Bayesian adaptive model selection for optimizing group sequential clinical trials.

Area of Science:

• Biostatistics
• Clinical Trial Design

Background:

• Conventional clinical trial designs may lack accuracy when the proportional hazards assumption is violated.
• Substantial deviations in actual power from nominal values can occur under such conditions.

Purpose of the Study:

• To develop an optimal group sequential clinical trial design for right-censored event times.
• To ensure targeted false-positive rates and power across diverse event time distributions.

Main Methods:

• Combines Bayesian decision theory, Bayesian model selection, and forward simulation (FS).
• Employs adaptive model selection at interim analyses to choose optimal decision bounds.
• Utilizes right-censored event time data.

Main Results:

• The proposed Bayesian group sequential design maintains targeted power and false-positive rates.
• Simulation studies show the method performs comparably to or better than conventional designs.
• The adaptive design can lead to significantly smaller clinical trials.

Conclusions:

• The novel Bayesian approach offers a robust method for group sequential clinical trial design.
• Adaptive decision-making improves trial efficiency and reliability, especially when standard assumptions are unmet.
• This method provides a more flexible and potentially smaller trial design for event-time studies.