Just How Effective are Antidepressant Medications? Results of a Major New Study

Abstract

The STAR*D (Sequenced Treatment Alternatives to Relieve Depression) project was the largest and most extensive trial of antidepressants
ever conducted. The study used state of- the-art methods to treat real patients coming to a hospital or psychiatric clinic
for relief from depression. Because the first antidepressant is often ineffective for most patients, a sequence of drug treatments
targeting various brain neurochemicals was carefully planned. This article will review the complex treatments and the various
outcomes, including the frequency of relapse during twelve months of follow-up care.

Full text

ORIGINAL PAPER
Just How Effective are Antidepressant Medications?
Results of a Major New Study
John J. Boren ÆAllan M. Leventhal Æ
H. Edmund Pigott
Published online: 6 September 2008
Springer Science+Business Media, LLC 2008
Abstract The STAR*D (Sequenced Treatment Alterna-
tives to Relieve Depression) project was the largest and
most extensive trial of antidepressants ever conducted. The
study used state of- the-art methods to treat real patients
coming to a hospital or psychiatric clinic for relief from
depression. Because the first antidepressant is often inef-
fective for most patients, a sequence of drug treatments
targeting various brain neurochemicals was carefully
planned. This article will review the complex treatments
and the various outcomes, including the frequency of
relapse during twelve months of follow-up care.
Keywords Depression Treatment Antidepressants 
Relapse
This article will review the findings of the largest and most
expensive trial of antidepressant medications yet con-
ducted. The research began in the year 2000 when the
National Institute of Mental Health funded the grant for
$35 million. Over a six-year period 4,790 depressed out-
patients were screened for inclusion in a multisite,
multistep study designed to see how well antidepressants
worked under conditions that simulated psychiatric treat-
ment in the real world. In a typical clinical practice, when
one antidepressant drug does not relieve depression, the
patient is likely to be given a second drug. If that doesn’t
work, a third drug is prescribed. If the third doesn’t work, a
fourth is prescribed. The effectiveness of carefully-selected
sequences is exactly what the STAR*D (‘‘Sequenced
Treatment Alternatives to Relieve Depression’’) project
was designed to study. The study’s outcome should tell us
how effectively a sequence of antidepressants can treat
depression under conditions that simulate clinical practice.
In the introduction to the first paper in the STAR*D
series (Trivedi et al. 2006b, pp. 28–29), the authors dis-
cussed several important considerations that shaped the
design of the study. First, less than half of depressed
patients get relief from a single course of antidepressant
medication. As the authors summarized it, ‘‘…most
depressed patients do not achieve remission with any initial
treatment (ibid.).’’ This unsatisfactory level of effective-
ness, little known to the general public and certainly not
mentioned in TV ads, strongly suggests a need for
improved effectiveness. The STAR*D study was designed
to gather evidence on whether greater effectiveness could
be achieved from a second (or if necessary, a third or
fourth) course of medication by switching to another drug
or by combining a second drug with the first. In an early
planning paper (Rush et al. 2004, p. 121), the authors
wrote, ‘‘Scientific evidence to define the most specific,
effective next-step treatment options for treatment-resistant
MDD [major depressive disorder] should improve clinical
outcomes and may reduce the cost of care.’’ The sequence
of drugs was carefully planned, with consideration of each
drug’s putative neurochemical mechanism of action. For
example, if a drug that inhibits serotonin reuptake failed to
work, then another drug with another mechanism, such as
inhibition of norepinephrine reuptake, was an option—and
so on with other drugs with other putative mechanisms.
With each treatment step, the goal was to find a drug or
J. J. Boren (&)
2409 Shallowford Lane, Chapel Hill, NC 27517, USA
e-mail: chch3515@yahoo.com
A. M. Leventhal
Silver Spring, MD, USA
H. Edmund Pigott
Clarksville, MD, USA
123
J Contemp Psychother (2009) 39:93–100
DOI 10.1007/s10879-008-9095-z

combination of drugs that would be effective with the
treatment-resistant patients.
A second important consideration: remission rather than
improvement was selected as the goal of treatment. In their
review of the outcome literature, the authors found that
patients who showed improvement (a ‘‘response’’ defined
as a 50% reduction in depressive symptoms), were more
likely to continue with disabling symptoms, to have poor
work productivity, to risk more alcohol/drug abuse, and to
relapse back into depression. Remission, on the other hand,
was associated with more normal psychosocial function,
better work performance, reduced risk of substance abuse,
and lower rates of relapse (Trivedi et al. 2006b, p. 23). The
goal of treatment, then, was to be remission, not
improvement.
A third important consideration in designing the study
was generalizability to the clinical practice of psychiatry.
Prior efficacy studies had typically enrolled mildly/moder-
ately depressed volunteers, often recruited by newspaper
ads, and excluded subjects with other coexisting psychiatric
diagnoses, general medical problems, chronic depression,
or substance abuse. These trials, frequently funded by drug
companies, were usually short in duration, had limited
range of dosages, and rarely involved follow-up to check on
relapse. In contrast, the STAR*D study focused on an
effectiveness goal so that the results would be applicable to
clinical practice. Real patients already seeking treatment for
depression at hospitals or psychiatric clinics were recruited
into the study. Exclusion criteria were kept minimal.
Medication was continued for 12–14 weeks with frequent
adjustment of dosage, and a 12-month follow-up was
scheduled to provide data on relapse.
Participants
The participants in the STAR*D trial were recruited from
outpatients who presented themselves for treatment at 18
primary care hospitals and 23 psychiatric specialty clinics
throughout the United States. To be eligible for the study,
patients had to be diagnosed with major depressive disor-
der. Recruitment was probably not a hard sell since the
costs of drugs, psychiatric fees, and other medical fees
were paid by the grant. Because a sequence of four treat-
ments was planned, with attrition expected after each
treatment, a very large number of patients, 4,790, were
screened for possible inclusion. Of this number 4,041 were
found eligible, and 2,876 moderately-to-severely depressed
patients were eventually selected for treatment. Although
2,876 constitutes the largest sample size ever enrolled in a
treatment study of depression, we still need to know what
population this sample represented. One selective factor
was that the sample included only people who sought
treatment at hospitals or psychiatric clinics. Most (62%)
came from psychiatric clinics. Depressed people who
called up a psychologist or social worker to seek psycho-
therapy had no opportunity to be represented in the sample.
Furthermore, the first treatment for every enrollee in the
study was a three-month course of antidepressant medica-
tion, so patients whose first choice of treatment was
psychotherapy may well have opted out. There is no
information on the magnitude of these selective factors.
We do know that, of the 4,790 screened, 613 refused
treatment, and 234 failed to return after the first interview.
The exclusion criteria were kept minimal in an effort to
study a sample of depressed patients likely to be treated in
primary care or psychiatric specialty settings. Patients with
chronic depression, other coexisting mental disorders, and
general medical disorders were accepted for the trial. At
study entry 65% had one or more comorbid disorders, with
the three most frequent being social phobia, generalized
anxiety, and PTSD disorders. Even substance abusers were
enrolled if they did not require detoxification. However,
patients with diagnoses of psychotic, bipolar, obsessive-
compulsive, or bulimic disorders were excluded, as were
women who were pregnant or breast feeding.
The study of the first treatment analyzed the results for
patients whose depression ranged from moderate to severe.
The 17-item Hamilton Rating Scale for Depression, often
considered the gold standard in trials of antidepressant
medication, was preselected as the primary tool to assess
depression. Patients whose depression was considered
mild, defined as a Hamilton score of 13 or less, were
excluded. The average Hamilton score of enrollees was
21.8.
A later paper in the STAR*D series summarizing fol-
low-up results (Rush et al. 2006a) revealed that 795 mildly
depressed patients had also been treated. The data for these
patients were not used in the analysis of the first treatment
(Trivedi et al. 2006b). However, their data was included in
the follow-up results.
Other characteristics of the selected sample: 24% were
minorities, 64% were female, 58% were divorced, never-
married, or widowed, 38% were unemployed, and 35% had
no medical insurance.
Level 1 Treatment
The first article of the outcome of the STAR*D series
(Trivedi et al. 2006b) covered the treatment methods and
the results of Level 1. All patients began with a 12–
14 week course of Celexa (generic name: citalopram), a
selective serotonin reuptake inhibitor (SSRI). Celexa was
chosen as a representative SSRI with the advantages over
other SSRI’s of minimal withdrawal symptoms, once-a-day
94 J Contemp Psychother (2009) 39:93–100
123

dosing, and less interaction with other drugs. The protocol
was designed to reach a fully adequate dose of Celexa for
each patient to increase the likelihood of achieving
symptom remission.
Visits with psychiatrists and other physicians were rec-
ommended at 2, 4, 6, 9, 12, and 14 weeks when the dosage
could be adjusted and other medical problems could be
treated. A treatment manual, clinical tests at each visit, and
guidance by a clinical coordinator aimed at state-of-the-art
care. Furthermore, as part of the standard protocol, physi-
cians treated general medical conditions, including sexual
dysfunction (a possible drug side effect or a symptom of
depression), and certain symptoms often accompanying
depression (sleeplessness, anxiety, and agitation).
This quality and extent of care in all probability
exceeded what most depressed patients receive in the real
world of treatment. Furthermore, the grant paid the psy-
chiatric fees, the medical fees, and the drug costs, thus
providing an extraordinary incentive for the patients not to
drop out. The high quality, cost-free, state-of-the-art care,
while beneficial to the patients, may have compromised the
original STAR*D goal of getting results that would gen-
eralize to clinical practice. The conditions were near
optimal for the medication treatment to be effective.
Acute Results of Level 1 Treatment with Celexa
The critical outcome measure, for reasons described above,
was remission, defined as a score of 7 or less on the
Hamilton scale. Note that this definition of remission does
not necessarily mean that a patient is free of all depressive
symptoms. For example, for the test item on ‘‘Work and
Interests’’ the lowest clinical rating was ‘‘Mild. Feels
incapable, listless, less efficient,’’ or for the item on
‘‘Suicide’’ the lowest rating was ‘‘Mild. Feels life is empty,
not worth living.’’ A patient would receive either or both of
these ratings and still have a score less than 7. Neverthe-
less, a Hamilton score of 7 is much lower than the average
intake score of 21.8 and represents a substantial reduction
in symptoms. Patients were designated as not achieving
remission when their exit Hamilton scores were missing.
Independent assessors, uninformed of the specific treat-
ment, administered the Hamilton scale by telephone in a
meritorious effort to reduce investigator bias.
The primary outcome of Level 1 treatment was that, of
the 2,876 depressed patients treated with Celexa for 12–
14 weeks, 790 or 27.5% reached the definition of remission
(Trivedi et al. 2006b, p. 33). This means that 2,086 patients
or 72.5% did not. Five hundred and ninety-two patients
(almost 21%) dropped out. Considering the near optimal
conditions of treatment, 27.5% is a strikingly modest rate
of remission.
Even this modest success rate may be due to several
factors, not just the antidepressive medication. Three other
factors may well have contributed to the observed remis-
sions. One is the cost-free, exceptional amount of
individualized psychiatric and medical care given at Weeks
2, 4, 6, 9, and 12. Physicians talked with the patients,
evaluated depressive symptoms and drug side effects,
adjusted the dosage of medications, and prescribed drugs
for most medical problems including anxiety, agitation,
sleep disturbances, and sexual dysfunction. Could this
exceptional degree of care have made a difference? A
Washington Post reporter (Vedatam 2006), based on
interviews of lead researchers of the study, wrote, ‘‘If
patients in this study had received the kind of care that
patients receive on average, the researchers said, the
remission rate probably would have been significantly
lower—perhaps even in the single digits.’’
A second factor is spontaneous recovery. Many
depressed people recover on their own, perhaps with the
aid of family, friends, exercise, a book on depression, or a
deliberate return to normal activities.
The third factor that probably contributed to remissions
is the placebo effect. A certain number of depressed people
will get better if they think they are getting an effective
drug, even when the pill is made of sugar. A placebo
control group, had it been included in the STAR*D study,
might well have benefited from the placebo effect plus the
exceptional medical and psychiatric care plus a 12–
14 week opportunity for spontaneous recovery.
The STAR*D authors recognized the difficulty in causal
interpretation near the end of the Level 1 paper (Trivedi
et al. 2006b, p. 37) when they stated, ‘‘Nonspecific treat-
ment effects undoubtedly accounted for some unknown
proportion of the acute response or remission rates.’’ How
much remission can be expected in a placebo control
group? Kirsch et al. (2002) performed a meta analysis of 47
placebo-controlled trials of six widely prescribed antide-
pressants. These trials, funded by pharmaceutical
manufacturers, were reported to the FDA to gain approval
of their drugs. Although the response to the antidepressants
was positive, Kirsch’s analysis showed that the response to
placebo was almost as great. The average difference was
only 2 points on the Hamilton scale. Kirsch concluded that
about 82% of the drug response could be accounted for by
the placebo effect. More than half of the 47 trials failed to
find any significant differences between placebo and drug.
In view of these findings, it is very hard to tell how many, if
any, of the remissions in Level 1 can be attributed to the
medication.
A recent article strongly confirmed the main thrust of
the Kirsch et al. findings. Turner et al. (2008) reviewed 74
trials of 12 antidepressant drugs (all eventually marketed)
that drug companies submitted to the FDA to win approval.
J Contemp Psychother (2009) 39:93–100 95
123

Every trial had a randomized, double-blind, placebo-con-
trolled design. According to the FDA reviews, only 38 of
the 74 trials (51%) showed positive effects of the drug to be
greater than the placebo. Turner et al. also found a pro-
found bias in which trials were eventually published.
Negative results were published very rarely or were inter-
preted to suggest positive effects. The published literature
appeared to show that 94% of the trials conducted were
positive, suggesting far greater efficacy of antidepressant
medications than was actually found.
A secondary goal of the Level 1 study was to determine
the characteristics of depressed people who did well on
antidepressant treatment and those who did poorly. Ques-
tionnaires and interviews revealed the usual suspects.
Worse outcomes were found among people who were
unemployed, less educated, non-Caucasian, unmarried,
male, living alone, and engaged in substance abuse. Better
outcomes were found in patients with less severe depres-
sion, more education, higher income, fewer concurrent
psychiatric and medical disorders, greater life satisfaction,
and a shorter current episode of depression. These are also
the characteristics of patients who are more likely to
achieve spontaneous remission.
Level 2 Treatment
The STAR*D authors had expected that more than 50% of
the patients (although perhaps not 72.5%) would be treat-
ment failures after a single course of medication. Indeed,
the main point of the research project was to evaluate
alternative treatments after the first treatment had failed.
The hope was to find the right medication that would
correct the presumed neurochemical defect causing the
patients’ depression. Only 1,439 patients were available to
continue with the second level of treatment. From the
original 4,041 screened as eligible, 1,475 had been moved
to follow-up (some after remission and others, who
declined further treatment, after ‘‘improvement’’) and
1,127 had exited the study. The many exits occurred in
spite of free treatment and free drugs.
Mirroring psychiatric practice when one drug fails to
work, the patients were offered two basic strategies for
continuing treatment: either switch to another treatment or
augment (add to) the first treatment. The patients in con-
sultation with their psychiatrist could choose which
strategy to follow or could even choose to be assigned to a
specific treatment. Full information about all options was
available, and informed consent procedures were followed.
The switch strategy offered four treatments: three medi-
cations or cognitive therapy. The augmentation strategy
offered two medications or cognitive therapy in combina-
tion with Celexa. Based on their informed consent, the
patients were assigned to one of the seven treatments,
randomized among several treatments if possible.
In the switch wing of Level 2 treatment (Rush et al.
2006b) two of the medications were believed to differ in
neurochemical mechanisms from Celexa. Since Celexa, an
SSRI drug, was demonstrably ineffective in these patients,
a drug with a different mechanism should theoretically be
particularly helpful. Effexor (venlafaxin) is a ‘‘dual action’’
drug, believed to act by increasing the levels of both
serotonin and norepinephrine. Wellbutrin (bupropion) is an
‘‘out of class’’ antidepressant believed to block the reup-
take of dopamine and norepinephrine. It is not an SSRI.
The third medication was Zoloft (sertraline), a frequently-
prescribed SSRI. Although logically it should not be par-
ticularly effective after Celexa, Zoloft was viewed as a
different SSRI that might work even if Celexa did not.
Seven hundred and twenty-seven patients were randomly
distributed among these three drug treatments.
In the augmentation wing of Level 2 treatment (Trivedi
et al. 2006a), patients continued to take Celexa but were
assigned to add either Wellbutrin (bupropion), Buspar
(buspirone), or cognitive therapy. Wellbutrin, as described
above, is an antidepressant that is not an SSRI while Bu-
spar is an antianxiety drug with mild sedative effects.
Presumably, the combination of an SSRI with drugs of
different mechanisms of action should enhance the out-
come. Five hundred and sixty-five patients were randomly
distributed among the two combined drug treatments.
The cognitive therapy (see Thase et al. 2007) was
delivered by 44 therapists selected at the 41 sites. The
training was not extensive, although the therapists were
given readings and a two-day workshop. Monthly group
supervision meetings were held. Therapy sessions were
scheduled twice weekly for four weeks, then once weekly
for the remaining 8 weeks, for a total of 16 sessions over
12 weeks. The therapists resembled what might be readily
available in the various communities, and the therapy itself
was short-term.
For unclear reasons, three-fourths of the patients who
entered Level 2 refused to accept cognitive therapy as an
option for treatment. A possible biasing factor was that
medication treatment was entirely paid by the grant and
cognitive therapy was not. Another factor may have been
that the choice of Level 2 treatment emerged from the
patient’s consultation with the psychiatrist who had treated
them with drugs in Level 1. In addition, cognitive therapy
had certain disadvantages. Patients had to go to the offsite
office of the assigned therapist, and therapy fees were not
reimbursed. Although medications were delivered under
near ideal conditions, cognitive therapy was not. As a
consequence of various complexities, only 36 patients
contributed to the outcome data for the switching wing of
Level 2 and only 65 contributed data for the augmentation
96 J Contemp Psychother (2009) 39:93–100
123

wing. Thus, Level 2 provided a relatively small N to
compare cognitive therapy with medication. The STAR*D
project, though a major study of medications, offered a
relatively minor study of psychotherapy.
Results of Level 2 Treatment
The results from switching to the three different drugs were
quite similar, in spite of dissimilar mechanisms of action.
The remission rates, assessed by the Hamilton scale, were
24.8% for Effexor, 21.3% for Wellbutrin, and 17.6% for
Zoloft. Statistically, there were no significant differences.
Of the 727 patients who were switched to other medica-
tions, we find the combined remission rate was a
disappointing 21.3% (155 patients). There was no support
for the theory that treating failed patients with drugs having
a different mechanism of action would enhance the
outcome.
For the 36 patients who switched to cognitive therapy
and whose data could be compared statistically with the
medications, the remission rate was 25%—not significantly
different from the switch medications (Thase et al. 2007,
Table 6). Thus, cognitive therapy was approximately as
effective as medication but not more so. The patients
attended an average of 11 sessions, with only one-fourth of
the patients completing the full 16-session protocol. The
patients’ failure to take full advantage of psychotherapy
may have been related to its disadvantages. The patients
had to go to an offsite office for therapy, and they were not
reimbursed for therapy fees or for copayment charges for
insurance-covered therapy.
In the augmentation wing of Level 2, remission differed
little among the various groups, and there were no statis-
tically significant differences among the various
augmentation treatments. For the group given cognitive
therapy plus Celexa, 23% (15 out of 65) remitted. For
patients given Wellbutrin plus Celexa 29.7% (83 out of
279) remitted, while for patients given Buspar plus Celexa
30.2% (86 out of 285 patients) remitted Trivedi et al.
2006a, Table 3). Even though the two drugs had quite
different mechanisms of action than Celexa, neither added
much positive effect. For the 564 patients treated with a
combination of medications, we find the combined remis-
sion rate to be 29.9%.
Although the augmentation strategy appeared to
engender slightly more remission (about 30%) than the
switching strategy (about 21%), the assignment of patients
to the two strategies was self-selected and nonrandom. The
assumptions of a statistical comparison of the two strate-
gies, therefore, could not be met. Furthermore, patients
who accepted the augmentation strategies were less
severely depressed and had less side effects from Celexa
than the patients who accepted the switch strategies (Rush
et al. 2006a, p. 1907).
Level 3 Treatment
Level 3 treatment offered yet another round of medications
in an effort to successfully treat the failures from Levels 1
and 2. Of the 4,041 patients originally eligible for the
study, only 377 (9.3%) were still available and willing to
accept the third step of treatment. These patients had either
been intolerant to side-effects in Level 2 or had failed to
reach remission. About 30% of Level 2 patients were
unavailable because they had refused further treatment and
had dropped out of the study.
The patients, with their physicians’ consultation, could
choose between a switch treatment and an augmentation
treatment. In the switch option (Fava et al. 2006) two drugs
were available: Remeron (mirtazamine), an atypical anti-
depressant believed to enhance levels of both
norepinephrine and serotonin, and Pamelor (nortriptyline),
an older tricyclic antidepressant. In the augmentation
option (Nierenberg et al. 2006), either lithium or T3 (tri-
iodothyromone) could be added to the ongoing medication.
Previous small studies had indicated that these drugs might
be effective as augmenters.
Results of Level 3 Treatment
The outcomes from the Level 3 treatments were notable for
a low rate of remissions and a high rate of dropout. In the
switch study, the remission rate for Remeron was 12.3%,
and for Pamelor it was 19.8%—with no statistical differ-
ence (Fava et al. 2006, p. 1168). In the augmentation study,
remission rates were 15.9% for lithium and 24.7% for T3.
Again, the difference between drugs was not statistically
significant (Nierenberg et al. 2006, p. 1526). A notable
44.8% of the patients who entered Level 3 dropped out of
the study and made themselves unavailable for treatment at
Level 4 (McGrath et al. 2006, p. 1533).
How effective was the third level of treatment? Out of
377 patients, we find that 67 or 17.8% reached remission.
On the other hand, 310 patients or 82.3% were not suc-
cessfully treated and continued to be depressed.
Level 4 Treatment
Level 4 treatment offered one last round of medication in
an effort to help patients who had been resistant to three
prior treatments. At this point only 109 patients consented
to further treatment. They were randomly assigned to
J Contemp Psychother (2009) 39:93–100 97
123

receive either Parnate (tranylcypromine), a monoamine
oxidase inhibitor, or a combination of Effexor (venlafax-
ine) and Remeron (mirtazapine). Previous small studies
suggested that these drugs could be effective in treatment
resistant patients.
Results of Level 4 Treatment
Remission rates were quite low for both the Parnate
group and the Effexor-plus-Remeron group, and the rates
were not significantly different. For Parnate the remission
rate was 6.9%, and for Effexor plus Remeron it was
13.7%. Of 109 patients, a total of 11 (10.1%) remitted
(McGrath et al. 2006, Table 2). Thus, the fourth and final
treatment effort, using drugs with various neurochemical
actions, benefited only a few patients. On the basis of
these data, it would be hard to justify treating patients in
clinical practice with a fourth round of antidepressant
medication.
Results from the 12-month Follow-up Care
Rush et al. (2006a) published the follow-up results in a
major paper summarizing the outcomes for all four treat-
ment levels. The summary paper unfortunately introduced
a confusing pair of changes. Change #1: depression was
now measured by the self-report scale (Quick Inventory of
Depressive Symptoms-Self-Report or QIDS-SR-16), and
measurement by the ‘‘gold standard’’ Hamilton scale
(HRSD-17) was abandoned. Since the self-report scale
typically yielded higher remission rates than the Hamilton
scale, the number of reported remissions was increased.
After Level 1 treatment, for example, the remission rate
was 27.5% based on the Hamilton scores versus 33% based
on the self-report scores (Trivedi et al. 2006b, p. 28). In
step 2’s switch comparisons, remission rates for bupropion,
sertraline, and venlafaxine were 21.3%, 17.6%, and 24.8%
respectively by Hamilton scores versus 25.5%, 26.6%, and
25%.7 by self-report scores (Rush et al. 2006b, p. 1,231) In
step 2’s augmentation comparisons, Hamilton remission
rates for bupropion and buspirone were 29.7% and 30.1%
versus self-report’s 39% and 32.9%.8 (Trivedi et al. 2006a,
p. 1243) In these comparisons, remissions based on self-
report scores were about 22% higher than remissions based
on Hamilton scores.
Readers need to be wary when the pre-specified primary
outcome measure is changed after the study is completed.
Despite the change, however, the relapse and dropout rates
during follow-up (see below) were so extraordinarily large
that the switch in test instrument could not have changed
the overall conclusion very much.
Change # 2: the enrollees selected to provide outcome
data in the summary paper include 795 mildly depressed
patients whereas the Level 1 paper reported only moder-
ately-to-severely depressed patients. ‘‘Mildly depressed’’
was defined by self-report scores ranging from 6 to 10
(comparable to Hamilton scores of 8–13). The number of
enrollees receiving treatment at Level 1 was now reported
as 3,671, rather than the 2,876 reported in the Level 1
paper.
Procedure During Follow-up Care
After each level of treatment, patients who achieved
remission were moved to a 12-month follow-up where
medications were continued and relapse could be evalu-
ated. Relapse was defined as a return to moderate or severe
depression, defined as a self-report score of 11 or more.
Patients who had showed improvement (defined as a
reduction in score of 50% or more) and who refused further
treatment were also moved to follow-up. The protocol
during the followup was to strongly recommend that all
patients continue taking the drug at the dose used at the end
of acute treatment. Visits to the hospital or clinic were
recommended every two months, A score on the self-report
scale was to be collected by telephone at one-month
intervals, using an an interactive voice response system.
Any medication change and even psychotherapy was per-
mitted, just as might happen in normal clinical treatment.
The follow-up care was designed to be naturalistic, even
though the care was not precisely defined.
Relapse during follow-up was evaluated by a survival
analysis. Table 1shows relapse data, after each level of
treatment, for patients entering follow-up in remission
(data taken from Fig. 3 in Rush et al. 2006a). The numbers
represent the remitted patients who ‘‘survived’’ (i.e., did
not relapse or drop out) for each three-month period, based
on at least one call in the three-month period to the auto-
mated assessment system. The bottom row of Table 1
summarizes the outcome for all treatment levels for the
1,518 patients entering follow-up in remission. The trend
was sharply downward with 851 (56.1%) patients calling in
Table 1 Relapse of patients who entered follow-up in remission.
Column ‘‘N’’ shows number of patients entering follow-up. Numbers
under ‘‘Months’’ are patients still in remission
Level N 0–3 Months 3–6 Months 6–9 Months 9–12 Months
1 1,085 628 (57.9%) 421 (38.8%) 290 (26.7%) 84 (7.7%)
2 383 199 (52.0%) 133 (34.7%) 79 (20.6%) 20 (5.2%)
3 35 16 (45.7%) 11 (31.4%) 8 (22.9%) 2 (5.7%)
4 15 8 (53.3%) 5 (33.3%) 5 (33.3%) 2 (13.3%)
Total 1,518 851 (56.1%) 580 (38.2%) 382 (25.2%) 108 (7.1%)
98 J Contemp Psychother (2009) 39:93–100
123

and not relapsing in the first quarter, 580 (38.2%) in the
second quarter, 382 (25.2%) in the third, and 108 (7.1%) in
the fourth. Thus, almost 93% of the remitted patients had
either relapsed or dropped out by the 9–12th month. Sim-
ilar trends with similar percentages were observed for all
four treatment levels, including Level 1 with the largest N
and no prior treatment failures.
Table 2, taken from Fig. 4 of the summary article (ibid.),
presents similar relapse information for patients who
entered follow-up not in remission. They had qualified for
follow-up because their self-report scores had improved
after treatment medication by 50% or more and they had
refused the next treatment level. As in Table 1, the trend
was sharply downward such that less that only 4% of the
patients had not relapsed or dropped out by the end of
12 months, regardless of the treatment level. Of the total
725 unremitted but improved patients, 305 (42.2%) had not
relapsed in the first three months, 171 (23.6%) in the sec-
ond three months, 102 (14.1%) in the third, and 24 (3.3%)
in the fourth. Every percentage entry in Table 2was
smaller than the corresponding entry in Table 1, indicating
that relapse and dropout was reliably greater for patients
who entered follow-up not in remission than for patients
who entered in remission. This result tends to support the
STAR*D authors’ contention that remission should be the
goal of treatment.
Summary and Commentary
Over the entire course of the STAR*D project, 3,671
depressed patients were given one to four steps of cost-free,
exemplary treatment with antidepressant medication. Of
the patients who entered follow-up in a remitted status,
only 108 were found not to have relapsed or dropped out by
the end of the 12-month follow-up. For a widely-used,
FDA-approved treatment, viewed in some quarters as the
treatment of choice, the long-term outcome was strikingly
unsatisfactory.
The STAR*D trial was particularly instructive because
the participants were broadly representative of real
patients, often with chronic depression and co-occurring
psychiatric and medical disorders. This choice of partici-
pants gave us a chance to see the effectiveness of
antidepressants with the types of patients that psychiatrists
treat in routine clinical practice.
If one thing is to be remembered from the massive,
complex, STAR*D study, it is that antidepressant drugs
were not very effective at treating depression in real
patients. This is especially so when you consider the
probable placebo effect combined with spontaneous
recovery. The trend from four sequential treatments was
clear. With each additional treatment a smaller and smaller
percentage of patients were successfully treated. Further-
more, even for patients who had a remission of symptoms,
only a small minority had not relapsed or dropped out by
the end of 12 months of follow-up care.
References
Fava, M., Rush, A. J., Wisniewski, S. R., Nierenberg, A. A., et al.
(2006). A comparison of mirtazapine and nortriptyline following
two consecutive failed medication treatments for depressed
outpatients: A STAR*D Report. The American Journal of
Psychiatry, 163, 1161–1172.
Kirsch, I., Moore, T. J., Scobabria, A., & Nicholls, S. S. (2002). The
emperor’s new drugs: An analysis of antidepressive medication
data submitted to the U. S. food and drug administration.
Prevention & Treatment, 5(23). http://content.apa.org/journals/
pre/5/1/33.
McGrath, P. J., Stewart, J. W., Fava, M., Trivedi, M. H., et al. (2006).
Tranylcypromine versus venlafaxine plus mirtazapine following
three failed antidepressant medication trials for depression: A
STAR*D Report. The American Journal of Psychiatry, 163,
1531–1541.
Nierenberg, A. A., Fava, M., Trivedi, M. D., Wisniewski, S. R., et al.
(2006). A comparison of lithium and T3 augmentation following
two failed medication treatments for depression: A STAR*D
report. The American Journal of Psychiatry, 163, 1519–1530.
Rush, A. J., Fava, M., Wisniewski, S. R., Lavori, P. W., et al. (2004).
Sequenced treatment alternatives to relieve depression
(STAR*D): Rationale and design. Controlled Clinical Trials,
25, 119–142.
Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Nierenberg, A. A.,
et al. (2006a). Acute and longer-term outcomes in depressed
outpatients requiring one or several treatment steps: A STAR*D
report. The American Journal of Psychiatry, 163, 1905–1917.
Rush, A. J., Trivedi, M. H., Wisniewski, S. R., Stewart, J. W., et al.
(2006b). Bupropion-sr, sertraline, or venlafaxine-xr after failure
of SSRIs for depression. The New England Journal of Medicine,
354, 1231–1242.
Thase, M. E., Friedman, E. S., Biggs, M. M., Wisniewski, S. R., et al.
(2007). Cognitive therapy versus medication in augmentation
and switch strategies as second-step treatments: A STAR*D
report. The American Journal of Psychiatry, 164, 739–752.
Trivedi, M. H., Fava, M., Wisniewski, S. R., Thase, M. E., et al.
(2006a). Medication augmentation after the failure of SSRIs for
depression. The New England Journal of Medicine, 354, 1243–
1252.
Trivedi, M. H., Rush, A. J., Wisniewski, S. S., et al. (2006b).
Evaluation of outcomes with citalopram for depression using
Table 2 Relapse of patients who entered follow-up not in remission.
Column ‘‘N’’ shows number entering follow-up. Numbers under
‘‘Months’’ are patients still in remission
Level N 0–3 Months 3–6 Months 6–9 Months 9–12 Months
1 388 173 (44.6%) 97 (25.0%) 56 (14.4%) 14 (3.6%)
2 237 99 (41.8%) 56 (23.6%) 35 (14.8%) 9 (3.8%)
3 66 21 (31.8%) 11 (16.7%) 7 (10.6%) 1 (1.5%)
4 34 12 (35.3%) 7 (20.6%) 4 (11.8%) 0 (0.0%)
Total 725 305 (42.1%) 171 (23.6%) 102 (14.1%) 24 (3.3%)
J Contemp Psychother (2009) 39:93–100 99
123

measurement-based care in STAR*D: Implications for clinical
practice. The American Journal of Psychiatry, 163, 28–40.
Turner, E. H., Matthews, A. M., et al. (2008). Selective publication of
antidepressant trials and its influence on apparent efficacy. The
New England Journal of Medicine, 358, 252–260.
Vedatam, S. (2006, March 23). Drugs cure depression in half of
patients: Doctors have mixed reactions to study findings. The
Washington Post, p. A1.
100 J Contemp Psychother (2009) 39:93–100
123