Improved Estimation of Restricted Mean Survival

• Activity Number: 403
• Type: Contributed
• Date/Time: Tuesday, July 31, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Nabihah Tayob*+ and Susan Murray
• Companies: and University of Michigan
• Keywords: Restricted mean survival ; Nonparametric ; Correlated data

Abstract

To date, a single year of data is used to estimate the number of days lived during a single year of follow-up. Our aim is to develop a nonparametric method to estimate the one-year restricted mean survival that incorporates additional follow-up time collected after the first year. We propose enriching the dataset with follow-up collected after the first year by introducing multiple observations per subject, corresponding to their survival experience in a series of observation windows with different starting times. This induces a correlated data structure which must be accounted for in the estimation of the variance. In simulation studies we compare different methods of variance estimation for the correlated data structure to determine which has the best performance. We also explore graphical methods for determining when additional follow-up data is useful and appropriate to use. In cases where additional follow-up is appropriate to use, preliminary study of our estimates show substantial improvement over existing methods. Additionally, we study the gains in precision associated with different choices of follow-up windows to aid medical researchers in optimal study design.