Measurable Residual Disease and Fusion Partner Independently Predict Survival and Relapse Risk in Childhood KMT2A-Rearranged Acute Myeloid Leukemia: A Study by the International Berlin-Frankfurt-Münster Study Group

PURPOSE A previous study by the International Berlin-Frankfurt-Münster Study Group (I-BFM-SG) on childhood KMT2A-rearranged (KMT2A-r) AML demonstrated the prognostic value of the fusion partner. This I-BFM-SG study investigated the value of flow cytometry-based measurable residual disease (flow-MRD) and evaluated the benefit of allogeneic stem-cell transplantation (allo-SCT) in first complete remission (CR1) in this disease. METHODS A total of 1,130 children with KMT2A-r AML, diagnosed between January 2005 and December 2016, were assigned to high-risk (n = 402; 35.6%) or non–high-risk (n = 728; 64.4%) fusion partner-based groups. Flow-MRD levels at both end of induction 1 (EOI1) and 2 (EOI2) were available for 456 patients and were considered negative (<0.1%) or positive (≥0.1%). End points were 5-year event-free survival (EFS), cumulative incidence of relapse (CIR), and overall survival (OS). RESULTS The high-risk group had inferior EFS (30.3% high risk v 54.0% non-high risk; P < .0001), CIR (59.7% v 35.2%; P < .0001), and OS (49.2% v 70.5%; P < .0001). EOI2 MRD negativity was associated with superior EFS (n = 413; 47.6% MRD negativity v n = 43; 16.3% MRD positivity; P < .0001) and OS (n = 413; 66.0% v n = 43; 27.9%; P < .0001), and showed a trend toward lower CIR (n = 392; 46.1% v n = 26; 65.4%; P = .016). Similar results were obtained for patients with EOI2 MRD negativity within both risk groups, except that within the non–high-risk group, CIR was comparable with that of patients with EOI2 MRD positivity. Allo-SCT in CR1 only reduced CIR (hazard ratio, 0.5 [95% CI, 0.4 to 0.8]; P = .00096) within the high-risk group but did not improve OS. In multivariable analyses, EOI2 MRD positivity and high-risk group were independently associated with inferior EFS, CIR, and OS. CONCLUSION EOI2 flow-MRD is an independent prognostic factor and should be included as risk stratification factor in childhood KMT2A-r AML. Treatment approaches other than allo-SCT in CR1 are needed to improve prognosis.


INTRODUCTION
Most pediatric AML study groups (SGs) currently base risk stratification on genetics, including fusion genes and molecular aberrations, and early treatment response, either morphologically or more accurately assessed by detection of flow cytometry-based measurable residual disease (flow-MRD). 1 Detection of flow-MRD in bone marrow (BM) after induction therapy is considered a strong indicator of relapse, [2][3][4][5][6][7] and can aid risk-directed postremission therapy, 2,5,6 including allogeneic stem-cell The benefit of allo-SCT in CR1 remains a debatable subject in pediatric AML, as its enhanced antileukemic activity needs to outweigh the risk of transplant-related mortality. 1,12,13 In previous pediatric KMT2A-r AML studies, allo-SCT in CR1 did not improve relapse risk, nor survival. 9,10 However, exposure to gemtuzumab ozogamicin (GO; Mylotarg, Pfizer, New York, NY) before transplantation seemed to improve the posttransplantation prognosis in these patients. 10 We aimed to evaluate the prognostic significance of end of induction 2 (EOI2) flow-MRD response and allo-SCT in CR1 in childhood KMT2A-r AML overall and within fusion partner-based risk groups.

Study Design and Patients
A retrospective study was conducted within the I-BFM-SG, including 15 pediatric AML SGs/countries. Eligible patients were younger than 19 years and were newly diagnosed with KMT2A-r AML between January 1, 2005, and December 31, 2016. Patients with a diagnosis of myeloid leukemia in Down syndrome, isolated myeloid sarcoma, or acute promyelocytic leukemia were excluded, as well as patients who were initially treated, for more than one week, for a diagnosis other than AML. Not all SGs/countries provided eligible patients for the entire study period, and treatment was given according to national or international pediatric AML SG trials (Data Supplement [ Table S1], online only), which were all cytarabine-/anthracycline-based. 2,4,6,[14][15][16][17][18][19][20][21][22][23][24] Institutional ethics committees approved these trials, and patients and/or parents provided written consent according to the Declaration of Helsinki. All data were checked for accuracy and corrected in consultation with the SGs/countries.

Assignment to Fusion Partner Groups and Risk Classification
The karyotype (G-, Q-, or R-banding), fluorescence in situ hybridization, and reverse transcription polymerase chain reaction (PCR) results were reviewed within the SGs/countries for the presence of KMT2A rearrangements. Patients were assigned to 10 different fusion partner groups or the KMT2A-other group, as previously reported by Balgobind et al. 9 The group assignment was reviewed by two authors (R.W. and C.H.). The fusion partners of KMT2A-other group patients were reviewed to identify novel groups, with a minimum of 10 patients. Patients with unidentified fusion partners or those occurring in less than 10 patients remained within the KMT2A-other group.

Flow-MRD Analysis
Flow-MRD analysis was implemented in most, but not yet in all, trials/treatment protocols used by the SGs/countries during the study period. Additionally, some SGs/countries could not provide flow-MRD data because of ongoing trials. Ten SGs/countries provided flow-MRD data, which were mainly detected using 4-to 10-color antibody panels. Two SGs also used the different from normal approach. Details on the BM cellularity were not collected.

Definitions and Statistical Analysis
CR was defined as ,5% BM blasts, the absence of cells with Auer rods and extramedullary disease, and peripheral blood cell regeneration. 1 Refractory disease was defined as $5% BM blasts, either morphologically, cytogenetically, or by a high positive PCR result, or proven extramedullary disease after induction therapy. Relapse was defined as $5% BM blasts or reappearance of blasts in peripheral blood, or the development of extramedullary disease after initial morphologic CR. 1 End of induction 1 (EOI1) and EOI2 BM responses were morphologically categorized as M1 (,5% blasts), M2 ($5%-,20% blasts), or M3 ($20% blasts). EOI1 and EOI2 flow-MRD responses ,0.1% were considered negative and $0.1% positive. Event-free survival (EFS) was defined as the time from diagnosis to induction failure, relapse, secondary malignancy, death in CR, or last follow-up, whichever occurred first. Induction failure was included as an event at t 5 0. Cumulative incidence of relapse (CIR) and nonrelapse mortality (NRM) were defined as the time from EOI1, for patients in CR, to relapse and to death without relapse, respectively. The competing event for CIR was death without relapse and for NRM death with relapse. Overall survival (OS) was defined as the time from diagnosis to death, or last follow-up.
Median (IQR) follow-up time of patients was 5.2 years (3.5-7.8). Differences in proportions were tested using Pearson's x 2 test. Medians between two groups were compared using the Mann-Whitney U test. EFS and OS estimates were calculated using the Kaplan-Meier method and compared using the log-rank test. The CIR was estimated by adjusting for competing risks and was compared using Gray's test. Hazard ratios (HR) were calculated using Cox proportional hazard models, wherein allo-SCT in CR1 was included as a time-dependent covariate. Covariates with two-sided P values , .05 in univariable analysis were included in multivariable analysis. To visually compare the CIR, NRM, and OS of allo-SCT versus no allo-SCT in CR1 overall and stratified by treatment era (2005-2010 v 2011-2016), a 90-day landmark was used. Multivariable analyses were performed excluding morphologic BM and flow-MRD response, referred to as the crude models containing the highest number of patients, and including morphologic BM and flow-MRD response separately from each other.
Analyses were performed using R version 4.0.3. Two-sided P values # .01 were considered statistically significant.

Univariable and Multivariable Analyses
In univariable analyses (Data Supplement,

DISCUSSION
After our previous I-BFM-SG study showing the prognostic value of the KMT2A fusion partner, 9 this novel study on childhood KMT2A-r AML is, to our knowledge, the first to demonstrate that EOI2 MRD positivity is an independent adverse prognostic factor in this disease, in addition to the fusion partner. This further demonstrates the need for risk stratification in childhood KMT2A-r AML, which constitutes 20%-25% of pediatric AML cases. In addition, allo-SCT in CR1 reduced relapse risk within the high-risk group but did not improve OS, as it  The pCIR could not be extrapolated to 5 years for patients within the high-risk group with EOI2 MRD positivity because all patients had an event before that time point. was counterbalanced by enhanced procedure-related toxicity.
We confirmed the independent adverse prognostic significance of previously defined high-risk KMT2A translocation partners (ie, 4q21, 6q27, 10p11.2, 10p12, and 19p13.3). 9,10 Therefore, our study serves as a consensus statement on fusion partner-based risk classification of childhood KMT2A-r AML, which will enable intergroup clinical trials and facilitate the performance of retrospective collaborative studies. The Children's Oncology Group (COG), as well as other SGs, incorporated these high-risk translocations into the initial risk stratification of their ongoing AAML1831 trial (ClinicalTrials.gov identifier: NCT04293562). 25,26 Hopefully, more SGs will follow. Interestingly, the EFS and OS rates of patients within the non-high-risk group with EOI2 MRD positivity were quite similar to those of patients within the high-risk group with EOI2 MRD negativity (EFS, 30.4% v 34.6%; OS, 43.5% v 52.3%). Notably, despite their good initial treatment response, the CIR rate of patients with EOI2 MRD negativity was 46%, which is markedly higher than the reported CIR rates of 17% and 32% of patients with EOI2 MRD negativity in pediatric AML in general. 2,5 This finding demonstrates the aggressive nature of childhood KMT2A-r AML. Relapses may be inherent to KMT2A genetic features, causing the emergence of leukemic stem cells not detected by flow-MRD after killing the leukemic cell bulk. 27 Leukemic stem-cell frequency assessment at diagnosis and during treatment may be included in future studies to further improve the identification of children with KMT2A-r AML at risk of relapse. 28,29 Alternatively, the limited sensitivity of the 4-to 10-color antibody panels used for flow-MRD detection in our study may have played a role. It needs to be determined whether flow-MRD detection on the basis of up-to-date methodology with 10 or more color approaches and centralized quality control, 30,31 PCR-based techniques, or next-generation sequencing (NGS) may at least partially overcome this limitation in sensitivity.
Among patients with EOI2 MRD negativity, the high-risk group showed significantly poorer outcomes than the nonhigh-risk group, consistent with a previous study on children with AML with MRD negativity that identified t(6;11) and t(10; 11)-breakpoints that were not further specified-as independent adverse prognostic factors. 32 By contrast, among patients with EOI2 MRD positivity, prognosis was not significantly influenced by the KMT2A risk group.
On the basis of our findings, patients within the high-risk group, irrespective of flow-MRD response, and patients within the non-high-risk group with EOI2 MRD positivity may benefit from high-risk-based treatment or novel treatment approaches, including experimental therapy. Allo-SCT in CR1 is generally used as high-risk-based treatment. The allo-SCT in CR1 rate in our study was slightly higher than in the previous I-BFM-SG study 9 (21% v 14%) but relatively low by modern standards, and given the high relapse risk of children with KMT2A-r AML. This may be explained by the fact that in the treatment protocols used during our study period, highrisk KMT2A translocations were not always considered a transplantation indication or that transplantation was restricted to those with available human leukocyte antigenmatched donors. To our knowledge, we are the first to show that allo-SCT in CR1 reduced the relapse risk in patients within the high-risk group, but not in patients within the nonhigh-risk group, nor within smaller groups on the basis of KMT2A risk group and EOI2 MRD response. This should be interpreted cautiously, as the overall transplantation rate was low and this study was not powered to assess the effect of allo-SCT in CR1. Furthermore, allo-SCT in CR1 did not improve OS, consistent with previous findings. 9,10 This is most likely due to insufficient eradication of the disease and, although significantly reduced in the most recent years, transplantation-related mortality.
Our results highlight the need for new treatment approaches to improve the prognosis of children with KMT2A-r AML. The COG demonstrated that GO added to induction therapy improved EFS and reduced relapse risk in these patients. 10 Additionally, GO seemed to improve post-transplantation prognosis. 10 These results need to be further confirmed to definitively establish which children with KMT2A-r Receiving allo-SCT in CR1 was not included in the multivariable analysis of EFS because, to receive allo-SCT in CR1, patients had to achieve CR, and EFS analysis also included induction failures. AML show the greatest benefit from intensified treatment strategies. In the current MyeChild01 (ClinicalTrials.gov identifier: NCT02724163) and Japanese Pediatric Leukemia/ Lymphoma SG-AML-20 (jRCTs041210015) trials, 33 GO added to induction and postinduction therapy is being studied, respectively. Other potentially promising agents are menin inhibitors, which have shown a significant reduction in leukemic cell load in mice engrafted with KMT2A-r AML. 34 Additionally, in two recent phase 1 studies, treatment with the menin inhibitors SNDX-5613 (revumenib) 35 and KO-539 (ziftomenib) 36 appeared safe and showed encouraging clinical responses in patients with relapsed/refractory KMT2A-r acute leukemia and AML, respectively.
Although flow cytometry is applicable in approximately 90% of children with AML, MRD cannot be detected in all patients because it requires extensive expertise to distinguish AML cells from normal, regenerating BM, which may only be available in larger experienced laboratories. 5,30,37 Considerable efforts are being made to standardize sample preparation and analyses. 5,30 This largest cohort of children with KMT2A-r AML serves as a highly valuable historical cohort for future pediatric AML SG collaborations in the pre-menin inhibitor era. Our study was limited by its retrospective design, the use of nonuniform treatment protocols with different risk stratifications across SGs, the overall low transplantation rate, and the lack of flow-MRD data at both time points in more than half of the patients. In some cases, MRD availability might have interacted with risk stratification and/or treatment allocation, including allo-SCT in CR1. Nonetheless, the relatively old and largely nonstandardized flow-MRD data from our cohort allowed us to discern flow-MRD-based risk associations independent of the fusion partner-based risk group.
We conclude that EOI2 flow-MRD response and fusion partner-based risk group should be included as risk stratification factors in childhood KMT2A-r AML. Patients within the non-high-risk group with EOI2 MRD negativity may be assigned to the standard-risk arm of treatment protocols but should be closely monitored for MRD after remission. All other patients (Data Supplement, Figure S3) should be assigned to the high-risk arm. Future studies should establish whether allo-SCT in CR1 will be the best risk-adapted treatment for all patients assigned to the highrisk-arm. New treatment approaches, including GO and menin inhibitors, are urgently needed for this subset of patients. The implementation of optimized stratification approaches, the increased availability of flow-MRD assays, quantitative PCR, and NGS among pediatric AML SGs worldwide, and the use of agents able to kill KMT2A-r AML cells more efficiently could contribute to improved survival of children with KMT2A-r AML, for whom international collaborative research remains indispensable.