A Comparison of Two Methods for Adaptive Multi-Arm Two-Stage Design

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methods

Background:

• Group sequential randomized clinical trials allow for interim analyses and adaptive design modifications.
• Comparing multiple treatment arms to a common control requires robust statistical methods to maintain error rates.

Purpose of the Study:

• To evaluate and compare two statistical procedures for analyzing two-stage group sequential randomized clinical trials.
• To assess the performance of the p-value combination method and the conditional error rate method in controlling family-wise error rate (FWER) and maximizing statistical power.

Main Methods:

• The study considers a two-stage group sequential design with adaptive sample size adjustments and interim arm dropping.
• Two methods, p-value combination and conditional error rate, were theoretically discussed and compared using simulation studies.
• The family-wise error rate (FWER) was controlled for both procedures.

Main Results:

• Both the p-value combination method and the conditional error rate method effectively controlled the family-wise error rate (FWER).
• The conditional error rate method demonstrated superior statistical power across various scenarios and alternative hypotheses compared to the p-value combination method.

Conclusions:

• The conditional error rate method is a more powerful approach for identifying efficacious treatments in two-stage group sequential randomized clinical trials.
• The findings suggest that the conditional error rate method should be considered for its enhanced ability to detect treatment effects while maintaining statistical rigor.