Control arm augmentation and hierarchical modeling in time-to-event trials: advantages and pitfalls

Area of Science:

• Biostatistics
• Clinical Trial Design
• Pharmacovigilance

Background:

• Borrowing external data in clinical trials can improve power but risks inflated Type I error rates with incompatible data.
• Traditional methods like power priors struggle with data heterogeneity, particularly in time-to-event (TTE) analyses with censored data.

Purpose of the Study:

• To develop a novel prior for TTE data that allows effective borrowing of information from external sources.
• To introduce a framework for incorporating external controls and borrowing treatment effect information between groups.

Main Methods:

• Development of the latent exchangeability prior specifically for TTE data.
• A new framework designed to integrate external control information and inter-group treatment effect borrowing.
• Simulation studies to evaluate operating characteristics under varying degrees of exchangeability.

Main Results:

• Borrowing information from external controls can enhance efficiency in clinical trials.
• Poor operating characteristics are observed when exchangeability assumptions are violated.
• The proposed latent exchangeability prior framework was applied to a real-world trial in metastatic colorectal cancer.

Conclusions:

• The latent exchangeability prior offers a method for leveraging external data in TTE clinical trials.
• Careful consideration of data exchangeability is crucial for reliable results when borrowing information.
• The approach shows potential for improving clinical trial efficiency, particularly in oncology settings.