Prospective Specification of a Conditional Power Prior for Clinical Trials

• Activity Number: 624 - Trial Design and Analysis Issues in Medical Devices
• Type: Contributed
• Date/Time: Thursday, August 3, 2017 : 8:30 AM to 10:20 AM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Tyson Rogers*
• Companies: NAMSA
• Keywords: Bayesian ; Virtual Patients ; Clinical Trial ; Prior

Abstract

Incorporating historical study data into a new clinical trial can be accomplished by specifying an informative prior in a Bayesian analysis. However an informative prior can result in substantial Type I error. We explore using an informative power prior where the power parameter is specified using a loss function that assesses the similarity between the observed results and the expected value of the informative prior, which has previously been applied to prior information from stochastic engineering models. This method results in borrowing more historical data if results are comparable to those previously observed and less if results differ. The method limits the degree of Type I error inflation. We demonstrate use of this prior in the context of a non-inferiority randomized controlled clinical trial with a focus on using simulation to prospectively specify the parameters of the loss function.