Identification of functional variants is one of the crucial steps to dissect the genetic mechanism of complex diseases. Family-based designs enriched with affected subjects and disease associated variants can increase statistical power for identifying functional rare variants. However, few rare variant analysis approaches are available for time-to-event traits in family designs. We developed novel pedigree-based burden and kernel association tests for time-to-event outcomes with right censoring for pedigree data, referred to FamRATS (Family-based Rare variant Association Tests for Survival traits). Cox proportional hazard models were employed to relate a time-to-event trait with rare variants (MAF < 0.05) with flexibility to encompass all ranges and collapsing of multiple variants. Robustness of violating proportional hazard assumptions was investigated for the proposed and four current existing tests including the conventional populationbased Cox proportional model and the burden, kernel and sum of squares statistic (SSQ) tests for family data. The proposed tests can be applied to large-scale whole-genome sequencing data. They are appropriate for the practical use under a wide range of misspecified Cox models, as well as for populationbased, pedigree-based or hybrid designs. In our extensive simulation study and data example, we showed that the proposed kernel test is the most powerful and robust choice among the proposed burden test and the existing four rare variant survival association tests. When applied to the Diabetes Heart Study, the proposed tests identified exome variants of the JAK1 gene on chromosome 1 to be associated with age at onset of type 2 diabetes (P < 4.82 × 10 – 5 ).
