Abstract
The α-spending function approach is a flexible interim analysis method for clinical trials, which can control the type I error probability without specifying the times and the number of analyses before the trial starts. However validity of the method can be lost when the times or the number of analyses depend on the results of the previous interim analysis because it may cause the inflation of type I error probability. Although data dependent interim analysis plans have not been recommended for this reason, it is possible to calculate the boundaries which can preserve the type I error probability for a certain simple data dependent plan.
In this article, we show the way to calculate the adjusted boundaries, and evaluate the power and expected stopping time of the trial for the adjusted and no adjusted boundaries. Then we will show that type I error adjustment for a simple data dependent analysis plan is admissible although the plan does not have significant advantages for the power and the expected stopping time of the trial.
