Family-based association tests for survival and times-to-onset analysis

In this paper, we discuss family-based association test (FBATs) relating genetic data to survival and time-to-onset data. We show how the standard logrank and Wilcoxon statistics can be used with family data to develop tests of association. We prove that the FBAT-logrank approach can be identical to the proportional hazard approach discussed in Mokliatchouk et al. (2000). Further, using simulation studies, we compare the power of the logrank, Wilcoxon and an approach developed for censored exponential data (Euro J Hum Gen 2001; 9:301-306). Based on the results of the simulation study, we suggest rules of thumb about which statistics to use in a given situation. An application of all three tests to an Alzheimer study illustrates the practical relevance of our discussion.