CMStatistics 2017: Start Registration
View Submission - CMStatistics
B1083
Title: Restricted residual mean lifetime and competing risks Authors:  Giuliana Cortese - University of Padua (Italy) [presenting]
Abstract: The typical modeling approach in time-to-event analysis is to consider regression on the hazard function, the main example being the Cox proportional hazards model. Some crucial issues about hazards models have been raised in the recent literature, and currently, there is increasing interest in global summary measures based on the survival function, such as the mean lifetime and the residual mean lifetime. Different regression models for these key measures have been proposed, following either a direct or indirect modeling approach. We present novel regression models based on the residual mean lifetime, and related inference, in presence of right-censoring and left-truncation, and a competing risks structure. We follow a direct approach based on generalized estimating equations, combined with either the IPCW technique or the pseudo-observations technique, to handle incomplete data. In the competing risks setting, the typical key quantity of interest is the cumulative incidence function (CIF). Global summary measures of CIFs would be highly beneficial to provide a clear and direct quantification of between-group differences. We explore different approaches to modeling cause-specific residual mean lifetimes and related measures in presence of competing risks. In particular, we provide a direct competing risks regression model that can handle time-dependent regression coefficients and covariates. The methods and models will be illustrated with applications to clinical data.