| Clinical Trials are the gold standard for assessing the efficacy of new treatments in the medical sciences. Efficacy is assessed by the effect of treatment on a clinically relevant endpoint (e.g. death, time to cancer recurrence etc). However, often times the assessment of efficacy is best measured by the effect of treatment on multiple endpoints. For example, a drug may be assessed by it's effect on both a clinical endpoint and a intermediate endpoint that is a mechanistically related to the clinical endpoint. The current trend is to design studies to detect treatment effects on the clinical endpoint, and treat the intermediate endpoint as a secondary endpoint. However, pre-planned stopping criteria based on the primary clinical endpoint are often modified ad-hoc by considering the joint results on both endpoints. This can distort the operating characteristics of the trial and lead to erroneous conclusions. We propose methodology for specifying bivariate decision criteria at the design stage based on information from both endpoints. We incorporate the mechanistic relationship between the endpoints using concepts of statistical mediation. We illustrate the use of this method in both Fixed Sample and Group Sequential Designs using an example from a clinical trial in Pulmonary Arterial Hypertension. |
