Bayesian decision sequential analysis with survival endpoint in phase II clinical trials

Area of Science:

• Biostatistics
• Clinical Trial Design
• Decision Theory

Background:

• Bayesian methods offer flexible approaches to clinical trial design.
• Traditional stopping rules for binary endpoints may not fully leverage event time information.
• Group sequential designs are essential for adaptive clinical trial management.

Purpose of the Study:

• To investigate the feasibility of computing exact, Bayesian, decision-theoretic time-to-event stopping rules.
• To apply these rules to single-arm group sequential non-inferiority trials.
• To compare Bayesian stopping rules with existing methods for clinical trial efficiency.

Main Methods:

• Utilizing a conjugate prior distribution and exponential failure time distribution.
• Employing backward induction to derive optimal Bayes stopping rules.
• Defining linear and threshold loss structures for decision-making.
• Computing frequentist operating characteristics (Type I error, power, run length).

Main Results:

• Optimal Bayes stopping rules were successfully computed for the specified trial design.
• The time-to-event approach demonstrated potential for earlier trial cessation.
• Frequentist operating characteristics were evaluated to assess rule performance.
• The methodology was shown to be feasible for practical application in clinical trials.

Conclusions:

• Bayesian time-to-event stopping rules are feasible and offer advantages for single-arm non-inferiority trials.
• This approach can lead to more efficient clinical trial designs by enabling earlier stopping.
• Further research into design issues and broader applicability is warranted.