Comparing and Summarizing Multiple Outcomes in a Clinical Trial

• Activity Number: 282
• Type: Topic Contributed
• Date/Time: Tuesday, August 5, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biometrics Section
• Author(s): Marc Buyse*+
• Companies: IDDI
• Keywords: Generalized pairwise comparison ; proportion in favor of treatment ; wilcoxon test ; gehan's generalized wilcoxon test

Abstract

A test and measure of treatment effect based on generalized pairwise comparisons

Marc Buyse, IDDI, Cupertino, CA and University of Hasselt, Belgium

We propose a test that extends the Mann-Whitney form of Wilcoxon's or Gehan's generalized test for two samples of (possibly censored) observations. The idea is to perform all pairwise comparisons between all patients in the treatment group and all patients in the control group. The observations can be individual patient outcomes for a single variable measured at a single time point, or a vector of outcomes for one or more variables repeatedly measured over time. A test statistic can be calculated using all pairwise comparisons if the variables and the measurement times can be prioritized in such a way that each pair of patients can be determined unequivocally to favor treatment, control, or neither. The advantage of the approach is that the variables can be of any type (discrete, continuous, time to event). A measure of treatment effect, called the "proportion in favor of treatment", can be calculated using all pairwise comparison, and will be shown to unify traditional measures of treatment effect.