Regression analysis of current status data in the presence of dependent censoring with applications to tumorigenicity experiments

Current status data frequently occur in many fields including demographic studies and tumorigenicity experiments. In these cases, the censoring or observation time may be correlated to the failure time of interest, the situation that is often referred to as dependent or informative censoring. Although several semiparametric methods have been developed in the literature for the situation, they either only apply to limited situations or may be computationally unstable. To address these, a frailty model-based maximum likelihood approach is proposed with the use of monotone splines to approximate the unknown baseline cumulative hazard function of the failure time. Also a novel EM algorithm, which is based on a three-stage data augmentation and can be easily implemented, is presented. The proposed estimators are proved to be consistent and asymptotically normally distributed. An extensive simulation study is performed to assess the finite sample performance of the proposed approach and suggests that it works well for practical situations. An application to a tumorigenicity study is provided.