Optimal futility stopping boundaries for binary endpoints

Area of Science:

• Clinical Trial Design
• Biostatistics
• Pharmaceutical Research

Background:

• Group sequential designs with futility stopping conserve resources in clinical trials.
• Simon's designs are common for binary endpoints in single-arm phase II studies but have limitations regarding false futility declarations and sample size efficiency.
• Optimizing futility boundaries is crucial for study performance, impacting power and accuracy of stopping decisions.

Purpose of the Study:

• To extend optimality criteria for futility boundaries to binary endpoints in single-arm phase II studies.
• To introduce an algorithm for deriving optimized futility boundaries.
• To compare the performance of these optimized boundaries against established Simon's designs and their modifications.

Main Methods:

• Developed an algorithm to derive optimized futility boundaries based on criteria by Schüler et al.
• Extended the approach to binary endpoints for single-arm phase II studies.
• Compared operating characteristics (e.g., sample size, power, futility error rates) with Simon's optimal and minimax designs and Kim et al.'s modified designs.

Main Results:

• Optimized futility boundaries maximize correct futility stopping while constraining power loss and wrong futility stops.
• Operating characteristics, including maximum and expected sample sizes, are comparable or superior to Simon's designs.
• Proposed boundaries are non-binding, offering flexibility compared to Simon's binding rules.

Conclusions:

• Futility boundary selection and interim analysis timing significantly impact study design performance.
• The proposed method offers a flexible, non-binding alternative to Simon's designs for phase II studies.
• Optimized futility boundaries minimize the probability of wrongly stopping for futility and avoid common drawbacks of Simon's designs.