A Shortcut in Bayesian Application

Abstract

Abstract For a well-established class of medical products, when a similar or improved version of a product in the class has been developed, to evaluate the safety and effectiveness of the new product, the Bayesian approach is gaining popularity to save resource and reduce trial duration by leveraging highly relevant external data. However, the Bayesian approach can appear daunting to practitioners, including statisticians without sufficient exposure to the Bayesian methodology. Instead of using the typical modeling approach to borrow evidence from external data, this paper proposes to quantify the effective sample size from the external data first and then combine the effective portion of the external data with the current trial data through a Bayesian scheme. Guided by the fundamental principles of the central limit theorem and sufficient statistics, the proposed method provides a shortcut to assess and implement the Bayesian methodology in a transparent and generalized fashion. For two selected case studies, the proposed method generates comparable analysis results with those from the typical modeling approaches.