DFG-Project (DI 2906/1-2)

Modelling and Quantifying effect sizes for survival data in factorial designs - Part II 

German Research Foundation (DFG), Project number: DI 2906/1-2, duration: 11/2022 - 10/2025.

Joint project with Prof. Dr. Markus Pauly (TU Dortmund University).

Project description

Due to missing alternatives for general factorial survival designs, classical regression methods (Cox- and Aalen-regression) are mainly used. In particular for crossing survival curves or crossing hazard rates, which are difficult to identify with the naked eye on the basis of estimated survival curves, the expressiveness of the (omnipresent) Cox-model is not guaranteed. There are already first positive developments of alternative analysis methods, which, however, focus mainly on the two and k-sample case. Extensions to more general factorial designs were generated in the first phase of the project. The main results include a flexible combination approach of weighted Logrank-tests for the two sample case and for more general factorial designs as well as methods based on alternative estimands (medians, probabilistic index and the RMST). Following on this successful part of the project, the aim is the extension of the current tools (and, in particular, of the R package GFDsurv and mdir.logrank) by (1) considering the RMST, which becomes more and more relevant, for factorial designs, (2) developing corresponding multiple contrast tests and simultaneous confidence intervals for contrasts of the different estimands, (3) investigating the application of the combination approach in the field of machine learning, (4) extending the application of all proceedings for competing risks.

Software

  • GFDsurv: avaibable on CRAN and as a SHINY-App. This packages contains three different tests for survival data in general factorial designs (Ditzhaus, Dobler and Pauly, 2021; Ditzhaus, Janssen and Pauly, 2020; Dobler and Pauly, SMMR, 2019) from the first project phase. The package will be updated stepwisely by the new methods devoloped during the current second project phase.
  • GFDrmst: available on CRAN and as a SHINY-App. This package consists multiple tests based on the restricted mean survival time (RMST) for general factorial designs as described in Munko et al. (2024, Stat. Med.) 

Publications

  • Munko, Ditzhaus, Dobler, Genuneit (2024). RMST-based multiple contrast tests in general factorial designs. Statistics in Medicine;1-18. doi: 10.1002/sim.10017 (Open Access).
    • Short Abstract:  The restricted mean survival time (RMST) is a popular example for an effect estimands that do not rely on the proportional hazards assumption. The first goal of the paper is to further extend the RMST-based two-sample permutation test for general factorial designs and general contrast hypotheses. Additionally, a groupwise bootstrap approach is considered. Moreover, multiple tests for the RMST are developed in a second step to infer several null hypotheses simultaneously.
  • Emura, Ditzhaus, Dobler, Murotani (2023). Factorial survival analysis for treatment effects under dependent censoring. Statistical Methods in  Medical Research 33(1); 61-79.
    • Short Abstract: All existing factorial analyses for survival data were developed under the independent censoring assumption, which is too strong for many applications. As a solution, the central aim of this article is to develop new methods in factorial survival analyses under quite general dependent censoring regimes. This will be accomplished by combining existing results for factorial survival analyses with techniques developed for survival copula models.
  • Dormuth, Liu, Xu, Pauly and Ditzhaus (2023). A comparative study to alternatives to the log-rank test. Contemporary Clinical Trials 128;107165.
    • Short Abstract: In this paper, we give updated recommendations on recent developments for two-sample alternatives to the (omnipresent) log-rank-test. Therefore, we perform a vast simulation study to compare tests that showed high power in previous studies with these more recent approaches. We thereby analyze various simulation settings with varying survival and censoring distributions, unequal censoring between groups, small sample sizes and unbalanced group sizes. 
  • Ditzhaus, Yu and Xu (2023+). Studentized Permutation Method for Comparing Restricted Mean Survival Times with Small Sample from Randomized Trials.Statistics in Medicine 42(13);  2226-2240. (Open Access)
    • Short Abstract: As pointed out by Horiguchi and Uno (2020), methods for the restricted mean survival time (RMST) based on asymptotic theory suffer from inflated type-I error under small sample sizes. Their permutation strategy leads to more convincing results in simulations. But it relies on exchangeable data and cannot be inverted to obtain valid confidence intervals. In this paper, we address these limitations by proposing a studentized permutation test and the corresponding permutation-based confidence intervals.

Preprints

  • Thurow, Dormuth, Sauer, Ditzhaus and Pauly (2023+). How to Simulate Realistic Survival Data? A Simulation Study to Compare Realistic Simulation Models. (arXiv:2308.07842)
    • Short Abstract: In most medical applications and especially for clinical trials in oncology, there is a lack of adequate benchmark data sets for method comparison. In this paper, we use reconstructed benchmark data sets as a basis for simulating survival data, which has the following advantages: the actual properties are known and more realistic data can be simulated. We investigate simulation models based upon kernel density estimation, fitted distributions, case resampling and conditional bootstrapping. In order to make recommendations on which models are best suited for a specific survival setting, we conducted a comparative simulation study. We focus on providing realistic simulation models for two-armed phase III lung cancer studies. To this end we reconstructed benchmark data sets from recent studies. We used the runtime and different accuracy measures (effect sizes and p-values) as criteria for comparison.
  • Ditzhaus, Fernandez and Rivera (2023+). A Multiple kernel testing procedure for non-proportional hazards in factorial designs. ArXiv-Preprint.  
    • Short Abstract: In this paper we propose a Multiple kernel testing procedure to infer survival data when several factors (e.g. different treatment groups, gender, medical history) and their interaction are of interest simultaneously. Our method is able to deal with complex data and can be seen as an alternative to the omnipresent Cox model when assumptions such as proportionality cannot be justified. 

Talks

  • How to correctly infer the progression-free-survival ratio under right-censoring, and an RMST-based alternative. Survival Analysis for Junior Researchers conference 2023 in Ulm (September 2023)
  • Multiple Contrast Tests for the RMST in General Factorial Designs. Survival Analysis for Junior Researchers conference 2023 in Ulm (September 2023)
  • Surviving the multiple testing problem: RMST-based tests in general factorial designs. CMStatistics 2023 in Berlin (December 2023)
  • Surviving the multiple testing problem: RMST-based tests in general factorial designs. Biometric Colloquium 2024 in Lübeck (February 2024)

Organized Sessions

  • "Beyond proportional hazards and standard survival",  CMStatistics 2022 in London (jointly organized with Prof. Dennis Dobler from Amsterdam)
  • Survival session at CMStatistics 2023 in Berlin  (jointly organized with Prof. Dennis Dobler from Amsterdam)

 Berlin1     Berlin2

Last Modification: 04.04.2024 - Contact Person: