A bayesian dynamic borrowing framework for improving the efficiency of clinical trials.

PURPOSE: To develop a Bayesian method for dynamic borrowing of information from historical clinical trials and real-world data that has the potential to improve the efficiency of clinical trials. METHODS: We propose a novel statistical metric to quantify heterogeneity among data sources. Based on this metric, a Multi-Source Dynamic Borrowing (MSDB) Bayesian prior framework was proposed, which can address baseline imbalance and dynamically discount information from each data source separately without assuming exchangeability. In a simulation study, we compared the proposed framework and metrics with historical methods. RESULTS: We evaluated performance by comparing power, type I error, bias, and mean squared error (MSE) across methods. Our approach outperformed existing methods by enhancing power and reducing MSE, while effectively controlling type I error and bias in the presence of heterogeneity and baseline imbalances. Moreover, an application involving isatuximab in relapsed and refractory multiple myeloma further validated the performance of the MSDB prior. CONCLUSION: The MSDB prior method dynamically adjusts borrowing information, allowing for a more robust and accurate incorporation of historical data. Its flexible framework allows it to play a crucial role in various applications, thereby accelerating drug development and supporting regulatory decision making.

Duke Scholars

Author Shein-Chung Chow Biostatistics & Bioinformatics, Division of Biostatistics