All Times EDT
In a clinical study to compare a new therapy with a control, the conventional practice is to define an endpoint of evaluating the treatment efficacy with a corresponding estimand. One then estimates the treatment effect for efficacy. For the harm assessment, we collect the safety data and summarize the difference between two arms. Treatment selection is then made using these two sets of summary measures. Unfortunately, this approach is not informative, that is, we do not know if the harm and benefit are associated with each other or they are unrelated at the individual patient level. Moreover, the adverse events and efficacy events may not be easily differentiated with each other. The occurrence of an adverse event would potentially “censor” the occurrence of the efficacy event (competing risk). This complexity makes the analysis rather difficult, if not impossible. An alternative is to consider composite endpoints which include the adverse and efficacy events at each patient level as a study endpoint. This approach reflects the clinical practice when treating patients in real world setting. The approach also allows us to identify a high subgroup of patients who would benefit from the new treatment, but with no or little harm. We will use a dataset from a recent study to illustrate this procedure.