The study of family traits and their impact on health outcomes has been a popular research area, particularly in the context of understanding how genetic similarities affect the risk of diseases within families. The kinship matrix measures the pairwise relatedness of individuals within the same family. Incorporating the kinship matrix into survival models for family data is difficult but this can be done through correlated frailty models. The popular method Coxme is based on a penalized partial likelihood approach but has never been tested for the analysis of complex pedigrees. Alternatively, a full likelihood approach can be proposed but is generally difficult to implement due to computational challenges. We compare these approaches using extensive simulations with or without ascertainment of the families through affected probands and also in an application to a series of families harboring BRCA1 mutations.