Sensitivity Analysis Vs Supportive Analysis Under Estimand Framework: a Case Study in Hematological Malignancies

• Activity Number: 496 - Estimand Framework and Its Impact on Drug Development in Oncology
• Type: Topic Contributed
• Date/Time: Wednesday, July 31, 2019 : 10:30 AM to 12:20 PM
• Sponsor: Lifetime Data Science Section
• Author(s): Steven Sun* and Hans-Jochen Weber and Marie-Laure Casadebaig and Emily Butler and Satrajit Roychoudhury and Kaspar Rufibach and Viktoriya Stalbovskaya
• Companies: J&J and Novartis and Celgene and GlaxoSmithKline and Pfizer Inc and F. Hoffmann-La Roche and Merus
• Keywords: Estimand; ICH E9; Sensitivity analysis; supplementary analysis ; hematological clinical trial

Abstract

The draft addendum to the ICH E9 guideline was released in 2017 to introduce the estimand framework. This framework targets to align the clinical study objectives with the design and statistical analyses by providing detailed definitions of the inferential quantity of interest, the estimand. One of the major impacts on statistical analysis is to ensure the estimate for the main estimator can be reliably interpreted based on the selected main analysis and sensitivity analyses. As every model has its assumptions, the guideline emphasizes importance of sensitivity analyses to justify the assumptions made. A unique estimand is determined by the choice of patient population, the definition of the primary endpoint, the strategies to handle intercurrent events, and the choice of population-level summary. Altering any of the attributes will lead to a different estimand and hence will target a different scientific question. These are supplementary analyses instead of sensitivity analyses. We present considerations on how the estimand framework can influence the analyses of clinical trials in hematology and how they should be interpreted.