I will address three aspects of designing clinical trials: (1) Using historical controls in addition to or in lieu of patients assigned to control in a randomized trial. I compare hierarchical modeling with modifying likelihoods with subjective assessment of prior distributions. There are circumstances in which each of these three approaches is appropriate. (2) Decision making in which explicit consideration is given to the reasons for conducting a clinical trial and the question is whether to continue the trial in view of the accumulating information. (3) Adaptive designs in which accumulating data influence the treatments used on patients presenting during the trial. The appropriateness of such designs depend on whether the trial is phase I, II, or III, on the prevalence of the disease, on the availability of alternative treatments, and on the endpoint.

Back to Workshop Schedule