Overview

The evidence included 52 randomized controlled trials (RCTs), presented in 67 articles, examining treatment outcomes. Of these, 48 trials concerned treatment for patients with BED and 4 concerned treatment for children with LOC eating. We found no studies of treatment for LOC eating among bariatric surgery patients. We assembled evidence concerning course of illness from 15 studies (23 articles).

Studies of BED therapies focus on pharmacological interventions, psychological and behavioral interventions, or on combinations of two or more approaches. We found no studies meeting inclusion criteria for any complementary and alternative medicine interventions.

We sought to include evidence of differences in treatment outcomes and course of illness for subgroups of individuals with BED and LOC eating, based on the demographic or patient characteristics noted above. We found evidence (albeit limited) to address these comparisons only in relation to treatment for LOC eating in children (Key Question [KQ] 13). Therefore, the five additional KQs (KQs 3, 5, 8, 10, 15) meant to address these comparisons in other treatment populations or in relation to course of illness for all populations will not be discussed further.

We limit our discussion to summarizing the strength of evidence for benefits of interventions, comparisons, and outcomes for which we had studies of at least low or medium risk of bias. We included studies with high risk of bias in sensitivity analyses of meta-analysis findings, as evidence of harms, and as sources of information for course of illness (because of the otherwise very limited body of available evidence).

We developed strength of evidence grades from ratings on five required domains: study limitations (based on individual study risk of bias), directness of the evidence or the comparisons, consistency, precision of estimates, and reporting bias. In grading cognitive behavioral therapy (CBT) versus waitlist control, we also applied the optional domain magnitude of effect.^129,130 We did not evaluate other optional strength of evidence domains because they were not relevant to our body of evidence (i.e., confounding and dose-response relationships). Strength of evidence can have one of four grades—high, moderate, low, or insufficient. Insufficient evidence arises when we had no studies addressing the particular topic; when we had only a single small study; when available studies were sufficiently inconsistent, indirect, or imprecise as to preclude drawing any conclusions; when differences in treatments appeared to show no difference among studies that may be underpowered; or when clinical thresholds for minimal differences have not been established.

KQ 1. Effectiveness of Treatments or Combinations of Treatments for Binge-Eating Disorder

For this KQ, we sought evidence for the effectiveness of pharmacological treatments, psychological and behavioral treatments, and combinations of pharmacological and psychological and behavioral treatments on a range of clinical outcomes, including frequency of binge eating and abstinence from binge eating, measures of eating-related and general psychological problems, and weight and other measures of physical health. We found data on many different general and eating-related psychological outcomes. A few—namely binge-eating-related obsessions and compulsions; dietary and cognitive restraint; eating, shape, and weight concerns; depression; and symptoms of general psychological distress—were fairly consistently reported across studies.

For outcomes of efficacy of pharmacological treatments (treatment compared with placebo), our findings are limited to outcomes at the end of treatment, as no studies followed patients beyond treatment unless to oversee medication taper for a brief period of time. Similarly, efficacy of psychological or behavioral treatments (treatment compared with waitlist or inactive controls) followed patients only to the end of treatment. By contrast, patients enrolled in comparative effectiveness trials comparing two or more psychological or behavioral treatments or intervention formats tended to undergo assessments beyond the end of treatment, most commonly less than 1 year but in some instances 2 years or more.

Pharmacological Interventions

Table 63 summarizes the pharmacological interventions on which we had low, moderate, or high strength of evidence for clinical outcomes. We found evidence for the effectiveness of second-generation antidepressants, as a class, based on meta-analyses. We also found evidence about effectiveness for one anticonvulsant medication (topiramate) and for one medication originally formulated for attention deficit hyperactivity disorder (ADHD) (lisdexamfetamine) based on meta-analysis and qualitative synthesis. All trials measured outcomes at the end of treatment. Meta-analysis allowed us to estimate magnitude of effect, whereas qualitative synthesis allowed for describing only direction of effect. To enhance comparisons of the magnitude of outcomes across treatments, we provide some limited descriptive data on ranges of outcomes from qualitatively synthesized bodies of evidence.

Table 63

Strength of evidence for pharmacological interventions to improve outcomes in binge-eating disorder.

Evidence concerning the efficacy of antidepressants in treating patients with BED differed by outcome. In relation to binge-eating outcomes, antidepressants were estimated to reduce the weekly frequency of binge-eating episodes by approximately two-thirds of a binge episode per week (high strength of evidence) and approximately one binge-eating day (moderate strength of evidence). Antidepressants were 1.67 times more likely than placebo to help patients achieve abstinence from binge eating (high strength of evidence). Even though patients improved, many did not achieve abstinence with antidepressants (only 41 percent of those receiving antidepressants compared with 23 percent of those receiving placebo).

We examined whether antidepressants were effective in treating psychological aspects and correlates of BED. The volume of evidence for these benefits was less than for binge-eating behavior, overall, the strength of evidence for benefit was low to moderate. Antidepressants helped reduce binge-eating-related obsessive thoughts and compulsions. That is, they provided some benefit in reducing the time that patients spend thinking about food, the degree to which they feel compelled to binge eat, the effort they exert to resist doing so, and the degree of distress associated with these mental processes. Before treatment, patients reported that their severity of obsessions and compulsions was approximately 20 on a 40-point scale. Collectively, obsessions and compulsions decreased approximately 4 points more with antidepressant treatment than placebo. We also found evidence of modest improvements in symptoms of depression as measured on the 52-point Hamilton Depression Rating Scale (HAM-D). Before treatment, patients reported relatively mild symptom levels (mean scores lower than 6 across studies). After treatment, those who received antidepressants experienced approximately a 2-point greater reduction in their HAM-D score than those who received placebo.

The range of responses in weight to antidepressant treatment was wide. Overall, overweight and obese patients treated with antidepressants did not lose significantly more weight during treatment than those who did not receive an antidepressant (low strength of evidence). Given the overall limited impact on weight and the short duration of treatment (6 to 12 weeks), finding no difference in the change in BMI at the end of treatment between those who received antidepressants and those who received placebo is not surprising.

Evidence was insufficient to demonstrate the effectiveness or comparative effectiveness of specific second-generation antidepressants in the treatment of BED. The main reason was that each medication (fluoxetine and sertraline) was studied in a single, small sample size trial or, at most, in two trials (of fluoxetine) that differed on key parameters such as doses or treatment duration.

The anticonvulsant topiramate reduced the frequency of binge eating by approximately 1 binge day per week more than placebo, and it helped approximately 30 percent more patients (58 percent vs. 28 percent) achieve abstinence from binge eating (moderate strength of evidence). By comparison, in a small study, 50 percent of patients treated with the anticonvulsant lamotrigine achieved abstinence, but this effect was not statistically significant because of an unusually high abstinence rate of 71 percent in patients treated with placebo (insufficient strength of evidence).²¹⁸

In addition, topiramate helped reduce binge-eating-related obsessive thoughts and compulsions by approximately 30 percent and more general psychological distress symptoms by approximately 23 percent more than placebo (moderate strength of evidence). Among overweight and obese patients, those treated with topiramate lost, on average, approximately 10 pounds more (equivalent to ~4 percent more total body weight) than those who received placebo (moderate strength of evidence). Topiramate had additional benefits including reductions in patients’ susceptibility to hunger as a trigger for binge eating and improvements in their general tendency to act less impulsively. Patients treated with topiramate also tended to experience increased sense of cognitive control over their binge eating and decreased disruptions in their social and family life compared with patients who received placebo. However, the strength of evidence for these benefits was low.

We combined evidence from three trials of lisdexamfetamine, a medication that was originally formulated to treat ADHD. Lisdexamfetamine is the only medication that has received U.S. Food and Drug Administration (FDA) approval for treating BED.¹⁴⁸ In one published Phase 2 trial, three separate dosage levels of lisdexamfetamine were compared with placebo (30, 50, and 70 mg/day). In two Phase 3 trials, only the two higher dosage levels were evaluated in a combined arm compared with placebo. We limited our analysis to the Phase 3 medication dosage levels. In relation to binge-eating outcomes, lisdexamfetamine was 2.61 times more likely than placebo to help patients achieve abstinence from binge eating (across studies, 40 percent in the treatment arm and 15 percent in the placebo arm) (high strength of evidence). Moreover, lisdexamfetamine patients experienced a greater reduction in binge-eating days per week (the point estimates of the differences in the two Phase 3 trials were 1.3 and 1.7 fewer days) (high strength of evidence). Lisdexamfetamine was also associated with superior eating-related psychopathology outcomes, as measured through the Yale-Brown Obsessive Compulsive Scale Modified for Binge Eating (YBOCS-BE) (high strength of evidence). Weight reduction was greater across the three studies (the point estimates of the differences in the two Phase 3 trials were 6.3 percent and 5.6 percent; virtually all the weight reduction was limited to the treatment arm) (high strength of evidence). However, data on depression and other psychological outcomes were too limited to be evaluated (insufficient strength of evidence).

Evidence was insufficient for benefits of other medications, including dietary supplements. Each medication that we identified in our literature searches was studied in only one small trial.

Psychological and Behavioral Interventions

Table 64 summarizes the psychological and behavioral interventions for which we had low, moderate, or high strength of evidence for treatment benefits. These included four forms of CBT: therapist-led CBT, partially therapist-led CBT, structured self-help CBT, and guided self-help CBT. These represent variations of therapist involvement and contact during the intervention in descending order of therapist participation. The first three forms were examined in group interventions and the fourth as individual therapy. We also evaluated evidence on the comparative effectiveness of different forms of CBT and the comparative effectiveness of CBT versus behavioral weight loss. We found evidence for all outcomes at the end of treatment and for some outcomes over periods as long as 6 years after treatment ended.

Table 64

Strength of evidence for psychological and behavioral interventions to improve outcomes in binge-eating disorder.

We found strong evidence of the effectiveness of CBT in reducing binge-eating-related outcomes, measured as binge frequency and achieved abstinence, compared with waitlist. These benefits were apparent for four forms of CBT (therapist-led, high strength of evidence; partially therapist-led, structured self-help CBT, and guided self-help CBT, low strength of evidence). Evidence of the benefits of therapist-led CBT was particularly compelling; meta-analyses estimated a 4.95 increased likelihood of abstinence compared with waitlist (across studies, 59 percent in the treatment arm and 11 percent in the placebo arm) and a reduction of 2.3 binge episodes per week. For reducing general and eating-related psychological symptoms, only therapist-led CBT and guided self-help CBT were superior to waitlist. Therapist-led CBT reduced patients’ susceptibility to hunger and eating concerns and improved their sense of control over eating (high strength of evidence), and guided self-help CBT helped patients reduce global eating-related psychopathology (low strength of evidence). However, across the various forms of CBT, treatment was generally no better than waitlist for reducing weight or symptoms of depression (low strength of evidence for no difference). We found very limited data comparing pure self-help CBT (which does not involve treatment facilitators) with waitlist; thus, we cannot comment on outcomes of this intervention. Collectively, this body of evidence suggests that various forms of CBT help patients with BED improve in several key behavioral and eating-specific psychological domains.

We examined evidence of the comparative effectiveness of three different forms of CBT with each other: therapist-led CBT, partially therapist-led CBT, and structured self-help CBT. These comparisons are of interest as they provide insight about the relative importance of therapist involvement in the effectiveness of CBT. Across comparisons, we found virtually no differences in binge-eating outcomes (low strength of evidence for no difference). Likewise, non-BED-specific outcomes did not differ across comparisons: neither BMI outcomes nor depression outcomes differed across comparisons of variations in therapist involvement in CBT interventions (low strength of evidence for no difference). Despite nonsignificant differences between CBT variations, the CBT variants that were studied were generally effective both at helping patients achieve binge abstinence and reduce binge frequency, most notably at end of treatment but throughout both short (6 month) and long-term (12 month) followup. Thus, although CBT variations generally did not show a difference in their ability to improve binge-eating-related outcomes, the treatments did produce a significant effect over time such that most patients (regardless of treatment arm) improved in important outcome domains.

We compared therapist-led CBT with therapist-led behavioral weight loss (BWL) treatment on outcomes assessed at the end of treatment and, in limited studies, for up to 6 years after treatment ended. CBT was superior to BWL for decreasing binge frequency at end of treatment and up to 12-month followup (low strength of evidence). Neither trial reported a significant difference between groups in abstinence, eating-related psychopathology, or depression at end of treatment or at 12-month or 6-year followup. Notably, the benefit of BWL over CBT was clear for reducing BMI at end of treatment (moderate strength of evidence); however, those receiving BWL tended to regain the weight they had lost during treatment.

Data were very limited about the effectiveness of various other psychological and behavioral therapies for BED compared with waitlist control. These treatments include several forms of CBT, such as pure self-help, and other therapies such as interpersonal psychotherapy (IPT), dialectical behavior therapy (DBT), and dietary approaches other than BWL therapy. Comparative effectiveness trials of these treatment options against more commonly studied treatments or each other were even more limited. Finally, we found only very limited evidence about the effectiveness of cognitive and behavioral treatments that are provided as adjunct therapy to existing hospital-based inpatient treatment for BED.

The primary limitation was the availability of only single trials for specific treatments. Secondarily, similar trials reported disparate outcome measures. For example, one trial reported binge eating and the other trial of a similar type reported only binge abstinence as an outcome. Thus, we are unable to comment on outcomes of these other treatment modalities.

A particular example of this limitation is in relation to the evidence of the effectiveness of interpersonal psychotherapy (IPT), which was limited to single trials because of diverse delivery formats. The results of these trials are notable given the impressive percentage of participants who achieved abstinence, both at end of treatment and long-term followup.^80,176 At end of treatment, 64 percent of participants receiving IPT were binge abstinent;^80,176 this benefit persisted at 2-year (67 percent abstinent)¹⁷⁶ and even 4-year followup (76 percent).¹⁸⁴ Thus, although the results of these trials were notable, we were unable to comment on the strength of evidence because of the differing ways the IPT interventions were given. Evidence for DBT, albeit promising, is even more restricted that than for IPT.

KQ 2. Evidence for Harms Associated With Treatments for Binge-Eating Disorder

We sought evidence of the potential harms or side effects that may occur with various treatment options. We anticipated finding some concerns because those are already well known in association with antidepressants, anticonvulsants, stimulants, and other medications. We also considered any others that authors of these trials might have reported. Table 65 summarizes the interventions for which we had low, moderate, or high strength of evidence for harms outcomes.

Table 65

Strength of evidence for side effects or harms associated with treatment for binge-eating disorder.

Consistent evidence showed that symptoms of sympathetic nervous system arousal were more common among patients who received topiramate than those who received placebo (moderate strength of evidence). For example, patients who received topiramate more frequently reported sweating, dry mouth, rapid heart rate, and similar physical side effects that are associated with anticonvulsant medications than patients who received placebo. Those treated with topiramate also reported a higher number of events, some relating to physical functioning and some to psychological or cognitive functioning. For example, patients who received topiramate reported more headaches and sleep disturbances (low strength of evidence) as well as a set of other symptoms including rash, high blood pressure, confusion, and taste aversion (moderate strength of evidence for the set of other events) than patients who received placebo. While topiramate is a known teratogen linked to a higher incidence of oral clefts; none of the included studies reported this harm. Patients treated with fluvoxamine reported symptoms of GI upset and sleep disturbances more frequently than patients who received placebo.

Patients treated with lisdexamfetamine more frequently reported GI upset, sympathetic nervous system arousal (including feeling jittery, increased heart rate, agitation, and irritability) and decreased appetite (moderate strength of evidence). Based on meta-analysis results, rates of insomnia and headache were significantly higher among patients being treated with lisdexamfetamine (high strength of evidence).

Evidence was insufficient for many of the specific types of events. The main reasons were that investigators were inconsistent in how they reported specific events across studies and that they often did not report events in an itemized fashion with clear attribution to treatment or placebo. These shortcomings in the body of evidence also limited our determination of whether patients receiving medication or combination treatments were more likely than those receiving placebo to discontinue treatment because of adverse events.

Thus, we could address harms only in a descriptive manner. We provided counts across categories of events with little assurance that those counts truly represented all adverse events that occurred in the included studies. Similarly, we could only summarize and describe the discontinuations attributed to serious harms and treatment differences in serious harms because so few serious adverse events were reported (N=10).

KQ 4. Course of Illness Among Individuals With Binge-Eating Disorder

We sought evidence on outcomes among individuals with BED 1 year or longer following their diagnosis (KQ 4). We identified 10 studies (trials or observational studies). None of the studies included cohorts of individuals identified in the community; rather, they were limited to individuals who had earlier participated in BED treatment studies. Two studies used a case series design, comparing outcomes in a treatment population with those in matched controls identified through a registry. Because the number of available studies was limited, we included three articles (reporting on two studies) that we had rated high risk of bias.

Binge-eating outcomes were commonly reported in these studies. However, studies differed in the characteristics that investigators hypothesized might be related to better outcomes; these variables included more rapid response to treatment, improvement in eating-related psychopathology, and improvement in non-eating-related psychopathology. One study found increased odds of miscarriage among women with BED. A review article of three studies and an additional study found no evidence of increased risk of suicide among BED patients 5 years after treatment (strength of evidence was moderate for no effect.) Strength of evidence was insufficient for all other comparisons and outcomes.

KQs 6, 7, 11, and 12. Effectiveness of Treatments and Harms Associated With Treatments for Loss-of-Control Eating

We sought evidence of the effectiveness of treatments or combinations of treatment for LOC eating among bariatric surgery patients and children. We found no evidence addressing treatment for LOC eating among bariatric surgery patients (KQs 6, and 7; insufficient strength of evidence).

Evidence about treating LOC eating among children was limited to four studies (KQ 11). Three studies focused on adolescents and another on children 8 to 12 years of age. All included children were overweight or obese. Studies differed in the criteria they used for defining LOC eating. One study focused on children eating in the absence of hunger; the other three examined children who had been binge eating (past 3 months) or experiencing LOC eating (once in the past month). Studies also differed in treatment comparisons. With the exception of no difference in weight outcomes, comparing IPT and non-BED health education “attention only” control (low strength of evidence for no difference), evidence is insufficient for all outcomes. No harms from treatment were reported in these studies (KQ 12).

KQ 9. Course of Illness Among Bariatric Surgery Patients With Loss-of-Control Eating

We sought evidence on outcomes among bariatric surgery patients with LOC eating, 1 year or longer following diagnosis. The two identified studies differed in the criteria used for defining LOC eating at baseline (i.e., before surgery). Strength of evidence is insufficient across all outcomes because of a lack of clear and consistent findings in more than one study.

KQ 14. Course of Illness Among Children With Loss-of-Control Eating

We sought evidence on outcomes among children with LOC eating and identified three longitudinal cohort studies. In two of the studies, eligibility was based on responses to survey questions concerning any binge or LOC-eating behavior in the past year. The third study used a more formal clinical assessment and the evaluation period was over the past 3 months. The body of evidence indicated that early adolescent binge or LOC eating predicted similar behavior in later adolescence (low strength of evidence). Evidence of additional outcomes was limited or inconsistent across studies.

Population

Findings about all BED treatment interventions are likely to be applicable to all adults above the age of 18 with the disorder. Patient populations were generally overweight or obese women. However, because of insufficient evidence, we cannot comment on treatment applicability as it pertains to specific subgroups of adults (even among women) or whether they extend to BED patients diagnosed based on DSM-5 criteria (which are less stringent than those for DSM-IV). Also unclear is whether our findings apply to adolescents or various minority group members with BED.

The evidence base about treatment for LOC eating in children was small and for bariatric surgery patients was nonexistent. The criteria used to define the condition varied across the studies of children and included different combinations of characteristics such as eating in the absence of hunger, objective and subjective binge episodes, and objective overeating episodes. Studies also differed in the required number of times the behavior had occurred and over what period of time Thus, although the evidence may be generally applicable, generalizing to child (or adolescent) patient populations is probably inappropriate (and is impossible for bariatric surgery patients). As noted, we had no useable information on these patient populations differentiated by sociodemographic or health characteristics. Finally, the appropriate diagnostic criteria that clinicians might reliably use to identify LOC eating have not been established.

With respect to the course of illness of BED, we found no evidence among an untreated population; we can, however, consider several hypotheses based on the ongoing concerns within the patient samples. Untreated BED could likely become a chronic condition, which could in turn result in deleterious mental and physical health effects. Left untreated, the condition may lead to or worsen other mental health concerns, such as depression or anxiety, or physical health conditions, such as diabetes or irritable bowel syndrome. Among children with LOC eating, evidence pointed to these children being at increased risk for excessive weight gain in later adolescence.

Interventions and Comparators

We present evidence on numerous possible treatments for BED, as long as those treatments were evaluated in trials that met our inclusion criteria and were not considered high risk of bias. We present evidence on medications, psychological and behavioral treatments, and combinations of treatments. Only one of the medications in our review, lisdexamfetamine, has been approved by the FDA for treating BED. Therefore, it is the only pharmacological treatment in which potential adverse events in a patient population with BED have received FDA review.

We found many single studies of treatments. Although we included these investigations in our review, we could not comment on the efficacy of these many interventions for BED patients. These included medications from many classes that are approved for treating depression and substance dependence. Similarly, we cannot comment on the applicability of any pharmacological interventions for LOC eating in children or bariatric surgery patients. Several of the behavioral trials treating LOC eating in children tested the effectiveness of IPT but, similar to the BED literature, the evidence for effectiveness was also limited for this patient group and virtually all results were insufficient.

In addition, we had planned to include complementary and alternative medicine approaches, but we could not find any studies that met our criteria. Thus, we have no evidence regarding the effectiveness of these particular treatments for any of these conditions (BED or LOC eating).

Outcomes

We did not limit the outcomes of interest but rather took a broad view of the kinds of benefits that might occur with treatments. Our primary focus was on reductions in commonly noted BED symptomatology, including binge frequency, eating-related obsessions and compulsions, restraint, shape and weight concerns, weight, and depression. As noted above, however, investigators used a considerable array of different measures or instruments to assess outcomes, constraining our ability to generalize findings across all of these outcome categories of interest. Also, we sought but did not find sufficient information to draw any conclusions about treatment effectiveness for more global measures such as quality of life or lost productivity. Finally, we found no evidence about treatment effectiveness as it relates to final health outcomes such as, for example, diabetes, gastric reflux, and irritable bowel syndrome.

Given the scarcity of information about LOC eating, we cannot conclude anything about applicability of these trials with respect to proposed or potential outcomes of treatment among bariatric surgery patients and little about treatment outcomes among children.

Timeframes

Studies varied in their length of followup periods. All trials of medications measured outcomes at the end of treatment (but many of these trials were relatively short); only two trials^195,222 reported longer-term followup beyond the end of treatment. Similar studies examining the efficacy of psychological and behavioral interventions measured outcomes at the end of treatment. Only comparative effectiveness studies comparing different psychological or behavioral interventions or different intervention formats were more likely to include both short- and long-term followup; one trial extended to 6 years after the end of treatment. Generally, the applicability of these studies for understanding the long-term impacts of treatment (benefits or harms) is relatively limited because the long-term efficacy of the individual treatments has not been established; the applicability of these studies (especially the pharmacological trials) for short-term benefits may be somewhat stronger.

Settings

The evidence base for both BED and LOC eating (in children) was largely outpatient care, which would be the standard of care in the United States. We found limited evidence about inpatient therapies; the patient populations in these studies, conducted in Italy, would unlikely be eligible to receive inpatient care in the United States. Of all the trials we included for either BED or LOC, most were conducted in clinical settings in North America (United States mainly, or Canada); evidence was also obtained from studies conducted in Scandinavia or elsewhere in Europe. Generally (apart from considerations relating to health systems and insurance for the few investigations done outside North America), results are applicable to US patient populations.

However, most studies were conducted in supervised settings generally associated with academic research and medical centers. In such settings, medication treatment was likely managed by a psychiatrist and psychological and behavioral treatments likely delivered by highly trained personnel, many specifically trained in treating individuals with binge-eating disorder. Whether our findings apply to the “real-world” settings in which individuals seek and receive treatment in their local community through contact with their primary care physician or other community-based providers who do not have specific expertise in BED treatment remains unclear.

Gaps in Subgroups Studied

We found no studies that addressed differences in treatment outcomes among important subgroups defined by age, sex, race, and other relevant patient characteristics. Observational and cross-sectional studies have shown that binge eating may be more common among certain racial minorities, for example, yet treatment studies have failed to address whether outcomes differ between groups defined by race. These gaps limit applicability to these important groups.

Secondary analyses of data from treatment studies have shed some light on factors that may be important for future consideration, including age and sex. Nevertheless, the specific analyses that were conducted did not address whether treatment effectiveness was the same, or different, in these subgroups. For instance, as in our earlier systematic review of eating disorders, we identified very little information about the impact of treatments on either men or boys.

Moreover, despite the high comorbidity between BED and depression and between BED and obesity, no studies specifically compared outcomes in groups of patients defined either by baseline level of depression or by baseline weight status. Based on our review, second-generation antidepressants have a small but significant impact on symptoms of depression in BED patients with low levels of depressive symptoms. Whether the small benefit of second-generation antidepressants is meaningful, or perhaps amplified, in BED patients with higher levels of depression warrants further study.

In light of growing awareness of LOC eating in children and concerns that LOC eating has negative health effects and predisposes to BED later in life, treatment studies focusing on children are needed. Similarly, lessons learned in BED treatment groups are likely to be hypothesis generating for bariatric surgery patients, but these individuals require treatment studies of their additional eating-related limitations and requirements following surgery.

Gaps in Outcomes Measured (Benefits or Harms)

The evidence base was deficient for outcomes related to social and occupational functioning (or quality of life more generally). It was similarly poor in relation to final health outcomes such as glucose intolerance or dysregulation that may predispose patients to diabetes and other chronic conditions. Also lacking is evidence of harms associated with psychological and behavioral treatments. A third critical gap exists in longer-term benefits and harms; this gap is especially evident for pharmacological treatments and combination treatments.

Gaps in Interventions

We found strong evidence that CBT is beneficial for patients with BED; however, that conclusion was related largely to therapist-led group CBT and to a lesser extent CBT conducted in other group formats. Much of the comparative effectiveness body of evidence for CBT constitutes a collection of disparate studies testing variations in format; furthermore, the rationale for comparing different formats is not consistently grounded in an a priori mechanism of action.

The number of therapists with expertise in CBT for BED is limited. This limitation poses a challenge for implementation of our findings. One useful step might be to compare directly (in adequately powered head-to-head trials) whether therapist-led CBT is equivalent to other CBT formats. If modified versions that require less therapist involvement can be shown to be equally effective as therapist-led CBT through equivalence or noninferiority trials, such information could help make CBT more scalable than it has been to this point. Findings might then guide the next generation of studies that are needed to move the field closer to an individualized approach to treatment. Those future studies should consider other psychological and behavioral interventions that have shown promise (IPT; DBT). In addition, they should be adequately powered to test for differences (or similarities, if appropriate) in outcomes across key subgroups (i.e., patient groups defined by age, sex, race, and weight status) for which a dearth of information still exists.

Second-generation antidepressants were beneficial in reducing symptoms of depression, and topiramate was beneficial for reducing symptoms of impulsivity. A head-to-head comparison of the effectiveness of these two treatment options on mood and impulse regulation outcomes might help clinicians and patients make first-line pharmacotherapy treatment choices based on individual patients’ needs and preferences. Further examination of lamotrigine may also be warranted despite the negative findings for abstinence in one small trial; in that trial, the lamotrigine response rate (50 percent) was similar to that of topiramate (58 percent) but the placebo response rate was extremely high (71 percent). Further examination of lamotrigine may also be justified because, owing to its unique biochemical structure and function relative to topiramate, it may also be associated with fewer sympathetic nervous system and other side effects.^224,225

Head-to-head comparisons involving pharmacological treatment, psychological treatment, and combination treatments are also needed to determine whether, as one study suggests,¹⁴⁰ gains persist longer following psychological (CBT) or combination (CBT+fluoxetine) treatment than for pharmacological (fluoxetine) treatment alone. This information would help patients and providers optimize their plans to address both short- and long-term goals of treatment.

In addition, the CBT comparative effectiveness evidence has focused on whether less specialized care can be as effective as more intensive services (e.g., those with substantial involvement of therapists). More studies of these comparisons are needed. In addition, studies of stepped-care models can elucidate whether and when combination treatments or (a shift to) higher levels of care (e.g., intensive outpatient, partial hospitalization, residential treatment, or inpatient) are warranted for patients who are not responding adequately to conventional outpatient treatment.

Despite current interest in complementary and alternative medicine, neutraceuticals, and mindfulness-based interventions for regulating appetite, eating behavior, and weight, the literature is deficient regarding these types of interventions for BED. We searched clinical trial registries to determine whether additional evidence was available from newly completed, but as-yet unpublished, studies. We also checked for evidence of studies that were selectively withheld from publication because of unfavorable outcomes (possible publication bias). Based on these activities, we did not determine that reporting bias was a concern in this evidence base.

We included evidence of the effectiveness of outcomes with lisdexamfetamine (Vyvanse®), a central nervous system stimulant that was originally formulated to treat patients with ADHD. The evidence from one Phase 2 trial and two Phase 3 trials was conducted by the same research team and combined in our analysis. Only information on the conduct and results of the Phase 2 trial was available through a peer-reviewed publication; we obtained data on the Phase 3 trials through the gray literature. Peer-reviewed publication of the Phase 3 trials would add to our confidence about the conduct and outcomes of these studies. In addition, the mechanism of action in lisdexamfetamine for treating BED is unknown. We do not know, therefore, whether similar results would emerge for other stimulants or other medications that are currently being used to treat patients with ADHD (aside from similar results from one small trial of the ADHD drug atomoxetine).

Deficiencies in Methods

Our 2006 review, “Management of Eating Disorders,”^103,120,123 identified several methodological issues within the BED treatment literature and recommended changes for future studies. Some but not all of the deficiencies we noted in 2006 persist; problems still include inadequate reporting of randomization and allocation concealment and insufficient attention to treatment group differences in the use of co-interventions. These and other factors led us to reconsider our risk of bias ratings for some studies; newer ratings, in turn, reduced the strength of the evidence for the current review.

In our 2006 review, we also highlighted several critical needs for advancing the field. These suggestions included conducting replication studies, doing longer-term followup studies, and streamlining and standardizing the outcome measures to eliminate reporting of false discoveries. Unfortunately, with few exceptions,^70,71,99,143 replication studies do not exist; thus, the evidence base remains insufficient to address whether gains achieved during short-term treatment persist after treatment ends. This gap is especially critical for pharmacological treatments, as patients and their providers seek to understand the need for on-going medical management to maintain treatment gains.

The field would benefit from the development of universally accepted definitions of remission and recovery.²²⁶ To reach this goal requires longer-term followup periods with periodic re-evaluation of a core set of psychological, behavioral, and physiological outcomes. Standard definitions of remission and recovery should consider a continuum approach rather than focus on just a fixed point in time. Toward this goal, we make two recommendations.

First, studies should implement a minimum 1-year followup period. Even longer periods of followup may be warranted to capture the remissions and improvements in illness that can occur over a longer period of time. Similarly, longer trials might help clarify what treatments are “better” for patients who do not fully recover but live with a chronic illness.

Second, future studies should include a reasonably limited set of eating-specific instruments (such as the Eating Disorder Examination questionnaire, the Three-Factor Eating Questionnaire, or the YBOCS-BE) and general psychological symptom (depression, anxiety, negative body image) self-report instruments. Binge-eating-specific adaptations of existing reliable and valid instruments⁹⁶ may help to move the field closer to an understanding of the core determinants of recovery and relapse, but such adaptations should be used only if they are clearly described so that others can replicate their use. Such descriptions should include basic information on the reliability, validity, and reproducibility of these newer instruments.

Additionally, considering the perspective of the patient in defining remission and recovery is crucial. Using such preferences or values in developing consistent definitions of these types of patient-centered outcomes would be a major advance in this clinical area. Interweaving this information with reliable, validated measures will allow researchers and clinicians to generate a comprehensive set of parameters by which remission and recovery could be measured. Consistent and thorough reporting of these outcomes (e.g., fully descriptive data at each major assessment point) will help improve calibration of these instruments against each other, which is ultimately needed for future efforts to use meta-analysis to evaluate treatment effect size.

Further, several etiological and treatment considerations merit further study to better elucidate the onset, maintenance, and treatment of BED. For example, given the prevalence of underlying metabolic disorders (e.g., diabetes mellitus, polycystic ovarian syndrome) in patients with BED, more fully examining the role of these disorders in the development and maintenance of BED would be useful. For treatment, evaluating treatment interventions originally developed for post-traumatic stress disorder (PTSD), given the incidence of BED in those with trauma histories, may yield helpful clinical insights.

Finally, we recommend that studies continue to measure and report binge frequency as both discrete binge episodes and binge days per week. More data are needed to resolve whether one or the other is the better choice for assessing treatment effects.