Global statistical tests for comparing multiple outcomes in rheumatoid arthritis trials

Publisher Website
Google Scholar
Abstract
Objective, To evaluate global statistical tests (GSTs) of treatment effectiveness for rheumatoid arthritis (RA) trials measuring multiple outcomes. Methods. Using outcome measures from American College of Rheumatology (ACR) core set variables available in 3 RA trials, GSTs were calculated using the O'Brien ranking procedure and a procedure for binary data. GSTs take correlations among outcomes into account. Power calculations using 1 trial data set provide comparisons of GSTs and ACR criteria for improvement. Results. Spearman correlations among outcomes ranged from 0.21 to 0.73. Erythrocyte sedimentation rate had the lowest correlation with other outcomes in all 3 trials. Within a trial, joint swelling and joint tenderness or patient and physician assessment had the highest correlations, depending on the trial. Results were consistent with results using the ACR criteria, although the GST was more powerful. Conclusion. GSTs are a useful tool for comparing treatment effects across multiple clinically meaningful outcome measures, The GST allows easy inclusion of validated, reliable new measures that are not a part of ACR criteria, such as quality of life, and can be computed with or without selecting a cutoff point defining patient improvement. GSTs should be considered for rheumatic disease treatment trials.