BEATS: Bayesian hybrid design with flexible sample size adaptation for time-to-event endpoints

Area of Science:

• Clinical Trial Design
• Biostatistics
• Drug Development

Background:

• Increasing reliance on historical trials and real-world evidence (RWE) in drug development.
• Existing methods often focus on analysis, risking unreliable estimates due to heterogeneity misspecification.
• Need for improved control arm augmentation using external data.

Purpose of the Study:

• Introduce a Bayesian hybrid design with flexible sample size adaptation (BEATS).
• Enable adaptive borrowing of external data to augment control arms during design and interim analyses.
• Extend Bayesian semiparametric meta-analytic predictive prior (BaSe-MAP) for time-to-event endpoints.

Main Methods:

• Calibrates sample size and randomization ratio based on external data heterogeneity.
• Implements flexible sample size adaptation at interim analyses to resolve control arm conflicts.
• Incorporates calibrated external data for final estimation and futility analysis.

Main Results:

• BEATS allows adaptive borrowing of external data based on heterogeneity levels.
• Extends BaSe-MAP to handle time-to-event endpoints for optimal borrowing.
• Provides flexible sample size adaptation to manage concurrent and historical control conflicts.

Conclusions:

• BEATS approximates an ideal randomized controlled trial with equal randomization.
• Enhances patient benefit and accelerates drug development by optimizing control arm size.
• Offers robust estimation and optimal power through adaptive sample size adjustments.