Improved methods of estimating survival probabilities applied to renal transplant data

In the evaluation of clinical studies of different kinds with survival time as the response variable to be analysed the estimation of survival probabilities plays an important role. The ordinary procedure in survival data analysis for estimating survival probabilities is the Kaplan-Meier product-limit estimator. However, in the case of heavy censoring or if the largest observed failure times are censored the product-limit method is known to be a biased estimator of the survival function. Recently, two improved methods of estimating survival functions, a semiparametric procedure and an approach using splines, were proposed (Klein JP, Lee SC and Moeschberger ML, Biometrics, 46 (1990) 795-811; Whittemore AS and Keller JB, Biometrics, 42 (1986) 495-506). These new methods are less biased than the product-limit estimator, especially for heavily censored data. A computer program based on the integrated statistical and graphical software package RS/1 was developed for the calculation and graphical representation of the new estimators. Their improved properties are illustrated by the application to renal transplant data.