Conditional power calculations as an aid in the decision whether to continue a clinical trial

It is shown how stochastically curtailed testing can be included in the process of deciding whether to continue a clinical trial in which survival times or times to some other event in two treatment arms are compared. A model with constant intensities is used. One examines the power function of the test which would be computed if the trial were prolonged, conditional on the course of the trial up to the time of decision.