OUP user menu

Evidence-based pharmacotherapy of eating disorders

Martine F. Flament, Hany Bissada, Wendy Spettigue
DOI: http://dx.doi.org/10.1017/S1461145711000381 189-207 First published online: 1 March 2012


The objective was to review scientific evidence for efficacy and safety of pharmacotherapy in adults or children with an eating disorder (ED). We conducted a computer search for all randomized controlled trials (RCTs) published between 1960 and May 2010 for treatment of anorexia nervosa (AN), bulimia nervosa (BN) or binge-eating disorder (BED). For drugs for which no RCT was found, open trials or case reports were retrieved. Clinically relevant RCTs in the treatment of AN have used atypical antipsychotics, selective serotonin reuptake inhibitors (SSRIs), and zinc supplementation. Olanzapine demonstrated an adjunctive effect for in-patient treatment of underweight AN patients, and fluoxetine helped prevent relapse in weight-restored AN patients in 1/2 studies. For treatment of BN, controlled studies have used SSRIs, other antidepressants, and mood stabilizers. In 9/11 studies, pharmacotherapy yielded a statistically significant although moderate reduction in binge/purge frequency, and some additional benefits. For BED, RCTs have been conducted using SSRIs and one serotonin norepinephrine reuptake inhibitor (SNRI), mood stabilizers, and anti-obesity medications. In 11/12 studies, there was a statistically significant albeit limited effect of medication. Meta-analyses on efficacy of pharmacotherapy for BN and BED support moderate effect sizes for medication, but generally low recovery rates. Treatment resistance is an inherent feature of AN, where treatment should focus on renourishment plus psychotherapy. For BN and BED, combined treatment with pharmacotherapy and cognitive behaviour therapy has been more effective than either alone. Data on the long-term efficacy of pharmacotherapy for EDs are scarce. Short- and long-term pharmacotherapy of EDs still remains a challenge for the clinician.

Key words
  • Anorexia nervosa
  • binge eating disorder
  • bulimia nervosa
  • paediatric
  • pharmacotherapy


Eating disorders (EDs) are widespread, disabling and often chronic psychiatric disorders. Anorexia nervosa (AN) and bulimia nervosa (BN) share a morbid preoccupation with weight and shape. AN is characterized by an ego-syntonic restriction of food intake resulting in a body weight lower than 85% of normal, and can be of restricting type (dieting and excessive exercise) or binge eating/purging type (associated binge/purge behaviours) (APA, 1994). The eating pattern in BN consists of recurring disinhibition of restraint, resulting in cycles of binge eating and compensatory behaviours to prevent weight gain, including self-induced vomiting and abuse of laxatives/diuretics in the purging type, or extreme exercise and fasting in the non-purging type. Most individuals with BN maintain an average body weight (Fairburn & Harrison, 2003). Binge-eating disorder (BED) is characterized by recurrent binge-eating episodes accompanied by a sense of distress, loss of control, and feelings of disgust, depression, or guilt (Spitzer et al. 1991). Most persons with BED are obese, and often seek treatment for weight loss rather than for disturbed eating behaviours. Night eating syndrome (NES) is a form of BED characterized by consumption of at least 25% of food intake after the evening meal, conscious nocturnal eating at least twice a week for 3 months, or both (Allison et al. 2010). The American Psychiatric Association is recommending modifications to some of the diagnostic criteria for EDs in the forthcoming revision of the DSM (DSM-V). Suggested changes include: ‘markedly low body weight’ to replace ‘body weight less than 85% of that expected’, and removal of the amenorrhea criterion for AN; listing of BED as a stand-alone diagnosis, and NES as a form of BED; restricting the diagnosis of BN to the purging form, with the non-purging form classified as BED; and a frequency criterion of at least once a week (rather than twice a week) for binge eating and/or inappropriate compensatory behaviours in the diagnosis of BN or BED (www.DSM5.org).

In young women, lifetime prevalence rates have been estimated at 0.3–0.9% for AN, 1–2% for BN, 3.5% for BED, and up to 10% for eating disorders not otherwise specified (EDNOS); among males, lifetime prevalence estimates are 0.3% for AN, 0.5% for BN, and 2.0% for BED (Grucza et al. 2007; Hoek, 2006; Hoek & Van Hoeken, 2003; Hudson et al. 2007; Machado et al. 2007). In children and adolescents, EDs profoundly affect physical, emotional and psychosocial development. In both youth and adults, psychiatric complications include social isolation, major depression, severe anxiety, substance abuse and suicide (Hudson et al. 2007), and medical complications encompass cardiac problems, osteopenia, and infertility (Mitchell & Crow, 2006). The mortality among women with an ED is as high as to 20% (Nielsen, 2001), and the impact of EDs on the healthcare system is enormous (Simon et al. 2005).

The objective of this paper is to review currently available evidence for efficacy and safety of pharmacotherapy in adults and children with EDs. We conducted a computer search of Medline and PsycINFO databases for pharmacotherapy of AN, BN and BED from 1960 to May 2010. The search included: randomized controlled trials (RCTs); if none were available (e.g. for paediatric EDs), open trials or case reports suggesting benefits; systemic reviews; meta-analyses; and guidelines. Pharmacotherapy for AN, BN, BED and NES are presented consecutively. Findings from clinically relevant RCTs are summarized in Tables 1–9. Other studies and meta-analyses are described in the text. For AN and BN, studies conducted in paediatric populations are reported separately (whenever available). Based on the evidence reviewed, we address three questions: (1) what is the first line pharmacotherapy of choice for EDs? (2) what is the evidence for maintenance treatment? and (3) what is the best approach to treatment resistance?

Pharmacotherapy of AN

Antipsychotics (Table 1)

It was initially argued that patients with AN should be treated with antipsychotics because their obsessions regarding weight and body shape could be viewed as delusional. Early RCTs have tested chlorpromazine (Dally & Sargant, 1966), pimozide (Vandereycken & Pierloot, 1982), and sulpiride (Vandereycken, 1984), with trends for higher weight gain, but no advantage on eating attitudes and behaviours. Given their potential for short- and long-term side-effects, and the fact that weight gain in anorexic patients tends not to persist if not accompanied by changes in attitudes towards eating and body shape, the treatment of AN with traditional antipsychotics is no longer recommended.

The atypical antipsychotics have a side-effect profile that often includes weight gain and an increased serotonin to dopamine blockade ratio, which make them potential candidates for pharmacotherapy of AN (see Table 1). In a small randomized trial, olanzapine was superior to chlorpromazine for reduction of anorexic ruminations (Mondraty et al. 2005). Brambilla et al. (2007) compared individual cognitive behavioural treatment (CBT) with and without olanzapine. There was a greater improvement in depression, anxiety, obsessive-compulsiveness, and aggressiveness in the olanzapine group. The increase in body mass index (BMI) was larger for participants with AN binge/purge type receiving olanzapine. Bissada et al. (2008) compared olanzapine to placebo, and concluded that olanzapine may be safely used in AN to achieve more rapid weight gain and reduce obsessive symptoms, as measured on the Yale–Brown Obsessive–Compulsive Rating Scale (YBOCS); there were no serious side-effects, and no evidence of impaired glucose tolerance. Quetiapine (50–800 mg/d) has been investigated in two small open studies, with significant changes in BMI and the Eating Disorder Examination (EDE) restraint subscale in one study (Bosanac et al. 2007), and decreases in measures of anxiety, depression, and the Positive and Negative Syndrome Scale (PANSS) score, but no significant change in BMI in another (Powers et al. 2007). Trunko et al. (2010) reported on eight patients (five with AN, three with BN) treated with aripiprazole, in addition to an antidepressant, for 4 months to 3 yr. They described a significant reduction in eating-specific anxiety as well as obsessional thoughts about food, weight, and body image, and full or partial weight restoration in all AN patients; the drug was well tolerated.

View this table:
Table 1

Antipsychotics in children and adolescents

Small case series using olanzapine in children and adolescents suggested: decrease in body image concerns, agitation, and anxiety regarding eating (Mehler et al. 2001); decreased agitation and anxiety, and improved weight gain (Boachie et al. 2003); and decreased eating anxiety and ruminations (Dennis et al. 2006). Barbarich et al. (2004a) conducted a 6-wk open trial of olanzapine in 17 adolescent or adult patients with AN. Olanzapine significantly reduced depression, anxiety, and ED symptoms, and significantly increased weight; there were no dropouts from the study, and no abnormal biological findings. In a retrospective chart review, Norris et al. (in press) compared 43 adolescents with AN treated with olanzapine (2.5–7.5 mg/d) to controls. Those on olanzapine had greater illness severity and higher rates of comorbidity. Side-effects (albeit mild) were reported in 56%, including abnormal lipid profiles in 29%. Two single case reports indicated possible benefits of risperidone for treating paediatric AN (Fisman et al. 1996; Newman-Toker, 2000), and Mehler-Wex et al. (2008) reported three cases treated successfully with adjunctive quetiapine.

Tricyclic antidepressants (TCAs)

Due to clinical and biological similarities between AN and depression (Halmi et al. 1991), the TCAs were seen, 30 yr ago, as the most promising drug treatment for AN. Unfortunately, findings from a few RCTs using amitriptyline (Biederman et al. 1985; Halmi et al. 1986) or clomipramine (Crisp et al. 1987; Lacey & Crisp, 1980) have not been encouraging. Given, on the one hand, the limited evidence for any clinical benefit and, on the other, the lethal risk with overdose and the potential for fatal arrhythmia at low body weight, particularly in youth, this type of medication is not recommended today for AN.

Selective serotonin reuptake inhibitors (SSRIs) (Table 2)

In the last decades, SSRIs have been increasingly used to treat AN, but only two have been tested in RCTs, with mostly negative results. Fluoxetine conveyed no advantage over placebo for underweight AN in-patients (Attia et al. 1998), nor for AN outpatients treated with fluoxetine plus serotonin precursors (Barbarich et al. 2004b). Kaye et al. (2001) randomly assigned restricting AN patients to fluoxetine or placebo after in-patient weight gain. In the following year, women receiving fluoxetine had a significantly lower rate of relapse than those treated with placebo. However, limitations of this study include its small size (only 13 completers), and the lack of standardized psychological treatment during the medication trial. In a larger RCT including 93 weight-restored outpatients randomized to CBT plus fluoxetine or placebo, Walsh et al. (2006) found no difference in relapse rate after 1 yr. Fassino et al. (2002) conducted a RCT comparing citalopram to a waitlist condition in restricting AN patients. Weight gain was similar in both groups, but citalopram appeared to improve depression, obsessive-compulsive symptoms, impulsiveness and trait-anger.

View this table:
Table 2

SSRIs in children and adolescents

Holtkamp et al. (2005) retrospectively compared 19 partially weight-restored adolescents with AN treated with fluoxetine (mean 35 mg/d), fluvoxamine (35 mg/d), or sertraline (100 mg/d) to 13 unmedicated adolescent patients. They found no group differences in BMI, ED, depression or obsessive-compulsive scores, and concluded that there was no effect of SSRIs on the course of illness. Given the limited evidence to support their use, SSRIs should be reserved for treating comorbid depression or anxiety in paediatric patients with AN who have been weight-restored. Due to concerns about possible increased suicidal ideation in young people on SSRIs, medication should be monitored closely. Currently, fluoxetine has the most evidence favouring its use for treating depression and anxiety in children and adolescents.

Other antidepressants

Mirtazapine is a non-SSRI sedative antidepressant which increases appetite and induces weight gain. One open trial by Schüle et al. (2006) reported weight gain on mirtazapine in five AN inpatients.

Mood stabilizers and anticonvulsants

Lithium is known to induce weight gain and has been used for AN patients in a small RCT, with greater average weight gain on lithium than placebo (Gross et al. 1981). However, the use of lithium is not recommended in AN, because sodium and fluid depletion may lead to reduced lithium clearance, resulting in increased potential toxicity. Only a few case reports were found for the use of anti-epileptic medications with mood-stabilizing properties. Available data are mixed, preventing definitive conclusions (McElroy et al. 2009).

Medications that improve gastric emptying (prokinetic agents)

Anorexic patients commonly complain of feelings of fullness, early satiety and bloating. Biological measures of gastric motility may be abnormal (Stacher et al. 1986). Prokinetic agents such as metoclopramide, bethanecol, cisapride, and domperidone, have been suggested as adjuvants to reduce post-prandial abdominal bloating in anorexic patients. Cisapride has been removed from the market because of concerns about potentially fatal cardiac effects.

Opiate antagonists

The opioid peptide system helps modulate appetite and feeding behaviours in animals, and disturbance in the opioid function has been shown in patients with AN (Baranowska et al. 1984).

A RCT by Marrazzi et al. (1995) assessed the efficacy of 200 mg/d of naltrexone in six patients with AN or BN. Although the amount of weight gain was not analysed, the binge/purge frequency was reduced on naltrexone. However, at the dose used (four times the recommended dose), there is a high risk for significant elevation in hepatic transaminases, reflected in the black box warning that ‘naltrexone can cause hepatocellular injury at excessive doses’.

Appetite enhancers

Cyproheptadine, a serotonin and histamine antagonist noted to produce weight gain in children with asthma, has been tested in controlled trials in anorexic patients, with disappointing results (Goldberg et al. 1979). Halmi et al. (1986) compared amitriptyline (160 mg/d), cyproheptadine (32 mg/d), and placebo in 72 females aged 13–36 yr. An unexpected finding was that cyproheptadine decreased the length of time to achieve a healthy weight for the restricting anorexic patients, but significantly impaired treatment efficiency for the binge/purge anorectic patients, compared to amitriptyline and placebo. Tetrahydrocannabinol (THC) has an appetite-stimulating effect and has been tested in a small RCT vs. diazepam in 11 anorexic adults; no significant effect of THC was found on weight gain, while side-effects included sleep disturbance, paranoia, and dysphoria (Gross et al. 1983). It should be emphasized that AN patients are not without appetite (Garfinkel & Garner, 1982), but are frightened about losing control over their hunger; thus, increasing appetite could potentially increase anxiety.

Other medications

AN patients often lose bone mass – 92% develop early osteopenia and 40% progress to osteoporosis – at a time (adolescence and young adulthood) when they should be optimizing bone growth (Grinspoon et al. 2000). Given the effectiveness of bisphosphonates and hormone therapy in the treatment of osteopenic post-menopausal women, their use in AN was evaluated. In a RCT using biphosphonates, the increase in bone mass density did not differ between active treatment and control groups (Golden et al. 2005). Mehler & MacKenzie (2009) reviewed three RCTs using oestrogens in AN. None demonstrated any significant improvement in bone mass density over the control groups.

Other medications in children and adolescents (Table 3)

Individuals with zinc deficiency seem to exhibit symptoms similar to AN, i.e. weight loss, depression, appetite and taste changes, and amenorrhoea (Zhu & Walsh, 2002). Given the nutritional deprivation in anorexic patients, zinc supplementation has been suggested. Three RCTs conducted in children and adolescents (Birmingham et al. 1994; Katz et al. 1987; Lask et al. 1993) have produced mixed results, and the role of zinc supplementation remains controversial.

View this table:
Table 3

Pharmacotherapy of BN


Seven RCTs have been published using TCAs for treatment of BN in adults (two with imipramine, four with desipramine, and one with amitriptyline). In all but one, the active drug (imipramine or desipramine) led to a mean reduction in frequency of binge eating (47–91%) significantly superior to that on placebo. However, because TCAs often cause unpleasant side-effects and have the potential to be fatal in overdose, this class of medication is not considered as a first-line pharmacological treatment for BN in adults, and should not be used to treat BN in children or adolescents.

Monoamine oxidase inhibitors (MAOIs)

Four RCTs have used MAOIs (two phenelzine, one isocarboxazid, one brofaromine) in patients with BN. In Walsh et al. (1988), phenelzine (60–90 mg/d) was superior to placebo for both reduction of binge frequency (64% vs. 5%) and the total number of participants achieving remission after 10 wk (35% vs. 4%). However, follow-up data indicated that all initial responders had relapsed at 6 months, and by 4 yr, 3/15 had experienced severe hypertensive episodes, including one with a fatal outcome (Fallon et al. 1991). In a RCT by Kennedy et al. (1988), 18 women with BN were assigned to either isocarboxazid 60 mg/d or placebo for 13 wk; on isocarboxazid, there was a significant reduction in binge eating and vomiting, and no serious adverse effects. A later RCT using phenelzine (75 mg/d) in 18 depressed bulimic patients resulted in significant reduction of both bulimic and depressive symptoms (Rothschild et al. 1994). Brofaromine is a reversible MAOI reported as somewhat effective in individuals with BN (Kennedy et al. 1993), but is not commercially available in the USA. Overall, due to the risk of adverse reactions or drug interactions associated with the use of MAOIs, and the potential for the ingestion of tyramine-containing food during a binge episode causing a hypertensive crisis, MAOIs are not a first-line treatment for BN.

SSRIs (Table 4)

Fluoxetine has been the most studied of the SSRIs. A large multicenter trial (Fluoxetine Bulimia Nervosa Collaborative Study Group, 1992) compared two dosages of fluoxetine to placebo. At 60 mg/d, fluoxetine was superior to placebo for reduction of binging (median reduction: 67% vs. 33% on placebo) and vomiting frequency (56% vs. 5%). At 20 mg/d, fluoxetine had an intermediate effect (median reduction: 45% for binge, and 29% for purge frequency). Fluoxetine at 60 mg also improved depression, carbohydrate binges, and eating attitudes, as demonstrated by reduced scores on most subscales of the Eating Attitudes Test (EAT) and the Eating Disorder Inventory (EDI). On the higher dosage of fluoxetine, side-effects were most frequent, and there was a mean weight loss of 1.7 kg. In another large RCT, subjects receiving 60 mg/d fluoxetine showed superior improvement in binge and purge behaviours, as well as significant decreases on the EDI total score and the drive for thinness (but not the other) subscale (Goldstein et al. 1995). Treatment of BN with fluvoxamine has yielded mixed results. In one study including 103 outpatients, fluvoxamine did not prove to be efficacious relative to placebo in reducing the frequency or intensity of any aspect of bulimic symptomatology (Flament et al. 1994). In 54 female patients discharged from intensive in-patient treatment, fluvoxamine had several significant effects in preventing relapse or deterioration of eating behaviours, as measured by patients and clinician ratings; however, it had no effect on body image or eating attitudes, and nine subjects in the fluvoxamine group discontinued treatment because of side-effects (Fichter et al. 1996). Milano et al. (2004) compared sertraline to placebo in 20 patients with BN, and found sertraline effective in decreasing binging and purging symptoms.

View this table:
Table 4

SSRIs for BN in children and adolescents

One open trial included 10 adolescents treated with fluoxetine (60 mg/d) for 8 wk (Kotler et al. 2003). At treatment end, binging and purging symptoms had decreased significantly, with 70% rated as improved or much improved, and another 30% slightly improved. There were no dropouts due to adverse effects. Given the strength of evidence for fluoxetine in treating adults with BN, along with its widespread use in treating paediatric depression, most clinicians would recommend fluoxetine as part of a treatment programme for adolescents with BN.

Mood stabilizers and anticonvulsants (Table 5)

In a RCT, lithium resulted in significant improvement, compared to placebo, only in depressed bulimic patients; however, plasma concentrations of lithium were relatively low (Hsu et al. 1991). Phenytoin and carbamazepine have been anecdotally tried in BN, with no significant effects (Kaplan et al. 1983; Wermuth et al. 1977). Topiramate has been found to induce weight loss in patients with seizure disorder (Langtry et al. 2003), and has been studied for BN in two RCTs (Hoopes et al. 2003; Nickel et al. 2005). In both, it was well tolerated and associated with improvement in binge and purge behaviours, and also weight loss.

View this table:
Table 5

Opiate antagonists

Despite preliminary work indicating that i.v. opiate antagonists decreased the duration and quantity of food eaten during a binge meal (Mitchell et al. 1986), naltrexone appeared ineffective for BN in two small RCTs (Igoin-Apfelbaum & Apfelbaum, 1987; Mitchell et al. 1989). In another, Jonas & Gold (1988) noted modest improvement in 16 bulimic patients who received high doses (200–300 mg/d) of naltrexone. As noted above, naltrexone should not be used at such a high dose because it can cause significant elevations in hepatic transaminases.

Other medications (Table 6)

In one RCT, the norepinephrine dopamine reuptake inhibitor (NDRI) bupropion was found superior to placebo, but was associated with unexplained occurrence of seizures in four (5.8%) participants (Horne et al. 1988). Therefore, bupropion is contraindicated in BN. Trazodone, an antidepressant with sedative properties, was superior to placebo for reducing binge-eating frequency; only drowsiness and dizziness were reported as side-effects (Hudson et al. 1989). Ondansetron, a 5-HT3 antagonist marketed for the treatment of nausea and vomiting induced by chemotherapeutic agents and radiation treatment, was administered to 26 severely bulimic women, with reduction of binge/purge frequency compared to no change on placebo (Faris et al. 2000).

View this table:
Table 6

Pharmacotherapy of BED


Despite some evidence of efficacy, TCAs are not considered a first-line pharmacological treatment for BED, because of their side-effect profile, which includes lethality in overdose.

SSRIs (Table 7)

RCTs against placebo have been conducted in BED patients using fluvoxamine (Hudson et al. 1998), sertraline (McElroy et al. 2000), fluoxetine (Arnold et al. 2002), citalopram (McElroy et al. 2003a), and escitalopram (Guerdjikova et al. 2008). All have been associated with significant reduction in the frequency of binge-eating episodes (except escitalopram), decrease in BMI, and overall clinical improvement. Ricca et al. (2001) randomized BED patients to fluoxetine, fluvoxamine, CBT, CBT+fluoxetine, or CBT+fluvoxamine. At treatment end, both BMI and total score on the EDE were reduced in the three groups treated by CBT, with or without medication, and the group receiving fluvoxamine was superior to the other two for reduction of the EDE score. At 1-yr follow-up, the BMI was significantly higher than at treatment endpoint, but still significantly lower than at baseline in the three groups which had received CBT, while the EDE score remained unchanged in all groups. Thus, the addition of fluvoxamine seemed to have enhanced the effects of CBT on eating behaviours, and the treatment effects were partially maintained post-treatment.

View this table:
Table 7

Norepinephrine uptake inhibitors and serotonin norepinephrine reuptake inhibitors

There has been one RCT indicating that the norepinephrine reuptake inhibitor (NRI) atomoxetine (40–120 mg/d) was superior to placebo and fairly well tolerated in BED patients (McElroy et al. 2007a) (see Table 7). In one retrospective review including 35 outpatients with BED and obesity, the serotonin norepinephrine reuptake inhibitor (SNRI) venlafaxine was associated with significant reduction in binge-eating frequency (Malhotra et al. 2002).

Mood stabilizers and anticonvulsants (Table 8)

Phenytoin was the first anticonvulsant used to treat BED. However, early promising results have not been replicated, and weight gain was a common side-effect. Because they also increase appetite, valproate and carbamazepine are of little value for BED (Carter et al. 2003). Topiramate induces appetite suppression and weight loss. In a flexible dose study comparing topiramate to placebo in obese outpatients with BED, topiramate was associated with a higher reduction of weekly and daily binge frequency, and with greater weight loss; no serious adverse events were reported (McElroy et al. 2003b). A larger RCT showed that topiramate significantly reduced frequency of binge-eating days and binge-eating episodes, weight and BMI (McElroy et al. 2007b). Claudino et al. (2007) compared CBT+topiramate to CBT+placebo. In the topiramate group, patients lost more weight, and were more likely to achieve binge remission. Zonisamide, which is approved for the treatment of epilepsy in the USA, has been associated with anorexia and weight loss (Gadde et al. 2003). In obese outpatients with BED, zonisamide was superior to placebo for reduction in binge frequency and weight (McElroy et al. 2006). In many countries, lamotrigine is approved for maintenance treatment of bipolar disorder (Goodwin et al. 2004). Guerdjikova et al. (2009) evaluated its efficacy and safety in BED associated with obesity. Due to an exceptionally high placebo response, the study was incapable of detecting drug–placebo differences. However, lamotrigine was associated with a numerically greater amount of weight loss.

View this table:
Table 8

Anti-obesity medications (Table 9)

In obese patients, Stunkard et al. (1996) showed that D-fenfluramine was superior to placebo for reducing binge frequency. Shortly after, D-fenfluramine was withdrawn from the market. The SNRI sibutramine is FDA-approved for inducement and maintenance of weight loss, and its efficacy for treatment of obesity has been well established (Arterburn et al. 2004). Appolinario et al. (2003) reported significant weight loss and reduction of binge frequency in BED patients receiving sibutramine, compared to placebo. In a larger RCT, sibutramine was superior to placebo for reduction in BMI, binge frequency and binge days, and was associated with greater global clinical improvement (Wilfley et al. 2008). However, in October 2010, Abbott Laboratories announced the voluntary withdrawal of sibutramine from the US market because of an increased risk of heart attack and stroke. Orlistat, a pancreatic lipase inhibitor available over-the-counter in the USA, has modest efficacy for treatment of obesity (Padwal et al. 2003). Golay et al. (2005) randomized patients with BED to placebo or orlistat, in conjunction with dietary modification, and reported greater weight loss and a significant decrease in ED symptoms for those taking orlistat. Although anti-obesity medications have been studied for the treatment of obesity in adolescents, trials are lacking for adolescent BED.

View this table:
Table 9

Pharmacotherapy of NES

The pharmacotherapy literature for NES only includes one RCT (see Table 7), in which the SSRI sertraline demonstrated significant improvement, compared to placebo, for night eating symptoms, general illness severity, and quality of life (O'Reardon et al. 2006). Case reports have suggested beneficial effects of fluoxetine (Howell & Schenck, 2009).

First-line pharmacotherapy for EDs



A Cochrane review of antidepressants for AN (Claudino et al. 2006) has included seven RCTs. Major methodological limitations such as small sample size and large confidence intervals were found to decrease the power of the studies to detect differences between treatments, and meta-analysis of data was not possible for the majority of outcomes. Four RCTs did not find evidence that antidepressants improved weight gain, ED or associated psychopathology. Isolated findings favouring amineptine (not available in the USA) and nortriptyline emerged from antidepressant vs. antidepressant comparisons, but could not be conceived as evidence of efficacy of a specific drug or class of antidepressant in light of the negative findings from the placebo comparisons.


Several meta-analyses of medication trials for BN concluded to a significant, albeit moderate efficacy of treatment. In the Whittal et al. (1999) meta-analysis including nine medication trials, the pooled effect sizes (ESs) were 0.66 for binge frequency, 0.39 for purge frequency, 0.73 for depression, and 0.71 for eating attitudes (self-reported restraint, weight/shape concern). The Nakash-Eisikovits et al. (2002) meta-analysis (16 studies) noted that ∼50% of potential participants were excluded from research trials because of psychiatric or medical comorbidity, but that only 10% of those included dropped out due to side-effects. The average ES also was moderate (Cohen's d=0.6) for both binge and purge episodes. Overall remission rate of bingeing episodes among completers was higher after treatment with MAOIs (35%) than with SSRIs (19%) or atypical antidepressants such as trazodone or bupropion (20%). Improvement rate was higher in studies using TCAs than in those using SSRIs, MAOIs, or atypical antidepressants. In intent-to-treat analysis, the recovery rate was 25% for remission of both binge and purge episodes, 17% for remission of bingeing episodes only, and 23% for remission of purging episodes only. A Cochrane review of antidepressants for BN (Bacaltchuk & Hay, 2003) has included 19 RCTs. Efficacy results were similar for all classes of drugs. The pooled risk ratio (RR) for remission of binge episodes was 0.87 (95% CI 0.81–0.93, p<0.001) favouring medication. The number needed to treat (NNT) for a mean duration of 8 wk was 9 (95% CI 6–16). For clinical improvement (50% reduction in binge episodes), the RR was 0.63 (95% CI 0.55–0.74), and the NNT was 4 (95% CI 3–6). Patients treated with antidepressants were more likely to interrupt treatment prematurely due to adverse events, except for fluoxetine.


A meta-analysis of pharmacotherapy for BED included 14 RCTs with a total of 1279 patients (Reas & Grilo, 2008). There were no significant differences between medication and placebo for attrition. Pharmacological treatments had a moderate though clinically significant advantage over placebo for short-term remission from binge eating (48.7% vs. 28.5%), and the average weight loss was −3.42 kg (95% CI −4.25 to −2.58).

Treatment guidelines

American Psychiatric Association's guidelines (APA, 2006)

In the acute treatment of AN, weight gain is critical and often requires a combined approach including psychological support, nutritional rehabilitation, CBT and, for children and adolescents, family interventions. The APA guidelines support with substantial clinical confidence (level I) that SRRIs do not confer advantage regarding weight gain in patients receiving treatment in an organized ED programme. The guidelines also state with moderate clinical confidence (level II) that adding fluoxetine after weight restoration does not confer additional benefits with respect to preventing relapse. On the basis of individual circumstances (level III), it is suggested that second-generation antipsychotics, particularly olanzapine, risperidone and quetiapine, may be useful in patients with severe, unremitting resistance to gaining weight, severe obsessional thinking, or denial that assumes delusional proportions. Based on level II evidence, pro-motility agents such as metoclopramide may be used for bloating and abdominal pains that occur during refeeding in some AN patients.

For BN, the APA guidelines convey level I evidence that antidepressants are effective as one component of the initial treatment for most BN patients. To date, fluoxetine is the best studied SSRI and is FDA-approved for BN; sertraline is the only other SSRI that has been shown effective. Dosages of SSRIs higher than those used for depression (e.g. fluoxetine 60 mg/d) are recommended in the treatment of bulimic symptoms. TCAs and MAOIs are not recommended as initial treatments.

Regarding BED, there is level I evidence that antidepressant medications, particularly SSRIs at the high end of the recommended dose, are associated with at least short-term reduction in binge-eating behaviour but, in most cases, not with substantial weight loss. The guidelines convey level II evidence that: the appetite-suppressant sibutramine (no longer available) is effective for binge suppression, at least in the short term, and associated with significant weight loss; the anticonvulsant topiramate is effective in binge reduction and weight loss, although adverse effects may limit its clinical utility in some individuals; and adding antidepressant medication to behavioural weight control and/or CBT may confer additional benefits in weight reduction.

National Institute for Clinical Excellence (NICE) guidelines

According to the 2004 NICE guidelines (www.nice.org.uk/CG009NICEguidelines), the key priorities for treating EDs include psychological treatments (CBT for BN), and family interventions for children and adolescents. Based on expert opinion/clinical experience (Grade C evidence), it is recommended that medication not be used as the sole or primary treatment for AN. A range of drugs may be used in the treatment of comorbid conditions, but caution should be exercised given the physical vulnerability of many people with AN, and because comorbid conditions such as depressive or obsessive-compulsive features may resolve with weight gain alone. For adults with BN, a trial of an antidepressant drug is proposed as an alternative or additional first step to using an evidence-based self-help programme, based on Grade B evidence (well-conducted clinical studies). Grade C evidence suggests that SSRIs (specifically fluoxetine) are the drugs of choice in terms of acceptability, tolerability and reduction of symptoms. No drugs, other than antidepressants, are recommended for the treatment of BN. For patients with BED, there is Grade B evidence that a trial of a SSRI may be offered as an alternative or additional first step to using an evidence-based self-help programme.


The category of EDNOS represent patients with subsyndromal AN or BN who meet most but not all DSM-IV criteria. Although evidence is still lacking, both the APA and the NICE guidelines recommend that patients with EDNOS receive treatment similar to that for the corresponding diagnosis which their illness most resembles.

Evidence for maintenance treatment

Although scientific evidence is scarce, clinical experience suggests that treatment interventions shown to date to have value in the acute management of the EDs do not guarantee long-term recovery. As discussed above, the utility of maintenance treatment with the SSRI fluoxetine in AN after weight restoration is still uncertain. For persons with BN, the Cochrane review of antidepressants vs. placebo (Bacaltchuk & Hay, 2003) demonstrated that the mean clinical improvement significantly favoured medication in the pooled shorter clinical trials (6–8 wk: RR 0.63, 95% CI 0.51–0.77) but not in the longer trials (>8 wk: RR 0.53, 95% CI 0.23–1.20). As shown in Tables 4–7, only a few of the acute medication trials for BN or BED have followed participants past treatment end. In a 4-month maintenance study (Pyle et al. 1990), 47% of bulimic patients who had responded to imipramine were randomized to imipramine alone, intensive group psychotherapy plus placebo, or combined treatment. Relapse rate was high (30%), but those patients receiving group psychotherapy alone or combined with imipramine, were the least likely to relapse. Romano et al. (2002) conducted a single-blind placebo-controlled study comparing fluoxetine (60 mg/d) and placebo in preventing relapse of BN during a 52-wk period following successful fluoxetine therapy. In acute treatment responders, continued treatment with fluoxetine improved outcome and decreased the likelihood of relapse. However, attrition rate was high. No data exist on long-term effects of pharmacotherapy for BED. Carter et al. (2003) recommend that once a successful response is achieved, medication should be continued for at least 6–12 months.

Best approaches for treatment resistance

AN is a disease characterized by denial of illness, frequent treatment refusal, and non-compliance with pharmacological treatment due to fear of gaining weight (McKnight & Park, 2010). The ego-syntonic nature of the illness leads to fear of and resistance to treatment that is a core feature of the disorder. It is often aggravated by cognitive impairment resulting from a state of starvation that deprives the brain of nutrients required to synthesize adequate amounts of neurotransmitters, a factor that also contributes to the poor medication response (Crow et al. 2009). Typically, anorexic patients will initially decline nutritional rehabilitation and pharmacotherapy out of fear of weight gain and loss of control. Motivational interviewing strategies may help move them from the pre-contemplative stage to a contemplative stage where they become less resistant to treatment. Addressing the patient's malnutrition status will also improve cognition and reduce the patient's psychological resistance. For patients who have a life threatening AN, an argument can be made for coercive treatment including involuntary hospitalization and enteral feeding if the patient adamantly refuses oral feeding. For paediatric AN, treatment is generally more aggressive, and involves family-based therapy to empower parents to compassionately renourish their child.

To date, medication alone is an imperfect treatment option for BN. Nakash-Eisikovits et al. (2002) calculated the rate of post-treatment symptoms based on 10 RCTs: at treatment termination, bulimic patients still binged, on average, 4.3 times per week, and still purged 6.2 times per week. For those receiving both pharmacotherapy and psychotherapy (nine studies), the average rates were 2.5 for both binge and purge frequency. Although these data indicate substantial improvement over conditions prior to treatment, they do not constitute a return to mental health, i.e. according to DSM-IV criteria, the average patient still has bulimia, even after an adequate trial of drug or combined treatment. According to Whittal et al. (1999), the ES of combined treatment (based on three trials) was significantly better than that of medication alone (binge frequency: 1.77 vs. 0.66 per week; purge frequency: 1.33 vs. 0.39), and significantly better than the ES of CBT for binge frequency (1.77 vs. 1.28) but not purge frequency (1.33 vs. 1.22). Nakash-Eisikovits et al. (2002) demonstrated a small advantage for combined treatment over medication for binge episodes, and a moderate advantage for purging episodes (d=0.2 and 0.5, respectively). In the review by Bacaltchuk et al. (2000), remission rates were 23% for antidepressants alone vs. 42% when combined with psychotherapy (p=0.06).

For BED, Carter et al. (2003) recommend the use of a SSRI at the upper dose approved for major depressive disorder; for patients unresponsive or intolerant to SSRIs, they suggest topiramate, starting at 25 mg nightly, then increased by 25 mg weekly. Reas & Grilo (2008) reviewed eight studies, including a total of 683 BED patients, studying the effects of pharmacotherapy combined with psychotherapy. Combined treatment failed to significantly enhance binge outcomes, although specific medications (orlistat, topiramate) enhanced weight losses achieved with CBT.

Conclusions and future directions for research

Despite significant progress over the past decades, pharmacotherapy research for EDs still has a long way to go. Methodological improvements are needed, notably to include a more comprehensive evaluation of the cognitive and emotional disturbances in EDs, to obtain both objective and subjective measures, and to assess non-purging as well as purging methods of weight control. Novel pharmacological compounds should target not only dietary restraint or binge eating, but also maladaptive weight and shape-related cognitions. Large-scale well-designed trials should investigate combination of treatments or treatment augmentations, in order to target both eating-related symptoms (eating behaviours, weight) and the underlying psychological and cognitive disturbances. Future research should also investigate the moderators and mediators of outcome, so that particular mechanisms of action can be enhanced. New trials should include patients with severe comorbid conditions, as frequently seen in clinical settings, and assess long-term outcomes. More research should be done on treatment of EDs during adolescence, in order to lower the number of chronic disorders in adults. The effects of starvation or malnutrition on treatment resistance should be further investigated and corrected. Research into the factors (historical, clinical, biological) differentiating ED patients who achieve remission from those who remain symptomatic, might provide clues to elucidate which treatment(s) can best target specific components of these complex biopsychosocial disorders.



Statement of Interest



View Abstract