A phase II, seamless single-arm to two-arm Bayesian design for a time-to-event endpoint

Area of Science:

• Clinical Trials Methodology
• Bayesian Statistics in Oncology
• Biostatistics

Background:

• Phase II oncology trials frequently encounter patient recruitment challenges and resource limitations.
• This often necessitates the use of single-arm designs lacking concurrent controls.
• Time-to-event endpoints are critical in oncology but pose analytical challenges in early-phase trials.

Purpose of the Study:

• To introduce a novel two-stage Bayesian design for phase II oncology trials.
• To bridge the gap between single-arm and randomized two-arm trial designs for time-to-event data.
• To optimize design parameters for minimizing expected sample size under the null hypothesis.

Main Methods:

• A two-stage Bayesian design incorporating an initial experimental treatment stage with futility monitoring.
• A randomized second stage comparing experimental and control arms, initiated based on first-stage progression.
• Optimization of design parameters to minimize expected sample size under the null hypothesis of no treatment effect.

Main Results:

• Extensive simulations show consistently lower expected sample sizes compared to conventional two-arm trials using log-rank tests.
• The proposed design achieves reduced maximum sample sizes versus sequential single-arm and two-arm trial implementations.
• Demonstrated practical implementation through an illustrative application in extensive-stage small cell lung cancer.

Conclusions:

• The novel two-stage Bayesian design offers a resource-efficient approach for phase II oncology evaluations.
• This methodology streamlines trial processes and expedites clinical decision-making for novel cancer treatments.
• The design provides a robust framework for analyzing efficacy by integrating data from both trial stages.