Cox regression analysis of dependent interval-censored failure time data

Many procedures have been proposed for regression analysis of interval-censored failure time data arising from the Cox or proportional hazards model. However, most of these existing methods only apply to the situation where the censoring mechanism generating censoring intervals is independent of the failure time of interest, and it is well-known that sometimes this may not be true in practice. To address this issue, a new approach that allows the dependence between the censoring mechanism and the failure time is proposed. More specifically, a situation where the dependence is through the length of censoring intervals is considered as it is often the case in follow-up studies. The asymptotic properties of the proposed estimators are established and the numerical studies are conducted for the assessment of the finite sample properties of the estimators.