As part of a collaboration with the company SEMEIA, conducted through Lucas Chabeau's CIFRE PhD thesis, we proposed several methodological studies (Chabeau et al., Statistics in Medicine, 2025; Chabeau et al., Clinical Epidemiology, 2026). This biostatistical work led to the development of a Python package for estimating a joint model for multivariate longitudinal data in the presence of competing risks (https://github.com/Semeia-io/jmrespy).
Following the TRIPOD guidelines, we subsequently developed, and internally and externally validated, a dynamic prediction tool for kidney graft failure based on repeated measurements of three longitudinal markers (serum creatinine, proteinuria, and donor-specific anti-HLA class II antibodies) along with baseline variables available at the time of kidney transplantation [GitHub Lucas] [App SEMEIA]). This project builds upon previous work conducted as part of Marie-Cécile Fournier's PhD thesis (Fournier et al., Nephrology Dialysis Transplantation, 2019; Lenain et al., Transplantation, 2021; https://dantan.shinyapps.io/DynPG).