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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Sunday, December 13, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Molecular Testing, Omics Technologies, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Drug Development, Therapy Sequence, Biological Therapies, Antibody Therapy, Transplantation, Checkpoint Inhibitor, Immunotherapy, Non-Biological Therapies, Chemotherapy, Adverse Events, Artificial Intelligence (AI), Deep Learning, Machine Learning, Biological Processes, Apoptosis, Pathogenesis, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Practice Models, Value, Diseases, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, CMML, Clinical Research, Epidemiology, Health Outcomes Research, Health Disparities Research, Real-World Evidence, Survivorship, Study Population, Human, Adult
PAPER

3453 An individualized continuous prognostic model for NPM1-mutated acute myeloid leukemia (AML) that incorporates genomic and clinical variables

Paloma Resende, MD, Andre Meireles, MD2, Tarcila Santos Datoguia, MD3, Amanda Ennes Inacio Dias dos Santos, MD4, Jade Zezzi Martins do Nascimento, MD5, Carolina Perrone Marques Ferreira, MD6, Kira Bucci, MD7, Juliana Dall Agnol Rocha, MD8, André Dias Américo, MD9, Luiz Ferreira Cortes10, Alexandre Eiji Eiji Kayano, MD11, Luana Nobrega, MD12, Frederico Honorato Honorato Jr, MD13, Claudia Bley, MD14, Elvira Velloso Sr, Dr15, Andreza Feitoza Ribeiro, MD16, Paulo Vidal Campregher, MD, PhD17, Nelson Hamerschlak Sr, MD PhD18, Fabio Pires S Santos, MD, PhD19
1 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
2 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
3 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
4 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
5 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
6 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
7 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
8 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
9 BP - A Beneficência Portuguesa de São Paulo, Sao Paulo, Brazil
10 Einstein Hospital Israelita, Sao Paulo, Brazil, Genesis Genomics, Sao Paulo, Brazil
11 Einstein Hospital Israelita, Sao Paulo, Brazil, Hospital das Clínicas da FMUSP, Sao Paulo, Brazil
12 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
13 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
14 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
15 Einstein Hospital Israelita, Sao Paulo, Brazil, Hospital das Clínicas da FMUSP, Sao Paulo, Brazil
16 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
17 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil, Genesis Genomics, Sao Paulo, Brazil
18 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil
19 Einstein Hospital Israelita, São Paulo, SP, Brazil, Sao Paulo, Brazil

Background: NPM1-mutated AML is a prognostically heterogeneous disease. Current stratification systems, such as the European LeukemiaNet, place substantial weight on the presence of co-occurring FLT3-ITD. More recent frameworks, such as the system developed by the HARMONY consortium, assign patients to one of three discrete genetic categories. Although easily implemented in clinical practice, these categorical systems do not capture continuous clinical effects, gene–gene interactions, or the impact of less common co-mutations. We hypothesized that an individualized, continuous prognostic model including both clinical variables and gene–gene interactions would predict overall survival (OS) more accurately than categorical classification systems. 

Methods: We analyzed intensively treated adults with NPM1-mutated AML from four cohorts with publicly available sequencing data: the UK-NCRI and AMLSG cohorts reported by Tazi et al. (Nat Commun, 2022), Beat AML (n=126), and TCGA (n=45). OS was measured from diagnosis and was not censored at transplant. We pre-specified a Cox model with restricted cubic splines for age and log2-transformed white blood cell count (WBC), AML type (de novo vs secondary), the main effects of genes commonly mutated in NPM1-mutated AML, including FLT3-ITD, DNMT3A, IDH1/2, TET2, NRAS, KRAS, PTPN11, and WT1, and three pairwise gene interactions: FLT3-ITD × DNMT3A, FLT3-ITD × IDH, and DNMT3A × IDH. Prediction used a ridge-penalized Cox model; adjusted genotype contrasts and confidence intervals (CIs) were estimated using the corresponding cohort-stratified unpenalized model. We measured discrimination using Harrell’s C-index, corrected optimism by bootstrap, tested transportability using leave-one-cohort-out (LOCO) cross-validation and benchmarked the model against ELN 2022 and HARMONY.

Results: Among 1,251 patients with complete covariates, 688 died. The individualized clinical-genomic model showed higher discrimination than the HARMONY genetic classification (C-index 0.660 vs 0.583; ΔC +0.077, 95% CI +0.054 to +0.101) and ELN 2022 (0.660 vs 0.558; ΔC +0.102, 95% CI +0.078 to +0.126). The bootstrap optimism-corrected C-index was 0.680. Age was the strongest clinical predictor (HR 2.82 for 70 vs 50 years, 95% CI 2.31–3.44), while a WBC of 100 vs 10 × 10⁹/L was associated with an HR of 1.28 (95% CI 1.05–1.55). Among the genomic terms, FLT3-ITD and DNMT3A showed a positive interaction (ratio of HRs 2.09, 95% CI 1.52–2.88). Patients carrying both mutations had twice the hazard of patients carrying neither mutation (HR 2.00, 95% CI 1.55–2.58). IDH mutations also interacted with DNMT3A. They were adverse in the presence of a DNMT3A mutation (HR 1.57, 95% CI 1.18–2.09) but had no detectable effect in DNMT3A-wild-type patients (HR 0.99, 95% CI 0.74–1.31). Among the main effects of individual genes, WT1 had the strongest association with worse outcomes (HR 1.75, 95% CI 1.30–2.36). Pre-specified analyses of TET2 mutational burden, RAS-pathway mutations, and DNMT3A mutation type showed no independent prognostic association. An online OS calculator is available at https://fabiopss23.github.io/npm1_model/.

Conclusions: Our results suggest that prognosis in NPM1-mutated AML is better captured as a continuous function of clinical variables, including age and WBC, and genomic architecture than by discrete categories. The pre-specified individualized model modestly improved discrimination over the HARMONY classifier and showed consistent performance across cohorts. WT1 was associated with adverse outcomes. Given the modest gain, the use of OS without censoring at transplant, and the lack of true external validation, prospective validation is warranted.

Paloma Resende, MD: BeOne - Honoraria; CSL Behring - Honoraria