Adaptive Constrained Weighted Estimation for Incorporating Multiple External Information Sources.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Data Science

Background:

• Growing interest in hybrid control approaches for clinical trials.
• Need for methods to augment concurrent control data using external sources.
• Challenges in managing heterogeneity across multiple information sources.

Purpose of the Study:

• To propose a novel frequentist method for borrowing information from multiple external sources in clinical trials.
• To manage heterogeneity across information sources while controlling power and Type I error rates.
• To dynamically optimize weights for external information sources without prespecified tuning parameters.

Main Methods:

• Constrained weighted maximum likelihood estimation.
• Concept of effective sample size for information borrowing.
• Dynamic optimization of weights for external information sources.
• Simulation studies and numerical examples to evaluate performance.

Main Results:

• The proposed method effectively manages heterogeneity across information sources.
• Power and Type I error rates are precisely controlled at nominal levels.
• Outperforms existing methods in simulations, showing greater robustness.
• Demonstrates adaptive weighting behavior and maintains effective sample size at a target level.

Conclusions:

• The proposed frequentist method offers a robust approach for hybrid control in clinical trials.
• Dynamic weight optimization enhances the utilization of external data.
• The method ensures reliable statistical performance and effective sample size management.