Unifying the classification of acute myeloid leukemia, myelodysplasia-related: A multi-institutional experience Free

WHO^5th and ICC²⁰²² include AML myelodysplasia-related (AML-MR) as a new diagnostic entity that incorporates the mutation profile, replacing AML with myelodysplasia-related changes. Although the diagnostic criteria are largely overlapping, there are notable differences between WHO^5th and ICC²⁰²², including inclusion of RUNX1 in the MR gene list, whether AML-MR based on cytogenetic abnormalities (CGA) or gene mutations (GM) should represent unique entities, and whether AML with TP53 mutations should be included. Additionally, the impact of AML ontogeny (history of cytotoxic therapy and antecedent MDS or MDS/MPN) on AML-MR outcome is uncertain. We aimed to address these questions to refine the definition of AML-MR as a distinctive AML subtype.

A total of 619 patients with AML, newly diagnosed and treated at Memorial Sloan Kettering Cancer Center (MSKCC, n=408; diagnosed in 2014-2025) and Mass General Hospital (MGH, n=211; diagnosed in 2012-2025) with complete cytogenetic and mutational profiling by targeted NGS, were retrospectively reviewed and included. Their clinical, histopathological and molecular characteristics were manually annotated. Overall survival (OS) was estimated by Kaplan-Meier method for entire cohort and genetic and clinical subgroups. Associations between genetic and clinical characteristics and OS were assessed using univariable and multivariable Cox proportional-hazard models while stratifying by centers (MSK vs MGH) or types of induction treatment.

The median age was 65 years (range: 18 to 92 years) with a M:F ratio of 3:2. All cases were classified based on WHO^5th and ICC²⁰²². Cases meeting diagnostic criteria for AML-MR (n=322) were subclassified into 4 groups based on GM and CGA carried: 1. RUNX1: RUNX1 mutations without other MR-GM; 2. MR: any MR-GM exclusive of RUNX1; 3: RUNX1-MR: mutations in RUNX1 and at least one other MR gene; 4: MR-CG: MR-CGA based on either WHO^5th or ICC²⁰²² classification. Four types of algorithms were used when assigning cases to the above 4 groups for statistical analysis: prioritizing MR-GM over MR-CGA or vice versa for classification, with MR-CGA assignment based on either WHO^5th or ICC²⁰²².

Estimates of OS by Kaplan-Meier analysis for AML diagnostic categories showed 4 distinctive outcome groups (p<0.001): AML with inv(16) or t(8;21) has the most favorable outcome (5 year survival rate: 82%; 95%Cl: 70%, 97%), followed by AML with bZIP CEBPA, NPM1 and AML-NOS (55%; 95%Cl: 46%, 65%), AML-MR (31%; 95%Cl: 25%, 39%), and AML with TP53 (4.1%; 95%Cl: 1.2%, 14%). Furthermore, no significant difference among 4 groups of AML-MR on OS was observed regardless of the algorithm used (p=0.3 to 0.9), indicating similar outcomes of AML with RUNX1, MR, MR-RUNX1 GM and MR-CGA.

Univariate analysis on all patients identified age, performance status, AML ontogeny, cytogenetic risk categories, complex karyotype (CK), stem cell transplant (SCT), and induction intensity as prognostic factors. Multivariate analysis after stratifying treatment and centers (MSK vs MGH) revealed that sex, performance status, AML ontogeny, CK, and SCT remained significant prognostic factors.

Univariate analysis only on the 4 MR groups identified sex, performance status, AML ontogeny, CK, and SCT as prognostic factors. In multivariate analysis, only SCT remained significant. Importantly, multivariate analysis stratified by treatment and centers revealed no significant difference on OS among the 4 groups irrespective of the algorithm used.

Subgroup analysis was performed based on initial treatment, high-intensity (e.g. 7+3, n=337) vs HMA/Ven regimen (n=276). When OS was compared among the 4 MR groups, no significant difference was seen in either high-intensity or HMA/Ven cohorts, irrespective of the algorithm used. The presence of CK in AML-MR was associated with inferior OS in HMA/Ven cohort on univariate analysis, however, only SCT but not CK was significantly associated with OS on multivariate analysis.

In summary, our findings support that: 1) RUNX1 should be included in MR gene list; 2) MR CGA and GM have similar prognostic impact; 3) CK and AML ontogeny do not represent risk factors that are independent of assignment to AML-MR based on the genetic profile. Our data provide evidence of unifying the classification of AML-MR as a prognostically distinctive subtype with superior prognosis to AML with TP53 mutation, but inferior to other AML diagnostic categories.