Graph Based, Adaptive, Multiarm, Multiple Endpoint, Two-Stage Designs.

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Methodology

Background:

• The graph-based approach offers an intuitive method for hierarchical multiple testing and type-1 error propagation.
• Traditional graph-based methods were developed for single-stage, nonadaptive designs.

Purpose of the Study:

• To extend the graph-based approach to adaptive two-stage clinical trial designs.
• To evaluate methods for preserving familywise error rate in adaptive designs.

Main Methods:

• Extension of the graph-based approach to accommodate two-stage adaptive designs.
• Implementation and comparison of the p-value combination method and the conditional error rate method.
• Large-scale simulation experiment to compare operating characteristics.

Main Results:

• The graph-based approach can be successfully extended to complex adaptive two-stage designs.
• Both the p-value combination and conditional error rate methods offer valid strategies for familywise error rate control.
• Simulation results provide insights into the operating characteristics of each method under adaptive scenarios.

Conclusions:

• Adaptive two-stage designs can be effectively managed using the extended graph-based multiple testing framework.
• The choice between p-value combination and conditional error rate methods depends on specific design goals and operating characteristics.
• This work provides a robust statistical framework for adaptive clinical trial design and multiple testing.