Expanding the High-Risk Definition for Children with Newly Diagnosed Acute Myeloid Leukemia

Introduction: The current Children’s Oncology Group (COG) phase III trial AAML1831 incorporates one of the most up to date risk classification systems for pediatric acute myeloid leukemia (AML) but ongoing evaluation and refinement is necessary. With larger sample sizes and improved next generation sequencing (NGS), novel variants as well as those previously considered to have a neutral impact on outcomes are emerging as new prognostic biomarkers. Stem cell transplant (SCT) in first complete remission (CR1) is the standard of care for patients with high-risk (HR) AML and systematic evaluation of an expanded HR definition is required prior to updating the contemporary risk classification. In this study, we identify candidate alterations for risk adjustment and assess the retrospective impact of an updated risk classification on treatment allocation in 2 large trials for de novoAML.

Methods: Diagnostic specimens from 2 sequential COG phase III trials for de novo pediatric AML (NCT01407757 and NCT01371981) were retrospectively interrogated with NGS. This data was coalesced with central review of structural and molecular alternations to identify candidates for risk adjustment. Alterations considered for risk adjustment needed to have a 5-year event-free survival (EFS) inferior to similarly treated patients with noninformative cytomolecular genetics (standard-risk, SR) with and/or without censoring for SCT in CR1. Additionally, these outcomes needed to be reproduced over both trials. Candidate alterations were then incorporated into an updated risk classification which was retrospectively applied to assess impact on treatment allocation.

Results: Of 1944 patients, 137 (7%) patients with 1 of 6 newly defined alterations met the predefined criteria for risk adjustment. These included CEBPA co-occurring with CSF3R (n=18, EFS 33%), CREBBP mutations (n=40, EFS 18%), TP53 mutations (n=20, EFS 0%), KMT2A-partial tandem duplications (n=21, EFS 43%), IDH1/2 mutations in the absence of NPM1 (n=36, EFS 31%), and fusions involving the ETS family, excluding previously high-risk defined members (n=13, EFS 43%). Thirty-eight (28%) had co-occurring HR alterations, 42 (31%) had co-occurring low-risk (LR) alterations and the remaining 57 (42%) were SR.

As a group, patients with candidate alterations had an EFS of 28% and relapse risk of 70%, which were worse than contemporarily defined SR patients and analogous to contemporarily defined HR patients (Figure 1a and 1b). When patients with co-occurring HR alterations were excluded, these patients maintained their poor EFS. Specifically, patients with co-occurring LR alterations had an EFS of 31% and previously considered SR patients had an EFS of 26%. As a group, patients with candidate alterations had an overall survival (OS) of 55% which was slightly better compared to contemporarily defined HR patients (41%). This was driven by candidate patients with co-occurring LR alterations that were largely salvageable, based on a favorable OS of 68%.

Following incorporation of these candidate alterations into an updated risk classification, there was a redistribution of patients in each risk group compared to contemporarily defined risk groups. There was an increase in HR patients (AAML0531 32% vs. 36%; AAML1031 33% vs. 39%) and compensatory decrease in LR (AAML0531 37% vs. 35%; AAML1031 38% vs. 36%) and SR patients (AAML0531 31% vs. 29%; AAML1031 29% vs. 25%). With incorporation of minimal residual disease to further stratify SR patients, this updated risk classification would have led to 75 (3.9%) more patients being allocated to SCT in CR1 compared to the contemporary system.

Conclusions: Retrospective incorporation of candidate HR alterations into COG’s contemporary risk classification system would have increased the number of patients allocated to allogenic donor SCT. While all 6 candidate alterations had a consistently inferior EFS over successive trials, patients w/ co-occurring LR alterations had a more favorable OS, suggesting they can be salvaged via SCT. Before these alterations can be considered HR, further investigation is required, including corroboration from other large pediatric and adult consortium trials and the benefit of this re-allocation will need to be validated prospectively.