A Model for the Joint Estimation of Interval-Censored Progression Time and Survival

• Activity Number: 387
• Type: Topic Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Dianne M Finkelstein*+ and David A Schoenfeld
• Companies: Massachusetts General Hospital/Harvard University and Harvard Medical School
• Keywords: interval censored ; cinical trial ; joint failure time model ; progression ; survival

Abstract

Cancer clinical Trials often assess efficacy by comparing progression-free survival (PFS). This outcome is the minimum of the first time progression is detected and mortality. This analysis has several issues. Disease progression is known to have occurred in the interval between examinations while survival time is known exactly or is right censored. Also, survival time could provide information on when progression actually occurred. We suggest instead using a joint model for progression and death where the effect of covariates on the hazard of progression, and of death after progression obey a proportional hazards model, with a piecewise constant underlying hazard. The model can be used to test whether treatment improves each of the following: the disease free interval, survival, time to progression in surviving patients and survival time in non-surviving patients. Finally we can test if treatment improves time to progression and does not worsen survival. Since the effect of covariates is parametric, all of these inferences can take into account prior knowledge.