David M. Garner, Ph.D., Wendi Rockert, B.A., Ron Davis, Ph.D., Maureen V. Garner, Ph.D., Marion P. Olmsted, Ph.D., and Morris Eagle, Ph.D.

Transcription

1 Regular Articles Comparison of Cognitive-Behavioral and Supportive-Expressive for Bulimia Nervosa David M. Garner, Ph.D., Wendi Rockert, B.A., Ron Davis, Ph.D., Maureen V. Garner, Ph.D., Marion P. Olmsted, Ph.D., and Morris Eagle, Ph.D. Objective: The authors compared the effectiveness of 4 months (1 8 sessions) of cognitivebehavioral and supportive-expressive therapy for bulimia. Method: Sixty patients obtained f rom clinical referrals to an eating disorders program who met modified DSM-III-R criteria f or bulimia nervosa were randomly assigned to the two conditions. Treatments were delivered in an individual format, on an outpatient basis, by experienced therapists using treatment manuals. The primary outcome measures were self-induced vomiting, binge eating, and attitudes toward body weight and shape, which were assessed by self-report and structured interview. Results: Fifty patients completed treatment, 25 in each condition. Both treatments led to significant improvements in specific eating disorder symptoms and in psychosocial disturbances. Supportive-expressive therapy was just as effective as cognitive-behavioral therapy in reducing binge eating. Where treatment differences were found, they favored cognitive-behavioral therapy. Cognitive-behavioral therapy was marginally superior in reducing the frequency ofself-induced vomiting; 36% ofthe patients who received cognitive-behavioral therapy and 1 2 % of those who received supportive-expressive therapy abstained from vomiting in the last month oftreatment. Cognitive-behavioral therapy was significantly more effective in ameloriating disturbed attitudes toward eating and weight, depression, poor self-esteem, general psychological distress, and certain personality traits. Conclusions: These results moderately favor cognitive-behavioral therapy over supportive-expressive therapy for bulimia nervosa, but follow-up is required to determine the durability of outcome with both modalities. The findings must be interpreted with caution since the selected clinical sample in this study may not represent the bulimia nervosa population. (Am J Psychiatry 1993; 150:37-46) Presented at the World Congress of Cognitive, Oxford, England, June 28 to July 2, From the Department of Psychiatry, Toronto General Hospital and University oftoronto. Address reprint requests to Dr. Garner, Department of Psychiatry, Michigan State University College of Human Medicine, East Lansing, MI The authors thank the therapists in the study, Daniela Bonatto, Ph.D., Robin Brooks-Hill, M.D., Barbara Dorian, M.D., David Goldbloom, M.D., Paul Isaacs, Ph.D.,Jon Innis, M.D.,Jane Irvine, D.Phil., and Randy Gangbar, M.D.; they also thank Lawrence Van Egeren, Ph.D., Jane Irvine, D.Phil., and Brian Mavis, Ph.D., for their comments on a draft of the manuscript and Kelly Vitousek, Ph.D., for contributions to the study. Supported by Health and Welfare Canada project grant , by NATO Grants for Collaborative Research, and by a Research Associate Award to Dr. D.M. Garner and a Research Fellowship to Dr. Davis from the Ontario Mental Health Foundation. Copyright 1993 American Psychiatric Association. S ince the original clinical description of bulimia nervosa by Russell in 1979 (1), there have been numerous reports of psychological and pharmacological treatments for bulimia nervosa. The predominant onentation in psychological treatment trials has been cognitive-behavioral therapy. Despite variation in the actual methods used and in the rigor of study designs, results have indicated that various forms of this intervention can significantly reduce binge eating, vomiting, and associated symptoms and that improvements endure for at least a year after treatment (2, 3). Only a minority of the treatment studies have used a design aimed specifically at testing the comparative effects of two or more credible treatments, and the results Am] Psychiatry 150:1, January

2 COMPARISON OF BULIMIA THERAPIES of these studies have been inconsistent. Three studies (4-6) showed few differences on most dependent measures between cognitive-behavioral therapy and the comparison treatment. The failure of comparative trials to find group differences may relate to the relative importance of nonspecific treatment factors compared to components unique to cognitive-behavioral therapy. Alternatively, it may relate to methodological issues such as inadequate power resulting from small subject groups or failure to adequately distinguish the components of the different treatment conditions. Two more recent studies provide stronger support for the superiority of cognitive-behavioral therapy over other methods. Fairburn et al. (7) compared cognitive-behavioral therapy, a simplified behavioral treatment, and interpersonal psychotherapy. The findings at the end of treatment indicated that, while all three treatments resulted in improvement on measures of psychopathology, cognitive-behavioral therapy was more effective than the other treatments in modifying extreme dieting, self-induced vomiting, and disturbed attitudes toward shape and body weight. In a carefully controlled comparison of imipramine and short-term intensive group treatment that incorporated many components of cognitive-behavioral therapy, Mitchell et al. (8) found that the intensive group treatment was superior to antidepressant treatment alone in modifying eating symptoms; however, both treatments led to significantly greater improvements in eating symptoms and mood disturbance than did placebo treatment. Adding the drug therapy to the group treatment led to additional improvement in measures of anxiety and depression. However, it did not significantly improve outcome related to eating behavior, and it resulted in a higher dropout rate. To date, several studies have yielded positive findings with psychodynamically oriented treatment (9-13); however, none has involved a comparison with an a!- ternative treatment and all have used such methods as self-monitoring, meal planning, education regarding the untoward effects of dieting, and behavioral contracting, which are generally considered more conceptually aligned with cognitive-behavioral therapy. Thus, it is not clear whether the observed changes in behavior were due to the specific components of the dynamic treatment or to dietary management. Very little improvement in binge eating and vomiting was reported after a 12-week psychodynamically oriented group program that did not include dietary management (14), but the brief group treatment format may not be a sufficiently potent test of the intervention. The aim of the present study was to compare cognitive-behavioral therapy (4, 7, 15) and a brief psychodynamic therapy (16), both delivered in an individual format, according to specific guidelines, and by experienced therapists. The psychodynamic condition selected for the present study was judged as particularly suitable because it is well conceptualized, manual based, and intended to be brief and it has been described as appropriate for individuals with impulse control problems. Efforts were made in the current study to use measurement techniques and a time frame for assessments that parallel those in a recent trial by Fairburn et a!. (7). A waiting list control group was not used since previous research has indicated that this control condition leads to minimal symptom change. We focused specifically on the magnitude of behavioral and psychological change with these two psychological treatments after 18 individual treatment sessions spaced over a minimum of 4 months. METHOD Subjects The study participants were 60 women who were selfreferred or referred by professionals to a hospital eating disorders program and who met the following inclusion/exclusion criteria: 1 ) the Russell criteria for bulimia nervosa ( 1 7) and the DSM-III-R criteria for the disorder with the exception that a minimum average of two binges a week involving large amounts of food was not required (criterion A), 2) a minimum of two episodes of vomiting a week for the past month, 3) a minimum duration of illness of 1 year, 4) a present body weight of between 85% and 120% of matched population mean weight (18), 5) age between 1 8 and 35 years, 6) no concurrent treatment for bulimia nervosa, and 7) written and informed consent to participate. The rationale for the reliance on the rather broad Russell criteria (1 7) for bulimia nervosa in favor of those specified by DSM-III-R was that there is no consensus on several aspects of the DSM-III-R criteria for bulimia nervosa, such as the definition of a binge and what constitutes a large amount of food (19). Several studies have indicated that almost one-half of the episodes labeled by bulimia nervosa patients as binges involve fewer than 1,000 calories (20-22). Nevertheless, the amount and type of food consumed during a binge was determined by a standardized interview, the Eating Disorder Examination (23). Four patients (two in each treatment condition) reported no objective episodes of binge eating (consumption of more than 1,000 calories of food not resembling a meal) in the month preceding the assessment interview (they vomited after consuming even small amounts of food), but all reported episodes that met their own subjective definitions of binge eating. The patients reported between 0 and 140 objectively defined binge eating episodes per month (mean=27.s, SD=2S.1) and between eight and 154 vomiting episodes a month (mean=42.2, SD=32.6) before the pretreatment assessment. During the 15-month entry phase of the study, 92 (31.0%) of the 297 patients receiving initial consultations for any form of eating disorder were referred to the present study. The referred patients were assessed by a clinician/research technician who determined whether they met the study criteria; 60 (65.2%) of these patients actually met the inclusion criteria and were entered into the trial. 38 Am J Psychiatry 150:1, January 1993

3 GARNER, ROCKERT, DAVIS, ET AL. The patients were stratified according to duration of illness (<3 years and 3 years), current weight (86%- 110% and >111% of matched population mean weight), and probable history of anorexia nervosa (i.e., adult weight <85% of matched population mean weight), which was derived from information provided on the study intake referral form. They were then randomly assigned by the nonblind study research technician to treatment conditions before the initial assessment interviews. Every attempt was made to adhere to the stratification procedures; however, in a few cases a patient who should have been assigned to one treatment was assigned to the other because therapists in the assigned condition were unavailable to accept a referral at the time. Any patient who dropped out was replaced by the next suitable patient, who was assigned to the same treatment cell, in order to obtain 25 patients who completed each treatment. Previous research indicates that attrition rates of approximately 15% can be expected in treatment studies of bulimia nervosa (2). Procedure The nature of the study was explained to prospective participants in an initial meeting, and written informed consent was obtained. Assessments were performed through interviews before treatment (within 3 weeks of the initial assessment), at midtreatment, after treatment, and 3 months, 6 months, and 1 year after the end of treatment by a clinician/research technician not involved in the patients clinical care. The patients were asked to keep records of the frequency of binge eating, vomiting, laxative abuse, and dieting and to complete the 26-item version of the Eating Attitudes Test (24) throughout the course of treatment as pant of the research protocol. The Eating Attitudes Test data were gathered once a week over the first month of treatment and every other week thereafter, and the symptom data were gathered for each week of treatment. The symptom summary forms and Eating Attitudes Test results were placed in a sealed envelope and deposited in a data collection box before treatment sessions. Because the rate of compliance with completing the symptom summany forms decreased markedly after the first 6 weeks of treatment, these data are not presented in the current report. To minimize potential effects of therapist access to study data, the patients were informed that the treatment and research aspects of the study were intended to be kept completely separate and their therapists would not be informed of the results from any of their assessment measures or their progress during treatment. Duning follow-up the patients were told that their therapists would be given only very general information about their clinical status if it was requested. At each assessment the patients were weighed, given a structured interview by the study research technician, and then asked to complete a battery of psychometric instruments. They were given instructions regarding the completion of symptom summary forms. The primary outcome measures were frequency of vomiting and binge eating. Measures of attitudes toward weight and body shape were also considered to be critical in examining the relative effectiveness of the two treatments. Secondary or exploratory measures were aimed at assessing psychological distress, personality features, and social adjustment. Frequency of vomiting and binge eating were determined before and after treatment with a standardized structured interview, the Eating Disorder Examination (23). Concerns regarding weight, body shape, and eating were also assessed before and after treatment by the drive for thinness, bulimia, and body dissatisfaction subscales of the Eating Disorder Inventory (25), the 26- item version of the Eating Attitudes Test (24), and the Eating Disorder Examination (23). The psychological measures completed before and aften treatment included the remaining subscales of the Eating Disorder Inventory, the SCL-90-R (26), the Borderline Syndrome Index (27), the Rosenberg Self-Esteem Scale (28), the Beck Depression Inventory (29), and the Millon Clinical Multiaxial Inventory (30). The subjects completed the entire Millon inventory, but only the results from the borderline and dysthymia subscales are presented since these subject domains are considered particularly relevant to eating disorders (31). The Social Adjustment Scale-Self-Report (32) was also administered. Although this scale yields scores for five areas of social functioning, only the aggregate score for all of these areas was used for the data analyses. A 25-item treatment satisfaction measure, based on the instrument described by Luborsky (16), was developed for the present study. This measure was administered at the posttreatment assessment and evaluated a broad range of factors related to satisfaction with the therapist and with the treatment provided. At the end of treatment, the patients were also asked to evaluate the treatment received in terms of the relative emphasis on 14 general content areas. Items for this measure were generated by the clinicians participating in the study and were based on themes that they judged to have theoretical relevance to one on both treatment conditions. Treatment Conditions The subjects were asked to attend I 9 individual treatment sessions, each minutes in duration, delivered over 1 8 weeks. The sessions occurred twice a week during the first month, once a week for the next 2 months, and once every other week for the final 6 weeks, in accordance with the model by Fairburn (15). The therapists were 10 experienced clinicians (five with M.D. degrees and five with Ph.D. degrees) who were each recruited for the present study to deliver only the type of treatment that was consistent with his or her usual clinical orientation. This was different from other comparative studies, in which the therapists delivered all available forms of treatment. While this design featune fails to control for potential sources of bias associated with therapist factors, it assures that each therapist is practicing a therapy that is consistent with the thera- AmJ Psychiatry 150:1, January

4 COMPARISON OF BULIMIA THERAPIES pist s orientation, and it also minimized potential therapist drift, which may result from delivering divergent forms of treatments. The therapists followed detailed manuals for both treatment conditions and attended weekly supervision meetings to encourage adherence to the treatment protocol when dealing with emergent clinical problems. (As a check on treatment fidelity, a Ph.D. clinician blindly rated 1 5 randomly selected tapes of therapy sessions from each treatment condition; the clinician was able to correctly classify each tape as representing either cognitive-behavioral therapy or supportive-expressive therapy. Each tape was also rated to determine whether at least three supportive-expressive or cognitive-behavioral therapy componentsextracted from the respective manuals-could be identified. Each of the tapes evaluated contained at least three components of the respective treatment, providing furthen evidence of treatment fidelity.) All therapists were instructed to inform patients that the research component of the study was independent from their treatment and that the therapists would not be informed of the research evaluation at any stage of treatment. The structure of the protocol for cognitive-behavioral therapy and the overall goals of each stage of treatment generally followed the manual described by Fairburn (15), supplemented by our own adaptation of cognitive-behavioral therapy principles for eating disorders (33-35). These principles were originally denived from techniques developed by Beck (36) and colleagues for the treatment of depressive and anxiety disorders. The patients were supplied with self-monitoning forms and asked to keep records of all food and liquid ingested; episodes of binge eating, vomiting, laxative abuse, and other weight-losing behaviors; and feelings and thoughts concerning eating. The supportive-expressive therapy used the treatment manual by Luborsky (16), supplemented by psychodynamic writings on eating disorders (37-39). The style of the supportive-expressive therapy was nondirective, and the emphasis was on listening to the patient and helping identify problems and their solutions. No specific advice was to be given at any stage. The therapist was instructed to assume a facilitative role and allow the patient to retain responsibility for change. Adapted to eating disorders, the supportive-expressive therapy approach was based on the assumption that the eating symptoms serve a functional role by disguising underlying interpersonal problems. Although information was gathered on present eating symptoms and each patient was given an information booklet outlining the dangers of binge eating, self-induced vomiting, and purgative abuse, the therapists were instructed to be careful to avoid giving patients specific advice and were told that if questions were asked by the patient, the question should be reflected back. Data Analysis The overall effects of the two treatment conditions were evaluated for each outcome variable by using a two-way repeated measures analysis of variance (ANOVA) with treatment group and time as the factors. The differences between the two treatment effects were determined by an analysis ofcovaniance (ANCOVA) for each outcome vanable with the pretreatment patient scones on the respective measures as the covaniates. Frequency of vomiting and binge eating as assessed by self-report and interview were considered the pnincipal bases for interpreting the comparative efficacy of the two treatments. Nevertheless, additional data are presented for measures of attitudes toward weight and shape and for secondary measures tapping a broad range of psychosocial areas of functioning. Exploratory analyses of a relatively large number of outcome measures were considered to be an important adjunct to formal hypothesis testing in the current study because of their potential for advancing the understanding of meaningful treatment variables (40). It was decided not to reduce the number of outcome measures by using data reduction procedures, such as principal component analysis, in spite of the problems imposed by the unfavorable subject-to-variable ratio, since this would result in difficulty in interpreting factors in relation to normative data and change scores for instruments widely used in previous research. Changes in scores on these instruments are of substantial interest to both clinicians and researchers despite the probable redundancy of some measures and the potential type I and type II errors due to multiple comparisons. While nonadjusted probabilities are presented in the tables to facilitate the examination of the pattern of group differences across all measures, probabilities adjusted to control for the family-wise error rate were calculated and are also included in the tables. RESULTS Dropouts and Completers Ten patients (five patients from each treatment condition) withdrew before the end of treatment and were not followed. The proportional distribution of dropouts across the two treatment conditions reduces, but does not eliminate, the possibility that the posttreatment comparisons were biased in favor of either of the treatment conditions. The representativeness of the patients who completed treatment was evaluated by collapsing the completers and dropouts across both treatment conditions and comparing these two groups on pretreatment age, weight and weight history variables, duration of illness, target eating symptoms, and scores on Eating Disorder Inventory subscales, Eating Attitudes Test subscales, SCL-90-R subscales, Borderline Syndrome Index, Beck Depression Inventory, Social Adjustment Scale total, and Rosenberg scale. The dropouts did not differ significantly from the completers on any of these variables except that their maximum adult weight, expressed as a percentage of matched population mean weight, was 40 Am J Psychiatry 1 50:1, January 1993

5 GARNER, ROCKERT, DAVIS, ET AL. higher than that of the completers (t=3.27, df=s6, p<o.oo2), and they showed more disturbance than the completers on the Beck Depression Inventory (t=2.34, df=58, p<o.o3), Eating Attitudes Test oral control subscale (t=2.79, df=s6, p<o.oo7), and SCL-90-R somatization subscale (t=2.58, df=s8, p<o.o2). These findings, along with the sizable proportion of referred subjects who were not actually entered into the trial, suggest that the findings from the present study may not be repnesentative of all bulimia nenvosa patients requesting treatment for the disorder. The remaining findings reported in the present study relate to the SO patients who completed treatment and include those who, by mutual consent, terminated treatment early because they had met the treatment objectives (this includes five patients from the cognitivebehavioral therapy condition who terminated after sessions 9, 10, 12, 13, and 18, respectively, and five patients receiving supportive-expressive therapy who terminated treatment at session 18 rather that 19). Demographic data and clinical features for these SO patients are summarized in table 1. Before treatment there were no significant differences between the cognitivebehavioral and supportive-expressive treatment groups in age, height, actual weight, weight expressed as pencentage of matched population mean weight, duration of illness (almost 6 years), previous treatment for an eating disorder (45.5% versus 52.2%), or history of an adult weight below 80% of matched population mean weight (28.0% in each group). The only psychometric instrument on which there was a significant pretreatment group difference was the maturity fears subscale of the Eating Disorder Inventory, on which the subjects who received supportive-expressive therapy had a higher mean initial scone (t=2.32, df=47, p<o.o3). The supportive-expressive therapy group also had a somewhat higher mean pretreatment score on the Borderline Syndrome Index (F=1.80, df=47, p<o.o8). Content Areas Discussed During Treatment At the posttreatment assessment, the patients were asked to rate on a 14-item, 6-point ( never to a!- ways ) Likert scale how often you and your therapist discussed specific content areas during treatment. The differences between the treatment groups were consistent with the theoretical expectations for the respective treatments and provided some reassurance that the treatment received was consistent with the intended model. There were six significant differences and two nearly significant differences (p<o.o9) between the ratings of 14 content areas by the subjects in the two treatment conditions, and all of these were in the expected direction. The subjects who received cognitive-behavioral therapy gave higher ratings than did the suppontive-expressive therapy subjects for discussions of preoccupation with weight and shape (t=9.68, df=40, p< ), bingeing (t=s.01, df=40, p<o.0001), pneoccupation with food (t=4.53, df=40, p<o.0001), TABLE 1. Baseline Features of Bulimic Patients Receiving Cognitive-Behavioral (N=25) or Supportive-Expressive (N=25) Cognitive- Behavioral Supportive- Expressive Characteristic Mean SD Mean SD Age (years) Height (in) Actual weight (Ib) Weight as percentage of matched population mean weight Current Maximum Minimum Duration of illness (months) vomiting (t=3.60, df=40, p zo.oo1), and specific strategies for overcoming your eating problems (t= 6.90, df=40, p<o.0001 ), and there was a nearly significant difference for the causes of your eating problems (t=i.86, df=40, p<o.o7). The ratings of the subjects who received supportive-expressive therapy were significantly higher for your relationship with your panents when you were a child (t=3.76, df=40, p<o.oi), and the difference was nearly significant for your relationship with your therapist (t=1.89, df=40, p< 0.07). There were no significant differences between treatments on your relationship with your parents, your relationship with men, your beliefs and attitudes in general, your feelings and emotions in general, your self-esteem, and your mood. Symptom Measures The two-way repeated measures ANOVAs indicated that there was a main effect of time for all outcome variables, indicating that the treatments had a significant effect on the symptom areas measured. Table 2 presents the mean frequencies of vomiting and objective binge eating, as determined by interview, for both treatment conditions. The two treatments were equally effective in their impact on binge frequency, but the group differences in vomiting frequency, in favor of the cognitive-behavioral therapy group, approached significance. There was an 81.9% reduction in vomiting frequency from pre- to posttreatment in the cognitive-behavioral therapy group, compared to a 62.1% reduction in the supportive-expressive therapy group. At the end of treatment, most patients had improvements in vomiting frequency of at least 50% (92.0% of the patients receiving cognitive-behavioral therapy and 68.0% of the patients receiving supportive-expressive therapy); however, only nine (36.0%) of the cognitivebehavioral therapy patients and three (12.0%) of the patients receiving supportive-expressive therapy had been abstinent from vomiting for the 28 days preceding the interview evaluation. Table 2 also presents the results of the ANCOVAs for the Eating Attitudes Test; cognitive-behavioral therapy led to significantly greater AmJ Psychiatry 150:1, January

6 COMPARISON OF BULIMIA THERAPIES TABLE 2. Eating Symptoms and Scores on the Eating Attitudes Test and Ea ting Disorder Examination for Bu limic Patie nts Before an d After Cognitive-Behavioral or Supportive-Expressive Cognitive-Behavioral Supportive-Expressive Score Score ANCOVAa Measure N Mean SD N Mean SD F df p Vomiting episodes in last 28 days 2S Binge eating episodes in last 28 days Eating Attitudes Test (26-item) Dieting Bulimia and food preoccupation Oral control Total Eating Disorder Examination Dietary restraint \i Attitudes toward shape Attitudes toward weight , n.s , , n.s , , , , athe family-wise error rate for 12 comparisons (all variables in table 2 and the first three subscales of the Eating Disorder Inventory in table 3) is 0.6 for p=o.os, 0.12 for p=o.ol, and for p=o.oo1. improvements in the scones on two Eating Attitudes Test subscales and on the total scone. Table 2 also contains findings for the same Eating Disorder Examination subscales reported by Fainburn et al. (7). In the current study, cognitive-behavioral therapy was significantly superior to suppontive-expressive therapy in effects on dietary restraint and shape concerns, and there was a nearly significant diffenence in the scone on the weight concerns subscale. Not reported in the table is the failure to achieve group differences on the Eating Disorder Examination bulimia subscale and the significant group difference, favoning cognitive-behavioral therapy, on the eating concerns subscale (F=S.97, df=i, 47, p<o.o2). Presented in table 3 are the results from the ANCO- VAs for the Eating Disorder Inventory subscales. There was a significant group difference on the bulimia subscale and a nearly significant difference on the drive for thinness subscale; both indicated superiority of cognitive-behavioral therapy. There was a significant group effect for the maturity fears subscale of the Eating Disorder Inventory; however, as indicated earlier, the group that received supportive-expressive therapy had higher pretreatment and posttreatment scores on this subscale than the cognitive-behavioral therapy group. A repeated measures Group by Time ANOVA mdicated that the patients experienced a significant weight increase during treatment (F=23.60, df=i, 47, p< ), and the Group by Time interaction was nearly significant (F=3.37, df=1, 47, p<o.o8), reflecting the fact that the cognitive-behavioral therapy group gained more weight during treatment than the suppontive-expressive therapy group (mean weight gain for the cognitive-behavioral therapy group was 6.6 lb. on 3.0 kg, to 100.4% of matched population mean weight; mean weight gain for the supportive-expressive therapy group was 3.0 lb, on 1.4 kg, to 97.6% ofmatched population mean weight). Table 4 summarizes data for both treatment groups on the Beck Depression Inventory, the Borderline Syndrome Index, the borderline and dysthymia scales of the Millon Clinical Multiaxial Inventory, the global SCL- 90-R, the Rosenberg Self-Esteem Scale, and the total Social Adjustment Scale. The number of subjects varies among the different measures owing to the facts that some patients did not complete certain psychometric instnuments and the decision to administer the Millon Clinical Multiaxial Inventory was made after the study had begun. The ANCOVAs showed better results for the cognitive-behavioral therapy condition, including 42 Am ] Psychiatry 1 50:1, January 1993

7 GARNER, ROCKERT, DAVIS, ET AL. significant posttneatment differences on the Beck Dcpnession Inventory, the borderline and dysthymia scales of the Millon Clinical Multiaxial Inventory, the global SCL-90-R, and the Rosenberg scale. Although not presented in table 4, there were also significant group differences in favor of cognitive-behavioral therapy on the schizotypal (F=S.79, df=1, 35, p<o.o2), anxiety (F= 7.55, df=1, 35, p<0.01), and somatoform (F=12.28, df=1, 35, p<o.oo1 ) subscales of the Millon Clinical Multiaxial Inventory. Treatment Satisfaction Satisfaction with treatment was rated by the patients at the posttreatment assessment with a 25-item scale adapted from Luborsky (16). This scale tapped a number of elements of treatment and had a reliability coefficient (alpha) of 0.95 for the patients completing treatment. On the basis of monthly frequency of vomiting at the end of treatment, the patients were classified as haying either good/moderate outcome (4 episodes per month) or poor outcome (>4 episodes per month). A Treatment by Outcome Group ANOVA for treatment satisfaction revealed a significant interaction (F=S.72, df=1, 46, p=o.o2). Cognitive-behavioral therapy patients with good outcomes were significantly (F=9.SS, df=3, 46, p<o.0001 ) more satisfied with treatment than were cognitive-behavioral therapy patients with poor outcomes or supportive-expressive therapy patients with either good or poor outcomes. DISCUSSION The major aim of the current study was to compare cognitive-behavioral therapy to a credible alternative matched on nonspecific therapeutic factors but not using methods considered integral to cognitive-behavioral therapy for bulimia nervosa. Both cognitive-behavioral therapy and supportive-expressive therapy led to significant improvements in specific eating disorder symptoms and in a broad range of measures of psychosocial functioning. Although previous comparative research has shown that treatments other than cognitive or behavioral therapies can lead to improvements in eating symptoms (4, 6), these results have been difficult to interpret because of a blurring between treatment conditions. This makes the supportive-expressive therapy findings in the current study, along with the findings of another recent study (7), of particular interest, since they suggest that treatments which do not specifically focus on eating behavior or concerns about eating or weight (unlike cognitive-behavioral therapy) are able to lead to significant improvements in these areas. In the current study, supportive-expressive therapy was just as effective as cognitive-behavioral therapy in reducing episodes of binge eating. This is consistent with the findings of the Fairburn et al. (7) study, in which interpersonal therapy compared favorably to cognitive-behavioral therapy in reducing binge eating episodes. TABLE 3. Scores on Subscales of the Eating Disorder Inventory for Bulimic Patients Before and After Cognitive-Behavioral (N=25) or Supportive-Expressive (N=24) Score Cognitive- Supportive- Subscale of Behavioral Expressive ANCOVA Eating (df=1, 48)a Disorder Inventory Mean SD Mean SD F p Drive for thinness Bulimia Body dissatisfaction n.s. Ineffectiveness n.s. Perfectionism n.s. Interpersonal distrust n.s. Interoceptive awareness n.s. Maturity fears n.s. athe family-wise error rate for eight comparisons is 0.4 for p=0.os, 0.08 for p=o.oi, and for p=0.ooi. A key finding in this comparative trial was that, where treatment differences did exist, they favored the cognitive-behavioral therapy intervention. Cognitivebehavioral therapy was marginally superior (p<o.o6) to supportive-expressive therapy in reducing vomiting frequency and also led to greater improvements in most measures of concern about eating and weight, symptom areas that are considered specific to bulimia nervosa. While the findings favoring cognitive-behavioral therapy in modifying vomiting frequency are similar to those of an earlier comparison of cognitive-behavioral therapy and interpersonal therapy (7), the overall percentage reduction in this symptom following cognitivebehavioral therapy in the current study (81.9%) and the proportion of patients who were abstinent from vomiting following cognitive-behavioral therapy (36.0%) are somewhat less impressive than the findings by Fairburn et al. (95% and 47%, respectively). These differences may be due to a number of factors, such as use of different selection criteria in the two studies (i.e., the presence of vomiting was required in the current study but was present in only 72% of the patients in the Fairburn et al. study), differences in the referral base for the two studies (the patients in the current study were selected Am J Psychiatry 150:1, January

8 COMPARISON OF BULIMIA THERAPIES TABLE 4. Psychological Functioning of Bulimic Patients Before and Afte r Cognitive-Behavioral or Supportiv e-expressiv e Cognitive-Behavioral Supportive-Expressive Score Score ANCOVAa Measure N Mean SD N Mean SD F df p Beck Depression Inventory Borderline Syndrome Index Rosenberg Self-Esteem Scale SCL-90-R Social Adjustment Scale-Self-Report Millon Clinical Multiaxial Inventory Borderline subscale Dysthymia subscale athe family-wise error rate for seven comparisons is 0.35 for p=o.os, 0.07 for p=o.ol, and for p=o.oo , n.s , , n.s , , from patients referred to a specialized tertiary eating disorder center, whereas there were no competing treatment centers in the Fairburn et a!. study), on the relative ineffectiveness of the cognitive-behavioral therapy offened in the current study. The relatively low rates of abstinence in the cognitive-behavioral therapy condition demonstrate that this treatment, delivered over the short term, is not curative for most bulimic patients seeking treatment at a tertiary care center specializing in eating disorders. Nevertheless, the percentage neduction in vomiting and the abstinence rates achieved in the current study are consistent with the median rates of improvement seen in other studies of the efficacy of cognitive-behavional therapy (2). The results for vomiting were paralleled by changes in measures of concern regarding weight, shape, and eating. Both treatment groups showed significant improvement, and at the end of treatment the group diffenences favored cognitive-behavioral therapy. The results on the Eating Attitudes Test are consistent with those of earlier comparative studies (4, 7), indicating that cognitive-behavioral therapy is superior to alternative forms of treatment in ameliorating attitudes and behaviors characteristic of eating disorders. The improvements in scores on the Eating Disorder Inventory in the current study are comparable to those reported in the study of drug treatment and group therapy by Mitchell et al. (8). The changes on the Eating Disorder Examination and the Eating Attitudes Test were similar to those reported by Fairburn et al. (7). This is reassuning to the degree that it confirms the clinical sensitivity of these instruments in detecting change. Perhaps of greaten importance than the change in eating symptoms in the current study is the superiority of cognitive-behavioral therapy in promoting changes on measures of depression, self-esteem, personality featunes, and general psychopathology, since it could be argued that these areas, rather than eating symptoms, represent the targets of change for suppontive-expressive therapy. These improvements may be attributed to direct therapeutic effects of cognitive-behavioral thenapy on these symptom areas, on they may be simply a reflection of the inverse association between chronic dietary chaos and psychological disturbance observed in bulimia nenvosa patients (41, 42). Nevertheless, the relative superiority of cognitive-behavioral therapy in promoting improvement across a number of psychological dimensions contradicts the notion that changes in eating symptoms are unrelated to more fundamental areas of psychological distress (43). The findings from the current study are limited by several factors. First, the cognitive-behavioral therapy and supportive-expressive therapy were delivered by different therapists, in contrast to earlier studies, in which the compared treatments were delivered by the same therapists. Cognitive-behavioral therapy received significantly higher patient satisfaction ratings than supportive-expressive therapy but only when there was a positive treatment outcome. Thus, while the therapists providing cognitive-behavioral therapy were not uniformly evaluated in a positive manner, the observed differences in treatment outcome in the current study could be due to other systematic sources of bias related to general therapist variables. Second, treatment fidelity 44 AmJPsychiatry 150:1,]anuary 1993

9 GARNER, ROCKERT, DAVIS, ET AL. and quality were not assessed for all sessions, leaving open to question whether the specified treatments were competently executed. Although by no means conclusive, some reassurance on these points comes from the following: 1 ) tape ratings for a subset of sessions mdicated that the treatment conditions could be distinguished, 2) patient ratings of treatment content were consistent with theoretical predictions, 3) the treatments were manual based, 4) they were delivered by experienced therapists practicing in their preferred onentations, and 5) the therapists were supervised by senion clinicians. Finally, the conclusions regarding the efficacy of the treatments evaluated in this trial are limited by the absence of follow-up data. While most previous research on psychological treatment of bulimia nenvosa has indicated that improvements at the end of treatment are maintained on augmented during the first year of follow-up (2), there is no assurance that this pattern would extend to the groups evaluated in the current study. We are currently gathering follow-up data on these treatments. At this time little is known about the active ingredients of treatment and the predictors of response to treatment. Moreover, the impact of nonspecific treatment factors on outcome in bulimia nervosa has nemained largely unexplored. Several studies (44-47) have evaluated specific components of cognitive-behavioral therapy by using dismantling designs and have begun to clarify the picture regarding the active ingredients of treatment; however, the ideal elements of treatment are not yet evident. Indeed, the optimal treatment is probably not genenalizable to all patients with bulimia nenvosa. Future refinements may point to sequencing on stepwise approaches to treatment taibred to meet the individual needs of members of the heterogeneous patient population (48). This would involve initially providing patients with an educationally oriented treatment (35) that may benefit those who have a more benign variant of bulimia nenvosa, followed by more intensive interventions for those who are resistant to change. Of interest in this regard is a comparison of the cognitive-behavioral therapy results in this study with our brief educational treatment (49), which indicated that on several important outcome indexes the two treatments were equally effective with the least symptomatic 2S%-4S% of patients. Decisions reganding the probable efficacy of the various treatment options are cleanly influenced by an understanding of predictors of outcome; however, few studies have canefully evaluated the range of psychosocial and biological factors that may be associated with outcome. This remains a potentially fruitful area for future research. REFERENCES 1. Russell G: Bulimia nervosa: an ominous variant of anorexia nervosa. Psychol Med 1979; 9: Garner DM, Fairburn CG, Davis R: Cognitive-behavioral treatment of bulimia nervosa: a critical appraisal. Behav Modif 1987; 11: Herzog DB, Keller MB, Lavori PW: Outcome in anorexia nervosa and bulimia nervosa: a review of the literature. J Nerv Ment Dis 1988; 176: Fairburn CG, Kirk J, O Connor M, Cooper PJ: A comparison of two psychological treatments for bulimia nervosa. Behav Res Ther 1986; 24: S. Freeman C, Barry F, Dunkeld-Turnbull J, Henderson A: Controlled trial of psychotherapy for bulimia nervosa. Br Med J 1988; 296: Kirkley BG, Schneider JA, Agras WS, Bachman JA: Comparison of two group treatments for bulimia. J Consult Clin Psychol 1985; 53: Fairburn CG, Jones R, Peveler RC, Carr SJ, Solomon RA, O Connor ME, Burton J, Hope RA: Three psychological treatments for bulimia nervosa: a comparative trial. Arch Gen Psychiatry 1991; 48: MitchellJE, Pyle RL, Eckert ED, Hatsukami D, Pomeroy C, Zimmerman R: A comparison study of antidepressants and structured intensive group psychotherapy in the treatment of bulimia nervosa. Arch Gen Psychiatry 1990; 47: Connors ME, Johnson CL, Stuckey MK: Treatment of bulimia with brief psychoeducational group therapy. Am J Psychiatry 1984; 141: Dixon KN, Kiecolt-Glaser J: Group therapy for bulimia. Hillside I Clin Psychiatry 1984; 6: I 1. Lacey H: Bulimia nervosa, binge eating, and psychogenic vomiting: a controlled treatment study and long term outcome. Br Med J 1983; 286: Stevens EV, Salisbury JD: Group therapy for bulimic adults. Am J Orthopsychiatry I 984; 54: Roy-Byrne P, Lee-Benner K, YagerJ: Group therapy for bulimia: a year s experience. mt J Eating Disorders 1 984; 3: Frommer MS, Ames JR. Gibson JW, Davis WN: Patterns of symptom change in the short-term treatment of bulimia. Int J is. Eating Disorders 1987; 6: Fairburn CG: Cognitive-behavioral treatment for bulimia, in Handbook of Psychotherapy for Anorexia Nervosa and Bulimia. Edited by Garner DM, Garfinkel PE. New York, Guilford Press, Luborsky L: Principles ofpsychoanalytic Psychotherapy: A Manual for Suppartive-Expressive Treatment. New York, Basic Books, Russell GFM: Anorexia nervosa and bulimia nervosa, in Handbook of Psychiatry, vol 4: The Neuroses and Personality Disorders. Edited by Russell GFM, Hersov L. Cambridge, Cambridge University Press, Health and Welfare Canada: Canadian Average Weights for Height, Age and Sex. Ottawa, Health and Welfare Canada, Nutrition Division, Garner DM, Shafer CL, Rosen LW: Critical appraisal of the DSM-III-R diagnostic criteria for eating disorders, in Child Psychopathology: Diagnostic Criteria and Clinical Assessment. Edited by Hooper SR, Hynd GW, Mattison RE. Hillsdale, NJ, Lawrence Erlbaum Associates, Davis R, Freeman RJ, Garner DM: A naturalistic investigation of eating behavior in bulimia nervosa. J Consult Clin Psychol I 988; 56: Rosen JC, Leitenberg H, Fisher C, Khasam C: Binge-eating episodes in bulimia nervosa: the amount and type of food consumed. Int J Eating Disorders 1 986; 5: Rossiter EM, Agras WS: An empirical test of the DSM-III-R definition of binge. Int J Eating Disorders 1990; 9: Cooper Z, Fairburn CG: The Eating Disorder Examination: a semistructured interview for the assessment of the specific psychopathology of eating disorders. Int J Eating Disorders 1987; 6: Garner DM, Olmsted MP, Bohr Y, Garfinkel PE: The Eating Attitudes Test: psychometric features and clinical correlates. Psychol Med 1982; 12: Garner DM: The Eating Disorder Inventory, 2: Professional Manual. Odessa, Fla, Psychological Assessment Resources, Derogatis LR: SCL-90-R: Administration, Scoring, and Procedures Manual, II. Towson, Md, Clinical Psychometric Research, Conte HR, Plutchik R, Karasu TB, Jerrett I: A self-report border- Am] Psychiatry 150:1, ]anuary

10 COMPARISON OF BULIMIA THERAPIES line scale: discriminant validity and preliminary norms. J Nerv Ment Dis 1980; 168: Rosenberg M: Society and the Adolescent Self-Image. Princeton, NJ, Princeton University Press, Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J: An inventory for measuring depression. Arch Gen Psychiatry 1961; 4: Millon T: Millon Clinical Multiaxial Inventory Manual, 3rd ed. Minneapolis, National Computer Systems, Swift WJ, Wonderlich SA: Personality factors and diagnosis in eating disorders: traits, disorders, and structures, in Diagnostic Issues in Anorexia Nervosa and Bulimia Nervosa. Edited by Garncr DM, Garfinkel PE. New York, Brunner/Mazel, Weissman MM, Prusoff BA, Thompson WD, Harding PS, Myers JK: Social adjustment by self-report in a community sample and in psychiatric outpatients. J Nerv Ment Dis 1978; 166: Garner DM: Cognitive therapy for anorexia nervosa, in Handbook of Eating Disorders: Physiology, Psychology, and Treatment ofobesity, Anorexia, and Bulimia. Edited by Brownell KD, Foreyt JP. New York, Basic Books, Garner DM, Bemis KM: A cognitive-behavioral approach to anorexia nervosa. Cognitive Res 1982; 6: Garner DM, Rockert W, Olmsted MP,Johnson CL, Coscina DV: Psychoeducational principles in the treatment of bulimia and anorexia nervosa, in Handbook of Psychotherapy for Anorexia Nervosa and Bulimia. Edited by Garner DM, Garfinkel PE. New York, Guilford Press, Beck AT: Cognitive and the Emotional Disorders. New York, International Universities Press, Goodsitt A: Self psychology and the treatment of anorexia ncrvosa, in Handbook of Psychotherapy for Anorexia Nervosa and Bulimia. Edited by Garner DM, Garfinkel PE. New York, Guilford Press, Lerner H: Contemporary psychoanalytic perspectives on gorgevomiting: a case illustration. IntJ Eating Disorders 1983; 3: Sugarman A, Quinlan D, Devenis L: Ego boundary disturbance in anorexia: preliminary findings. J Pers Assess 1 982; 46: Van Egeren LF: The analysis of continuous data, in Handbook of Research Methods in Cardiovascular Behavioral Medicine. Edited by Schneiderman N, Weiss SM, Kaufmann PG. New York, Plenum, Fairburn CG, Cooper PJ, Kirk J, O Connor M: The significance of the neurotic symptoms of bulimia nervosa. J Psychiatr Res 1985; 19: Garner DM, Olmsted MP, Davis R, Rockert W, Goldbloom D, Eagle M: The association between bulimic symptoms and reported psychopathology. Int J Eating Disorders 1990; 9: Norman DK, Herzog DB, Chauncey 5: A one-year outcome study of bulimia: psychological and eating symptom changes in a treatment and non-treatment group. Int J Eating Disorders 1986; 5: Agras WS, SchneiderJA, Arnow B, Raeburn SD, Telch CF: Cognitive-behavioral and response-prevention treatments for bulimia nervosa. J Consult Clin Psychol 1989; 57: Leitenberg H, Rosen JC, Gross J, Nudelman 5, Vara LS: Exposure plus response-prevention treatment of bulimia nervosa. J Consult Clin Psychol 1988; 56: Wilson GT, Rossiter E, Kleifield EL, Lindholm L: Cognitive-behavioral treatment of bulimia nervosa: a controlled evaluation. Behav Res Ther 1986; 24: Wilson GT, Eldridge KL, Smith D, Niles B: Cognitive-behavioral treatment with and without response prevention for bulimia. Behay Res Ther 1991; 29: Garner DM, Garfinkel PE, Irvine Mi: Integration and sequencing of treatment approaches for eating disorders. Psychother Psychosom 1986; 46: Olmsted MP, Davis R, Garner DM, Rockert W, Irvine MJ, Eagle M: Efficacy of a brief group psychoeducational intervention for bulimia nervosa. Behav Res Then 1991; 29: Am J Psychiatry 150:1, January 1993