Restricted mean survival time based on Wu-Kolassa estimator compared to Kaplan-Meier estimator

Abstract

Objective

The purpose of the paper is to introduce the Wu-Kolassa estimator (WKE) and the RMST based on it for its less biased estimation and substantial power gain, compared to the Kaplan-Meier estimator (KME), to researchers working in medical and health sciences to evaluate and compare patient survival times.

Results

The seven numerical studies showed that the power gain in WKE-based RMST analysis can reach more than 80 %, depending on the size of the study and the trend of failure rate. For the phase III study of iniparib plus gemcitabine and carboplatin (GCI) versus gemcitabine and carboplatin (GC) in patients with metastatic triple-negative breast cancer, GCI is superior to GC demonstrated by WKE-based RMST analysis (point estimate of treatment difference in RMST = 0.807; 95 % CI: 0.0214 to 1.592, 1.5 times higher than the estimate based on KME = 0.542; 95 % CI: −0.385 to 1.470). For the phase III trial that studied ovarian suppression (os) with tamoxifen or exemestane, tamoxifen plus os is superior to tamoxifen alone using WKE-based RMST (point estimate of difference = 0.228; 95 % CI: 0.016 to 0.439, more than 2-fold higher than the estimated difference based on KME = 0.113; 95 % CI: −0.008 to 0.234).

Conclusions

The Wu-Kolassa estimator is superior to the Kaplan-Meier estimator as it reduces the estimation bias and increases the power in the RMST analysis when the censoring rate is high or when the reference group has more censoring data than the test group.