A1585
Title: Inferring median survival under dependent censoring using the copula graphic estimator
Authors: Takeshi Emura - Hiroshima University (Japan) [presenting]
Dennis Dobler - RWTH Aachen University (Germany)
Abstract: The median is a popular nonparametric measure to calibrate patient survival and represents clinical benefit of treatments. A key difficulty in estimation of median survival is handling right censoring. Existing inference methods have been developed under the fundamental assumption of independent censoring, which is often too strong for clinical applications. Novel statistical methods are proposed for inferring median survival when data are subject to dependent censoring. To accommodate dependent censoring schemes, the copula graphic estimator is utilized leading to the estimation and significance test for medians in the one sample and two sample settings. A Wald test for a general factorial design is also developed. Simulation studies are performed to investigate the performance of the proposed methods and to check the reliability of the R package. A real dataset on gastric cancer patients is analyzed to illustrate the proposed methods for median survival.
