Introduction
The Yale-Brown Obsessive Compulsive Scale modified for Binge Eating (Y-BOCS-BE), which assesses the obsessiveness of binge eating (BE) thoughts and compulsiveness of BE behaviors [
1], is a modified version of the Yale-Brown Obsessive Compulsive Scale [
2]. The Y-BOCS-BE has been used in studies of binge eating disorder (BED) to assess the efficacy of pharmacotherapy [
3?
6]. Across studies, reductions in BE were accompanied by statistically significant reductions in Y-BOCS-BE total score [
3?
6].
Psychometric testing and analysis of the Y-BOCS-BE is being conducted as a multi-stage process to optimize the characterization of BED. A preliminary validation of the Y-BOCS-BE [
1] was conducted using data from a phase 2 study of lisdexamfetamine dimesylate (LDX) in adults with moderate-to-severe BED [
7]. This analysis demonstrated that the Y-BOCS-BE had high internal consistency (Cronbach?s
?, 0.81) at baseline [
1]. The Y-BOCS-BE also exhibited good construct validity in relation to multiple reference measures, including the Three-Factor Eating Questionnaire (TFEQ) and Binge Eating Scale (BES), but correlations with reference measures at baseline were lower than for score changes from baseline to end-of-study [
1]. The estimated range of Y-BOCS-BE total score reductions indicative of minimal clinically important change was 4 to 17 points [
1]. These data suggested the Y-BOCS-BE is a reliable and valid measure of treatment benefit in BED [
1].
This study extends the preliminary psychometric validation of the Y-BOCS-BE by using two phase 3 trials in adults with BED for validation and another phase 3 trial for characterization of treatment benefit on developed scores [
6,
8]. LDX was superior to placebo in reducing BE days/week in the two short-term efficacy studies [
6] and in prolonging time to relapse in the randomized-withdrawal maintenance-of-efficacy study [
8]. The Y-BOCS-BE was included as a secondary efficacy endpoint in each of the studies. In the short-term efficacy studies, LDX was superior to placebo in reducing Y-BOCS-BE total score [
6]. In the maintenance-of-efficacy study, LDX maintained Y-BOCS-BE total score reductions during the randomized-withdrawal phase [
8]. These analyses further examine the dimensionality, item-level properties, scoring, and minimal clinically important improvement (MCII) of the Y-BOCS-BE and LDX treatment effects on Y-BOCS-BE scores from the aforementioned studies.
Discussion
This study described the psychometric evaluation of the Y-BOCS-BE based on phase 3 clinical data from two independently conducted and identically designed short-term efficacy studies and a maintenance-of-efficacy study of LDX in adults with BED. These analyses demonstrated that the Y-BOCS-BE possesses a bifactor structure composed of a general binge eating severity domain and three subdomains (obsessive/compulsive, restraint, and control) and exhibits strong internal consistency. Test?retest reliability from baseline to week 4 and convergent validity were poor at baseline but substantially better from week 4 to week 12 for test?retest reliability and at week 12 for convergent validity, potentially due to LDX treatment effects and improved insight into BED at later points during the studies. MCII estimators based on raw metrics displayed considerable heterogeneity, which was reduced with score standardization. Estimated treatment effects were significant, consistent with published results [
6,
8], and accounted for a substantial percentage of the variance in change scores.
The Y-BOCS-BE measured BED severity via items assessing BE thoughts and actions. These analyses demonstrated that the Y-BOCS-BE can be decomposed into three domains (obsessive/compulsive [6 items], restraint [2 items], and control [2 items]) rather than the hypothesized two domains (obsessiveness of binge thoughts, compulsiveness of binge actions) [
1]. This is consistent with previously published findings, which also reported that a two-domain structure did not adequately describe the Y-BOCS-BE [
1]. In these analyses, a bifactor solution of the three-domain structure fits the data optimally and was carried forward. Under the bifactor solution, subdomains are assumed to arise because of idiosyncratic effects, such as shared item phrasing, that are assumed to be noise and best explained by the common unidimensional domain of BED severity.
Inter-item correlations, internal consistency, test?retest reliability, and convergent validity were poor for analyses conducted at baseline but substantially improved at later time points. These findings are also consistent with those of Deal and colleagues, who reported lower inter-item correlations and convergent validity at baseline than end-of-study [
1]. Studies of emotional function/processing in BED have shown high levels of alexithymia (inability to identify and describe emotions), impaired interoceptive awareness (ability to recognize and respond to emotional states and visceral sensations), and impaired emotion regulation, including low cognitive reappraisal [
18,
19]. These findings suggest that an inability to identify and describe emotions and visceral states may negatively impact self-awareness of BED and its severity in individuals with BED. Therefore, it is possible that the weaker baseline psychometric properties of the Y-BOCS-BE are related to poor disease insight, including an inability to discriminate BED symptom severity. However, as the studies progressed, participants gained insight into their disorder through LDX treatment effects and the experience of evaluating their symptoms.
Consistent with a previous Y-BOCS-BE validation [
1], MCII estimates varied within estimator across scores and within scores across estimators so meaningful agreement on a representative meaningful change estimate could not be established in either short-term efficacy study. The MCII estimates associated with the most reasonable achievement rates (50% and 63.5%) were obtained from the ?
??Baseline SD and ?
??Baseline SEM estimators, respectively. Therefore, the estimates in score reductions of 12 to 17 points were taken to represent the best estimates of clinically meaningful improvement. This estimate range is narrower than the 4- to 17-point range reported by Deal and colleagues [
1]. While a meaningful change estimate of up to 17 points is large relative in regard to the potential change score range (±?40), in fact, this reflects only 21% of the total range. Moreover, given the large treatment effect of LDX, lesser meaningful change estimates produced achievement rates of???80%.
Examination of CDF curves for all scores displayed substantial separation between LDX and placebo, with estimated treatment effects being significant in all three studies and consistent with previous publications [
6,
8]. The observed treatment effects accounted for a substantial percentage of the change score variance for all scores. Based on these findings, ?
??Baseline SEM yielded a narrower CI range, and would be the recommended metric for defining responders in a clinical trial.
Several limitations should be considered when interpreting these data. First, including only participants with protocol-defined moderate-to-severe BED without psychiatric comorbidities limits the generalizability of these findings to the more heterogeneous general population of individuals with BED. Second, although the Eating Disorder Examination Questionnaire is considered a gold standard for assessing eating disorder symptoms, it was not used to validate the Y-BOCS-BE because it was not assessed at the baseline or post-baseline visits. Third, MCII estimators based on standardized data displayed increased homogeneity and may be advantageous in educational testing and real-world evidence contexts, for which larger sample sizes can be used to generate standardized estimates for the purposes of comparison. However, they are sample dependent and do not permit individual patient benchmarking and thus may be impractical in clinical practice due to the need to focus on patient-centric and individualized assessment in the clinical practice setting. Last, when considering treatment effects in the maintenance-of-efficacy study, it is a limitation that the analysis only included participants who did not relapse during the randomized-withdrawal phase.
These analyses from large placebo-controlled studies of LDX indicate that the Y-BOCS-BE is a valuable tool for assessing BED symptoms. These analyses demonstrated that the Y-BOCS-BE can be decomposed into three distinct domains (obsessive/compulsive, restraint, and control) rather than the previously hypothesized two domains of the Y-BOCS-BE. However, a bifactor solution of the three-domain structure best fits the data and our findings were consistent with the local dependence findings contained in the supplemental material of Deal et al. [
1]. Since the psychometric properties of the Y-BOCS-BE were substantially stronger at later study time points than at baseline, it may be valuable for clinicians to implement regular self-reports of BE thoughts and behaviors into their clinical practices in order to enhance patient insight into BED and possibly improve long-term outcomes. Our findings based on MCII estimators for Y-BOCS-BE total score and standardized scores set the stage for normalizing the Y-BOCS-BE and increasing the understanding of the clinical significance of Y-BOCS-BE scores and score changes to be useful both for clinical practice and clinical research.
Acknowledgements
Under the direction of the authors, Craig Slawecki, PhD, an employee of Complete Healthcare Communications, LLC (CHC; North Wales, PA), a CHC Group company, provided writing and formatting assistance for this manuscript. Shire Development LLC, Lexington, MA, a member of the Takeda group of companies, provided funding to CHC for support in writing and editing this manuscript.
Compliance with ethical standards
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.