Cognitive Behavioral Therapy and Fluoxetine for Binge Eating Disorder: Two-year Follow-up
The costs of publication of this article were defrayed, in part, by the payment of page charges. This article must, therefore, be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
New York State Psychiatric Institute, Unit 116, 1051 Riverside Drive, New York, NY 10032. E-mail: firstname.lastname@example.org
Objective: This study assessed the long-term effects of group behavioral treatment plus individual cognitive behavioral therapy (CBT) and/or fluoxetine in binge eating disorder (BED) patients.
Research Methods and Procedures: A total of 116 individuals were randomized to an initial five-month trial and were followed up over two years. Assessments, including binge frequency, weight, and self-report measures, were administered at pre-treatment, post-treatment, and ∼6, 12, 18, and 24 months after initial treatment.
Results: Across treatment groups, there was overall improvement over 29 months in binge frequency and in binge abstinence. The odds of binge abstinence 2 years post-treatment were 1.373 times the odds of binge abstinence immediately post-treatment. There was no significant change in weight over the two-year period. Subjects who received individual CBT evidenced lower binge frequency over the two-year follow-up period than patients who had not received individual CBT. Similarly, CBT was associated with increased rates of binge abstinence. There were no main effects of treatment assignment on weight over the two-year follow-up period. There was a significant advantage for fluoxetine assignment over the two-year follow-up period on depressive symptoms.
Discussion: The major significance of the study rests in its examination of the long-term effects of standardized interventions for BED. Our findings provide support for the ideas that short-term treatment may confer long-term benefit and that not all treatments are equivalent in the benefits they confer.
Binge-eating disorder (BED)1 is a commonly occurring syndrome that is characterized by recurrent uncontrolled consumption of large amounts of food. Although obesity is not required for the diagnosis of BED, treatment samples are almost uniformly obese. The binge eating behavior of individuals with BED is, by definition, distressing in its own right and, therefore, an appropriate target of treatment. However, since most individuals seeking treatment for BED are obese, it is not surprising that weight loss, either in the short term or the long term, is often, but not always, an explicit additional goal of treatment.
Treatment studies for BED conducted to date have used various forms of group and/or individual psychotherapy, medication, or combined medication and psychotherapy. In general, such studies have reported large short-term effects on binge eating. Effects on weight have been more variable, and significant weight losses have, for the most part, been observed only in: 1) studies using medications that are known to produce weight loss such as sibutramine (1), topiramate (2), and orlistat (3); or 2) studies carried out in weight control centers in which weight loss was the primary goal of treatment (4, 5, 6).
Although several short-term treatment studies for BED have been conducted, there have been few controlled studies of the long-term effects of treatment for BED on binge eating and on weight (7, 8, 9). In two studies using psychological approaches, treatments were associated with abstinence in a substantial percentage of patients, with good maintenance over one year. One study in which treatment included a weight management component reported weight loss during treatment but regain during one-year follow-up (7), while the other study did not include weight management treatment and did not detect significant weight change either at post-treatment or at one-year follow-up (8). In a third study, McElroy et al. (9) reported excellent maintenance of binge reduction and sustained weight loss in subjects who continued to take topiramate over the course of one year; however, a significant number dropped out during this time period. A six-year follow-up study of intensive inpatient treatment for BED, not emphasizing weight loss, reported good overall outcome in a majority of patients, but with only modest weight loss (10).
There are several important considerations in interpreting these results. First, the pattern of gradual weight regain after initial treatment-associated weight loss in treatments that emphasize weight management is not different in obese individuals with BED compared with those without BED (4). However, there have been hints that binge abstinence is associated with improved weight course over time (5, 7, 8, 11) and, conversely, that untreated binge eating may be associated with weight gain over time (12). Additionally, findings from treatment studies must be viewed in the context of the natural course of BED, which, in one five-year community study, was found to be rather unstable (12). Given the unstable natural course of binge eating, variable findings in the few studies that have been done, and uncertainty regarding short- and long-term effects of binge reduction on weight, further systematic study is needed.
The present study was designed to measure the added benefit attributable to two adjunctive treatments, individual cognitive behavioral therapy (CBT) and fluoxetine, when administered in the context of a group behavioral treatment for BED. The short-term effects of the two adjunctive treatments have been reported previously (13). To assess the long-term course of patients with BED after initial treatment as well as to study the possible ongoing effects of the two adjunctive interventions, we followed the clinical course of study participants over a two-year period after the end of their initial five-month treatment.
Research Methods and Procedures
The initial trial included overweight or obese men and women between the ages of 18 and 70 years, with a maximum weight of 350 lbs. (159 kg). All subjects met DSM criteria for BED for at least 6 months. Exclusion criteria (13) included current medical illness not adequately controlled with diet or medication, certain psychiatric disorders, concurrent eating/weight control treatments, and current use of antidepressants, mood stabilizers, or appetite suppressants.
All 116 subjects with BED who were randomized to the controlled trial were asked to participate in the follow-up study. Subjects, including those who had prematurely discontinued treatment (42 of 116 subjects; 36%), were contacted by telephone by a member of the research staff at 6-month intervals over a 2-year period. Individuals were paid $100 for completing the full battery of interviews and self-report assessments. Subjects who mailed in self-report assessments but did not complete the interviews either in person or by telephone were paid $50. Weight was also obtained on individuals for whom in-person evaluations were conducted.
After the 5-month initial phase of treatment, subjects who attained a reduction in the frequency of binge days of at least 75% were asked to enter the 2-year maintenance phase. These individuals continued to attend monthly maintenance groups and were asked to continue double-blind study medication for 18 of the 24 months, which provided an opportunity to observe patients for six months off fluoxetine. They were also asked to not engage in other eating disorder or weight loss treatment during the two-year follow-up period. However, a number of these patients did participate in such treatments during the follow-up period (see “Results”). Non-responders were discontinued from study medications and were referred to other treatments, as appropriate, but were also allowed the option of continuing to attend monthly maintenance groups.
In the initial trial, all subjects received group behavioral weight control treatment based on the LEARN program of Brownell (14) for 5 months. Within each behavioral weight control group, approximately one half of subjects were assigned to fluoxetine and one half to placebo. Medication assignment and management were conducted in a double-blind fashion. In addition, approximately one half of the subjects were randomly assigned to receive individual CBT, based on the manual of Fairburn et al. (15), in addition to the group treatment. The other one half of subjects received group treatment only, with no individual therapy. Each group as a whole was randomly assigned to receive either additional individual CBT or no additional psychotherapy, to avoid inadvertent dissemination of components of individual CBT to group co-members who were not assigned to individual CBT. See Devlin et al. (13) for a more detailed description of the initial study design and treatments.
The monthly maintenance group sessions were based on the LEARN program, but rather than reviewing specific chapters in detail, a different component was featured each month (Lifestyle, Exercise, Attitudes, Relationships, Nutrition, Overview/Synthesis). Subjects also discussed progress toward individually set monthly goals. Before each group session, subjects who were continuing with double-blind medication met with their study doctor for medication management. A research assistant also obtained binge frequencies and weights from each subject.
Assessments were administered at pre-treatment, post-treatment, and ∼6, 12, 18, and 24 months after cessation of treatment. For all study participants, the diagnosis of BED was established by semi-structured interview, using items from the Eating Disorder Examination (EDE), 12th edition (16), with additional items to assess BED diagnostic criteria. In addition, in keeping with the DSM (17) Appendix B criteria, the duration of binge eating was revised from three months to six months before the assessment, and the number of binge days (days on which one or more objective binge episodes occurred) rather than the number of individual binge episodes was determined. Other Axis I psychiatric disorders were diagnosed using the Structured Clinical Interview for DSM (18). End-of-treatment and follow-up diagnostic assessments were determined by an interviewer through use of the EDE-BED and an abbreviated version of the Structured Clinical Interview for DSM. Interviewers were blind to the subject's medication assignment but were not blind to their therapy assignment, as each group as a whole was randomly assigned to receive either additional individual CBT or no additional psychotherapy. Assessments also included weight measured using a balance beam scale, height measured using a stadiometer (pre-treatment only), BMI, waist-to-hip ratio, and self-report instruments detailed below.
Subjects were asked to complete the following self-report questionnaires at all time-points: the Beck Depression Inventory (BDI) (19), the Body Shape Questionnaire (BSQ) (20), the Binge Eating Scale (21), the Brief Symptom Inventory (BSI) (22), and the Three-Factor Eating Questionnaire (TFEQ) (23).
The study was designed to assess the long-term effects of the initial five-month treatment. Post-treatment and follow-up values for primary and secondary outcomes were modeled as linear functions of time, medication (fluoxetine vs. placebo), individual CBT (yes vs. no), and all 2- and 3-way interactions among these three factors, adjusting (as a main effect only) for the pre-treatment value of that outcome measure. To account for the correlation between repeated observations on a subject over time and for the clustering effect due to subjects belonging to the same group-therapy cohort, mixed effects models (24) were used. The outcomes binge frequency, BSI, and BDI, which follow Poisson distribution, and binge abstinence, which is a binary variable, were modeled with generalized estimating equation using the appropriate link. Inferences about the effect of time, medication, and therapy are based on the final model for each of the outcomes, where the final models are such that all higher order terms are statistically significant and all lower order terms, contained in the higher order interactions, are kept in the model regardless of their statistical significance, and all main effects are retained. A backward elimination procedure was used to arrive at the final model for each outcome. Everywhere, statistical significance was judged at a level of α = 0.05. PROC MIXED and PROC GLIMMIX in SAS (25) were used to fit these models.
Two of the original 116 subjects were excluded from follow-up analyses of weight and BSQ because they underwent bariatric surgery during the follow-up period and were, therefore, likely to be outliers on these variables. However, re-running the analyses including these subjects yielded no substantive differences in observed effects.
On average, subjects were 43 (± 12) years old, had a mean weight and BMI of 115.0 (± 21.8) kg and 40.9 (± 6.9) kg/m2, respectively, and had experienced binge eating for 27.3 (± 14.7) years). Of the 116 subjects, 89 (77%) were white, 14 (12%) were black, 12 (10%) were Hispanic, and 1 (1%) was mixed or other. The gender breakdown was 90 (78%) female and 26 (22%) male. See Devlin et al. (13) for a more detailed description of the subjects and the selection process.
Assessment Completion Rates
Efforts were made to obtain assessments at each time-point from all 116 subjects randomized to treatment. Complete data sets (post-treatment plus all four follow-up points) were obtained from 72 participants (62%); three follow-ups were obtained from 13 participants; two follow-ups were obtained from 5 participants; and one follow-up was obtained from 12 participants. No follow-up data were collected from 14 participants (12%). The number of follow-up assessments completed did not differ by treatment [(χ2)12 = 6.747, not significant]. Although 102 (88%) subjects completed at least one follow-up visit, the maximum follow-up rate at any particular follow-up visit was 88 (76%). This reflects the fact that a number of individuals completed follow-up visits at some time-points but not others (see Table 1 for N at each time-point). Thus, the proportion of individuals contributing at least some follow-up data was greater than the proportion attending any given follow-up visit.
Table 1. . Pre-treatment (Pre-Tx), post-treatment (Post-Tx), and follow-up outcome measures (raw data)
|Binge frequency 28-day||116; 16.3 (6.5)||116; 4.5 (6.7)||88; 1.8 (3.6)||88; 3.3 (5.6)||87; 2.2 (5.2)||87; 1.6 (4.0)|
|Binge abstinence [N (%)]||0 (0%)||54 (47%)||60 (68%)||54 (61%)||61 (70%)||64 (74%)|
|Weight (kg)||116; 115 (21.8)||116; 113 (23.6)||82; 111 (23.2)||79; 113 (23.2)||63; 113 (23.2)||62; 115 (23.6)|
|BDI||115; 15.3 (9.0)||115; 8.3 (8.4)||86; 7.3 (6.8)||88; 7.6 (8.6)||81; 7.5 (7.8)||81; 7.6 (8.9)|
|BSQ||115; 132 (31.5)||115; 106 (36.5)||86; 98.6 (32.4)||88; 101 (37.8)||82; 97.7 (33.2)||81; 94.5 (33.1)|
|TFEQ restraint||115; 7.28 (3.5)||115; 10.8 (4.2)||86; 11.6 (4.5)||87; 10.5 (4.5)||83; 10.2 (4.4)||81; 10.5 (4.7)|
|TFEQ hunger||115; 10.2 (3.1)||115; 7.4 (3.9)||86; 7.5 (3.9)||87; 7.5 (4.1)||83; 8.2 (3.9)||81; 8.2 (3.8)|
|TFEQ disinhibition||115; 13.4 (2.1)||115; 10.3 (4.1)||86; 9.9 (3.6)||87; 10.3 (3.8)||83; 10.6 (3.4)||81; 10.5 (3.6)|
|BSI||115; 42.4 (29.7)||115; 25.6 (29.2)||86; 21.4 (26.2)||87; 22.5 (31.9)||84; 27.8 (32.2)||81; 22.8 (29.4)|
Of subjects who complied with follow-up assessments, in-person assessments, including weight, were obtained on 82 of 88 subjects (93%) at the 6-month follow-up, on 79 of 88 subjects (90%) at the 12-month follow-up, on 63 of 87 subjects (72%) at the 18-month follow-up, and on 62 of 87 (71%) subjects at the 24-month follow-up. For the remaining subjects, interviews were conducted by telephone, and self-report assessments were obtained by mail.
Treatments During Follow-up Phase
Treatment during the follow-up phase included attendance at monthly maintenance groups, with study medication management for those continuing double-blind medication, treatment outside the study, or both. For the 74 subjects who attended at least one maintenance group, the mean duration of attendance was 16.7 ± 8.0 months (range, 1 to 24 months). The mean number of sessions attended was 11.3 ± 6.4 sessions (range, 1 to 23 sessions). Of the 74 subjects who attended maintenance sessions, 37 had been randomized to individual CBT and 37 to no CBT (χ21 = 0, p = 1.0). Of these 74 subjects, 42 were randomized to fluoxetine and 32 to placebo (χ21 = 1.351, p = 0.245).
Regarding medication compliance, 57 subjects had their blind broken at or before the end of the initial phase of treatment (including the 42 initial phase dropouts), 33 subjects had their blind broken after the end of active treatment but before the 18-month follow-up visit, and 26 subjects had their blind broken at the 18-month follow-up visit.
Regarding treatment outside the study during the two-year follow-up period, of the 116 subjects randomized, 61 individuals had no known additional treatment, 21 had medication treatment, 19 had some form of psychosocial treatment (commercial weight management, individual or group psychotherapy, or self-help), and 15 had both medication and psychosocial treatment either in combination or at different time-points during follow-up. Of the 36 patients who reported medication treatment, the majority (29) reported taking antidepressants, 4 reported taking weight loss medications (including over-the-counter), and 3 reported taking both antidepressants and weight loss medications, in combination or at different times during the follow-up period, or other psychotropic medications. Two patients had bariatric surgery procedures.
Response to Treatment: Overall Effects of Time
Mean values of primary and secondary outcome measures for subjects who completed follow-up assessments at each time-point are reported in Table 1, collapsed over all four treatment groups. Model-based estimates of changes over the 24-month follow-up period are reported in Table 2, second column. In cases where there was no statistically significant interaction between a treatment (medication and/or individual CBT) and time, i.e., when the final model contained only main effect of time, the “estimated change” column reports the time effect common for all treatment groups. Across treatment groups, there was overall improvement over time in binge frequency and in binge abstinence. The ratio of frequency of binge episodes 2 years post-treatment as compared with binge frequency immediately post-treatment is 0.689, i.e., a 31% decrease in frequency. The odds of binge abstinence 2 years post-treatment were 1.373 times the odds of binge abstinence immediately post-treatment. There was no significant change in weight over the two-year period. For secondary variables, there was overall improvement over time in BSQ and overall worsening over time in TFEQ hunger. The time effect on BSQ score was different depending on whether or not subjectsreceived individual CBT, with BSQ scores during the two-year follow-up improving more rapidly for subjects who did not receive individual CBT than for those who received individual CBT. The time effect on TFEQ restraint depended on medication, with restraint scores for subjects on fluoxetine decreasing significantly more over the two years of follow-up than the scores of subjects on placebo.
Table 2. . Model-based estimates of treatment effects over 24 months
|Binge frequency 28-day||0.689 (0.049)†||F test||0.01||10.45||1.20||27.80|| || |
| || ||NDF/DDF||1/87.87||1/94.3||1/90.21||1/73.84|| || |
| || ||p||0.908||0.002||0.277||<0.0001|| || |
|Binge abstinence||1.373 (0.104)‡||F test||0.56||4.28||0.16||17.52|| || |
| || ||NDF/DDF||1/99.33||1/99.31||1/98.66||1/228.8|| || |
| || ||p||0.456||0.041||0.689||<0.0001|| || |
|Weight (kg)||0.484 (0.309)||F test||1157.13||0.01||0.01||2.46|| || |
| || ||NDF/DDF||1/131||1/132||1/132||1/137|| || |
| || ||p||<0.0001||0.911||0.914||0.119|| || |
|BDI||0.967 (0.027)†||F test||38.15||1.62||4.27||1.43|| || |
| || ||NDF/DDF||1/103.8||1/92.98||1/91.7||1/82.62|| || |
| || ||p||<0.0001||0.207||0.041||0.235|| || |
|BSQ||Individual CBT −0.302 (1.975)||F test||109.56||1.39||3.72||10.77||4.00|| |
| || ||NDF/DDF||1/105||1/208||1/105||1/353||351|| |
| ||No Individual CBT −2.642 (0.805)||p||<0.0001||0.238||0.056||0.001||0.046|| |
| || || || || || || || || |
|TFEQ restraint||Fluoxetine −0.412 (0.368)||F test||28.24||1.84||0.73||0.18|| ||4.95|
| || ||NDF/DDF||108||109||233||139|| ||137|
| ||Placebo 0.065 (0.153)||p||<0.0001||0.177||0.392||0.670|| ||0.027|
| || || || || || || || || |
|TFEQ hunger||0.225 (0.073)||F test||60.42||0.39||0.63||9.32|| || |
| || ||NDF/DDF||1/112||1/113||1/113||1/367|| || |
| || ||p||<0.0001||0.536||0.427||0.002|| || |
|TFEQ disinhibition||0.102 (0.091)||F test||19.72||2.00||0.13||1.25|| || |
| || ||NDF/DDF||1/109||1/110||1/110||1/156|| || |
| || ||p||<0.0001||0.159||0.720||0.265|| || |
|BES||−0.156 (0.194)||F test||23.01||1.84||0.11||0.65|| || |
| || ||NDF/DDF||1/106||1/105||1/105||1/356|| || |
| || ||p||<0.0001||0.177||0.741||0.422|| || |
|BSI||1.001 (0.035)†||F test||117.86||0.60||3.16||0.00|| || |
| || ||NDF/DDF||1/113.6||1/102.3||1/101.7||1/80.78|| || |
| || ||p||<0.0001||0.441||0.078||0.967|| || |
Effects of Treatment
Model-based treatment effects and time-by-treatment interactions over the two-year period beginning at post-treatment are also summarized in Table 2. Subjects who had received adjunctive individual CBT evidenced lower binge frequency over the two-year follow-up period than patients who had not received individual CBT. There was no CBT-by-time interaction, suggesting that the superior post-treatment status of subjects who received CBT over that of subjects who did not was maintained over the two-year follow-up period, and their post-treatment trajectories were not different. Similarly, CBT was associated with increased rates of binge abstinence during the one month preceding follow-up assessments from post-treatment to 24-month follow-up; once again, there was no CBT-by-time interaction. There were no main effects of treatment assignment on weight over the two-year follow-up period and no treatment-by-time interactions.
As was the case for acute treatment response, there were relatively few effects of treatment assignment on secondary variables over a two-year period beginning at post-treatment. For BSQ, there was a marginally significant advantage for fluoxetine assignment (p = 0.056) and a CBT-by-time interaction, with subjects who had received no individual treatment showing an improvement over time, in contrast to subjects who had received CBT, whose scores remained relatively constant over time, suggesting a possible floor effect. For BDI, as was the case for the acute treatment effect, there was a significant advantage for fluoxetine assignment over the two-year follow-up period. For TFEQ restraint, there was a fluoxetine-by-time interaction, with subjects who had been assigned to fluoxetine showing less restraint over time, while the placebo group's scores remained stable. There were no significant main effects of treatment or interactions for other variables, including TFEQ disinhibition score, Binge Eating Scale score, BSI score, Rosenberg Self-Esteem score, and Inventory of Interpersonal Problems score.
In our initial description of the results of a randomized controlled trial of individual CBT and fluoxetine as adjuncts to group behavioral treatment for BED (13), we reported that, across treatment cells, there was substantial improvement from pre-treatment to post-treatment in binge eating and in most other outcome variables, with the notable exception of weight. The current analysis was designed to gauge the ongoing effects of initial treatment and time-related trends over the first two years after initial treatment.
In the current report, when the period of outcome assessment was extended from a single post-treatment time-point to a two-year period after treatment, we found, once again, that participants across treatment cells showed a dramatic reduction in the frequency of binge eating from pre-treatment and, in fact, showed slight continued improvement over the two-year follow-up period. We continued to observe no significant overall treatment effect on weight and a beneficial effect of the overall treatment program on self-report measures relating to mood, eating, body image, and general psychopathology that, for the most part, neither deteriorated nor improved over time. It is worth noting that the lack of observed effect on weight may be attributable to the modifications made to the LEARN program for this study, emphasizing binge cessation over weight loss.
We also studied the benefits of CBT and fluoxetine treatments as adjuncts to group behavioral weight control over the two-year follow-up period. As found previously when binge eating was examined at a single post-treatment time-point, the benefit of having received individual CBT in the initial phase of the study was apparent over two years of follow-up. Despite a relatively large number of premature medication discontinuations and crossovers, the beneficial effects of initial assignment to fluoxetine treatment on depressive symptoms also continued. However, there were no beneficial effects of fluoxetine on binge eating; this replicates findings reported by Grilo et al. (26).
The major significance of the study rests in its examination of the long-term effects of standardized interventions for BED. Since BED is a condition that is well known to be unstable over time (27), short-term treatment effects must be viewed with caution. Ideally, follow-up periods of even longer than two years would be of benefit in fully assessing the impact of treatment interventions for binge eating, including their impact on psychological and medical sequelae of binge eating and obesity.
The study design and analysis did not focus on the effects of the maintenance intervention per se. The monthly maintenance groups were designed as informal sessions to encourage continued study participation, offer minimal support over the two-year period, and provide a venue for continued medication monitoring. We cannot know from this study the degree to which even a minimal ongoing group intervention, along with continuation of study medication, may have contributed to the sustained improvements that we observed. However, the variable attendance at monthly maintenance groups and the fact that only a small minority of patients who entered the study actually continued double-blind medication for the prescribed 18 months of follow-up argue against a large contribution of ongoing study-associated treatments to the sustained benefits we observed. Previous follow-up studies have reported good maintenance of reductions in binge eating at one-year post-treatment (7, 8). Our findings are consistent with these results and suggest that the beneficial effects of treatment for binge eating may persist for even longer periods.
There were several limitations to this study. As previously discussed (13), the decision to adapt the standard group behavioral weight control treatment outlined in the LEARN manual to prioritize normalization of eating rather than weight loss may have limited the effect of the overall treatment program on weight loss. In addition, the initial treatment dropout rate was significant, although it is of note that many patients who had dropped out of initial treatment or did not attend maintenance groups nonetheless returned for follow-up assessments. Moreover, there were many premature medication discontinuations and crossovers, and many patients received treatment outside the study during the follow-up period.
Although a substantial majority (88%) of subjects returned for at least one follow-up visit, and our analytic methods made use of all available data, missed follow-up visits represent a potential source of error. Of particular concern is the possibility of differential effects of follow-up participation on observed differences among treatment groups. While this cannot be ruled out, the likelihood that this was a significant source of systematic error is lessened by the similar rates of follow-up participation among treatment groups.
Another factor to be considered is the degree to which treatment during the follow-up period was contingent on response to initial treatment, which differed among treatment groups. Specifically, subjects who responded well were asked to continue double-blind medication and encouraged to attend maintenance groups. However, subjects who did not respond well, after their blind was broken, were both referred for alternative psychosocial and/or pharmacological treatment and invited to attend maintenance groups if they wished. Thus, we think that it is unlikely that differential opportunities for treatment affected long-term outcome.
In summary, this study begins to address the need for investigation of the long-term maintenance of change after treatment in obese patients with BED. While our findings provide support for the ideas that short-term treatment may confer long-term benefit and that not all treatments are equivalent in the benefits they confer, further studies of the long-term course of BED with and without treatment, particularly studies using state-of-the-art weight loss interventions, are clearly needed.
This work was supported in part by National Institute of Mental Health Grant MH54612. M.J.D. received research support from Eli Lilly and Co. in the form of Prozac and matching placebo provided for this study. B.T.W. received research support from Eli Lilly and Co. The authors thank Dr. Pamela S. Raizman, Dr. Sara Wolk, Dr. Laurel Mayer, Dr. Janel Carino, Dr. Dara Bellace, Claudia Kamenetz, Ilyse Dobrow, and Gillian Boudreau.
Nonstandard abbreviations: BED, binge-eating disorder; CBT, cognitive behavioral therapy; EDE, Eating Disorder Examination; BDI, Beck Depression Inventory; BSQ, Body Shape Questionnaire; BSI, Brief Symptom Inventory; TFEQ, Three-Factor Eating Questionnaire.