Measurable residual disease monitoring in AML with FLT3-ITD treated with intensive chemotherapy plus midostaurin

Internal tandem duplications of the FLT3 gene (FLT3-ITD) are found in ∼10% to 15% of adult patients with newly diagnosed acute myeloid leukemia (AML).^1-4 ITD mutations have been shown to be associated with poor prognosis because of a high relapse rate, in particular in cases with a high mutant to wild-type allelic ratio (AR; ≥0.5),^5-9 an insertion site located in the beta-1 sheet of tyrosine kinase domain-1,^9-11 and in patients without concomitant mutations in NPM1.^7-9,12,13 

In the 2022 European LeukemiaNet (ELN) risk classification, AML with FLT3-ITD (without adverse-risk genetic lesions) is now categorized as intermediate risk, irrespective of the AR or concurrent presence of NPM1 mutation (NPM1^mut).¹⁴ This revision to the 2017 ELN classification was based on methodologic issues with standardizing the assays for measurement of the FLT3-ITD AR, the modifying impact of midostaurin-based therapy on FLT3-ITD,^11,13,15 and the increasing role of measurable residual disease (MRD) in treatment decisions.^14,16 MRD monitoring allows response assessment and the early detection of relapse and therefore can be used for treatment decision-making and early intervention. In addition, MRD data have contributed significantly to refining relapse risk. Moreover, MRD monitoring is a powerful tool to assess kinetics and the depth of response during therapy, which is particularly informative for the evaluation of treatment effects within clinical trials that are investigating novel therapies.^14,16,17 Based on these many clinical implications, MRD is currently being considered to serve as a surrogate end point in clinical trials, which may accelerate the approval of new drugs.

The currently most widely used methods for MRD assessment are multiparameter flow cytometry and quantitative polymerase chain reaction (qPCR).¹⁸ When compared with other molecular targets in AML, such as the recurrent gene fusions RUNX1::RUNX1T1, CBFB::MYH11, and PML::RARA, as well as NPM1^mut, FLT3-ITD MRD monitoring using qPCR has been hampered by the heterogeneity of the ITD mutation types that are determined by the broad variety of ITD lengths and insertion sites. Recent advances in next-generation sequencing (NGS) have been shown to overcome these limitations and now offers the opportunity for MRD monitoring in FLT3-ITD–positive (FLT3-ITD^pos) AML.^19-23 The 2021 update consensus document set forth by the ELN MRD Working Party acknowledged these efforts and included technical specifications for NGS-based MRD testing and integrative assessment of MRD, irrespective of technique.¹⁶ 

The objective of our study was to prospectively evaluate the prognostic impact of NGS-based MRD monitoring of FLT3-ITD in a cohort of 157 patients with FLT3-ITD^pos AML who received intensive chemotherapy in combination with midostaurin, followed by midostaurin maintenance within the German-Austrian Acute Myeloid Leukemia Study Group 16-10 (AMLSG16-10) treatment trial (ClinicalTrials.gov identifier: NCT01477606).^24,25 

All patients were enrolled in the AMLSG16-10 treatment trial.²⁵ Patients were selected based on the following criteria: (1) achievement of complete remission (CR) or CR with incomplete blood count recovery (CRi) after 2 cycles of intensive chemotherapy (Cy2) combined with midostaurin, (2) availability of a diagnostic bone marrow (BM) or peripheral blood (PB) sample, and (3) ≥1 subsequent sample (BM after Cy2 and/or at end of treatment [EOT] and/or BM or PB during 3 to 12 months follow-up [FU]). Following these criteria, 157 patients were evaluated at diagnosis, 142 after Cy2, 116 at EOT, and 148 during the defined FU period (supplemental Figure 1). All patients gave informed consent according to the Declaration of Helsinki. Approval was obtained from the institutional review boards of the participating AMLSG institutions.

At the time of diagnosis, mutation analysis for FLT3-ITD and NPM1, as well as NPM1^mut MRD assessment, was performed as previously described.^15,17,26 

The NGS libraries were paired-end sequenced on an Illumina NGS platform according to the manufacturer’s recommendation (Illumina, San Diego, CA) with minor modifications to our previously described assay that exhibited a variant allele frequency (VAF) sensitivity of 10⁻⁴ to 10⁻⁵.²⁰ Details on the experimental procedures are available in the supplemental Data. The raw sequencing data were analyzed using the bioinformatics program getITD.²⁰ 

CR/CRi, partial remission, overall survival (OS), relapse-free survival, and cumulative incidence of relapse (CIR) were defined according to standard criteria.¹⁴ Using landmark analyses, survival times for the time point after Cy2 were calculated from the date of first CR (CR1) or from the date of allogeneic hematopoietic-cell transplantation (HCT) in CR1 and were determined from the date the MRD sample was obtained for the EOT and FU time points. Patients who did not experience the event of interest at the end of FU were censored at the date of last contact. CIR was computed according to the method described by Gray.²⁷ The median FU for survival was calculated using the reverse Kaplan-Meier estimate.²⁸ Logistic regression and Cox proportional hazards models were used to identify prognostic variables for CR/CRi and OS.²⁹ CIR was analyzed using cause-specific Cox models in which death during CR was considered a competing event. Additional covariates in the multivariate analysis were sex, FLT3-ITD AR, and NPM1 mutation status as dichotomous variables and BM blast count, white blood cell (WBC) count (log₁₀ transformed), and age (10-year increase) as continuous variables; HCT during CR1 and MRD status over time were included as time-dependent variables. Missing values for covariates (WBC and BM blasts) were addressed by multiple imputation using chained equations. Mann-Whitney U tests were used to compare quantitative variables between patient subgroups; categorical variables were compared by means of Fisher exact tests. Associations between continuous variables were analyzed using the Spearman rank correlation coefficient. Survival distributions were estimated using the Kaplan-Meier method, and differences between groups were analyzed using 2-sided log-rank tests. An effect was considered significant if its P value was <5%. The analyses were not adjusted for multiple testing. All statistical analyses were performed using IBM SPSS Statistics 28.0.1.0, statistical software R (version 4.2.2), using the packages survival (version 3.5-0) and cmprsk (version 2.2-11), and/or GraphPad Prism7.

Table 1 summarizes the baseline characteristics of the 157 patients with FLT3-ITD^pos AML and of the 142 patients according to the FLT3-ITD MRD status after Cy2.

To determine tissue-dependency on VAF, we compared 29 paired PB and BM samples at diagnosis (n = 10) and after Cy2 (n = 19). Although the median FLT3-ITD VAF was slightly higher in BM than in PB (29.24% vs 23.59%; P = .065) at diagnosis, the median VAF (BM, 0.11% vs PB, 0.03%; P < .001) and MRD negativity (MRD^neg; BM, 0/19 [0%] vs PB, 7/19 [37%]; P = .008) differed significantly after Cy2 (supplemental Figure 2), clearly indicating the higher sensitivity in BM. Therefore, subsequently, only BM samples were selected for MRD assessment after Cy2 and at EOT. PB samples were analyzed only at diagnosis (n = 37) and during FU (n = 29).

A total of 465 ITDs were identified in 157 patients with FLT3-ITD^pos AML at diagnosis. Of the 465 ITDs, the median ITD length was 51 nucleotides (range, 9-285) and the median ITD VAF was 0.312% (0.006-92.26) (supplemental Figure 3). In total, 108 patients (69%) exhibited >1 ITD (median, 2; range, 1-16). The median total ITD VAF per patient (determined as sum of individual ITD VAFs) was 31.54% (0.46-92.26). Total ITD VAF per patient was correlated positively with higher WBC count (Rho, 0.287; P < .001), BM blast count (Rho, 0.196; P = .020), PB blast count (Rho, 0.264; P = .001), and lactate dehydrogenase level (Rho = 0.316; P < .001) and correlated inversely with the number of ITD clones (Rho, −0.226; P = .004). There was no correlation with age, sex, or NPM1 mutation status. The NGS-based calculated ITD AR (∑VAF/[100 − ∑VAF]) correlated positively with AR as determined by a GeneScan analysis (Rho, 0.855; P < .001; supplemental Figure 4).

At diagnosis, FLT3-ITD VAF a as log₂ transformed continuous variable did not have an impact on OS (hazard ratio [HR], 1.15; 95% confidence interval [CI], 0.92-1.42; P = .218) or CIR (HR, 0.95; 95% CI, 0.78-1.17; P = .646).

The median FU time of the 157 patients with FLT3-ITD^pos AML was 47.4 months (95% CI, 40.9-53.9); 59 of the 157 (37.6%) patients died. The median OS and 2-year OS rate were 68.9 months (95% CI, 51.5% to not applicable) and 0.72 (95% CI, 0.64-0.79), respectively. Allogeneic HCT during CR1 was performed in 122 of 157 (78%) patients and was performed at any time during the disease course in 135 of 157 (86%) patients.

All FLT3-ITDs identified after Cy2 were already detectable at diagnosis. In relation to diagnosis, the median log₁₀ reduction in the total FLT3-ITD VAF was 4.6 (range, 0.42-5.17; P < .001); FLT3-ITD MRD^neg, defined as undetectable FLT3-ITD, was achieved in 111 of 142 (78%) patients (Figure 1A). Patients with FLT3-ITD MRD^neg and MRD positivity (MRD^pos) differed significantly in terms of concurrent NPM1^mut and the ELN 2017 risk classification (Figure 1B; Table 1). The only favorable factor for achievement of FLT3-ITD MRD^neg after Cy2 was concurrent NPM1^mut (odds ratio [OR], 10.45; 95% CI, 3.40-32.07; P < .001); adverse factors were WBC count (OR for 10-fold increase, 0.31; 95% CI, 0.10-0.92; P = .035) and the administration of a second induction cycle (administered in 37 patients who achieved partial remission only after the first induction; OR, 0.19; 95% CI, 0.06-0.58; P = .004; supplemental Table 1).

A higher log₁₀ reduction in FLT3-ITD VAF was significantly associated with a lower CIR rate (HR for 10-fold better VAF reduction, 0.56; 95% CI, 0.43-0.71; P < .001) and improved OS (HR, 0.75; 95% CI, 0.60-0.93; P = .010; supplemental Table 2).

We next examined the prognostic impact of MRD^neg, which was achieved in 111 of 142 (78%) patients. In the univariate analysis, achieving MRD^neg was predictive of a superior 4-year CIR rate (4y-CIR; 26% vs 46% for MRD^pos; HR, 0.33; 95% CI, 0.17-0.64; P = .001) and 4-year OS (4y-OS; 70% vs 44%; HR, 0.47; 95% CI, 0.26-0.85; P = .012) (Figure 2A-B; supplemental Table 2). Figure 3 illustrates the course of events for every individual patient according to FLT3-ITD MRD status after Cy2. Of note, for patients with FLT3-ITD MRD^pos status, the risk for relapse correlated with the MRD burden, in particular ≥0.1%, the threshold provisionally used to define NGS-MRD test positivity according to the 2021 ELN MRD Working Party (Figure 4; supplemental Table 2). We additionally evaluated the impact of a ≥3 log₁₀ reduction in FLT3-ITD VAF (MR^3.0). Achieving MR^3.0 after Cy2 was associated with a lower CIR rate (4y-CIR, 30% vs 54%; HR, 0.31; 95% CI, 0.16-0.62; P < .001) but not with improved OS (4y-OS, 68% vs 45%; HR, 0.58; 95% CI, 0.30-1.10; P = .097) (supplemental Figure 5; supplemental Table 2).

Next, we performed a multivariate Cox regression analysis in different models that included the FLT3-ITD MRD log₁₀ reduction, an FLT3-ITD VAF <0.1%, achievement of MR^3.0, or FLT3-ITD MRD^neg status. In all models, a higher log₁₀ reduction, FLT3-ITD VAF <0.1%, achievement of MR^3.0, and MRD^neg were the only consistent favorable variables for risk of relapse and OS with the exception of MR^3.0 for OS. NPM1^mut and HCT during CR1 were favorable factors only for CIR (Table 2; supplemental Table 3).

Of the 142 patients assessed for FLT3-ITD MRD after Cy2, 107 patients underwent HCT during CR1 with 84 patients undergoing HCT immediately after Cy2 and 23 patients after receiving additional consolidation therapy with high-dose cytarabine. In total, 81 of 107 patients (76%) were FLT3-ITD MRD^neg before HCT; the median time from MRD assessment after Cy2 and HCT was 30 days (range, 7-88). FLT3-ITD MRD^pos before HCT was associated with an increased risk for relapse (4y-CIR, 39% vs 14%; HR, 4.90; 95% CI, 1.87-12.83; P = .001) and inferior outcome (4y-OS, 49% vs 71%; HR, 2.17; 95% CI, 1.06-4.43; P = .029) (Figure 2C-D). As shown in supplemental Figure 6, patients with FLT3-ITD MRD^pos status before HCT had a comparable outcome as those who were FLT3-ITD MRD^neg after Cy2 and were treated with conventional consolidation. Of the 26 patients with FLT3-ITD MRD^pos status before HCT, 21 patients were also eligible at EOT, and 15 (71%) patients achieved FLT3-ITD MRD^neg after HCT.

All FLT3-ITDs detected at EOT were already detectable at diagnosis. When compared with diagnosis, the median log₁₀ reduction in the total FLT3-ITD VAF was 4.7 (range, 1.55-5.19; P < .001); FLT3-ITD MRD^neg status was achieved in 109 of 116 (94%) patients (Figure 1A). In the Cox regression analysis, MRD^neg status at EOT was significantly associated with improved OS and trended toward a reduced risk for relapse, likely because of the high rate of patients with FLT3-ITD MRD^neg statusat EOT (Table 2).

To assess the risk for relapse after completion of intensive therapy, we analyzed MRD at least at 1 time point between 3 and 12 months after EOT during FU in 148 patients (BM, n = 119; PB, n = 29). Of these 148 patients, 117 (79%) had started maintenance with midostaurin with a median of 9 cycles (range, 1-12). Overall, 22 (15%) patients were FLT3-ITD MRD^pos during FU (all FLT3-ITDs were detected at diagnosis; Figure 3C). FLT3-ITD MRD persistence, defined as MRD^pos at all time points (range, 3-4) assessed, was detected in 6 patients and all relapsed; 16 patients (including 9 under maintenance) converted from FLT3-ITD MRD^neg to MRD^pos (median VAF, 0.058%; range, 0.006%-91.965%) within 5.78 months (median time from last MRD^neg to first MRD^pos sample; range, 2.07-10.0 months). Of the 16 patients, 13 (81%) (including 7 after HCT) relapsed within a median time of 7 days (measured from time point of MRD conversion in FU [MRD^conv_FU] to hematologic relapse; range, 0-197 days), translating into a significantly increased relapse risk (2y-CIR, 81% vs 16%; HR, 11.00; 95% CI, 5.44-22.28; P < .001) and inferior OS (2y-OS, 31% vs 80%; HR, 4.31; 95% CI, 2.22-8.38; P < .001) (supplemental Figure 7). Three of the 16 converted patients remained in continuous remission after allogeneic HCT, and all 3 exhibited low FLT3-ITD MRD levels (<0.008%); 2 of them received maintenance with midostaurin, and all 3 became MRD negative later on.

To analyze the impact of the FLT3-ITD MRD status over time, we performed 2 Cox regression models with MRD status as a time-dependent covariate. One model considered the impact of MRD^pos at any time point (after Cy2, at EOT, and/or during FU) and the second addressed MRD conversion (MRD^conv, MRD^neg to MRD^pos) after achievement of MRD^neg after Cy2. In both models, MRD^pos was the only consistent unfavorable variable for risk of relapse and OS (Table 3).

In total, 111 of the 157 (71%) patients had concomitant NPM1^mut. Concurrent NPM1^mut favorably impacted the log₁₀ VAF reduction (median, 4.7 vs 3.7 for NPM1^wt; P < .001) and the achievement of FLT3-ITD MRD^neg status (88% vs 53%; P < .001) after Cy2 (Figure 1B). This translated into a lower CIR rate (4y-CIR FLT3-ITD MRD^neg/NPM1^mut vs FLT3-ITD MRD^neg/NPM1^WT, 19% vs 68%; HR, 0.23; 95% CI, 0.10-0.50; P < .001) and a trend toward improved OS (4y-OS, 72% vs 59%; HR, 0.52; 95% CI, 0.25-1.09; P = .071; supplemental Figure 8). No additional benefit was observed for NPM1^mut at EOT, but NPM1^mut was associated with FLT3-ITD MRD^neg during FU (91% vs 70%; P = .002; Figure 1B).

According to the ELN MRD Working Party, in NPM1^mut AML, MRD should be assessed, preferentially in PB, after Cy2 and in BM at EOT and during FU. In line with this recommendation, we correlated NGS-based FLT3-ITD MRD assessed in BM with qPCR-based NPM1^mut MRD assessed in PB after Cy2 (Figure 5A). Of the 82 eligible patients, 41 (50%) were FLT3-ITD MRD^neg/NPM1^mut MRD^neg, 33 (40%) were FLT3-ITD MRD^neg/NPM1^mut MRD^pos, 7 (9%) were FLT3-ITD MRD^pos/NPM1^mut MRD^pos, and 1 (1%) was FLT3-ITD MRD^pos/NPM1^mut MRD^neg. It should be noted that an FLT3-ITD MRD^neg status was associated with a lower relapse risk and improved outcome irrespective of NPM1^mut MRD status (Figure 5B-C). Because of the small number of events in this subgroup (relapses, n = 13; deaths, n = 14) Cox regression analysis on relapse-free survival was performed and confirmed the beneficial impact of FLT3-ITD MRD^neg status regardless of the NPM1^mut MRD status (supplemental Table 4).

At EOT and during FU, an FLT3-ITD MRD^neg status was more frequent than an NPM1^mut MRD^neg status. In addition, all 3 patients with FLT3-ITD MRD^pos statuswere also NPM1^mut MRD^pos at EOT. Furthermore, of the 30 patients with NPM1^mut MRD^pos, 26 exhibited NPM1^mut MRD at low level provisionally defined as <2%. Similar findings were also observed during FU. The proportion of FLT3-ITD MRD^neg/NPM1^mut MRD^neg increased from 50% after Cy2 to 68% during FU (Figure 5).

In this study, we performed FLT3-ITD MRD monitoring in 157 adult patients with FLT3-ITD^pos AML using a highly sensitive NGS-based assay as previously reported by us.²⁰ The data revealed that FLT3-ITD MRD was a highly significant risk factor for relapse and OS and even outperformed known risk factors, such as NPM1 mutational status and FLT3-ITD AR.

The particular strengths of our study are twofold. First, sequential biosampling was done prospectively within a controlled clinical trial at defined time points, that is, after Cy2, at the EOT, and during FU. Second, all patients received intensive chemotherapy in combination with the FLT3 inhibitor midostaurin, which has improved outcomes and is now considered standard of care in patients with FLT3-ITD^pos AML.^15,25 Recently, addition of the second-generation FLT3 inhibitor quizartinib to intensive chemotherapy has also been approved for the treatment of patients with newly diagnosed FLT3-ITD^pos AML.

Three recently published studies reported on the clinically relevant impact of FLT3-ITD MRD monitoring using NGS-based assays.^21-23 In contrast with our study, these analyses were restricted to a single time point (after 1 or 2 induction cycles) and were performed in heterogeneous patient cohorts, particularly with respect to treatment with an additional FLT3 inhibitor.^21-23 Furthermore, 2 of these studies evaluated the impact of FLT3-ITD MRD specifically before HCT; the Pre-MEASURE study analyzed the PB of 608 patients with FLT3-ITD^pos AML before HCT during CR1²³ and the study by Loo et al and Dillon et al analyzed samples from 104 patients with FLT3-ITD^pos AML, irrespective of CR1, CR2, or molecular relapse.²² In the HOVON study, 93 of 161 (58%) intensively treated patients underwent HCT during CR1.²¹ Regarding concurrent NPM1^mut, the prevalence of 75% in the study by Loo et al and Dillon et al was comparable with ours (71%) and lower than that in the HOVON (57%) and Pre-MEASURE studies (58%).

After 2 cycles of intensive chemotherapy is the first time point at which MRD assessment in the BM is considered to be clinically relevant.¹⁶ In our study, a higher reduction in FLT3-ITD VAF and achieving FLT3-ITD MRD^neg after Cy2 in the BM were statistically significant favorable prognostic factors in terms of both relapse risk and OS and therefore enable a refined risk assessment of patients in hematologic response. A concurrent NPM1 mutation and HCT during CR1 were additional significant favorable factors for CIR, whereas FLT3-ITD MRD^neg was the only significant factor for OS. Concurrent NPM1^mut correlated with a deeper molecular response, as reflected in a better FLT3-ITD VAF reduction and a significantly higher rate of FLT3-ITD MRD^neg after Cy2.

Our observation that FLT3-ITD MRD^neg status after Cy2 is of high prognostic relevance is in line with the data recently published by the HOVON group.²¹ In their study, achievement of MRD^neg after 2 induction cycles identified patients with a significantly lower 4y-CIR rate (33% vs 75%; P < .001) and improved OS (4y-OS, 57% vs 31%; P < .001). Similar results were reported for the QuANTUM-first study that showed that patients who achieved CR/composite CR after Cy2 and MRD levels <10⁻⁴ had a significantly improved OS when compared with patients above that MRD cutoff.³⁰ The current ELN Working Party on MRD proposed a cutoff of 0.1% for NGS-based MRD assessment but also stated that this cutoff is provisional and not based on robust data. As shown in the 3 published studies and this study, the evaluation of different cutoffs is limited by the small patient numbers that underline the current uncertainty regarding thresholds and highlight the need for future studies (eg, meta-analysis) to define this.^16,21-23 For MRD-based prognostication and prediction and for comparability between different study populations and techniques, the value of cutoffs beyond MRD^neg, such as the log reduction in the transcript level between diagnosis and after induction, are under evaluation in clinical trials, as are other well-established MRD targets.¹⁶ 

In FLT3-ITD^pos AML, concurrent NPM1^mut has been shown to be a favorable prognostic factor for all survival endpoints.^7,8,12,13,22 Consistent with the HOVON data, concurrent NPM1^mut was associated with a significantly higher percentage of FLT3-ITD MRD^neg after Cy2. Although, in the HOVON study, identical relapse rates were observed for the entire FLT3-ITD MRD^neg cohort and the FLT3-ITD MRD^neg/NPM1^mut cohort (4y-CIR 33% each), we found a lower CIR rate for FLT3-ITD MRD^neg/NPM1^mut patients than for the entire cohort of patients with FLT3-ITD MRD^neg (4y-CIR, 19% vs 29%), underlining the favorable effect of a concurrent NPM1^mut in terms of achieving a deeper molecular response.

Allogeneic HCT has been shown to improve the outcome of patients with FLT3-ITD^pos AML.^8,9,13,14 In our study, HCT during CR1 was an independent favorable factor for CIR at all time points, demonstrating that HCT during CR1 is an important pillar for treatment of FLT3-ITD^pos AML. Of note, FLT3-ITD MRD status before HCT provided an additional prognostic impact. An FLT3-ITD MRD^neg status before HCT was associated with a lower risk for relapse and superior outcome (Figure 2). In line with the recently published studies that demonstrated the prognostic impact of FLT3-ITD MRD^neg status before HCT and considering previous studies on various molecular MRD markers, these data underline the prognostic value of molecular remission before HCT.^21-23,31-34 Moreover, recent results from the MORPHO study demonstrated that FLT3-ITD MRD^neg before and after HCT was associated with improved outcomes and only patients who were MRD positive significantly benefited from maintenance treatment with gilteritinib.³⁵ 

The proportion of patients who achieved FLT3-ITD MRD^neg status after Cy2/before HCT varied slightly across the published cohorts (63%,²² 71%,²¹ and 86%²³) and our cohorts (78%). Differences might be explained by patient characteristics and the tissues assessed, the various prevalence rates of concurrent NPM1^mut, and/or the additional treatment with midostaurin in our cohort.

To our knowledge, this was the first study to evaluate the impact of FLT3-ITD MRD status over time and MRD conversion during early FU. In patients with FLT3-ITD MRD conversion (MRD^neg to MRD^pos) during FU, we observed subsequent relapses in 13 of 16 (81%) patients, which translated into a significantly increased relapse risk. However, the FLT3-ITD MRD status was not systematically evaluated during the FU period, and FLT3-ITD MRD monitoring for relapse surveillance needs to be confirmed in additional studies. The 3 patients who converted to MRD^pos during FU but who remained in continuous remission exhibited very low VAF levels. This reflects the clinical challenge to interpret MRD^pos status during FU and underscores the indispensability to discriminate molecular persistence at (very) low VAF level from molecular progression and molecular relapse by analyzing a second sample as recommended by the ELN MRD Working Party.¹⁶ 

In line with the 2 previous studies, combined FLT3-ITD and NPM1^mut MRD positivity was associated with a poor outcome.^21,22 Favorable outcomes were observed for patients with FLT3-ITD MRD^neg status irrespective of the NPM1^mut MRD status. This implies that within FLT3-ITD^pos/NPM1^mut AML, the FLT3-ITD MRD status may further refine the individual prognosis by better discrimination of patients at high and low risk for relapse than when using NPM1^mut MRD detection alone.

Considering the different assays used for FLT3-ITD MRD monitoring, an international consensus on standardizing the approach is needed, which is also currently discussed in the ELN MRD Working Party.

In summary, beyond the known risk factors, NGS-based FLT3-ITD MRD monitoring allows for the identification of patients at high risk for relapse and death. Currently, randomized studies that are comparing midostaurin and second-generation FLT3 inhibitors are ongoing (eg, NCT04027309). FLT3-ITD MRD analysis will be informative in assessing whether these more selective inhibitors will increase the depth of molecular remission and if these deeper responses are associated with an improved outcome.

The authors acknowledge Julia K. Herzig, Susanne Lux, and Laura K. Schmalbrock for technical support during next-generation sequencing data analysis and the members of the German-Austrian AML Study Group for providing patient samples and clinical information.

This work was supported, in part, by the Collaborative Research Center (SFB 1074), projects B3 (L.B. and K.D.), B12 (K.D.), and Z1 (H.D.).

Contribution: F.G.R., L.B., H.D., and K.D. designed the study; S.C., S.S., T.J.L., A.C., V.I.G., A.M., S.A., and F. Theis performed the experiments and validated the data; F.G.R., J.K., E.P., and A.B. performed the statistical analyses; F.G.R., D.W., F.S., A.S., W.F., H.R.S., G.W., H.S., T.S., K.S.G., M.W.M.K., M.L., R.F.S., F. Thol, M.H., A.G., H.D., and K.D. collected, assembled, analyzed and interpreted data; F.G.R., H.D., and K.D. wrote the first draft of the manuscript; and all authors undertook manuscript writing, editing and approval, revised the manuscript, and reviewed and approved the final version.

Conflict-of-interest disclosure: F.G.R. reports receiving honoraria from and serving as a consultant for Jazz Pharmaceuticals, Novartis, and Bristol Myers Squibb (BMS)/Celgene, and receiving travel support from Jazz Pharmaceuticals. L.B. reports receiving honoraria from AbbVie, Amgen, Astellas, BMS/Celgene, Daiichi Sankyo, Gilead, Janssen, Jazz Pharmaceuticals, Menarini, Novartis, Pfizer, Roche, and Sanofi, and receiving research support from Bayer and Jazz Pharmaceuticals. V.I.G. reports serving in an advisory role for Jazz Pharmaceuticals, AbbVie, and Boehringer Ingelheim; serving on the speakers' bureau of Pfizer, Janssen, and AbbVie; and receiving travel support from AbbVie. F.S. reports receiving honoraria from and serving as a consultant for AOP Orphan Pharmaceuticals, MorphoSys, BMS/Celgene, Incyte, Novartis, and Pfizer. W.F. reports receiving personal fees and nonfinancial support from AbbVie; receiving grants, personal fees, and nonfinancial support from Amgen and Pfizer; receiving personal fees from Jazz Pharmaceuticals, Celgene, MorphoSys, Incyte, Stemline Therapeutics, Clinigen, Daiichi Sankyo, Otsuka, and Servier outside the submitted work; receiving research support from Apis; filing a patent with Amgen; and receiving support for medical writing for Amgen, Pfizer, and AbbVie. H.S. reports receiving honoraria from AbbVie, Amgen, AstraZeneca, BMS/Celgene, Genzyme, GlaxoSmithKline, Janssen, Oncopeptides, Pfizer, Roche, Sanofi, Stemline Therapeutics, and Takeda, and travel expenses from Amgen, BMS/Celgene, Janssen, and Sanofi. K.S.G. reports serving in an advisory role for BMS, Jazz Pharmaceuticals, Pfizer, and AbbVie. M.W.M.K. reports receiving honoraria from and serving as a consultant for Pfizer, Kura Oncology, Jazz Pharmaceuticals, BMS/Celgene, and AbbVie, and serving on the speakers bureau of Gilead. M.L. reports serving in an advisory role for AbbVie, Astex Pharmaceuticals, Imago BioSciences, Janssen, Otsuka, and Syros, and receiving research support from Janssen and Cheplapharm. R.F.S. reports serving in an advisory role or as a consultant for Daiichi Sankyo, Pfizer, Astellas, and Novartis; receiving research funding from PharmaMar, AstraZeneca, Pfizer, Roche, Boehringer Ingelheim, Daiichi Sankyo, and Recordati; and receiving funding for travel, accommodation, and expenses from Daiichi Sankyo. F. Thol reports serving in an advisory role for Novartis, BMS, AbbVie, Menarini, and Rigel. M.H. reports receiving honoraria from Certara, Jazz Pharmaceuticals, Janssen, Novartis, and Sobi; serving as a paid consultant for AbbVie, Amgen, BMS/Celgene, Glycostem, Delbert Lab, Pfizer, Pinotbio, and Servier; and receiving research funding to his institution from AbbVie, Agios, Astellas, BMS/Celgene, Glycostem, Jazz Pharmaceuticals, Karyopharm, Loxo Oncology, and Pinotbio. H.D. reports serving in an advisory role for AbbVie, Agios, Amgen, Astellas, AstraZeneca, Berlin-Chemie, BMS/Celgene, Daiichi Sankyo, GEMoaB, Gilead, Janssen, Jazz Pharmaceuticals, Novartis, Servier, Stemline Therapeutics, and Syndax, and receiving research funding from AbbVie, Agios, Amgen, Astellas, BMS/Celgene, Jazz Pharmaceuticals, Kronos Bio, Novartis, and Pfizer. K.D. reports serving in an advisory role for Amgen, BMS/Celgene, Daiichi Sankyo, Janssen, Jazz Pharmaceuticals, Novartis, and Roche, and receiving research funding from Agios, Astex, Astellas, BMS/Celgene, and Novartis. The remaining authors declare no competing financial interests.

Correspondence: Konstanze Döhner, Department of Internal Medicine III, University Hospital of Ulm, Albert-Einstein-Allee 23, D-89081 Ulm, Germany; email: konstanze.doehner@uniklinik-ulm.de.