The present study is a two-branches study: one is a short (6-months) and medium (12-months) term therapeutic outcome study, the other a 15-years long-term follow-up study.
It was considered for recruitment in the study a sample of 149 female adolescent outpatients from the Child Neuropsychiatry Outpatient Service for the Eating Disorders the Regina Margherita Hospital, Department of Public Health and Pediatric Sciences of the University of Turin, first evaluated between January 1st, 2003 and December 31st, 2005 (T0 time of this study) who received a full diagnosis of AN (Restricter – ANR or Bingeing-purging –ANBP) or Eating Disorder Not Otherwise Specified (EDNOS Anorexia Nervosa type) according to diagnostic criteria of DSM IV and IV-TR. The diagnosis was established by a child neuropsychiatrist during the first examination at the intake in the center (T0) using the Structured Clinical Interview for Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition (SCID-I) (48). It was revised by another expert child neuropsychiatrist at the follow-up 15 years later (T3) using the SCID-5-CV for DSM 5 criteria (49), thus the study was conducted according to the DSM 5 criteria.
Inclusion criteria for the study were: 1) female participants enrolled for a 1-year treatment at outpatient service for Eating Disorders; 2) participants filled in the psychopathology and personality measures at intake in the therapeutic program; 3) absence of documented intellectual disability or neurodevelopmental disorders at intake defined according to the DSM criteria; 4) absence of a psychosis or bipolar disorder at any time-point of the study.
Among 149 eligible subjects, 73 fulfilled all the required issues for inclusion in the study: 34 did not complete the assessment at the enrollment in the study, 13 received a psychosis, pervasive development disorder or bipolar disorder diagnosis during assessment phase, and in 29 cases the results of the assessment collected for the study were lacking or incomplete. The 73 participants were contacted to be enrolled in the long-term follow-up research study. Ten subjects were out of reach. Among the other 64 participants, 21 calls contacted their parents who did not release the phone number of their daughters to protect them. Among the residual 43 participants, 16 refused to take part in the study for personal reasons, and 8 agreed with participation in the study but they did not give back the package of tests or filled them incorrectly. Finally, 19 subjects with AN participated in the 15-year follow-up study. DSM 5 criteria were applied to assess diagnostic outcome at follow-up, using the SCID-5-CV applied by an expert child neuropsychiatrist (FA author).
Ethical issues
All participants and their parents gave their written informed consent for participation in clinical research at intake in the outpatient service. All contacted participants who agreed for participation in the study gave their written informed consent at follow-up. The Intercompany Review Board of Torino (CEI) approved this study with the protocol number 36931.
Treatment approach
All selected participants were voluntarily enrolled for a 1-year treatment at the outpatient service for Eating Disorders of the Regina Margherita Hospital in Turin. The treatment program was multimodal and multidisciplinary and included: a first assessment visit conducted by an experienced child psychiatrist assisted by a specializing doctor, self-administration of assessment instruments at home, a second assessment visit in which the results of the tests, modality and timing of the program were explained to the participant and to her parents and possible drug prescription (antidepressant or anxiolytic). Neuropsychiatric control visits once/month. First joint visit by a dietician and a dietitian with a diet-therapy prescription followed by diet-therapy follow up once/month. Assignation to two cycles of time-limited, brief (20 sessions) Adlerian Psychodynamic Psychotherapy (B-APP) once/week according to the manualized model of Fassino and coworkers (50).
2.1 Outcome Measures
ED diagnosis according to DSM 5 (Anorexia Nervosa Restricting type - ANR, Anorexia Nervosa Bingeing-purging type - ANBP, Other Specified Feeding or Eating Disorders - OSFED) and clinical data (height, weight, BMI, bingeing/purging per week, hours of physical activity per week, and minor psychiatric symptoms) of participants were collected at their first access into the center (T0) and at the time points of clinical outcome and follow-up. In consideration of their relevance for the study, of the diagnostic composition of the final sample, and data missing they were considered as clinical outcome measures ED diagnosis and BMI.
The time points were defined at intake (T0), at 6 months after the beginning of the treatment (T1), at 12 months corresponding to the end of the treatment program (T2), and at 15-years follow-up (T3).
At T0 and T2 participants were assessed with a battery of psychometric tests including: Temperament and Character Inventory (TCI) (51), Eating Disorder Inventory-2 (EDI-2) (52), Young Self-report Questionnaire (YSR); while their parents filled in the Child Behavior Checklist (CBCL).
At T1 only the Eating Disorder Inventory-2 (EDI-2), Young Self-report Questionnaire (YSR), and the Child Behavior Checklist (CBCL) were administered because the changes in personality traits within six months were considered unstable (51).
At T3 the Temperament and Character Inventory (TCI) and the Eating Disorder Inventory-2 (EDI-2) were administered because the CBCL and the YSR are measures validated only for minors.
Materials
All the participants were administered the same battery of psychometric tests including:
The Temperament and Character Inventory (TCI), which provides a clinical classification of different personality traits according to the Cloninger model (53). It includes four temperament subscales: Novelty Seeking, Harm Avoidance, Reward Dependence, Persistence, and three character subscales: Self-directedness, Cooperativeness, Self-transcendence. Each dimension is represented by sub-dimensions which better specify the meaning of the scale, nevertheless for the aims of the present research only main dimensions were used. Concerning reliability and validity, its psychometric properties support its clinical usefulness in the assessment of personality psychopathology (54). Cronbach’s alpha for Italian Version = 0.72.
The Eating Disorder Inventory-2 (EDI-2), a self-administered 91-item questionnaire that evaluates symptoms and characteristics typical of patients affected with ED (51). It consists of 11 scales: Drive to Thinness, Bulimia, Body Dissatisfaction, Inadequacy, Interpersonal Distrust, Perfectionism, Asceticism, Interoceptive Awareness, Impulsiveness, Social Insecurity, and Maturity Fears. All the scales were applied in the present research. Cronbach’s alpha for EDI-2 Italian Version = 0.85.
The Youth Self-Report (YSR) for ages ranging from 11 to 18 and Child Behaviour Checklist (CBCL) for parents (55) measures perceived competencies, adaptive functioning and problems of adolescents within the past 6 months. They have been shown to have adequate reliability and validity (55). The questionnaires include 132 items, of which 20 are competence items (social, activity and academic competence score) and 112 measure eight symptom subscales: withdrawn, somatic complaints, anxiety/depression (grouped into the internalizing problems cluster), aggressive behavior and rule-breaking behavior (grouped into the externalizing problems cluster) and three subscales measuring problems that are both internalizing and externalizing (thought, attention and social problems). Each item is scored on three levels (0 = not true, 1 = sometimes true, 3 = always true). The total problem scale consists of the accumulation of the scores on the 8 symptom subscales and one subscale called “other problems”. The item scores can also be converted into 6 DSM-oriented scales: anxiety, affective, somatic, conduct, oppositional-defiant and attention deficit/ hyperactivity problems. Mean scores on the YSR/CBCL subscales can be compared with the scores of normal controls of the same age and gender, obtaining a T score, that are considered in normal range (T score < 65), borderline clinical range (T score between 65 and 70) and clinical range (T score > 70). Cronbach’s alpha for YSR Italian Version = 0.81; Cronbach’s alpha for CBCL Italian Version = 0.82.
2.2 Statistical Analysis
In the present study they were considered three diagnostic groups according to DSM-5 definition: ANR, ANBP and OSFED. The prevalence of each diagnosis at T0, T1, T2 and T3 were compared with the χ2 test. Descriptive statistics concerning BMI distribution according to DSM levels of severity was performed.
The T0 BMI, personality and psychopathology measures of the follow-up sample recruited at T3 were compared to the whole initial sample using bivariate oneway ANOVA to explore its representativeness.
The GLM ANOVA for repeated measures was applied to compare clinical measures (BMI), personality traits (TCI), self-rated (YSR) and parent rated (CBCL) general functioning, and eating psychopathology (EDI-2) among the treatment endpoints (T0-T1-T2). The diagnostic subgroups were not considered separately because of the numeric exiguity. The Student’s t-test for repeated measures was applied as a post-hoc test between the treatment endpoints (T0-T1, T0-T2, T1-T2).
Based on the diagnostic evolution at T2 it was intended to subgroup the subjects into five groups: 1) “healed” group (i.e. who did not display any of the DSM 5 criteria for an eating disorder at T2); 2) “improved” group (with a diagnostic improvement with respect to T0 but persistence of any DSM 5 criteria for ED diagnosis); 3) “stable” group (those who did not improve or worsen their diagnosis); 4) “worsened” group (those who displayed a worsening of the diagnosis or BMI with respect to T0); 5) those who dropped out from treatment (i.e. between T0 and T1 or between T1 and T2). Since no participant worsened after treatment with respect to T0 the 4th group was empty. Moreover, since there were only four improved participants with minimal residual symptoms at T2, the healed and improved groups were considered together as an “improved” group. Hence the personality and psychopathology scores at T0 of the improved, stable, and dropped-out group were compared by ANOVA to evidence of possible dropout predictors and medium-term prognostic factors for treatment.
Due to the non-normal distribution of the variables in the follow-up sample the Wilcoxon rank test was performed between T2 and T3 for the follow-up group for all available variables (BMI, TCI, EDI-2).
Based on the persistence of diagnostic criteria at T3 (follow-up) they were defined as two subgroups: healed (without any diagnostic criteria) and non-healed (with full or partial ED diagnosis). These were compared using the Mann-Whitney U test due to the non-normal distribution of the samples to evidence possible long-term prognostic factors for treatment.
To evidence the relationships among the changes in BMI, personality and psychopathology measures after multimodal treatment, the deltas of the measures which significantly changed after treatment were calculated and correlated with each other and with the measures at T0 using a multivariate regression analysis. According to a hierarchy descending from personality changes to BMI changes through psychopathology changes and from T0 measures to after-treatment changes, BMI deltas were used as dependent variables with respect to both personality and psychopathology deltas and T0 measures (including BMI at T0). Psychopathology deltas were used as dependent variables with respect to personality deltas and T0 measures. Personality deltas were used as dependent variables with respect to T0 measures.
Statistical analysis was carried out with SPSS 27 for Windows. A Bonferroni correction to p < 0.05 level of significance was applied to the GLM ANOVA analysis on repeated measures to reduce the type 1 error due to the high number of considered variables for each test, according to the number of variables of each test (7 for TCI with p < 0.007, 11 for EDI-2 with p < 0.005, 17 for YSR and 16 for CBCL with p < 0.003).
Instead, in consideration of the explorative nature of the analysis and of the low numerosity of the compared sub-samples it was considered a p < 0.05 significance threshold in the analysis on the FU group and on outcome predictors.
Finally, in consideration of the high number of variables included but also of the data reduction selecting only the significant deltas of BMI, personality and psychopathology measures for correlation and regression analysis a p < 0.010 was considered acceptable for this analysis.