Zeitschriftenartikel

New models for survival studies of twins

Abstract

In this paper we discuss applications of survival models to genetic epidemiological studies which further develop the ideas descibed by Yashin and Iachine (A.I.Yashin and I.A. Iachine, 1995, Genet. Epidemiol. 12, 529-538). A new family of correlated frailty models of bivariate survival for the genetic analysis of duration traits is suggested. This family exploits the properties of the three-parametric distribution of frailty which includes gamma, inverse Gaussian and positive stable distributions as particular cases. The semiparametric representations for the bivariate marginal survival functions associated with different frailty distributions are derived. Such representations are used in estimation procedures which do not require parametric assumptions about conditional survival distributions. Several such procedures are suggested and analysed using simulated data. The identifiability and other properties of these models are discussed. The statistical analysis of survival data on Danish twins is performed. The results of analysis for different models are compared. The models are used in the genetic studies of susceptibility to disease and death. The heritability of individual frailty is estimated for different models. The modification of these models in the case of observed covariates in the genetic studies of related individuals is considered. The applications of these models to the analysis of cross-sectional and longitudinal data are discussed. (AUTHORS)