Improving early phase oncology clinical trial design: A case study.

Area of Science:

• Clinical Trial Design
• Biostatistics
• Oncology Drug Development

Background:

• Early phase oncology trials often face challenges in dose selection.
• The traditional "3+3" design has limitations when dose-toxicity information is scarce.
• Model-assisted designs offer potential improvements for dose escalation studies.

Purpose of the Study:

• To present a first-in-human case study using a Bayesian Optimal Interval design.
• To evaluate the operating characteristics of the design, including amendments.
• To provide practical recommendations for designing early phase oncology trials.

Main Methods:

• Bayesian Optimal Interval design implemented for a first-in-human trial.
• Simulation studies to assess design performance under various scenarios.
• Analysis of study amendments and their impact on trial characteristics.

Main Results:

• The Bayesian Optimal Interval design demonstrated favorable operating characteristics.
• Simulations indicated the robustness of the design despite study amendments.
• Model-assisted designs were found to be superior to the "3+3" design in simulations.

Conclusions:

• Bayesian Optimal Interval designs are practical, flexible, and perform better than "3+3" designs.
• Model-assisted designs should be considered the standard for early phase oncology trials with limited dose-toxicity data.
• This case study offers valuable insights for optimizing early phase trial design.