Weighted multiple testing correction for correlated tests.

Area of Science:

• Biostatistics
• Clinical Trial Methodology
• Statistical Significance

Background:

• Clinical trials frequently involve multiple, correlated endpoints.
• Existing methods for controlling family-wise type I error rate (FWER) often ignore endpoint correlations, potentially reducing statistical power.
• Current methods like Bonferroni, Holm, and fixed-sequence procedures have limitations in handling correlations or computational complexity.

Purpose of the Study:

• To propose a novel weighted multiple testing correction method designed for correlated endpoints.
• To address the limitations of existing methods, particularly computational difficulties and suboptimal power.
• To enhance the statistical power of hypothesis testing in clinical trials with multiple correlated endpoints.

Main Methods:

• Development of a weighted multiple testing correction algorithm.
• Utilized the 'mvtnorm' package in R for computational implementation, enabling scalability to numerous endpoints.
• Comparative analysis through simulations against established methods like Flexible Fixed-Sequence (FFS) and Alpha-Exhaustive Fallback (AEF).

Main Results:

• The proposed weighted method demonstrates higher power for the first hypothesis compared to FFS and AEF.
• It retains the advantage of FFS and AEF by providing high power for subsequent hypotheses in the testing sequence.
• Achieves superior power for individual hypotheses over the weighted Holm procedure, particularly when endpoint correlations are high.

Conclusions:

• The proposed weighted multiple testing correction is an effective approach for correlated endpoints in clinical trials.
• It offers improved statistical power across all tested hypotheses, surpassing existing methods.
• The method's computational efficiency allows application to a large number of endpoints.