Patients
Overall, 24 youths (23 females, one male) were included in the pilot study of whom 16 (all female) completed the posttreatment and 13 the post follow-up assessment of BEAT (see Fig 1). The recruitment of patients in the present sample took place between May 2018 and February 2020 at the Division of Clinical Psychology and Psychotherapy of the University of Fribourg (Switzerland). The recruitment was promoted via public advertisements on webpages and fitness centers, media, as well as cooperating clinicians, healthcare institutions and foundations. Inclusion criteria were the presence of LOC at least once during the last 6 months (based on findings that LOC of low frequency is associated with poorer mental health outcomes compared to no LOC; e. g. 9) up to threshold BED according to the DSM-5 (1), age between 14 and 24 years (according to the term youth defined e.g. by the UN (2)) and written informed consent. Exclusion criteria were the presence of another medical or psychological condition requiring prior treatment (e.g. acute substance abuse, psychosis, suicidality), current Bulimia Nervosa or Anorexia Nervosa, pregnancy, the lack of sufficient German language skills, and concurrent participation in a diet or weight loss program or in an eating disorder psychotherapy.
Insert Fig 1 here
Study design and procedure
The BEAT pilot study represents a repeated-measures (within-subjects) waitlist control design. After giving informed consent, patients completed pretreatment assessment (week 0) in which online questionnaires were provided and mental disorders assessed in a clinical interview. Thereafter, eligible patients completed a two-week waiting time (up to week 2; for a detailed overview of study weeks, assessments and study periods, see Table 1) before starting with the first session of the active treatment. The active treatment lasted nine weeks (up to week 11), including nine weekly sessions (one session per week) and the posttreatment assessment, in which the same online questionnaires were provided as for the pretreatment assessment. The follow-up period included 49 weeks (up to week 59) with four follow-up assessments 3, 11, 23, and 47 weeks after the posttreatment assessment (i. e. 1, 3, 6, and 12 months after the last session of the active treatment at weeks 14, 22, 34 and 58, respectively), and the post follow-up assessment (week 59).
Insert Table 1 here
Face to face workshops during the active treatment were conducted in groups of at most three patients or in a single setting to avoid long waiting times. Email-guided self-help sessions were processed by patients at home. The BEAT program is based on an evidence-based CBT treatment manual for adults with BED that was developed by our research group (31, 32, 47, 48) and adapted to youth in terms of simplification and adequacy of language and interventions. BEAT further encompasses a training in interpersonal emotion regulation such as coping with rejection and appearance-based rejection sensitivity. During the three workshops that lasted approximately 90–180 minutes (depending on the workshop and the group- or single setting), all eating disorder specific interventions as well as interventions on interpersonal emotion regulation were discussed and prepared. The email-guided self-help sessions lasted approximately 30–60 minutes per session and were thought to support patients in implementing interventions that were discussed and prepared during the workshops in daily life. The content of each BEAT session is summarized in Table 2. All treatment sessions were manualized and standardized. However, for some of our youngest patients, sessions had to be further adapted in terms of length, without changing the session’s content.
Insert Table 2 here
Each patient was assigned to one of four therapists for guidance. All therapists were postgraduate psychologists in CBT training, supervised by SM. After each email-guided self-help session, patients sent their notes, questions and worksheets to their therapist via email. Therapists provided written feedback via email within three days according to standardized topic and text templates (available from the authors) that were derived and adapted for BEAT from previous email- and online based BED treatment programs for adults (32, 49). All feedback messages were then individualized to the specific needs of the individual patient.
The BEAT pilot study (DRKS00014580; date of registration: 21/06/2018) was approved by the local ethic committee in Switzerland (study ID of the cantonal ethics approval: 2018-00230) and conforms the Declaration of Helsinki. All patients gave written informed consent prior their study participation (50).
Measures
Sociodemographics
At pretreatment assessment patients were asked to provide age, gender, nationality and occupation status.
Diagnostic interview for mental disorders, short version (Mini-DIPS; 51)
LOC and further mental disorders were assessed by the Mini-DIPS, a structured interview to assess mental disorders according to the DSM-5 (1), which was conducted by phone before the start of BEAT (pretreatment) and one week after the active treatment (posttreatment). The Mini-DIPS has good reliability and validity in outpatient, inpatient and community samples (52, 53).
The following two online self-report questionnaires were assessed weekly during the waiting-time period, before each session of the active treatment, at posttreatment, before each follow-up session, and at post follow-up (total of 17 assessment points).
Weekly Binges Questionnaire (WBQ; 54)
The WBQ was used to assess the frequency of self-reported weekly LOC episodes by asking youth about the number of episodes in which they experienced loss of control over eating a perceived unusual large amount of food. It is important to note that self-report assessment of objectively large amounts of food has shown to be less reliable than interview-based assessment (3). Similarly, the definition of an objective large amount of food and its differentiation from smaller respectively normative food amounts, not fulfilling the criteria of a binge-eating episode according to the DSM-5 (1), is difficult due to different growth-dependent needs of energy intake and age-dependent limited access to high quantities of food (9). Therefore, in youth samples the WBQ assesses LOC (loss of control eating irrespective of the amount of food) rather than binge-eating. The WBQ shows high convergent validity relative to ecological momentary assessment (55).
Beck Depression Inventory-Fast Screen (BDI-FS; 56)
The BDI-FS is a short version of the Beck Depression Inventory (57), consisting of seven items and assesses depressive symptoms during the last two weeks. To assess weekly depressive symptoms, we adapted the symptom review period to the past seven days. The BDI-FS has satisfactory internal consistency (Cronbach’s α = .84) and convergent validity (r = .67; 57). Cronbach’s α in the present sample was .81 at pretreatment assessment.
The following two online self-report questionnaires were assessed four times (at pre- and posttreatment, 6-month follow-up, and post follow-up).
Eating Disorder Examination-Questionnaire (EDE-Q; 59)
The EDE-Q assesses eating disorder pathology during the last 28 days. It consists of 28 items, of which 22 items can be assigned to four subscales (restraint eating, eating concerns and shape-and weight concerns) and a global score. For the present study, only the global score was used. The abstainer rate from LOC was defined as no LOC episodes during the last month, and was derived from item 15 of the EDE-Q, which assesses the number of LOC episodes similarly to the WBQ but during the last 28 days (“over the past 28 days, on how many days have you eaten an unusually large amount of food and have had a sense of loss of control at the time?”). The EDE-Q was also used to assess self-reported weight and height to calculate age and gender adjusted BMI (kg/m2) standard deviation scores (BMI-SDS) according to the LMS method (60). The EDE-Q subscales and the global score showed good internal consistency with Cronbach’s α values ranging from .70 to .94 (61). In the present study, Cronbach’s α of the EDE-Q global score at pretreatment assessment was .93.
Appearance-based rejection sensitivity Scale (ARS-D; 62)
To assess appearance-based rejection sensitivity, we applied a shortened and adapted version of the German version of the ARS. In the present version of the ARS-D, 10 interpersonal situations are presented to patients (originally the ARS-D includes 15 situations). The ARS-D has good convergent and divergent validity and satisfying internal consistency with Cronbach’s α of .90 (62). Cronbach’s α of the present adapted ARS-D version at pretreatment assessment was .97.
Clinical Global Impression Scale (CGI; 63)
The CGI assesses the global clinical impression of patients by clinicians before and after treatment, applying three measures: the global severity of illness measure (CGI-S), the global improvement measure (CGI-I) and an efficacy index. The CGI-S was rated at pre- and posttreatment assessment by therapists and additionally by patients. The CGI-I was rated at posttreatment assessment by therapists. In the present study, side effects assessed within the efficacy index are reported. Findings regarding the validity of the CGI are inconsistent. While some studies supported the validity of the CGI in clinical trials (e.g. 64), others criticized the CGI of being inconsistent, biased and too general (65-67).
Patient’s subjective evaluation of the BEAT treatment program (own items)
Patients reported their satisfaction with BEAT at posttreatment assessment applying eight self-developed items based on the Working Alliance Inventory (WAI; 68, “overall, how satisfied were you with BEAT?”, “how much did BEAT help you to cope with LOC and the feeling of losing control while eating?”, “do you think, BEAT is efficacious?“, how much did you like the mix of email-guidance and workshops?”, “how satisfied have you been with the support you received from your therapist?”, “I think, another treatment would have been better for me”, “would you participate again in BEAT?”, “would you recommend BEAT to peers?”). Items were rated on a scale from 0 (not at all) to 10 (very much). While each item has been analyzed separately, items were additionally summarized to a mean treatment satisfaction score. Cronbach’s α in the present sample of the total score at posttreatment assessment was .91.
Statistical analyses
WBQ and BDI-FS were assessed regularly (altogether 17 times), thereby allowing us to analyze their weekly temporal course using a discontinuous multilevel model (69), covering three study periods for WBQ and BDI-FS (waiting-time, active treatment and follow-up) and two study periods for EDE-Q and ARS-D (active treatment and follow-up). A linear weekly time course was estimated for each phase, resulting in three (WBQ and BDI-FS) or two (EDE-Q and ARS-D) fixed effects. A random intercept and, if this improved model fit, random slope coefficients for the three phases were also included in the model. For the BMI-SDS, no discontinuous multilevel model was set up since previous studies did not point to a decrease in BMI during active treatment (i.e. no specific temporal trend could be specified beforehand). Instead, we set up a multilevel model with time as sole fixed factor (four levels: pretreatment, posttreatment, 6-month follow-up, and post follow-up), plus a random intercept.
To assess the change in the CGI-S scale between pre- and posttreatment, we used the Wilcoxon-test. For the WBQ, the Number Needed to Treat (NNT) for a significant treatment outcome during the active treatment relative to the waiting-time was calculated according to (70). The WBQ was transformed (ln[x+1]) prior to analyses and predicted means from the multilevel models were back-transformed for reporting. All other outcomes were left untransformed.
We used descriptive statistics, i.e. means and standard deviations, to analyze patient’s subjective evaluation of the BEAT program (own items).
All analysis were performed using R for statistical computing, version 3.2 (71). The level of significance was set at .05.