in the pathophysiology of de-
pression and in antidepressant effects, as well as studies linking
specific lesion locations to dysregulation of mood (Bench et al.
1992, 1995; Brody et al. 2001; George et al. 1995a, 1993a, 1993b;
Kimbrell et al. 2002; Mayberg 2002, 2003; Nobler et al. 2001;
Sackeim 2001a; Sackeim 2001b; Robinson et al. 1984, 1988).
Meta-analyses conducted of this literature have largely con-
curred that either a slow rate of stimulation (? 1 pulse/sec) over
the right DLPFC or fast stimulation (at 5–20 pulses/sec) over the
left DLPFC have greater antidepressant effects than matched
sham stimulation conditions (Burt et al. 2002; Couturier 2005;
Kozel and George 2002; Martin et al. 2002; McNamara et al.
2001). Such effects, however, have been questioned in terms of
clinical significance, with at least one meta-analysis, using con-
delphia; Department of Psychiatry (HBS), Stanford University, Palo Alto,
ter, Chicago, Illinois; Mayo Clinic College of Medicine (SS), Rochester,
Minnesota; Department of Psychiatry (KEI), Washington University
School of Medicine, St. Louis, Missouri; Department of Psychiatry and
Behavioral Sciences (WMM), Emory University, Atlanta Georgia; Depart-
ment of Psychiatry and Behavioral Sciences (DA), Seattle, Washington;
Centre (PBF), Melbourne, Australia; the Alfred and Monash University
Psychiatry (CL), University of NSW, Sydney, Australia; Department of
Psychiatry (ZN, MSG), Medical University of South Carolina, Charleston,
Address reprint requests to John P. O’Reardon, M.D., Laboratory for Trans-
cranial Magnetic Stimulation, University of Pennsylvania, 3535 Market
Street, Suite 4005, Philadelphia, PA 19014.
Received September 29, 2006; revised November 25, 2006; accepted Janu-
ary 19, 2007.
BIOL PSYCHIATRY 2007;62:1208–1216
© 2007 Society of Biological Psychiatry
servative criteria, finding no clear benefit from active TMS (Martin
et al. 2002). All meta-analyses have acknowledged the limitations
of prior work including insufficient intensity and duration of
treatment, inferior sham condition, single-center study design,
and inadequate sample size.
We conducted a large, multicenter, randomized controlled
trial of TMS to address key prior shortcomings. We administered
TMS at fixed treatment parameters over the left DLPFC and
compared it with a sham TMS intervention that mimicked the
active condition to a greater extent than has been the case in
previous research. The TMS parameters in the study were
selected to achieve the highest feasible dose consistent with
current safety guidelines (Wassermann 1998) and patient tolera-
bility. The duration of the TMS course (4–6 weeks) was also
longer than in previous studies, providing greater opportunity for
cumulative therapeutic and adverse effects to be expressed.
Methods and Materials
Eligible subjects were antidepressant medication-free outpa-
tients, aged 18–70, with a DSM-IV diagnosis of MDD, single
episode or recurrent, with a current episode duration of 3 years
or less. The episode had a Clinical Global Impressions Severity of
Illness (CGI-S) score of at least 4 and a total score of at least 20
on the 17-item Hamilton Depression Rating Scale (HAMD17).
Symptom stability was required during a 1-week no-treatment
lead-in period, with a HAMD17 total score of at least 18 and a
decrease in score of 25% or less from that observed at the
screening assessment. Prior antidepressant treatment during the
current episode was assessed using the Antidepressant Treat-
ment History Form (ATHF; Sackeim 2001b). Patients were re-
quired to have failed at least one but no more than four adequate
antidepressant treatments in this or the most recent episode.
Alternatively, patients were eligible if they had marked intoler-
ance to antidepressants as demonstrated by four failed attempts
to tolerate an adequate medication trial (lifetime).
Exclusionary criteria for study participation included a life-
time history of psychosis, bipolar disorder, or obsessive–com-
pulsive disorder; posttraumatic stress disorder and eating disor-
ders (if present in the past year); lack of response to an adequate
trial of electroconvulsive therapy (ECT); prior treatment with
TMS or a vagus nerve stimulator implant; pregnancy; a personal
or close family history of a seizure disorder; presence of neurologic
disorder or medication therapy known to alter seizure threshold; or
presence of ferromagnetic material in or in close proximity to the
head. Routine laboratory studies (complete blood count, chemistry,
thyroid stimulating hormone), urine toxicology screen, and electro-
cardiogram were performed at study screening, and subjects were
required to be medically stable before entry.
The study was conducted at 23 study sites in the United States
(20 sites), Australia (2 sites), and Canada (1 site), with active
enrollment extending from January 2004 through August 2005.
Institutional review board approval was obtained at all sites. The
study was conducted under an Investigational Device Exemption
from the U.S. Food and Drug Administration (FDA). All subjects
signed an informed consent document before undergoing any
The study had three phases: a lead-in phase (1 week, no
treatment), a 6-week acute treatment period (daily treatment with
TMS or sham), and a taper phase (3 weeks reduced frequency of
TMS or sham, start of antidepressant). Patients were randomized
1:1 to either active TMS or sham TMS. During the acute treatment
phase, TMS sessions were scheduled daily in a 5-day sequence,
for a maximum of 30 sessions (6 weeks), and typically adminis-
tered on a Monday through Friday schedule. Then TMS was
tapered in a blinded manner in six sessions across 3 weeks
during which all patients were titrated onto monotherapy with an
antidepressant medication. After 4 weeks of participation in the
acute phase, if patients failed to show meaningful clinical benefit
(i.e., ? 25% reduction in baseline symptoms on the HAMD17),
they could crossover to an open-label, acute treatment extension
study. The specific criterion for early study exit was concealed
from the investigators’ knowledge.
Study Device Description, TMS Session Procedures,
The TMS sessions were delivered using the Neuronetics
Model 2100 Therapy System investigational device (Neuronetics
Inc., Malvern, Pennsylvania). Three separate magnetic coils,
similar in weight, external appearance, and acoustic properties
when actively pulsed, were used at each site, with one coil
unblinded and used as the known active coil to determine motor
thresholds. The other two coils differed in that the sham coil had
an embedded magnetic shield. The latter limited the magnetic
energy reaching the cortex to 10% or less than the active coil but
nevertheless allowed the active and sham coils to have similar
appearance, placement, and acoustic properties. All treatment
personnel were blind as to coil assignment. All efficacy outcome
measures were assessed by blinded study personnel (raters) who
were not permitted access to the treatment sessions. Raters
underwent certification in which their study participation was
contingent on demonstrating adequate reliability in the conduct
and scoring of interviews to derive HAMD and MADRS scores.
Quality of ongoing ratings was assessed through the use of video
monitoring reviewed by an independent expert. Patients were
instructed not to disclose any details of the treatment session
with the study raters during rating sessions. Furthermore, neither
raters nor other study personnel at the specific centers were
aware of the primary efficacy measure during the trial. Ratings
were administered at baseline and at weeks 2, 4, and 6–9.
Treatment was fixed at 120% magnetic field intensity relative
to the patient’s observed resting motor threshold (MT), at a
repetition rate of 10 magnetic pulses/sec, with a stimulus train
duration (on time) of 4 sec and an intertrain interval (off time) of
26 sec. The left DLPFC was the treatment location and was
determined by movement of the TMS coil 5 cm anterior to the
motor threshold location along a left superior oblique plane with
a rotation point about the tip of the patient’s nose (George et al.
1995b). Spatial coordinates were recorded with a mechanical coil
positioning system to ensure placement reproducibility. The MT
estimation was repeated weekly by visual observation of thumb
or other finger movement (Pridmore et al. 1998) using the MT
Assist (Neuronetics Inc.). The latter is a standardized, software-
based mathematical algorithm that provides an iterated estimate
of the MT.
During the first week of the acute phase only, treatment
intensity could be adjusted to 110% of MT for tolerability but was
then required to return to 120% MT from week 2 onward. A
treatment session lasted for 37.5 min for a total of 3000 magnetic
pulses delivered per session.
J.P. O’Reardon et al.
BIOL PSYCHIATRY 2007;62:1208–1216 1209
All patients were free of antidepressants or other psychotropic
medications directed at treating depression. Patients were allowed
only limited use of either hypnotics or anxiolytics for treatment-
emergent insomnia or anxiety, respectively. Up to 14 daily doses
(lorazepam 2 mg/day equivalent) only were permitted (of either or
both types of medications) during the acute phase.
The primary efficacy outcome was the difference between active
and sham TMS using the last visit MADRS score through week 4 of
the acute phase. Secondary outcome measures were the MADRS
score at 6 weeks, 24-item and 17-item HAMD scores at 4 and 6
weeks, and categorical endpoints using MADRS, HAMD17, and
HAMD24 at 4 and 6 weeks. Response was defined as at least 50%
reduction from baseline score. Remission was defined by an abso-
lute scale-specific score, as indicated in Figure 2). Several standard-
ized HAMD factor scores were also derived: including the Depres-
sion Core factor (Items 1, 2, 3, 7, 8), the Maier subscale (Items 1, 2,
7, 8, 9, 10), the Gibbons subscale (Items 1, 2, 3, 7, 9, 10, 11, 14), the
Anxiety/Somatization factor (Items 10, 11, 12, 13, 15, 17), the
Retardation factor (Items 1, 7, 8, 14), and the Sleep factor (Items
4–6). Global clinical status was assessed using the observer rated
Clinical Global Impressions Severity of Illness Scale (CGI-S).
Patient-reported outcomes were obtained using the Inventory
of Depressive Symptoms—Self Report version (IDS-SR), and the
Patient Global Impressions Improvement Scale (PGI-I). Various
functional status and quality-of-life measures were collected and
will be reported elsewhere.
Safety was assessed at every treatment visit by recording
spontaneous adverse event reports that were coded using the
current version of the Medical Dictionary for Regulatory Activi-
ties. Additional safety evaluations included targeted assessment
of air-conduction auditory threshold at baseline, week 4, and
week 6. Cognitive function was assessed periodically and will be
the subject of a separate report.
Sample size was determined by requiring 90% power and a
two-tailed 5% level in detecting a difference between the active
and sham conditions with a putative MADRS effect size of
approximately .40, based on the standard t test method. Efficacy
analyses were performed on the strict intent-to-treat sample of all
evaluable patients, defined in the protocol as those with a
baseline and at least one postbaseline observation available for
analysis. The null hypothesis for the primary outcome was tested
with an analysis of covariance, using baseline score and ATHF
medication resistance level as fixed-effect covariates, adjusting
for site differences using a random effect (SAS Institute, Cary,
North Carolina). Secondary outcome analyses for continuous
measures were conducted in a similar fashion. All analyses were
conducted in a last-observation carried forward (LOCF) manner
through the indicated time points.
Of the 325 patients randomized to TMS or sham, 301
(92.6%) had at least one postbaseline assessment, and their
data comprised the a priori–specified analysis set. There were
no baseline clinical or demographic differences between the
24 nonevaluable patients and the 301 evaluable subjects.
Nonevaluable patients were evenly distributed between the
active (n ? 10) and the sham (n ? 14) conditions, and there
was no systematic difference between these groups in the
reasons for discontinuation. Reasons given for study discon-
tinuation through week 4, the primary efficacy time point, are
provided in Figure 1.
Through the primary efficacy time point of 4 weeks, the
overall discontinuation rate was low and similar in the active TMS
(7.7%) and sham TMS (8.2%) treatment groups. Discontinuation
because of adverse events of TMS was rare, similar across
treatment conditions (4.5% of active TMS vs. 3.4% of sham TMS
patients), and most commonly due to scalp discomfort. Beyond
the primary efficacy time point, 74 (47.7%) patients in the active
Figure 1. Reasons for study discontinuation through the primary efficacy time point (week 4).
1210 BIOL PSYCHIATRY 2007;62:1208–1216
J.P. O’Reardon et al.
TMS group and 92 (63.0%) patients in the sham TMS treatment
group elected to enter the open-label extension study.
Demographic and clinical features were similar in both
groups. The CGI-S severity score of 4.7 for the study cohort
corresponded to a depression severity between moderately and
markedly ill. A minority of the sample reported a current episode
duration greater than 2 years. The degree of treatment resistance
in the current episode did not differ between groups, with an
average of 1.6 failed adequate antidepressant treatments in the
current episode and about half the study population failing to
benefit from at least two treatments (by ATHF criteria). The
degree of functional impairment in the study cohort was sub-
stantial, with about half of the sample unemployed and one third
on disability because of their mood disorder.
As indicated in Table 1, baseline symptom severity at random-
ization differed between groups on the primary outcome mea-
sure (MADRS, p ? .036) but not on any of the other symptom
scales. This difference arose in the context of the protocol not
specifying a minimal symptom severity score on the MADRS
for study inclusion, differing in this respect from the HAMD17
(for which a minimum score of 20 was required at the screening
visit). Six participants had baseline MADRS scores in the mild
range (? 20, range 14–19) and were randomized unevenly
across the two groups (i.e., 4 to active, and 2 to sham). This
resulted in a small but significant difference in baseline symptom
scores (i.e., 1.1 point).
Continuous Efficacy Outcome Measures
Efficacy results for continuous outcomes with the MADRS,
HAMD24, and HAMD17 are shown in Figure 2.
At the primary efficacy time point, week 4, the baseline to
endpoint change on the MADRS showed a statistical trend
favoring active TMS in the a priori–specified evaluable study
population (p ? .057). Given the observed baseline imbalance
on the MADRS, a supplementary analysis was conducted on a
study sample that included only those patients (n ? 295) with a
minimum baseline score of 20 on the MADRS, excluding the six
patients noted earlier. In this analysis, the baseline to endpoint
change was statistically significant at week 4 (p ? .038). At the
secondary efficacy time point, week 6, the baseline to endpoint
change on the MADRS continued to show a statistical trend
favoring active TMS in the a priori–specified evaluable study
population (p ? .057) and in the subset of patients with a
baseline MADRS ? 20 (p ? .052).
Table 1. Demographic Features, Clinical Features, and Baseline and End Point (Weeks 4 and 6) Symptom Scores of
the Study Sample
(n ? 155)
(n ? 146)
N (%) female
Age (years ? SD)
Ethnic origin, n (%)
Employment status, n (%)
Receiving disability compensation
Recurrent illness course (%)
Duration of current episode in months, Mean (SD)
n (%) of population with current episode ? 2 years
Number of fully adequate antidepressant treatments in current episode
Baseline Symptom Scores
MADRS total score (SD)
HAMD17 total score (SD)
HAMD24 total score (SD)
IDS-SR total score (SD)
Week 4 Symptom Scores
MADRS total score (SD)
HAMD17 total score (SD)
HAMD24 total score (SD)
Week 6 Symptom Scores
MADRS total score (SD)
HAMD 17 total score (SD)
HAMD 24 total score (SD)
47.9 ? 11.0
48.7 ? 10.6
CGI-S, Clinical Global Impressions Severity of Illness; HAMD, Hamilton Depression Rating Scale (17 and 24 item);
IDS-SR, Depressive Symptoms—Self Report version; MADRS, Montgomery–Asberg Depression Rating Scale; TMS,
transcranial magnetic stimulation.
p values indicate the contrast of the within group changes from baseline.
J.P. O’Reardon et al.
BIOL PSYCHIATRY 2007;62:1208–1216 1211
HAMD17 and HAMD24
Results for the HAMD24 and HAMD17 are presented for the a
priori–specified evaluable data set only (n ? 301). No baseline
imbalance in symptom severity occurred with these scales, and
the findings were unaltered when the six patients with low
scores on the MADRS were excluded (data available from the
authors on request). At the primary efficacy time point, week 4,
the baseline to endpoint change on both the HAMD17 and the
HAMD24 yielded a significant main effect of treatment group
favoring active TMS (p ? .006 and p ? .012, respectively). This
outcome was sustained at the secondary efficacy time point,
week 6, with a significant advantage in favor of active TMS (p ?
.005 for HAMD17, p ? .015 for HAMD24). As seen in Figure
2A–C, it is notable that the time course of clinical effect showed
a sustained improvement across the acute treatment phase for
the active TMS group, whereas the sham TMS group showed a
pattern of early change that dissipates at the later time points.
Clinician-rated global illness severity showed greater im-
provement with active TMS compared with sham TMS as early as
week 2 of the acute treatment phase (p ? .047) and continued
through the primary efficacy time point of week 4 (p ? .009) and
the secondary efficacy time point of week 6 (p ? .012).
primary efficacy time point, the response rate as defined by the
three symptom-rating scales (MADRS, HAMD17, and HAMD24)
was higher with active compared with sham TMS. This effect was
sustained through the secondary efficacy time point at week 6.
Remission Rates. As seen in Figure 3A–C, at the week 4
primary efficacy time point, a significant difference in remission
rates was not detected. At the secondary efficacy time point of
week 6, however, remission rate was higher with active com-
pared with sham TMS on the MADRS and HAMD24.
HAMD Factor Scores. The pattern of change in the HAMD17
scale was examined further by the assessment of previously
established factor scores for the HAMD. These results are shown
in Table 2 and demonstrate significantly better outcomes for
active TMS compared with sham TMS on core depression
symptoms (Core Depression Factor, Maier Factor, and Gibbons
Factor), anxiety symptoms (Anxiety/Somatization Factor), and
vegetative symptoms (Retardation Factor) at weeks 4 and 6.
Patient-Reported Mood Symptoms and Global Improve-
ment. Patient-reported mood change and global improvement
were assessed using the IDS-SR and the PGI—Improvement
scales. The pattern of change followed the clinician-reported
measures but showed a less robust effect. The IDS-SR exhibited
a trend toward improved outcome in active compared with sham
TMS at the primary efficacy time point, week 4 (p ? .058), which
was maintained at the secondary efficacy time point, week 6 (p ?
.053). There was no statistically significant separation between
the groups on the PGI-I at either time point.
Antidepressant Effects During TMS Taper.
phase, response and remission rates of the active TMS group
improved incrementally. With the MADRS, the response rate
increased from 23.9% to 27.7% and the remission rate from 14.2%
to 20.6% for. In contrast, addition of medication to the sham TMS
group produced little meaningful clinical change during this
phase: a change from 12.3% to 13.7% for the response rate and
from 5.5% to 8.9% for remission on the MADRS.
As seen in Figure 3A–C, at the week 4
Spontaneous Adverse Events. There was a higher incidence
of scalp discomfort and pain with active than sham TMS (Table 3).
These events were generally reported as mild or moderate in
severity and diminished rapidly in incidence after the first
treatment week. Scalp discomfort had the potential to compro-
mise the study blind, and a separate analysis was conducted to
examine the relationship between clinical outcome and the
experience of cutaneous discomfort. The findings were negative
regarding an association between any of these adverse event
terms and the primary outcome measure (data not shown).
Figure 2. (A) Montgomery–Asberg Depression Rating Scale total score
change from baseline during the acute treatment phase. (B) Hamilton De-
pression Rating Scale (HAMD; 17 item) total score change from baseline
during the acute treatment phase. (C) HAMD (24-item) total score change
from baseline during the acute treatment phase.
1212 BIOL PSYCHIATRY 2007;62:1208–1216
J.P. O’Reardon et al.
Serious Adverse Events. There were no deaths in this study,
and no seizures were reported. During the acute treatment
phase, 16 serious adverse events were reported, 9 in the active
TMS group and 7 in the sham TMS group. Events reflecting
disease-related exacerbation were the most common serious
adverse events. These included suicidality (1.9% with sham vs.
.6% with active TMS), exacerbation of depression (1.9% with
sham vs. .6% with active), and a single suspected suicide gesture
(in the sham group). Overall risk of exacerbation of suicidality
was evaluated by determining the proportion of patients in either
group who increased in score on the suicide item of the HAMD
(item 3) from a value of 0 or 1 at baseline to a value of 3 or 4 at
any time point during the acute treatment phase. Cumulatively,
10 events meeting this criterion were observed in the sham TMS
group compared with 1 event in the active TMS group.
Audiometry. All subjects used earplugs during the treatment
sessions. No differences in air-conduction thresholds were de-
tected between or within treatment groups across the acute
treatment phase of the study (data not shown, to be presented in
a subsequent report).
This is the first large, multisite, randomized controlled trial of
daily left prefrontal TMS in medication-free patients with major
depression who had failed to receive adequate benefit from prior
antidepressant treatment. The findings indicate that TMS, admin-
istered at these parameters for a period of 4–6 weeks, is safe and
effective in the treatment of major depression.
The pattern of symptom improvement was consistent across
the various study outcome measures. Active treatment with TMS
was significantly superior to sham TMS treatment for the change
in mean symptom score using the HAMD17 and HAMD24 at
weeks 4 (p ? .006, p ? .012 respectively) and 6 (p ? .005, p ?
.015). After correction for the baseline score imbalance, the
MADRS also showed this pattern (p ? .038). Clinically important
change, as reflected in terms of the categoric outcomes of
response and remission, was also achieved in a substantial
portion of patients. At 6 weeks, the active TMS group was about
twice as likely to have achieved remission compared with the
sham TMS group (MADRS:14.2% vs. 5.2%, HAM-D17: 15.5% vs.
7.1%, HAMD24; 17.4% vs. 8.2%). Improvement on the self-report
IDS scale was at the level of a trend (p ? .058 week 4; p ? .053
week 6). Clinical outcomes were observed in the setting of a
favorable tolerability profile, with less than 5% of patients on
active TMS discontinuing treatment because of adverse effects by
the primary efficacy endpoint of 4 weeks.
The trajectory of improvement, as indicated in Figure 2,
implies that more than 2 weeks of TMS, compared with sham, is
required in this population before a significant improvement is
detected. Similarly, it appears that an additional 2 weeks of TMS
beyond the initial 4 weeks, as indicated in Figure 3, can have an
important clinical impact. The remission rates doubled during
that period of time.
come assessments during the acute treatment phase. (B) Hamilton Depres-
ments during the acute treatment phase.
Table 2. HAMD Factor Scores—Contrasts Between Active TMS and Sham
TMS during the Acute Treatment Phase
HAMD Rating Scale Factor Score Week 2Week 4Week 6
Core depression factor
HAMD, Hamilton Depression Rating Scale; TMS, transcranial magnetic
stimulation. All contrasts, except for the sleep factor, reflect a superior
treatment condition. p values shown for analysis of covariance model; post
hoc contrast between active and sham TMS at the time points indicated.
J.P. O’Reardon et al.
BIOL PSYCHIATRY 2007;62:1208–1216 1213
In this study, TMS was well tolerated and safe. The dropout
rate for any reason from active TMS was 7.7% at 4 weeks, and
discontinuation specifically because of side effects was 4.5%.
This is lower than the discontinuation rates generally reported
from clinical trials of marketed antidepressants. Adverse events
reported were principally limited to scalp discomfort or pain
within the confines of the TMS session itself and were mostly
transient phenomena in the first weeks of the TMS course. In fact,
the incidence of headache did not differ between active and
sham TMS conditions.
This study also suggests that the upper limit on the safe
administration of TMS may be somewhat greater than we cur-
rently suspect. Despite TMS being administered here at 120% of
motor threshold and 3000 pulses/session, an elevated rate of
serious adverse events relative to sham was not detected. The
most serious side effect that has been reported with TMS is a
seizure (Anderson et al. 2006), and none were observed in this
study. No adverse effects on mood in terms of either treatment-
emergent suicidality or treatment-emergent mania or hypomania
were observed. In fact, the frequency of treatment-emergent
suicidal ideation was numerically lower in the active TMS group
(1 event observed with active vs. 10 events observed with sham
during the acute treatment phase).
How can the clinical significance of these results be placed in
context for the practitioner? In answering this question, it is impor-
tant to consider the treatment resistance of the population of
patients studied. On average, the patients included had an average
of 1.6 failed adequate trials in the current episode of major depres-
sion, and nearly half had failed to benefit from 2 or more treatments.
The importance of prior resistance to antidepressant treatment is to
diminish significantly the likelihood of responding to subsequent
interventions. This pattern has been demonstrated with respect to
outcomes with electroconvulsive therapy (Prudic et al. 1990, 1996;
Sackeim et al. 1990), pharmacotherapy (Trivedi et al. 2006), and
vagus nerve stimulation (Sackeim et al. 2001).
The most comprehensive demonstration of this in the recent
literature is from the large, open-label, seminaturalistic STAR*D
trial (Trivedi et al. 2006). In that program, there was a progressive
reduction in remission rates with each stage of resistance. In fact,
patients who had failed two prior antidepressant treatments
experienced a mean remission rate (defined as a HAMD17 score
? 8) in level 3 treatment of only 16.2% after a 10- to12-week
course of treatment (Fava et al. 2006). In the controlled trial
results reported here, at the end of 6 weeks of treatment with
TMS alone, patients experienced a remission rate of 15.5%, and
this increased to 22.6%, by the same HAMD17 criteria, after 9
weeks during the taper phase of the study. Given that controlled
trials generally report somewhat lower clinical response and
remission rates than are seen in open-label experience, the
results reported here compare favorably to those seen in similarly
treatment-resistant patients in the STAR*D reports.
Controlled trials, where an active antidepressant medication is
measured against its relevant within-study control, provide the
most appropriate comparison for the TMS results. In the large
analysis of the FDA database of approved antidepressant medi-
cations, Khan and colleagues (2000) noted that the mean per-
centage reduction from baseline in total HAMD17 score across
the entire data set of antidepressants was 40.7% for active
treatment and 30.9% for placebo. On average, this represents an
overall relative advantage of about 10% in the reduction of total
score from baseline on the HAMD17 when comparing active
antidepressant to placebo treatment. By comparison, in this
study, at 4 weeks, active TMS treatment resulted in a reduction in
HAMD17 score of approximately 23% compared with 15% in the
sham group. This represents an overall relative advantage of
about 8% for active TMS compared with sham in reduction of the
total HAMD17 score. Similarly, application of the metric of
number needed to treat (NNT; Kraemer and Kupfer 2006) as an
indicator of effect size, yielded an NNT for TMS of 11 at week 4
and 9 at week 6. Thase et al. (2005), recently reported a pooled
estimate of response rates in a large sample (N ? 1795) of
patients treated with bupropion, various selective serotonin
reuptake inhibitors, or placebo and noted an overall response
rate in active-treated patients of 62.8% compared with 50.8% in
placebo treated patients, yielding an NNT in that sample of eight.
Thus, by these metrics, the antidepressant efficacy of TMS is
comparable to that of standard pharmacotherapy.
This study has several limitations that should be considered
when interpreting the results. Although a rigorous method was
used to assess prior treatment resistance, namely, the ATHF, this
method relies on retrospective report. Therefore, greater cer-
tainty regarding the level of treatment resistance would have
been obtained from a prospective, open-label antidepressant
lead-in phase and subsequently enrolling only nonresponders to
that intervention. This prospective method, however, requires
considerably more time and cost to implement and is therefore
not commonly used. The data documented in this trial support a
moderate level of treatment resistance for the study population.
Moreover, the observed response and remission rates with the
sham TMS intervention were very low, providing additional
evidence of the treatment resistance of this patient population.
The study employed an innovative approach to sham meth-
odology that represents a clear advance over prior work. Special
efforts were made to match the active and sham conditions in
procedure, sound, and sensation as closely as possible, while
substantially limiting the exposure of the cortex to the actual
magnetic field, as discussed in the study methods. Formal query
of patients and treaters to assess the adequacy of the blind,
however, was not conducted. Results from prior studies indicate
that previously TMS-naïve patients primarily base judgment of
whether they received an active or sham TMS procedure on
clinical outcome (Fitzgerald et al. 2003). A secondary analysis
(results to be presented in a subsequent report) also indicated
that scalp discomfort with active TMS did not correlate with
Table 3. Adverse Events Occurring in the Active Treatment Group at a
Rate of 5% or More and at Least Twice the Rate for Sham (with ME-Coded
Preferred Terms Shown)
(n ? 165)
(n ? 158)
Gastrointestinal Disorders Toothache
General Disorders and Site Administration
Application site discomfort
Application site pain
Musculoskeletal and connective tissue disorders
Skin and subcutaneous tissue disorders
Pain of skin
34 (20.6)5 (3.2)
14 (8.5) 1 (.6)
1214 BIOL PSYCHIATRY 2007;62:1208–1216
J.P. O’Reardon et al.
treatment outcome. Thus, unblinding of the active condition is an
unlikely explanation for the therapeutic advantage of active TMS.
The primary efficacy endpoint in this study was at week 4 of
the 6-week acute treatment phase. At or after that time point, if
clinically justified, patients were eligible to enter the open-label
extension study. Complete randomization was thus only truly
present through the primary efficacy time point. Therefore,
outcomes at time points after week 4 should be interpreted in
this context. On the other hand, because the treatment assign-
ment blind was maintained throughout the study, it is worth
noting that active TMS subjects continuing through the end of the
acute phase and into the taper phase of the study showed
persistent and perhaps accumulating benefit in comparison to
their continuing counterparts in the sham TMS group. Finally, the
method of coil positioning used to identify the left DLPFC was
not based on guidance by means of an magnetic resonance
imaging–assisted neuronavigational method. Rather, we used a
probabilistic surface anatomy approach targeting 5 cm anterior to
the motor threshold location, like most clinical trials of TMS. This
approach may not fully account for individual differences in
brain anatomy (Herwig et al. 2001, 2003).
In conclusion, TMS administered over the left DLPFC using
the parameters reported here for a period of up to 6 weeks was
effective in treating major depression and with a good tolerability
profile. These results indicate that TMS offers clinicians a novel
alternative in the treatment of this disorder (Sackeim 2001a).
The authors thank the patients who participated in this
clinical trial. In addition, we thank the Steering Committee for
this trial (Mark George, MUSC; Harold Sackeim, Columbia
University; Alan Schatzberg, Stanford), the principal investiga-
tors who were members of the Neuronetics TMS Study Group, and
their staff; Scott Aaronson, Sheppard Pratt Health System; David
Avery, University of Washington Medical Center; Randolph Can-
terbury, University of Virginia; Z. Jeffrey Daskalakis, Centre for
Addiction and Mental Health; James Ferguson, Radiant Re-
search; Paul Fitzgerald, The Alfred Psychiatry Research Centre;
William Gilmer, Northwestern University; Rosben Gutierrez, Psy-
Care, Inc.; Mustafa Husain, University of Texas—Southwestern
Medical Center; Keith Isenberg, Washington University School of
Medicine; Philip G. Janicak, Rush University Medical Center;
Andrew Krystal, Duke University Medical Center; Sarah H.
Lisanby, New York State Psychiatric Institute; Colleen Loo, Uni-
versity of New South Wales; Daniel Maixner, University of
Michigan Medical Center; Lauren Marangell, Baylor College of
Medicine; William McDonald, Emory University; Ziad Nahas,
Medical University of South Carolina; John P. O’Reardon, Uni-
versity of Pennsylvania; Elliott Richelson, Peter Rosenquist, Wake
Forest University Health Sciences; Shirlene Sampson, Mayo Clinic;
Brent Solvason, Stanford University.
Dr. Avery has received grant support from the National
Institute of Mental Health (NIMH) and Neuronetics; has acted as
a consultant to Bristol Myers Squibb (BMS), Cyberonics, Glaxo-
SmithKline (GSK), Eli Lilly, Janssen, Pharmaceutica Products,
Neuronetics, Performance Plus, and Takeda; and has been a
member of speakers bureaus for BMS, Cyberonics, GSK, Eli Lilly,
Janssen, and Pfizer. Dr. Demitrack is an employee of Neuronet-
ics, the study sponsor, and owns equity in Lilly, Wyeth, and
Neuronetics. Dr. Fitzgerald has received research support from
Neuronetics. Dr. George has received grant support from Abbott,
Cephos, Cortex, Cyberonics, Dantec, DarPharma, DARPA, GSK,
Jazz, National Institutes of Health (NIH), NARSAD, Neotonus,
NeuroPace, Neuronetics, and the Stanley Foundation; has acted
as consultant to Abbott, Aspect Biomedical, Aventis, Jazz Phar-
maceuticals, Argolyn Pharmaceuticals, Neuralieve, Neuronetics,
and NeuroPace; and has been a member of speakers bureaus for
Janssen, Lilly, GSK, Parke Davis, Picker Int, Cyberonics, and
Mediphysics/Amersham. Dr. Isenberg has received grant support
from NIMH and Neuronetics and has acted as a consultant to
Wellpoint and Barnes Jewish-Christian Behavioral Health. Dr.
Janicak has received grant support from Astra Zeneca, BMS,
Janssen, Neuronetics, and Solvay; has acted as consultant to
Astra Zeneca, BMS, Janssen, Neuronetics, and Shire; and is a
member of speaker bureaus for Abbott, Astra Zeneca, BMS,
Janssen, Pfizer, and Shire. Dr. Loo has received research support
from National Health and Medical Research Council (Australia),
Neuronetics, NSW Schizophrenia Fellowship, Pfizer Neuroscience
Research Grant scheme, and Rebecca Cooper Medical Research
Foundation. Dr. McDonald has received grant support from the
American Foundation for Suicide Prevention, Fuqua Founda-
tion, National Institute of Mental Health, National Institute of
Neurological Disease and Stroke, National Alliance for Research
in Schizophrenia and Depression, Neuronetics, and Janssen; has
acted as consultant to Neuronetics, Janssen, Forest, BMS; has
been a member of speakers bureaus for BMS, Forest, Janssen, and
Solvay; and has equity in Amgen, Teva, Pfizer, and Abbott. Dr.
Nahas has received grant support from Cyberonics, Eli Lilly,
Integra, Medtronic, Neuronetics, NeuroPace, and NIMH; has
acted as a consultant to Avanir Pharmaceutical, Aventis Phar-
maceutical, Cyberonics, Neuronetics, and NeuroPace; and has
been a member of the speakers bureau for Cyberonics. Dr.
O’Reardon has received grant support from BMS, Cyberonics,
Lilly, Magstim, Neuronetics, Pfizer, and Sanofi; acted as consul-
tant for Lilly and Neuronetics; and is a member of speakers
bureaus for BMS, Cyberonics, and Lilly. Dr. Sackeim has received
research support from the Cyberonics, MECTA, National Alliance
for Research in Schizophrenia and Depression, and NIMH and
has acted as a consultant to Cyberonics, Eli Lilly, Magstim,
MECTA, Neuronetics, NeuroPace, Novartis, and Pfizer. Dr.
Sampson has received research support from Neuronetics and
Pfizer and has acted as a consultant to NeuroPace. Dr. Solvason
has received grant support from Astra Zeneca, BMS, Forest, GSK,
and Neuronetics and has served as a consultant to Cephalon,
Cyberonics, Forest, Lilly, NeuroPace, and Sepracor.
Clinical trial is posted on www.clinicaltrials.gov. Listing No.
NCT 00104611. Supported by a grant from Neuronetics Inc.
Amassian VE, Eberle L, Maccabee PJ, Cracco RQ (1992): Modelling magnetic
coil excitation of human cerebral cortex with a peripheral nerve im-
mersed in a brain-shaped volume conductor: The significance of fiber
bending in excitation. Electroencephalo Clin Neuro 85:291–301.
MS (2006): Tolerability and safety of high daily doses of repetitive trans-
cranial magnetic stimulation in healthy young men. J ECT 22:49–53.
Avery DH, Holtzheimer PE 3rd, Fawaz W, Russo J, Neumaier J, Dunner DL, et
al. (2005): A controlled study of repetitive transcranial magnetic stimu-
lation in medication-resistant depression. Biol Psychiatry, September 1
[Epub ahead of publication].
Bench CJ, Frackowiak RS, Dolan RJ (1995): Changes in regional cerebral
blood flow on recovery from depression. Psychol Med 25:247–261.
anatomy of melancholia—focal abnormalities of cerebral blood flow in
major depression. Psychol Med 22:607–615.
Brody AL, Saxena S, Stoessel P, Gillies LA, Fairbanks LA, Alborzian S, et al.
findings [see comment]. Arch Gen Psychiatry 58:631–640.
J.P. O’Reardon et al.
BIOL PSYCHIATRY 2007;62:1208–1216 1215
Burt T, Lisanby SH, Sackeim HA (2002): Neuropsychiatric applications of Download full-text
transcranial magnetic stimulation: A meta analysis. Int J Neuropsychop-
Couturier JL (2005): Efficacy of rapid-rate repetitive transcranial magnetic
stimulation in the treatment of depression: A systematic review and
meta-analysis [see comment]. J Psychiatry Neurosci 30:83–90.
Fava M, Rush AJ, Wisniewski SR, Nierenberg AA, Alpert JE, McGrath PJ, et al.
(2006): A comparison of mirtazapine and nortriptyline following two
consecutive failed medication treatments for depressed outpatients: A
(2003): Transcranial magnetic stimulation in the treatment of depres-
sion: A double-blind, placebo-controlled trial. Arch Gen Psychiatry 60:
George MS, Ketter TA, Parekh PI, Horwitz B, Herscovitch P, Post RM (1995a):
George MS, Ketter TA, Post RM (1993a): SPECT and PET imaging in mood
George MS, Ketter TA, Post RM (1993b): SPECT and PET imaging in mood
George MS, Nahas Z, Molloy M, Speer AM, Oliver NC, Li XB, et al. (2000): A
[see comment]. Biol Psychiatry 48:962–970.
George MS, Wassermann EM (1994): Rapid-rate transcranial magnetic stim-
ulation and ECT. Convuls Ther 10:251–254; discussion 255–258.
George MS, Wassermann EM, Kimbrell TA, Little JT, Williams WE, Danielson
itive transcranial magnetic stimulation in patients with depression: A
placebo-controlled crossover trial. Am J Psychiatry 154:1752–1756.
al. (1995b): Daily repetitive transcranial magnetic stimulation (rTMS)
improves mood in depression. Neuroreport 6:1853–1856.
Greden JF (2001): The burden of recurrent depression: Causes, conse-
quences, and future prospects. J Clin Psychiatry 22:5–9.
Herwig U, Satrapi P, Schonfeldt-Lecuona C (2003): Using the international
10-20 EEG system for positioning of transcranial magnetic stimulation.
Herwig U, Schonfeldt-Lecuona C, Wunderlich AP, von Tiesenhausen C,
Thielscher A, Walter H, Spitzer M (2001): The navigation of transcranial
magnetic stimulation. Psychiatry Res 108:123–131.
Khan A, Warner HA, Brown WA (2000): Symptom reduction and suicide risk
Kimbrell TA, Ketter TA, George MS, Little JT, Benson BE, Willis MW, et al.
(2002): Regional cerebral glucose utilization in patients with a range of
severities of unipolar depression. Biol Psychiatry 51:237–252.
Klein E, Kreinin I, Chistyakov A, Koren D, Mecz L, Marmur S, et al. (1999):
netic stimulation in major depression: A double-blind controlled study
Kozel A, George M (2002): Meta-analysis of left prefrontal repetitive trans-
to clinical research and practice. Biological Psychiatry 59:990–996.
Loo C, Mitchell P, Sachdev P, McDarmont B, Parker G, Gandevia S (1999):
Double-blind controlled investigation of transcranial magnetic stimula-
tion for the treatment of resistant major depression. Am J Psychiatry
Loo C, Sachdev P, Elsayed H, McDarmot B, Mitchell P, Wilkonson M, et al.
(2001): Effects of a 2- to 4-week course of repetitive transcranial mag-
netic stimulation (rTMS) on neuropsychologic functioning, electroen-
(2002): Transcranial magnetic stimulation for treating depression. Co-
Mayberg HS (2002): Modulating limbic-cortical circuits in depression: Tar-
gets of antidepressant treatments. Sem Clin Neuropsychiatry 7:255–268.
McNamara B, Ray JL, Arthurs OJ, Boniface S (2001): Transcranial magnetic
stimulation for depression and other psychiatric disorders [see com-
ment]. Psychol Med 31:1141–1146.
Murray C, Lopez A (1996): The Global Burden of Disease: A Comprehensive
tors in 1990 and Projected to 2020. Cambridge, MA: Harcourt University
metabolism after ECT. Am J Psychiatry 158:305–308.
nial magnetic stimulation of left dorsolateral prefrontal cortex in drug-
resistant depression [see comment]. Lancet 348:233–237.
Pridmore S, Fernandes Filho JA, Nahas Z, Liberatos C, George MS (1998):
neurophysiological method and a visualization of movement method.
Prudic J, Haskett RF, Mulsant B, Malone KM, Pettinati HM, Stephens S, et al.
response to ECT [see comment]. Am J Psychiatry 153:985–992.
Prudic J, Sackeim HA, Devanand DP (1990): Medication resistance and clini-
cal response to electroconvulsive therapy. Psychiatry Res 31:287–296.
and depression after brain injury: Causal factors. Am J Psychiatry 145:
Robinson RG, Kubos KL, Starr LB, Rao K, Price TR (1984): Mood disorders in
stroke patients: Importance of location of lesion. Brain 107:81–93.
Roth BJ, Cohen LG, Hallett M (1991a): the electric field induced during
magnetic stimulation. In Lecy WJ, Cracco RQ, Barker AT, Rothwell J,
editors. Magnetic Motor Stimulation: Basic Principles and Clinical Experi-
ence (EEG Suppl. 43). Amsterdam: Elsevier Science, 268–278.
the electric field induced in the cortex during magnetic stimulation.
al. (2006): Bupropion-SR, sertraline, or venlafaxine-XR after failure of
Sackeim HA (2001a): Functional brain circuits in major depression and re-
Sackeim HA (2001b): The definition and meaning of treatment-resistant
Sackeim HA, Prudic J, Devanand DP, Decina P, Kerr B, Malitz S (1990): The
relapse following response to electroconvulsive therapy in major de-
Sackeim HA, Rush AJ, George MS, Marangell LB, Husain MM, Nahas Z, et al.
(2001): Vagus nerve stimulation (VNS) for treatment-resistant depres-
Remission rates following antidepressant therapy with bupropion or
selective serotonin reuptake inhibitors: a meta-analysis of original data
from 7 randomized controlled trials. J Clin Psychiatry 66:974–981.
Trivedi MH, Rush AJ, Wisniewski SR, Nierenberg AA, Warden D, Ritz L, et al.
(2006): Evaluation of outcomes with citalopram for depression using
measurement-based care in STAR*D: Implications for clinical practice.
Wassermann EM (1998): Risk and safety of repetitive transcranial magnetic
stimulation: Report and suggested guidelines from the International
June 5–7, 1996. Electroencephalogr Clin Neurophysiol 108:1–16.
1216 BIOL PSYCHIATRY 2007;62:1208–1216
J.P. O’Reardon et al.