Sample Size Calculation for Comparing Time-to-event Data

In clinical research, the occurrence of certain events (e.g., adverse events, disease progression, relapse, or death) is often of particular interest to the investigators, especially in the area of cancer trials. In most situations, these events are undesirable and unpreventable. In practice, it would be beneficial to patients if the test treatment could delay the occurrence of such events. As a result, the time-to-event has become an important study endpoint in clinical research. When the event is death, the time-to-event is defined as the patient’s survival time. Consequently, the analysis of time-to-event data is referred to as survival analysis. The statistical method for the analysis of time-to-event data is very different from those commonly used methods for other types of data, such as continuous and binary response, for two major reasons. First, timeto-event data are subject to censoring (e.g., right, left, or interval censoring). In other words, the exact value of the response is unknown; however, it is known that the value is larger or smaller than an observed censoring time or within an observed censoring interval. Second, the time-to-event data are usually highly skewed, which makes many standard statistical methods designed for normal data not applicable. In this article, the authors will focus on procedures for sample size calculation based on time-toevent data, which is subject to right censoring. Three commonly employed testing procedures, namely exponential model, Cox’s