A Randomized Controlled Trial of Psychoanalytic Psychotherapy or Cognitive-Behavioral Therapy for Bulimia Nervosa

We conducted a randomized controlled trial in which 70 patients with bulimia nervosa were randomly selected to 2 years of weekly psychoanalytic psychotherapy or 20 sessions of CBT delivered over 5 months. We used a randomized block design with a block size of four and a 1:1 allocation ratio, the strata being the need for ongoing psychopharmacological treatment and the diagnosis of a personality disorder. The randomization was conducted using sequentially numbered envelopes by an independent researcher who had no involvement in any other aspects of the study. The study protocol was submitted to the Research Ethics Committee for the Capital Region of Denmark, who decided that it did not require ethical approval.

The study took place in a university outpatient clinic. The participants were recruited by advertisements and by referrals from local clinics. There was an initial telephone-based screening interview followed by a detailed assessment interview with those participants who appeared eligible, which confirmed the participant’s eligibility and provided detailed written and oral information about the trial. Those participants who were eligible to take part and who gave written informed consent then underwent the pretreatment baseline assessment. The remaining individuals were referred for alternative forms of treatment. The participant flow through the study is depicted in Figure 1.

The inclusion criteria were age at least 18 years, being available for the duration of the longer of the two treatments, and meeting DSM-IV criteria for bulimia nervosa. The exclusion criteria were severe physical and psychiatric conditions that would interfere with treatment (e.g., psychosis), pregnancy, current psychotherapeutic treatment, and difficulty speaking or understanding Danish. Individuals already receiving psychopharmacological treatment as well as individuals meeting ICD−10 criteria for moderate or severe depression but who were otherwise considered eligible for the trial were referred to a consulting psychiatrist. When a stable dose of medication had been reached, the assessment procedure was continued. Patients in psychopharmacological treatments were monitored regularly by the consulting psychiatrist. Patients were withdrawn from the trial if their physical health became a cause for concern.

Two sets of therapists were used, one for each treatment. Psychoanalytic psychotherapy was provided by six female and two male clinical psychologists. On average, the therapists had 17 years of psychotherapeutic experience. All the therapists had at least 2 years of training in various psychoanalytic or psychodynamic treatment modalities, and all were supervised biweekly by the first and second authors, who devised the treatment (9).

CBT was provided by one male and two female clinical psychologists and one female psychiatrist. On average, the therapists had 8 years of psychotherapeutic experience. All the therapists had completed a 1-year generic training in CBT, and all had participated in an initial 2-day training workshop in CBT given by the last author. The therapists were supervised biweekly via videoconferencing by the last author, who devised the treatment (10, 22).

The main assessment points were at baseline, after 5 months of treatment (coinciding with the end of CBT), and 19 months later (at the end of psychoanalytic psychotherapy). The assessments were conducted by trained psychologists who had no involvement with treatment and were blind to the patient’s treatment condition.

Eating disorder psychopathology was assessed using the Eating Disorder Examination interview, version 14.4 (23), which was administered at baseline and 5 and 24 months after starting treatment. All interviews were recorded and subsequently consensus rated.

General psychopathology was assessed at baseline using the Present State Examination (24). The Symptom Checklist-90-Revised (25), the Beck Depression Inventory-II (26), and the State-Trait Anxiety Inventory (27) were also administered at baseline and 5, 12, 18, and 24 months after starting treatment.

Personality disorder status was assessed at baseline using the Structured Clinical Interview for DSM-IV Axis II Personality Disorders (28). The interviews were conducted by trained psychologists, and all interviews were recorded and subsequently consensus rated by members of a study group comprising psychologists and an external psychiatric consultant.

Interpersonal functioning was assessed at pretreatment and 5, 12, 18, and 24 months after beginning treatment with the Inventory of Interpersonal Problems (29).

Patients’ expectations of treatment outcome were rated before treatment randomization on a 5-point Likert scale. Adherence to CBT was assessed using a revised version of the Cognitive-Behavioral Therapy Treatment Protocol Adherence Scale (30), a measure developed specifically for CBT for bulimia nervosa. This consisted of 20 items rated on a 7-point scale reflecting general aspects of therapy, phase-specific interventions, and the use of noncognitive behavioral techniques. Adherence to psychoanalytic psychotherapy was measured by the Adherence Scale for Psychoanalytic Psychotherapy for Bulimia nervosa (unpublished 2011 manuscript by S. Lunn et al.), a scale specifically developed for the present study. It was rated using the same scheme as the CBT adherence scale. Both adherence measures included three items designed to monitor the use of interventions characteristic of the opposite treatment modality. The adherence scales were completed by six graduate psychology students and a clinical psychologist who had participated in the development of the measures and served as the criterion rater. The students were trained for 30 hours. Before making the study ratings, they had to achieve acceptable levels of interrater reliability compared with the criterion rater (intraclass correlation [ICC] ≥0.60) on a subset of seven training cases for each therapy type (mean ICC was 0.79 for psychoanalytic psychotherapy [range, 0.73–0.89] and 0.73 for CBT [range, 0.62–0.86]). Subsequently, two treatments from each of the eight psychoanalytic psychotherapy therapists and four treatments from each of the four cognitive-behavioral therapists were randomly selected. Within each treatment, an early, middle, and late full session was rated. Accordingly, 96 sessions were rated, 48 from each therapy type. The mean scores of the respective measures were used for the data analyses.

The funding of the trial permitted a sample size of only 70 patients. This sample size provided 80% power at two-sided p<0.05 to detect a difference of around 35% in proportions and an effect size of 0.70. As this sample size is small for a trial comparing two active treatments for bulimia nervosa (13), the study was underpowered and subject to type II error.

The primary outcome variable was cessation of binge eating and purging, defined as no binge eating or purging over the previous 28 days. Because of the different lengths of the two treatments, three comparisons were made: status 5 months after beginning treatment (corresponding to the end of CBT), status 24 months after beginning treatment (corresponding to the end of psychoanalytic psychotherapy), and status at the end of CBT (i.e., after 5 months) compared with status at the end of psychoanalytic psychotherapy (19 months later).

Group comparisons for binary outcomes are presented as odds ratios with 95% confidence intervals. Group comparisons for secondary outcome measures are presented as estimated marginal means with 95% confidence intervals. Sample means and expectancy and adherence ratings for the two treatment groups were compared using t tests or Fisher’s exact tests. All primary outcome analyses were conducted by intention-to-treat using logistic regression, with treatment expectancy as covariate. Proportions of patients achieving a global Eating Disorder Examination score below one standard deviation above the community mean were compared by logistic regression, controlling in posttreatment analyses for initial global Eating Disorder Examination scores and treatment expectancy. Frequencies of binging and purging (both square root transformed) and values of the other secondary outcome measures were compared using mixed-effect model repeated measures analysis with time, treatment expectancy, personality disorder, psychopharmacological medication, baseline values of the variable × time, and therapy type × time specified as fixed effects and with random intercepts for patients. Categorical analyses were conducted with missing data accounted for by last observation carried forward. All analyses were repeated with data based on multiple imputations. None had a material effect on the overall pattern of results. All analyses were conducted using SPSS, version 20.

Patients were recruited between October 15, 2004, and April 17, 2008. The demographic and clinical characteristics of the sample are summarized in Table 1. The majority of the patients were in their mid-20s and had eating disorder features for a mean of 12.3 years (SD=6.2). One-third had a history of anorexia nervosa and over half had received prior treatment for an eating disorder. The two treatment groups did not differ significantly on any of the baseline characteristics.

The baseline expectancy ratings were high for both treatments and did not differ significantly (CBT: mean=4.14, SD=0.75; psychoanalytic psychotherapy: mean=4.10, SD=0.76; t=0.50, p=0.62). Among the participants randomly assigned to CBT, no significant differences in expectancy ratings for the two treatments were observed (CBT: mean=4.14, SD=0.72; psychoanalytic psychotherapy: mean=4.31, SD=0.58; t=−1.64, p=0.11). In the psychoanalytic psychotherapy condition, however, expectations for CBT (mean=4.15, SD=0.78) were significantly higher than those for psychoanalytic psychotherapy (mean=3.88, SD=0.88; t=2.06, p<0.05).

The global adherence ratings were high for both treatments (CBT: mean=4.59, SD=0.60; psychoanalytic psychotherapy: mean=4.98, SD=0.73) with particularly high scores on the items differentiating the two treatments (CBT: mean=6.64, SD=0.78; psychoanalytic psychotherapy: mean=6.45, SD=0.64).

In all, 34 patients were randomly selected for psychoanalytic psychotherapy and 36 for CBT. Twenty-four (70.6%) of the patients in psychoanalytic psychotherapy completed their 2 years of treatment, and 28 (77.8%) of the CBT patients completed their 5 months of treatment. Two of the patients in psychoanalytic psychotherapy were started on antidepressant medication during the treatment phase. During follow up, 11 of the CBT completers received additional treatment. Three of these patients were in psychiatric outpatient treatment for an eating disorder, one of whom was prescribed antidepressant medication. Eight patients received some form of psychotherapy. It is not known whether this additional psychotherapy was directed at the eating disorder or at other psychological difficulties.

Both treatments resulted in an improvement in the patients’ state, although the improvement was greater among those who received CBT. At their respective endpoints, 15 (42%) of the patients in CBT had ceased to binge eat and purge compared with five (15%) of the patients in psychoanalytic psychotherapy (odds ratio=4.41; 95% CI=1.20–16.15; p=0.03). A marked difference was also observed in the time course of the response to the two treatments. This is depicted in Figure 2. CBT had a rapid and marked effect whereas the effect of psychoanalytic psychotherapy was slower and more modest. At the 5-month assessment point, 42% (15/36) of the patients in CBT had stopped binge eating and purging compared with 6% (2/34) of those receiving psychoanalytic psychotherapy (odds ratio=13.40; 95% CI=2.45–73.42; p<0.01), and at the 2-year assessment point the cessation rates were 44% (16/36) and 15% (5/34), respectively (odds ratio=4.34; 95% CI=1.33–14.21; p=0.02). Of the 15 patients in CBT who were free from binge eating and purging at the end of treatment, 10 (66%) were binge and purge free at 24 months, while six who were binge eating and purging at month 5 had ceased to do so at month 24. The outcome of CBT completers who received additional treatment did not differ significantly from the outcomes of patients who did not, neither at the end of treatment nor 19 months later.

Substantial improvements were also observed in global eating disorder psychopathology (Table 2). These improvements had occurred by the respective endpoints of the two treatments and to a similar extent (Figure 2). However, the improvements took place more rapidly in the CBT group (time-by-treatment interaction for global Eating Disorder Examination score, p<0.01); after 5 months, the patients in CBT achieved lower (better) scores than patients in psychoanalytic psychotherapy, but the difference between the groups 2 years after starting treatment was not statistically significant.

A similar pattern was observed with regard to the other secondary outcome measures (Table 2). Thus, while patients in CBT had significantly lower values after 5 months on all measures except BMI and state anxiety, after 2 years the significant differences remained only for number of binging and purging episodes. However, improvements occurred more rapidly in the CBT group for the majority of variables (time-by-treatment interaction for all measures except purging, BMI, and state and trait anxiety, p<0.05). The proportion of patients achieving a global Eating Disorder Examination score below one standard deviation above the community mean did not differ between the two treatments at either 5 months or 2 years (Table 3).