Abstract
Recent clinical trials have underscored the value of repeated minimal residual disease (MRD) measurements as a highly sensitive method for detecting subclinical disease and enabling dynamic risk stratification in hematologic malignancies. Despite its clinical potential, the complex and heterogeneous nature of MRD kinetics presents significant challenges for interpreting and integrating it into routine clinical decision-making. In this study, we present a comprehensive, model-based workflow for the longitudinal analysis of MRD trajectories designed to improve relapse risk prediction. We applied this newly developed workflow to a cohort of patients with mantle cell lymphoma (MCL). MRD measurements were collected from both bone marrow (BM) and peripheral blood (PB) over time, stored in the Fondazione Italiana Linfomi MCL0208 clinical trial. Using our functional MRD workflow, we defined four MRD dynamics that collapsed into two clinically relevant groups: favorable (rapid, sustained negativization) and unfavorable (persistent or fluctuating MRD). Patients with unfavorable profiles showed significantly shorter time to progression (TTP), with hazard ratio (HR) = 4.18 (95% CI: 2.44-7.14) in BM and HR = 5.71 (95% CI: 2.86-11.42) in PB. External validation in the European MCL Network "Younger trial" confirmed the predictive power of this stratification, with Kaplan-Meier analyses demonstrating significant prognostic discrimination. The most informative temporal windows for patient clustering vary by tissue. Early-phase BM assessments offer greater discriminatory power, whereas late-phase assessments are most informative in PB. These findings indicate that longitudinal MRD assessment in PB represents a clinically actionable strategy that could reduce dependence on invasive BM procedures.