A predictive probability design for phase II cancer clinical trials

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Cancer Research

Background:

• Phase II cancer trials commonly use rigid two- or three-stage designs.
• These designs allow early termination for inefficacious treatments but require fixed patient evaluations.
• Rigidity in traditional designs can hinder adherence due to fixed evaluation points.

Purpose of the Study:

• To develop an efficient and flexible clinical trial design.
• The goal is to create a design with desirable statistical properties for phase II cancer studies.

Main Methods:

• A novel flexible design was constructed using Bayesian predictive probability and the minimax criterion.
• A three-dimensional search algorithm was implemented for determining optimal design parameters.

Main Results:

• The new design effectively controls Type I and Type II error rates.
• It allows for continuous monitoring, enabling earlier trial termination for non-efficacious treatments and reducing expected sample size.
• Simulation studies confirm the predictive probability design's good operating characteristics.

Conclusions:

• The predictive probability design offers enhanced efficiency and adaptability compared to traditional multi-stage designs.
• It demonstrates robustness in controlling error rates even with deviations from the original plan.
• While computationally intensive and potentially yielding biased efficacy estimates, it is easier to implement with provided S-PLUS/R programs.