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, Omics Technologies, Bioinformatics, Computational Biology, Profiling, Treatment Considerations, Combination Therapy, Biological Therapies, Transplantation, Non-Biological Therapies, Chemotherapy, Artificial Intelligence (AI), Machine Learning, Biological Processes, Apoptosis, Pathogenesis, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Microbiome, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Clinical Guidelines, Diseases, Bone Marrow Failure Syndromes, Acquired Marrow Failure Syndromes , Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Fundamental Science, AML, APL, Translational Research, MDS, Clinical Research, Real-World Evidence, Registries, Study Population, Human, Adult, Elderly, Pediatric, Young Adult, Animal Model
PAPER

5238 Pretreatment DNA methylation profiles identify a decitabine-specific response signature: Results of the EORTC phase III intergroup trial AML21

Geoffroy Andrieux, PhD, Michal Kicinski2, Joop H. Jansen, Ph.D.3, Tobias Ma4, Aniek O De Graaf, Ph.D.5, Sylvain Chantepie, MD6, Walter J.F.M. Van der Velden, MD7, Laimonas Griškevicius, MD8, Emanuele Ammatuna, MD, Ph.D9, Adriano Venditti, MD10, Pierre W. Wijermans, MD, PhD11, Melanie Boerries12, Gerwin A. Huls, MD, PhD13, Michael Lübbert, MD14
1 Institute of Medical Bioinformatics and Systems Medicine, Freiburg, Germany
2 Eortc, Brussel, Belgium
3 Radboudumc, Nijmegen, Netherlands
4 Uniklinik Freiburg, Freiburg im Breisgau, Germany
5 Radboudumc, Nijmegen, Netherlands
6 CHU Caen, Caen, France
7 Radboudumc, Nijmegen, Netherlands
8 Santaros Klinikos, Vilnius, Lithuania
9 UMC Groningen, Groningen, Netherlands
10 Università Tor Vergata, Roma, Italy
11 Haga Ziekenhuis, The Hague, Netherlands
12 Institute of Medical Bioinformatics and Systems Medicine, Freiburg, Germany
13 UMC Groningen, Groningen, Netherlands
14 Uniklinik Freiburg, Freiburg im Breisgau, Germany

DNA-hypomethylating agents (HMAs) are an established backbone for older or unfit patients with acute myeloid leukemia (AML), yet pretreatment biomarkers predicting benefit from single-agent HMA therapy remain poorly defined. Because HMAs act through mechanisms distinct from cytarabine-based intensive chemotherapy (IC), we hypothesized that baseline DNA methylation profiles may identify treatment-specific response signatures. We investigated this question in the randomized phase III AML21 trial of the EORTC Leukemia Group, GIMEMA, and German MDS Study Group, which compared standard IC (“7+3”) with single-agent decitabine (DAC) in patients aged ≥60 years with newly diagnosed AML who were considered fit for IC.

Baseline DNA methylation profiling performed on AML blasts using Illumina EPIC arrays was performed on 405 pretreatment bone marrow samples from both treatment arms. Clinical and molecular characteristics were well-balanced between arms, including age, sex, ELN risk category, mutation profile, marrow blast percentage, and white blood cell count. CR/CRi rates were higher after IC than DAC, whereas overall survival and allogeneic hematopoietic cell transplantation rates and outcome were similar (Lübbert et al., Lancet Haemat. 2023). CpGs were aggregated into correlated genomic regions to reduce redundancy and capture coordinated methylation programs. Supervised machine-learning models were developed to predict CR/CRi separately within treatment arms. Models were trained using feature selection, dimensionality reduction, and repeated cross-validation with gradient-boosted decision trees.

Unsupervised regional methylation profiling identified six robust epigenetic AML subgroups. These clusters recapitulated major biologic and clinical disease features: one cluster was enriched for IDH2-mutated AML, a second for DNMT3A-mutated/IDH2-wild-type cases, a third for TP53-mutated AML, and another for NPM1-mutated and ELN favorable-risk disease. Two additional clusters were associated primarily with clinical characteristics, including lower marrow blast percentage and lower white blood cell count, and were enriched for patients treated in the IC arm. These findings confirm that pretreatment methylation landscapes capture clinically meaningful AML heterogeneity beyond individual mutations.

A methylation-based model trained to predict response to DAC achieved a ROC AUC of 0.70 in held-out DAC-treated samples. In contrast, the same DAC-derived signature showed reduced performance in the IC arm (ROC AUC of approximately 0.60), supporting a treatment-specific methylation predictor rather than a general marker of treatment sensitivity. Regions selected by the DAC model showed more pronounced methylation differences between responders and non-responders in the DAC arm than in the IC arm. Differential methylation analysis confirmed that DAC signature regions were preferentially regulated in DAC-treated responders versus non-responders, whereas background regions showed substantial overlap between arms. Conversely, a model trained in the IC arm performed best within IC-treated patients (ROC AUC 0.71) and showed attenuated performance when applied to DAC-treated patients. A third model trained on both treatment arms performed comparably within its training context but relied on a largely distinct and only partially overlapping set of genomic regions. Comparison of model-selected regions demonstrated that most predictive regions were treatment arm-specific, with limited overlap between DAC-derived, IC-derived, and combined-arm signatures.

These data demonstrate that pretreatment DNA methylation profiles can predict response to single-agent decitabine in older, fit patients with newly diagnosed AML. Importantly, the predictive methylation architecture differs between DAC monotherapy and cytarabine-based IC, indicating that HMA response is not captured by a generic response signature but reflects treatment-specific epigenetic biology. This study provides a clinically relevant framework for methylation-guided selection of AML therapy and supports further validation of treatment arm-specific methylation biomarkers to identify patients most likely to benefit from HMA-based approaches.

Michael Lübbert, MD: No relevant conflicts of interest to declare.