In the following sections, the available data on the efficacy of treatments for eating disorders are reviewed. For several reasons, interpreting the meanings and significance of these studies for patients seen in clinical practice is often difficult. Most studies have consisted of 6- to 12-week trials designed to evaluate the short-term efficacy of treatments. Unfortunately, few data exist on the long-term efficacy of treatment for patients with eating disorders, who often have a chronic course and variable long-term prognosis. Many studies also inadequately characterize the phase of illness when patients were first treated (e.g., early or late), which may have an impact on outcomes. Particularly for studies of psychosocial therapies that may consist of multiple elements, the precise interventional elements responsible for treatment effects may be difficult to identify. Furthermore, in comparing the effects of psychosocial treatments among studies, important variations may exist in the nature of the treatments delivered to patients. In addition, most studies have examined the efficacy of treatments only on eating disorder symptoms, with few reporting the efficacy on associated features and comorbid conditions such as the persistent mood, anxiety, and personality disorders that are common in "real world" populations.

A variety of outcome measures are used in trials for patients with eating disorders. Outcome measures used in studies of patients with anorexia nervosa often include the amount of weight gained within specified time intervals or the proportion of patients achieving a specified percentage of expected body weight, as well as whether those with secondary amenorrhea experience a return of menses. Measures of the severity or frequency of eating disorder behaviors have also been reported. In studies of bulimia nervosa, outcome measures include reductions in the frequency or severity of eating disorder behaviors such as binge eating, vomiting, and laxative use and the proportion of patients achieving remission from or a specific reduction in eating disorder behaviors.

With regard to approaches to promoting weight gain, the evidence does not show that giving a patient a warming treatment or growth hormone injections significantly increases weight gain or decreases the length of hospitalization. In a randomized controlled trial by Birmingham et al. (599), 10 female patients with anorexia nervosa received a warming therapy, consisting of a heating vest set at medium heat for 3 hours/day for 21 days, and a control group of 11 patients received the vest when it was not turned on. Among the 18 patients who completed the study, there was no difference in change in BMI; the study authors concluded that warming did not increase the rate of weight gain. In a double-blind study by Hill et al. (600), 15 inpatients with anorexia nervosa, ages 12–18 years, were randomly assigned to receive recombinant human growth hormone (0.05 mg/kg, s.c.) or an equivalent volume of placebo daily for 28 days. At the end of the study, the growth hormone and placebo groups did not differ significantly in admission weight, BMI, or daily caloric intake.

For weight maintenance, Kaye et al. (601) found that weight-restored patients with anorexia nervosa often require 200–400 calories more than sex-, age-, weight-, and height-matched control subjects to maintain weight. The energy wasting of malnourished anorexia nervosa patients results in higher than normal resting energy expenditure (REE; measured in kilocalories per kilogram per day) (602), resulting in resistance to weight gain. In malnourished anorexia nervosa patients, a high REE has been independently linked to anxiety level; abdominal pain and vomiting; physical activity, including exercise, fidgeting, and other non-exercise-related energy expenditure; and cigarette smoking (603, 604). Of note, REE measured as kilocalories per day is lower in patients with anorexia nervosa and returns to normal with refeeding (604). These different measurement and reporting techniques may be the source of some confusion among study results, and reports concerning REE must be read carefully to fully understand exactly what is being measured (602).

Research that addresses the optimal length of hospitalization or the optimal setting for weight restoration is sparse. There is no available evidence to show that brief stays for anorexia nervosa are associated with good long-term outcomes. Several studies have reported that hospitalized patients who are discharged at a weight lower than their target weight subsequently relapse and are rehospitalized at higher rates than those who achieve their target weight before discharge (605). Baran et al. (606) assessed weight, height, eating disorder symptoms, and severity of depressive and anxiety symptoms in 22 women with anorexia nervosa at hospital admission and at follow-up occurring a mean of 29 months after patient discharge. The patients who were discharged while severely underweight reported significantly higher rates of rehospitalization and endorsed more symptoms than those who had achieved normal weight before discharge. Commerford et al. (607) had similar observations in their 5-year follow-up study (by telephone interview) of 31 patients with anorexia nervosa and bulimia nervosa who met specific criteria for discharge. That group reported that patients being discharged while at a low weight was associated with brief lengths of stay and that the closer patients were to a healthy weight at the time of discharge from the hospital, the lower their risk of relapse.

In a study by Watson et al. (103) of 397 patients admitted to an inpatient service over a 7-year period, patients who were admitted involuntarily showed the same short-term rates of weight gain as those who were admitted voluntarily. Moreover, most of those who were involuntarily treated later affirmed the need for and exhibited a better attitude toward the treatment process.

In ambulatory settings, most programs find weight gain goals of 0.5–1 lb/week to be realistic, although gains of up to 2 lb/week have been reported in a partial hospitalization step-down program in which patients previously treated as inpatients are treated 12 hours/day, 7 days/week (113). The opinion of the clinicians running this program is that it would not be as effective for never-hospitalized patients.

Although psychosocial interventions, including psychoeducation, individual therapy, family therapy, and (in some settings) group therapy, are considered to be the mainstay of effective treatment for anorexia nervosa, supporting evidence is sparse. Instead, this perspective is derived primarily from considerable clinical experience (608) and patient reports. In a review of 23 studies reporting surveys of people who have had an eating disorder to determine which treatments patients find helpful, support, understanding, and empathic relationships were rated as critically important, psychological approaches were rated as the most helpful, and medical interventions focused exclusively on weight were viewed as not helpful (609).

The concept of "readiness for change," which is widely used in the treatment of substance use disorders (610), has garnered increasing interest and use with patients with anorexia nervosa to improve their motivation for treatment and potentially improve treatment efficacy. The Anorexia Nervosa Stages of Change Questionnaire, developed with a population of patients age 14 years and older, has been designed to be specific for anorexia nervosa. It is reliable and valid (611, 612) but has not yet been used to study treatment effectiveness at various stages. The Readiness and Motivation Interview, which was developed as an assessment tool for adult patients (613), has been shown to predict clinical outcomes, such as the decision to enroll in treatment, dropping out from intensive residential treatment, posttreatment symptom change, and relapse (613–615); however, it has not yet been evaluated to determine its effectiveness in helping patients move from precontemplation stages to higher stages of readiness for treatment. After a quantitative and qualitative analysis of results from a survey of 278 patients with anorexia nervosa, Jordan et al. (616) attempted to develop a staging measure associated with recovery from anorexia nervosa; they concluded that the most meaningful measure was one that measured progress in readiness to stop restricting, bingeing, and purging behaviors. The development and use of validated tools that assess readiness are important because clinicians have been shown to be poor at estimating patients' readiness for change (617) and consequently are ill-equipped to make treatment recommendations tailored to patients' readiness status.

Most inpatient programs use one of many behaviorally formulated interventions, individual and family psychotherapy, empathic nursing approaches, nutritional counseling, and several group therapies designed to improve the patient's knowledge about and attitude toward eating, exercise, and body image (618, 619). These behavioral programs implement a variety of strategies derived from social learning theory that include reinforcement and contingency management (e.g., empathic praise, exercise-related limits and rewards, bed rest, privileges linked to achieving weight goals and desired behaviors). Behavioral programs have been shown to produce good short-term therapeutic effects (620). One review comparing behavioral psychotherapy programs with medication treatment alone found that behavior therapy resulted in more consistent weight gain among patients with anorexia nervosa as well as shorter hospital stays (620). Studies of consecutively admitted inpatients with anorexia nervosa (621, 622) found that "lenient" behavioral programs that use initial bed rest and the warning of returning the patient to bed if weight gain does not continue are as effective as, and in some situations possibly more effective than, "strict" programs in which meal-by-meal caloric intake or daily weight is tied precisely to a schedule of privileges (e.g., time out of bed, time off the unit, permission to exercise or receive visitors). Some evidence suggests that the use of a supervised graded exercise program, such as nonaerobic yoga, may be of benefit in the inpatient treatment of anorexia nervosa (623–625).

Although there is debate about the value of supplemental feedings and formula feedings during the early weight-gain phase in anorexia nervosa, emerging evidence suggests that this strategy may sometimes be helpful. In a short-term study of 100 adolescent Caucasian female patients, one center reported that over a comparable period of time, patients voluntarily treated during hospitalization with supplemental nocturnal nasogastric refeeding had greater and more rapid weight gain than patients treated with traditional oral refeeding alone (121). Other centers have reported similar experiences (626). High-calorie supplements have also been shown to lead to more rapid weight gain (627). However, further study is needed to assess the short- and long-term effectiveness of this approach (121, 628).

Adolescents with anorexia nervosa may have the best outcomes after structured inpatient or partial hospitalization treatment. For example, one study in Norway (629) found that among 55 patients who had received systematic (usually inpatient) treatment based on close cooperation among parents and the pediatric and child and adolescent psychiatry departments, outcome after 3–14 years was good. No patient had died and 82% of the patients had no eating disorder; however, 41% had other axis I diagnoses (most commonly depression or anxiety disorders).

Among adults with anorexia nervosa who receive inpatient treatment, outcome is not usually as favorable. For example, in another report from Norway of 24 adult patients, 42% of patients had improved by the 1-year follow-up, whereas the outcome was poor in 58% (630).

Attempts have been made to determine factors that predict relapse after hospitalization, but identifying such features with certainty has proved challenging. Strober et al. (19) were unable to clearly identify factors associated with posthospitalization relapse among adolescents. One study (631) found that a young age (<15 years), markedly abnormal eating attitudes at admission, and a low rate of weight gain during hospitalization predicted readmission.

During the acute phase of treatment, the efficacy of specific psychotherapeutic interventions for facilitating weight gain remains uncertain. A randomized controlled trial by McIntosh et al. (7) of 56 acutely ill adult women with anorexia nervosa treated as outpatients showed that 20 weekly sessions of nonspecific clinical management (a manual-based therapy delivered by individuals knowledgeable about the treatment of eating disorders and consisting of advice, support, and education) were as effective as or more effective than 20 weekly sessions of CBT or IPT. Of the 56 women, 70% either did not complete treatment or made small or no gains; only about 10% had a very good outcome and 20% improved considerably by the end of these treatments.

With regard to psychotherapy after weight gain, in the first clear demonstration of the efficacy of CBT, 33 adult patients with anorexia nervosa were randomly assigned after weight gain to 1 year of outpatient CBT or nutritional counseling (136). The group receiving nutritional counseling relapsed significantly earlier and at a higher rate than the group receiving CBT, and modified Morgan Russell criteria for "good outcome" were met by significantly more of the patients receiving CBT (44%) than those receiving nutritional counseling (7%).

In another study of outpatient treatment by Dare et al. (85), 84 adult outpatients with anorexia nervosa were randomized to one of four treatments: 1 year of focal psychoanalytic psychotherapy, 7 months of cognitive-analytic therapy, 1 year of family therapy, or low-contact, routine treatment as usual. At 1-year follow-up, only modest symptomatic improvement was seen in the whole group of patients, and several patients remained significantly undernourished. Although improvements were quite modest for all groups, psychoanalytic psychotherapy and family therapy were superior to the control treatment; cognitive-analytic therapy (which was shorter in duration) tended to show benefits over the control treatment as well.

In practice, individual psychotherapies, family therapies, nutritional counseling, and group therapies are often combined during hospital treatment and in comprehensive follow-up care. As of yet, no systematic data have been published regarding outcomes of using these combined approaches, which experienced clinicians often view as superior to a single-therapy approach.

Evidence suggests that family therapy is frequently useful for reducing symptoms and dealing with family relational problems that may contribute to an eating disorder's persistence, particularly in adolescents. Russell et al. (154) demonstrated family therapy to be superior to individual therapy in adolescents with anorexia nervosa for <3 years. Follow-up studies showed that this superiority was maintained 5 years later (155). In an outpatient study by Eisler et al. (86), 40 adolescent patients were randomly assigned to conjoint family therapy or separated family therapy. Both therapies were found to be equally effective on global measures of outcome, but symptomatic change was more marked in the separated family group, but only if the parents were highly critical of the patient, whereas psychological change was more prominent in those receiving conjoint family therapy.

In a randomized inpatient trial, Geist et al. (130) compared the effects of two family-oriented treatments, family therapy and family group psychoeducation, on 25 girls with newly diagnosed anorexia, restrictive type. At 4 months, significant improvement in weight was noted in both groups compared with baseline—77.7% versus 89.1% for family therapy and 77.2% versus 90.4% for family group psychoeducation—with no significant difference between the two groups. In all patients, no significant changes were noted on any self-report measures of specific or nonspecific eating disorder psychopathology. This study was uncontrolled, so it is also difficult to determine the specific results of the treatments in the context of other treatments received by the patients. However, the study results suggest that family therapy and family psychoeducation may be equally helpful with respect to weight gain in the course of IPT in patients with anorexia nervosa.

In a randomized study of 37 adolescents with anorexia nervosa by Robin et al. (126), behavioral family systems therapy (BFST), in which eating and distorted beliefs are targeted during family therapy, was compared with ego-oriented individual therapy (EOIT). At the end of treatment and at 1-year follow-up, groups receiving either treatment had significant weight gain, resumed menstruation, and showed improvements in eating attitudes, depression, and eating-related family conflict. BFST produced a quicker response initially, but by 1-year follow-up, the outcome in the two groups was similar. This study was uncontrolled, and the loss of participants at follow-up could have biased the results. However, the study suggests that both BFST and EOIT may be helpful.

Systematic studies of the Maudsley model of family therapy that are currently under way are receiving considerable interest (87). In these interventions, families are "put in charge" of their children's eating. Lock et al. (87, 129) found that short-term family therapy was as helpful as longer-term family therapy and that, overall, adolescents with anorexia nervosa did well. The results suggested that those who had high levels of OCD symptoms or a nonintact family tended to need longer courses of more intense therapy. A multisite study of this approach is currently being conducted.

For adults, family therapy is less promising. Russell et al. (154) found that individual therapy tended to be superior to family therapy by the end of active treatment but that this difference disappeared at follow-up. Dare et al. (85) observed that specific treatments, including family therapy, were superior to generic treatments and treatments rendered by inexperienced clinicians. Family therapy was no less effective than the other types of therapy, but, as mentioned above, in this study results were modest for all active treatments.

Some programs attempt to blend features of addiction models, such as the 12 steps, with medical model programs that use cognitive-behavioral approaches (632). However, no systematic data exist regarding the effectiveness of these approaches for patients with anorexia nervosa.

Expert opinion suggests that benefits are likely to accrue from support groups; however, at present no data are available that systematically assess the contribution of support groups led by professionals or advocacy organizations that provide patients and their families with mutual support, advice, and education about eating disorders.

Studies of the effectiveness of antidepressants on weight restoration are limited. In two studies (174, 175), the addition of fluoxetine to the nutritional and psychosocial treatment of hospitalized, malnourished patients with anorexia nervosa did not appear to provide any advantage with respect to either the amount or the speed of weight recovery. Attia et al. (174) conducted a randomized, placebo-controlled, double-blind study of fluoxetine at a target daily dose of 60 mg in 31 women with anorexia nervosa receiving treatment for their eating disorder on a clinical research unit. At 7 weeks, there were no significant differences in body weight or measures of eating behavior or psychological state between patients receiving fluoxetine and those receiving placebo. Similar results were reported in a 6-week open-label trial by Strober et al. (175), in which the response to fluoxetine in adolescents hospitalized for the treatment of anorexia nervosa was investigated. Patients were drawn from consecutive admissions to a specialty treatment service and received fluoxetine as an add-on to their multidisciplinary treatment regimen at 3 weeks to 1 month after intake. Analyses of global clinical severity ratings of eating behaviors and weight phobia failed to show any beneficial or detrimental effect of fluoxetine in the patients when compared with matched historical case-control subjects. An uncontrolled trial by Gwirtsman et al. (633) of six patients with chronic, refractory anorexia nervosa treated with fluoxetine reported positive results, including weight restoration. Overall, however, the little evidence that is available does not support the use of antidepressant medications for weight restoration in severely malnourished patients with anorexia nervosa who are being treated in well-structured hospital-based eating disorder programs.

After patients have gained weight and when the psychological effects of malnutrition are resolving, preliminary evidence suggests that SSRI antidepressants may be helpful with weight maintenance. In a double-blind, placebo-controlled trial by Kaye et al. (177), 35 patients with restricting-type anorexia nervosa were randomly assigned to receive fluoxetine (n = 16; average 40 mg/day) or placebo (n = 19) after they had gained weight in an inpatient hospital program and were discharged; they were then observed as outpatients for 1 year. The dropout rate from the trial was much higher in the placebo (84%) than in the fluoxetine (37%) group. Patients continuing to take fluoxetine for 1 year had a reduced rate of relapse, as determined by a significant increase in weight and a reduction in symptoms. They also showed a reduction in depression, anxiety, and obsessions and compulsions. However, these study results are problematic because some patients' weight had not been restored when the study started and the study design was complex, with many exceptions and multiple raters. In contrast, preliminary analysis of data from a 5-year two-site study funded by the National Institute of Mental Health on relapse prevention for anorexia nervosa involving 93 patients ages 16 years and older did not favor fluoxetine when CBT was administered (138). After weight restoration to at least 90% of ideal body weight, all patients received CBT and were also randomized to either fluoxetine 60 mg/day or placebo. The survival analysis showed no difference between those receiving CBT plus medication and those receiving CBT alone with respect to time to relapse (138).

In an open outpatient study by Bergh et al. (634), underweight adolescent patients with anorexia nervosa treated with psychotherapy plus citalopram did worse (losing several kilograms) than did patients treated with psychotherapy alone (losing about 0.2 kg during the period of observation). Another study by Fassino et al. (176) compared 52 adult female patients with anorexia nervosa, restricting type, who received citalopram (n = 26) or were assigned to a waiting-list control group (n = 26). The randomization method in this study was not clearly defined. After 13 patients dropped out, 19 and 20 patients remained in the citalopram and control groups, respectively. Although no differences were found in weight gain between the groups, after 3 months of treatment, those receiving citalopram showed modest advantages regarding symptoms of depression, obsessive-compulsive symptoms, impulsiveness, and trait anger, as assessed by rating scales.

Data on the efficacy of tricyclic antidepressants are even more limited. In a study by Halmi et al. (635), 72 patients with anorexia nervosa were randomly assigned to receive cyproheptadine hydrochloride, a weight-inducing drug; amitriptyline hydrochloride, a tricyclic antidepressant; or placebo, using a double-blind method. Lower-weight patients with the restricting subtype of anorexia who were receiving intensive inpatient treatment seemed to benefit more, albeit to a modest degree, from either amitriptyline or cyproheptadine, compared with patients who were receiving placebo. In another double-blind, controlled study by Lacey and Crisp (636) of 16 patients with anorexia nervosa, no significant beneficial effect was observed from adding clomipramine to the usual treatment (although dosages of only 50 mg/day were used).

Small open-label studies in adults suggest that low doses of second-generation antipsychotic medications such as olanzapine may improve weight gain and psychological indicators, but controlled studies are needed to confirm this. Barbarich et al. (190) reported that 17 hospitalized patients with anorexia nervosa given open-label treatment with olanzapine for up to 6 weeks had a significant reduction in depression, anxiety, and core eating disorder symptoms and a significant increase in weight. Malina et al. (191) retrospectively questioned 18 patients with anorexia nervosa about their response to open treatment with olanzapine; the patients reported a significant reduction in anxiety, difficulty eating, and core eating disorder symptoms after taking olanzapine. Powers et al. (192) reported that of 14 patients with anorexia nervosa who completed a 10-week open-label study of olanzapine at 10 mg/day in an outpatient setting, 10 gained an average of 8.75 pounds, with 3 of those attaining their healthy body weight, and 4 lost a mean of 2.25 pounds. The patients also received weekly drug monitoring sessions and weekly group medication adherence sessions in which psychoeducation was provided.

The second-generation antipsychotic quetiapine, examined in an open-label study, had only a small benefit in terms of weight gain but some benefit in eating disorders–related preoccupation and depression (195, 196, 637).

In an open trial, 13 severely ill outpatients with anorexia nervosa, restricting type received low-dose (1–2 mg) haloperidol in addition to standard treatment and were reported to benefit (significant weight gain and improved insight) (198). Another study suggested no significant benefit for pimozide (638).

Although these pilot studies of antipsychotic medications are promising and suggest that these medications may be useful during the weight restoration phase, no controlled studies have been reported. In addition, few of the available studies have included male patients, only limited numbers of adolescents have been studied, and only case reports are available regarding prepubertal children. Again, controlled studies are needed.

Few controlled studies have been published on the use of other psychotropic medications for the treatment of anorexia nervosa. In one study (639), the use of lithium carbonate resulted in no substantial benefit. Uncontrolled studies of other somatic treatments, including vitamins, hormone treatments, and ECT, have also demonstrated no specific benefit (199).

In addressing associated features of anorexia nervosa, other medications have been used with apparent benefit, although evidence for their effect is limited. For example, antianxiety agents have been used selectively before meals to reduce anticipatory anxiety concerning eating (200, 201).

Although supplemental estrogen-progestins, calcium, and vitamin D are often prescribed in routine practice in an effort to minimize or ameliorate osteopenia or osteoporosis (203), they have not been shown to meaningfully prevent or reverse skeletal deterioration. Rather, nutritional rehabilitation during the period of bone growth is the only practical intervention shown to potentially reverse bone loss (413, 502, 640). The only controlled trial to date that has examined the effects of estrogen administration on adult women with anorexia nervosa (n = 44) showed that estrogen-treated patients had no significant change in bone mass density compared with control patients (204). However, the six estrogen-treated patients whose initial body weight was <70% of their healthy weight had a 4.0% increase in mean bone density, whereas the 10 subjects of comparable body weight not treated with estrogen had a further 20.1% decrease in bone density. This finding suggests that hormone replacement therapy may help a subset of low-weight women with anorexia nervosa. At the same time, artificially inducing menses carries the risk of supporting or reinforcing a patient's denial that she does not need to gain weight.

Experimental approaches to bone revitalization using recombinant human insulin-like growth factor, bone growth factors (641), and biphosphonates (642) have also been attempted, but these approaches cannot be recommended for routine practice. Studies concerning these agents and other investigative treatments are now under way.

CBT specifically directed at the eating disorder symptoms and underlying cognitions in patients with bulimia nervosa is the psychosocial intervention that has been most intensively studied in adults and for which there is the most evidence of efficacy (209, 212, 643–654). Although there has been considerable variation in the way in which CBT has been implemented, several controlled trials used short-term, time-limited interventions, such as 20 individual psychotherapy sessions over 16 weeks, with two scheduled visits per week for the first 4 weeks (212, 643, 644, 653, 655–662). Significant decrements in binge eating, vomiting, and laxative abuse have been documented among some patients receiving CBT; however, the percentage of patients who achieve full abstinence from bingeing/purging behavior is variable and often includes only a small number of patients (212, 643, 646, 648–655, 663–668). Among studies with control arms, CBT has been shown to be superior to waiting list (663–665, 667), minimal intervention (668), nutritional counseling alone (652), or nondirective control (666) conditions. In most of the published CBT trials, significant improvements in either self-reported (665, 669) or clinician-rated (670) mood have been reported. To date, there are no published treatment trials of adolescents with bulimia nervosa, although two studies are nearing completion (unpublished study of U. Schmidt et al. in London and unpublished study of D. le Grange et al. in the United States). Case series concerning CBT treatment for adolescents with bulimia nervosa are now available or in press (671, 672).

In practice, many other types of individual psychotherapy are used in the treatment of bulimia nervosa, such as interpersonal, psychodynamically oriented, or psychoanalytic approaches. Clinical experience and naturalistic survey research suggest that these approaches can help in the treatment of the co-occurring mood, anxiety, personality, interpersonal, and trauma- or abuse-related disorders that frequently accompany bulimia nervosa (153, 673). Evidence for the efficacy of these treatments for bulimia nervosa comes mainly from case reports and case series. Some modes of therapy, including the interpersonal and psychodynamic approaches, have been studied in randomized trials as comparison treatments for CBT or in separate trials (212, 643, 645, 674). In general, these and other studies have shown IPT to be helpful as an acute treatment. However, one trial examined the efficacy of IPT in patients who did not respond to CBT and found a low response rate coupled with a high dropout rate (211).

Although the specific forms of focused psychodynamic psychotherapy that have been studied in direct comparison with CBT have generally not been as effective as CBT in short-term trials (656, 657), it must be understood that in these studies the therapists were proscribed from undertaking any discussion of the patients' symptoms in the 20 sessions of psychodynamic treatment they provided.

The behavioral technique of exposure (e.g., to bingeing foods) plus response prevention (e.g., inhibiting vomiting after eating) has also been considered as treatment for bulimia nervosa. However, data on the efficacy of this approach are conflicting, as studies have reported enhanced (675), not significantly altered (676), and reduced (655) responses to CBT when this type of behavioral therapy was used as an adjunct. On the basis of results from a large clinical trial, and given the trial's logistical complexity, exposure treatment does not appear to have additive benefits over a solid core of CBT (648, 649), a finding supported by a recent authoritative meta-analysis (651).

Few studies have directly compared the effectiveness of various types of individual psychotherapy in the treatment of bulimia nervosa. One study by Fairburn et al. (643) that compared CBT, IPT, and behavior therapy showed that all three treatments were effective in reducing binge eating symptoms by the end of treatment but that CBT was most effective in improving disturbed attitudes toward shape and weight and restrictive dieting . However, at long-term follow-up (mean of 5.8 years), the study found equal efficacy for IPT and CBT on eating variables, attitudes about shape and weight, and restrictive dieting (659), which suggests that the IPT patients had "caught up" in terms of benefits over time. A second multicenter study that compared CBT with IPT suggested that CBT worked more rapidly (653) and showed greater efficacy than IPT at the end of 20 weeks of treatment (212). However, at 1-year follow-up, no significant difference was noted between the two treatment groups (212).

Another trial reported the efficacy of dialectical behavior therapy (677), and an ongoing multicenter study (678) has reported similar findings. In a 6-week controlled study involving 50 patients assigned to guided affective imagery therapy or a control group, those receiving the active treatment showed substantially more improvement in binge eating and purging behaviors and eating disorder–related attitudes (679).

Group psychotherapy approaches have also been used to treat bulimia nervosa. Many clinicians favor a combination of individual and group psychotherapy. Psychodynamic and cognitive-behavioral approaches may also be combined. Group therapy may help patients to more effectively deal with the shame surrounding their disease as well as provide additional peer-based feedback and support. A meta-analysis of 40 group treatment studies suggested moderate efficacy for group therapy, with those studies that provided 1-year follow-up data reporting that improvement was typically maintained (680). There is some evidence that group treatment programs that include dietary counseling and management as part of the program are more effective than those that do not (208) and that more frequent visits (e.g., sessions several times a week) throughout treatment (646) or early in treatment (644, 646) result in improved outcome.

A direct comparison of group and individual CBT failed to find evidence of substantially different outcomes, with patients in both treatments doing equally well at follow-up (681). However, in a meta-analysis, the data showed clear advantages of delivering CBT in an individual versus a group format (216).

Family therapy has been reported to be helpful in the treatment of bulimia nervosa in a large case series of adults (682), but more systematic studies are not available. A systematic study of family therapy for adolescents with bulimia nervosa is currently under way (89), but no results have been reported yet.

Family therapy should be considered whenever possible, especially for adolescents who still live with their parents, older patients with ongoing conflicted interactions with parents, or patients with marital discord. For women with eating disorders who are mothers, parenting help and interventions aimed at assessing and, if necessary, aiding their children should be included (386–388).

Some patients have found Overeaters Anonymous and similar groups to be helpful as adjuncts to initial treatments or for prevention of subsequent relapses (223, 683), but no data from short- or long-term outcome studies of these programs have been reported.

A growing body of literature has suggested that CBT can be administered successfully through self-help or guided self-help manuals, at times in association with pharmacotherapy (220, 222, 651, 684–689). A review of self-help treatments with or without guidance by a professional has recently been published (690). Although such techniques are not yet sufficiently developed to recommend their acceptance as a primary treatment strategy, developments in this area may prove of great importance in providing treatment to patients who otherwise might not have access to adequate care. Clinicians unfamiliar with the CBT approach may benefit from acquainting themselves with these CBT treatment manuals and obtaining specialized training in CBT to further help their bulimia nervosa patients (656, 691–696).

Early observations that individuals with bulimia nervosa exhibit an elevated lifetime prevalence of mood disorders, together with an elevated prevalence of mood disorders in their first-degree relatives, prompted initial trials of antidepressants for the acute treatment of bulimia nervosa (233). In these trials, antidepressants appeared to be effective for bulimia nervosa regardless of whether or not the patient was clinically depressed. Subsequent randomized trials confirmed that nondepressed patients responded to these medications and that the baseline presence of depression was not a predictor of medication response (234, 235, 697, 698). Although wide variability exists across studies, reductions in binge eating and vomiting rates in the range of 50%–75% have been achieved with active medication (181, 224, 230, 241, 699–714).

Specific antidepressant agents that have demonstrated efficacy among patients with bulimia nervosa in double-blind, placebo-controlled studies include trazodone (708); tricyclic compounds such as imipramine (233, 236), desipramine (234, 237, 238, 240), and amitriptyline (for mood but not eating variables) (235); the SSRIs fluoxetine (224, 230, 698, 703) and sertraline (227) but not fluvoxamine (715); and several MAOIs, including phenelzine (697) and isocarboxazid (716) but not moclobemide (717). Treatment with bupropion was also efficacious (181), but its use is not recommended because of the association between bupropion treatment and seizures in purging bulimic patients. The results of one study (709) suggest that patients with atypical depression and bulimia nervosa may preferentially respond to phenelzine in comparison with imipramine. However, because MAOIs are potentially dangerous in patients with chaotic eating and purging habits, they should be used cautiously in bulimia nervosa patients.

Dosages of tricyclic and MAOI antidepressants for treating patients with bulimia nervosa parallel those used to treat patients with depression, although fluoxetine at dosages higher than those used for depression may be more effective for bulimic symptoms (e.g., 60–80 mg/day). The first multicenter fluoxetine study (230) demonstrated that 60 mg/day was clearly superior to 20 mg/day on most variables, and in a second study (224) all subjects receiving active medication started with 60 mg/day. A third trial used 60 mg/day as a maintenance therapy, but the dropout rate from the study was very high (226). The medication was surprisingly well tolerated at this dosage, and many clinicians initiate treatment for bulimia nervosa with fluoxetine at a higher dosage, titrating downward if necessary to manage side effects.

Three trials have examined the effectiveness of antidepressant maintenance therapy. One trial with fluvoxamine (701) demonstrated an attenuated relapse rate versus placebo in patients with bulimia nervosa who were on a maintenance regimen of the medication after leaving an inpatient treatment program. However, in the continuation arm of a clinical trial with desipramine (240), 29% of the patients entering that phase experienced a relapse within 4 months. A 1-year maintenance trial using fluoxetine or placebo found evidence for a lower relapse rate on active drug, but the dropout rate was quite high in both treatment arms (226). An open-label trial found evidence that fluoxetine could be effective in adolescents (225).

Two studies have examined the efficacy of medication in patients who did not respond to IPT and/or CBT (211, 250). The results were somewhat inconsistent in that one study found medication to be quite useful, whereas the other found low rates of response and high dropout rates.

A number of other medications have been used experimentally for bulimia nervosa without evidence of efficacy, including fenfluramine (700) and lithium carbonate (241). Fenfluramine has now been taken off the market because of links between its use (mainly in combination with phentermine) and cardiac valvular abnormalities. Lithium may occasionally be used concurrently for the treatment of co-occurring conditions. The opiate antagonist naltrexone has been studied in three randomized trials at dosages used for treating narcotic addiction and preventing relapse among alcohol-dependent patients (50–120 mg/day). The results consistently show that the medication is not superior to placebo in reducing bulimic symptoms (699, 704, 707). In a small double-blind, crossover study involving higher dosages (e.g., 200–300 mg/day), naltrexone did appear to have some efficacy. Further studies using these dosage ranges are needed. However, there have been mixed reports concerning the risk of hepatotoxicity with the use of high dosages of naltrexone (705, 706, 718).

Other agents that have been shown to be efficacious in treating bulimia nervosa symptoms in randomized trials are topiramate, an anti-epileptic agent (242, 243), and ondansetron, an antinausea drug that decreases afferent vagal neurotransmission through its action as a 5-HT₃ antagonist (719); however, odansetron requires multiple daily administrations and is quite expensive. In a randomized controlled trial, carbamazepine showed efficacy in only one patient, and that patient had a history of bipolar disorder (720).

The relative efficacy of psychotherapy, medication, or both in the treatment of bulimia nervosa has been examined in six studies. In the first study (253), intensive group cognitive psychotherapy (45 hours of therapy over 10 weeks) was superior to imipramine alone in reducing binge eating/purging and depressive symptoms. Imipramine plus intensive group CBT did not improve the outcome on eating variables but did improve depression and anxiety variables. In the second study (254), patients in group CBT improved more than those receiving desipramine alone. Some advantage was also seen for combination therapy on some variables, such as dietary restraint. The results of the third study (255), which compared fluoxetine treatment, CBT, and combination therapy, favored CBT alone and suggested little benefit for combination therapy. These results, however, are difficult to interpret because of the high attrition rate (50% by the 1-month follow-up). In the fourth study (251), CBT was superior to supportive psychotherapy, and active medication (consisting of desipramine followed by fluoxetine if abstinence from binge eating and purging was not achieved) was superior to placebo in reducing eating disorder behaviors. The combination of CBT and active medication resulted in the highest abstinence rates. The use of sequential medication in this study addressed a limitation of earlier studies in that when one antidepressant fails, a clinician typically tries other agents that often result in better antidepressant efficacy than the first medication alone. In the fifth study (702), no advantage was found for the use of fluoxetine over placebo in an inpatient setting, although both groups improved significantly. In the sixth study (721), combination treatment with desipramine and CBT was terminated prematurely because of a high dropout rate. A recent review concluded that the combination treatment was superior on some variables (711).

In general, studies show the importance of achieving abstinence from binge eating and vomiting regardless of the interventions used in the treatment of bulimia nervosa, and they confirm that longer-term outcome is better when abstinence is achieved after short-term interventions (722).

Some studies of treatments for binge eating disorder have prioritized weight loss as the primary goal, whereas others have prioritized cessation of binge eating. Both types of studies will be addressed below.

In patients with binge eating disorder, very-low-calorie diets alone have been associated with substantial initial weight losses, with >33% of these patients maintaining their weight loss 1 year after treatment (723–726). In some studies using low-calorie diets, either significant weight loss did not occur (727) or weight was partially regained during the first year (290, 728). The pattern of weight regain after initial weight loss is common in all general medical and psychological treatments for obesity and not only for obesity associated with binge eating disorder. In one study, the presence of subthreshold or full syndrome binge eating disorder at baseline did not appear to adversely affect weight loss in programs using behavioral weight control, a low-calorie diet, and aerobic and strength training (729). Likewise, in a study using telephone- and mail-based behavioral weight control for obesity, binge eating status at baseline was not associated with outcome (730). Treatments with psychotherapies that do not specifically address weight control, such as CBT, IPT, and dialectical behavior therapy, ordinarily do not yield significant weight reduction (see below), but adding exercise and 6-month biweekly maintenance to CBT may improve weight loss in patients with binge eating disorder (731).

Binge eating is substantially reduced in programs using very-low-calorie diets, but a small number of individuals may experience a reemergence of binge eating when regular meals are reintroduced (585, 723, 724, 726). Among individuals who do not manifest binge eating prior to treatment, behavioral weight control with a low-calorie diet does not appear to promote the emergence of binge eating (732). There is as yet little evidence regarding patient characteristics (e.g., age at onset of binge eating, magnitude and frequency of weight fluctuation) that predict differential response to programs that prioritize weight versus those that prioritize binge cessation.

CBT, behavior therapy, dialectical behavior therapy, and IPT have all been associated with binge frequency reduction rates of 67% or more and significant abstinence rates during active treatment (272, 276, 733–739). Deterioration during the follow-up period has been observed with all three forms of psychotherapy; however, in some cases, maintenance of change at 1-year follow-up has been substantial (271, 272, 731). Several studies have examined various treatments intended to augment standard CBT. The addition of exercise appears to augment both binge and weight reduction (731), whereas spouse involvement in treatment does not significantly improve outcome (88). One group reported promising effects on binge eating with a novel virtual reality modification of standard treatment (739, 740).

Nondiet approaches that focus on self-acceptance and healthy lifestyle rather than weight loss per se may reduce binge eating, depression, anxiety, bulimia, the drive for thinness, body dissatisfaction, total and LDL cholesterol, and systolic blood pressure while increasing moderate physical activity levels (741, 742). One study failed to find a difference between dieting and nondieting approaches in reducing binge eating and weight. In an expected observation, however, even the dieting treatment did not yield significant weight loss in this study, calling into question the integrity of the treatments (742, 743).

Self-help programs using self-guided, professionally designed manuals have been effective in reducing the symptoms of binge eating disorder in the short run for some patients and may have long-term benefit (273–277). One recent study found that guided self-help CBT was superior to guided self-help behavioral weight loss treatment for binge remission (46% vs. 18%, respectively), although neither treatment produced significant weight loss (277).

A variety of SSRIs (citalopram, fluoxetine, fluvoxamine, and sertraline) (744–747) and tricyclic antidepressants (desipramine and imipramine) (699, 748) have been found to be associated with significantly greater decreases in binge frequency than placebo. Fluvoxamine was not superior to placebo in a controlled study, in part due to a high placebo response (749). A retrospective chart review study of the serotonin-norepinephrine reuptake inhibitor venlafaxine in obese patients with binge eating disorder reported beneficial effects on eating, weight, and mood (750). As is the case for bulimia nervosa, in most studies using SSRIs for binge eating disorder, the dosages used have been at or near the high end of the recommended dosage range. Where follow-up data were reported, it appears that patients tend to relapse after medication is discontinued (289, 748); however, most medication studies to date do not report follow-up data. For the most part, treatment with antidepressants has not been demonstrated to yield clinically significant weight loss in this population, although one study reported an estimated weight loss of 5.6 kg with sertraline treatment compared with 2.4 kg in the placebo group (747). It should be noted that the treatment of other psychiatric disorders (e.g., depression) with SSRIs has at times been associated with weight gain, particularly in the long term (751).

The appetite-suppressant medication sibutramine has also shown promise in the treatment of binge eating disorder. In a randomized controlled trial (284), sibutramine was shown to have significant beneficial effects on binge eating and weight loss, with remission rates of 40% and 27% in the sibutramine and placebo groups, respectively, and a weight decrease of 7.4 kg versus a weight increase of 1.4 kg in these groups, respectively. A laboratory feeding study reported that subjects with binge eating disorder treated with sibutramine versus placebo for 4 weeks in a crossover design consumed less in a laboratory binge meal and lost more weight (3.2 vs. 0.5 kg) after sibutramine than after placebo treatment (752). Although the appetite-suppressant medications fenfluramine and dexfenfluramine have also been found to significantly reduce binge frequency (289), their use has been associated with serious adverse events, including a 23-fold increase in the risk of developing primary pulmonary hypertension when used for >3 months (753). Studies have suggested that patients taking the combination of fenfluramine and phentermine may be at greater risk of heart valve deformation and pulmonary hypertension; as a result, fenfluramine has been withdrawn from the market (753–756).

Most recently, the anticonvulsants topiramate and zonisamide have been studied in patients with binge eating disorder. Two open studies, one a retrospective review of patients with affective disorders and co-occurring binge eating disorder (757) and the other an open-label prospective study (758), as well as one randomized, double-blind, placebo-controlled study (286), found topiramate to be effective for both binge suppression and weight loss. The latter study reported remission in 64% of the topiramate group versus 30% of the placebo group, with weight loss of 5.9 and 1.2 kg in the topiramate and placebo groups, respectively. A 1-year open-label extension of this study (287) found that these effects were largely maintained over the study period; however, only a small number of the patients (10 of 61) remained in the study for the full year, and adverse effects, including paresthesias, dry mouth, cognitive problems, headache, dizziness, somnolence, fatigue, and dyspepsia, led to discontinuation in about 33% of patients. An open-label study of the anticonvulsant zonisamide (288) suggests that it may have similar effects, both in clinical response and in adverse events.

Finally, naltrexone has been associated with a decrease in binge frequency similar to that reported with antidepressant medications, although the response rate did not differ from that of placebo (699). This observation underscores the fact that high placebo response rates are found in many studies of binge eating disorder, so caution is required in evaluating the claims of effective treatments, particularly those using a waiting-list control condition (289, 699, 749).

In some studies, coadministration of medication with psychotherapy or dietary counseling has been found to be associated with significantly more weight loss than has psychotherapy or dietary counseling alone (290–292). A recent study found that in patients receiving guided self-help CBT for binge eating disorder, the addition of orlistat yielded significantly higher rates of binge remission after active treatment but not at 3-month follow-up and significantly greater weight loss both after treatment and at 3-month follow-up compared with placebo (295). However, other studies suggest that concomitant administration of SSRI antidepressants adds little benefit in binge reduction or weight loss for patients treated with CBT for binge eating disorder (293, 294, 759). One recent study in which subjects with binge eating disorder received individual CBT plus fluoxetine, individual CBT plus placebo, fluoxetine, or placebo found that individual CBT but not fluoxetine demonstrated efficacy for reducing binge eating but not weight (293). A similar study, in which subjects with binge eating disorder were randomized to receive the same four treatments as adjuncts to group behavioral treatment, found that adjunctive individual CBT, but not fluoxetine, resulted in significant additional binge reduction, whereas fluoxetine appeared to augment the reduction in depressive symptoms. Although neither adjunctive treatment contributed significantly to weight loss, the 54 subjects who achieved binge remission lost an average of 6.2 kg, whereas the 62 subjects who did not achieve remission gained 0.7 kg (294).

There are few available studies of treatments for night eating syndrome. One open-label study of sertraline at dosages of up to 200 mg/day for night eating syndrome found improvements in both the number of awakenings and the nocturnal ingestions, with full remission in 29% of subjects (302). A subsequent double-blind study of sertraline for night eating syndrome reported that in a group of 24 patients, 75% of those treated with sertraline versus 25% of those who received placebo were considered to have responded to the treatment. Response was sustained during the 6-month open-label follow-up period (760). A small case series of four patients, two with night eating syndrome and two with the related condition of nocturnal sleep-related eating disorder, reported that topiramate treatment was helpful (761). Abbreviated progressive muscle relaxation training may be useful in treating night eating syndrome (301). Finally, other treatments reported to be helpful in sleep-related eating disorder are carbidopa/L-dopa, bromocriptine, codeine, and clonazepam (762, 763).