Severity of Depression and Response

to Cognitive Behavior Therapy

Michael E. Thase, M.D., Anne D. Simons, Ph.D., Jack Cahalane, M.S.W.,

Janice McGeary, B.S.N., and Tim Harden, B.S.

Objective: The authors studied the relationship between clinical severity ofdepression and
response to cognitive behavior therapy. Method: Fifty-nine outpatients with major depres-
sion with endogenous features, according to Research Diagnostic Criteria, were stratified
into more severe (score of2O or more on the Hamilton Rating Scale for Depression; N=38)
or less severe (Hamilton score of 1 9 or less; N=21) subgroups. Patients were treated with a
1 6-week, 20-session cognitive behavior therapy protocol. Outcome was assessed with the
Hamilton scale, the Global Adjustment Scale, and the Beck Depression Inventory. Results:
The more severe group was significantly more symptomatic across the 1 6-week protocol and
had a significantly lower response rate on the Beck inventory (50% versus 81 %). However,
the groups did not significantly differ at end point on any of the three measures, and they
showed comparable rates of symptomatic improvement (i.e., percent change in scores and
interactions between severity classification and time). Conclusions: These results partially
replicate the National Institute of Mental Health ‘s Treatment of Depression Collaborative
Research Program ‘s findings of poorer response to cognitive behavior therapy in patients
with Hamilton scale scores of 20 or more. However, both groups experienced robust and
clinically significant reductions in depressive symptoms, and the response of the more severe
patients in the current study could hardly be considered poor. While these findings do not
support the view that a Hamilton scale score of2O or more is a relative contraindication for
cognitive behavior therapy, the symptoms ofthe more severely depressed patients did tend to
remit less completely (particularly on the Beck inventory) and thus these patients may benefit
f rom a more intensive or extended course of therapy.
(AmJ Psychiatry 1991; 148:784-789)

A lthough cognitive behavior therapy is perhaps the mine)

more
was
severe
found
depression
to be significantly
than in less severe
bess effective
depression.
in
best studied psychological treatment of unipolar
major depression (1), significant questions persist Further, cognitive behavior therapy was not found to
about its efficacy in more severe depressive states (2, differ significantly from a placebo-clinical management
3). Such concern may be reinforced by the recently condition (4). These results, which represent the first
reported findings of the Treatment of Depression Cob- published negative findings pertaining to cognitive be-
labonative Research Program of the National Institute havior therapy, appear to be at variance with findings
of Mental Health (4), in which cognitive behavior ther- of earlier trials (5, 6) that examined the efficacy of
apy (unlike interpersonal psychotherapy and imipra- cognitive behavior therapy in outpatients with moder-
ate to severe depression. Moreover, the results did not
seem consistent with our clinical experience. Neverthe-
less, because the Collaborative Research Program
Received April 30, 1990; revision received Nov. 1, 1990; accepted study was a well-controlled, randomized, mubtisite
Jan. 17, 1991. From the Department of Psychiatry, University of
clinical trial conducted under the auspices of the Na-
Pittsburgh School of Medicine, Western Psychiatric Institute and
Clinic. Address reprint requests to Dr. Thase, Department of Psy- tionab Institute of Mental Health, such findings are
chiatry, University of Pittsburgh School of Medicine, Western Psy- likely to have a significant impact on the field.
chiatric Institute and Clinic, 381 1 O’Hara St., Pittsburgh, PA 15213. We therefore examined the relationship between se-
Supported in part by NIMH grants MH-41884 and MH-30915.
verity of depression and clinical outcome in a consec-
The authors thank Joseph Howell, Ph.D., Lisa Keller, M.S.N.,
utive series of depressed outpatients treated with cog-
Jean Smith, M.S.N., and Lisa Stupar for their assistance with this
research. nitive behavior therapy. We used the definitions of
Copyright © 1991 American Psychiatric Association. severity and response employed by the Collaborative

784 Am j Psychiatry 148:6, June 1991

 THASE, SIMONS, CAHALANE, ET AL.

TABLE 1. Clinical Characteristics of 59 Depressed Patients Treated With Cognitive Behavior Therapy

Total Mo re Severe Depres sion Less Severe Depress ion’

(N=S9) (N=38) (N=21) .
Statistical
0/
Item N Mean SD N % Mean SD N 0/ Mean SD Analysis
Age (years)
Current 37.7 8.5 39.8 7.9 34.0 8.5 t2.6”
Atonset 31.0 10.9 32.3 11.1 28.7 10.5 t1.2
Sex 203
Male 15 25 8 21 7 33
Female 44 75 30 79 14 67
Marital status 258
Single, never
married 17 29 10 26 7 33
Married 29 49 16 42 13 62
Separated/divorced 10 17 9 24 1 5
Widowed 3 5 3 8 0 0
Employment status x2 10.1’
Employed outside
home 44 75 29 76 15 71
Homemaker S 8 S 13 0 0
Student 4 7 0 0 4 19
Other 6 10 4 11 2 10
Education (years) 14.3 2.4 14.5 2.6 13.9 2.1 t0.9
High school 14 24 8 21 6 29
Some college 22 37 14 37 8 38
College graduate 11 19 6 16 5 24 22 7
Postgraduate work 12 20 10 26 2 10 X
Previous episodes of
depression 1.5 1.4 1.4 1.5 1.6 2.5 t0.3
RDC subtype
Endogenous 230
Probable 9 15 3 8 6 29
Definite SO 85 35 92 15 71
Occurrence 2l9
Single episode 28 47 15 39 13 62
Recurrent 31 53 23 61 8 38
Pretreatment scores
Hamilton depres-
sion scale 21.1 3.4 23.3 3.0 17.1 1.8 t=9.9’1
GAS 52.4 8.0 50.6 7.9 55.5 7.2 t=2.4’
Beck Depression
Inventory 26.5 7.5 28.9 6.6 22.4 7.0 t”3.6”
More severe=Hamilton depression score of 20 or more; less severe=score of 19 or less.
b<001
Cp<0#{149}05#{149}

d<0001

Research Program, as well as several alternative meth- less, our protocol recruits from the same patient pop-
ods for classifying treatment outcome. ubation as the earlier collaborative study and was the
principal referral site for research-eligible depressed
patients during the 24-month period in which this
METHOD study group was collected.
Patients were selected for inclusion in the study on
The patients included in this report were 59 subjects the basis of the following syndromab and symptomatic
consecutively admitted to the Psychobiology of Recov- criteria: DSM-III-R diagnosis of major depression, di-
ery in Depression protocol, an ongoing longitudinal agnosis of probable or definite endogenous subtype of
investigation of the recovery process in unmedicated primary major depression (nonbipohan, nonpsychotic)
depressed outpatients. This protocol is being con- according to the Schedule for Affective Disorders and
ducted in the Cognitive Therapy Clinic of the Mood Schizophrenia and Research Diagnostic Criteria (RDC),
Disorders Module of the Western Psychiatric Institute and Hamilton Rating Scale for Depression (7) scone of 15
and Clinic. The Mood Disorders Module was also one or more. Chronically depressed patients were excluded
of the three clinical sites of the Collaborative Research from the study; all patients’ index episodes were 18
Program study. However, our study protocol began months or less in duration. These criteria were developed
about 1 year after the collaborative study ended, and to yield a group of moderately to severely depressed out-
there is no overlap in personnel on patients. Neventhe- patients with a relatively high probability of psychobio-

Am J Psychiatry 148:6, June 1991 785

DEPRESSION AND COGNITIVE BEHAVIOR THERAPY

TABLE 2. Test Scores at End of Treatment for 59 Depressed Patients Treated With Cognitive Behavior Therapy

More Severe Depression’ Less Severe Depressiona

(N=38) (N=21)
Statistical
Item N 0/ Mean SD N 0/ Mean SD Analysis
Categorical response
Hamilton depression score6 24 63 18 86 x22.3
Beck depression score9 19 50 17 81 x24.2b
End-point score
Hamilton depression scale 6.0 5.7 3.9 5.1 t1.S
GAS 81.7 14.7 87.5 12.2 t=1.S
Beck Depression Inventory 9.2 8.6 5.2 6.8 t1.8
Percent change in score
Hamilton depression scale 73.6 26.2 77.9 27.1 t0.6
GAS 64.5 36.5 60.0 31.4 t=0.S
Beck Depression Inventory 66.6 30.9 76.4 28.9 t1.2
aMore severe=Hamilton depression score of 20 or more; less severe=score of 19 or less.
m’p<O.OS.

logical disturbance. Patients were not selected on the ba- pretreatment symptoms on both the GAS and the Beck
sis of suitability for cognitive behavior therapy. inventory. In addition, there were trends (p<O.lO) sug-
The criteria were applied to 1 10 unipoham depressed gesting that the more severe group was more likely to
outpatients who were consecutively admitted during a be separated on divorced and was more likely to meet
24-month period; 66 outpatients were considered eli- the RDC for definite endogenous subtype.
gible for participation following clinical evaluation. The relationship between severity of depression and
The principal reasons for excluding patients were response to cognitive behavior therapy was evaluated
chmonicity of the index episode on insufficient number by a series of six anahyses. First, simple correlations
of endogenous symptoms for the RDC diagnosis. Two were performed for the relationship between pre- and
eligible patients declined to participate in favor of posttmeatment scores on the Hamilton scale, GAS, and
pharmacological treatment, and two other potentially Beck inventory. Second, response at termination was
eligible patients were referred for immediate treatment defined by the two outcome criteria used in the collab-
because of their degree of incapacity or suicidality. Six- orative study: Hamilton scale score of 6 or less and
ty-two patients provided written informed consent for Beck inventory scone of 9 or less. Response rates were
research participation. calculated for all 59 patients (by using end-point scones
After initial diagnosis and medical screening, persis- for the nine patients who terminated the study prema-
tence of depression was evaluated over a 14-day drug turely) ; chi-squane statistics (one-tailed, corrected for
and alcohol washout period. Sixty patients still met discontinuity) were used to test the significance of dif-
entry criteria following the washout. They next began femences in response mates between the more and less
a 16-week, 20-session course of cognitive behavior severe groups. Third, pretreatment Hamilton scale,
therapy conducted by experienced therapists who me- GAS, and Beck inventory scones were compared in pa-
ceived ongoing supervision. Clinical response was tients categorized as responders or nonmespondems by
evaluated by independent raters according to the these definitions, through a series of two-tailed t tests.
Hamilton depression scale and the Global Assessment Outcome also was assessed by comparing week 16 (on
Scale (GAS) (8) and by the patient-rated Beck Depres- end-point) scores and percent change in scores on de-
sion Inventory (9). Neither the raters nor the therapists pression measures (Hamilton scale, GAS, and Beck in-
were informed of the main hypotheses of the study, ventory) for the two study groups; two-tailed t tests
including plans to test clinical response as a function of were used.
severity. One patient withdrew from the protocol after Finally, analyses of covaniance (ANCOVA) (covani-
one session of therapy. The remaining 59 patients ate=age) with repeated measures for time were per-
completed a minimum course of 10 sessions of treat- formed by comparing the more and less severe groups
ment with cognitive behavior therapy; 50 (85%) com- on the Hamilton scale, GAS, and Beck inventory
pleted the entire protocol. The nine patients who ten- across the 16-week protocol. When significant main
minated the protocol prematurely did not differ from effects or interactions (Severity Group by Time) were
the SO who completed the protocol on any of the base- observed, Student-Newman-Keubs tests were planned
line measures. to further examine the difference between groups on
In keeping with the methods of Elkin et al. (4), the across time points. With a total of 59 patients (and
study group was stratified into more severe (Hamilton harmonic mean cell sizes of N=27 patients), the
scale score of 20 or more; N=38) and less severe planned parametric analyses had 80% statistical
(Hamilton scale score of 19 or less; N=21) groups. As power to detect effect sizes of d=0.45 (abpha=0.0S,
summarized in table 1, the more severe group was sig- two-tailed test) (10). This meant that we had the abil-
nificantly olden and manifested significantly greaten ity to detect in a reliable manner between-group dif-

786 Am J Psychiatry 148:6, June 1991

 THASE, SIMONS, CAHALANE, ET AL.

ferences of 3 points on the Hamilton scale and about FIGURE 1. Test Scores Over 16 Weeks for 59 Depressed Patients
4.5 points on the Beck inventory at posttmeatment on, Treated With Cognitive Behavior Therapy
alternatively, to detect a between-group difference of
about 15% in change in scones. Thus, we considered w 30
1
the size of the study group to be sufficient to evaluate MORE SEVERE (n=38)
4 -.--

our hypothesis that more severely depressed patients C.) ‘ -0-- LESS SEVERE (n=21)
would respond to cognitive behavior therapy as well as zCl) 20
less severely depressed patients. oz
I-0
=!C1)
gio
RESULTS Xi
0.
w
Pretreatment scores on both the Hamilton scale and
Beck inventory did not correlate significantly with end- oa 2a 4a 6 8 i’o’ i i4 16
of-treatment scores on these measures (n=0.13 and 100
0.15, respectively, df=57, pO.2O). By contrast, a
weak but statistically significant correlation was found
between pne- and posttneatment GAS scores (r=0.27, 80
df=57, p=O.O4). C,)
A total of 42 patients (71%) scored 6 on bess on the 4
Hamilton scale at posttneatment, and 36 patients
60
(61%) scored 9 on less on the Beck inventory. As sum-
marized in table 2, significantly fewer patients in the
more severe group met the treatment response criteria 40 i- I I I I 1

of the Beck inventory. In addition, fewer patients in the oa 2 4 6 8 10 12 14a16

more severe group had final Hamilton scale scores of 6
on less; however, this difference was not statistically z
significant (p=O.lS). 0
C,)>.
Patients who were subsequently classified as nonne- Cl)120
sponders on the Hamilton scale were significantly wo
ii-
more symptomatic, according to pretreatment GAS
scores, than responders (mean±SD scores=49.1 ±6.9 LIJW
> 10
versus 53.7±8.1; t2.0, df=57, p0.OS). The pa-
tients classified as treatment responders on the Hamib- C.)
ton scale did not differ on pretreatment assessments on w
either the Hamilton scale on the Beck inventory. Re-
0 oa 2 4 6 8 10 12 14 16
spondens according to Beck inventory criteria were sig-
WEEKS OF TREATMENT
nificantly bess symptomatic at pretreatment than non-
responders on both the Hamilton scale (scones=20.2± ‘Significant between-group differences (Student-Newman-Keuls
3.9 versus 22.4±3.8; t2.1, df=57, p=0.04) and the tests).
GAS (54.8±7.7 versus 48.6±6.9; t=3.1, df=57, p<
0.01). Pretreatment Beck inventory scores did not dif-
ferentiate between patients classified as responders ac- ton scale: F=7.4, df=1, 56, pO.OO9; GAS: F4.7,
cording to either the Hamilton scale on the Beck
df=1, 56, pO.O3; Beck inventory: F=4.0, df=1, 56,
inventory.
p=O.OS). However, theme were no significant interac-
Symptomatic status at the end of treatment, as re-
tions between severity classification and time.
flected by mean end-point scones on the Hamilton
Because the groups appeared to improve at relatively
scale, GAS, and Beck inventory, did not significantly
parallel rates, the ANCOVAs were repeated after co-
differ between groups (table 2). However, in each case,
varying for week 0 (pretreatment) severity on each de-
the less severe group manifested less pathological mean
pendent measure. This permitted a reexamination of
scores. The percent change in scones, which reflects
change relative to pretreatment status, did not signifi- the between-group effects after taking into account mi-
candy differ between the more and less severe groups tial differences in severity. The results of the repeat
(table 2). analyses indicated that the more and less severe groups
Across the 16-week protocol, Hamilton scale, GAS, did not differ significantly in subsequent outcome on
and Beck inventory scores improved significantly (ac- any of the dependent measures after the pretreatment
cording to ANCOVA) in both groups (figure 1; differences were controlled (ANCOVA group effects;
F=70.6, 66.5, and 66.5, respectively, df=8, 456, all p Hamilton scale: F0.13, df=1, 55, p=O.72; GAS:
values <0.0001). Significant main effects for severity F=0.97, df=1, 55, pO.33; Beck inventory: F0.S0,
group also were found on all three measures (Hamil- df=1, 55, p=O.48).

Am J Psychiatry I 48:6, June 1991 787

DEPRESSION AND COGNITIVE BEHAVIOR THERAPY

DISCUSSION of 50 patients [40%]; 2=10.7, df=1, p<O.OO1). Sec-

ond, the relative responsiveness of the more severely
We found some further evidence of a relationship depressed subgroup was higher in our study. Several
between pretreatment severity of depression and factors might have accounted for the difference in re-
poorer response to cognitive behavior therapy. How- sults. In the absence of placebo- and tnicyclic-tneated
ever, the findings were not consistent across outcome comparison groups, it is impossible to gauge the inher-
indicators and appeared dependent upon which method ent potential for positive outcome of this subgroup.
of analysis was used. For example, several outcome clas- Further, patients in the present study were treated in a
sifications based on absolute score values (i.e., responder specialty clinic devoted to research using cognitive be-
versus nonresponder comparisons, the proportion of pa- haviom therapy, whereas patients in the collaborative
tients who met arbitrary response criteria, and the anal- study were randomly assigned to treatment in a clinic
ysis of variance main effect for severity group) indicated that was not committed to a particular type of therapy.
that the more severe group had a less complete response Treatment response mates are generally higher in
to treatment. The major difference between severity “open-babel” trials (11). Other differences might in-
groups was reflected in the repeated-measures analysis of dude our exclusion of patients with RDC diagnoses of
assessment scores across the course of treatment. These nonendogenous depression, the inclusion of more
between-group differences were negated when level of
chronically depressed patients in the collaborative
pretreatment severity was covanied. Further, analyses
study, on subtle differences in therapist competence.
based on rates of change or relative differences (i.e., pen-
We hope to examine these possibilities, as well as other
cent change in scores on the ANCOVA interaction term)
explanations, in collaboration with our colleagues
showed the groups to have equivalent responses. In es-
from the Pittsburgh component of the Collaborative
sence, the more severe group experienced a rate of symp-
Research Program.
tom reduction during cognitive behavior therapy that
Until such data are forthcoming, we propose that
was comparable to the rate for the less severe patients,
the combination of ongoing, weekly therapy supervi-
although the former group did not fully make up for the
sion and the provision of treatment in a specialty clinic
difference in pretreatment symptoms. We therefore con-
milieu made the critical difference in treating more se-
dude that we have only partially replicated an associa-
verely depnesse1 patients in the present study. The
tion between severity of depression and poorer response
weekly supervision received by the therapists in the
to cognitive behavior therapy.
present study may be particularly salient, since Shaw
Are the differences in outcome of the more and less
(15) has noted that patient difficulty may adversely
severely depressed patients clinically significant? At
affect the technical quality of cognitive behavior then-
one level, our findings do not support the notion that
apy. Thus, therapists may require more intensive su-
a simple severity criterion, such as a Hamilton scale
pervision in order to maintain optimal cognitive be-
score of 20 or more, can be used to identify patients
haviom therapy with more severely depressed patients.
who respond poorly to cognitive behavior therapy and
Alternatively, the espnit de corps of a specialty clinic
thus should be treated with an alternative modality,
may help the therapist and patient to continue to en-
such as phammacothemapy. The outcome of the more
gage in therapy until a more severe depressive episode
severe group in this study is quite typical of depressed
begins to remit, despite a somewhat slower course of
patients treated with a variety of active therapies, in-
recovery. The batten explanation touches on a clinically
cluding antidepressant medications. Indeed, high pre-
relevant variant of experimental bias, in which both
treatment severity matings also have been associated
the therapists’ and patients’ expectations of positive
with poorer response to pharmacotherapy in some
outcome may influence the subsequent response mate.
studies ( 1 1 ). This phenomenon may simply reflect a
In summary, our findings suggest that in an open
decline in response to placebo as severity of depression
treatment setting, cognitive behavior therapy is nearly
increases, with a relatively constant difference between
as effective a treatment of moderately severe depres-
drug and placebo. And yet, the fact that our more
sion (i.e., Hamilton scale scones of 20-25) as it is in
severely depressed patients did manifest higher mesid-
milder depression. Nevertheless, more severely de-
ual scores at termination of therapy certainly suggests
pressed patients did finish treatment with somewhat
that their recovery was less complete. The finding of
less complete remissions and might benefit from longer
higher residual depressive severity alone may convey
on more intensive courses of therapy. The efficacy of
a higher risk for subsequent relapse (12, 13). Although
cognitive behavior therapy in even more severely de-
a combined psychothemapy/pharmacotherapy regimen
pressed patients (e.g., inpatients on melancholic outpa-
might reduce such a risk (14), so might a longer or more
tients with Hamilton scale scones greater than 25), of
intense course of cognitive behavior therapy (13).
course, remains to be determined.
Our findings differ from those of the Collaborative
Research Program (4) in several respects. First, the
overall response to cognitive behavior therapy in the REFERENCES
total patient group was greaten in our study (e.g., 1. Shea MT, Elkin I, Hirschfeld RMA: Psychotherapeutic treat-
Hamilton scale scone of 6 on less: our study42 of 59 ment of depression, in American Psychiatric Press Review of
patients [71%], Collaborative Research Pnognam=20 Psychiatry, vol 7. Edited by Frances A, Hales RE. Washington,

788 Am J Psychiatry 148:6, June 1991

 THASE, SIMONS, CAHALANE, ET AL.

DC, American Psychiatric Association Press, 1988 atric disturbance. Arch Gen Psychiatry 1976; 33:766-771
2. Williams JMG: Cognitive-behavior therapy for depression: 9. Beck AT, Ward CH, Mendelson M, et al: An inventory for
problems and perspectives. Br J Psychiatry 1984; 145:254-262 measuring depression. Arch Gen Psychiatry 1961; 4:561-571
3. Simons AD, Thase ME: Mood disorders, in Handbook of Out- 10. Kraemer HC, Thiemann S (eds): How Many Subjects? New-
patient Treatment of Adults. Edited by Thase ME, Hersen M, bury Park, Calif, Sage, 1987
Edelstein BA. New York, Plenum, 1990 1 1 . Thase ME, Kupfer DJ: Characteristics of treatment resistant
4. Elkin I, Shea MT, Watkins JT, et al: National Institute of Men- depression, in Treating Resistant Depression. Edited by Zohar
tal Health Treatment of Depression Collaborative Research J, Belmaker RH. New York, PMA, 1987
Program: general effectiveness of treatments. Arch Gen Psychi- 12. Simons AD, Murphy GE, Levine JL: Relapse after treatment
arty 1989; 46:971-982 with cognitive therapy and/or pharmacotherapy: results after
S. Blackburn IM, Bishop 5, Glen AIM, et al: The efficacy of cog- one year. Arch Gen Psychiatry 1986; 43:43-48
nitive therapy in depression: a treatment trial using cognitive 13. Thase ME: Relapse and recurrence in unipolar major depres-
therapy and pharmacotherapy, each alone and in combination. sion:short term and long term approaches. J Clin Psychiatry
BrJ Psychiatry 1981; 139:181-189 1990; S1(June suppl):S1-57
6. Rush AJ, Beck AT, Kovacs M, et al: Comparative efficacy of 14. Frank E, Kupfer DJ, Perel JM, et al: Three year outcomes for
cognitive therapy and pharmacotherapy in the treatment of de- maintenance therapies in recurrent depression. Arch Gen Psy-
pressed outpatients. Cognitive Ther Res 1977; 1:17-31 chiatry 1990; 47:1093-1099
7. Hamilton M: A rating scale for depression. J Neurol Neurosurg 15. Shaw BF: Specification of the training and evaluation of cogni-
Psychiatry 1960; 23:56-62 tive therapists for outcome studies, in Psychotherapy Research:
8. Endicott J, Spitzer RL, Fleiss JL, et al: The Global Assessment Where We Are and Where Should We Go? Edited by Williams
Scale: a procedure for measuring the overall severity of psychi- JBW, Spitzer RL. New York, Guilford Press, 1984

Am J Psychiatry 1 48:6, June 1991 789