Generalized pairwise comparisons of prioritized outcomes in the two-sample problem

Area of Science:

• Biostatistics
• Statistical Methods
• Nonparametric Statistics

Background:

• The Wilcoxon-Mann-Whitney test is a standard nonparametric method for comparing two independent groups.
• Existing methods often focus on single outcome variables, limiting comprehensive analysis when multiple outcomes are present.

Purpose of the Study:

• To extend the U-statistic concept for generalized pairwise comparisons between two groups.
• To introduce a general measure of group difference, the 'proportion in favor of treatment' (Δ).
• To demonstrate the utility of generalized pairwise comparisons in analyzing complex data from randomized clinical trials.

Main Methods:

• Extension of the U-statistic for generalized pairwise comparisons.
• Application to single or multiple prioritized outcome variables of various types (discrete, continuous, time-to-event).
• Illustration using data from two randomized clinical trials.

Main Results:

• Generalized pairwise comparisons were shown to encompass well-known nonparametric tests.
• The method effectively analyzes data with multiple, prioritized outcomes.
• A general measure of group difference, Δ, was derived and related to traditional effect measures.

Conclusions:

• Generalized pairwise comparisons offer a flexible and powerful approach for analyzing differences between two groups, especially with multiple outcomes.
• The 'proportion in favor of treatment' (Δ) provides a valuable, interpretable measure of treatment effect.
• This methodology enhances the analysis of data from clinical trials and other comparative studies.