Comparison of Desipramine and
Cognitive/Behavioral Therapy in the
Treatment of Elderly Outpatients With
Mild-to-Moderate Depression
Larry W. Thompson, Ph.D., David W. Coon, Ph.D.
Dolores Gallagher-Thompson, Ph.D., Barbara R. Sommer, M.D.
Diana Koin, M.D.

The authors evaluated the efficacy of desipramine-alone, vs. cognitive/behavioral therapy-alone (CBT) vs. a combination of the two, for the treatment of depression in older
adult outpatients. Patients (N‫ )201ס‬meeting criteria for major depressive disorder
were randomly assigned to one of these three treatments for 16 to 20 therapy sessions.
All treatments resulted in substantial improvement. In general, the CBT-Alone and
Combined groups had similar levels of improvement. In most analyses, the Combined
group showed greater improvement than the Desipramine-Alone group, whereas the
CBT-Alone group showed only marginally better improvement. The combined therapies were most effective in patients who were more severely depressed, particularly
when desipramine was at or above recommended stable dosage levels. The results
indicate that psychotherapy can be an effective treatment for older adult outpatients
with moderate levels of depression. (Am J Geriatr Psychiatry 2001; 9:225–240)

D

epression is one of the most common psychiatric
disorders among older adults,1–3 with estimates of
at least 15%–20% of community elderly persons suffering from a type and degree of depression that needs
clinical and public health attention.4 Although pharmacotherapy is known to be effective in the treatment
of depression, various forms of psychotherapy also have
an essential role, in that they address the psychosocial
and functional antecedents and consequences of this
disorder.5–10

Controlled trials evaluating psychotherapeutic interventions for the treatment of late-life depression have
most often featured brief approaches, including timelimited psychodynamic therapy, interpersonal psychotherapy (IPT), or cognitive/behavioral therapy (CBT). A
meta-analysis of 17 psychosocial treatment studies involving 765 depressed elderly patients demonstrated
that these interventions were more effective than notreatment or placebo-control conditions, yielding an
overall effect size of 0.78.11 Furthermore, effect sizes

Received September 13, 1999; revised July 6, 2000; accepted August 21, 2000. From the Geriatric Research, Education and Clinical Center (GRECC),
Veterans Affairs Palo Alto Health Care System, Palo Alto, CA, and the Dept. of Medicine, Stanford University School of Medicine, Stanford, CA.
Address correspondence to Dr. Thompson, Pacific Graduate School of Psychology, 935 East Meadow, Palo Alto, CA 94304. e-mail: larrywt@leland.stanford.edu
Copyright ᭧ 2001 American Association for Geriatric Psychiatry

Am J Geriatr Psychiatry 9:3, Summer 2001

225

Desipramine vs. Psychotherapy
varied negligibly between self-rated and clinician-rated
outcome measures, as well as across studies with only
major depressive disorder or with subclinical varieties
of depression.
Correspondingly, only a handful of studies to date
have directly compared psychotherapies to pharmacotherapies in treatment of late-life depression, and thus
an insufficient amount of data precludes researchers
and clinicians from drawing firm conclusions. Sloane
and colleagues12 found that IPT was comparable to nortriptyline in the acute treatment of depressed elderly
subjects and was associated with lower dropout rates.
Likewise, Beutler and his fellow researchers,13 in their
comparison of group cognitive therapy-plus-alprazolam,
group cognitive therapy-plus-placebo support, alprazolam-plus-support, and placebo-plus-support, found that
patients assigned to group cognitive therapy improved
more on self-reported depressive symptoms and objective sleep efficiency than patients not assigned to group
cognitive therapy. No differences between alprazolam
and placebo were found, regardless of whether or not
patients received group cognitive therapy. Finally, Reynolds and colleagues14,15 have demonstrated in several
reports that a combination treatment of IPT with nortriptyline can be quite efficacious in both the acute and
maintenance treatment of elderly patients with recurrent depression.
Although not targeted toward elderly patients, the
highly publicized results of the Treatment of Depression
Collaborative Research Program (TDCRP), sponsored
by the National Institute of Mental Health,16 produced
significant implications for clinicians and clinical researchers treating depressed elderly patients. The
TDCRP compared imipramine-plus-clinical management (IMI-CM), IPT, CBT, and placebo-plus-clinical management (PLA-CM) for the treatment of major depressive disorder in adult outpatients. Initial findings
reported a significant reduction in depressive symptoms
and functional improvement over time across all four
conditions. At the time, the authors also pointed out a
consistent ordering of treatments at termination, with
IMI-CM generally doing the best and demonstrating a
more rapid effect, PLA-CM the worst, and the two psychotherapies in between.16–18 More recent post-hoc
analyses of TDCRP data have used random regression
models to investigate the role of initial severity in treatment outcome, as measured by scores on the Hamilton
Rating Scale for Depression (Ham-D),19 the Beck Depression Inventory (BDI),20 or the Global Assessment

226

Scale (GAS).21 Results have suggested significantly better outcomes on depressive symptoms in IMI-CM for
patients who were severely depressed at baseline, when
compared with similar patients in each of the other
treatments. However, treatment differences were less
evident in patients who were less severely depressed
on study entry.22
Even though the TDCRP project has sparked considerable debate about site differences, therapeutic allegiances, placebo-control conditions, and treatment fidelity issues,11,23–28 the project’s results were still
instrumental in shaping treatment guidelines prepared
for both the American Psychiatric Association (APA)29
and the Agency For Health Care Policy and Research
(AHCPR).30 For example, the resultant APA guidelines
suggest that CBT may not be any more effective than
pill-placebo coupled with clinical management. Similarly, AHCPR guidelines state that there is strong evidence for the efficacy of medication in the treatment of
severe depressions and little or none for the efficacy of
psychotherapy alone. Contrary to these positions, a recent meta-analysis reported in 1999 on four randomized
clinical trials investigating the acute outcomes of antidepressant medication and CBT in severely depressed
outpatients does not support the position that antidepressant medication is superior to CBT in these patients.31
Thus, the question of the relative efficacy of pharmacotherapy, psychotherapy, and their combination remains unanswered. Researchers summarizing points
from the 1991 Consensus Conference pointed out that
“both antidepressant medications and brief structured
psychotherapies have efficacy in the acute treatment of
older adult depressed outpatients with major unipolar,
nondelusional depression, but the relative efficacy of
specific medications and psychotherapies is unclear (p
485).”32 Similarly, as part of the Consensus Update,
Niederehe noted, in a review of randomized trials of
psychotherapy conditions, that the “weaknesses lie not
so much with the quality of specific studies, but with
the cumulative body of evidence (p S76).”10 Furthermore, he makes the point that “combination-treatment
studies deserve much greater emphasis, and [an] integration of psychopharmacological and psychotherapy
research should prove beneficial to both areas (p
S75).”10
Previous research in patients with late-life depression has shown the efficacy of CBT and brief psychodynamic treatments alone.24 The present study was de-

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
signed to evaluate the efficacy of a specific medication,
desipramine, and a specific structured psychotherapy,
CBT, or a combination of both in the treatment of older
adult outpatients with major depressive disorder. At the
time that this study was first conceived, tricyclic antidepressants were the pharmacologic treatments of
choice for both young adult and elderly patients with
severe depression. Although newer selective serotonin
reuptake inhibitors (SSRIs) are now the preferred firstline treatments, tricyclics continue to play a role as both
primary and augmenting treatments of this disorder.33,34
Tricyclics may even be treatments of first choice for
medically stable patients with very severe depression33,35 or for patients with both depression and
chronic neuropathic pain syndromes. This project
therefore remains timely in its attempt to evaluate these
specific treatments for depression in elderly patients.

METHODS
Subjects
The study was conducted at the Veterans Affairs
Palo Alto Health Care System and Stanford University
School of Medicine, in a clinic specializing in the treatment of older adults with affective disorders. Study subjects were required to meet the following criteria: 1)
age 60 years or older; 2) a current diagnosis of major
depressive disorder (MDD) as determined by the Research Diagnostic Criteria (RDC)36 applied to the Schedule for Affective Disorders and Schizophrenia (SADS)37;
3) a score of 14 or greater on the Ham-D; 4) a score of
16 or greater on the BDI; 5) a score of 26 or greater on
the Mini-Mental State Exam38; 6) no medical or psychiatric contraindications; 7) no evidence of serious alcohol abuse; 8) no evidence of a psychotic disorder; 9) no
evidence of bipolar disorder; 10) no immediate suicide
risk; 11) not taking another medication for the treatment of depression; and 12) adequate transportation to
reach the outpatient clinic.
Potential patients were either self-referred in response to media advertisements or were referred directly to the project by community physicians, mental
health organizations, and social service agencies. Patients diagnosed as being in an episode of MDD according to RDC upon initial screening (N‫ )712ס‬were then
seen by a staff psychiatrist in order to confirm the diagnosis. A few prospective participants were already re-

Am J Geriatr Psychiatry 9:3, Summer 2001

ceiving medication for depression. In these instances,
the medication was withdrawn after consultation and
approval from their primary care physician, and rescreening occurred after a 2-week washout. Those with
a confirmed diagnosis by the project psychiatrist were
scheduled for admission to the hospital overnight for
medical screening to determine whether there were
any contraindications for using desipramine. Patients
were then re-screened to see if they still met criteria for
the study. Because of scheduling issues, the time interval for this entire screening process ranged from 2 to 6
weeks. Approximately half of these subjects (n‫)511ס‬
were not entered into the study because they no longer
met criteria for MDD or they were below the cutoff
score on the Ham-D or BDI at the time of re-screening.
Ultimately, 102 patients met inclusion and exclusion requirements and were enrolled in the study. Two
of these subjects dropped out after randomization but
before receiving any treatment. They are included in the
initial descriptive analyses, but not in the outcome analyses of the 100 patients presented here. Ineligible or
uninterested subjects were offered referrals to local clinics, physicians, psychiatrists, psychotherapists, and
other research studies to help address their needs. Although this was not a placebo-controlled study, the
length of time from initial rating scales to study entry
was about 1 month, on average. It was our expectation
that this waiting period would exclude patients who
might improve as a result of the inquiry itself.
Procedures
Patients were assigned randomly to receive treatment with either Desipramine-Alone, CBT-Alone, or a
Combination of the two (Desipramine ‫ ם‬CBT). Patients
in all three treatment conditions were seen for 16 to 20
sessions over a 3- to 4-month period. During the first 4
weeks, they were seen twice per week, and then once
per week for the next 8 to 12 weeks.
Medication treatment. Six psychiatrists (3 women, 3
men) with a primary interest in geriatrics and geropsychiatry participated as therapists in this research program. All had received specialized training in geriatrics
and geriatric psychiatry.
Desipramine was chosen in consultation with project psychiatrists because of its low side-effect profile,
including low sedative effect and lower anticholinergic
effect when compared with other tricyclic and hetero-

227

Desipramine vs. Psychotherapy
cyclic agents.39,40 In the Desipramine-Alone condition,
the psychiatrists followed the drug condition protocol
developed for the NIMH-TDCRP,41 with some minor
modifications to make the protocol more suitable for
elderly patients. (The manual for this modified protocol
is available upon request from the first author.) Patients
were seen individually for approximately 30 minutes
per session on the frequency schedule described previously. Each session began with a brief inquiry about
the patient’s current level of distress. Psychiatrists maintained a supportive attitude in talking with patients
about their feelings and problems and continued to emphasize the importance of the medication in improving
patients’ depression. A review of other symptoms and
side effects currently experienced by the patient followed this discussion and included a count of the number of desipramine pills the patient had remaining.
Blood pressure (sitting and standing) and weight were
also monitored during every session. Sessions ended
with a brief discussion about the medication regimen
to be followed until the next session, a brief summarization of treatments prescribed to counteract any disquieting side effects, and words of continued encouragement by the psychiatrist regarding the efficacy of
desipramine in treating patients’ depression.
The general strategy for prescribing medication was
to start with very low doses (10 mg) of desipramine and
increase this amount gradually, depending on the magnitude of side effects and the patients’ general willingness to tolerate them. A physician, board-certified in
Internal Medicine and Geriatrics, was available as
needed for consultation in managing these side effects.
Desipramine plasma levels were obtained at 1 month
after initiation of treatment or at any point in the course
of treatment that the psychiatrist felt this information
would be helpful in determining appropriate daily dosage. In some patients, repeated levels were measured
to assist in establishing preferred dosage levels or check
on compliance. The mean stable daily dose of desipramine in this study was 90.0‫ 00.36ע‬mg, which is below
the dosage level recommended by Potter42 for use with
young patients, but slightly above the daily dosage range
(10 mg–75mg) recommended by Salzman for use with
geriatric patients.43 The mean plasma level of desipramine was 123 ng/mL which approximates the recommended therapeutic blood level of 125 ng/mL or
greater,44 but values ranged from barely detectable to
466‫ 331ע‬ng/mL. The design of this study did not call
for plasma levels to be obtained consistently at stable

228

dose levels; therefore it is not known whether patients
were in the recommended steady-state therapeutic level
of 125 ng/ml. Actual plasma levels were not related to
outcome in therapy, but it should be noted that plasma
determinations were not obtained according to a suitable protocol specifically designed to address this question. Moreover, although appropriately designed studies
have generally found a significant relationship between
serum levels of desipramine and outcome, this has not
always been the case.42
Cognitive/behavioral therapy. Eight clinical psychologists (4 men, 4 women) were therapists in the CBTAlone condition. All therapists had at least 1 year of
experience in working with geriatric patients who evidenced psychiatric symptoms.
The CBT-Alone condition essentially followed the
conceptual model and treatment program developed by
Aaron Beck and his colleagues,45 with specific modifications for older individuals.46 These involved strategies
to facilitate learning, such as repeated presentation of
information using different modalities, slower rates of
presentation, greater use of practice, and so forth, along
with greater use of structure and modeling behavior.
Briefly, patients were taught to identify, monitor, and
ultimately challenge negative cognitions about themselves or their situations and then to develop more
adaptive and flexible cognitions in their stead. Where
appropriate, emphasis was also placed on teaching patients to monitor and increase daily pleasant events in
their lives by use of behavioral treatment procedures.47,48 Each session lasted for 50 to 60 minutes and
followed a structured manual developed specifically for
this research program.49
Drug and CBT combined. In the Combined condition, patients saw both a psychiatrist and a cognitive–
behavioral therapist in two separate contacts for each
of the 16–20 therapeutic sessions. Typically, these drug
and CBT contacts were conducted back-to-back, although, occasionally, scheduling difficulties would require that one contact be held on the next day. In the
drug administration component of this condition, the
psychiatrist completed a brief (approximately 15 minutes) review of symptoms and a check on medication
usage. Any issues besides concerns about side effects or
questions about medication that arose during this time
were deferred until the next CBT contact. In the CBT
component, the therapist followed the protocol de-

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
scribed in the CBT-Alone condition. Any questions regarding side effects or medication usage were either deferred until the next meeting with the psychiatrist or, if
there were indications of possible serious complications, referred immediately to the treating psychiatrist.
Measures
During project screening and enrollment, SADS interviews were used to formulate a diagnosis of major
depressive disorder, and Ham-D ratings were made.
Also, all patients completed self-report scales before
treatment to assess their level of symptoms and general
functioning, their characteristic ways of dealing with
stressful situations, and their personality. At the conclusion of treatment, these measures were repeated. A
posttreatment RDC diagnosis was also made, using the
SADS-Change interview,50 along with the Ham-D. Information about pre–post diagnostic status and mean levels on measures of symptoms and functioning have
been previously reported at professional conferences
(Gerontological Society of America Convention, November 1991, San Francisco, CA; American Psychological Association Convention, August 1994, Los Angeles,
CA), and also reviewed at the 1991 Consensus Conference51 (papers are available upon request). The primary
focus of this report is on the application of a recently
developed method for the analysis of change over time,
described in the Statistical Measures section, below. We
selected, a priori, two measures of depressive symptoms for these outcome analyses. They are the BDI (a
self-report measure) and the 17-item version of the (interviewer-rated) Ham-D; both were used to evaluate the
intensity of depression. We used the long form of the
BDI during initial screening reported here and for the
pre–post-treatment comparisons reported in earlier
data referred to above. During the actual treatment, patients completed the short form of the BDI (BDI-SF) at
every session.52,53 We selected the BDI-SF rather than
the long form (LF), as well as the Ham-D, for the growth
model analyses reported in this paper. Both measures
were obtained immediately before patients entered the
first treatment session. The BDI-SF was collected at each
subsequent treatment session, including termination,
whereas Ham-D observer ratings were collected at
every other session. When patients discontinued treatment, BDI and Ham-D ratings were obtained within 10
days of termination. On the rare occasions when patients were unable to complete Ham-D ratings at a

Am J Geriatr Psychiatry 9:3, Summer 2001

scheduled visit, these ratings were obtained at the patient’s next appointment. Ham-D interviews were audiotaped, and over 50% were randomly selected for review
by a project psychiatrist or the principal investigator for
quality-control purposes. All independent interviewers
were trained to proficiency by the principal investigator, and interviewer training ratings were cross-compared with training proficiency standards applied in an
independent clinical research laboratory at Stanford
University School of Medicine. Dual ratings were obtained on approximately 50% of Ham-D interviews,
yielding a mean interrater reliability for the Ham-D
across the study of 0.97, with a range of 0.91 to 0.99.
Statistical Methods
Growth modeling54,55 was used to investigate the
differential treatment effects on the reduction of patient
depression as measured by patient responses on the BDI
and the Ham-D. This method is similar to random-regression model methods applied to analyses of the
TDCRP data set.17,22 Also, our interest in both intraindividual as well as interindividual and group change in
depression over the course of treatment lent itself to
the use of this particular methodology. Moreover, in
contrast to traditional analytic methods, such as endpoint analysis and multivariate repeated-measures analysis, growth modeling permits the number and spacing
of assessments to vary across subjects. Patients who
missed some evaluation points or who discontinued
treatment are still considered in the analyses. This is of
particular importance, given illness, transportation, and
other concerns that can interfere with older adults’ being able to keep scheduled appointments. Since all 100
study subjects who entered treatment had completed
some outcome measurements both before and during
treatment, we were able (with this statistical method)
to include all data collected on these patients whether
or not they completed the full trial.
In this growth modeling framework, a linear mixedeffects (random coefficients) model56 was fitted to the
data. The basic functional form of the model specifies
two linear growth parameters of interest: the intercept,
or pre-treatment level of depression, and the slope, or
the per-session rate of improvement (the current study’s
dependent variable as measured by the reduction of depression levels in the Ham-D and BDI). These two parameters are estimated as random effects that are allowed to vary across individuals, where a straight line

229

Desipramine vs. Psychotherapy
with an initial depression level and the improvement
rate specific to the individual then describes each patient.
Each subject’s treatment trajectory is thus captured
by the linear regression of the dependent variable on
time. In this study, time is indexed by the progression
of therapy sessions. Treatment efficacy in the reduction
of depression levels is estimated by introducing the
treatment-group variable as a fixed effect into the linear
mixed model, where the distribution of each parameter
across study participants is described by its mean and
variance. Systematic differences among participants
due to different treatments can then be estimated by
examining the distributions of the growth parameters
across the three different treatment groups. Restricted
maximum-likelihood technique, as implemented in the
SAS MIXED procedure, was used to obtain estimates of
the model parameters and their standard errors (see
Verbeke and Molenberghs57 for technical details). All
other data analyses were completed using the Systat for
Windows, Release 5.0 program.58
A brief summary of pre– post-treatment means of
the Ham-D and the BDI-Long Form for the full sample
(N‫ )001ס‬has also been included. Because there is concern that our Ham-D cutoff score for exclusion was too
low, we also have included a post-hoc analysis on only
those subjects with initial Ham-D scores above 16.
These analyses were completed using a 3 (treatment)
‫( 2 ן‬pre–post-treatment) multivariate analysis of variance design for repeated measures.

RESULTS
Sociodemographic Characteristics
Table 1 lists the sociodemographic characteristics
for the total number of patients who received treatment
(N‫ ,)001ס‬grouped according to treatment condition.
There were no significant differences among the groups
for age, education, occupation, gender, or marital status
either at the start or at the endpoint of treatment. Subjects were mostly in their mid- to late-sixties. Most were
white; almost one-third of them still had some form of
employment; and all were reasonably well-educated for
this age cohort. The female-to-male ratio was 2:1. As
evidenced in previous work with depressed elderly patients,59 the mean duration of the current depressive
episode for this age group was greater than 2 years.

230

Pretreatment Ham-D mean scores ranged between 18.6
and 18.8 for the three groups and were not significantly
different. Thus, the sample was generally in the low end
of the geriatric age range, was relatively high-functioning, and was experiencing mild-to-moderate major depression.

Treatment Discontinuation vs. Completion
After randomization and across the course of treatment, 31 of 102 patients (30%) discontinued. This rate
is slightly lower than the dropout rate of 32% reported
in the TDCRP.16 There were no significant differences
in discontinuation due to treatment condition
(v2[2]‫ ;92.1ס‬N‫ ;201ס‬P‫ .)25.0ס‬Twelve patients (34%)
dropped from the Desipramine-Alone condition, and 12
(33%) from the Combined condition, whereas 7 patients (23%) discontinued from the CBT-Alone condition. There was no significant difference in discontinuation due to treatment dissatisfaction across the three
treatments (v2[2]‫ ;99.2ס‬N‫ ;201ס‬P‫ .)22.0ס‬No significant gender, age, education, occupation, or maritalstatus differences were found between project Completers and Dropouts across the three treatment groups.
Thus, the rate of discontinuation does not appear to be
associated with sociodemographic characteristics or
with the type of therapy.
A comparison of Dropouts and Completers was
also made on the pretreatment measures of depressive
symptoms. There were no significant differences between the two groups on either the BDI-LF (t[100]‫;77.0ס‬
P‫ )544.0ס‬or the Ham-D (t[100]‫ ;89.0ס‬P‫ .)823.0ס‬Also,
initial severity of depressive symptoms was explored by
use of the pretreatment Ham-D ratings split into Low
(scores Յ18) and High (scores Ն19) categories. This
split is slightly different from the Ham-D severity groups
of Յ19 and Ն20 chosen by the TDCRP. Although we
considered using the TDCRP’s Ham-D cut-off, the distribution of initial scores in the current study supported
our choice. Moreover, there appears to be no empirical
basis for the TDCRP cut-off. A log-linear analysis yielded
no significant differences in Completers and Dropouts
across treatment conditions on the basis of initial severity of depressive symptoms (LR v2[7]‫ ;97.5ס‬P‫.)565.0ס‬
Thus, the rate of discontinuation does not appear to be
associated with overall depressive symptoms or with
initial severity of those symptoms before patients entered treatment.

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
Treatment Outcomes

Growth-curve analysis. In contrast to the pretreatment BDI scores just discussed, all BDI scores used in
the analysis of treatment outcomes are based on the
BDI-SF. Table 2 summarizes the results for the Ham-D
and BDI-SF change per session. Overall, results investigating the treatment effects indicated significant persession improvement in depressive symptoms as
measured by the Ham-D in all three conditions. Improvement in symptoms as measured by the BDI-SF was
also evident, as reflected in negative per-session rates of
change or slopes, but only the slopes for the CBT-Alone
and Combined conditions were statistically significant
at the 0.05 level.
A significant differential treatment effect was also
found in a comparison of rates of change among the
three groups. Looking first at Ham-D (Table 2) the persession rate of change for the Combined condition was
significantly greater than for the Desipramine-Alone
TABLE 1.

condition. There was no significant difference between
the Combined and CBT-Alone conditions. Although
CBT-Alone showed a larger per-session rate of decline
than the Desipramine-Alone condition, this difference
was not significant. For the BDI-SF, both the Combined
and the CBT-Alone conditions showed a significantly
greater per-session rate of change than the DesipramineAlone condition. There was no significant difference between the Combined and the CBT-Alone conditions.
Post-hoc analyses regarding initial level of severity
and treatment outcome. One concern with this study
has been the low cutoff score of 14 on the Ham-D for
inclusion. Although we think this cutoff was justified,
given the long delay in the screening process and the
careful re-screening that was done to ensure that patients were still clinically depressed, the argument
could be made that our patients were not sufficiently
depressed to provide a fair evaluation of treatments for
MDD. We addressed this issue first by comparing Highand Low-Severity groups in the growth-curve analysis.

Sociodemographic characteristics and initial depression levels for patients entering treatment, by treatment condition
Desipramine
(n؄33)

Variable
Age, years
Education, years
Occupation, category
Sex, n
Men
Women
Marital status, n
Married
Not married

CBT-Alone
(n؄31)

Combined
(n؄36)

Total
(N؄100)

F or v2[df]

P

66.8‫2.6ע‬
14.6‫4.2ע‬
3.6‫7.1ע‬

66.5‫1.5ע‬
14.3‫4.2ע‬
3.2‫4.1ע‬

67.2‫4.6ע‬
14.8‫0.2ע‬
3.4‫8.1ע‬

66.8‫9.5ע‬
14.6‫3.2ע‬
3.4‫6.1ע‬

0.138[2,97]
0.485[2,97]
0.510[2,97]

0.878
0.617
0.602

12
21

10
21

11
25

33
67

0.274[2]

0.872

16
17

14
17

14
22

44
56

0.668[2]

0.716

Note: Values are mean‫ע‬standard deviation, unless otherwise specified. CBT‫ס‬cognitive–behavioral therapy. For occupational category:
1‫ס‬executive/professional; 2‫ס‬business/managerial; 3‫ס‬administrative; 4‫ס‬clerical/sales; 5‫ס‬skilled labor; 6‫ס‬semiskilled labor; 7‫ס‬unskilled
labor.

TABLE 2.

Comparison of change in Ham-D and BDI-SF scores across therapy sessions for the three treatment conditions: rate of
change per session
Ham-D

BDI-SF

Estimate (SE)
Desipramine
CBT
Combined
Desipramine vs. CBT
Desipramine vs. Combined
CBT vs. Combined
Variance
Residual

t

Estimate (SE)

t

–0.20 (0.06)
–0.36 (0.06)
–0.41 (0.06)
0.16 (0.09)
0.21 (0.09)
0.05 (0.09)
0.07 (0.02)
11.67 (0.69)

–3.10**
–5.81***
–6.54***
1.82
2.39*
0.59

–0.10 (0.07)
–0.31 (0.07)
–0.44 (0.07)
0.21 (1.10)
0.34 (1.10)
0.13 (1.10)
0.13 (0.02)
8.56 (0.32)

–1.32
–4.26***
–6.15***
2.08*
3.39***
1.30

Note: Ham-D‫ס‬Hamilton Rating Scale for Depression; BDI-SF‫ס‬Beck Depression Inventory–Short Form; SE‫ס‬standard error; CBT‫ס‬cognitive–
behavioral therapy; t values test the null hypotheses that parameter‫.0ס‬
*PϽ0.05; **PϽ0.01; ***PϽ0.001.

Am J Geriatr Psychiatry 9:3, Summer 2001

231

Desipramine vs. Psychotherapy
In the TDCRP, analyses that considered initial level
of severity found that patients with severe depression
responded more favorably to medication than to psychotherapy, but this effect was not evident in patients
with mild-to-moderate depression.16,22 The recent interest and controversy on this issue23,25–27,60,61 has further
encouraged us to conduct a post-hoc exploratory evaluation of the effect of initial severity on treatment outcome. As described previously, the current study assigned patients with initial Ham-D scores of Յ18 to the
Low Severity group and those with Ham-D scores of
Ն19 to the High Severity group. Pretreatment Ham-D
means for these two groups are included in Table 3. The
range for the Low group was 14–18; for the High group,
19–34.
Growth-curve analyses revealed an interaction between initial level of severity and type of treatment for
both the Ham-D (F[6, 647]‫ ;90.41ס‬P‫ )100.0ס‬and the
BDI-SF (F[6, 1486]‫ ;64.01ס‬P‫ )100.0ס‬measures. Comparisons between treatments for the High- and Low-Severity groups are summarized in Table 3. Looking first at
the Ham-D, there was little difference in improvement
between High- and Low-Severity groups, as reflected in
slope measures, for either the Desipramine-Alone or the
CBT-Alone treatments. In the Combined condition, the
slope of change was noticeably greater for the HighSeverity group than for the Low-Severity group. As
shown in Table 3, post-hoc comparisons, however, did
not reveal significant differences between High and
TABLE 3.

Low groups for any of these treatments. In contrast to
the Ham-D results, post-hoc comparisons for the BDI-SF
in the Combined treatment indicated that the expected
per-session reduction in BDI-SF for subjects with high
initial severity of depression was significantly greater
(0.58 per session) than the rate of improvement (0.29
per session) reported for subjects with low initial severity of depression (t[1486]‫ ;91.2ס‬PϽ0.05). Again,
there was little difference in improvement between the
High- and Low-Severity patient groups in the Desipramine-Alone and CBT-Alone treatment conditions. The
overall pattern of results for both the Ham-D and the
BDI-SF are quite similar, suggesting that the lack of statistical significance of the post-hoc tests on the Ham-D
is likely due to lower power, since these ratings were
obtained at every other intervention session.
Further analyses investigating treatment effects
within High- and Low-Severity groups conveyed results
similar to those described in the investigation of main
treatment outcomes. Table 4 reports comparisons of the
mean rate of change in Ham-D and BDI-SF across treatment conditions, grouped according to initial severity
level. In the High group, depressive symptoms as measured by the Ham-D improved –0.50 per session for the
Combined condition. This is significantly greater than
the –0.19 per-session improvement reported in the Desipramine-Alone condition (t[234]‫ ;59.1ס‬PϽ0.05). Similarly, depressive symptoms in the High group as captured by the BDI-SF improved at a significantly greater

Difference in mean rate of change in Ham-D and BDI-SF between Low- and High-Initial-Severity groups by treatment
condition

Effect
Pretreatment Ham-D, mean‫ע‬SD
Rate of change per session for Ham-D
Desipramine
CBT-Alone
Combined
Variance
Residual
Rate of change per session for BDI-SF
Desipramine
CBT-Alone
Combined
Variance
Residual

Low
(n؄53)

High
(n؄47)

Low vs. High

16.0‫5.1ע‬

21.9‫3.3ע‬

t[647]

–0.20
–0.36
–0.41
0.07
11.67

–0.20
–0.35
–0.31
0.05
9.62

–0.19
–0.36
–0.50
0.10
14.06

0.12
0.28
1.22

–0.10
–0.31
–0.44
0.13
8.56

–0.09
–0.30
–0.29
0.06
6.99

–0.10
–0.33
–0.58
0.12
12.55

0.06
0.11
2.19*

Total

Note: For differences in mean rate of Ham-D and BDI-SF change between Low- and High-Initial-Severity groups across treatment conditions:
F[6, 647]‫ ;90.41ס‬PϽ0.001 and F[6, 1486]‫ ;64.01ס‬PϽ0.001, respectively.
Ham-D‫ס‬Hamilton Rating Scale for Depression; BDI-SF‫ס‬Beck Depression Inventory–Short Form; SD‫ס‬standard deviation; CBT‫ס‬cognitive–
behavioral therapy.
*PϽ0.05.

232

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
rate, –0.58 per session, as opposed to –0.10 per session
for the Desipramine-Alone condition (t[643]‫;21.3ס‬
PϽ0.01). There were no significant differences between the CBT-Alone and the Combined, nor between
the CBT and Desipramine-Alone treatments for the
High-Severity groups.
In contrast, no significant differences in the rate of
change in depressive symptoms as measured with HamD were found across treatments within the Low-Severity
group. However, the estimated reduction in depressive
symptoms in this group as measured by the BDI-SF was
significantly greater in separate comparisons of Desipramine-Alone with CBT-Alone (t[844]‫ ;54.2ס‬PϽ0.05) and
with the Combined treatment (t[844]‫ ;31.2ס‬PϽ0.05). In
the Low group, the estimated per-session changes of
–0.30 in the CBT-Alone and –0.29 in Combined treatment are over three times the –0.09 rate of change reported in the Desipramine-Alone condition. No significant differences were found between the CBT-Alone
and Combined conditions for the Low-Severity groups.
Intent-to-treat analysis: pre–post change in mean
Ham-D and BDI-Long Form scores. Mean scores before and after treatment are also being presented to provide a clearer description of absolute levels of depression. Table 5 contains the means and standard
deviations (SD) of the BDI and the Ham-D from pretest
to endpoint for subjects in the three treatment groups
who received some treatment (N‫ .)001ס‬Both Dropouts
TABLE 4.

and Completers were included in these analyses. “Post-”
evaluations were completed within 10 days of the final
treatment session for all dropouts. The multivariate analysis of variance (MANOVA) for repeated measures is reported in Table 6. An estimate of Wilks’ lambda yielded
an F ratio indicating that the improvement in level of
depression across time was highly significant
(F[2,96]‫ ;37.74ס‬PϽ0.000. A comparison of all three
groups for the Treatment ‫ ן‬Time interaction yielded
an estimate of Wilks’ lambda of borderline significance
(F[2,96]‫ ;70.2ס‬P‫ .)680.0ס‬Turning to the univariate analyses, the F ratio was significant for the BDI (P‫,)120.0ס‬
but not for the Ham-D (P‫ .)001.0ס‬Table 6 also reports
post-hoc comparisons of the three treatment-group
pairs. Looking at the CBT-Alone treatment compared
with Desipramine-Alone, the MANOVA for repeated
measures was highly significant for change across time
(F[2,61]‫ ;15.23ס‬PϽ0.000, and not significant
(F[2,61]‫ ;35.2ס‬P‫ 880.0ס‬for the Treatment ‫ ן‬Time interaction. The univariate analyses for the BDI and HamD yielded a significant interaction for the BDI
(P‫ )040.0ס‬and a marginal value for the Ham-D
(P‫ .)350.0ס‬There was clearly no significant Treatment
‫ ן‬Time interaction in the comparison of the CBT-Alone
and Combined treatment conditions, but, as expected,
the improvement across time was highly significant. A
comparison of the Combined treatment and the Desipramine-Alone again showed the improvement across
time to be highly significant (F[2,66]‫ ;78.82ס‬PϽ0.000).

Comparison (t values) between treatment conditions in the mean rate of change per session by Low- and High-InitialSeverity level for both Ham-D and BDI-SF scores
Ham-D

BDI-SF

Low Severity
(df؄322)

Desipramine vs. CBT
Desipramine vs. Combined
Combined vs. CBT

Low Severity
(df؄844)

High Severity
(df؄643)

1.51
1.01
–0.37

Effect

High Severity
(df؄234)
1.01
1.95*
0.86

2.45*
2.13*
–0.11

1.33
3.12**
1.61

Note: Ham-D‫ס‬Hamilton Rating Scale for Depression; BDI-SF‫ס‬Beck Depression Inventory–Short Form; CBT‫ס‬cognitive–behavioral therapy.
*PϽ0.05; **PϽ0.01.

TABLE 5.

Pre- and endpoint meansؓstandard deviations for the Beck Depression Inventory–Long Form (BDI–LF) and the
Hamilton Rating Scale for Depression (Ham-D), grouped by treatment condition (N؄100)
BDI–LF

Condition

n

Pre-

Desipramine-Alone
Cognitive–Behavioral
Combined treatmenta

33
31
36

22.8‫8.4ע‬
24.3‫5.7ע‬
25.5‫4.7ע‬

Ham-D
Post18.6‫2.7ע‬
16.4‫7.9ע‬
15.7‫6.8ע‬

Pre18.8‫7.4ע‬
18.9‫7.3ע‬
18.7‫3.3ע‬

Post15.0‫2.6ע‬
12.5‫0.7ע‬
12.0‫9.6ע‬

a

Combined treatment includes both cognitive–behavioral therapy and desipramine.

Am J Geriatr Psychiatry 9:3, Summer 2001

233

Desipramine vs. Psychotherapy
The Treatment ‫ ן‬Time interaction for these two groups
was also significant (F[2,66]‫ ;67.3ס‬P‫ .)920.0ס‬Univariate
analyses showed that the Treatment ‫ ן‬Time interaction
effect for the BDI and the Ham-D were both significant
(P‫ 700.0ס‬and 0.048, respectively).
In summary, the intent-to-treat analyses, in which
both Dropouts and Completer subjects are included,
indicated that all three treatments resulted in substantial
improvement. The Combined treatment resulted in significantly greater improvement than the DesipramineAlone condition. Although the univariate analyses suggested that the CBT-Alone condition showed greater
improvement than the Desipramine-Alone condition,
the multivariate analysis did not support this interpretation.
As noted earlier, in order to adjust for the patients
with mild depression, consideration of initial severity
was introduced by comparing High- and Low-Severity
groups by using a cutoff of 19. However, the numbers
in the two severity groups were relatively small, and this
may have resulted in an absence of clear effects. For the
endpoint analysis, we selected 16 as the cutoff point
and analyzed only those data above this point. In doing
so, we had a larger group (n‫ )87ס‬with descriptive statistics more in line with other samples. The resulting
Ham-D mean and SD scores for the three treatment
groups were 19.8‫ ,5.2ע3.02 ,3.3ע‬and 20.0‫ 5.4ע‬for
the CBT-Alone, Combined, and the Desipramine-Alone
groups, respectively.
TABLE 6.

A MANOVA for repeated measures revealed a significant overall Treatment ‫ ן‬Time interaction (F[4,
148]‫ ;528.2ס‬P‫ .)720.0ס‬Univariate analyses indicated
that both the BDI-LF and the Ham-D were significant
(P‫ 600.0ס‬and 0.031, respectively). Post-hoc comparison of the treatment pairs indicated that greater improvement occurred in the Combined than in the Desipramine-Alone condition (F[2,49]‫ ;922.5ס‬P‫.)900.0ס‬
Both the BDI-LF and the Ham-D were significant
(PϽ0.002 and 0.014, respectively). The improvement
for the CBT-Alone was marginally greater than for the
Desipramine-Alone (F[2,49]‫ ;920.3ס‬P‫ .)850.0ס‬Univariate analysis yielded significant differences for the BDILF (PϽ0.031) and the Ham-D (PϽ0.026). There was no
difference between the Combined and the CBT-Alone
conditions.
Interaction of stable dose level with treatment outcome. Although the mean stable dose of desipramine
in this study was consistent with dose levels recommended in past psychopharmacologic reviews,43,44
the argument could be made that the stable dose for
some of our subjects was not at a therapeutic level.
Unfortunately, plasma serum levels at stable dose levels
are not available to address this issue. To allay concerns
about the treatment regimen, we grouped the subjects
according to those who were receiving a stable dose
of 100 mg or more and those receiving less than 100
mg, and then compared their outcomes for the

Multivariate and univariate analyses of variance with repeated measures comparing three treatment groups on the
BDI-LF and the Ham-D: endpoint analysis (N؄100)
Treatment ‫ ؂‬Time Interaction

Change Over Time
Comparison: Analysis Type
All three treatments
Wilks’ lambda (F statistic)
BDI-SF
Ham-D
CBT vs. Desipramine-Alone
Wilks’ lambda (F statistic)
BDI-SF
Ham-D
CBT vs. Combined
Wilks’ lambda (F statistic)
BDI-SF
Ham-D
Combined vs. Desipramine-Alone
Wilks’ lambda (F statistic)
BDI-SF
Ham-D

F[df]

P

F[df]

P

47.73[2,96]
77.61[1,97]
86.31[1,97]

0.000
0.000
0.000

2.07[4,192]
4.03[2,97]
2.36[2,97]

0.086
0.021
0.100

35.51[2,61]
45.78[1,62]
59.69[1,62]

0.000
0.000
0.000

2.53[2,61]
4.39[1,62]
3.90[1,62]

0.088
0.040
0.053

35.49[2,64]
63.22[1,65]
63.15[1,65]

0.000
0.000
0.000

0.52[2,64]
0.67[1,65]
0.04[1,65]

0.597
0.418
0.840

28.87[2,66]
47.68[1,67]
55.25[1,67]

0.000
0.000
0.000

3.76[2,66]
7.62[1,67]
4.07[1,67]

0.029
0.007
0.048

Note: N includes both Dropouts and Completers. Wilks’ lambda is multivariate F statistic.
BDI-SF‫ס‬Beck Depression Inventory–Long Form; Ham-D‫ס‬Hamilton Rating Scale for Depression; CBT‫ס‬cognitive–behavioral therapy.

234

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
three Treatment conditions. Table 7 shows the mean
BDI-LF and Ham-D scores for subjects with pretreatment Ham-D levels at Ն16, grouped according to stable dose level. A MANOVA for repeated measures for
subjects in the High-Dose group revealed that the overall Treatment ‫ ן‬Improvement interaction was highly
significant (F[4,98]‫ ;310.4ס‬P‫ .)500.0ס‬The effect was
apparent for both the BDI-LF (P‫ )100.0ס‬and the HamD (P‫ .)820.0ס‬Post-hoc comparison of treatment pairs
indicated that patients in the Combined treatment
showed greater improvement than the CBT-Alone
(F[2,35]‫ ;715.3ס‬P‫ )140.0ס‬or the Desipramine-Alone
condition (F[2,24]‫ ;434.01ס‬P‫ .)100.0ס‬There was no significant difference between the CBT-Alone and the Desipramine-Alone conditions (F[2,38]‫ ;818.1ס‬P‫,)671.0ס‬
although mean changes for both measures were greater
in the CBT-Alone condition than in the DesipramineAlone condition (P‫ 950.0ס‬for the Ham-D).
The overall comparison of the Treatment ‫ ן‬Improvement interaction for subjects who were receiving
100 mg or less was not significant (F[4,94]‫;519.0ס‬
P‫ .)954.0ס‬The pre–post-treatment change was highly
significant for all comparisons made in all three treatment conditions (PϽ0.001 for all). This analysis was repeated on the full sample, and all results were similar,
except that the difference in rate of improvement between the Combined and CBT-Alone conditions was
marginally significant (F[2,43]‫ ;846.2ס‬P‫.)280.0ס‬
In summary, the data indicated that all three treatTABLE 7.

ment conditions resulted in significant improvement in
depressive symptoms. Comparisons among the treatments showed that the Combined treatment consistently resulted in greater improvement than the
Desipramine-Alone condition. In most analyses, the difference between the CBT-Alone and the combined conditions was negligible. The difference between the CBTAlone and the Desipramine-Alone conditions was less
clear. Some analyses suggested that CBT resulted in
greater improvement than desipramine, particularly as
reflected in the self-report measure, rather than the
clinical rating. Controlling for initial level of depression
and stable medication level did not noticeably affect the
rate of improvement in the medication condition. Posthoc analyses did suggest that the consistently more positive effect of the combined treatment, when compared
with desipramine alone, was due in large measure to
the interaction of severity of depression and size of the
stable medication dose. Thus, more severely depressed
patients in the Combined treatment condition showed
greater improvement than those in either the CBT or
Desipramine-Alone condition when they were on a stable dose of desipramine that was equal to or greater
than 100 mg. Also, the suggested differences between
Desipramine-Alone and CBT-Alone were not apparent in
this select group of high-stable-dose patients. When patients were on a lower stable dose of medication, there
were no significant differences between the three conditions.

Change in depression symptoms in the three treatments for subjects with pretreatment Ham-D scores Ն16, grouped
by level of stable medication dosage
High Medication Dose Ն100 mg
Desipramine
(n؄15)

CBT
(n؄26)

Combined
(n؄12)

P

PreMANOVA
Ham-D
BDI-LF

Post-

Pre-

Post-

Pre-

Post-

Time

Group ‫ ؂‬Time

19.6‫7.3ע‬
23.9‫1.4ע‬

15.5‫4.5ע‬
20.4‫2.5ע‬

19.8‫3.3ע‬
24.8‫9.7ע‬

12.9‫1.7ע‬
16.8‫5.9ע‬

20.7‫0.3ע‬
28.1‫9.7ע‬

10.0‫6.7ע‬
12.2‫0.01ע‬

0.000
0.000
0.000

0.005a
0.028b
0.001b

0.000
0.000
0.000

0.459a
0.187b
0.360b

Low Medication Dose Ͻ100 mg
(n؄11)
MANOVA
Ham-D
BDI-LF

20.6‫7.5ע‬
23.5‫9.5ע‬

17.5‫9.8ע‬
20.1‫0.9ע‬

(n؄26)
19.8‫3.3ע‬
24.8‫9.7ע‬

12.9‫1.7ע‬
16.8‫0.01ע‬

(n؄14)
20.0‫0.2ע‬
25.6‫4.6ע‬

14.3‫1.7ע‬
18.0‫1.7ע‬

Note: Values are mean‫ע‬standard deviation. Ham-D‫ס‬Hamilton Rating Scale for Depression; CBT‫ס‬cognitive–behavioral therapy;
MANOVA‫ס‬multivariate analysis of variance; BDI-LF‫ס‬Beck Depression Inventory–Long Form.
a
Multivariate analysis of variance was used, and P levels reflect an F ratio estimate of Wilks’ lambda. All treatments showed improvement at
PϽ0.005 or better in both multivariate and univariate tests. Group ‫ ן‬Time significance in the High Medication group is due to greater
improvement in the Combined than in either treatment alone.
b
Univariate tests.

Am J Geriatr Psychiatry 9:3, Summer 2001

235

Desipramine vs. Psychotherapy
DISCUSSION
Our findings indicate that over 16–20 weeks of individual treatment, the total sample of older adult outpatients in a current episode of MDD showed substantial
improvement in all three treatment conditions. CBT
combined with desipramine resulted in greater improvement in depression than desipramine administered alone. There was little difference between CBT
alone and CBT combined with desipramine. Looking
just at the self-report measure, the data suggest that patients receiving only cognitive–behavioral therapy improved more than those receiving only desipramine.
However, this effect was not apparent in the Ham-D
ratings.
The low cutoff point for inclusion on the initial
Ham-D might suggest that our sample comprised primarily mildly depressed patients, which may have contributed to the lack of effect between CBT and Desipramine-Alone. To address this question, we completed
post-hoc analyses to control for initial level of depression. Looking first at the growth-curve analysis, the
Ham-D ratings indicate that the Combined treatment is
superior to Desipramine-Alone, but not CBT-Alone in
the High-Severity group (Ham-DϾ18). The trend was
similar for the Low-Severity patients (Ham-DϽ19), but
not statistically significant. For the self-report measure,
the Combined group was again superior to Desipramine-Alone, and not CBT-Alone in the High-Severity
group. In the Low-Severity group, both the CBT-Alone
and the Combined groups were favored over the Desipramine-Alone. The intent-to-treat analysis showed a
similar picture using this severity cutoff, but the data
are not presented here, as explained below.
A question concerning the adequacy of the treatment with desipramine could also be raised. The high
proportion of patients who were on relatively low stable medication levels, in combination with the highly
variable plasma levels of desipramine, suggested that
many of the patients may not have been at therapeutic
level. In attempting to address this question, we first
compared patients with plasma levels Ͼ125 ng/ml with
those who were below this level. We found no significant differences. These data were not reported here because most subjects were not at steady-state medication
levels when the samples were obtained. We decided to
address the question by grouping the patients according
to High stable-dose level (Ն100 mg) or Low stable-dose

236

level (Ͻ100 mg). We assumed that patients in the High
stable medication group were more likely to be at therapeutic level than those in the Low stable medication
group. We evaluated this effect in the intent-to-treat
analysis, using the same severity cutoff mentioned
above, but the cell sizes were small and variable. We
decided to decrease the severity cutoff to 16, which
would give us larger cell sizes in the High-dose group
and yet still have depression levels similar to those reported in other studies. We analyzed the data in the
High-dose and Low-dose group separately. Patients in
the Combined treatment, who were receiving a high
stable dose, clearly showed greater improvement than
patients in either treatment alone. Patients receiving a
high stable dose of desipramine-only were not favored
over those receiving CBT-Alone. In the Low-stable-dose
group, there was no significant difference among the
three groups.
To our knowledge, there are no randomized trials
of depressed older adult outpatients involving pharmacotherapy and individual CBT with which we could
compare our findings. However, Hollon et al.62 have reviewed a number of such studies involving younger patients. They reported that pharmacotherapy-alone did
not prove to be clearly superior to cognitive therapyalone in any of these studies and, further, that a combined treatment, although never inferior, was not
clearly superior to either single modality. Results of the
TDCRP also showed minimal differences over a 16week time-span between pharmacotherapy-plus-clinical
management and either cognitive–behavioral or interpersonal therapy if the effects of initial level of severity
were excluded from the analyses.16,17 When initial severity was considered, TDCRP results suggested better
outcomes on depressive symptoms in the imipramineplus-clinical management condition for patients who
were both severely depressed and functionally impaired. In contrast, DeRubeis and colleagues’ recent
meta-analysis,31 which also included TDCRP data, found
no statistically significant advantage for either antidepressant medication or CBT in the treatment of severely
depressed outpatients.
On the other hand, Thase and his colleagues,63 in
their meta-analysis of several studies comparing psychotherapy and a psychotherapy–pharmacotherapy combination in younger patients, found the combined condition to be superior to psychotherapy in patients with
high severity levels, but not in those with low severity.
Our results are consistent with the finding that a com-

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
bination of psychotherapy–pharmacotherapy is superior to either alone when treating severely depressed
patients. Our data also suggest, however, that if, for
some reason, the dosage is below therapeutic level, this
may not be the case. These findings, in combination
with results from the TDCRP,22 argue that initial level of
severity should be considered as a primary variable in
the design of future studies to evaluate treatment effects
in the older adult. Furthermore, particular attention
should be given to plasma levels of medication at steadystate dose levels.
The discussion of initial severity warrants mention
of our general clinical impression that patients who participated in the current study seemed to have more
complications than patients did in our previous projects
and did not show as much improvement as seen in earlier studies.59,64 A possible contributing factor may be
the delay from initial clinic contact to treatment start
date mentioned earlier. The delay was a direct result of
the extensive pretreatment work-up implemented to
ensure that patients were suitable treatment risks for
desipramine. As we reported earlier, over 40% of the
potential subjects were not enrolled in the study because they no longer met MDD diagnosis or Ham-D and
BDI criteria at re-screening. Therefore, these delays may
have greatly reduced the number of spontaneous responders and rapid responders who entered the treatment program—two groups that often may be included
in clinical trial reports.
A number of questions in addition to this clinical
impression need to be raised regarding the generalizability of these results. For example, the argument could
be raised that the present study does not provide an
adequate appraisal of “pharmacotherapy” as practiced
in clinical settings. Clinical lore supports the importance of individual differences in response to different
antidepressants, and a clinician might reasonably try
several antidepressants before finding one that results
in a positive patient response. Although desipramine is
still useful in practice and is associated with fewer adverse side effects for elderly patients than other TCAs
(e.g., amitriptyline and imipramine), a number of selective serotonin reuptake inhibitors (SSRIs) and other
non-tricyclic agents with even fewer side effects have
been introduced since the study began. The effectiveness of SSRIs (as indicated by their lower dropout rates
among older patients) may ultimately prove somewhat
better than TCAs. Still, there are too few clinical trials
involving older adults that compare SSRIs directly with

Am J Geriatr Psychiatry 9:3, Summer 2001

TCAS to confirm this assumption.9 Furthermore, this
study has been limited to outpatients, most of whom
are in a moderate range of depression, and similar results may not be obtained with more severely depressed
inpatients. Nevertheless, these data do support the feasibility of alternatives to the use of medication in the
treatment of MDD in older adult outpatients. This may
be particularly helpful in instances where the use of
antidepressants is contraindicated because of medical
problems or in situations where older adult patients are
reluctant to comply with a drug treatment regimen because of disturbing side effects.
Our results also may not apply to more socially, ethnically, and clinically diverse groups of depressed older
adults. The majority of patients recruited into the study
were well-educated, middle-class, self-identified white,
older adults in their 60s, who were without serious cognitive or physical impairment. Thus, our results may apply only to psychologically-minded older adults with
milder forms of depression, seen in outpatient settings.
Future studies, as emphasized in the recent Consensus
Update Conference,6,7 need to investigate more heterogeneous study populations encompassing ethnic and racial diversity, persons with disabilities, and a variety of
medical illnesses and comorbidities, to help ensure
clinical generalizability. The Conference also stressed
the need for studies of patients 70 years or older, and
the need to broaden outcomes beyond depressive
symptoms alone to quality-of-life components, healthcare utilization, social and occupational functioning, interpersonal relationships, and coping skills. Studies examining these various issues will help clinical research
and practice move toward systematic treatment selection by identifying critical individual-difference variables that influence treatment adherence, course, and
outcome, and help determine which techniques work
best with which patients.
Finally, caution is needed in interpreting the efficacy of desipramine on the basis of these data, since
determination of plasma levels of desipramine at stable
medication levels were not available for all patients. The
plasma level at 1 month was obtained before most patients were receiving a stable dose. Many individuals
had dose levels below the recommended range, and it
may be that they were not at the recommended therapeutic level. However, in this particular sample, the psychiatrists met resistance from many of the patients in
attempting to increase dose levels above those attained,
despite frequent consultations with the internist for

237

Desipramine vs. Psychotherapy
remedies to counteract side effects. Therefore, the psychiatrists proceeded, using their best clinical judgment
as to what would be the most likely regimen to achieve
maximal benefit without raising side effects to a level
such that patients would discontinue treatment. Ironically, this issue may reflect an important confound pertaining to individual differences that is often neglected
in drug outcome studies. Individuals who refuse to comply with a medication regimen, and therefore may not
attain therapeutic plasma levels, are more likely to be
the kind of patients who benefit minimally from any
type of therapy. On the other hand, patients who are
willing and able to cope with the uncomfortable symptoms of side effects of medication are more likely to be
the more resilient individuals, who with proper guidance and support, can overcome immense difficulties
and disruptions in their lives.
In summary, our study has found that CBT in combination with desipramine is more efficacious than desipramine-alone in reducing symptoms of depression in
older adults, but the efficacy of CBT-alone over desipramine-alone is less consistently demonstrated in the various analyses completed. These results are in contrast to
the comments of the 1991 Consensus Conference regarding the usefulness of psychotherapy. As Niederehe10 noted, the Consensus Conference “appeared to
emphasize [psychotherapy’s] adjunctive role” or its use
“as a second-line choice when antidepressant medications are not suitable (e.g., because of medical complications or medication intolerance) or simply not preferred,” and “suggested that antidepressant medications
are useful across the range of severity in depression and
are required in its more severe forms, but that the use-

fulness of psychotherapy may be limited primarily to
cases of mild-to-moderate depression (p S69).” Our results provide additional evidence for the assertion that
psychosocial treatments have an essential role in the
treatment of late-life depression, without relegating
these interventions to adjunctive positions. Although
the combination treatment appeared most efficacious
in our findings, on both the Ham-D and the BDI-SF, the
only instance where the combined treatment was significantly different from the CBT-alone occurred in the
more severely depressed group, who were receiving
higher levels of medication. Some researchers have suggested that combined treatments may well be considered the standard of care in clinical practice.9,10 Reynolds and his colleagues14,15 have found that a
combination of IPT and nortriptyline was efficacious in
both the acute and the maintenance treatment of elderly patients experiencing recurrent depression. However, more studies are needed to reach a definitive conclusion on this matter. Nevertheless, the current study
adds to the growing body of evidence suggesting that
CBT, either alone or in combination with appropriate
psychotropic medication, deserves greater attention in
the treatment of late-life depression.
The authors acknowledge the generous assistance
of Leo Hollister, M.D., in the planning of this research,
as well as Jerome Yesavage, M.D., Richard Vieth, M.D.,
and Lissy Jarvik, M.D., Ph.D., for their helpful consultation on issues pertaining to the medication regimen.
The desipramine was provided by Merrill-Dow
Chemical.
This work was supported principally by Grant
#MH37196 from the National Institute of Mental
Health.

References
1. Parmelee PA, Katz IR, Lawton P: Depression among institutionalized aged: assessment and prevalence estimation. J Gerontol
1989; 44:22–29
2. Myers JK, Weissman MM, Tischler GL: Six-month prevalence of
psychiatric disorders in three communities, 1980–1982. Arch
Gen Psychiatry 1984; 41:959–967
3. Blazer DG, Hughes DC, George LK: The epidemiology of depression in an elderly community population. Gerontologist 1987;
27:281–287
4. Gurland GJ, Cross PS, Katz S: Epidemiological perspectives on
opportunities for treatment of depression. Am J Geriatr Psychiatry 1996; 4(suppl1)):S7–S13
5. Schneider LS, Reynolds CF III, Lebowitz BD, et al (eds): Diagnosis
and Treatment of Depression in Late Life: Results of the NIH
Consensus Development Conference. Washington, DC, American Psychiatric Press, 1994
6. Katz IR, Alexopoulos GS: Consensus Update Conference: Diag-

238

nosis and Treatment of Late-Life Depression. Am J Geriatr Psychiatry 1996; 4(suppl1):S1–S10
7. Lebowitz BD, Pearson JL, Schneider LS, et al: Diagnosis and treatment of depression in late life: Consensus Statement Update.
JAMA 1997; 278:1186–1190
8. Friedhoff AJ: Consensus Development Conference Statement: Diagnosis and treatment of depression in late life, in Diagnosis and
Treatment of Depression in Late Life: Results of the NIH Consensus Development Conference. Edited by Schneider LS, Reynolds
CF III, Lebowitz BD, et al. Washington, DC, American Psychiatric
Press, 1994, pp 493–511
9. Schneider LS: Pharmacologic considerations in the treatment of
late-life depression. Am J Geriatr Psychiatry 1996; 4(suppl1):S51–
S65
10. Niederehe G: Psychosocial treatments with depressed older
adults. Am J Geriatr Psychiatry 1996; 4(suppl1):S66–S78
11. Scogin F, McElreath L: Efficacy of psychosocial treatments for

Am J Geriatr Psychiatry 9:3, Summer 2001

Thompson et al.
geriatric depression: a quantitative review. J Consult Clin Psychol
1994; 62:69–74
12. Sloane RB, Staples FR, Schneider LS: Interpersonal therapy vs.
nortriptyline for depression in the elderly, in Clinical and Pharmacological Studies in Psychiatric Disorders. Edited by Burrows
GD, Norman TR. London, UK, John Libby, 1985, pp 344–346
13. Beutler LW, Scogin F, Kirkish P, et al: Group cognitive therapy
and alprazolam in the treatment of depression in older adults. J
Consult Clin Psychol 1987; 55:550–556
14. Reynolds CF III, Frank E, Perel JM, et al: Combined pharmacotherapy and psychotherapy in the acute and continuation treatment of elderly patients with recurrent major depression: a preliminary report. Am J Psychiatry 1992; 149:1687–1692
15. Reynolds CF III, Frank E, Perel JM, et al: Nortriptyline and interpersonal psychotherapy as maintenance therapies for recurrent
major depression: a randomized controlled trial in patients older
than 59 years. JAMA 1999; 281:39–45
16. Elkin I, Shea T, Watkins J, et al: National Institute of Mental Health
Treatment of Depression Collaborative Research Program: general effectiveness of treatments. Arch Gen Psychiatry 1989;
46:971–982
17. Gibbons RD, Hedeker D, Elkin I, et al: Some conceptual and statistical issues in analysis of longitudinal psychiatric data. Arch
Gen Psychiatry 1993; 50:739–750
18. Watkins JT, Leber WR, Imber SD, et al: Temporal course of change
in depression. J Consult Clin Psychol 1993; 61:858–864
19. Hamilton M: Development of a rating scale for primary depressive illness. Br J Social Clin Psychol 1967; 6:278–296
20. Beck AT, Ward CH, Mendelson M, et al: An inventory for measuring depression. Arch Gen Psychiatry 1961; 4:561–571
21. Endicott J, Spitzer RL, Fleiss JL, et al: The Global Assessment
Scale: a procedure for measuring overall severity of psychiatric
disturbance. Arch Gen Psychiatry 1976; 33:766–771
22. Elkin I, Gibbons RD, Shea MT, et al: Initial severity and differential
treatment outcome in the NIMH Treatment of Depression Collaborative Research Program. J Consult Clin Psychol 1995;
63:841–847
23. Elkin I, Gibbons RD, Shea MT, et al: Science is not a trial (but it
can sometimes be a tribulation). J Consult Clin Psychol 1996;
64:92–103
24. Gallagher-Thompson D, Thompson LW: Psychotherapy with
older adults in theory and practice, in Comprehensive Textbook
of Psychotherapy. Edited by Bongar B, Beutler L. New York, Oxford University Press, 1994, pp 359–379
25. Jacobson NS, Hollon SD: Cognitive/behavior treatment vs. pharmacotherapy: now that the jury’s returned its verdict, it’s time
to present the rest of the evidence. J Consult Clin Psychol 1996;
64:74–80
26. Jacobson NS, Hollon SD: Prospects for future comparisons between drugs and psychotherapy: lessons from the CBT-vs.-pharmacotherapy exchange. J Consult Clin Psychol 1996; 64:104–108
27. Klein DF: Preventing hung juries about therapy studies. J Consult
Clin Psychol 1996; 64:81–87
28. McNally RJ: Methodological controversies in the treatment of
panic disorder. J Consult Clin Psychol 1996; 64:88–91
29. American Psychiatric Association: Practice Guidelines for Major
Depressive Disorder in Adults. Am J Psychiatry 1993; 150(April
suppl.)
30. Agency for Health Care Policy and Research: Clinical Practice
Guideline (No. 5). Depression in Primary Care, Vol. 2: Treatment
of Major Depression. Washington, DC, U.S. Government Printing
Office, 1993
31. DeRubeis RJ, Gelfand LA, Tang TZ, et al: Medications vs. cogni-

Am J Geriatr Psychiatry 9:3, Summer 2001

tive behavior therapy for severely depressed outpatients: metaanalysis of four randomized comparisons. Am J Psychiatry 1999;
156:1007–1013
32. Reynolds CF III, Schneider LS, Lebowitz BD, et al: Treatment of
depression in elderly patients: guidelines for primary care, in
Diagnosis and Treatment of Depression in Late Life. Edited by
Schneider LS, Reynolds CF III, Lebowitz BD, et al. Washington,
DC, American Psychiatric Press, 1994, pp 463–490
33. Keller MB, Harrison WM, Fawcett JA, et al: Treatment of chronic
depression with sertraline or imipramine: preliminary blinded
response rates and high rates of undertreatment in the community. Psychopharmacol Bull 1995; 31:205–212
34. Bushell J, Newton W: SSRI or tricyclics for depression? J Fam
Pract 1996; 43:345–346
35. Klein DN, Schatzberg AF, McCullough JP, et al: Early- vs. late-onset
dysthymic disorder: comparison in outpatients with superimposed major depressive episodes. J Affect Disord 1999; 52:187–
196
36. Spitzer RL, Endicott J, Robins E: Research diagnostic criteria: rationale and reliability. Arch Gen Psychiatry 1978; 35:773–782
37. Endicott J, Spitzer RL: A diagnostic interview for affective disorders and schizophrenia. Arch Gen Psychiatry 1978; 35:837–844
38. Folstein MF, Folstein SE, McHugh PR: Mini-Mental State: a practical method for grading the cognitive state of patients for the
clinician. J Psychiatr Res 1975; 12:189–198
39. Alexopoulos GS: Treatment of depression, in Clinical Geriatric
Psychopharmacology. Edited by Salzman C. Baltimore, MD, Williams & Wilkins, 1992, pp 137–174
40. Jenike MA: Drug treatment of depression. Topics in Geriatrics
1983; 2:9–11
41. Fawcett J, Epstein P, Fiester SJ, et al: Clinical management: imipramine/placebo administration manual: NIMH Treatment of Depression Collaborative Research Program. Psychopharmacol Bull
1987; 23:309–324
42. Potter WZ: Treatment of depression: a 25-year review of tricyclic
antidepressants with Norpramin (desipramine hydrochloride tablets, USP) as a model. Princeton, NJ, Excerpta Medica, 1989, pp
1–16
43. Salzman C: Clinical Geriatric Psychopharmacology. New York,
McGraw-Hill, 1984
44. Jenike MA: Handbook of Geriatric Psychopharmacology. Littleton, MA, PSG Publishing Co., Inc., 1985
45. Beck AT, Rush J, Shaw BF, et al: Cognitive Therapy of Depression.
New York, Guilford, 1979
46. Thompson LW, Gantz F, Florsheim M, et al: Cognitive–behavioral
therapy for affective disorders in the elderly, in New Techniques
in the Psychotherapy of Older Patients. Edited by Myers WA.
Washington, DC, American Psychiatric Press, 1991, pp 3–20
47. Lewinsohn P: A behavioral approach to depression, in The Psychology of Depression: Contemporary Theory and Research. Edited by Friedman R, Katz M. New York, Wiley, 1974, pp 157–176
48. Gallagher D, Thompson LW: Depression in the Elderly: A Behavioral Treatment Manual. Los Angeles, CA, Univ. of Southern California Press, 1981
49. Thompson LW, Gallagher-Thompson D, Dick L: Cognitive–Behavioral Therapy for Late-Life Depression: A Therapist’s Manual.
Palo Alto, CA, VA Palo Alto Health Care System and Stanford University, 1996
50. Spitzer RL, Endicott J: SADS–Change Interview. New York, New
York State Psychiatric Institute, 1977
51. Teri L, Curtis J, Gallagher-Thompson D, et al: Cognitive–behavior
therapy with depressed older adults, in Diagnosis and Treatment
of Depression in Late Life: Results of the NIH Consensus Confer-

239

Desipramine vs. Psychotherapy
ence. Edited by Schneider LS, Reynolds CF III, Lebowitz BD, et
al. Washington, DC, American Psychiatric Press, 1994, pp 279–
291
52. Beck AT, Beck RW: Screening depressed patients in family practice: a rapid technique. Postgrad Med 1972; 52:81–85
53. Beck AT, Steer RA, Garbin MG: Psychometric properties of the
Beck Depression Inventory: twenty-five years of evaluation. Clin
Psychol Rev 1988; 8:77–100
54. Rogosa DR, Brandt D, Zimowski M: A growth curve approach to
the measurement of change. Psychol Bull 1982; 92:726–748
55. Rogosa DR, Saner HM: Longitudinal data analysis examples with
random coefficient models. J Educ Behav Stat 1995; 20:149–170
56. Laird NM, Ware JH: Random-effects models for longitudinal data.
Biometrics 1982; 38:963–974
57. Verbeke G, Molenberghs G: Linear mixed models in practice: an
SAS-oriented approach, in Lecture Notes in Statistics, 126. New
York, Springer-Verlag, 1997
58. Systat: Systat for Windows: Statistics, Version 5. Evanston, IL, Systat, Inc., 1992

240

59. Thompson LW, Gallagher D, Breckenridge JS: Comparative effectiveness of psychotherapies for depressed elders. J Consult Clin
Psychol 1987; 55:385–390
60. Elkin I: The NIMH Treatment of Depression Collaborative Research Program: where we began and where we are, in Handbook of Psychotherapy and Behavior Change. Edited by Bergin
AE, Garfield SL. New York, Wiley, 1994, pp 114–138
61. Thase ME: After the fall: perspectives on cognitive–behavioral
treatment of depression in the “post-collaborative” era. The Behavior Therapist 1994; 17:48–52
62. Hollon SD, DeRubeis RJ, Evans MD, et al: Cognitive therapy and
pharmacotherapy for depression: singly or in combination? Arch
Gen Psychiatry 1992; 49:774–781
63. Thase ME, Greenhouse JB, Frank E, et al: Treatment of major
depression with psychotherapy or psychotherapy–pharmacotherapy combinations. Arch Gen Psychiatry 1997; 54:1009–1015
64. Gallagher D, Thompson LW: Treatment of major depressive disorder in older adult outpatients with brief psychotherapies. Journal of Psychotherapy: Theory, Research, and Practice 1982;
19:482–490

Am J Geriatr Psychiatry 9:3, Summer 2001