An Approach to Design Adaptive Clinical Trials With Time-to-Event Outcomes Based on a General Bayesian Posterior...

Area of Science:

• Clinical Research
• Biostatistics
• Medical Informatics

Background:

• Adaptive clinical trials offer ethical and efficiency advantages over standard designs.
• Current adaptive trial design relies on simulations with potentially misspecified data-generating processes.
• Misspecification can lead to suboptimal trial performance, impacting statistical power and error rates.

Purpose of the Study:

• To propose a novel approach for designing adaptive clinical trials with time-to-event outcomes.
• To develop a method that avoids explicit definition of the data-generating process.
• To enhance the robustness and reliability of adaptive trial designs.

Main Methods:

• Utilized a general Bayesian framework for trial design.
• Employed partial likelihood for robust inference on treatment effects.
• Designed adaptive trials with implicitly defined data-generating processes.

Main Results:

• Demonstrated the benefits of the proposed approach through an illustrative example.
• Successfully redesigned a motivating clinical trial using the new methodology.
• Showcased robustness to the baseline hazard function's form.

Conclusions:

• The proposed Bayesian approach facilitates adaptive clinical trial design for time-to-event outcomes without explicit data-generating process assumptions.
• This method improves robustness and efficiency in adaptive trial design.
• The approach is applicable to real-world clinical trial scenarios, including vaccine trials.