33 Nonparametric frequentist proposals for monitoring comparative survival studies

Abstract

This chapter discusses the nonparametric frequentist proposals for monitoring comparative survival studies. The chapter reviews several frequentist innovations for monitoring clinical trials. The frequentist viewpoint is vigorously challenged by Cornfield, who presents a Bayesian alternative. Anscombe, too criticized the frequentist approach, and presented a decision theoretic approach to monitoring. Over the past fifteen years, a number of methods for monitoring the results of clinical trials are developed in an effort to shorten the trials and reduce the numbers of patients required without greatly diminishing the quality of the scientific information gained. Clinical trials for comparing survival on two treatment groups are often costly and time-consuming, and they may impose psychological and physical burdens on participating patients. Yet these trials offer a uniquely sound methodology for acquiring new medical knowledge. To obtain convincing scientific evidence from such a trial requires that one follows a substantial number of patients for a sufficient time interval.