An investigation of key symptoms that account for the early response effect during psychological therapy for eating disorders

Objective: The early response effect, defined as a reliable symptomatic improvement during the initial phase of treatment, is the most robust predictor of recovery following eating disorder treatment. This study aimed to investigate which symptom domains mostly influence the early response effect.

Methods: Data from N = 232 adult patients (90.8% females; mean age = 29.97, SD = 10.67) treated in an outpatient eating disorder psychotherapy service were randomly partitioned into training (N = 161) and test (N = 71) samples. A Bayesian network model was developed in the training sample, modelling early changes (sessions 1–4) and interactions among symptoms measured by the Eating Disorder Examination Questionnaire (EDE-Q). A variable selection approach was applied to include only the most important variables in the model (i.e., reliable predictors of recovery). The trained model was externally validated by applying it to predict post-treatment recovery status in the test sample. Prediction accuracy was evaluated using the AUC statistic.

Results: The model identified a network of six interrelated eating disorder symptoms which were the most important predictors of recovery. The model was reliable in predicting recovery status and showed good generalisability to a test sample (training AUC = 0.81 vs. test AUC = 0.77). Early changes in six areas (ranked by importance) reliably predict recovery after therapy: [1] avoidance of body exposure; [2] feelings of ‘fatness’; [3] preoccupation with food, eating or calories; [4] fear of losing control over eating; [5] dissatisfaction with body shape; [6] dietary rules.

Discussion: The identification of early response domains associated with eventual recovery could help to inform targeted interventions strategies for patients with eating disorders. Future replication is warranted in more diverse and larger samples, including the applicability of these findings to different diagnostic groups.