Approximate Bayesian Analysis for Borrowing External Controls for Randomized Controlled Trials With Dynamic Borrowing...

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Statistical Inference

Background:

• Augmenting control arms in small randomized controlled trials (RCTs) with external data is common.
• Differences between external and RCT populations can introduce bias, potentially invalidating combined data analysis.
• Dynamic borrowing and pre-adjustment for prognostic factors can mitigate bias in external data integration.

Purpose of the Study:

• To propose a robust Bayesian approach for integrating external control data into RCTs.
• To address variability from dynamic borrowing and covariate pre-adjustment using a Bayesian bootstrap (BB) framework.
• To develop a computationally efficient BB-algorithm for approximate posterior sampling.

Main Methods:

• Integrated Bayesian approach utilizing Bayesian bootstrap (BB) for dynamic borrowing.
• Covariate balancing (CB) for pre-adjustment of external data using Euclidean or entropy distances.
• Development of a BB-algorithm for efficient approximate posterior sample generation.

Main Results:

• The proposed BB-based approach with CB is shown to be a valid approximate Bayesian method.
• Demonstrated validity even when CB uses distances different from probability-based approaches.
• The BB-algorithm provides an efficient way to generate approximate posterior samples for statistical inference.

Conclusions:

• The novel Bayesian bootstrap approach effectively integrates external control data in RCTs.
• The method offers valid statistical inference by mitigating bias through dynamic borrowing and covariate balancing.
• The proposed algorithm is computationally efficient and practical for real-world applications, as shown in an acute myeloid leukemia trial example.