Simultaneously leveraging individual- and summary-level external data in clinical trials

This is the authors' abstract. We don't add key points for this paper.

With the growing availability of external data (e.g., real-world data (RWD) and historical data), there is increasing interest in integrating external data into Randomized Controlled Trials (RCT), especially for rare diseases with high unmet medical needs where traditional clinical trials would be impractical and/or unethical. However, despite the increasing availability of external data, the level of detail varies across sources, as some provide individual-level patient records while others offer only aggregated summary statistics, often due to privacy and regulatory constraints. Most existing statistical methods focus on leveraging either individual-level or summary-level external data exclusively, with relatively few approaches addressing the simultaneous integration of both data types. Therefore, we propose a Bayesian method that integrates existing approaches for borrowing individual-level and summary-level external data to augment the control arm of RCT, enabling their joint use through Bayesian updating. Specifically, we combine inverse probability weighting (IPW) with Bayesian power prior for individual-level data, and the unit information prior (UIP) for summary-level data. Simulation results demonstrate that the proposed method maintains type I error control at the nominal level and improves statistical power, while remaining straightforward to implement. Its flexible structure also allows the components of the proposed method to be replaced with alternative existing methods that leverage either individual-level or summary-level external data, thereby enabling the integration of both data types to effectively synthesize external information.