Revista de biometría y bioestadística

A Maximum-Type Association Test for Censored Time-to-Event Data

Abstract

Esther Herberich and Ludwig A Hothorn

Background: Testing the association between a diallelic marker and a censored time-to-event trait is a specific problem in population-based association studies. For a certain gene, the mode of inheritance may be of particular interest. Therefore, the principle of maximum-type tests (or minimum p procedure) is modified for continuous traits, especially for censored time-to-event data.
Results: We propose a Marcus-type multiple contrast test for a single censored time-to-event trait in a populationbased study assuming a Cox proportional hazard model. Using simulations we worked out the limitation of this asymptotic approach: sufficient sample sizes and non-rare alleles are required. A user-friendly implementation of this method is available in the survival and multcomp packages of the statistical software R.
Conclusions: The proposed approach can be used for the analysis of individual SNPs when censored time-to-event data in population-based association studies are of interest. The approach allows both a global claim of association and determination of the particular underlying mode of inheritance. The mode-specific hazard ratios and their lower simultaneous confidence limits provide information about statistical significance and genetic relevance.