32 – Recent Advances in Survival Analysis
Assessing Life Extension from Medical Interventions
Javier Cabrera
Rutgers University
Jerry Cheng
Robert Wood Johnson Medical School
John Kostis
Robert Wood Johnson Medical School
Kezhen Liu
Rutgers University
Comparing outcomes assessing performance of several types of treatments or interventions is an important task in clinical trials as well as in observational studies. Among various measurements in assessing life extension, the gain in life expectancy is one of performance measurements of interest. In this paper, we propose a framework for estimating this quantity by calculating the area between estimated survival curves from two comparative treatments respectively, for example, active treatment and control. We estimate the survival curves �rst via the non-parametric Kaplan-Meier estimator to reflect the observed survival probabilities in the study. We then use semi-parametric Cox proportional hazard model and obtain the direct adjusted survival curves. By doing this, we can adjust for any imbalance of covariates between the two treatments. In order to assess the variability of our estimate, we propose a new Bootstrap method for obtaining a bootstrap con�dence interval for this quantity. We also propose the corresponding bootstrap testing procedure to test the null hypothesis that two treatments have the same expected survival. We conduct simulation studies to evaluate the effectiveness of this method and use it in a real data application.