Background: In sequential and adaptive trials, the delay that happens after the trial is stopped, by a predetermined stopping criterion, takes the name of overrunning. Overrunning consists of extra data, collected by investigators while awaiting results of the interim analysis (IA). The inclusion of such extra data in the analyses is scientifically appropriate and follows regulatory advice. Nevertheless, its effect from a broader perspective is unclear.
Methods: This article aims at clarifying the overall impact of including such overrunning data, providing first a revision, and then a comparison of the several approaches proposed in the literature for treating such data. A simulation study is performed based on two real-life examples.
Results: The paper shows that overrunning inclusion could seriously change the decision of an early conclusion of the study. It also shows that some of the methods proposed in the literature to include overrunning data are more conservative than others.
Conclusion: The choice of a more or a less conservative method could be considered more appropriate depending on the endpoint type or the design type.
Keywords: Combining p values; Deletion method; Overrunning; Repeated confidence interval.