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
Access, Representation, Population Health, Technology and Procedures, Molecular Testing, Clinical Procedures, Treatment Considerations, Drug Development, Therapy Sequence, Biological Therapies, Antibody Therapy, Transfusion, Transplantation, Non-Biological Therapies, Chemotherapy, Devices, Pharmacology, Radiation Therapy, Adverse Events, Artificial Intelligence (AI), Deep Learning, Biological Processes, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Clinical Practice (Health Services and Quality), Clinical Guidelines, Diseases, Genetic Disorders, Hemoglobinopathies, Sickle Cell Disease, Sickle Cell Trait, Immunology, GVHD, Infectious Diseases, Viral, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Metabolic Disorders, Measurable Residual Disease, ALL, Non-Hodgkin Lymphoma, B Cell Lymphoma, Research, AML, Translational Research, MDS, CMML, Clinical Research, Clinical Trials, Epidemiology, Patient-Reported Outcomes, Real-World Evidence, Registries, Survivorship, Study Population, Human, Adult, Elderly, Young Adult
PAPER

4196 Identification of post-transplantation acquired cardiac dysfunction using artificial intelligence-enabled analysis of electrocardiogram waveforms

Husayn Mohamed Ladha, Yiyu Xie, MD2, Miguel-Angel Perales, MD3, Sergio Giralt, MD4, Ann Jakubowski, PhD, MD5, Esperanza Papadopoulos, MD6, Roni Tamari, MD7, Craig Sauter, MD8, Doris M Ponce, MD9, Moneeza Walji, MD10, Jeffery J. Auletta, MD11, Steven Michael Devine, MD12, Jennifer Liu13, Anthony Yu, MD14, Ioanna Kosmidou, MD15, Brian C Shaffer, MD, MS16
1 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States
2 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States
3 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
4 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
5 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
6 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
7 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
8 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
9 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
10 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
11 CIBMTR® (Center for International Blood and Marrow Transplant Research), NMDP, Minneapolis, Minnesota, United States
12 CIBMTR® (Center for International Blood and Marrow Transplant Research), NMDP, Minneapolis, Minnesota, United States
13 Cardiology Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
14 Cardiology Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
15 Cardiology Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States
16 Adult Bone Marrow Transplant Service, Memorial Sloan Kettering Cancer Center, New York, New York, United States, Department of Medicine, Weill Cornell Medical College, New York, New York, United States

Background: Post-transplant cyclophosphamide (PTCy) is an effective means to prevent graft-versus-host disease (GVHD) after allogeneic hematopoietic cell transplantation (allo-HCT). While Cy is a known cardiotoxin, whether PTCy increases the risk of post-transplant heart failure due to reduced left ventricular ejection fraction (rEF) remains unclear. To address this problem, we developed an artificial intelligence model to detect a signature of rEF using digitized 12-lead ECG waveforms (AI-ECG-rEF). We then used this model to estimate rEF among HCT recipients using PTCy versus other GVHD prophylaxis regimens and assessed the relationship between early model estimates of rEF and later mortality.

Methods: A compact convolutional transformer deep learning model was trained on digitized 12-lead ECG waveforms to generate a scaled probability of rEF, defined as left ventricular ejection fraction ≤40%. ECGs obtained within ±14 days of transthoracic echocardiography (TTE) demonstrating rEF were used as cases for model training, whereas ECGs within ±14 days of normal EF were controls. Patients with any cancer diagnosis and an available TTE at our center (n=88,339) were partitioned patient-wise into training, validation, and test cohorts in a 7:1:2 ratio. Model selection was based on validation F1-score, and the optimal high-risk threshold was determined by validation cohort Youden’s index. The trained model was then applied to ECGs from a distinct cohort of adult patients undergoing first allo-HCT from 2020-2025 who had at least one ECG between days 0-28 post-HCT. Survival was estimated by the Kaplan-Meier method and cumulative incidence by non-parametric Aalen-Johansen estimation. Multivariable Cox and competing risk regression models were built using stepwise parameter selection based on minimization of Akaike Information Criterion. Two-sided p<0.05 was considered significant.

Results: In the test cohort (n=13,251; rEF prevalence 2.7%), AI-ECG-rEF achieved an area under the receiver operating characteristic curve of 0.933 (95% confidence interval: 0.920-0.944), accuracy of 88.5% (88.0-89.0%), sensitivity of 81.2% (77.0-85.2%), specificity of 88.7% (88.2-89.3%), and diagnostic odds ratio of 33.9 (25.0-48.0).

The allo-HCT cohort included 750 patients: 414 received standard dose PTCy (50 mg/kg/d) and 336 received non-PTCy GVHD prophylaxis, including calcineurin inhibitor-based regimens (n=244) and ex vivo graft manipulation (n=92). Median age was 63.6 years (interquartile range: 52.4-69.0), 411 patients (55%) were male, and 326 (43%) had high-risk HCT-CI scores (≥3). Median follow-up was 20.0 months. The most common indications for HCT were acute myeloid leukemia (42%), myelodysplastic or myeloproliferative syndromes (30%), and non-Hodgkin lymphoma (13%). Myeloablative conditioning was used in 238 patients (32%). The most used regimen was fludarabine/melphalan (47%). Donors were HLA 8/8 matched in 485 patients (65%). Pre-HCT EF was similar across different GVHD prophylaxis programs.

By day 28 post-HCT, the cumulative incidence of AI-ECG-rEF+ was higher among PTCy recipients than non-PTCy recipients: 34.3% (29.7-38.9%) versus 24.1% (19.5-28.7%). After adjustment for age, disease, and conditioning intensity, PTCy remained associated with increased risk of AI-ECG-rEF+ (HR, 1.51; 1.15-1.97; p=0.003). In a day-28 landmark analysis of survivors (n=729), early AI-ECG-rEF+ was associated with increased late mortality after adjustment for conditioning intensity, age, and sex (HR, 1.81; 1.36-2.41; p<0.001). This excess mortality was driven by non-relapse mortality (adjusted HR, 2.64; 95% CI, 1.80-3.88; p<0.001), without a significant difference in relapse risk (adjusted HR, 1.03; 95% CI, 0.70-1.51; p=0.883). Findings were robust in sensitivity analyses to exclusion of recipients of ex vivo graft manipulation. In an exploratory analysis, the day-28 cumulative incidence of AI-ECG-rEF+ was lower among 34 patients receiving reduced-dose PTCy (17.6%; 4.6-30.7%) compared to standard-dose PTCy (unadjusted p=0.002).

Conclusions: We propose AI-ECG-rEF as a novel and scalable approach to automated monitoring for acquired cardiac dysfunction after allo-HCT. Early AI-ECG-rEF occurred more frequently after PTCy and was independently associated with increased late mortality, driven by non-relapse mortality. In preliminary analyses, reduced-dose PTCy was associated with a lower incidence of AI-ECG-rEF, warranting further prospective evaluation.

Husayn Mohamed Ladha: No relevant conflicts of interest to declare.