Confidence interval for the difference between two median survival times with semiparametric transformation models

Abstract

In medical studies, we usually are interested in comparing the treatment effects of the drug according to the difference of two median survival times. In this paper, we consider the problem of constructing conditional confidence interval for the difference of two median survival times given the covariates under a general class of the semiparametric transformation models with right-censored data. We propose two methods for constructing the conditional confidence intervals. One is based on the estimating equations (EE) estimator of Chen, Jin and Ying’s (2002) and the other on the nonparametric maximum likelihood estimator of Zeng and Lin (2006). Simulation results indicate that both methods provide satisfactory coverages for finite sample. We illustrate the proposed method using a real data set in a two-arm non-small cell lung cancer study.

Keywords:

• Confidence interval
• Estimating equation
• Maximum likelihood
• Median survival time
• Semiparametric transformation model
• Weighted Breslow

Acknowledgements

This work was supported by the Ministry of Science and Technology of Taiwan under Grant MOST 105-2118-M-029-002.