Digital twins and Bayesian dynamic borrowing: Two recent approaches for incorporating historical control data.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Health Informatics

Background:

• Increasing interest in using external control data for randomized clinical trials (RCTs).
• Potential benefits include reduced costs, shorter trial durations, and feasibility for small populations.
• Bayesian dynamic borrowing (BDB) and Digital Twins (DT) are emerging methods for historical data utilization.

Purpose of the Study:

• To analyze and compare Bayesian dynamic borrowing (BDB) and Digital Twins (DT) methods for RCTs.
• To evaluate their performance using analytic derivations and simulations.
• To identify fundamental differences and practical considerations for their application.

Main Methods:

• Analytic derivations and simulation studies were employed.
• Bayesian dynamic borrowing (BDB) approach was examined.
• Digital Twins (DT) method, utilizing prognostic scores from historical data within an ANCOVA framework, was analyzed.

Main Results:

• Both BDB and DT aim to leverage historical data but possess distinct underlying mechanisms.
• A significant concern identified for BDB is the potential inflation of type 1 error rates.
• The tangible benefits of DT in actual randomized clinical trials require further empirical validation.

Conclusions:

• BDB and DT, despite similar goals, have critical differences impacting their suitability for specific RCTs.
• Type 1 error inflation is a notable drawback of BDB that necessitates careful management.
• Further research and evidence are needed to establish the practical value and reliability of DT in real-world clinical trial settings.