2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Imaging, Molecular Testing, Newborn Screening, Omics Technologies, Pathology, Serologic Tests, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Therapy Sequence, Biological Therapies, Antibody Therapy, Transplantation, Bispecific Antibody Therapy, (CAR)-T Cell Therapies, Monoclonal Antibody Therapy, Non-Biological Therapies, Chemotherapy, Surgical, Adverse Events, Artificial Intelligence (AI), Computer Vision, Deep Learning, Generative AI, Machine Learning, Natural Language Processing, Biological Processes, Pathogenesis, CHIP, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Clinical Guidelines, Diseases, Hemoglobinopathies, Sickle Cell Disease, Immunology, GVHD, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Plasma Cell Disorders, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, ALL, CLL, Non-Hodgkin Lymphoma, B Cell Lymphoma, Indolent Lymphoma, Aggressive Lymphoma, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, CMML, Clinical Research, Clinical Trials, Health Outcomes Research, Patient-Reported Outcomes, Real-World Evidence, Registries, Survivorship, Study Population, Human, Adult, Elderly, Pediatric
PAPER

6092 Artificial intelligence-based risk stratification for overall survival after allogeneic HCT in chronic myelomonocytic leukemia: A registry study on behalf of the chronic malignancies working party of the EBMT

Marie Robin, MD, Adrian Mosquera, MD2, Luuk Gras3, Francis Ayuk4, Thomas Schroeder5, Joe Tuffnell6, Linda Koster7, Raynier Devillier, PhD8, Ibrahim Yakoub-Agha9, Annoek Broers10, Peter Dreger, Prof. Dr. med.11, Francesca Kinsella, PhD12, Johannes Schetelig13, Cristina Castilla-Llorente, MD14, Georg-Nikolaus Franke Sr, MD15, Henrik Sengeloev16, Tobias Gedde-Dahl17, Matthias Stelljes18, Robert Zeiser19, Gesine Bug, MD20, Carlos Perez Miguez, Bioinformatics Software/Biomedical Data Engineer21, Ana Tarrío Pérez22, Franca Fagioli23, Nicola Polverelli, MD24, Giuliano Filippini Velázquez25, Kavita Raj26, Carmelo Gurnari, MD,PhD27, Francesco Onida28, Donal McLornan29
1 Assistance Publique - Hôpitaux de Paris, Paris, France
2 University Hospital of Santiago de Compostela, Santiago de Compostela, Spain
3 EBMT Leiden Study Unit, Leiden, Netherlands
4 University Hospital Eppendorf, Hamburg, Germany
5 University Hospital | Essen, Essen, Germany
6 EBMT Leiden Study Unit, Leiden, Netherlands
7 EBMT Leiden Study Unit, Leiden, Netherlands
8 Programme de Transplantation&Therapie Cellulaire, Marseille, France
9 CHU de Lille, Lille, France
10 Erasmus MC Cancer Institute, Rotterdam, Netherlands
11 University of Heidelberg, Heidelberg, Germany
12 Birmingham Centre for Cellular Therapy and Transplant (BCCTT), Birmingham, United Kingdom
13 Universitaetsklinikum Dresden, Dresden, Germany
14 Gustave Roussy Cancer Campus, Villejuif, France
15 Medical Clinic and Policinic 1, Leipzig, Germany
16 Rigshospitalet, Copenhagen, Denmark
17 Oslo University Hospital, Rikshospitalet, Oslo, Norway
18 University of Muenster, Münster, Germany
19 University of Freiburg, Freiburg, Germany
20 University Hospital Frankfurt - Goethe University, Frankfurt Main, Germany
21 Group of Computational Hematology & Genomics (GrHeCo-Xen), Health Research Institute of Santiago de Compostela (IDIS), Santiago de Compostela, Spain
22 Group of Computational Hematology & Genomics (GrHeCo-Xen), Health Research Institute of Santiago de Compostela (IDIS), Santiago de Compostela, Spain
23 Ospedale Infantile Regina Margherita, Torino, Italy
24 Unit of Bone Marrow Transplantation - Unit of Hematology, Fondazione IRCCS Policlinico San Matteo,, Pavia, Italy
25 Augsburg University Hospital and Medical Faculty, Augsburg, Germany
26 University College London Hospitals NHS Trust, London, United Kingdom
27 Department of Biomedicine and Prevention, University of Rome Tor Vergata, Rome, Italy, Translational Hematology and Oncology Research Department, Taussig Cancer Center, Cleveland Clinic, Cleveland, OH, USA, Cleveland, United States
28 Fondazione IRCCS - Ca’ Granda - Ospedale Maggiore Policlinico IRCCS, Milan, Italy
29 University College Hospital London, London, United Kingdom

Background: Allogeneic hematopoietic cell transplantation (allo-HCT) is the only potentially curative treatment for chronic myelomonocytic leukemia (CMML), but outcomes remain heterogeneous because of relapse and non-relapse mortality. Most established CMML prognostic scores were developed outside the transplant setting and have limited ability to predict post-transplant outcomes. We therefore used the EBMT registry to develop and geographically evaluate a model based on routinely available pre-transplant variables to identify patients at increased risk of death after allo-HCT.

 Methods: We studied 2,346 patients undergoing allo-HCT for CMML. Patients transplanted in Germany formed a prespecified external validation cohort (n = 595), while the remaining patients constituted the training cohort (n = 1,751). From 63 candidate routinely available pre-transplant variables, a missingness filter — whose threshold was tuned as a hyperparameter by nested cross-validation in the training cohort (retaining variables with ≤80% missingness) — kept 52, which were further reduced to a set of 9 (age, Karnofsky performance status, peripheral-blood blasts, hemoglobin, red-cell transfusion dependence, cytogenetic risk, IPSS category, HLA matching, and donor–recipient sex mismatch) using random survival forest screening and LASSO selection. Cox-based models, random survival forest, XGBoost, and a stacked meta-learner were developed to predict overall survival. All preprocessing, imputation, variable selection, and model optimization were restricted to the training cohort. A prespecified threshold defining the highest-risk 15% of the training distribution was fixed before evaluation in the external validation cohort.

 Results: Baseline characteristics were broadly comparable between the training and validation cohorts. In the German validation cohort, external discrimination for overall survival was modest and comparable across model specifications, whether using the reduced 9-variable set or the full 52-variable set (c-index: 0.58–0.61 across algorithms). Reduction to 9 variables did not degrade performance: the reduced stacked meta-learner gave the highest discrimination overall (c-index 0.603, vs 0.586 for the full-variable meta-learner). In the external validation cohort, the optimal cut-off identified 71 of 595 patients (11.9%) as high-risk. These patients were clinically distinct: they were older (mean 66.2 vs 59.5 years; p < 0.0001) and more frequently transplanted from HLA-mismatched donors (52.1% vs 11.6%; p < 0.0001), whereas time from diagnosis to HCT (13.4 vs 15.6 months; p = 0.31) and the proportion with HCT-CI ≥ 3 (25.4% vs 25.0%; p = 1.00) were similar. These findings suggest that the model captured prognostic information not fully accounted for by the HCT-CI. Notably, 11% of patients with low-risk IPSS in the training cohort were reclassified as high risk, highlighting the model’s complementary value beyond disease-based staging. This additional risk stratification was associated with marked differences in clinical outcomes across both cohorts. In the training cohort, 5-year overall survival was approximately 2% among high-risk patients compared with 45% among standard-risk patients (HR, 5.11; 95% CI, 4.37–5.97; p<0.0001). Significant differences were also observed in the external validation cohort, in which the corresponding 5-year overall survival rates were 19% and 47%, respectively (HR, 2.43; 95% CI, 1.77–3.34; p<0.0001). The same cutoff also identified increased non-relapse mortality in both cohorts (training: subdistribution HR, 3.81; 95% CI, 3.11–4.66; external validation: 2.74; 95% CI, 1.87–4.02; both p<0.0001). In contrast, relapse incidence was similar between groups, suggesting that the survival disadvantage in the high-risk group was driven primarily by non-relapse mortality.

Conclusions: In this large multinational EBMT registry study, a model based on routinely available pre-transplant variables identified a subgroup of patients with CMML undergoing allo-HCT who had substantially higher post-transplant mortality. This practical high-risk/standard-risk classification may support pre-transplant risk stratification and individualized patient counseling and help identify patients who warrant closer clinical assessment and peri-transplant surveillance.

Marie Robin, MD: Abbvie - Honoraria; Taiho - Honoraria; Asthex - Honoraria; Medac - Honoraria