A0859
Title: Effect measures for comparing consecutive survival times
Authors: Dennis Dobler - TU Dortmund University (Germany) [presenting]
Marc Ditzhaus - Otto-von-Guericke University Magdeburg (Germany)
Merle Munko - Otto-von-Guericke University Magdeburg (Germany)
Dominic Edelmann - DKFZ Heidelberg (Germany)
Simon Mack - TU Dortmund University (Germany)
Abstract: The progression-free-survival ratio (PFSr) is a popular endpoint in oncology trials, where patients serve as their own control. Various inference methods have been proposed for this endpoint. However, many of them rely on unrealistic assumptions or fail to incorporate censoring correctly, leading to biased results. The close connection between the PFSr and the relative treatment effect is pointed out. Recent results about the latter are applied to multivariate survival data based on techniques from competing risk analysis. In this way, a mathematically correct inference strategy is developed for the PFSr under random right-censoring. To further consider the rising importance of interpretable estimands in medical research, effect measures are proposed based on the restricted mean survival times as an alternative in this context. For both endpoints, valid resampling procedures are presented to improve the performance for small sample sizes. The novel methods are exemplified by an extensive simulation study and a real data analysis.
