Abstract
Background: Risk stratification is essential in guiding treatment decisions in acute myeloid leukemia (AML). While the European LeukemiaNet (ELN) classifications provide a standardized framework, their utility in nonintensively treated patients remains unclear. This retrospective study evaluates the prognostic performance of ELN 2022 and ELN 2024 in AML patients receiving nonintensive frontline therapy.
Methods: Patients treated with nonintensive treatment from May 2020 to May 2025 were included in this retrospective analysis. Patients were categorized according to ELN 2022 and ELN 2024 criteria. Discrimination between risk groups was assessed using Kaplan-Meier survival curves, Harrell's concordance index (C-index), and time-dependent area under the ROC curve (AUC). Cox regression analysis was performed to identify prognostic factors, and a scoring system was subsequently constructed based on the multivariable model.
Results: A total of 66 patients with newly diagnosed AML were treated with nonintensive regimens. All patients received nonintensive regimens containing venetoclax (47 with azacitidine, 16 with decitabine, and 3 with low-dose cytarabine). Median age was 77 years (range, 51–89), and 43 patients (65.2%) were male. Among 61 evaluable patients, 18 (29.5%) achieved complete remission (CR), 19 (31.1%) achieved CR with incomplete hematologic recovery (CRi), 9 (14.8%) had partial response, and 15 (24.6%) had no response. The composite CR rate (CR + CRi) was 60.7%. Median overall survival was 15.0 months (95% confidence interval, [95% CI], 12.1-17.9), 9.6 months (95% CI, 7.7-12.1), 2.5 months (95% CI, 1.4-3.7) for favorable, intermediate, and adverse risk groups by ELN 2024.
Both ELN systems showed discernible stratification of survival outcomes. However, ELN 2022 showed limited separation between intermediate and adverse risk groups, whereas ELN 2024 showed overlapping survival curves between favorable and intermediate groups. The C-index was 0.789 and 0.734 for ELN 2022 and 2024, respectively, with the difference not reaching statistical significance (p = 0.469). Time-dependent ROC analysis revealed superior AUCs for ELN 2022 across all time points; at 1 year, the AUC was 0.730 for ELN 2022 and 0.670 for ELN 2024, with the difference not reaching statistical significance (p = 0.476).
In univariable Cox regression analysis, several factors were associated with overall survival, including TP53 mutation, complex karyotype (CK), and IDH1 mutation. These three variables were subsequently included in a multivariable Cox proportional hazards model, in which TP53 (HR, 2.58; 95% CI, 1.02–6.50; p = 0.045) and IDH1 (HR, 0.12; 95% CI, 0.016–0.93; p = 0.042) remained significant. A prognostic score was generated by assigning points based on normalized β coefficients: 2 points for TP53, 1 point for CK, and –3 points for IDH1. Patients were stratified into high-risk (score > 1) and low-risk (score ≤ 1) groups based on the optimal cut-off. Median overall survival was 13.7 months (95% CI, 10.0–20.2) in the low-risk group (n = 53) and 2.5 months (95% CI, 1.5–not estimable) in the high-risk group (n = 13), with a significant difference in survival (log-rank p <0.001). The risk model showed excellent discriminatory ability with a Harrell's C-index of 0.81 (95% CI, 0.73–0.89).Conclusion: In this cohort of nonintensively treated AML patients, neither ELN 2022 nor ELN 2024 clearly delineated all three prognostic categories. While each system demonstrated some prognostic relevance, their ability to distinctly separate favorable, intermediate, and adverse risk groups was limited. These findings highlight the need for improved risk categorization tailored specifically to patients receiving nonintensive treatment.
