AUTHORS
Brian Chernak, MD, Ali Kamali, PhD2, Xin Wang, MD, MSCE3, Kuo-Kai Chin, MD4, Alexander Chan, MD5, Wenbin Xiao, MD, PhD6, Greg Goldgof, MD7, Raajit K. Rampal, MD, PhD8
AFFILIATIONS
1 Memorial Sloan Kettering Cancer Center, New York, New York, United States
2 Memorial Sloan Kettering Cancer Center, New York, New York, United States, Memorial Sloan Kettering Cancer Center, New York, New York, United States
3 Memorial Sloan Kettering Cancer Center, New York, New York, United States
4 Memorial Sloan Kettering Cancer Center, New York, New York, United States
5 Memorial Sloan Kettering Cancer Center, New York, New York, United States
6 Memorial Sloan Kettering Cancer Center, New York, New York, United States
7 Memorial Sloan Kettering Cancer Center, New York, New York, United States, Memorial Sloan Kettering Cancer Center, New York, New York, United States
8 Memorial Sloan Kettering Cancer Center, New York, New York, United States
ABSTRACT
Chronic myelomonocytic leukemia (CMML) is a myeloid malignancy with both myeloproliferative and myelodysplastic features and associated with high rates of transformation to acute myeloid leukemia (AML). Due to its heterogeneity, features associated with disease transformation are not well characterized. This study aimed to characterize the clinical features and molecular landscape of disease transformation with paired patient samples.
Methods: Patients with AML transformation on or after January 1, 2020, were screened for documented antecedent CMML-1 (bone marrow blasts <10%; peripheral blood blasts <5%). Cases were identified using the Hematologic Multimodal Analytics Platform (HeMAP), a large language model-enabled system that reconstructs longitudinal AML disease trajectories from electronic health records. HeMAP was used as a candidate-screening tool to identify patients with evidence of prior CMML in clinicopathologic documentation. The platform’s architecture and validation metrics were presented at EHA 2026 and have been submitted separately to the ASH 2026 Annual Meeting. All candidate cases subsequently underwent manual chart review to confirm the diagnosis, pathological reports, and outcomes.
Results: HeMAP identified 60 potential cases of AML arising after CMML-1. Following manual review, 26 patients met study criteria and had bone marrow evaluations available at both CMML diagnosis and AML transformation. Most excluded cases lacked sufficient documentation to confirm antecedent CMML diagnosis. Median age at CMML diagnosis was 68 years (range: 51-84). 22 (84.6%) were male. Median time from diagnosis of CMML to AML was 11 months (range: 1-71). 53.9% of patients had myeloproliferative subtype-CMML (MP-CMML).
Of bone marrow biopsies at CMML-1 diagnosis, 25 had available cytogenetic and molecular data. CPSS-Mol Genetic risk and CPSS-Mol favored higher risk groups (Genetic: Intermediate-2/High-risk: n=15/26; CPSS-Mol: Intermediate-2/High-risk: n=17/26). Prior to AML transformation, interval treatments included hypomethylating agents (n=13), hydroxyurea (n=3), JAK inhibitor (n=3), allogenic stem cell transplant (HSCT) (n=2), or observation only (n=12).
The most common mutations at CMML-1 were TET2 (64%), SRSF2 (64%), ASXL1 (44%), and RUNX1 (44%). At AML transformation, the most common mutations were SRSF2 (68%), ASXL1 (68%), TET2 (64%), and RUNX1 (60%). Genes mutated at AML transformation with a frequency of at least 10% in this cohort were compared to de novo AML patients (The Cancer Genome Atlas, 2013); enrichment of SRSF2, ASXL1, TET2, RUNX1, CBL (all p<0.0001), KRAS (p=0.0001), EZH2 (p=0.0050), and NF1 mutations (p=0.0027) were noted.
The most common acquired mutations at transformation were ASXL1 (CMML: 44%; AML: 68%), RUNX1 (CMML: 44%; AML: 60%) and NPM1 (CMML: 4%; AML: 20%). When analyzing the high-risk mutations by CPSS-Mol (ASXL1, NRAS, RUNX1, SETBP1), ASXL1 was noted at increased frequency at transformation by Exact McNemar test (p = 0.031). Gain of a high-risk mutation at transformation occurred in 14 patients, but was not associated with inferior overall survival (OS) (p>0.05).
6 patients had paired mutant variant allelic frequencies (VAFs) available at the time of their earliest known biopsy and AML transformation. The mean VAF change was minimal amongst driver mutations of at least 2 events (ASXL1: n=4, mean VAF change: -3.4%; RUNX1: n=3, -7.4%; TET2: n=11, -5.8%; SRSF2: n=5, -3.9%; CBL, n=3, -15.9%). All six patients acquired new mutations at transformation with VAF of ≥20%: Pt-1: SETBP1 (VAF 28.4%), ETV6 (VAF: 26.0%), ROS1 (30.1%), FAT1 (31.1%), Pt-2: STAG2 (20.1%), Pt-3: EP300 (22.4%), Pt-4: ARID4A (26.6%), Pt-5: ASXL1 (23.5%), KRAS (24.4%), Pt-6: MED12 (39.8%).
Median OS was 0.97 years at AML transformation. Receipt of HSCT after AML was associated with improved OS by Cox proportional hazards: (Hazard Ratio (HR): 0.19; 95% confidence interval (CI): 0.05-0.73, p =0.016). Receipt of intensive chemotherapy (including Vyxeos) was not associated with improved OS (HR 1.50; 95% CI: 0.6-3.9, p=0.31).
Conclusion: Transformation from CMML-1 to AML is a heterogenous process with poor survival and clonal complexity, but acquisition of CMML-associated high-risk mutations may not impact outcomes. Serial molecular testing in an expanded cohort may aid in defining clonal transformation and the harms and benefits of HSCT, intensive chemotherapy, and alternative approaches.
DISCLOSURES
Brian Chernak, MD: Cogent Biosciences - Honoraria; Refined Oncology - Honoraria; Integritas Communications - Honoraria; Plexus Communications - Honoraria; Sobi - Consultancy (Includes expert testimony); Geron Corporation - Consultancy (Includes expert testimony); Blueprint Medicines - Consultancy (Includes expert testimony)