Eating Disorders
Predictors of Eating Disorder Relapse: A Meta-Analysis.
Margaret Sala, Ph.D.
Assistant Professor
Ferkauf Graduate School of Psychology
Riverside, Connecticut
Ani C. Keshishian, B.A.
Clinical Psychology Doctoral Student
University of Louisville
Louisville, Kentucky
Sarah Song, B.A.
Doctoral student
Ferkauf Graduate School of Psychology
Flushing, New York
Rivka Moskowitz, None
Student
Yeshiva University
Bergenfield, New Jersey
Cynthia M. Bulik, Ph.D.
Distinguished Professor
UNC Chapel Hill/Karolinska Institutet
Chapel Hill, North Carolina
Corey R. Roos, Ph.D.
Assistant Professor
Yale University School of Medicine
New Haven, Connecticut
Cheri A. Levinson, Ph.D.
Associate Professor
University of Louisville
Louisville, Kentucky
Eating disorders (EDs) have high rates of relapse. However, it is still not clear what factors are the strongest predictors of ED relapse, and the extent to which predictors of relapse may vary due to study and individual differences. We conducted a meta-analysis to quantify and compare the factors that predict relapse in EDs and evaluate various potential moderators of these relations (e.g., ED subtype, sample age, length of follow-up, timing of predictor assessment). A total of 35 papers (effects = 315) were included. We used a multilevel random-effects model to estimate summary study-level effect sizes, and multilevel mixed-effects models to examine moderator effects. Higher level of care, having psychiatric comorbidity, and higher severity of ED psychopathology were associated with higher odds of relapse. Higher leptin, higher meal energy density/variety, higher motivation for change, higher body mass index (BMI)/weight/body fat, better response to treatment, anorexia nervosa (AN)-restricting (vs. AN-binge purge) subtype diagnosis, and older age of ED onset were associated with lower odds of relapse. Several moderators were identified. A variety of variables can predict ED relapse. Furthermore, predictors of ED relapse vary among ED subtypes, sample ages, lengths of follow-up, and timing of predictor assessments. Future research should identify the mechanisms by which these variables may contribute to relapse.