Accrual by groups in trials with discrete-time survival endpoints

Abstract

In many fields of science, event status is often recorded in intervals or at discrete points in time and can be investigated in experimental settings. Conducting such trials requires thorough planning before they are actually performed. To investigate accrual by groups in a trial with discrete-time survival endpoints and to describe how to choose the number of accrual groups, the size of the accrual groups, and the duration of the trial to achieve a sufficient power level. In trials with multiple time periods, the event status is recorded at the end of each period, but the event may occur at any time between the time points the measurements are taken. Therefore, time is recorded discretely, but the underlying process is continuous. To find the risk of event occurrence in each time interval, a continuous-time survival function is used and the generalized linear model is applied. It is observed that the combination of the number of accrual groups, the size of the accrual groups, and the duration of the trial that gives a sufficient power level depends on the shape of the continuous-time survival function, the proportion of subjects who have experienced the event after a fixed number of time periods, and the size of the treatment effect. The results of the study are only presented graphically, because there is no simple closed-form expression for finding the variance of the treatment effect. The authors provide MATLAB software to perform the power calculations. More subjects should be recruited in each accrual group or more accrual groups should be included if the effect size or the proportion of the subjects who have experienced the event after a fixed number of time periods decreases, or the probability of the event occurrence is concentrated toward the end of the study duration.