Psychiatry Research, 37:245-260 Elsevier
245
Clinical, Cognitive, and Demographic Predictors of Response to Cognitive Therapy for Depression: A Preliminary Report Robin B. Jarrett, A. John Rush
G. Gregory Eaves, Bruce D. Grannemann,
and
Received November 5. 1990; revised version received March 28. 1991; accepted April 27, 1991. Abstract. This preliminary study evaluated prognostic indicators or predictors of response to cognitive therapy. The sample included 37 unipolar outpatients with moderate to severe major nonpsychotic depressive disorder, according to Research Diagnostic Criteria. Demographic characteristics (sex, age, marital status, and education), pretreatment severity measures (Hamilton Rating Scale for Depression [HRSD] and Beck Depression Inventory [BDI]), pretreatment cognitive measures (Dysfunctional Attitudes Scale [DAS] and Attributional Style Questionnaire Failure Composite [ASQ-F]), and historical features (length of illness, length of current episode, number of episodes, and age of onset) were used in multiple regression models to predict response. In accord with previous findings, patients who had higher (rather than lower) pretreatment HRSD, BDI, or DAS scores and were single (rather than married) showed a poorer response to cognitive therapy, according to the HRSD. Furthermore, married outpatients with high DAS scores or single patients with low DAS scores showed an intermediate response to cognitive therapy, while single patients with high DAS scores responded the least. Generally, effects were stronger when response was assessed according to clinician-rated severity measures rather than patient self-reports. Key Words. Depressed outpatients, psychotherapy, depression. In psychotherapy,
has been: “What specific problem,
cognitive
therapy,
predictors,
short-term
the critical question confronting the field for more than 20 years treatment, by whom, is most effective for this individual with that under which set of circumstances?” (Paul, 1967, p. 1II). With
respect to the field of mood disorders, this question is not unique to psychotherapy, as it also applies to pharmacotherapy. To improve the treatment of depressed patients, we must answer the question, “Which depressions respond best to what treatments under what conditions?”
An earlier version of this report was presented at the Annual Meeting of the Society for Psychotherapy Research, Santa Fe, NM, June 1988. Robin B. Jarrett, Ph.D., is Associate Professor of Psychiatry; G. Gregory Eaves, Ph.D., is Clinical Assistant Professor of Psychology; Bruce D. Grannemann, M.S., is a Research Associate; and A. John Rush, M.D., holds the Betty Jo Hay Chair in Mental Health and is Director of the Mental Health Clinical Research Center. (Reprint requests to Dr. R.B. Jarrett, Mental Health Clinical Research Center, Dept. of Psychiatryj9070, U.T. Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235, USA.) 0165-1781/91/$03.50
@ 1991 Elsevier Scientific
Publishers
Ireland
Ltd
246 This question presumes that effective treatments are available. To date most of the literature on unipolar major depressive disorder has focused on identifying which treatments are effective. There is evidence that a subset of patients with unipolar major depression respond to placebo (e.g., White et al., 1985; Fairchild et al., 1986); interpersonal, cognitive, and behavior therapies (for a review, see Jarrett and Rush, 1986; Dobson, 1987; Shea et al., 1988); tricyclic antidepressants and monoamine therapy (Fink, oxidase inhibitors (Brotman et al., 1987); and electroconvulsive 1987). The response rate typically cited for either tricyclic antidepressants or cognitive therapy ranges from approximately 45% to 85% (Bielski and Friedel, 1976; Rush and Shaw, 1983; Rush, 1984), and placebo response rates typically range from 15% to 30% (Rush et al., 1983) for nonpsychotic, nonbipolar outpatients. While no single treatment is ubiquitously effective, only a few studies have tried to identify which depressions respond to what treatments. Hollon and Najavits (1988), in their attempt to locate predictors of treatment response, distinguish between prescriptive and prognostic indicators of response to treatment. Prescriptive indicators reveal which patients respond best to Treatment A compared with Treatment B (i.e., a between-subjects factor involving patient characteristics is crossed with a factor involving different types of treatment) and thus aid in selecting between two or more treatments. Prognostic indicators reveal the extent to which different types of patients respond to a single treatment (i.e., a between-subjects factor involving patient characteristics is crossed with a single treatment) and thus aid in deciding to whom a particular treatment should be given. This study attempted to identify prognostic indicators or predictors of response to a single treatment, namely cognitive therapy. Previous studies that characterize those patients who respond to cognitive therapy for depression included those by Keller (1983) and Shea (1987), both of whom found that depressed outpatients with high scores on the Dysfunctional Attitudes Scale (DAS; Weissman, 1979) were less likely to respond to cognitive therapy. Miller et al. (1990) showed that depressed inpatients with high scores on the DAS and the Cognitive Bias Questionnaire (CBQ; Hammen and Krantz, 1976; Krantz and Hammen, 1979) were more likely to respond to a combination of pharmacotherapy and cognitive-behavior therapy (compared to pharmacotherapy alone), whereas patients with low scores on these measures responded equally to either treatment. It is noteworthy that the Miller et al. (1990) study differs from the other two studies by including inpatients and by comparing cognitive therapy and social skills training combined with pharmacotherapy to pharmacotherapy alone. Blackburn et al. (198 1) reported that the endogenous/ nonendogenous dichotomy did not differentiate responders from nonresponders to cognitive therapy alone (as did Kovacs et al., 1981), to medication alone, and to the combination of cognitive therapy and medication. On the other hand, Gallagher and Thompson (1983) reported pilot data from depressed geriatric patients which suggested that more nonendogenous (80%) than endogenous (33%) patients improved with cognitive or brief dynamic therapy. It has been suggested that although depressed patients with melancholia (as defined by DSM-III or DSM-III-R) may not be any more likely than patients with nonmelancholic depression to respond to antidepressants, they may be particularly
247 unresponsive to nonsomatic “treatment” (i.e., whether it is placebo or psychosocial intervention) (Nelson et al., 1990). In our search for prognostic indicators of response to cognitive therapy, we attempt to predict which unipolar depressed adult outpatients respond to cognitive therapy on the basis of their severity and history of illness, demographic characteristics, and pretreatment cognitive characteristics.
Methods Subjects. The preliminary data reported here are from the first 46 (of 80 planned) subjects who entered an ongoing study designed to evaluate which subtypes of depressions respond to cognitive therapy. When we had collected approximately half of the data, we analyzed the data and reported our preliminary findings. The preliminary evaluation of other predictors was reported separately and showed that neither the latency of rapid eye movement sleep (Jarrett et al., 1990) nor pretreatment levels of learned resourcefulness (Jarrett et al., submitted) predicted response to cognitive therapy. The subjects analyzed here and reported in Jarrett et al. (1990) and Jarrett et al. (submitted) were not independent samples. Forty subjects completed at least six sessions of cognitive therapy, which we defined a priori as a minimally adequate exposure to cognitive therapy. (Six patients dropped out before receiving at least six sessions. In two cases the treating clinician recommended hospitalization; three patients requested alternative treatment, and one patient discontinued cognitive therapy for practical reasons.) Of the 40 subjects who had at least six sessions of cognitive therapy, 37 subjects had complete data on all but one or two of the predictors examined in this report. The family history of six patients could not be determined. Table 1 summarizes the severity measures, demographic data, cognitive measures, history of illness, family history, depressive subtypes, and laboratory findings, which were gathered before treatment for the 37 subjects, and were included in analyses conducted for this report. Note that t tests showed no significant differences between the 37 subjects (included in this report) and the 9 subjects omitted due to the fact that they (I) did not complete a minimally adequate exposure to cognitive therapy (n = 6) or (2) had missing data on the variables used as predictors (n = 3). Evaluation. All outpatients presented at the Mood Disorders Program, University of Texas Southwestern Medical Center, Dallas, Texas, between 1983 and 1988 with the chief complaint of depression. Patients were evaluated using the Schedule for Affective Disorders and Schizophrenia-Lifetime Version (SAD%L; Endicott and Spitzer, 1978) and strictly applying the Research Diagnostic Criteria (RDC; Spitzer et al., 1978). Subjects were diagnosed by RDC as having nonpsychotic, unipolar, major depressive disorder. Patients with bipolar I or II disorder, psychotic depression, organic affective disorder, or concurrent medical disorders were excluded. Structured interviews were conducted at initial presentation; diagnoses and symptom severity were reevaluated within approximately 5-7 days at a followup interview. When followed longitudinally after successful completion of the protocol. two patients developed hypomanic episodes. Interviewers gathered demographic data, history of illness (age of onset of first depressive episode, duration of the current episode, length of illness, and number of episodes), and family history (Winokur et al., 1978; Winokur, 1979) from each patient and, when available, from a relative or close friend. Family history was documented by the evaluator who reviewed the lifetime psychiatric status of all first degree biological relatives (i.e., parents, siblings, and offspring) by interviewing the patient. In these analyses family history was coded as “positive” when the patient reported the presence of a treated mood disorder, drug abuse, or sociopathy or convincingly described symptoms of these disorders. Depressions were rigorously subtyped by RDC into endogenous/ nonendogenous and primary/ secondary. All diagnoses of endogenous subtypes were rendered by strictly applied RDC for
248 the nadir of the current episode. As such, the patient’s mood, for example, had to be virtually unreactive to positive events rather than simply less reactive than during the prior nondepressed state. The RDC require 6 of 10 key symptoms to classify the episode as definite endogenous. For those patients deemed “probable endogenous” by RDC, a nonendogenous diagnosis was rendered. Primary depression was coded when depression was the first major psychiatric illness as defined by RDC that had occurred during the patient’s lifetime. To rule out occult medical disorders, all patients underwent routine laboratory screens, including complete blood count, liver, thyroid and renal function tests. glucose level (SMA20), and microscopic urinalysis. After a 14-day medication washout phase, patients underwent a routine polysomnographic evaluation and the dexamethasone suppression test. Clinical rating scales to assess symptom severity included the 17-item Hamilton Rating
Table 1. Demographic and clinical resoonders, and the total samole
comparison
Responders (n = 26) Variable
Mean
SD
of responders,
Nonresponders (n= 11) Mean
SD
non-
Total sample (n = 37) Mean
SD
Pretreatmentseverity measures HRSD BDI Demographic
17.4
2.7
19.4
4.0
17.9
3.2
23.4
6.5
31.4
9.0
25.7
8.1
Age
36.2
9.4
38.1
9.8
36.8
9.4
15.7
2.0
16.4
2.5
15.9
Education
(yr)
2.1
Female
69.2%
54.6%
64.9%
Single
42.3%
81.8%
54.1%
Pretreatment cognitive measures DAS ASQ-F History of illness Age at onset Length of episode
(mo)
Length of illness (yr) Number
of episodes
(n = 34)
147.9
27.7
168.2
32.8
5.0
0.8
4.8
0.6
28.3
9.2
23.7
9.5
27.0
9.4
13.7
16.7
32.4
33.5
19.3
24.1
8.1
6.9
13.7
10.8
9.8
8.5
2.7
1.7
2.1
1.0
2.5
1.5
30.4
153.9 4.93
0.7
Family history subtypes 30.8%
18.2%
27.0%
Depression spectrum disease
30.8%
18.2%
27.0%
Sporadic depressive disorder
19.2%
36.4%
24.3%
3.9%
9.1%
5.4%
15.4%
18.2%
16.2%
Familial pure depressive
disease
Bipolar family Unknown
Depressive subtypes Primary depression RDC endogenous Laboratory measures DST nonsuppressors Reduced
REM latency
100%
84.6% 15.4%
(n = 34) (n = 36)
Mean REM latency (n = 36)
74.8
89.2%
27.3%
18.9%
11.5%
0.0%
8.1%
42.3%
36.4%
40.5%
22.5
80.2
Note. HRSD = Hamilton Rating Scale for Depression. BDI = Beck Depression
30.4
76.3
24.6
Inventory. DAS = Dysfunctional Altitudes Scale (Form 8). ASQ-F = Attributional Style Questionnaire-Failure Composite Score. RDC = Research Diagnostic Criteria. DST = dewmethasone suppression test REM = rapid eye movement.
249 Scale for Depression (HRSD; Hamilton, 1960) and the 28-item Inventory of Depressive Symptomatology-Clinician Version (IDS-C; Rush et al., 1986). A minimum score of 14 on a 17-item HRSD collected at the time of the follow-up interview was required to enter the study. Self-report questionnaires to assess severity included the 21-item Beck Depression Inventory (BDI; Beck et al., 1961) and the 28-item IDS-SR (Rush et al., 1986). With the exception of the BDI score, therapists and patients were unaware of the results of pretreatment self-report and polysomnographic (PSG) measures. Treatment. Cognitive therapy was based on the procedures detailed by Beck et al. (1979). By protocol, patients received up to twenty 50-min sessions, twice weekly for a total of 10 weeks. The mean number of therapy sessions was 17.7 (SD = 4.2) and the mean number of weeks in treatment was 12.8 (SD = 2.6) for the 37 patients included in this report. Note that only two subjects had fewer than 10 sessions of cognitive therapy. During treatment, no patient received psychoactive medication of any type and patients’ medication status was reviewed weekly. Therapists included two Ph.D. clinical psychologists (1 male and 1 female), one male psychiatric fellow, one male psychiatric resident, and one male clinical social worker. All had previous training and experience in treating depressed patients with cognitive therapy and received ongoing supervision throughout the protocol. The competence of the therapist, as well as the integrity of the intervention, was assessed using the Cognitive Therapy Scale (CTS; Young and Beck, 1980; Vallis et al., 1986). The following CTS scores were assigned by an outside consultant (B.F. Shaw) on the basis of periodic random samples of videotapes. CTS scores were averaged over 2 to 18 occasions: Therapist 1 = 39.4 + 7.9 (SD), Therapist 2 = 30.3 f 7.9 (SD), Therapist 3 = 43.1 + 7.1 (SD), Therapist 4 = 36.5 f 28.9 (SD), and Therapist 5 = 43.5 + 6.4 (SD). A CTS score of 39 or better has been used to designate competent cognitive therapy (Shaw, 1984). Predictors. The following pretreatment variables (see Table 1) were used to predict posttreatment response to cognitive therapy. Demographic characteristics included sex, age, marital status, and education. These were gathered by the evaluator before treatment. Marital status was coded as single (never married, separated, divorced, or widowed) or as married (married or cohabiting). The “married” sample was composed of 100% married and 0% cohabiting. The “single” sample was composed of 70% never married, 0% separated, 20% divorced, and 10% widowed. Pretreatment cognitive distortion was assessed using the DAS and the Attributional Style Questionnaire-Failure Composite (ASQ-F; Peterson et al., 1982). These measures are described below. The DAS was used to assess the “silent assumptions” or rules hypothesized as important in the development and maintenance of depression (Beck et al., 1979). The ASQ-F was used to assess the extent to which the patient attributed negative events to internal, global, and stable factors. Dysfunctional Attitudes Scale (DAS). The DAS (Form B) (Weissman, 1979) is a 40-item found in depression and other psychiatric self-report measure of “silent assumptions” conditions. Items, including perfectionistic standards, concern about approval from others, requirement for being happy, or feeling inadequate, are rated on a 7-point scale from “agree very much” to “disagree very much.“The average score for students with depressive symptoms (BDI > 10) was 130.26 (SD = 29.60) and 114.46 (SD = 25.10) for nondepressed students. The DAS has demonstrated a test-retest reliability correlation of 0.84 over an S-week period and coefficient a’s ranging from 0.89 to 0.92 (Weissman, 1979). The concurrent validity of the DAS is supported by its correlation of 0.36 and 0.47 (p’s < 0.001) with two measures of depressive symptoms in a student population (Weissman, 1979) and with BDI scale scores (Dobson and Breiter, 1983). In addition, significant differences have been found between depressed and nondepressed psychiatric patients and nondepressed volunteers on this measure (Hamilton and Abramson, 1983). Attributional Style Questionnaire (ASQ). The ASQ (Peterson et al., 1982) is a self-report
250 measure of an individual’s tendency to attribute negative and positive events to internal versus external, stable versus unstable, and global versus specific causes. Subjects are asked to generate a cause for 12 hypothetical events (50% positive and 50% negative). Cronbach’s a’s for the positive and negative composite scores are 0.75 and 0.72, respectively (Peterson et al., 1982). The six-item subscales reflecting each of the attributional dimensions had reliabilities ranging from 0.44 to 0.69 with an average of 0.54. Test-retest correlations for the three attributional dimension and composite scores for good and bad events ranged from 0.57 to 0.70 (p < 0.001). Response to Treatment. Response to treatment was the primary dependent variable and was assessed using the 17-item HRSD and the 21-item BDI scores collected at the final cognitive therapy session. HRSD. The 17-item HRSD was collected by the treating therapist and recorded on videotape. Interobserver agreement between the therapist and an independent evaluator was calculated based on 51% of HRSDs (with pretreatment and posttreatment HRSD ratings combined). The corater, who was unaware of the patient’s diagnosis, laboratory findings, and pretreatment or posttreatment status, viewed a videotaped interview. Intraclass correlations equalled ‘1 = 0.82 (n = 20 observations) at pretreatment, rt = 0.94 (n = 18 observations) at posttreatment, and rt = 0.69 (n = 38 observations on 27 subjects) overall. The computation of the intraclass correlation was based on Tinsley and Weiss’ (1975) formula in which the between-judges variance was not included in the error term. This method treats judges as a fixed factor and is appropriate because the reported reliability is meant to represent only the reliability of the particular group of judges and not the general reliability of the HRSD. Measures of kappa were not made because the HRSD ratings were not experimentally independent. In categorical analyses, 70% responded to cognitive therapy when an end-oftreatment HRSD score
245
Clinical, Cognitive, and Demographic Predictors of Response to Cognitive Therapy for Depression: A Preliminary Report Robin B. Jarrett, A. John Rush
G. Gregory Eaves, Bruce D. Grannemann,
and
Received November 5. 1990; revised version received March 28. 1991; accepted April 27, 1991. Abstract. This preliminary study evaluated prognostic indicators or predictors of response to cognitive therapy. The sample included 37 unipolar outpatients with moderate to severe major nonpsychotic depressive disorder, according to Research Diagnostic Criteria. Demographic characteristics (sex, age, marital status, and education), pretreatment severity measures (Hamilton Rating Scale for Depression [HRSD] and Beck Depression Inventory [BDI]), pretreatment cognitive measures (Dysfunctional Attitudes Scale [DAS] and Attributional Style Questionnaire Failure Composite [ASQ-F]), and historical features (length of illness, length of current episode, number of episodes, and age of onset) were used in multiple regression models to predict response. In accord with previous findings, patients who had higher (rather than lower) pretreatment HRSD, BDI, or DAS scores and were single (rather than married) showed a poorer response to cognitive therapy, according to the HRSD. Furthermore, married outpatients with high DAS scores or single patients with low DAS scores showed an intermediate response to cognitive therapy, while single patients with high DAS scores responded the least. Generally, effects were stronger when response was assessed according to clinician-rated severity measures rather than patient self-reports. Key Words. Depressed outpatients, psychotherapy, depression. In psychotherapy,
has been: “What specific problem,
cognitive
therapy,
predictors,
short-term
the critical question confronting the field for more than 20 years treatment, by whom, is most effective for this individual with that under which set of circumstances?” (Paul, 1967, p. 1II). With
respect to the field of mood disorders, this question is not unique to psychotherapy, as it also applies to pharmacotherapy. To improve the treatment of depressed patients, we must answer the question, “Which depressions respond best to what treatments under what conditions?”
An earlier version of this report was presented at the Annual Meeting of the Society for Psychotherapy Research, Santa Fe, NM, June 1988. Robin B. Jarrett, Ph.D., is Associate Professor of Psychiatry; G. Gregory Eaves, Ph.D., is Clinical Assistant Professor of Psychology; Bruce D. Grannemann, M.S., is a Research Associate; and A. John Rush, M.D., holds the Betty Jo Hay Chair in Mental Health and is Director of the Mental Health Clinical Research Center. (Reprint requests to Dr. R.B. Jarrett, Mental Health Clinical Research Center, Dept. of Psychiatryj9070, U.T. Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX 75235, USA.) 0165-1781/91/$03.50
@ 1991 Elsevier Scientific
Publishers
Ireland
Ltd
246 This question presumes that effective treatments are available. To date most of the literature on unipolar major depressive disorder has focused on identifying which treatments are effective. There is evidence that a subset of patients with unipolar major depression respond to placebo (e.g., White et al., 1985; Fairchild et al., 1986); interpersonal, cognitive, and behavior therapies (for a review, see Jarrett and Rush, 1986; Dobson, 1987; Shea et al., 1988); tricyclic antidepressants and monoamine therapy (Fink, oxidase inhibitors (Brotman et al., 1987); and electroconvulsive 1987). The response rate typically cited for either tricyclic antidepressants or cognitive therapy ranges from approximately 45% to 85% (Bielski and Friedel, 1976; Rush and Shaw, 1983; Rush, 1984), and placebo response rates typically range from 15% to 30% (Rush et al., 1983) for nonpsychotic, nonbipolar outpatients. While no single treatment is ubiquitously effective, only a few studies have tried to identify which depressions respond to what treatments. Hollon and Najavits (1988), in their attempt to locate predictors of treatment response, distinguish between prescriptive and prognostic indicators of response to treatment. Prescriptive indicators reveal which patients respond best to Treatment A compared with Treatment B (i.e., a between-subjects factor involving patient characteristics is crossed with a factor involving different types of treatment) and thus aid in selecting between two or more treatments. Prognostic indicators reveal the extent to which different types of patients respond to a single treatment (i.e., a between-subjects factor involving patient characteristics is crossed with a single treatment) and thus aid in deciding to whom a particular treatment should be given. This study attempted to identify prognostic indicators or predictors of response to a single treatment, namely cognitive therapy. Previous studies that characterize those patients who respond to cognitive therapy for depression included those by Keller (1983) and Shea (1987), both of whom found that depressed outpatients with high scores on the Dysfunctional Attitudes Scale (DAS; Weissman, 1979) were less likely to respond to cognitive therapy. Miller et al. (1990) showed that depressed inpatients with high scores on the DAS and the Cognitive Bias Questionnaire (CBQ; Hammen and Krantz, 1976; Krantz and Hammen, 1979) were more likely to respond to a combination of pharmacotherapy and cognitive-behavior therapy (compared to pharmacotherapy alone), whereas patients with low scores on these measures responded equally to either treatment. It is noteworthy that the Miller et al. (1990) study differs from the other two studies by including inpatients and by comparing cognitive therapy and social skills training combined with pharmacotherapy to pharmacotherapy alone. Blackburn et al. (198 1) reported that the endogenous/ nonendogenous dichotomy did not differentiate responders from nonresponders to cognitive therapy alone (as did Kovacs et al., 1981), to medication alone, and to the combination of cognitive therapy and medication. On the other hand, Gallagher and Thompson (1983) reported pilot data from depressed geriatric patients which suggested that more nonendogenous (80%) than endogenous (33%) patients improved with cognitive or brief dynamic therapy. It has been suggested that although depressed patients with melancholia (as defined by DSM-III or DSM-III-R) may not be any more likely than patients with nonmelancholic depression to respond to antidepressants, they may be particularly
247 unresponsive to nonsomatic “treatment” (i.e., whether it is placebo or psychosocial intervention) (Nelson et al., 1990). In our search for prognostic indicators of response to cognitive therapy, we attempt to predict which unipolar depressed adult outpatients respond to cognitive therapy on the basis of their severity and history of illness, demographic characteristics, and pretreatment cognitive characteristics.
Methods Subjects. The preliminary data reported here are from the first 46 (of 80 planned) subjects who entered an ongoing study designed to evaluate which subtypes of depressions respond to cognitive therapy. When we had collected approximately half of the data, we analyzed the data and reported our preliminary findings. The preliminary evaluation of other predictors was reported separately and showed that neither the latency of rapid eye movement sleep (Jarrett et al., 1990) nor pretreatment levels of learned resourcefulness (Jarrett et al., submitted) predicted response to cognitive therapy. The subjects analyzed here and reported in Jarrett et al. (1990) and Jarrett et al. (submitted) were not independent samples. Forty subjects completed at least six sessions of cognitive therapy, which we defined a priori as a minimally adequate exposure to cognitive therapy. (Six patients dropped out before receiving at least six sessions. In two cases the treating clinician recommended hospitalization; three patients requested alternative treatment, and one patient discontinued cognitive therapy for practical reasons.) Of the 40 subjects who had at least six sessions of cognitive therapy, 37 subjects had complete data on all but one or two of the predictors examined in this report. The family history of six patients could not be determined. Table 1 summarizes the severity measures, demographic data, cognitive measures, history of illness, family history, depressive subtypes, and laboratory findings, which were gathered before treatment for the 37 subjects, and were included in analyses conducted for this report. Note that t tests showed no significant differences between the 37 subjects (included in this report) and the 9 subjects omitted due to the fact that they (I) did not complete a minimally adequate exposure to cognitive therapy (n = 6) or (2) had missing data on the variables used as predictors (n = 3). Evaluation. All outpatients presented at the Mood Disorders Program, University of Texas Southwestern Medical Center, Dallas, Texas, between 1983 and 1988 with the chief complaint of depression. Patients were evaluated using the Schedule for Affective Disorders and Schizophrenia-Lifetime Version (SAD%L; Endicott and Spitzer, 1978) and strictly applying the Research Diagnostic Criteria (RDC; Spitzer et al., 1978). Subjects were diagnosed by RDC as having nonpsychotic, unipolar, major depressive disorder. Patients with bipolar I or II disorder, psychotic depression, organic affective disorder, or concurrent medical disorders were excluded. Structured interviews were conducted at initial presentation; diagnoses and symptom severity were reevaluated within approximately 5-7 days at a followup interview. When followed longitudinally after successful completion of the protocol. two patients developed hypomanic episodes. Interviewers gathered demographic data, history of illness (age of onset of first depressive episode, duration of the current episode, length of illness, and number of episodes), and family history (Winokur et al., 1978; Winokur, 1979) from each patient and, when available, from a relative or close friend. Family history was documented by the evaluator who reviewed the lifetime psychiatric status of all first degree biological relatives (i.e., parents, siblings, and offspring) by interviewing the patient. In these analyses family history was coded as “positive” when the patient reported the presence of a treated mood disorder, drug abuse, or sociopathy or convincingly described symptoms of these disorders. Depressions were rigorously subtyped by RDC into endogenous/ nonendogenous and primary/ secondary. All diagnoses of endogenous subtypes were rendered by strictly applied RDC for
248 the nadir of the current episode. As such, the patient’s mood, for example, had to be virtually unreactive to positive events rather than simply less reactive than during the prior nondepressed state. The RDC require 6 of 10 key symptoms to classify the episode as definite endogenous. For those patients deemed “probable endogenous” by RDC, a nonendogenous diagnosis was rendered. Primary depression was coded when depression was the first major psychiatric illness as defined by RDC that had occurred during the patient’s lifetime. To rule out occult medical disorders, all patients underwent routine laboratory screens, including complete blood count, liver, thyroid and renal function tests. glucose level (SMA20), and microscopic urinalysis. After a 14-day medication washout phase, patients underwent a routine polysomnographic evaluation and the dexamethasone suppression test. Clinical rating scales to assess symptom severity included the 17-item Hamilton Rating
Table 1. Demographic and clinical resoonders, and the total samole
comparison
Responders (n = 26) Variable
Mean
SD
of responders,
Nonresponders (n= 11) Mean
SD
non-
Total sample (n = 37) Mean
SD
Pretreatmentseverity measures HRSD BDI Demographic
17.4
2.7
19.4
4.0
17.9
3.2
23.4
6.5
31.4
9.0
25.7
8.1
Age
36.2
9.4
38.1
9.8
36.8
9.4
15.7
2.0
16.4
2.5
15.9
Education
(yr)
2.1
Female
69.2%
54.6%
64.9%
Single
42.3%
81.8%
54.1%
Pretreatment cognitive measures DAS ASQ-F History of illness Age at onset Length of episode
(mo)
Length of illness (yr) Number
of episodes
(n = 34)
147.9
27.7
168.2
32.8
5.0
0.8
4.8
0.6
28.3
9.2
23.7
9.5
27.0
9.4
13.7
16.7
32.4
33.5
19.3
24.1
8.1
6.9
13.7
10.8
9.8
8.5
2.7
1.7
2.1
1.0
2.5
1.5
30.4
153.9 4.93
0.7
Family history subtypes 30.8%
18.2%
27.0%
Depression spectrum disease
30.8%
18.2%
27.0%
Sporadic depressive disorder
19.2%
36.4%
24.3%
3.9%
9.1%
5.4%
15.4%
18.2%
16.2%
Familial pure depressive
disease
Bipolar family Unknown
Depressive subtypes Primary depression RDC endogenous Laboratory measures DST nonsuppressors Reduced
REM latency
100%
84.6% 15.4%
(n = 34) (n = 36)
Mean REM latency (n = 36)
74.8
89.2%
27.3%
18.9%
11.5%
0.0%
8.1%
42.3%
36.4%
40.5%
22.5
80.2
Note. HRSD = Hamilton Rating Scale for Depression. BDI = Beck Depression
30.4
76.3
24.6
Inventory. DAS = Dysfunctional Altitudes Scale (Form 8). ASQ-F = Attributional Style Questionnaire-Failure Composite Score. RDC = Research Diagnostic Criteria. DST = dewmethasone suppression test REM = rapid eye movement.
249 Scale for Depression (HRSD; Hamilton, 1960) and the 28-item Inventory of Depressive Symptomatology-Clinician Version (IDS-C; Rush et al., 1986). A minimum score of 14 on a 17-item HRSD collected at the time of the follow-up interview was required to enter the study. Self-report questionnaires to assess severity included the 21-item Beck Depression Inventory (BDI; Beck et al., 1961) and the 28-item IDS-SR (Rush et al., 1986). With the exception of the BDI score, therapists and patients were unaware of the results of pretreatment self-report and polysomnographic (PSG) measures. Treatment. Cognitive therapy was based on the procedures detailed by Beck et al. (1979). By protocol, patients received up to twenty 50-min sessions, twice weekly for a total of 10 weeks. The mean number of therapy sessions was 17.7 (SD = 4.2) and the mean number of weeks in treatment was 12.8 (SD = 2.6) for the 37 patients included in this report. Note that only two subjects had fewer than 10 sessions of cognitive therapy. During treatment, no patient received psychoactive medication of any type and patients’ medication status was reviewed weekly. Therapists included two Ph.D. clinical psychologists (1 male and 1 female), one male psychiatric fellow, one male psychiatric resident, and one male clinical social worker. All had previous training and experience in treating depressed patients with cognitive therapy and received ongoing supervision throughout the protocol. The competence of the therapist, as well as the integrity of the intervention, was assessed using the Cognitive Therapy Scale (CTS; Young and Beck, 1980; Vallis et al., 1986). The following CTS scores were assigned by an outside consultant (B.F. Shaw) on the basis of periodic random samples of videotapes. CTS scores were averaged over 2 to 18 occasions: Therapist 1 = 39.4 + 7.9 (SD), Therapist 2 = 30.3 f 7.9 (SD), Therapist 3 = 43.1 + 7.1 (SD), Therapist 4 = 36.5 f 28.9 (SD), and Therapist 5 = 43.5 + 6.4 (SD). A CTS score of 39 or better has been used to designate competent cognitive therapy (Shaw, 1984). Predictors. The following pretreatment variables (see Table 1) were used to predict posttreatment response to cognitive therapy. Demographic characteristics included sex, age, marital status, and education. These were gathered by the evaluator before treatment. Marital status was coded as single (never married, separated, divorced, or widowed) or as married (married or cohabiting). The “married” sample was composed of 100% married and 0% cohabiting. The “single” sample was composed of 70% never married, 0% separated, 20% divorced, and 10% widowed. Pretreatment cognitive distortion was assessed using the DAS and the Attributional Style Questionnaire-Failure Composite (ASQ-F; Peterson et al., 1982). These measures are described below. The DAS was used to assess the “silent assumptions” or rules hypothesized as important in the development and maintenance of depression (Beck et al., 1979). The ASQ-F was used to assess the extent to which the patient attributed negative events to internal, global, and stable factors. Dysfunctional Attitudes Scale (DAS). The DAS (Form B) (Weissman, 1979) is a 40-item found in depression and other psychiatric self-report measure of “silent assumptions” conditions. Items, including perfectionistic standards, concern about approval from others, requirement for being happy, or feeling inadequate, are rated on a 7-point scale from “agree very much” to “disagree very much.“The average score for students with depressive symptoms (BDI > 10) was 130.26 (SD = 29.60) and 114.46 (SD = 25.10) for nondepressed students. The DAS has demonstrated a test-retest reliability correlation of 0.84 over an S-week period and coefficient a’s ranging from 0.89 to 0.92 (Weissman, 1979). The concurrent validity of the DAS is supported by its correlation of 0.36 and 0.47 (p’s < 0.001) with two measures of depressive symptoms in a student population (Weissman, 1979) and with BDI scale scores (Dobson and Breiter, 1983). In addition, significant differences have been found between depressed and nondepressed psychiatric patients and nondepressed volunteers on this measure (Hamilton and Abramson, 1983). Attributional Style Questionnaire (ASQ). The ASQ (Peterson et al., 1982) is a self-report
250 measure of an individual’s tendency to attribute negative and positive events to internal versus external, stable versus unstable, and global versus specific causes. Subjects are asked to generate a cause for 12 hypothetical events (50% positive and 50% negative). Cronbach’s a’s for the positive and negative composite scores are 0.75 and 0.72, respectively (Peterson et al., 1982). The six-item subscales reflecting each of the attributional dimensions had reliabilities ranging from 0.44 to 0.69 with an average of 0.54. Test-retest correlations for the three attributional dimension and composite scores for good and bad events ranged from 0.57 to 0.70 (p < 0.001). Response to Treatment. Response to treatment was the primary dependent variable and was assessed using the 17-item HRSD and the 21-item BDI scores collected at the final cognitive therapy session. HRSD. The 17-item HRSD was collected by the treating therapist and recorded on videotape. Interobserver agreement between the therapist and an independent evaluator was calculated based on 51% of HRSDs (with pretreatment and posttreatment HRSD ratings combined). The corater, who was unaware of the patient’s diagnosis, laboratory findings, and pretreatment or posttreatment status, viewed a videotaped interview. Intraclass correlations equalled ‘1 = 0.82 (n = 20 observations) at pretreatment, rt = 0.94 (n = 18 observations) at posttreatment, and rt = 0.69 (n = 38 observations on 27 subjects) overall. The computation of the intraclass correlation was based on Tinsley and Weiss’ (1975) formula in which the between-judges variance was not included in the error term. This method treats judges as a fixed factor and is appropriate because the reported reliability is meant to represent only the reliability of the particular group of judges and not the general reliability of the HRSD. Measures of kappa were not made because the HRSD ratings were not experimentally independent. In categorical analyses, 70% responded to cognitive therapy when an end-oftreatment HRSD score
