Operational characteristics of generalized pairwise comparisons for hierarchically ordered endpoints.

Area of Science:

• Biostatistics
• Statistical Methods
• Clinical Trial Design

Background:

• The Wilcoxon-Mann-Whitney test is a standard non-parametric method for comparing two groups.
• Generalized pairwise comparisons (GPC) offer a multivariate extension for hierarchical endpoints.
• GPC accounts for endpoint correlations and uses a 'net benefit' measure.

Purpose of the Study:

• To evaluate the impact of endpoint correlation on the type-I error and power of the GPC test.
• To investigate the behavior of GPC with two hierarchically ordered endpoints.
• To understand how correlations influence the power of GPC in statistical analysis.

Main Methods:

• A simulation study was conducted using two hierarchical endpoints.
• The study assessed type-I error probability and statistical power under varying correlation scenarios.
• The generalized pairwise comparisons (GPC) method was applied.

Main Results:

• Including a secondary endpoint in hierarchical GPC analysis modifies the power for the primary endpoint.
• The change in power is dependent on the correlation between the hierarchical endpoints.
• Power can decrease irrespective of any marginal treatment effect on the secondary endpoint.

Conclusions:

• The overall power of hierarchical GPC procedures is complexly related to the variance-covariance structure of outcomes.
• Factors like clinical relevance thresholds, dropouts, and censoring schemes must be considered in trial design.
• Endpoint correlation is a critical factor influencing the performance of GPC statistical tests.