|
Abstract:
|
Illness-death models provide a framework for characterizing the progression of a disease in the survival analysis. We propose a class of semiparametric models to study gap times between successive events under an illness-death model, in which some subjects experience ``illness'' then ``death'', while others experience only ``death''. The dependence between events is of interest, as is the marginal distribution of event times. In particular, death time after illness is censored by a dependent variable related to the occurrence time of illness, and is observable only if the first event has occurred. To accommodate such dependent censoring effect, a copula model is employed for the successive events and semiparametric transformation models are used for the marginal distributions. The estimation of model parameters are developed through the nonparametric maximum likelihood method and the optimization algorithms are presented to show the effectiveness of the computation. Simulations are conducted to demonstrate performances of the proposed analysis and an application of clinical study with chronic myeloid leukemia is reported to illustrate its utility.
|
