Nomogram-based prediction models for clinical outcomes in pediatric RUNX1::RUNX1T1-positive acute myeloid leukemia: a retrospective analysis from AML-CAMS serial trials

Objective To identify prognostic factors and develop nomograms predicting short-term mortality and relapse in pediatric RUNX1::RUNX1T1 -positive AML, thereby enabling individualized risk assessment and optimizing clinical management. Methods We retrospectively analyzed 136 pediatric patients with RUNX1::RUNX1T1 -positive AML who achieved morphologic complete remission (CR) after one induction course under AML-CAMS-2009 or AML-CAMS-2016 regimen. Least absolute shrinkage and selection operator (LASSO) and multivariable Cox regression identified independent predictors of 3-year overall survival (OS) and relapse-free survival (RFS). Nomograms were built from these predictors. Model performance was assessed by time-dependent receiver operating characteristic (ROC) curves, calibration plots, decision curve analysis (DCA), and concordance index (C-index), with internal validation performed by bootstrap resampling. Results High-Risk measurable residual disease (MRD), treatment regimen, and diagnostic white blood cell (WBC) group (≥20×10⁹/L vs. <20×10⁹/L) independently predicted OS. For RFS, independent predictors were percentage of bone marrow blasts by flow cytometry (BM blasts [FCM]), extramedullary infiltration (EMI), High-Risk MRD, treatment regimen, and WBC group. Nomograms demonstrated strong discrimination and calibration with superior clinical net benefit versus any single predictor. Nomogram-derived scores stratified patients into prognostically distinct subgroups with significant differences in OS and RFS. Conclusions This study established internally validated 3-year OS and RFS nomograms for pediatric RUNX1::RUNX1T1 -positive AML with excellent discrimination and clinical utility. Prospective multicenter validation is warranted to confirm the robustness and facilitate clinical adoption.