Review and Meta-analysis of Pharmacotherapy for Binge-eating Disorder

Authors

  • Deborah L. Reas,

    1. Institute of Psychiatry, Faculty of Medicine, University of Oslo, Oslo, Norway
    2. Regional Eating Disorders Service (RASP), Psychiatric Division, Ullevål University Hospital, Oslo, Norway
    Search for more papers by this author
  • Carlos M. Grilo

    Corresponding author
    1. Department of Psychiatry, Yale University School of Medicine, New Haven, Connecticut, USA
    2. Department of Psychology, Yale University, New Haven, Connecticut, USA
    Search for more papers by this author

(carlos.grilo@yale.edu)

Abstract

This study evaluated available controlled treatment studies to determine utility of pharmacotherapy for binge-eating disorder (BED). The authors identified randomized placebo-controlled trials testing pharmacotherapy-only treatments and controlled trials testing pharmacotherapy with psychotherapy treatments. Meta-analysis was performed on placebo-controlled trials with data for attrition, remission, and weight loss. Qualitative review was performed on remaining controlled treatment literature. A total of 33 studies were considered of which 14 studies with a total of 1,279 patients were included in the meta-analysis of pharmacotherapy-only treatment and 8 studies with a total of 683 patients were included in the qualitative review of pharmacotherapy combined with psychotherapy interventions. No evidence suggested significant differences between medication and placebo for attrition. Evidence suggested that pharmacological treatments have a clinically significant advantage over placebo for achieving short-term remission from binge eating (48.7% vs. 28.5%) and for weight loss, although weight losses are not substantial. No data exist to allow evaluation of longer-term effects of pharmacotherapy-only treatment for BED. Combining medications with psychotherapy interventions failed to significantly enhance binge outcomes, although specific medications (orlistat, topiramate) enhanced weight losses achieved with cognitive behavioral therapy and behavioral weight loss. In summary, BED patients can be advised that certain pharmacotherapies may enhance likelihood of stopping binge eating short term, but that longer-term effects are unknown. Although some weight loss may occur, it is unlikely to be substantial with available medications. Combining medications with cognitive or behavioral treatments is unlikely to enhance binge outcomes, but specific medications (orlistat, topiramate) may enhance weight losses, albeit modestly.

Introduction

Binge-eating disorder (BED), a research category in the Diagnostic and Statistical Manual of Mental Disorders IV (DSM-IV) (1), is characterized by recurrent binge eating without inappropriate weight-control behaviors. BED is a prevalent (2) and stable problem (3) associated with heightened psychological (4), psychosocial (2), psychiatric (5), and medical (6) impairment. BED is strongly associated with obesity (2), although obese persons with BED have significantly greater eating and psychological disturbances than obese persons without BED (7), and new evidence suggests that BED represents a distinct familial phenotype in obese individuals (8).

The first generation of treatment studies for BED identified some effective psychological treatments, as highlighted by critical qualitative reviews (9) and quantitative meta-analytic studies (10). In contrast, although a growing number of pharmacotherapy treatment studies have reported that certain medications have efficacy for BED, the clinical significance of this emerging literature is difficult to interpret due to mixed findings and methodological limitations. An early meta-analysis reported cautious support for selective serotonin reuptake inhibitor (SSRI) antidepressant medications (10), but the continued publication of randomized controlled trials (RCTs) testing various pharmacotherapies, including recent larger-scale studies, warrants a current critical review. The present study aimed to systematically evaluate the available controlled treatment research findings to determine the utility of pharmacotherapy-only treatments and pharmacotherapy combined with psychotherapy treatments for BED. This evaluation used a meta-analysis on placebo-controlled, pharmacotherapy-only trials that produced data for major outcomes (attrition, remission, and weight loss). Pharmacotherapy trials that included psychotherapy interventions were evaluated using a qualitative critical review.

Methods and Procedures

Search strategy

We aimed to systematically identify and synthesize existing evidence for the use of pharmacological agents in the treatment of BED. Due to the relatively recent addition of BED as a research category to the DSM-IV (1), literature published prior to 1985 offered little to no clinical data specific to the current definition of BED. Accordingly, we attempted to locate all studies published between 1985 and 31 January 2008. To enhance the specificity of this review and to improve upon earlier shortcomings of qualitative reviews, we excluded literature pertaining to mixed eating disorder samples, nonpurging bulimia nervosa, atypical eating disorders, eating disorders not otherwise specified except for the specific example of BED, or the ICD-10 category, “overeating associated with other psychiatric condition” (11). Only randomized placebo-controlled trials evaluating the effects of a pharmacological agent as a primary treatment for BED (pharmacotherapy-only) or RCTs comparing pharmacotherapy vs. or combined with psychosocial treatments (additive or combined designs) were considered. Thus, all open-label trials and case series studies were excluded. A previously published RCT investigating d-fenfluramine (12) was also not considered in the meta-analysis because this medication was withdrawn from the market due to adverse events and safety concerns.

All eligible studies were examined for outcome data. As will be summarized below, data usable for meta-analysis for three major clinical outcomes (attrition, remission from binge eating, and weight loss) were reported by most pharmacotherapy-only studies. Due to the particularly wide scope of the additive or combined trials of pharmacotherapy and psychotherapy treatments, a critical qualitative review was performed in lieu of meta-analysis. Attrition was consistently defined as withdrawal or noncompletion for any reason, and remission was defined as 100% reduction in binge-eating episodes, although studies varied in the trial duration or required time frame. Attrition is a useful proxy for the general acceptability of a treatment. Remission from binge eating is a useful overall clinical outcome, although it does not necessarily capture other important aspects such as cognitive features of eating disorder psychopathology or associated psychological distress. Weight-change (continuous) data were reported by several studies and were included in the meta-analysis.

Relevant studies were identified with major computerized literature databases including: Medline, PubMed, PsychLit, Cochrane Library, and Science Direct. Relevant search terms included “binge”, “binge eating”, and “binge eating disorder.” Studies were also identified by cross-referencing and hand-searching obesity, eating disorder, psychiatry, and pharmacology journals. A total of 33 studies were considered: 14 studies were included in the meta-analysis of pharmacotherapy-only studies and 19 were excluded for various methodological reasons (see Table 1). A summary of characteristics of pharmacotherapy-only studies used for the meta-analysis is provided (see Table 2). A total of eight controlled studies were qualitatively reviewed for outcomes to determine the utility of pharmacotherapy and psychotherapy interventions in additive/combined designs (see Table 3).

Table 1.  Characteristics of clinical trials excluded from the meta-analysis of controlled pharmacotherapy-only trials
inline image
Table 2.  Randomized placebo-controlled trials of pharmacotherapy-only treatment for binge-eating disorder
inline image
Table 3.  Summary of additive or combined trials of pharmacotherapy and psychotherapy treatment for BED
inline image

Data collection and analysis

Data were entered into RevMan 4.2.7, a statistical program developed by the Cochrane Collaboration (Review Manager, Version 4.2 for Windows; Cochrane Collaboration, Oxford, UK) for systematic reviews. Effect sizes for dichotomous data (remission, attrition) were evaluated by the relative risk, also referred to as risk ratio (RR), with 95% confidence intervals (CIs). Effect sizes for weight loss data were expressed as weighted mean differences (WMDs) using Hedges g (ES = X1 − X2/Spooled) with a 95% CI. To increase power and precision, a summary variance-weighted effect size was calculated to lend more weight to larger studies. Whenever possible, intention-to-treat analyses were performed. For dichotomous remission data, missing data were treated based on the assumption of a negative outcome (“no remission achieved”) and baseline values were carried forward. For continuous data, intention-to-treat data were used whenever possible; any completed-only data was clearly labeled to draw a distinction.

Heterogeneity was examined using a standardized χ2 test and an I2 test of heterogeneity, such that I2 < 30% indicated mild heterogeneity, I2 = 30–50% indicated some heterogeneity, and I2 > 50% indicated considerable heterogeneity. Results were synthesized using a fixed effects model for homogenous data or a random-effects model for heterogeneous data. If significant heterogeneity was found via graphical inspection or by I2 tests, trials contributing to considerable heterogeneity were removed in sensitivity analyses and potential reasons accounting for such heterogeneity were discussed. Subgroup analyses were performed to compare efficacy across different medication classes. Forest plots are used to display the effect estimates and CIs for individual studies and meta-analyses; a statistically significant summary statistic (treatment effect size) is visually displayed by a CI falling entirely within the clinically relevant range and does not cross the line of no effect.

Results

Pharmacotherapy-only studies: overview

Fourteen studies met the inclusion criteria and provided data eligible for at least one outcome measure of interest. Table 2 provides a summary of each of the study's details including the nature of the sample, the design including dosing information and duration of the trial, and major outcomes such as attrition and clinical effects. A total of 1,279 participants (mostly women) were included across the 14 trials. Twelve of the studies were conducted in the United States, and one each in Brazil and Italy. Recruitment of participants was almost exclusively through media advertisement. Twelve of the fourteen RCTs were funded by the pharmaceutical industry (i.e., supported by the drug manufacturer), one (13) RCT did not disclose or acknowledge any funding source, and one (14) was funded by the National Institutes of Health, although medication and placebo pills were provided by the drug manufacturer.

Types of interventions included three main classes of drugs, including antidepressant SSRI/ serotonin-norepinephrine reuptake inhibitor medication, antiepileptic medication, and obesity medication. Seven trials were identified, which compared an SSRI antidepressant with placebo, including two studies of fluoxetine, two studies of fluvoxamine, one study of sertraline, one study of citalopram, and one study of escitalopram (14,15,16,17,18,19), and one trial compared a serotonin-norepinephrine reuptake inhibitor antidepressant with placebo (20). Three trials compared antiepileptic medications (topiramate and zonisamide) with placebo (21,22,23). Three trials compared obesity medication (sibutramine) vs. placebo (13,24,25). Some trials used flexible dosing of medication whereas others used fixed dosing throughout the course of treatment (Table 2 describes dosing for all medications). The trials were of short duration, averaging 12.2 weeks with a range from 6 (refs. 17,18,26) to 24 weeks (25). To date, none have reported follow-up data after medication discontinuation.

Attrition: meta-analysis findings

Figure 1 summarizes attrition data which were provided by 13 of the 14 RCTs. In the placebo-treated condition, the overall attrition rate was 31.5%, compared with an overall attrition rate of 30.4% in the medication-treated condition. There was no evidence to suggest a clinically significant difference exists between medication and placebo for attrition (N = 13; n = 1,254; RR = 1.02: 95% CI: 0.79–1.32). For the SSRI subgroup, substantial heterogeneity was noted (I2 = 57.7%) and several sensitivity analyses were performed. The heterogeneity appeared due to the trial by Arnold et al. (26), which reported 57% attrition for the placebo group. Excluding this trial from the analysis resulted in a slight advantage to placebo for this subgroup analysis (RR = 1.66; 95% CI: 1.01–2.74; Z = 1.99; P =.05), but did not affect the pooled meta-analysis.

Figure 1.

Attrition rate: medication vs. placebo. Pearlstein et al. (15) reported an overall (5/25) 20% dropout rate, but did not specify attrition by group (denoted by an asterisk).

Effect of treatment on binge-eating remission: meta-analysis findings

Figure 2 summarizes remission data which was reported in 13 of the 14 RCTs. Two studies (14,16) defined remission as zero binges for past 28 days, whereas remaining studies used a less conservative definition of zero binges at end of treatment. Pooling all studies together, 48.7% of patients receiving medication achieved 100% remission from binge eating by the end of treatment, compared to 28.5% of patients receiving placebo. Overall, there was evidence to suggest a clinically significant difference exists between medication and placebo for remission, with medication being superior (N = 13; n = 1,254; RR = 0.74; 95% CI: 0.66–0.84). A relative risk of 0.74 indicates that medication reduces nonremission rates by ∼26%.

Figure 2.

Remission rate: medication vs. placebo.

When effect sizes for specific classes of medication were considered separately, there was also evidence to suggest a significant treatment effect relative to placebo, although the level of efficacy varied across subgroups. For SSRI medication (N = 7; n = 335; RR = 0.81; 95% CI: 0.70–0.94), antiepileptic medication (N = 3; n = 515; RR = 0.63; 95% CI: 0.51–0.78), and obesity (sibutramine) medication (N = 2; n = 364; RR = 0.80; 95% CI: 0.69–0.94), statistically significant summary statistic (RR) values were found with CIs that did not cross the line of no effect. The overall effect observed for antiepileptic agents with an RR of 0.63 suggested that this broad class of medication reduces nonremission rates by 37%. Especially noteworthy in regard to antiepileptic agents given their different structures and mechanisms, are the RRs calculated for topiramate (0.56 and 0.59) in the two studies (21,23)—which suggest reduction of nonremission rates by 44–41%, respectively—being clearly stronger than the RR value (0.88) for zonisamide (e.g., 12%) (22). Findings for atomoxetine (serotonin-norepinephrine reuptake inhibitor) indicated a promising RR of 0.43, yet findings are limited to one trial (n = 40). A sensitivity analysis performed to exclude topiramate and atomoxetine yielded an overall RR of 0.81 (95% CI: 0.73–0.90; I2 = 0%], which suggests an average reduction of nonremission rates of 19% for 11 of the 14 pharmacotherapy-only RCTs.

Effect of treatment on BMI/weight: meta-analysis findings

Figure 3 summarizes data regarding effects on weight loss (kg) based on 8 of the 14 RCTs that provided usable data (i.e., weight-change data (m/s.d.) or actual pre- and post-treatment weight or BMI data (m/s.d.)). Overall, the pooled meta-analysis showed evidence of a significant additional weight reduction of 3.4 kg (7.5 lb) by medication compared to placebo (N = 8; n = 1,236; WMD = −3.42; 95% CI: −4.25 to −2.58). The effect sizes for the different medication classes varied considerably, from modest effects (WMD = −1.7) for SSRIs to larger effects for antiepileptic (WMD = −4.6) and antiobesity (WMD = −3.6) medications. Average mean weight loss for medication was 3.56 kg vs. 0.08 kg for placebo across the 14 trials. Overall, the average pretreatment weight (kg) for participants was 110 kg, or 240 lb, across studies; thus, observed percent weight loss achieved with medication equaled 3.2% (of original body weight).

Figure 3.

Weight loss (kg): medication vs. placebo. An asterisk (*) denotes completer data. When possible, missing or unavailable s.d. for change scores (C) were imputed using the general formulas: inline image. Pounds (lb) were converted to kilograms (kg) by multiplying values by 0.45.

Pharmacotherapy: combined or additive or relative effects

Eight RCTs that evaluated pharmacotherapy combined with psychotherapy interventions were identified. Table 3 provides a summary of each of the study's details. A total of 683 patients (∼90% women) were included across the eight trials. Four of the studies were conducted in the United States, two in Switzerland, and one each in Brazil and Italy. Five studies were at single site, two had two sites, and one was multisite. Recruitment was through media advertisement in five studies and via clinic referral in three studies. Three of the eight trials were funded by the pharmaceutical industry (drug manufacturer), two did not disclose any funding, and three were funded by the National Institutes of Health or Medical Foundations.

The eight RCTs were diverse in their designs and treatment methods. Six studies provided double-blind data regarding medication vs. placebo delivered in addition to either cognitive behavioral therapy (CBT) (14,27,28), to diet (29,30), or to multimodal diet plus psychosocial treatment (31). Two studies provided data regarding medication vs. CBT, delivered in combination with behavioral weight loss (BWL) (29,32) and one study provided separate data regarding two different medications (fluoxetine and fluvoxamine) vs. CBT and the addition of those two medications to CBT vs. CBT without medication (33). The trials tested various antidepressants including desipramine (32), imipramine (31), fluoxetine (14,29,33), fluvoxamine (33), obesity medication (orlistat) (28,30), and antiepileptic medication (topiramate) (27). Some trials utilized flexible dosing of medication, whereas others used fixed dosing throughout the course of treatment. The trials were of moderate duration, averaging 20.1 weeks of active acute treatment, with a range of 8 (ref. 31) to 36 (ref. 32) weeks. Five of the eight trials reported follow-up data for periods ranging 3–24 months following completion of treatment.

The heterogeneity of the designs of these studies precluded meaningful meta-analysis, although a qualitative review and synthesis of the findings (detailed in Table 3) follows. Overall, the addition of pharmacotherapy to CBT did not enhance binge-eating outcomes but might have enhanced weight loss outcomes. Specifically, the medications added to CBT in double-blind fashion (topiramate (27), fluoxetine (29), orlistat (28), imipramine (31)) did not significantly improve binge-eating outcomes, nor did the open-label addition of two SSRIs (fluoxetine and fluvoxamine) (33). In contrast, the addition of pharmacotherapy to CBT or to BWL resulted in mixed findings in terms of improving weight loss. Antidepressants including desipramine (32), fluoxetine (14,29,33), and fluvoxamine (33) did not significantly enhance weight losses in either CBT or BWL. In contrast, topiramate (27) and orlistat (28) enhanced weight losses in CBT, and imipramine (31) and orlistat (30) enhanced weight losses in diet and behavioral treatments. Noteworthy is that the addition of topiramate to CBT (27), and orlistat to diet (30), resulted in clinically significant (not just statistically) weight losses. Last, in terms of direct comparative data, two studies found that CBT (without medication) was superior to pharmacotherapy-only therapy with fluoxetine (14,33) or fluvoxamine (33).

Discussion

The evidence base for pharmacotherapy for BED is growing, but remains limited both in terms of number of studies and particularly by the lack of follow-up findings regarding maintenance or durability of effects. Our tentative conclusions must be viewed cautiously in light of this relatively small literature and limited evidence base. Only 14 RCTs (placebo controlled) were identified that tested pharmacotherapy-only therapy as the primary intervention and only 8 RCTs were identified that tested pharmacotherapy relative to, or in combination with, psychotherapy interventions. There was no evidence to suggest a significant difference exists between medication and placebo for attrition, suggesting acceptability. Evidence exists to suggest that pharmacological treatments have a clinically significant advantage over placebo for achieving remission from binge eating and for producing weight loss, although the weight losses are not substantial (average percentage of weight loss was ∼3% vs. 0%). There are no data to allow evaluation of longer-term effects or durability of pharmacotherapy-only therapy for BED. Combining medications with psychotherapy interventions failed to significantly enhance binge-eating outcomes, although promising findings have been reported for specific medications (orlistat and topiramate) to enhance weight losses, albeit modestly, achieved with CBT and BWL treatments.

The literature on pharmacotherapy for BED is relatively small and should be cautiously interpreted within the context of potential methodological limitations. The RCTs for BED were of short duration (range 6–24 weeks) and none, to date, have reported follow-up data after medication discontinuation. This represents perhaps the major methodological limitation in the pharmacotherapy treatment literature for BED. The only study of pharmacotherapy-only therapy category with follow-up data (albeit for an obesity agent since withdrawn from the market due to adverse events) reported high relapse rates very soon after medication discontinuation (12). The literature varies in terms of following the reporting standards outlined by the CONSORT Group (34) for reporting RCT methodology (e.g., blinding methods) and detailing analyses although recent larger studies have generally improved in this regard. Pharmacotherapy RCTs, geared primarily toward addressing the important question, “Is there a drug effect?”, tend to rely on statistical methods testing time to reduction or change in symptoms and sometimes do not report endpoint data. Most studies have not reported data regarding the effects of pharmacotherapy on eating disorder psychopathology (other than the behavioral feature of binge eating) characteristic of BED, such as unhealthy and chaotic eating patterns, body image dissatisfaction, and cognitive features such as the overvaluation of shape and weight (35).

The RCTs recruited almost exclusively from the community through media advertisements and enrolled primarily women (90% of participants were women), a departure from prevalence data suggesting a more even gender ratio (2). This raises the possibility of a clinic bias to the extent that media respondents to specialist research clinics might differ (in severity or complexity) from patients who present to traditional clinical services. Prevailing clinical lore is that recruited participants for such research studies are less severely disturbed and have fewer comorbidities than “real” clinic patients. Although surprisingly little evidence exists to address this potential criticism (36), inspection of exclusionary criteria in pharmacotherapy RCTs suggests that such concerns of a clinic bias may exist. Consider, for example, the two largest RCTs. McElroy and colleagues (23) excluded participants with current clinically significant depression, with current or recent substance abuse or dependence, and with personality disorders if there were concerns regarding assessment or compliance, and did not report a descriptive summary of psychiatric comorbidity in their subject description. Similarly, Wilfley and colleagues (25) excluded patients with drug or alcohol abuse during the past year and current major depressive disorder and did not report a descriptive summary of psychiatric comorbidity in their subject description. These exclusionary practices potentially limit generalizability, particularly because both epidemiological (2) and clinical (37,38) studies of psychiatric comorbidity in BED have reported that those exclusions are frequently co-occurring problems. Lastly, 12 of the 14 studies of pharmacotherapy-only category reported they were funded by the drug manufacturer. This state of affairs perhaps reflects, in part, difficulties faced by researchers in obtaining funding to perform treatment studies from the National Institutes of Health given its current severe budgetary shortfalls. Industry sponsorship may be relevant for clinicians and consumers of the literature (39,40) in light of recent reviews concluding that RCTs with pharmaceutical industry sponsorship and conflicts of interest were substantially more likely to report positive results in psychiatric (41) and biomedical (42,43) studies. Chan and colleagues (40) previously suggested that readers of original reports of RCTs funded by industry, as well as of reviews such as ours, should be cautious in interpreting conclusions given the potential for overestimation of efficacy.

With these potential methodological issues in mind, we cautiously offer implications for clinical practice and research. Clinically, our meta-analytic findings indicate patients with BED can be conservatively advised that certain pharmacotherapies may enhance the likelihood of remission from binge eating over the short term but that the longer-term effects of pharmacotherapy-only therapy are unknown. Patients should be advised that although some weight loss may occur, the amount is unlikely to be substantial with available pharmacological agents. Our meta-analytic findings generally echo the general conclusions regarding the potential utility of pharmacotherapy offered previously in the National Institutes of Clinical Excellence (NICE) guidelines (10), but suggest important changes in specific recommendations. That is, our findings highlight the potential utility of antiobesity (sibutramine) and antiepileptic (topiramate) medications, but suggest more limited utility of SSRIs, given their smaller effects on binge eating and absence of effect on weight. Alternative (nonpharmacological) treatments such as specific forms of psychotherapy (e.g., CBT and BWL) should be offered (9,10). Similarly, recent meta-analytic (44) and critical qualitative (45) reviews provided cautious support for the use of certain self-help approaches. Self-help CBT provides an alternative first-step option to intensive psychosocial interventions that—like pharmacotherapy—is readily available. Our qualitative review highlighted that combining specific medications (orlistat and topiramate) with cognitive behavioral or BWL treatments, although unlikely to enhance binge-eating outcomes, may enhance weight losses, albeit modestly.

Implications for research include the need for additional large studies and longer studies with more comprehensive assessment protocols (46). In particular, there is a pressing need for RCTs of medications to include follow-up periods and to provide longer-term outcome data to address questions regarding the durability or maintenance of clinical effects. Such studies are needed to provide guidance regarding whether or when to discontinue pharmacotherapy. There is also a pressing need for secondary analyses of pharmacotherapy RCTs to explore for predictors, moderators, and mediators of outcome (9). Findings regarding predictors and moderators would have implications to guide rationale prescription of treatment, whereas findings regarding mediators would provide clues regarding mechanisms of action and the pathophysiological nature of BED.

Acknowledgment

C.M.G. is supported by grants from the National Institutes of Health (R01 DK49587, R01 DK073542, and K24 DK070052) and Donaghue Medical Research Foundation. No additional funding was received for completion of this work or review.

Disclosure

The authors declared no conflict of interest.

Ancillary