Validation of Early Functional Change Based Prognostic Groups in Predicting Long Term Depression Outcomes: Findings from COMED Trial

• Activity Number: 506 - Statisticians' Approaches to the Realities of Clinical Trials
• Type: Contributed
• Date/Time: Wednesday, August 2, 2017 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Abu Minhajuddin* and Manish K Jha and Madhukar Trivedi
• Companies: University of Texas Southwestern-Medical Center At Dallas and University of Texas Southwestern Medical Center and University of Texas Southwestern Medical Center
• Keywords: Clinical Trial ; Depression ; Validation ; Growth Model ; Logistic Regression ; Mixture Model

Abstract

Identifying prognostic groups based on short-term trajectories of change that are more (or less) likely to achieve longer term favorable outcome(s) is of interest in many longitudinal treatment trials. We used changes in psychosocial function measured by WSAS during the first 6 weeks of antidepressant treatment to identify two distinct subgroups among participants of CO-MED trial via SAS PROC TRAJ. The subgroups were derived using a randomly selected training sample of n = 334 and final model was selected using lower AIC and BIC values, parsimony, membership size, and interpretability. We then classified participants in the other half of the CO-MED sample (test sample, n = 331) into one of two subgroups derived in the training sample based on the minimum mean squared difference between the observed and predicted WSAS score over the first 6 weeks. Compared to participants with gradual change in WSAS, those with early improvement had significantly higher rates of remission at both 3 months and 7 months even after controlling for the above-mentioned baseline variables and remission status at week 6. These findings are similar to those obtained based on the training sample.