A Randomized Comparison of Group Cognitive-Behavioral Therapy and Group Interpersonal Psychotherapy for the Treatment of Overweight Individuals With Binge-Eating Disorder FREE

BINGE-EATING disorder (BED) is characterized by frequent and persistent overeating episodes that are accompanied by feelings of loss of control and marked distress, in the absence of regular compensatory behaviors.¹ The prevalence of BED in the general population is1.5% to 2.0%.^2- 3 Individuals with BED typically present to treatment with the multiple problems of binge eating, eating disorder psychopathology (eg, extreme concerns about eating, shape, and weight), psychiatric symptoms, and overweight.^4- 5 Accordingly, treatments for BED need to impact these multiple domains in both the short- and long-term.

Cognitive behavioral therapy (CBT) is the most well-established psychotherapeutic treatment for BED.^6- 8 Interpersonal psychotherapy (IPT) has been examined as an alternative treatment to target BED by directly addressing the social and interpersonal deficits observed among these individuals.^9- 10 Both CBT and IPT for BED provide greater short-term efficacy than wait-list control groups for the reduction of binge eating.^10- 13 The only comparison between active psychological treatments for BED that included follow-up to 12 months after treatment cessation has been between CBT and IPT.¹⁰ No immediate or long-term differences between IPT and CBT for binge-eating outcomes were found.

The present study evaluates whether the prior study's findings comparing CBT and IPT in the treatment of BED¹⁰ were reliable, especially given its modest sample size. Using a substantially larger sample than previously examined, we also sought to expand on prior BED research by simultaneously (1) using the most valid assessment of BED and specific eating disorder psychopathology, (2) evaluating time course repeatedly across multidimensional symptomatology through the 1-year follow-up, and (3) assessing the proposed mode specificity of these procedurally distinct treatments in reducing binge eating (ie, CBT through reducing problematic eating-, shape-, and weight-related attitudes and behaviors; IPT through improvements in interpersonal functioning, negative mood, and self-esteem).^10,14

This study was conducted at eating disorder clinics at Yale University(New Haven, Conn) and at San Diego State University (San Diego, Calif) and approved by the institutional review board at each university. Recruitment was conducted through media publicizing "compulsive overeating" treatment. Individuals telephoning the clinics were preliminarily screened for inclusion criteria (Table 1). Consent was obtained at the initial clinic visit; DSM-IV BED diagnosis and other study eligibility criteria were confirmed via interviews. Nine hundred seventy-four individuals expressed initial interest; 320 remained eligible based on telephone screens, and 195 met criteria after being interviewed. Of these, 162 were interested, eligible, and randomized (Figure 1). Participants were randomly assigned within each of the9 cohorts of 18 people to form 2 groups (IPT and CBT) of 9 participants each. Specifically, the 18 slots for a given cohort were blocked by sex and then randomly designated either IPT or CBT. Next, eligible participants were assigned to a given slot based on the order in which they were accepted into the study.

Both treatments consisted of twenty 90-minute, weekly group sessions and 3 individual sessions specifically addressing each participant's goals and progress. These standardized supplemental meetings occurred at critical time points (ie, pretreatment, midtreatment, and posttreatment) and were consistent with the stage and focus of each treatment. Patients also received weekly personalized, written feedback detailing progress. Groups were led by 2 therapists, at least one of whom was at the doctoral level (the second being either at the doctoral level or a psychology doctoral student). Therapists followed treatment manuals and were trained by Bruce Rounsaville, MD (IPT) and G. Terence Wilson, MD (CBT) (D.E.W. et al, unpublished data, 1996; D.E.W., unpublished data, 1993; and reference 15). One of us (D.E.W.) provided session-by-session supervision and feedback on each session's audiotapes to ensure manual adherence. All IPT groups and 6 of the 9 CBT groups were led by a therapist who conducted each treatment at some point.

Group CBT for BED is a triphasic, focal psychotherapy. In the first phase (sessions 1-6), behavioral strategies (eg, self-monitoring) help patients identify episodes of overrestriction and underrestriction and encourage normalization of eating patterns. During the second phase (sessions 7-14), patients learn cognitive skills to counter negative thoughts identified as predisposing binge eating. Cognitive restructuring helps patients challenge harsh stereotyped views of overweight and promotes acceptance of diverse body sizes. In the third phase (sessions 15-20), relapse prevention techniques, such as problem solving and coping with high-risk situations, are presented to help with maintaining changes. Patients are encouraged to identify reasonable goals and strategies for weight loss that will not promote binge eating.

Interpersonal psychotherapy is a brief, focused treatment¹⁶ adapted for BED and group format.^10,15,17 Interpersonal psychotherapy focuses on problem resolution within 4 social domains: grief, interpersonal role disputes, role transitions, and interpersonal deficits. The initial phase (sessions 1-5) involves examination of a patient's interpersonal history to identify the interpersonal problem area(s) associated with BED onset and maintenance, and a detailed plan is provided for the patient to work on specified problem area(s). Interpersonal deficits was the primary problem area for many IPT patients (49 [60.5%] of 81), followed by interpersonal role disputes (24 [29.6%] of 81), grief (5 [6.2%] of 81), and role transitions(3 [3.7%] of 81). During the intermediate phase of treatment (sessions 6-15), strategies are implemented to help patients make changes in identified problem areas. In the termination phase (sessions 16-20), patients evaluate and consolidate gains, detail plans for maintaining improvements, and outline remaining work.

Measures were administered at all time points, except assessment of non–eating disorder diagnostic psychiatric comorbidity, which occurred only at pretreatment. Posttreatment assessment occurred immediately following treatment cessation (median, 0.5 months), and follow-up assessments were approximately4 (median, 4.7), 8 (median, 8.7), and 12 (median, 12.6) months after treatment cessation. Structured clinical interviews were conducted by experienced assessors trained specifically in the interviews. The assessors (bachelor level or higher) had no therapeutic relationship with any of the participants they assessed. Although we attempted to keep assessors unaware of group assignment, this was not possible in all cases. To minimize this potential problem, assessors received ongoing supervision to ensure standardized administration of the interviews and underwent extensive training from the developers of the Eating Disorder Examination (EDE) (Christopher Fairburn, MD) and the Structured Clinical Interview for the DSM-III-R (Michael First, MD, Columbia University, New York, NY). Dr Fairburn also provided annual calibration workshops and ongoing EDE consultation. Interviews were audiotaped and randomly selected to evaluate coding accuracy at weekly supervisions by 2 of us (D.E.W. and R.R.W).

The EDE, 12th edition (12.0D),¹⁸ is an interview that assesses eating disorder psychopathology. It was adapted in accordance with DSM-IV research criteria to diagnose BED and was used to track changes in binge eating and eating disorder psychopathology. Binge-eating days were considered to be the number of days during the previous28 on which at least 1 objective bulimic episode (OBE) occurred, as defined by the consumption of an unusually large amount of food given the circumstances, accompanied by loss of control over eating.¹⁸ Eating disorder psychopathology was assessed by the EDE subscales of dietary restraint, eating concern, shape concern, and weight concern. Interrater reliability, assessed via 36 (18 pretreatment, 18 follow-up) randomly selected EDE interviews, revealed average interrater reliability (intraclass correlations) for subscales and number of binge-eating days from 0.83 to 0.99 (all P<.001). The Cohen κ for pretreatment diagnosis of BED was1.00 (P<.001).

The Structured Clinical Interview for the DSM-III-R^19- 20 is a semistructured interview designed to assess current and lifetime psychiatric disorders. The Symptom Checklist-90–Revised²¹ assesses psychiatric symptoms, the mean of which was standardized based on outpatient sex-specific clinical norms.

Weight was assessed on a Detecto (Cardinal Scale Manufacturing Co, Webb City, Mo) balance-beam scale; height was measured with a stadiometer. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters.

The Rosenberg Self-Esteem Questionnaire²² provided a measure of self-esteem. The Inventory of Interpersonal Problems²³ measured level of interpersonal problems. The Social Adjustment Scale²⁴ assessed current social functioning.

Before being informed of treatment assignment (already made at this point), participants read brief treatment descriptions and indicated their treatment preference. Using Likert-type scales from 1 to 10, participants also rated the degree to which each treatment made sense and their confidence in each treatment's success in reducing binge eating.

Following all treatment, 2 independent raters coded 9 CBT and 9 IPT audiotaped sessions using 30 items adapted from an integrity scale used to differentiate IPT and CBT for depression (S. D. Hollon, PhD, unpublished data, 1991). Neither rater was a therapist in this study, but both were trained in the delivery of the 2 treatments. Therapists were blind to which session was to be selected for rating treatment integrity, which was session 11 from each cohort. Raters were 100% accurate in judging actual treatment modality(ie, IPT vs CBT), and the treatment-specific indices significantly differentiated the treatments, whereas the nonspecific index did not (Table 2).

The completer sample size at posttreatment (n = 78 for CBT, n = 80 for IPT) provided 80% power to detect a treatment difference of greater than or equal to 0.8 binge days (d = 0.45), and at 12-month follow-up (n = 67 for CBT, n = 71 for IPT), a difference of greater than or equal to 1.8 binge days(d = 0.48).

Statistical analyses, using STATA version 6,²⁵ were based on the generalized estimating equation (GEE) approach,^26- 27 an extension of generalized linear models. With the GEE approach, participants with missing data at some, but not all, time points remain in the analyses. The GEEs tested hypotheses about treatment effects, time course, and treatment × time interactions, with linear, quadratic, and cubic components of time as the within-subjects factors, and treatment and interactions between time components and treatment as between-subjects factors. Higher-order terms were only included in the analyses of the follow-up period because the analyses of change from pretreatment to posttreatment were based on only 2 assessment time points. All higher-order terms were tested in the presence of lower-order terms. Interactions were analyzed using tests with 1 degree of freedom per interaction, comparing the model with the interaction term with a model that had all of the same terms except the interaction. An exchangeable error variance-covariance matrix was assumed in all cases. A2-tailed α level of .05 was used for all statistical tests. Data were considered missing if not collected by 17.5 months following treatment cessation, to avoid the effects of a possible assessment-availability bias on treatment effects and time course. This was the case for 5 participants whose 12-month data were excluded from all completer analyses.

The primary analyses included the posttreatment and follow-up time points for 3 outcomes: recovered (ie, no OBEs in the last month), improved to subclinical binge eating (ie, fewer than 4 OBE days in the last month),^28- 29 and being at or below a comparative level of eating disorder attitudes and behaviors. The latter rating was made based on whether the global scale of eating disorder psychopathology was at or below the mean of the 4 EDE subscales reported for a non-BED, overweight, treatment-seeking sample of 115 participants(20% male) with a mean BMI of 36.3 kg/m² and age of 40.8 years.³⁰ Given the dichotomous nature of the variables we defined for the outcomes in these analyses (eg, recovered vs not recovered), changes over time were analyzed using GEE logistic regression models (ie, logit link function and binomial error distribution). Intent-to-treat (Figure 2, Figure 3, and Figure 4)and completer analyses were used to test treatment differences at each time point for each of these 3 categorical outcomes.

For each secondary outcome, 2 GEEs were conducted: 1 evaluating change from pretreatment to posttreatment and 1 evaluating change from posttreatment through 12-month follow-up. Analyses with binge eating (days/episodes) were modeled based on an identity link function and the Poisson error distribution, to accurately reflect the count nature of these outcomes, including the high proportion of zero values at posttreatment and follow-up time points. Analyses with other outcomes were modeled based on the identity link function and a standard Gaussian error distribution. Completer analyses included data from all available assessment time points for each participant, irrespective of treatment completion. Data are given as mean ± SD unless otherwise indicated.

Treatment groups did not significantly differ on demographics (Table 3) or on pretreatment level of any outcome (all P values ≥.01) (Table 4). In addition, there were no significant site effects on any outcome at baseline or across time.

Therapy credibility ratings for CBT (8.8 ± 1.7) and IPT (8.6 ± 1.9) indicated that both treatment rationales made sense to patients, with no significant differences (t₁₅₇=0.95, P = .34) between the 2 treatments. In addition, there were no significant differences (t₁₅₇= 0.97, P = .34) in how confident patients felt that CBT (7.0 ± 2.3) or IPT (6.8 ± 2.2) would be successful. Ratings of treatment preference indicated that 46.5% of participants preferred CBT, while 53.5% preferred IPT.

Sixteen patients (9.9%) dropped out of treatment: 9 (11.1%) from the CBT group and 7 (8.6%) from the IPT group, a nonsignificant difference (χ²₁= 0.28, P = .6). There were no significant treatment differences in compliance (t₁₅₈= −1.91, P = .06), with CBT participants attending 16.6 ± 3.7 sessions (83.0%) and IPT participants attending17.7 ± 3.7 sessions (88.5%). Posttreatment interviews were conducted with 158 (97.5%) of the 162 participants. For the 4 through 12-month follow-ups,155 participants (95.7%) were available for at least 1 follow-up interview,147 (90.7%) for at least 2 follow-ups, and 133 (82.1%) completed all 3 follow-ups. The number of follow-up assessments completed (2.7 ± 0.8 of 3) did not differ by treatment (t₁₆₀= −0.93, P = .35).

Generalized estimating equation analyses revealed that there were no significant treatment × time interactions for any of the 3 categorical outcomes (all P values ≥.15). The predicted probability of being recovered decreased from approximately 78% to 65% from posttreatment to about 9 months following treatment cessation, remaining stable through12 months following treatment cessation. For recovery, there was a significant quadratic time effect (P = .03) for the follow-up period, reflecting a slight decline in the probability of being recovered toward the end of the 12-month follow-up assessments. The probability of binge eating at less than a clinically significant level decreased from approximately91% to 84% from posttreatment to 12 months following treatment cessation. For this outcome, there was a significant linear time effect (P = .007) for the follow-up period, reflecting a slight decrease in the probability of bingeing at or below the subclinical level. The probability of being at or below the normative level of global eating disorder psychopathology remained stable from posttreatment across the follow-up period (all P values ≥.73).

Intent-to-treat (Figure 2, Figure 3, and Figure 4) and completer rates for these 3 outcomes did not differ by treatment at any time point (all P values ≥.100); intent-to-treat rates are reported in Figure2, Figure 3, and Figure 4. Completer rates indicated that64 (82%) of 78 CBT patients and 59 (74%) of 80 IPT patients were abstinent from binge eating at posttreatment; 48 (72%) of 67 and 50 (70%) of 71, respectively, were abstinent at the 12-month follow-up. In addition, at posttreatment, 73(94%) of 78 CBT and 72 (90%) of 80 IPT patients were binge eating at less than a clinically-significant level of 4 days per month; 56 (84%) of 67 and63 (89%) of 71, respectively, were so improved at 12-month follow-up. Finally, 66 (85%) of 78 CBT and 60 (75%) of 80 IPT patients were found at posttreatment to have global eating disorder psychopathology at or below a sample of patients who were obese and not bingeing, thus substantially improved from baseline rates of 23 (28%) of 81 for CBT and 22 (27%) of 81 for IPT. At 12-month follow-up, these rates were 54 (82%) of 66 and 56 (79%) of 71, respectively.

For completers, binge eating decreased by 96% from pretreatment to posttreatment for CBT and by 94% for IPT. At the 12-month follow-up, binge eating was reduced by 90% and 93% for CBT and IPT patients, respectively, from pretreatment levels. Cognitive-behavioral therapy and IPT were both effective in reducing number of binge days during treatment (β = −.27; SE = 0.007; z = −11.68, P<.001), with no significant treatment-by-time interaction (z = 0.11, P = .91). Significant quadratic (β = −.02; SE = 0.004; z = −4.82, P<.001) and cubic (β = .002; SE = 0.007; z = 3.11, P = .002) time effects occurred through follow-up, with no significant linear or higher-order treatment-by-time interactions (all P values ≥.2). The pattern of results indicated an increase during the first 6 months of follow-up from about 0.5 to 2.0 binge days per month across treatments, a maintained level of binge eating 6 through 12 months following treatment cessation, and a slight upward trajectory toward the end of 12-month follow-up assessments (similar pattern detected for binge episodes). Observed mean numbers of binge days are presented along with SDs (Table 4), but the medians of 16 at pretreatment and zero at all posttreatment and follow-up time points are a better measure of central tendency than the mean because of high skewedness.

All other secondary outcomes showed a significant improvement from pretreatment to posttreatment (linear time effects, all P values <.001) except for BMI, which remained stable during the course of treatment. Results indicated that both CBT and IPT participants significantly reduced their levels of dietary restraint during treatment but had a different time course from posttreatment through the follow-up period. Specifically, CBT was stable during the follow-up period, while IPT was significantly greater than that of CBT at posttreatment and 4-month follow-up; however, the treatment groups were indistinguishable on this outcome thereafter. All other outcomes showed no treatment main or interaction effects from pretreatment to posttreatment or posttreatment through follow-up. In addition, all of these outcomes remained stable across follow-up, except interpersonal problems, which continued to improve, and BMI, which decreased (Table4). The number of patients who received additional treatment for binge eating (CBT, n = 3; IPT, n = 8) was not significantly different (χ²₁= 2.52, P = .11). These patients received individual psychotherapy, except one IPT patient, who received pharmacotherapy. In post hoc analyses of binge eating, removing those patients who received additional treatment did not change results.

Binge-eating abstinence was significantly related to BMI change, as those participants who were abstinent at posttreatment had reduced their BMI by 0.5 ± 1.5 kg/m² during the course of treatment, while those who were nonabstinent at posttreatment had increased their BMI by 0.4 ± 2.0 kg/m² (t₁₅₁= −2.33, P = .02). Similarly, of the participants who were abstinent at posttreatment, those who were abstinent at 12-month follow-up (n = 77) had further decreased their BMI by 1.0 ± 3.0 kg/m², whereas those who were no longer abstinent at 12-month follow-up (n = 26) had increased their BMI by 0.7 ± 2.9 kg/m² (t₁₀₄= −2.61, P = .01).

Interpersonal psychotherapy and CBT demonstrated equivalent, substantial improvements in the short- and long-term across the core symptomatology and associated problematic psychosocial functioning that characterizes BED. Treatment efficacy for binge eating was best at posttreatment, with slight increases in binge eating across both treatments during follow-up. At 12 months' posttreatment, intent-to-treat rates of recovery from binge eating (98 [61%] of 162), and the percentage of participants who had reduced binge eating to less than once weekly (123 [76%] of 162), did not differ significantly by treatment modality. Interpersonal psychotherapy took longer to achieve its full effects on dietary restraint than did CBT; but by the last 2 follow-up assessments, IPT had reached parity with CBT on this outcome. The equivalence of IPT and CBT on binge-eating replicates an earlier IPT and CBT BED treatment comparison study.¹⁰

The lack of differences in binge-eating change or time course between groups IPT and CBT for BED likely reflects a true similarity in their efficacy. The relatively large BED treatment sample and high retention allowed power to detect clinically meaningful treatment differences. In addition, outcome concordance occurred despite marked differences in the focus and delivery of CBT and IPT, as monitored throughout treatment and confirmed by blind independent ratings of treatment integrity.

Results extend prior findings by documenting similarity between CBT and IPT in the degree of and time course for decreases in most of the BED-related eating- and weight/shape-related attitudinal disturbances. The exception to treatment similarity—CBT's faster action on dietary restraint—is consistent with CBT's direct focus on reducing behaviors and cognitions that constitute excessive restriction of intake. This time course difference suggests that these distinct treatment modalities may be operating differently, although it is interesting to note that both treatments achieved high efficacy for multiple problems, even those not directly targeted by the intervention. For example, CBT resulted in changes in interpersonal functioning, as did IPT, despite this area being addressed only in IPT. In the same way, IPT resulted in sustained decreases in cognitive disturbances about eating, shape, and weight even though the intervention did not directly address these attitudinal problems. Moreover, both treatments incurred equivalent short- and long-term improvements in general psychological functioning.

During follow-up, CBT and IPT resulted in statistically, but not clinically,³¹ significant decreases in BMI on average across the entire sample. Nevertheless, weight maintenance is itself a positive outcome since individuals with BED likely experience a trajectory of weight gain.³² Furthermore, approximately one fourth of treated individuals lost more than 5% of their body weight, which has been associated with health benefits.³¹ Our findings and those of others³³ also suggest that patients with BED who cease binge eating tend to lose the most weight. For instance, among individuals recovered from binge eating at posttreatment, those who remained recovered at the 12-month follow-up had lost weight during the course of follow-up(−5.3 lb [2.4 kg]), whereas those who were no longer recovered at the end of follow-up had gained weight (+4.6 lb [2.1 kg]).

A comparative design testing active treatments (CBT, IPT) for BED was selected for this study because both treatments have demonstrated superiority to wait-listed control conditions, which are marked by persistent levels of binge eating and BED-related symptoms.^{10- 13,34- 35} However, the comparative design prevents determining whether CBT and IPT had specific effects for the treatment of BED. Unlike wait-listed conditions, a nonspecific treatment condition would control not only for assessments and the passage of time but also for patient expectation effects and other nonspecific therapeutic influences. The issue of treatment specificity has become increasingly relevant to BED treatment since some^36- 38(but not all³⁹) pharmacological studies of BED have evidenced short-term placebo response. Also, a recent natural-course study suggests relatively low long-term stability of BED among young, and partly subthreshold, individuals with DSM-IV BED.³²

Some recent specificity findings for the treatment of BED do exist. Among obese patients with BED, cognitive therapy has been found to result in higher binge-eating abstinence than behavioral therapy 6 months following treatment.⁴⁰ In addition, CBT was significantly more effective than a credible nonspecific psychotherapy treatment in a pilot study of recurrent binge eaters (most of whom met DSM-IV BED criteria) (J. Kenardy, PhD, et al, unpublished data, 2000). Nonetheless, larger specificity studies are still needed as are those examining the specificity of IPT. Also unknown are the characteristics of individuals with BED who may be more or less responsive to nonspecific intervention or require a specific treatment approach. Further, BED treatment research needs to examine generalizability across different samples since most studies to date have enrolled samples that are predominantly well educated and white.⁴¹

More information is clearly needed about the mechanisms by which CBT and IPT treat the multiple problems inherent in BED. The remarkable similarity in effectiveness and time course of CBT and IPT raises the possibility that they operate through shared mechanisms in the treatment of BED but does not necessarily indicate a similarity in treatment processes between them. In this study, we examined the mode-specific effects of each treatment at posttreatment and across follow-up; it may be that specificity occurred earlier, as recent research on bulimia nervosa has demonstrated.⁴² Clearly, future studies are required to assess potential mediators throughout the course of treatment.

The present reductions in binge eating and abstinence rates posttreatment and through the follow-up period are among the highest reported in the treatment research literature for BED.⁸ There were no long-term treatment differences in binge eating, dietary restraint, or other associated psychopathology. While both IPT and CBT facilitated slight average decreases in relative weight, abstinence from binge eating resulted in the most substantial weight loss. Binge-eating disorder was shown among earlier studies to be highly responsive to CBT,^10,12- 13 and our study indicates that IPT is an equivalent treatment alternative. Group CBT and group IPT have positive effects across the multiple domains of problematic functioning that characterize this disabling disorder.

Submitted for publication November 20, 2000; final revision received September 17, 2001; accepted October 12, 2001.

This research was supported by grants R29MH51384 and R29MH138403 from the National Institute of Mental Health, Bethesda, Md.

This study was presented in part at the Eating Disorders Research Society Annual Meeting, San Diego, Calif, November 20, 1999.

We thank Susan Beren, PhD, Michael Friedman, PhD, Anna Jost, BA, Danyte Mockus, MPH, Marlene Schwartz, PhD, and Marian Tanofsky, MA, for their assistance in the execution and analysis of this study. We are indebted to Christopher G. Fairburn, MD, and G. Terence Wilson, PhD, for their invaluable training of assessors and therapists, respectively. We give special thanks to Kelly D. Brownell for his support of the project and to Bruce J. Rounsaville, MD, for his significant contribution to the scientific quality of this project and helpful comments on a previous draft of this manuscript. We are grateful to Jamie Manwaring, BA, and Andrea Waldron, BA, for their help in the preparation of this manuscript.

Corresponding author and reprints: Denise E. Wilfley, PhD, San Diego State University/University of California–San Diego Joint Doctoral Clinical Psychology Training Program, 6363 Alvarado Ct, Suite 103, San Diego, CA 92120(e-mail: dwilfley@psychology.sdsu.edu).