"Estimating the Average Effect on the Survival Function of a Time-Dependent Treatment "
Douglas Schaubel, Ph.D. , University of Michigan
January 19, 2010 @ 3:30 pm - 4:30 pm
Location: BRB 253
Biostatistics
Title: Estimating the Average Effect on the Survival Function of a
Time-Dependent Treatment
Abstract:
In several areas of medicine, the treatment of interest is time-dependent. For example, patients with end-stage renal disease typically receive dialysis for a period of time before undergoing kidney transplantation. When the time to failure (e.g., death) is potentially censored, analyses of such data has traditionally involved Cox regression using time-dependent treatment indicators, with non-proportionality addressed through a time-dependent treatment effect. However, investigators usually prefer treatments to be contrasted in terms of survival (not hazard) functions; particularly in cases where the treatment effect on the hazard function is not constant over time. We develop semiparametric methods to estimate the effect of a time-dependent treatment on survival and restricted mean residual lifetime. Asymptotic properties of the proposed estimators are derived, with finite-sample characteristics evaluated through simulation. The proposed methods are applied to end-stage renal disease data from a national organ transplant registry. This is joint work with Jack Kalbfleisch.
