Participants
This cross-sectional study was performed from August 2018 to November 2019. In this study, 110 patients with BPD and 110 healthy people were selected. The sample size was calculated based on the study performed by Eizaguirre et al. [16]. With the assumption of r = 0.3, the probabilities of type I error and type II error were obtained as (α = 0.05 and Z1- α/2 = 1.96) and (β = 0.2 and Z1- α/2 = 0.84), respectively. Also, the value of the correlation coefficient approximated to the normal distribution, and the value of r in normal distribution was calculated by the formula C = 0.5 × ln [(l + r) / (l ̶ r)]. Finally, the total sample size of 220 people was obtained based on the formula N = [(Ζα + Ζβ) ÷ C] 2 + 3 [20]. The patients with BPD were selected by systematic random sampling among the persons who referred to Baharan psychiatric hospital in Zahedan, Iran, with the sampling interval of 3. Also, the healthy people were selected out of their relatives through one-to-n matching (i.e., one case dedicated to one control). The inclusion criteria were as follows: (1) getting a score above 10 in borderline personality inventory (BPI) and approved diagnosis of the disorder based on structured clinical interview for DSM-5 personality disorders (SCID-5-PD) by a psychiatrist; (2) age range of 18-35 years; (3) ability to read and write alongside reading comprehension; (4) for healthy people, getting a score of ≤22 in general health questionnaire (GHQ-28) and approved mental health based on structured clinical interviews for DSM-5: research version (SCID-5-RV) by the psychiatrist. The exclusion criteria included the followings: (1) severe and acute physical illness; (2) brain traumatic injury; (3) comorbidity of schizophrenia and other psychotic disorders; (4) epileptic disorder; (5) intellectual disability; (6) mixed personality disorder; (7) taking any drug or substance that causes anorexia and bulimia; (8) failing to fill the questionnaires properly.
In this study, 220 participants were evaluated in three groups, namely BPD (n = 38), BPD+FEDs (n = 72), and healthy control (n = 110). Table 1 presents the participants’ sociodemographic information. According to the table, there was no significant difference in sociodemographic factors between the study groups.
Procedures
After the approval of the research project by the ethics committee of the Medical Faculty of the ZAUMS Zahedan (IR.ZAUMS.REC.1398.212), the informed consent forms were distributed among the participants. In order to follow the Declaration of Helsinki, participation in the study was optional, and the participants could leave the study for any reason. After obtaining informed consent from the participants, EAT-26, TAS-20, BAI, and BDI-II were distributed among them. Next, the psychiatrist evaluated all of the patients with BPD (with or without a score ≥20 in EAT-26) using SCID-5-RV to identify different types of FEDs. The questionnaires were anonymous to keep the participants’ information private.
Measures
The following measures were used in this study (in general, the Cronbach’s alpha values of 0.70 or higher indicate acceptable internal consistency) [21]:
TAS-20
Alexithymia was assessed with the Persian version of the TAS-20, a 20-item self-report questionnaire scored based on a five-point (1-5) Likert scale. The minimum and maximum scores are 20 and 100, respectively, and scores of ≥ 61 represent alexithymia [22]. In our study, the Cronbach’s alpha coefficients for the TAS-20 subscales of difficulty identifying feelings, difficulty describing feelings, and externally oriented thinking were 0.85, 0.82, and 0.75, respectively, while it was 0.72 for the total sample.
BAI
Anxiety symptoms were assessed with the Persian version of the BAI, a self-report 21-item questionnaire scored based on a four-point (0-3) Likert scale. The minimum and maximum scores are 0 and 63, respectively [23]. In our study, the Cronbach’s alpha coefficient for the BAI was 0.90.
BDI-II
Depressive symptoms were assessed with the Persian version of the BDI-II, a self-report 21-item questionnaire scored based on a four-point (0-3) Likert scale. The minimum and maximum scores are 0 and 63, respectively [24]. In our study, the Cronbach’s alpha coefficient for the BDI-II was 0.88.
EAT-26
Eating-disordered behaviors were assessed with the Persian version of the EAT-26. In this 26-item questionnaire, the minimum and maximum scores are equal to 0 and 78, respectively. A score above 20 stands for the probability of being affected by FEDs [25]. In our study, the Cronbach’s alpha coefficients for the EAT-26 subscales were as follows: “drive for thinness” = 0.90, “restrained eating” = 0.77, “perceived social pressure to eat” = 0.87, “food preoccupation and oral control” = 0.75, and “bulimia” = 0.71.
BPI
In this 53-item questionnaire (answered by yes or no), if the person’s score for the 20 items of the cutoff score is above 10, the person is most likely to be affected by BPD [26]. In our study, the Cronbach’s alpha coefficients for the BPI subscales of identity diffusion, primitive defense mechanisms, reality testing, and fear of closeness were 0.78, 0.82, 0.74, and 0.77, respectively, while it was 0.80 for the total sample.
SCID-5-PD
SCID-5-PD is a semi-structured clinical interview used by researchers and clinicians, which evaluates DSM-5 personality disorders under three clusters of A, B, and C, and other specific personality disorders. Several studies have reported acceptable reliability and validity of SCID-5-PD [27].
SCID-5-RV
SCID-5-RV is a semi-structured interview for major DSM-5 diagnoses, which is performed by a trained clinician or health expert familiar with the diagnostic criteria and classification of disorders in DSM-5. Several studies have reported acceptable reliability and validity of SCID-5-RV [28].
GHQ-28
GHQ-28 is a 28-item questionnaire in which items are scored in the range of 0-3. The overall score ranges between 0 and 84. A score of ≤ 22 indicates a person’s mental health [29]. In our study, the Cronbach’s alpha coefficients for the GHQ-28 subscales of somatic symptoms, anxiety and insomnia, social dysfunction, and severe depression were 0.76, 0.84, 0.71, and 0.88, respectively, while it was 0.90 for the total sample.
Data analysis
Statistical analysis was performed using descriptive statistics, including mean and standard deviation. Chi square test and Kruskal-Wallis test were conducted for a sociodemographic comparison of the three study groups. Also, an analysis of variance (ANOVA) was used to compare the mean scores of BAI, BDI, TAS-20, and EAT-26. In ANOVA, the Scheffé test was applied to the post-hoc analysis. Subsequently, in both BPD only and BPD+FEDs groups, the Pearson correlation coefficient was used to evaluate the correlation between the study variables. Also, in these groups, the mediation analysis was conducted to investigate the mediating effect of anxiety and depression on the relationship between alexithymia and eating-disordered behaviors using the Hayes’ PROCESS macro for SPSS [30]. As outlined in Preacher and Hayes [31], mediation emerges when the indirect effect is significant and the confidence intervals do not contain zero. Furthermore, given the relationship of sociodemographic factors (including age, gender, marital status, education level, and income) with anxiety, depression [32-35], and FEDs [36, 37] obtained in previous studies, the above-mentioned factors were considered as covariates in the mediation analysis. Meanwhile, data analysis was performed by SPSS 25, and the significance level was set at P ˂ 0.05.