Joint modelling of two sequential times to event with longitudinal information

Abstract

In survival analysis, the lifetimes may be observed in some speci¯ed order, where the time to event Tk, cannot be observed until T1; :::; Tk¡1 have been observed. The present work proposes a joint model of two sequential times to events together with longitudinal information, extending the joint model of Wolfsohn and Tsiatis (1997) for one time to event and one longitudinal variable. We apply the model to the clinical trial called TIBET, in which an intermit- tent therapeutic strategy has been assigned to each patient. Of special clinical interest is the lifetime that a patient needs before restarting treatment given the progression of biological markers recorded during the followup period.