**4. Discussion**

The aim of the present study was to explore treatment responses in the different FA profiles identified by [18], considering only the ED sample. Clinical characteristics of these ED-focused clusters are similar to those previously found and were relevant for treatment outcome as well. As we hypothesized, the functional cluster (C3), do present better treatment response and lower dropout rates than the moderate (C2) and dysfunctional (C1) clusters. Several aspects of these results must be highlighted.

First, as in the prior study [18], FA levels were higher in the moderate cluster, followed by the dysfunctional one, and lower in the functional. The composition of each cluster regarding the diagnosis of the patients was maintained for the dysfunctional and moderate clusters, however, the composition of the functional one changed. In the prior study, this subgroup was highly represented by patients with obesity but no ED, while in this study non-ED participants were excluded. However, the clinical characteristics of the present and previous clusters were similar. This is, dysfunctional cluster (C1) had higher presence of BN and OSFED patients, higher severity of the disorder and worst psychopathological state, as well as low self-directedness and high harm avoidance. The functional cluster (C3) had more equilibrated proportion of diagnosis subtypes, with BN being more prevalent, and higher self-directedness and persistence, with lower levels of harm-avoidance. Finally, the moderate cluster (C2) had a heightened presence of BN (72.5%) followed by BED (24.6%), therefore, this cluster was particularly represented by binge ED subtypes; as well, this cluster had the highest levels of FA as in the first study.

Thus, what differentiates the dysfunctional cluster (C1) from the other clusters is the severity of it clinical characteristics (except FA), while the moderate (C2) group differs from the functional (C3) and dysfunctional cluster (C1) by a higher severity of FA, and the functional cluster (C3) differ from the dysfunctional (C1) and moderate (C2) one by the low severity of its clinical profile.

Treatment outcome was explored as well, not only to relate it with the presence of FA, being that other studies have already approach the subject [14,32], but to add to a better characterization of FA construct in well-defined phenotypes that consider FA presence and other clinical variables.

Low levels of full remission and higher rates of dropouts in the dysfunctional cluster (C1) were found. This subgroup was highly represented for OSFED patients, which have been reported to present low harm avoidance and self-directedness, as well as higher severity of ED symptomatology, aspects identified as predictors of high drop-outs and low full remission rates [28]. Additionally, similar personality traits that imply difficulties in following goals and higher levels of anxiety levels have been found in BN patients (also present in this cluster) [33]. This has also been associated with low levels of full remission after cognitive behavioural treatment (CBT) [26]. Therefore, patients within this cluster may benefit from treatments that target the reduction of the ED symptomatology and general distress, as well as favour the improvement in the establishment and following of objectives. It is also important to mention that younger patients with an earlier onset of the disorder were particularly present in this cluster; therefore, these aspects could be added as indicators of a more dysfunctional profile. Of note, early onset of the disorder has already been mentioned as a predictor of a longer maintenance of the ED [34].

In the moderate cluster (C2), the highest dropout rates were found, as well medium rates of full remission in comparison with the dysfunctional (C1) and functional clusters (C3). This cluster was characterized by the presence of binge spectrum ED patients and by the higher levels of FA, both aspects that could be involved in the response to treatment of the participants in this cluster. It is possible that the higher presence of FA symptomology in binge spectrum ED (relative to non-binge ED) may reflect the more frequent binge eating episodes and food craving associated with FA [35,36]. In the same line, it could be hypostatized that the high levels of FA could be related to the higher drop out percentage found here. This is consistent with prior studies that have found that FA predicts worse intervention response in BN patients [32], and that FA can act as a mediator between severity of ED in BN and BED patients and treatment outcome [14]. However, being that this cluster did not present levels of psychopathology and severity of the ED as high as the dysfunctional cluster (C1), this moderate cluster (C2) was more likely to reach full remission than the patients in C1. Therefore, it may be important to screen for severe FA (particularly for patients with diagnoses more represented in this cluster), to implement approaches that could reduce FA symptomatology. In this regard, several authors have suggested additional therapy aims of craving managements and increasing inhibitory control [37], and psychoeducation about the dietary patterns implicated in addictive eating [35,38]. As well, BED patients and FA patients present high levels of impulsivity, therapies aimed to reduce food related impulsivity could be implemented as well [39].

Finally, the functional cluster (C3) presents the higher levels of full remission and the lowest of drop out, as well as had the lowest levels of severity of FA. Further, this cluster is clinically speaking the most functional, presenting low ED severity and general psychopathology. It also had the highest levels of self-directedness and persistence, which may be associated with good compliance with the treatment. Patients within this subgroup may respond best to traditional CBT treatment and not need additive therapies to help them succeed. This groups overall FA severity level is low and other studies have found that FA symptoms can remit after traditional CBT [32]. Thus, specific targets for addictive mechanisms may not be needed for this cluster.
