Design and analysis of randomized controlled trials using a hybrid control approach: a case when external data is derived as summary statistics

Abstract

International momentum to promote the use of external control data in clinical trials is growing. The use of external control data to augment the control arm in randomized controlled trials (RCTs), termed as a hybrid control approach, is one of the most attracting areas for actual applications. In this article, we focus on a situation where an endpoint of interest is binary and the summary statistics of external control are available to this end. The purpose is to review design and analysis methods for RCTs incorporating a hybrid control approach. The methods of interests are test-then-pool, power prior and its extended versions, Pocock’s method, commensurate prior, meta-analytic predictive prior and its robustified version, unit-information prior, elastic prior, and so on. For better understanding of Bayesian methods, a brief overview of recent proposals regarding an effective sample size is also presented. Most of these methods are applied to an actual placebo-controlled RCT for ankylosing spondylitis patients which applied a hybrid control approach by using summary statistics of past RCTs.