Abstract Venetoclax (VEN) has shown promising efficacy when combined with intensive chemotherapy regimens—such as FLAG-IDA (DiNardo et al. 2025), DA (Mantzaris et al. 2025), and others—in newly diagnosed acute myeloid leukemia (AML). While current risk classification systems such as the European LeukemiaNet 2022 (ELN2022) for intensive chemotherapy and ELN2024 for VEN plus hypomethylating agents (VEN+HMA) provide standardized frameworks, they may not fully capture the prognostic heterogeneity among patients receiving VEN-based intensive therapy. To address this gap, we conducted a real-world study to construct a refined prognostic model based on molecular features for patients receiving VEN-intensive regimens (ICV) in the first two treatment cycles. Eligible patients were newly diagnosed with AML and received venetoclax in combination with intensive chemotherapy during the first and/or second induction cycle. Therapy regimens included venetoclax with or without a hypomethylating agent (VEN±HMA), combined with one of the following: homoharringtonine-based therapy, anthracycline-based regimens, purine analog-containing regimens, or intermediate-/high-dose cytarabine-based protocols. A total of 409 newly diagnosed AML patients were included. The median age was 49 years (range 11–76), with 37.9% diagnosed with acute monocytic leukemia. Complex or monosomal karyotypes (CK/MK) were observed in 8.1%. Fusion genes included RUNX1-RUNX1T1 (14.2%), CBFB-MYH11 (8.8%) and inv(3) (1.0%). Additional cytogenetic abnormalities included -17/del(17p) (1.5%), -7 (4.2%), and -5/del(5q) (2.0%). Frequently mutated genes included FLT3-ITD (21.8%), NPM1 (17.8%), CEBPA bZIP (9.0%), KMT2A rearrangement (7.3%), RUNX1 (6.1%), and TP53 (3.7%). Univariable analysis revealed that inv(3), -5/del(5q), -7, CK/MK, and mutations in TP53, FLT3-ITD and RUNX1 were significantly associated with inferior OS, whereas RUNX1-RUNX1T1, CBFB-MYH11, and mutated NPM1 without FLT3-ITD conferred favorable survival. The prognosis was not significantly associated with monocytic development (p=0.640). Using a bootstrap-based variable selection approach with the random survival forest algorithm and Cox proportional hazards models, we identified several top significant “beneficial” and “harmful” genetic and cytogenetic features. Based on these results, we developed a novel risk model (ICV-CHN) which stratified patients into four molecularly defined risk groups. The favorable-risk category comprised patients with CBFB-MYH11, RUNX1-RUNX1T1, or mutated NPM1 without FLT3-ITD. Patients with intermediate-risk features were further divided based on their risk gene profiles. The Intermediate-1 group comprised patients not fitting other risk categories, representing standard/unclassified risk. The Intermediate-2 group included those with moderately adverse features such as CK/MK, -7, -5/del(5q), or mutations in RUNX1 or FLT3-ITD—known to attenuate outcomes despite potential sensitivity to targeted therapy. And the adverse-risk category included patients harboring inv(3) or TP53 mutations. According to the ICV-CHN model, the 2-year OS rates of low-risk/intermediate-risk-1 patients were 87.0% with chemotherapy alone and 90.2% with transplantation (p=0.350), while those of intermediate-risk-2/high-risk patients were 47.8% and 73.2%, respectively (p0.001). Comparative analyses demonstrated that the ICV-CHN model outperformed ELN-based classifications. Under ELN2022, pairwise log-rank tests showed no significant OS difference between intermediate- and adverse-risk groups (p=0.440), with a 2-year time-dependent area under the curve (AUC) of 0.66 and concordance index (c-index) of 0.644. ELN2024 yielded statistically significant pairwise differences but had a lower 2-year AUC of 0.62 and c-index of 0.617. In contrast, the ICV-CHN 2025 model demonstrated significant OS separation across all strata (2-year OS: Fav. vs Int1 vs Int2 vs Adv: 91.4% vs 82.9% vs 68.1% vs 24.3%, p0.001), with a mean 2-year AUC of 0.73 and mean c-index of 0.72, indicating improved prognostic performance. The model was further verified in an external validation cohort. In concclusion, the ICV-CHN model offers a molecularly grounded and clinically applicable risk stratification framework for AML patients undergoing VEN-intensive therapy. This model enhances outcome prediction beyond current ELN classifications and may support more personalized treatment approaches in real-world settings.
Pei et al. (Mon,) studied this question.