Critical reflections on clinical trials

There is often concern about the large number of patients needed for a particular clinical trial. However, one could be equally concerned that a trial is designed with a number of patients that is too small to obtain a precise enough estimate of the difference between the success rates of the treatments. The design of the trial should ensure that statistical significance is only reached if the treatment difference exceeds a specified value of clinical importance. Designs with this property generally require more than the usual number of patients. It is hoped that discussion on the ethical acceptability of this consequence will be stimulated by this paper.