Abstract:
|
The key to successful statistical analysis is a well-written protocol. In turn, the key to a well-written protocol are carefully crafted objectives. Good protocol objectives should directly reflect the scientific purpose of the protocol, which, as elucidated in the draft ICH-E9(R1) guidance, leads to the estimand of interest. Good objectives inform the protocol endpoints and their timing, the patient population, the primary and sensitivity analyses, the use of ancillary variables in the analysis, missing data strategies, and multiple comparison strategies. In this roundtable, we will review the critical link between the protocol objectives and the estimand, discuss how to write good objectives, and give examples of objectives with the aim of examining the analysis strategies to which they lead. We will also discuss how statisticians can best engage protocol writing teams to ensure that the main clinical question of interest is incorporated into protocol objectives that can be translated to effective statistical analyses.
|