Data-Driven Prior Distributions for a Phase II COPD Dose-Finding Clinical Trial

• Activity Number: 372
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Shuyen Ho* and Steven Novick
• Companies: GSK and GlaxoSmithKline
• Keywords: Bayesian ; Prior Distribution ; Phase 2 ; Clinical Trials ; Dose Finding ; Historical Data

Abstract

The prior distribution reflects knowledge and uncertainty of the modeled parameter space. Determining a reasonable prior distribution for a dose-finding clinical trial can be influential in its design and analysis. In planning a phase-2 trial for COPD with a dose-response curve, we illustrate the use of relevant historical data for the nonlinear curve mean-model parameters, residual variance parameter, consideration for terms to characterize between-trial variability, and the potential bias in the curve. Through a predictive inference exercise, we illustrate how we derived an informative prior distribution for the future study. Some background information on clinical trials for COPD will be given. We share our strategies on how to obtain informative Bayesian priors using relevant historical data and the associated issues.