Probability of Success Based on Confidence Distribution and Its Application in Clinical Development

• Activity Number: 660
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yabing Mai*+ and Jiajun Liu and Yue Shentu
• Companies: Merck Research Laboratories and Merck Research Laboratories and Merck
• Keywords: POS ; Confidence Distribution ; Go/No-go Decision

Abstract

The assessment of Go/No-go decision criteria is critical for clinical development and the Probability of Success (POS) provides a quantitative approach for the decision making process. In general the POS is derived by combining the classical hypothesis testing framework with the Bayesian formulation of prior knowledge for study design parameters. In order to maintain the integrity and unbiasedness of the decision process, it is imperative to achieve data-based objectivity in the calculation of POS. Therefore we propose the POS based on Confidence Distribution (CD), which is a data-driven alternative to the Bayesian POS under the Frequentist framework. CD is a modern concept that bridges the gap between objective Bayesian methods and Frequentist paradigm. It is a sample-dependent distribution estimator of parameter(s) that contains all relevant information for point estimate, hypothesis testing for any value of interest, and confidence intervals at any level of significance. In this presentation we will review the concept of CD, introduce the derivation of CD-based POS, and investigate its application in decision making of clinical development.