ABSTRACT
The chapter describes sample size calculations for clinical trials and the study is concerned with investigating the survival experience of the patients. Usually, this survival experience is expressed in terms of survival status (e.g. alive or dead; recurred or recurrence free) and survival time (e.g. time to death; time to recurrence).
It is highlighted how if the event of interest was observed in all subjects, then the analysis, and hence design, would be relatively straightforward, as we would have a continuous primary endpoint with continuous methodologies applicable. However, most studies usually finish some fixed time after the start of the study such that the event of interest is not observed in all subjects. For this reason, specific methods for survival data need to be applied to estimate the sample size.
The chapter further highlights how the planned survival analysis, and hence sample size calculation, depends on whether the event of interest is negative (e.g. death), where a proportional hazard approach would be applied, or positive (e.g. cure), where an accelerated failure time approach would be applied. For the former, it is desirable to delay the time to when the event occurs and a formal statistical test would be done through a Log-rank test. The chapter describes sample size calculations, where the statistical test will be a log-rank test.
The concentration in the chapter is on sample size calculations that estimate the number of events in the study and not the number of patients.
