Graphical approaches using a Bonferroni mixture of weighted Simes tests

Area of Science:

• Biostatistics
• Clinical Trial Design
• Statistical Inference

Background:

• Graphical approaches to multiple testing procedures offer flexibility and ease of communication, particularly in clinical trials.
• The R package gMCP facilitates these graphical methods, but lacks specific functionalities for certain complex hypothesis structures.

Purpose of the Study:

• To describe the calculation of adjusted p-values for nonparametric testing procedures based on a Bonferroni mixture of weighted Simes tests.
• To optimize the generation of weights for intersection hypotheses within graph-based multiple testing procedures to reduce computation time.
• To demonstrate the validity of the proposed Simes test approach across various data types and clinical trial scenarios.

Main Methods:

• Development and description of adjusted p-value calculations for a Bonferroni mixture of weighted Simes tests.
• Optimization of weight generation for intersection hypotheses in graph-based multiple testing procedures.
• Validation of the Simes test for diverse data types (normal, binary, count, time-to-event) and testing scenarios (treatment vs. control, noninferiority, superiority, subgroup analysis).

Main Results:

• The study details methods for calculating adjusted p-values not currently available in the gMCP package.
• Optimized weight generation significantly reduces computing time for simulation-based power calculations.
• The Simes test approach is shown to be valid for various clinical trial comparisons and data types.

Conclusions:

• The proposed enhancements to graphical multiple testing procedures provide valuable tools for clinical trial design and analysis.
• The optimized methods improve efficiency in statistical power calculations, enabling more robust trial designs.
• The validated Simes test approach offers a flexible and powerful method for complex hypothesis testing in confirmatory clinical trials.