C050 | VALIDATION OF ELN 2024 AND BEAT-AML2024 RISK STRATIFICATION SYSTEMS IN PATIENTS WITH ACUTE MYELOID LEUKEMIA NOT ELIGIBLE FOR INTENSIVE CHEMOTHERAPY IN A MULTICENTRIC ITALIAN COHORT

Acute myeloid leukemia (AML) frequently affects older adults, with a peak incidence at age 68 and two-thirds of cases diagnosed after age 55. Treatment decisions depend on disease biology and patient-specific factors such as comorbidities and performance status. The European LuekemiaNet (ELN)2022 risk model, developed for fit patients receiving intensive therapy, is less suitable for unfit patients. To address this, two new models—ELN 2024 and BEAT-AML 2024—have been proposed for those receiving non-intensive treatment. This study evaluates which model best predicts 2-years survival in unfit AML patients in Italy, highlighting strengths, limitations, and potential strategies to identify candidates for further reduced-intensity therapies. We analyzed 171 unfit AML patients from 8 Italian centers. Molecular and cytogenetic data were used to apply ELN22, ELN24, and BEAT-AML24 classifications. Mutations were assessed via NGS or conventional methods; cytogenetics followed ISCN guidelines. Patients received hypomethylating agents (HMA) with or without venetoclax. Risk distributions were analyzed using contingency tables. Survival was evaluated with Kaplan Meier curves, log-rank tests and Cox regression. Model performance was assessed using the C-index. Median age was 74.8 years. Of the patients, 132 received azacitidine plus venetoclax, 31 HMA alone, and 8 other treatments. Patients were stratified according to the ELN22, ELN24, and BEAT-AML24 risk models. ELN22 and ELN24 showed significant survival differences only between adverse and other groups but failed to distinguish between favorable and intermediate risks. Conversely, the BEAT-AML24 model significantly differentiated all three risk groups (p<0.0001), with 2-year OS of 48.7% (CI:35.5–66.9%) for favorable, 25.6% (CI:15.8–41.6%) for intermediate, and 5.2% (CI:0.8–34.4%) for adverse risk (Fig1). The C-index was highest for BEAT-AML24, confirming its better predictive performance (0.606). The BEAT-AML24 model proved to be the only statistically significant tool with a higher C-index, outperforming ELN22 and ELN24 in stratifying unfit AML patients. Nonetheless, the study has limitations: only two centers routinely analyzed mutations in DDX41, BCOR, and STAG2. Moreover, these models require both karyotyping and NGS, which are costly and not yet supported by definitive therapeutic benefits. Further research is needed to refine stratification and identify patients who may benefit from reduced-intensity regimens.