Multiplicity and Type I error control. Laura L. Fernandes, DBV, OB, OTS, CDER, FDA

Clinical trials in oncology often have key primary endpoints like progression free survival (PFS), overall survival (OS) that aim to capture the efficacy of the study drug. These endpoints are usually tested using a hierarchical testing procedure so as to control the overall type I error. Given that OS is captured over a longer duration, studies are usually powered for both PFS and OS as primary endpoints or with OS as the first key endpoint to be tested after PFS. Since PFS is used as a surrogate endpoint for OS, the OS interim analyses (IA) are timed to coincide with the final PFS analyses while the final OS analysis is to be done when the required OS events are achieved. In recent years there has been an emphasis on including patient reported outcomes (PROs) as key endpoints as opposed to utilizing them for exploratory analyses. These endpoints could focus on a particular aspect of the instrument, for example time-to-deterioration of cough, and would have to be included in the hierarchy of the overall testing procedure. The timing for the analyses (IA versus Final) and the placement of these endpoints in the overall hierarchy is crucial so as to ensure success on maximum number of endpoints. This talk will focus on the challenges faced when reviewing such applications that have many key endpoints, using various graphical methods for the multiplicity and in addition having multiple analyses at different time points (IA1, IA2, Final analyses).