SAMPLE SIZE CALCULATION FOR COMPARING （为比较样本大小的计算）.pdf

SAMPLE SIZE CALCULATION FOR proportional hazard model, and the log-rank COMPARING TIME-TO-EVENT DATA test, are discussed. The remaining of this article is orga- HANSHENG WANG nized as follows. In Section 2, sample size Guanghua School of Management, formulae based on exponential model are Peking University derived. In Section 3, Cox’s proportional haz- Department of Business Statistics ard model is discussed. In Section 4, sample Econometrics size formulae based on log-rank test are pre- Beijing, P. R. China sented. For each model, the sample size SHEIN-CHUNG CHOW formulae for testing equality, superiority, Millennium Pharmaceuticals, Inc. and equivalence/non-inferiority are derived. Cambridge, Massachusetts Finally, this article concludes with a general discussion. 1 INTRODUCTION 2 EXPONENTIAL MODEL In clinical research, the occurrence of certain events (e.g., adverse events, disease progres- In this section, the sample size formulas sion, relapse, or death) is often of particular for testing equality, superiority, and equiv- interest to the investigators, especially in the alence/noninferiority are derived based on area of cancer trials. In most situations, these exponentially distributed time-to-event data. events are undesirable and unpreventable. In