3.2.1. Methodological Issues

The reviews and meta-analyses selected in this overview reviewed from six [51] to 66 studies [45] (see Table 3). This wide range results from numerous factors (aim of the review; types of study considered: ranging from only RCTs to all types of study; period; methodology for selection of studies: ranging from all published studies to a selection based on quality guidelines; type and number of databases used). For example, Brockmeyer et al. used the Cochrane Handbook and the National Institute of Heath Criteria for a quality assessment of controlled intervention studies [51]

between October 2011 and the end of 2016 and selected only three good quality RCTs on psychotropic medication. Conversely, Franck et al. [45] were exhaustive and included all studies (except case reports) published in the past 50 years. Miniati et al. [48] underlined the impossibility of applying quality criteria to select studies in their review. In addition, samples were very different in terms of age (see Table 3), some focused on adolescents, others on adults, and some on both.

Psychotropic medications have been used since the 1960s to treat AN. For the studies included in the reviews, Miniati et al [48] underlined the paucity of reports, including the small number of RCTs, the small numbers of subjects included in the studies (one to maximum 93 in this review), the heterogeneity of sample composition, treatments, and treatment settings, and the small number of males to enable comparisons based on gender, which are all aspects that impact the conclusions.

The variable settings (inpatient/outpatient) had an impact on patient characteristics; for example in the studies, these authors observed that the age of inpatients was significantly younger than that of outpatients [48]. In addition, in cases of positive results in favour of a medication, the results had never been replicated. The conclusions of these studies were based on small samples and the fact that they derived no significant results could be partly explained by a lack of power; the results were often extrapolated across patient groups of different ages, illness durations, and severity. Another important limitation of these studies is their duration [48]. Most of the studies were based on short term follow-up in an illness requiring long-term management [48]. The majority of the studies were conducted on inpatient samples, whereas the majority of patients are treated on an ambulatory basis [48]. Not all symptom dimensions of AN, nor all the comorbidities, nor the impact of the medication on these elements were systematically reported.

A small number of studies specially focused on children and adolescents. The majority of the studies were conducted in the acute phase of AN and a minority in the maintenance phase [40]. Below we provide information in terms of dose, duration of treatment, and sample size, for comparative studies only.

#### 3.2.2. Antidepressants (AD)

The rationale for treating AN with antidepressants was initially that AN and depression had clinical and biological similarities, including comorbidity and symptom overlap with anxiety disorders, obsessive compulsive disorders and depression, and a hypothetical dysfunction in the serotonergic and noradrenergic systems in the pathophysiology of AN. The earliest studies mostly concerned tricyclics and monoamine oxydase inhibitors and the more recent mostly concerned selective serotonin reuptake inhibitors (SSRI) [6,37,45,48]. We found one small meta-analysis [41] pooling two studies on tricyclics [54,55], two on selective serotonin reuptake inhibitors [56,57], and five reviews concerning each different class of antidepressant separately [6,37,40,45,48]. No other recent study was found.
