From Statistical Power to Statistical Assurance: It's Time for the Paradigm Change in Clinical Trial Design

• Activity Number: 659
• Type: Contributed
• Date/Time: Thursday, August 13, 2015 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Din Chen* and Shuyen Ho
• Companies: University of Rochester and GSK
• Keywords: assurance ; conditional and unconditional probability ; sample size determination ; Monte-Carlo simulation ; statistical power ; Bayesian prior

Abstract

Clinical trials should be well designed. An aspect of good design of clinical trial protocol is to determine an appropriate number of patients (i.e. sample size) with adequate statistical power to address the clinical objectives. The statistical power is traditionally defined as the probability of rejecting the null if the true clinical trial treatment effect equals the prerequisite value which is then a conditional probability to this unknown prerequisite value. In practice, this value is not unambiguous and a newly proposed alternative to this conventional statistical power is to use assurance which is defined as the unconditional probability of rejecting the null hypothesis. The assurance can be then obtained as the expected power where the expectation is to the prior probability distribution of the prerequisite value which leads to the Bayesian paradigm. In this paper, we review the transition from conventional statistical power to assurance and discuss the computations of assurance using Monte-Carlo simulation-based approach.