Title: A progressive three-state model for the restricted residual mean time
Authors: Giuliana Cortese - University of Padua (Italy) [presenting]
Davide Serafin - University of Padova (Italy)
Abstract: The residual restricted mean lifetime (RRML) is an alternative robust summary measure that captures a global profile of the survival curve over $[t, t_0]$ by the integration of the survival function. This measure describes the remaining life expectancy up to $t_0$ of an individual surviving up to a specific time $t$, and it permits to make easier and better assessments from a clinical point of view than those based on the hazard ratio from a proportional hazards model. Direct regression models for RRML have gained attention recently, and different inferential approaches have been proposed to handle both right censoring and left truncation. However, the multi-state setting has not yet been fully explored in regression models for the RRML. Therefore, a progressive three-state model is presented to handle situations where subjects may experience two different events, involving three possible states visited by a subject. For each transition, we assess the effect of covariates on the residual mean lifetime scale rather than on transition intensities. The inferential procedure underlying the proposed model, based on IPCW weights, is investigated by simulation studies. An application on cutaneous melanoma is illustrated to evaluate the effects of possible biomarkers on the residual mean lifetime to tumor recurrence and death after recurrence.