Risk of bias in randomized controlled trials of psychological treatments for bulimia nervosa and binge eating

Purpose: In the context of Cochrane systematic reviews/meta-analyses of randomized clinical trials, risk of bias (RoB) is assessed using categorical indicators (low, unclear, or high RoB). This study sought to evaluate the indicators of the Cochrane RoB tool available for construct validity as applied to randomized clinical trials of psychological treatments for bulimia nervosa and binge eating. Methods: Bayesian confirmatory factor analysis was used to test the construct validity of the measurement model underlying the set of five categorical items, and the reliability of these indicators to measure RoB. Results: In 48 primary randomized clinical trials, the model showed good fit indices and factor loadings higher than 0.4. Conclusions: The results support the construct validity of the Cochrane RoB tool and the reliability of three of five items in this health intervention context. (C) 2018 Elsevier Inc. All rights reserved.