Weighted Multiple Testing Correction for Correlated Endpoints in Survival Data

Abstract

Multiple correlated time-to-event endpoints often occur in clinical trials and some time-to-event endpoints are more important than others. Most weighted multiple testing adjustment methods have been proposed to control family-wise type I error rates either only consider the correlation among continuous or binary endpoints or totally disregard the correlation among the endpoints. For continuous or binary endpoints, the correlation matrix can be directly estimated from the corresponding correlated endpoints. However, it is challenging to directly estimate the correlation matrix from the multiple endpoints in survival data since censoring is involved. In this chapter, we propose a weighted multiple testing correction method for correlated time-to-event endpoints in survival data, based on the correlation matrix estimated from the WLW method proposed by Wei, Lin, and Weissfeld. Simulations are conducted to study the family-wise type I error rate of the proposed method and to compare the power performance of the proposed method to the nonparametric multiple testing methods such as the alpha-exhaustive fallback (AEF), fixed-sequence (FS), and the weighted Holm-Bonferroni method when used for the correlated time-to-event endpoints. The proposed method and others are illustrated using a real dataset from Fernald Community Cohort (formerly known as the Fernald Medical Monitoring Program).