Closed-Form Approximation of Correlation Matrix Among Fleming Harrington Test Statistics in MaxCombo Test: Comments on “Robust Design and Analysis of Clinical Trials With Nonproportional Hazards: A Straw Man Guidance From a Cross-Pharma Working Group”

Abstract

Roychoudhury et al. recently proposed a robust MaxCombo test that considers the maximum of four “correlated” test statistics from FH(0, 0), FH(0, 1), FH(1, 0), and FH(1, 1) tests. Their formulation needs estimation of correlations of these test statistics, and they obtained the correlation matrix via simulation. I have described a closed-form solution to obtain this correlation matrix under the “no censoring” assumption that depends only on sample size and planned number of events. This approach has been extended for group-sequential design as well. Our correlation matrix is very close to the one obtained by Roychoudhury et al. via simulation. In addition, we have also made remarks on the work of Roychoudhury et al.

Keyword:

• Closed-form solution
• Correlation matrix
• Fleming–Harrington test
• Group-sequential
• MaxCombo test
• Nonproportional hazards

Funding

The author(s) reported there is no funding associated with the work featured in this article.