Sample size reestimation and Bayesian predictive probability for single-arm clinical trials with a time-to-event...

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Accurate sample size estimation is crucial for study power.
• Uncertainty in survival data distribution parameters complicates planning.
• Early stopping rules can improve clinical trial efficiency.

Purpose of the Study:

• To evaluate the internal pilot study (IPS) approach for sample size reestimation in Weibull survival analysis.
• To present Bayesian predictive probability methods for interim analyses in single-arm trials with unknown Weibull shape parameters.
• To provide evidence for early trial termination based on efficacy or futility.

Main Methods:

• Investigated the utility of the internal pilot study (IPS) approach for sample size adjustment.
• Developed Bayesian predictive probability calculations for interim analyses.
• Utilized Weibull distribution for time-to-event endpoints.
• Proposed methods using posterior mode or the full posterior distribution of the shape parameter.

Main Results:

• The IPS approach can rescue study power but may double the required sample size.
• Bayesian predictive probability calculations are feasible even with an unknown Weibull shape parameter.
• Incorporating the entire posterior distribution of the shape parameter is recommended to manage uncertainty.

Conclusions:

• The IPS approach offers flexibility but requires careful consideration of practical constraints.
• Bayesian predictive probability provides a robust framework for interim decisions in clinical trials with Weibull survival data.
• Accounting for shape parameter uncertainty is vital for reliable interim analyses.