Data monitoring in clinical trials: the case for stochastic curtailment

Interim analyses have become an essential part of the monitoring process of clinical trials. Stochastic curtailment has been used in such analyses. This procedure allows for calculation of the probability of rejecting the null hypothesis at the end of a trial given the current data and assuming the null or an alternative hypothesis for the remainder of the trial. Such information can be used to decide whether a trial should continue or be stopped early due to either treatment benefit or harm or because of lack of power to show an effect. Using stochastic curtailment, stopping rules for one- or two-sided test trials can be easily visualized by constructing boundaries based on the null and alternative hypotheses. Interim Z test statistics falling above or below these boundaries can aid in interim monitoring decisions. Methods for constructing boundaries, expected trial times and examples of clinical trials in cardiovascular and vision research where stochastic curtailment was used are presented.