Title: Dynamic prediction for time-to-death under the joint frailty-copula model
Authors: Takeshi Emura - Chang Gung University (Taiwan) [presenting]
Abstract: A copula-based approach is considered to construct dynamic prediction formulas of predicting time-to-death for a patient. The suggested prediction formulas incorporate the genetic and clinical covariates collected on the patient entry as well as time-to-tumour progression (TTP) available after the entry. We first review the framework of dynamic prediction by introducing prediction formulas, such as the conditional failure function and conditional hazard function. We then demonstrate how the parameters in the prediction formulas are estimated by fitting semi-competing risks data to a model. The joint frailty-copula model is employed for the marginal Cox regression models on TTP and time-to-death as well as for the copula model between the two endpoints. For illustration, we derive dynamic prediction formulas to predict time-to-death for ovarian cancer patients.