Modeling and quantification of effect measures for factorial survival outcome - Part II

• Applicants: Professor Dr. Marc Ditzhaus (†); Professor Dr. Markus Pauly
• Subject Area: Epidemiology and Medical Biometry/Statistics
• Term: since 2017
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 352692197

The aim of this project is to continue the successful first project phase and to (further) develop biostatistical methods for effect quantification in complex designs with time-to-event endpoints. These are motivated by interdisciplinary collaborations of the PIs with medical colleagues from national university hospitals as well as emerging problems with existing methods from the literature. Thus, a special focus is on well-interpretable estimands such as the RMST (restricted mean survival time) as well as situations with non-proportional hazards and/or competing risks. Such situations, e.g., can occur in oncology (especially with novel immunotherapies) or in case of certain autoimmune diseases (such as multiple sclerosis) or chronic respiratory diseases (such as childhood asthma). To develop trustworthy biostatistical inference methods (point estimates, confidence intervals and regions, and tests) for such settings we apply and combine several methods: permutation- and bootstrap-based methods, modern techniques of nonparametric statistics, multiple testing, and machine learning, for example. In extensive simulation studies, clinical data analyses with medical cooperation partners and reconstructed data from published studies, we optimize the methodology with respect to practicability and efficiency. Subsequently, we will make all methods available in R-packages and user-friendly shiny apps for a flexible analysis of complex time-to-event data. Detailed guidelines as well as accessible journal articles allow easy and immediate access to the software.

• DFG Programme: Research Grants