Using Bayesian Method to Predict Responders Based on Clinical Scores in the Initial Weeks of Treatment in a Phase II Multiple Sclerosis Trial

• Activity Number: 414
• Type: Contributed
• Date/Time: Tuesday, August 2, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Dazhe Wang*+ and Lynn Wei
• Companies: sanofi-aventis and sanofi-aventis
• Keywords: Bayesian predictive model ; Responders prediction ; ROC curve analysis

Abstract

In the treatment for Multiple Sclerosis (MS), responders and non-responders to study medication may have different profiles for the clinical outcome of interest, and the separation of the profiles may start at the early stage of the treatment due to a possible early onset of drug action. The ability to predict responders based on early observations from patients can help make timely decision to stop treatment for those who are predicted as non-responders. The aim of this research was to develop Bayesian prognostic models for allocation of a patient to either responder or non-responder group at study end on the basis of partial data collected at the initial stage of the study. Three independent models (logistic, longitudinal and autoregressive) were considered using the data from a phase II MS clinical trial. Their predictive performance was then evaluated using the receiver operating characteristic (ROC) curve analysis.