Meta-analyses

The meta-analyses (see Table 3 for details) concerned all antidepressant RCTs versus placebo evaluating the impact on either weight restoration or maintenance, with various doses, durations and evaluation criteria. It pooled four RCTs. One was on clomipramine (50 mg, duration not mentioned, there were eight placebo subjects and there were eight AD subjects, evaluation criterion: weight gain in kg) [54], one was on amitriptyline (160 mg max, for 32–45 days, there were 25 placebo subjects, and 24 AD subjects, evaluation criterion: weight gain per day) [55]. Two were on fluoxetine [56,57] (respectively 60 mg and 20 to 60mg, for 7 weeks and 1 year, 16 placebo and 44 subjects, 15 AD and 49 subjects, evaluation criteria, respectively: ideal body weight and body mass index) involving 96 patients in the antidepressant groups and 94 patients in the placebo groups. No impact of antidepressants on weight gain or maintenance in AN was found [41]; because of the small sample sizes no meta regression or subgroup analyses were conducted.

#### Tricyclics

On the basis of an analysis of three RCTs versus placebo [45,48], there is no clear evidence for the general use of tricyclics among patients with AN in terms of weight gain and depressive symptoms, according to two studies on amitryptiline (160 mg max and 2.8 ± 0.3mg/kg, respectively for 32–45 days and five weeks, placebo 25 subjects and 11, AD 24 and 11 subjects, evaluation criteria: weight gain per day and weight plus psychological outcome) [58] and one with clomipramine (50 mg, 76 days clomipramine and 72 days placebo, placebo eight and AD eight subjects, evaluation criterion: weight gain in kg) [54,59]. These drugs only showed some impact on hunger, appetite, and energy intake at the beginning of treatment, but with no effect on weight, as mentioned previously. Although clomipramine tended in one study [54] to be associated with lower weight gain compared to the placebo, this was hypothesised to be linked to more physical activity. An open trial comparing paroxetine to clomipramine (respectively 18.4 ± 4.7 mg and 75.3 ± 7.6 mg, duration 39 ± 26 and 58 ± 30, 39 and AD 57 subjects, evaluation criteria: BMI, duration to weight gain) found no difference in terms of weight gain in an adolescent sample, but duration to obtain weight gain was significantly shorter for paroxetine (72 versus 97 days) [45,60]. In addition, because of the lethal risk with overdose in suicide attempts and the potential for fatal arrhythmia at low body weight, particularly among young subjects, this type of medication is no longer studied and is not recommended today for AN [6,37,40,45,48].

#### Selective Serotonin Reuptake Inhibitors (SSRIs)

There is today no clear evidence supporting the use of SSRIs in AN [48].

The three RCTs [56,57,61] on fluoxetine versus placebo provided controversial results (respectively 60 mg for the first and for the others 20 to 60mg, for seven weeks, one year, and 52 weeks, placebo 16, 44, and 16 subjects AD 15, 49, and 19 subjects). Most RCTs with placebo (2/3) reported negative results concerning the effect on eating psychopathology and weight gain [56] and on weight maintenance on a large sample at 12 months [45,48,57]. The only one that had positive results for the maintenance of weight at 12 months and for anxiety [61] had only 13 completers for the treatment and no associated psychological treatment [48]. Some authors hypothesised that fluoxetine inefficacy could be linked to malnutrition, and in particular, to a lack of dietary tryptophan. Tryptophan supplementation in addition to fluoxetine was evaluated in a double-blind controlled trial versus placebo (fluoxetine 20 to 60 mg, for six months, AD and placebo 11 subjects, AD with nutritional supplementation 15 subjects). Barbarich, et al. [62], did not show benefits from this supplementation on weight gain, anxiety or obsessive compulsive symptoms [45]. Three RCTs compared fluoxetine to nortriptyline, amineptine [63,64] (respectively for fluoxetine 60 mg, versus nortriptyline 75 mg, amineptine 300 mg for 16 weeks, 22 subjects and 13 subjects) or clomipramine / amisulpride [65] (respectively for fluoxetine 28 mg, clomipramine 58 mg/ amisulpride 50 mg for 12 weeks, 35 subjects), and failed to demonstrate any difference [40].

The effect of citalopram was only investigated in three open trials [66–68]. The authors of the third study with a control group (citalopram 20 mg, for 12 weeks, 19 subjects compared to 20 patients on a waiting list) found no improvement in terms of weight, but an improvement in depression, obsessive-compulsive symptoms, impulsiveness and trait-anger [48], and body dissatisfaction [45].

A small open-label study that compared sertraline with the placebo [69] reported that sertraline improved depressive symptoms, perceptions of ineffectiveness, a lack of interoceptive awareness, and perfectionism, but not weight gain [45]

#### Other Antidepressants

Monoamine oxydase inhibitors were evaluated in an open study among six patients for six weeks [70]. Mood and anxiety improved [45] but these drugs are no longer used, as a result of both their inefficacy on weight and their unfavourable side effect profile [48]. Venlaxine has only been studied in an open trial [71], in comparison with fluoxetine and in association with cognitive behavioural therapy, with no differences in weight or behaviour outcomes [45]. Mirtazapine and duloxetine used in AN have only been described in case reports on adults [48]. Concerning adolescents, mirtazapine was not found to be superior to other medications nor to no medication in AN [45].

In conclusion, antidepressants have no impact on weight gain [45,48], and their impact on eating symptoms or psychopathology is not clear. Tricyclics and monoamine oxydase inhibitors have adverse side effect profiles and should no longer be used in AN [45,48].

#### Antipsychotics

The rationale for treating AN with antipsychotics was initially linked to the hypothesis that obsessions regarding weight and body shape in AN (abnormal beliefs that are ego-syntonic and characterised by an acute lack of insight that persists even when the affected person's health status is endangered) could be viewed as delusional ideas and consequently could result from dopamine receptor hypersensitivity in AN [45]. Other arguments for their use in AN were based on the effect of these drugs on reward system regulation, their potential efficacy in controlling problematic frequency of physical activity (usually called hyperactivity in AN) and the weight gain side effects observed in other disorders [48]. Another recent argument is that second generation drugs also act on the salience network, known to be impaired in eating disorders, and that they could thus act by enhancing the reactivity of the anterior cingulate cortex and the salience network in the response to the reward value of food in AN. These drugs are mainly dopamine D2 and serotonin 5HT2A antagonists [71]. Finally, more recently, Frieling et al. [72] cited by Miniati et al. [48], have postulated the existence of an altered expression of the dopaminergic genes among patients exhibiting psychomotor hyperactivity. These drugs are dopamine D2 receptor antagonists with severe adverse effects. For second-generation drugs, the arguments also include positive effects on depression and anxiety symptoms arising from eating disorders [48]. First-generation antipsychotics (typical antipsychotics) and second-generation (atypical) antipsychotics have all been studied.

Four meta-analyses [27,38,41,43] and six reviews concerned each class of antipsychotic separately [6,37,40,45,48,51]. The reviews are considered for first-generation drugs, as they are mainly not included in meta-analyses, and also for second-generation drugs after 2012 (the most recent data in the meta-analyses). Since the end of the last review [48] five study reports have been published about antipsychotics in AN, two retrospective chart reviews on adults and adolescents with AN respectively [45] not reviewed here, one open-label study among adolescents [73] and one among adults [74] not reviewed here, and two RCTs [75,76]. Details concerning antipsychotic studies are presented in Table 4.



#### Meta-Analyses

The meta-analyses compared antipsychotics, mainly second-generation, to placebo in six to eight studies, but never exactly the same drugs, always with a majority of studies on olanzapine. 2/4 pooled first and second-generation antipsychotics [27,41], the two others only included second-generation drugs [38,43]. They concerned respectively:





No impact of antipsychotics was found in comparison to placebo on weight gain, whether antipsychotics were pooled or considered individually (evaluated in 4/4 meta-analyses), nor on eating disorder symptoms (evaluated in 3/4 meta-analyses [38]), nor was it found for akathisia [27], dropout [27], or glucose levels [27]. A recent paper by Attia et al. [76] reported an RCT on olanzapine increased to the maximum of 10 mg/day in four weeks compared to placebo. This study was conducted on a large sample of 152 adults AN (83 completers) with a long illness duration (mean age for placebo group 28 ± 10.9 and for olanzapine group 30.0 ± 11.0 years old and illness duration respectively 10.5 ± 9.5 and 12.6 ± 11.7). It showed a modest therapeutic effect of four months of olanzapine, compared with placebo on BMI (with an increase difference of 0.165 BMI points more per month). This study found no difference between groups in terms of clinical global impressions, obsessionality, anxiety and depressive symptoms, nor eating disorders symptoms (except an increase in shape concerns that was observed in the olanzapine group).

Drowsiness/sedation occurred significantly more often with antipsychotics than with placebo/usual care in the pooled analyses, and especially for olanzapine in an individual analysis [27]. Attia et al [76] found no significant differences in the frequency of the abnormal blood test to assess metabolic abnormalities between the olanzapine and placebo groups.

For the effect on anxiety and depressive symptoms, there was no apparent efficacy of antipsychotics according to the largest meta-analysis on the topic (pooling four studies) [27]. A recent small RCT study lasting three months among 30 adult outpatients with AN mentioned a superiority of Olanzapine (2.5 mg the first month and 5 mg the two following months) combined with cognitive behavioural therapy (CBT) versus placebo combined with CBT in improving obsessiveness-compulsivity, depression, anxiety and especially hostility (but not weight gain or specific aspects related to the AN eating pathology) without showing results [75].

Typical Antipsychotics other than those Explored in the Meta-Analyses

Haloperidol and chlorpromazine have generally not been evaluated in RCTs, and evidence levels for their efficacy and safety are poor [48]. Haloperidol as an adjunct to psychotherapy was found to be associated with weight over six months in one open study [45,48,79]. In an RCT, there was no difference in terms of weight gain between chlorpromazine and olanzapine [80] but olanzapine was superior for the reduction of anorexic ruminations [37].

Other Second-Generation Antipsychotics than those Explored in the Meta-Analyses

Amisulpride was studied in one study [45]. It was found to be superior to fluoxetine and clomipramine in a single blind RCT for weight gain (not for weight phobia, body image disturbance or amenorrhea) [65].

Aripiprazole was considered only in a case series of adults and young people [48].

In conclusion, according to the four meta-analyses considered here, antipsychotics had no impact on weight gain compared to placebo, and their impact on eating symptoms and psychopathology is not clear. Olanzapine had a modest effect on weight gain in one RCT including adults with a long duration of AN [76]. Typical antipsychotics have an adverse side effect profile and should not be used in AN, except perhaps haloperidol in severe AN [45]. Atypical antipsychotics could alleviate some symptoms, such anorexic ruminations, anxiety or depressive symptoms, but levels of evidence are low, and based only some small studies [75] and not supported by meta-analyses on the topic [27].

#### 3.2.3. Lithium

The rationale for using lithium among patients with AN is mainly related to the fact that it induces weight gain in other disorders [37,48].

We found four reviews concerning lithium [6,37,48] all mentioning one RCT [89], but we found no other recent study.

This small RCT, involving lithium among adults (dose not mentioned, 16 patients and controls, evaluation criterion: weight gain at weeks three and four) showed a significant difference in terms of weight for the lithium-treated group, but not for other psychological dimensions [89]. However, the use of lithium is not recommended in AN, even for patients with severe and resistant forms. Sodium and fluid depletion are frequent in AN and could reduce lithium clearance, which could lead to lithium poisoning [45,48], as renal complications are frequent in AN [90].

#### 3.2.4. Appetite Enhancers

#### Antihistamines

Four reviews [6,37,45,48] mentioned antihistamines but we did not find any more recent studies. Cyproheptadine (CYP), a serotonin and histamine antagonist reputed to produce weight gain among children with asthma, was tested in two RCTs [37,48]. The first RCT in four arms compared Placebo and CPY both with and without cognitive behavioural therapy (12 mg to 32 mg maximum, duration not mentioned, 81 subjects in four groups: CYP and behavioural therapy or placebo, placebo and behavioural therapy, placebo; evaluation criterion: weight gain) [91]. The second compared placebo, CYP and amitriptyline (CPY 32 mg maximum, amitriptyline 160 mg maximum, duration: to 5% of target weight, 72 subjects in three groups, evaluation criterion: duration to 5% target weight) [92]. There was no clinically significant effect on weight gain with CYP in the first RCT. In the second, cyproheptadine marginally decreased the length of time to reach the weight gain objective for restricting AN patients, but it significantly impaired treatment efficacy for the binging/purging anorectic patients, compared to amitriptyline and placebo [45,48]. Thus antihistamines remain non-indicated in AN and should be avoided. No further study was found.

#### Opiates

One review [37] mentioned a study about cannabinoids and we did not find any more recent studies.

The opioid peptide system has been implicated in appetite and feeding regulation, linked to the hedonic value of food in both animals and humans. It has been hypothesised that both anorexia nervosa and bulimia nervosa could be opioid-mediated addictions [37]. In line with these hypotheses opiates and opiate antagonists have been used in order to stimulate eating in AN or to deactivate the suspected auto-addictive properties of food restriction. No study has been conducted in AN exclusively to test this hypothesis. Only one review by Aigner et al. [6] mentions one RCT that was conducted in a mixed sample of 19 AN binging/purging subtype and bulimia nervosa patients using naltrexone, 100 mg for 6 weeks [93]. Bingeing and purging behaviours decreased among both AN and bulimia nervosa patients. Naltrexone has no indication in AN.

#### Cannabinoids

Given the well-known impact of cannabis on appetite, cannabinoids have also been used and tested in AN. [45,51].

Two reviews mentioned cannabinoids, but we did not find any more recent studies [45,51].

Two RCTs evaluated cannabinoids. An earlier four-week double-blind crossover study [45] compared delta-9-terahydrocannabiol (delta-9-THC) to diazepam (delta-9-THC, 7.5 to 30 mg versus diazepam 1 to 15 mg, for four weeks, 11 subjects, evaluation criteria: weight gain, daily calorie intake), and showed no benefit of delta-9-THC [94]. In addition, three patients experienced severe dysphoric reactions under 9-Tetrahydrocannabinol administration. The second four-week crossover RCT (5 mg, 4 weeks, 24 subjects, evaluation criteria: weight gain, daily calorie intake), [95] compared low doses of dronabinol (a synthetic form of delta-9- tetrahydrocannabinol) to a placebo and observed a small significant gain of 0.73 kg for dronabinol. It had no impact on the total duration of physical activity but increased the average intensity of this physical activity [51].

In conclusion, cannabinoids have no proof of their efficacy nor of their safety in AN, they need further evaluation in AN, and should not be used in routine care practice.
