Prognostic Validation of IPSS-M Versus ELN 2022/2024 in MDS/AML Patients: A Survival and Treatment Response Analysis.

The 2022 International Consensus Classification (ICC) introduced myelodysplastic syndrome/acute myeloid leukemia (MDS/AML) as a distinct entity for patients with 10%-19% marrow blasts lacking AML-defining genetic abnormalities. The comparative prognostic utility of the Molecular International Prognostic Scoring System (IPSS-M), derived from MDS, and the European LeukemiaNet 2022/2024 (ELN 2022/2024) classification, designed for AML, within this overlap syndrome remain undefined. We retrospectively analyzed 110 adults with ICC-defined MDS/AML treated at West China Hospital (2019-2024). Baseline risk was assessed using IPSS-M and ELN 2022. Treatment-matched exploratory analyses compared IPSS-M with ELN 2022 in intensively treated patients and with the 2024 ELN Less-Intensive genetic-risk classification in patients receiving HMA monotherapy or HMA plus venetoclax. Prognostic discrimination was assessed using the time-dependent area under the curve (AUC) and category-based Harrell's concordance index. Most patients were at high risk according to the IPSS-M (high/very high, 90.0%) and ELN 2022 (adverse, 79.1%). During a median follow-up of 36 months, the median OS was 25 months. In the full cohort, IPSS-M showed better discrimination than ELN 2022 did (C-index, 0.578 vs. 0.518). In the primary HMA-based cohort (n = 62), ELN 2024 classified 42 patients as favorable, 7 as intermediate, and 13 as adverse; the median OS was 33, 18, and 12 months, respectively (log-rank p = 0.005). ELN 2024 showed greater discrimination than IPSS-M did (C-index, 0.605 vs. 0.562), but the difference was not significant. Among the 91 response-evaluable patients, the highest observed ORR (83.3%) and CR rate (54.2%) were obtained for HMA plus venetoclax. IPSS-M provided better overall discrimination than ELN 2022 did in this MDS/AML cohort, whereas the 2024 ELN Less-Intensive classification produced clinically relevant risk separation and numerically higher discrimination in HMA-treated patients. Model performance was treatment context dependent, and subgroup results require validation in larger prospective cohorts.