Keywords: safety, restricted mean survival time, hazard ratio
Event-driven safety outcome trials are typically designed to rule out a risk margin using hazard ratios (HR). HR has limitations; it requires hazards to be proportional and does not incorporate information from non-events. Consequently, safety trials with low event rates often require a large sample size. Therefore, interest in using alternative measures in safety trials is increasing. We will present the pros and cons of alternative measures, such as restricted mean survival time (RMST) difference and risk difference (RD), compared to HRs. Also, we will present simulation results that assess the Type 1 error rate and power of HR, RMST difference and RD corresponding to different trial truncation time points, lengths of trial, background event rates and risk margins. Lastly, we apply RD and RMST difference to data from a cardiovascular outcome trial that used HR, and investigate the required sample size to obtain the same conclusion using RD and RMST difference.