Benchmarking progression-free survival ratio as primary endpoint in precision oncology clinical trials

作者：Federico Nichetti, Jennifer Hüllein, Pauline du Rusquec, Erin Pleasance, Li Chen, Andreas Möck, Peter Horak, Paolo Ambrosini, Simone Rota, Monica Niger, Luca Agnelli, Gabriele Tinè, Dominic Edelmann, Christophe Le Tourneau, Janessa Laskin, Giancarlo Pruneri, Chi Wang, Daniel Hübschmann, Filippo de Braud, Stefan Fröhling, Luigi Mariani · 发表于：npj Precision Oncology · 年份：2025 · DOI：10.1038/s41698-025-01231-x · 被引用次数：2 · 研究领域：Statistical Methods in Clinical Trials、Advanced Causal Inference Techniques、Cancer Treatment and Pharmacology

Abstract Progression Free Survival Ratio (PFSratio), as defined as the ratio between PFS on investigational treatment (PFS2) and PFS on the last prior therapy (PFS1), is a popular endpoint in precision oncology (PO) studies. In this work, five methodologies for PFSratio-based trial analysis (count-based, Kaplan Meier, Kernel-based Kaplan Meier, parametric and midrank) and two for trial design (GBVE and Weibull) are benchmarked. The Kernel-based Kaplan Meier analysis is most recommended, as it handles informative censoring and does not require PFS1/PFS2 distribution assumptions. Sample size and power calculation methods perform best when applied to settings with expected high PFS1/PFS2 correlation and median ratio. Analysis of five clinical trials (MOSCATO 01, WINTHER, MASTER, SHIVA and POG570) from >800 patients revealed an overall weak PFS1/PFS2 correlation (Kendall’s τ range 0.17-0.35), and an asymptotically unbiased median S PFSratio ( δ =1.3) = 33% by means of the Kernel-based analysis, while other methods considerably deviated in studies with censoring rate>10%. This methodology is implemented in the PROPHETS R package and Shiny app.