Prediction of overall survival in acute myeloid leukemia using supervised machine learning
17 affiliations
- 1Department of Clinical Medicine, Center for Clinical Data Science, Aalborg University and Aalborg University Hospital, Aalborg, Denmark
- 2Department of Molecular Medicine, Aarhus University Hospital, Aarhus, Denmark
- 3Department of Clinical Medicine, Aarhus University, Aarhus, Denmark
- 4Department of Pathology, Aalborg University Hospital, Aalborg, Denmark
- 5Department of Hematology, Odense University Hospital, Odense, Denmark
- 6Department of Clinical Research, University of Southern Denmark, Odense, Denmark
- 7Department of Pathology, Odense University Hospital, Odense, Denmark
- 8Department of Pathology, Copenhagen University Hospital, Herlev, Denmark
- 9Department of Hematology, Zealand University Hospital, Roskilde, Denmark
- 10Department of Hematology, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark
- 11Biotech Research and Innovation Center (BRIC), University of Copenhagen, Copenhagen, Denmark
- 12Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
- 13Department of Clinical Genetics, Rigshospitalet, Copenhagen University Hospital, Copenhagen, Denmark
- 14Department of Hematology, Clinical Cancer Research Center, Aalborg University Hospital, Aalborg, Denmark
- 15Department of Clinical Medicine, Aalborg University, Aalborg, Denmark
- 16Department of Hematology, Aarhus University Hospital, Aarhus, Denmark
- 17Department of Computer Science, Aalborg University, Aalborg, Denmark
Abstract
Background: Risk stratification in acute myeloid leukemia (AML) is traditionally based on cytogenetic and genetic features alone and does not incorporate patient-related clinical variables. This study evaluated whether machine learning (ML) models integrating clinical, cytogenetic, and genetic features improve prediction of 1-year and overall survival (OS) compared with guideline-based risk stratification.
Patients and methods: We conducted a study of 785 contemporary real-world patients with AML. Patients were stratified by treatment intensity (intensive and non-intensive). Several ML models for binary 1-year survival and time-to-event OS were trained using three feature sets: clinical alone; clinical plus cytogenetics; and combined clinical, cytogenetic, and genetic. Performance was evaluated using area under the receiver operating characteristic curve (ROC-AUC) and C-index and compared with current models by European LeukemiaNet (ELN) 2022 and 2024.
Results: The binary classification performance was similar across treatment groups with ROC-AUC between 0.5 and 0.7. Predictive performance for OS varied by treatment group, where ML models for non-intensively treated patients achieved better performance (C-index >0.75), consistently outperforming ELN2024 (C-index ∼0.57). In intensively treated patients, performance was lower and did not outperform ELN2022 although best results were obtained using combined clinical and cytogenetic features. Adding genetic features did not consistently improve model performance across settings. Models using the full feature set indicate that clinical variables have the greatest impact.
Conclusions: ML-based models yielded better or equal prognostic accuracy compared with ELN 2022 and 2024, highlighting the value of integrating clinical features. Limited performance in intensively treated patients suggests a need for dynamic predictors, such as treatment response measures, to enhance risk stratification further.
Related
- HepaKineticNet: Deep Learning of CT-Derived Intratumoral Heterogeneity for Overall Survival Prediction in BCLC Stage B-C Hepatocellular Carcinoma Treated with Locoregional Therapy Plus First-Line TKI and Anti-PD-1 Therapy.
- A transcriptional signal associated with mesenchymal stromal cells predicts overall survival in B cell acute lymphoblastic leukemia
- Deep learning prediction of overall survival in glioblastoma patients using pre-chemoradiotherapy MRI and clinical data
- Epigenetic age acceleration enhances prediction of overall survival in early-stage non-small cell lung cancer
- Development and temporal validation of nomograms and machine learning models for predicting overall and cancer-specific survival in primary distal cholangiocarcinoma: a SEER-based study
BibTeX
@article{brondum2026prediction,
title = {{Prediction of overall survival in acute myeloid leukemia using supervised machine learning}},
author = {R.F. Brøndum and M.S. Andersen and M. Knudsen and L. Grubach and C. Marcher and B. Preiss and M.L. Bibi and E. Høgdall and T.S. Poulsen and V. Skov and A.D. Ørskov and K. Grønbæk and J.W. Hansen and C. Schöllkopf and J.B. Cowland and M. Klarskov and M.T. Severinsen and C. Vejgaard and O.H. Larsen and S. Vang and M. Bøgsted and J.M. Rodriguez and A.S. Roug and D.T. Kristensen},
journal = {ESMO Real World Data and Digital Oncology},
year = {2026},
month = oct,
doi = {10.1016/j.esmorw.2026.100806},
url = {https://doi.org/10.1016/j.esmorw.2026.100806},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - JOUR
TI - Prediction of overall survival in acute myeloid leukemia using supervised machine learning
AU - R.F. Brøndum
AU - M.S. Andersen
AU - M. Knudsen
AU - L. Grubach
AU - C. Marcher
AU - B. Preiss
AU - M.L. Bibi
AU - E. Høgdall
AU - T.S. Poulsen
AU - V. Skov
AU - A.D. Ørskov
AU - K. Grønbæk
AU - J.W. Hansen
AU - C. Schöllkopf
AU - J.B. Cowland
AU - M. Klarskov
AU - M.T. Severinsen
AU - C. Vejgaard
AU - O.H. Larsen
AU - S. Vang
AU - M. Bøgsted
AU - J.M. Rodriguez
AU - A.S. Roug
AU - D.T. Kristensen
PY - 2026
DA - 2026/10/28
JO - ESMO Real World Data and Digital Oncology
DO - 10.1016/j.esmorw.2026.100806
UR - https://doi.org/10.1016/j.esmorw.2026.100806
AB - Background: Risk stratification in acute myeloid leukemia (AML) is traditionally based on cytogenetic and genetic features alone and does not incorporate patient-related clinical variables. This study evaluated whether machine learning (ML) models integrating clinical, cytogenetic, and genetic features improve prediction of 1-year and overall survival (OS) compared with guideline-based risk stratification. Patients and methods: We conducted a study of 785 contemporary real-world patients with AML. Patients were stratified by treatment intensity (intensive and non-intensive). Several ML models for binary 1-year survival and time-to-event OS were trained using three feature sets: clinical alone; clinical plus cytogenetics; and combined clinical, cytogenetic, and genetic. Performance was evaluated using area under the receiver operating characteristic curve (ROC-AUC) and C-index and compared with current models by European LeukemiaNet (ELN) 2022 and 2024. Results: The binary classification performance was similar across treatment groups with ROC-AUC between 0.5 and 0.7. Predictive performance for OS varied by treatment group, where ML models for non-intensively treated patients achieved better performance (C-index >0.75), consistently outperforming ELN2024 (C-index ∼0.57). In intensively treated patients, performance was lower and did not outperform ELN2022 although best results were obtained using combined clinical and cytogenetic features. Adding genetic features did not consistently improve model performance across settings. Models using the full feature set indicate that clinical variables have the greatest impact. Conclusions: ML-based models yielded better or equal prognostic accuracy compared with ELN 2022 and 2024, highlighting the value of integrating clinical features. Limited performance in intensively treated patients suggests a need for dynamic predictors, such as treatment response measures, to enhance risk stratification further.
SN - 2949-8201
ER -