ABSTRACT
To avoid a parametric survival distribution assumption, a proportional hazards (PH) model is usually used to derive the sample size calculation with the well-known nonparametric log-rank test. This chapter illustrates proportional hazards model by following the lung cancer clinical trial. It assumes that there are two treatment groups, the control and experimental groups, which are designated by groups 1 and 2, respectively. The chapter derives the Schoenfeld formula under the proportional hazards model. Among the available sample size formula, the Schoenfeld formula is widely used in clinical trial design. The Schoenfeld formula has been implemented in the commercially available software EAST 6, but its use is limited to the piecewise exponential model. The chapter also illustrates how the sample size can be calculated under various parametric and nonparametric proportional hazards models. Zhang and Quan discussed sample size calculation for the log-rank test with a time lag in the treatment effect.
