Extending landmarking to mixture cure models with time-varying covariates

Marta Cipriani, Sapienza University of Rome

Co-authors: Marco Alfò, Sapienza University of Rome; Mirko Signorelli, Leiden University

Abstract: Mixture cure models are used in survival analysis when a subset of individuals is considered cured and no longer at risk. Dynamic prediction models refine survival predictions using longitudinal and baseline data. For mixture cure models, a landmarking approach based on the last observation carried forward method has been proposed. However, this approach does not make efficient use of the available longitudinal data, because it discards all repeated measurements except the last one. To overcome this limitation, we propose a novel framework for dynamic prediction with mixture cure models to efficiently model longitudinal data.