AUTHORS
Sophia Maharry Roush, PhD, Deedra Nicolet, MS2, Preem Geeta Torlapati, MD, MPH3, Krzysztof Mrózek, MD, PhD4, Kristin Koenig, MD5, Yara Edor6, Daniel Pollyea, MD7, William Blum, MD8, Andrew Carroll, PhD9, Richard Stone, MD10, John Byrd, MD11, Electra D Paskett, PhD12, Alexander Joseph Ambinder, MD, MPH13, Andrew T Kent, MD, PhD14, Jesse J Plascak, PhD, MPH15, Elaine Mardis, PhD16, Ann-Kathrin Eisfeld, MD17
AFFILIATIONS
1 University of North Carolina, Chapel Hill, United States
2 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
3 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
4 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
5 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
6 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
7 University of Colorado, Aurora, Colorado, United States
8 Winship Cancer Institute of Emory University, Atlanta, Georgia, United States
9 The University of Alabama at Birmingham, Birmingham, Alabama, United States
10 Dana Farber Cancer Institute, Boston, Massachusetts, United States
11 University of Pittsburgh Medical Center, Pittsburgh, Pennsylvania, United States
12 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
13 Johns Hopkins University, Baltimore, Maryland, United States
14 University of Colorado, Aurora, Colorado, United States
15 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
16 Nationwide Children's Hospital, Columbus, Ohio, United States
17 The Ohio State University Wexner Medical Center, Columbus, Ohio, United States
ABSTRACT
Background: AML predominantly affects older adults and children, leaving AYA pts in a therapeutic gap where care is guided by protocols derived from older pts. While several studies have suggested differences in disease biology, a comprehensive characterization of AYA AML and assessment of current prognostic algorithms in this population is lacking.
Methods: We analyzed 1,389 AYA pts (age 18–39y) treated at academic centers through the Myeloid Malignancy Association on Rapid Research Outcomes Working Group (MARROW; n=393, 2005–2024), frontline Alliance for Clinical Trials in Oncology protocols (n=456, 1986–2012), and in the real-world setting using the Flatiron Health database (n=540, 2014–2025). Analyses were performed on all pts and separately by age group (18–29y vs 30–39y). Multivariable Cox regression (DFS, OS) and logistic regression (CR) were performed in 7+3-treated pts. Clinical and molecular features of AYA pts were also compared with those of 2,882 pts aged 40–60y.
Results: The median age was 31y, 50% of pts were female, 27% were non-White, 45% had a normal karyotype, 25% CBF-AML and 8% KMT2A rearrangements (KMT2A-r). The most frequent mutations (mut) were in NRAS (26%), FLT3-ITD (23%), NPM1 (21%), DNMT3A (13%), WT1 (12%) and KRAS (11%). There were no significant molecular differences between 18–29y and 30–39y pts. Compared with 40–60y pts, AYA pts had more NRAS, KRAS, KIT, and WT1 mut and fewer NPM1, DNMT3A, IDH2, TET2, TP53 and SRSF2 mut (all p≤.004). Median OS for all AYA pts was 6.2y and DFS was 2.6y, and survival was comparable between 18–29y and 30–39y pts.
Remarkably, ELN 2022 did not seem to effectively stratify pts. Among 18–29y pts, outcomes of Intermediate-risk and Adverse-risk pts were nearly identical, whereas in 30–39y pts, Intermediate-risk pts had outcomes resembling Favorable-risk pts. On direct comparison, Intermediate-risk pts aged 18–29 had worse DFS than those aged 30–39 (p=.04), and Adverse-risk pts aged 30–39 had worse OS than their 18–29 counterparts (p=.002). Across both AYA age subgroups, FLT3-ITD/WT1 double-mut pts had particularly poor DFS and OS.
For CR, NPM1 mut (OR 4.87, p<.001) and CBF-AML (OR 3.66, p<.001) were associated with higher rates, while FLT3-ITD (OR 0.41, p=.001) and PTPN11 mut (OR 0.33, p=.006) were associated with lower rates. For DFS, complex karyotype (CK; HR 3.84, p<.001) and WT1 mut (HR 1.66, p=.005) predicted shorter DFS, whereas PTPN11 mut was protective (HR 0.51, p=.03), almost exclusively in the setting of co-mutated NPM1. For OS, CK (HR 2.34, p<.001) and KMT2A-r (HR 1.56, p=.007) predicted worse outcomes, while HCT (HR 0.42, p<.001), CBF (HR 0.33, p<.001), and NPM1 mut (HR 0.65, p=.001) were protective.
On multivariable analysis within each age group, BCOR mut emerged as an additional worse outcome predictor in 18–29y pts (DFS: HR 2.73, p=.02; OS: HR 3.32, p=.003), while DNMT3A (DFS: HR 1.62, p=.03) and NRAS mut (DFS: HR 1.42, p=.05; OS: HR 1.66, p=.03) predicted worse survival in 30–39y pts. We therefore proposed a refined ELN model for all AYAs that added WT1/FLT3-ITD double-mut and PTPN11 mut without NPM1 mut as Adverse, WT1 mut or FLT3-ITD alone as Intermediate, and NPM1/PTPN11 co-mut as Favorable, with subgroup-specific refinements for BCOR mut (18–29y), and DNMT3A and NRAS mut (30–39y). This model separated risk groups more effectively than ELN 2022 for both OS and DFS, overall and within each age subgroup.
Among non-HCT pts, CK and KMT2A-r predicted worse OS, and CBF-AML shorter DFS. Among HCT recipients, CK (HR 2.78, p=.001) and WT1 mut (HR 2.04, p<.001) predicted shorter DFS, U2AF1 mut (HR 3.06, p=.007) was associated with worse OS, and CBF-AML (HR 0.52, p=.03) was associated with improved OS. Residential zip code social deprivation index, race, and ethnicity were not associated with outcomes overall, but in the 18–29y subset White pts had better OS than non-White pts (HR 0.59, p=.01) and high social deprivation index conferred worse OS in 30–39y pts (HR 1.71, p=.03).
Conclusions: AYA AML represents a biologically and clinically distinct pt population that is incompletely characterized by current genetic risk stratification. ELN 2022 performance varied by age subgroup, with the most pronounced deficits in the Intermediate-risk group, carrying potential implications for HCT allocation. These data support AYA-specific risk stratification and prospective validation to guide treatment decisions in this underserved population.
DISCLOSURES
Sophia Maharry Roush, PhD: No relevant conflicts of interest to declare.