Borrowing in Clinical Trials: Design and Analysis Perspectives

The concept of information borrowing in clinical trials is widely used but lacks a unique interpretation. Traditionally, borrowing refers to the incorporation of external data at the end of a study to improve inferential effciency, for example by augmenting the information on the control arm. More broadly, however, borrowing may also occur during the course of an ongoing trial, as in an adaptive randomization framework where information can be shared across studies. While the potential benefts and risks are common to both perspectives, the distinction between them remains ambiguous in the literature. In this paper we provide an overview of borrowing in adaptive experiments, discuss key methodological and practical challenges, and argue that this remains an underexplored area warranting greater attention.