A randomized two-stage design for phase II clinical trials based on a Bayesian predictive approach.

Area of Science:

• Clinical Trials
• Biostatistics
• Oncology

Background:

• Phase III oncology trials exhibit high failure rates, often linked to insufficient Phase II studies.
• Randomized designs in Phase II trials are increasingly recommended to overcome limitations of historical controls.

Purpose of the Study:

• To propose a novel two-arm, two-stage randomized design for Phase II oncology trials.
• To enhance the probability of detecting a truly more effective experimental treatment using a Bayesian predictive approach.

Main Methods:

• A Bayesian predictive approach is utilized for a two-arm, two-stage trial design.
• The design is a randomized adaptation of Sambucini's two-stage single-arm design.
• Comparison with Jung's minimax and optimal designs is performed.

Main Results:

• The proposed design aims to provide substantial posterior evidence for experimental treatments.
• Analysis of design features across varying parameters is conducted.
• An illustrative example is provided as supplementary material.

Conclusions:

• The novel Bayesian randomized design offers a promising approach to improve Phase II oncology trial efficiency.
• This method can potentially reduce the high failure rates observed in subsequent Phase III trials.
• The design provides a robust framework for evaluating experimental cancer therapies.