Bayesian and Frequentist Promising Zone Designs for Adaptive Sample Size Modification in Clinical Trials

• Activity Number: 25 - Medical Devices and Diagnostics Speed Session
• Type: Contributed
• Date/Time: Sunday, August 8, 2021 : 1:30 PM to 3:20 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Apurva C Bhingare* and Cyrus Mehta and Pralay Senchaudhari and Lingyun Liu
• Companies: Cytel Inc. and Cytel Inc and CYTEL INC and Vertex
• Keywords: Bayesian; Adaptive Design; Promising Zone; Sample Size Re-estimation; Trial Optimization; Conditional and Predictive Power

Abstract

Clinical trials with adaptive sample size reassessment based on the unblinded data at the interim characterized by promising zone have become popular in the recent times. Mehta and Pocock (2011) and Jennison and Turnbull (2015) provide examples of such adaptive designs. However, the question of how to determine the promising zone and the corresponding sample size re-estimation function based on some underlying principles rather than trial and error is not straightforward to address. Conditional power based on the observed interim results provides an insight into the chance of a successful trial and hence it would be reasonable to base the choice promising zone and sample size re-estimation rule on the observed conditional power. To this end we propose the Constrained Promising Zone (CPZ) approach to construct promising zone designs based on the intuitively plausible notion that any additional investment of sample size at an interim analysis should be contingent on a minimal acceptable return on the investment, expressed in terms of guaranteed conditional power. To incorporate the uncertainty around the design parameters, we also provide a Bayesian alternative to the CPZ approach.