Power: Conditional, Unconditional, and Predictive

Having developed a unified statistical framework for the monitoring of trials involving different kinds of endpoints, we apply it to the calculation of different types of power. We focus primarily on trials not designed to allow early stopping for benefit. For example, in a short-term feeding trial in people with mild hypertension, we do not feel ethically compelled to stop early for benefit even if the intervention is superior to the control. On the other hand, if it becomes clear that the new diet is not worthwhile, we may not want to continue the expensive feeding. Thus, stopping for futility may still be important.