OBJECTIVE:Cognitive-behavioral therapy (CBT) is efficacious for a range of eating disorder presentations, yet premature dropout is one factor that might limit CBTs effectiveness. Improved understanding of dropout from CBT for eating disorders is important. This meta-analysis aimed to study dropout from CBT for eating disorders in randomized controlled trials (RCTs), by (a) identifying the types of dropout definitions applied, (b) providing estimates of dropout, (c) comparing dropout rates from CBT to non-CBT interventions for eating disorders, and (d) testing moderators of dropout. METHOD:RCTs of CBT for eating disorders that reported rates of dropout were searched. Ninety-nine RCTs (131 CBT conditions) were included. RESULTS:Dropout definitions varied widely across studies. The overall dropout estimate was 24% (95% CI = 22-27%). Diagnostic type, type of dropout definition, baseline symptom severity, study quality, and sample age did not moderate this estimate. Dropout was highest among studies that delivered internet-based CBT and was lowest in studies that delivered transdiagnostic enhanced CBT. There was some evidence that longer treatment protocols were associated with lower dropout. No significant differences in dropout rates were observed between CBT and non-CBT interventions for all eating disorder subtypes. CONCLUSION:Present study dropout estimates are hampered by the use of disparate dropout definitions applied. This meta-analysis highlights the urgency for RCTs to utilize a standardized dropout definition and to report as much information on patient dropout as possible, so that strategies designed to minimize dropout can be developed, and factors predictive of CBT dropout can be more easily identified.