A multistate model for the genetic analysis of the ageing process

In this paper a multivariate frailty model is suggested that can be used in the genetic analysis of the ageing process as a whole, simplified to consisting of the states 'healthy', 'disabled' and 'deceased'. The model allows us to evaluate simultaneously the relative magnitude of genetic and environmental influences on frailty variables corresponding to the period of good health and to the life span. The frailty variables can be interpreted as susceptibility to illness or death. The model can be applied to data on groups of related individuals (twins, siblings, a litter). One of the major advantages of this model is that it allows one to include groups of individuals where some or all members of the group are already deceased at the time of observation. The current health status of the living individuals and the exact life span of individuals who are already deceased is the only information necessary for the application of the model. Questions concerning the identifiability of the model based on current health status data and estimation strategies are discussed in the context of specifying the model for twins. Finally, the results of a sample analysis of twin data on prostate cancer are presented.