*2.3. Statistical Analysis*

Statistical analysis was carried out with Stata15 for Windows. The comparison of quantitative variables between the groups was based on analysis of variance adjusted for the participants' age, education level, and civil status (ANCOVA). The estimation of the effect size of the pairwise comparisons was based on Cohen's-d coefficients (|d| > 0.20 was considered low, |d| > 0.5 was considered moderate, and |d| > 0.8 was considered high) [52]. In addition, Finner's procedure (a familywise error rate stepwise method which has demonstrated more powerful than Bonferroni correction) controlled the increase in Type-I error due to multiple comparisons [53].

Linear multiple regressions stratified by sex estimated the predictive capacity of clinical measures (defined as the independent variables) on ER (defined as the criterion, DERS total score). Each regression was adjusted in five blocks/steps: (a) First block-step entered and set the covariates participants' age, education, and civil status; (b) Second block added ED-related variables (EDI-2 total, onset of the ED, and duration of the ED); (c) The third block included global psychopathological state (SCL-90R GSI); (d) The fourth block entered NSSI (0 = absent; 1 = present); and (e) The fifth block included personality traits (TCI-R scale scores). The specific predictive capacity of each step/block was measured as the increase in the *R*<sup>2</sup> coefficient (Δ*R*2).

Pathways analysis assessed the underlying mechanisms of the following study variables: Participants' sex and age, personality traits, EDI-2 total score, SCL-90-R GSI and DERS scale scores. This method constitutes an extension of multiple regression modeling, which aims to estimate the magnitude and significance of hypothesized associations in a set of variables with the advantage of allowing for the testing of mediational links (direct and indirect effects) [54]. Structural equation modeling (SEM) was used by defining the maximum-likelihood estimation of parameter estimation and testing goodness-of-fit through standard statistical measures: The root mean square error of approximation (RMSEA), Bentler's Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), and the standardized root mean square residual (SRMR). Adequate model fit was considered non-significant by χ<sup>2</sup> tests and if the following criteria were met [55]: RMSEA < 0.08, TLI > 0.9, CFI > 0.9 and SRMR < 0.1. In this study, ER was defined as a latent variable defined by DERS scale scores, and the personality profile as a latent class defined by TCI-R scale scores.

#### **3. Results**

#### *3.1. Sample Characteristics*

Table 1 includes the description and the comparison between the four groups of the study defined by ED diagnosis and sex. Differences emerged with regards to civil status, education and age.



SD: Standard deviation. \* Bold: Significant comparison (0.05 level). ED: Eating disorder; HC: Healthy control.

#### *3.2. ER and Negative Affect Measures and Comparison between Groups*

The first block of Table 2 includes the results of the ANCOVA (adjusted forage, civil status, and education) comparing the four study groups (ED-women, ED-men, HC-women, and HC-men) with regard to DERS scales, EDI-2 scales, and the binge-eating/purging levels (these two last measures were only compared between ED groups). Pairwise comparisons between ED-women and ED-men reached significance in all measures (more ER difficulties for ED-women), except for DERS awareness and the EDI-2 interpersonal distrust (no differences between the two groups were obtained). ED-women registered higher mean scores in all the measures compared to HC-women. The same occurred with ED-men compared to HC-men (except for on EDI-2 perfectionism). No differences between the two HC groups (women and men) were found.

The second block of Table 2 contains the prevalence of NSSI and the comparison between the groups (comparison between the groups was based on logistic regression adjusted by the participants' age, education, and civil status). The proportion of ED-women who reported the presence of this behavior was higher than the proportion reported by ED-men (44.2% vs. 16.1%, *p* < 0.001), as well as the proportion reported by the HC-women (44.2% vs. 21.8%, *p* < 0.001). No significant differences were found comparing the HC groups (women and men) or between ED-men and HC-men.

Figure 1 includes a radar-chart for the study variables in the four groups. To allow for easy interpretation, z-standardized means were plotted.



SD: Standard deviation; NSSI: Non-suicidal self-injury; ED: Eating disorder; HC: Healthy control; EDI-2: Eating Disorder Inventory-20; DERS: Difficulties in Emotion Regulation Scale. Bold: Significant comparison (0.05 level); † Effect size in the moderate (|d| > 0.50) to large range (|d| > 0.80); 1 Results obtained in logistic regression. — Binge and purging episodes werenotregisteredfortheHCgroup.

\* **Figure 1.** Radar-chart with the z-standardized means by group (*n* = 1082). DERS: Difficulties in Emotion Regulation Scale; ED: Eating disorders; EDI-2: Eating Disorders Inventory-2; HC: Healthy controls; TCI-R, SCL-90R: Symptom Checklist-90 Items-Revised; Temperament and Character Inventory revised.

## *3.3. Comparison of ER between ED Subtypes*

Table 3 includes the ANCOVA (also adjusted for age, education, and civil status) comparing DERS scores between the ED types (AN, BN, BED, and OSFED), stratified by sex. In the female subsample as a whole, greater ER difficulties were associated with BN, followed by BED and OSFED. The lowest DERS scores were found in AN. In the male subsample, greater ER difficulties were registered in OSFED group, followed by the BN and BED groups. AN males had the lowest DERS scores. Results obtained in the men subsample must be interpreted with caution due to the low sample size of the groups.

## *3.4. Predictive Capacity of the Study Variables on ER*

Table 4 includes the final models of the two multiple regressions measuring the predictive capacity of study variables on the DERS total score. In the ED-females model, emotion regulation difficulties were predicted by higher EDI-2 total scores, more pronounced psychopathology, higher levels in the novelty seeking and cooperativeness traits, and lower levels in the reward dependence and self-directedness traits. No significant predictive contribution of the NSSI on the DERS-total was found in the ED-females group.

For the ED-males model, DERS-total scores increased for men who reported higher scores on the EDI-2, those with higher psychopathology and lower levels in persistence.



BED: Binge eating disorder; OSFED: Other specified feeding or eating disorder.



self-injury. SCL-90R GSI: Global Severity Index of the questionnaire; Symptom Checklist-90 Items-Revised; TCI-R: Temperament and Character Inventory-Revised; B: Non-standardized B-coefficient; SE: standard error; Beta: Standardized B-coefficient; 95% CI: 95% confidence interval.

#### *3.5. Pathways Analysis*

Figure 2 includes the path-diagram with the standardized coefficients of the SEM obtained in the ED group (Table S1, supplementary material, includes the complete results valuing direct, indirect. and total effects). Goodness-of-fit was obtained (all the fit statistics were in the adequate range). The latent variable measuring ER difficulties (labeled as DERS in the figure) was directly increased for patients who presented higher ED severity (higher EDI-2 total), higher psychopathology (higher SCL-90R GSI), and who were younger. Higher scores in the latent variable measuring the personality construct (labeled as TCI-R in the figure) were also direct predictors of greater ER difficulties. ED severity and the psychopathology levels mediated the relationships between personality measures and ER, as well as between sex and ER: Higher levels in the TCI-R construct and being female increased EDI-2 interoceptive awareness and SCL-90R scores, which contributed to increases on the DERS.

**Figure 2.** SEM: Standardized coefficients (ED subsample; *n* = 718). Continuous line: Significant parameter (0.05 level). Dash line: Nonsignificant parameter. GSI: Global Severity Index.

#### **4. Discussion**

The present study attempted to address a relevant issue in the psychopathology of male patients with ED. It aimed to provide a better knowledge regarding ER in this clinical population, analyzing and comparing ER difficulties between male and female patients with ED and HC, which was rarely studied before. Findings from this study provide new information for the treatment approach of male patients with ED, a minority in the field of ED that runs the risk of being overlooked.

Our first main finding confirmed that patients with ED, both males and females, showed greater global ER difficulties than HC. Although these results are not in accordance with prior research indicating that males with ED did not differ from HC males in emotion regulation strategies, such as emotion recognition [34], they are in line with previous studies which found that negative affect and difficulties in ER predicted disordered eating in both males and females in community samples [26,27]. These discrepancies may be due to the fact that the study by Goddard et al. [34] focused on emotional recognition and not on ER. Moreover, our results support previous findings in clinical samples that

have reported decreased effective ER strategies among female patients with ED when compared with HC [10,56], suggesting that a lack of effective ER skills may prompt individuals to use disordered or abnormal eating behaviors to regulate negative affect [57], as well as contribute to body dissatisfaction and disordered eating in males [30]. Therefore, ER difficulties may act as an important etiological feature [57] or risk factor for the occurrence of EDs [23]. Although previous studies have focused primarily on females with ED, our findings also offer the possibility of generalizing these findings to males with ED.

When comparing male and female patients with ED, female patients with ED engaged in more dysfunctional ER strategies than males with ED, displaying greater scores on all DERS scales, except for DERS emotional awareness. There were no differences between male and female controls with regard to ER difficulties. First, these findings confirmed our hypothesis that both males and females with ED displayed a lack of emotional awareness. Second, the fact that female patients with ED scored higher in the most of the DERS scales, such as nonacceptance of emotional responses, limited access to emotion regulation strategies, lack of emotional clarity, and difficulties in engaging in goal-directed behavior or in controlling impulsive behaviors when experiencing negative emotions suggests that there are indeed gender-related patterns of ER in ED. However, we cannot fully exclude gender-related response bias, since males may have had a tendency to minimize or underestimate (intentionally or unintentionally) the difficulties related to their ER in order to prevent their culturally imposed, self-perceived masculinity ideals from being threatened [58]. Furthermore, males with ED appear to more often use externalizing behaviors (e.g., hetero-aggression) or to engage in drug or alcohol use/abuse to deal with emotions whereas females with ED tend to use more internalizing behaviors, such as NSSI [41]. Our results support these observations, with females with ED in the present study showing significantly more NSSI behaviors than males with ED.

Regarding ED types, our findings showed higher ER difficulties in patients with binge eating-related behaviors (BN, BED, and OSFED) compare to patients with restrictive behaviors (AN), in both males and females with ED. These results are consistent with previous studies reporting more ER difficulties among patients with binge-eating behaviors [44,59], but they are discrepant to other studies reporting less severe ER difficulties in BED patients and no significant differences between AN-R and other ED subtypes [5]. However, while females with BN showed the greatest ER deficits compared to females with other ED diagnoses, males diagnosed with OSFED were those who displayed the most ER difficulties. These differences suggest that females and males with ED engage in different disordered eating behaviors for alleviating negative affect and emotional instability. Females with ED seem to present more binge eating and purging behaviors for ER, whereas males with ED are prompt to use more heterogeneous ED-related symptomatology for alleviating emotional distress (e.g., high levels of exercise).

In terms of primary predictors, higher general psychopathology was the shared factor associated with ER difficulties in both males and females with ED. However, ED severity and different personality traits were identified as differential predictors in females and males with ED. Increased ED severity, higher novelty seeking, higher cooperativeness, lower reward dependence, and lower self-directedness were related to higher ER difficulties in females with ED, while lower persistence was associated with ER difficulties in males with ED. Thus, in females with ED, difficulties in ER were associated with a tendency to be more impulsive and intolerant of routine, and which are linked with seeking little emotional support, the unwillingness to be sociable, and having difficulty in expressing feelings and thoughts [60]. On the other hand, in males with ED, difficulties with ER were associated with low persistence, that is, a tendency to be less perseverant in situations of frustration and fatigue [60]. In light of our results, our findings suggest that personality differences may impact ER difficulties, therefore, it would be important to assess for personality traits and consider potential gender-related differences [61,62]. In this regard, it may also be useful to apply ER-based adjuvant treatments focused on reducing impulsivity and increasing self-directedness and reward dependence for females with ED,

and specific treatment approaches for males with ED where increased persistence management are specifically addressed.

Finally, another emergent finding was that both ED severity and general psychopathology mediated the relationships between personality and ER difficulties. This may open a new line of research that allows for knowing if the improvement in the ED symptomatology could establish changes in emotion dysregulation. In this sense, a previous study in females with ED found ER improvements after CBT (treatment as usual, without any specific module addressing ER), especially in patients with BN. This study found that improvements in ER were the largest in those with a better treatment outcome [23]. In this line, our results reinforce this concept, suggesting that ED severity and psychopathology may be associated with ER difficulties. In addition, although our study is transversal and does not allow us to analyze the causality, we suggest the existence of a bidirectional pathological process that has ER difficulties acting as a maintenance factors for the ED. However, these findings do not exclude the possibility that ER is also a vulnerability factor for ED. The lack of longitudinal studies analyzing individuals before developing the ED does not allow us to identify if the ER is an etiopathogenic factor of the disorder or if, on the contrary, difficulties in ER are aggravated with the ED. It is most likely that ER is probably acting in both directions, both as a vulnerability factor and as a maintenance factor for the disorder (which is aggravated by psychopathology). With this in mind, we hypothesize that treatment enhanced with a module aimed at improving ER skills could benefit the treatment outcome of ED patients. Further studies should address this point.

Also, the results suggest that, a more dysfunctional personality profile and being female increased the risk of higher ED severity and general psychopathology, which contributed to an increase in ER difficulties in patients with ED. In this vein, a recent study found that depression moderated the association between ER difficulties and binge eating in patients with BED, suggesting that individuals who experience more intense emotions are more affected by difficulties in ER [8]. Again, the above is consistent with the need for treatment based on addressing the difficulties of ER in ED patients, since, although being aware of one's own emotions is not sufficient for an adaptive emotional regulation [44], it is the first step to improving it.
