Predictors of remission with placebo using an integrated study database from patients with major depressive disorder

Abstract

Objective:To identify predictors of remission with placebo treatment in double-blind randomized controlled trials (RCTs) in major depressive disorder (MDD).Method:A total of 1017 placebo-treated patients with baseline Hamilton Depression rating scale (HAMD) total ≥15 from eight duloxetine RCTs were included. Remission was defined as endpoint (7–8 weeks) HAMD total ≤7. Data were randomly split into training data (N = 813, 80%) for model selection and test data (N = 204, 20%) for validation. Logistic regression and classification and regression tree (CART) methods were used to identify predictors of remission. Predictive accuracy of models was assessed by Receiver Operator Characteristic (ROC) curves.Results:Baseline predictors for remission with placebo consistently identified with the logistic regression and CART analysis were less severe depressive symptoms (based on HAMD core symptoms), younger age, less anxiety (based on HAMD anxiety/somatization), and shorter current MDD episode duration. Associated cut-off values from the CART method characterized patient groups according to their remission likelihood. However, the predictive accuracy was modest for both methods with areas under the ROC curve of 0.6–0.65 based on test data.Conclusion:The derived models, although of limited value for predicting remission in individual patients, may be useful for adjusting for placebo effects in clinical trials.