Understanding differential technologies for detection of MRD and how to incorporate into clinical practice

A female patient of 65 years of age was referred to our hospital. Bone marrow (BM) aspiration showed a high proportion of myeloblasts by morphology and multiparameter flow cytometry (MFC), which led to the diagnosis of acute myeloid leukemia (AML). The disease was classified as de novo AML, not otherwise specified.¹  The patient was stratified as European Leukemia Network (ELN) 2022 intermediate risk²  based on normal karyotype, an FLT3/ITD mutation (at low allelic ratio, although ELN 2022 does not require allelic ratio), and a DNMT3A mutation. At diagnosis, the leukemia was characterized by MFC and revealed a leukemia-associated phenotype (LAIP) with the combination of cluster of differentiation markers (CD) of CD34⁺CD13⁺CD7⁺.

The patient's age, medical history, and overall health indicated her eligibility for intensive 7 + 3 chemotherapy (cytarabine continuously for 7 days, along with short infusions of an anthracycline on each of the first 3 days). She was enrolled in a randomized trial comparing standard 7 + 3 chemotherapy with midostaurin to 7 + 3 chemotherapy including a different (more targeted) tyrosine kinase inhibitor. After the first induction cycle, the patient reached morphologic complete remission (CR; <5% blasts in BM by microscopy). MFC–measurable residual disease (MRD) showed 0.15% LAIP cells of the total white blood cells. Considering the applied MFC-MRD assay, this was deemed MRD positive (≥0.1%). Retrospective analysis using next-generation sequencing (NGS)–MRD confirmed MRD positivity for FLT3/ITD at 0.36%. A donor search was initiated, and she received the second chemotherapy cycle. After this second cycle, MFC-MRD showed MRD negativity of 0.04% (MRD negative), while the post hoc analysis identified the patient MRD positive for FLT3/ITD with a variant allele frequency (VAF) of 0.09%. The patient received myeloablative conditioning and peripheral blood (PB) stem cells from a matched unrelated donor for allogeneic stem cell transplantation (alloSCT), and the patient achieved CR_MRD negative by both MRD techniques and resumed the FLT3 inhibitor after blood count recovery.

Disease monitoring was performed at 1 and 3 months, at which time MFC and NGS identified MRD relapse by MRD conversion from negative to positive. In the BM, the MFC-MRD was 0.2% of a new LAIP (CD34⁺CD13⁺CD56⁺), while the original LAIP was absent, and FLT3/ITD was detectable with a VAF of 1.6%. Another BM after 2 weeks confirmed MRD relapse (MFC 0.3%, NGS 2.6%). This coincided with a mixed donor/patient T-cell chimerism detected in PB.³  The immunosuppression was decreased and donor lymphocyte infusions were started.^4,5  This resulted in a decrease in the LAIP cells, and MRD was cleared entirely after 3 months. After 2 years, the MRD monitoring was stopped, and the patient was followed up by clinical visits and standard laboratory evaluation.

The ongoing trial the patient was enrolled in is designed to evaluate MRD as a secondary end point for surrogacy of the clinical end-point event-free survival (EFS). In case the primary end-point EFS is improved by the new tyrosine kinase inhibitor and MRD is indeed significantly lower after a defined period of treatment, it is ideally suited to establish surrogacy of MRD for faster and more efficient evaluation of new treatment strategies.

In this review, we discuss the pros and cons of different MRD technologies, MRD results, and their clinical implications based on the available data.

AML is a heterogeneous disease in both biology and clinical outcome. Main reasons for poor outcome are refractory disease (no CR) or relapse. To tailor therapy, estimating the risk of poor response to treatment is now imperative. Therefore, after the initial assessment of AML, the diagnostic workup needs to be more detailed to determine the prognostic risk group. This is mainly based on cytogenetic and molecular aberrations, leading to 3 ELN risk groups (favorable, intermediate, and adverse).²  First-line treatment of patients with AML aims at reaching CR by eradicating the majority of leukemia cells. This commonly consists of 1 or 2 cycles of intensive induction chemotherapy. For FLT3/ITD-mutated patients, addition of the multitargeted kinase inhibitor midostaurin significantly increased EFS and overall survival (OS)⁶  and is now standard of care. The induction phase is then followed by consolidation treatment to extend the response, and the intensity is commonly selected based on risk group. First-line treatment options with increasing intensity are (1) up to 3 cycles of chemotherapy (commonly intermediate- dose AraC), (2) myeloablative chemotherapy rescued by autologous stem cell transplantation, and (3) allogenic stem cell transplantation (alloSCT). Increasing treatment intensity will improve antileukemia efficacy but comes with adverse side effects and increased nonrelapse mortality (Figure 1). According to ELN, alloSCT is the preferred consolidation treatment for patients with an estimated relapse risk exceeding 35% to 40%.^2,7  It needs to be emphasized that intensity of treatment is not solely based on laboratory testing but on patient-related factors as well, such as older age or comorbidities.

The ELN-AML guideline²  recommends refinement of morphologic response assessment by more sensitive MRD assessment, and the ELN-MRD guidelines⁸  provide the tools for standardized MRD evaluation (Table 1). Flow cytometry is commonly used for AML diagnosis, and hence the technique is widely available. MFC-MRD testing still needs expertise and strict criteria due to the rather extensive and subjective gating strategies. In addition, MFC may not be sensitive enough (10⁻³-10⁻⁴) to detect minimal numbers of residual cells. Molecular techniques such as quantitative polymerase chain reaction (PCR) or digital PCR are more sensitive (10⁻⁶) but only suitable for patients with specific molecular aberrations (NPM1 mutation or core binding factor [CBF] translocations). This may be resolved when several mutations can be followed simultaneously using NGS.⁹ FLT3/ITD has been a difficult target to monitor MRD due to the variable length of the insertions and the duplicity of sequences.¹⁰  However, NGS techniques have been applied¹¹  and the getITD analysis algorithm is increasingly used for MRD monitoring.¹² 

For accurate risk classification,²  the leukemic cells are molecularly characterized by cytogenetic analysis and DNA sequencing. When BM is available (or, if not, PB), MFC-MRD can be performed to characterize the patient-specific LAIP at diagnosis, thereby allowing to follow the LAIP in the MRD setting. A LAIP is defined as CD marker expression combinations present on >10% of blasts that are not found in BM from healthy individuals. A LAIP can be identified in about 90% of patients.¹³  Measurement of MFC-MRD is recommended when patients do not harbor NPM1 mutations or CBF translocations. Otherwise, patients should be monitored by quantitative PCR or digital PCR for NPM1 mutations or CBF translocations.⁸  Based on the recent data on FLT3/ITD monitoring by NGS, we expect a rapid uptake of this approach to refine or replace MFC-MRD monitoring as in our patient.^14,15  Furthermore, mutations such as DNMT3A, TET2, and ASXL1 (DTA) mutations may reflect clonal hematopoiesis of indeterminate potential and can be often present with a relatively high VAF in nonmalignant clones and therefore cannot be used for MRD assessment.^14,16 

For accurate detection of MFC-MRD after chemotherapy, BM material is required and patients should be in morphologic CR. For the MFC-MRD measurement, the LAIP approach would require only measuring the flow cytometry panel tube that contains the LAIP found at diagnosis. However, during therapy, small cell populations that were undetectable at diagnosis may be selected and proliferate, causing an immune phenotypic shift that the LAIP approach can miss. Another MFC-MRD approach can detect these emerging cell populations and is referred to as different-from-normal, which can also be used when no diagnostic sample is available and is essential to recognize clonal evolution.¹⁷  ELN recommendations advocate the combined use by measuring all antibody panels at follow-up to assess MRD as LAIP-based different-from-normal,⁸  allowing to monitor the dominant LAIP cell population and the possible emergence of a new LAIP.

Often, the search for a donor for alloSCT starts at diagnosis, while some argue the time point after the first cycle of chemotherapy would suffice, since very few patients who are MRD negative become MRD positive after the second cycle and the effort of searching a donor could be spared for nonresponding and MRD-positive patients.¹⁸  For adverse risk patients, it is debated whether patients with MRD-positive CR should receive a second cycle of chemo or immediately undergo alloSCT.¹⁹ 

MFC-MRD with a cutoff of 0.1% after induction therapy is now established as important prognostic factor for outcome.⁸  The clinical implementation of MRD to guide further postremission strategies varies widely among treatment centers. Recently, several studies indicated that alloSCT could safely be omitted^19-21  for many intermediate-risk patients, and those who relapsed could be salvaged effectively with alloSCT.²²  The data of Zhang and collaegues¹⁹  suggest that alloSCT in favorable- and intermediate-risk patients may only benefit MRD-positive patients. Published data of the 3 trials were highly comparable, as shown in Figure 2. To gather further evidence that MRD-negative intermediate-risk patients can be safely treated with non-alloSCT treatment, a randomized trial would be required showing noninferiority of the non-alloSCT treatment approach compared to alloSCT.

Although some studies recently showed its potential prognostic relevance,^15,23 FLT3/ITD MRD is not as well clinically validated as NPM1.^24,25  Current data show that FLT3/ITD is a reliable MRD marker before alloSCT at a ≥0.01% VAF cutoff (Table 1).^15,23,26  This also applied to patients treated with an FLT3 inhibitor, as shown in the Quantum first trial, which confirmed the 0.01% cutoff as an ideal cutpoint.²⁷  Residual FLT3/ITD MRD should be considered an indication for alloSCT, while it is unclear how the unfavorable prognostic effect of MRD before alloSCT may be mitigated.

Should every FLT3/ITD-positive patient undergo alloSCT? While this is common practice in the United States,^28,29  there are data that question this approach. A post hoc analysis of the Ratify trial identified 318 patients with available data comparing allogeneic hematopoietic cell transplantation and non–allogeneic hematopoietic cell transplantation treatment approaches.³⁰  Patients were grouped according to ELN 2017 risk groups based on NPM1 mutation status and FLT3/ITD allelic ratio. Patients with favorable risk (NPM1 mutated and FLT3/ITD low allelic ratio) had long-term survival at 75% when treated with midostaurin independent of alloSCT. Although the survival of the intermediate-risk group was worse with approximately 50% long-term survival, there was no OS difference with or without alloSCT in the midostaurin group. Only the ELN adverse risk group had a significant benefit from alloSCT, especially in the midostaurin-treated group. The ELN 2022 classification has abandoned the FLT3/ITD allelic ratio and groups all FLT3/ITD-positive patients in the intermediate-risk group. The recent studies evaluating FLT3/ITD NGS-MRD show a long-term survival for MRD-negative patients of 50% to 75%.^15,23,26  It is thus conceivable that the mutation profile at diagnosis (eg, NPM1 mutated, FLT3/ITD low allelic ratio), the FLT3/ITD MRD status after 2 cycles of chemotherapy, or a combination of both can identify FLT3/ITD-mutated patients who will not benefit from alloSCT. Recently, the results of the BMT-CTN 1506 (MORPHO) study were presented, which evaluated giltertinib maintenance after alloSCT in FLT3/ITD-mutated patients.³¹  Gilteritinib treatment was associated with an improved relapse-free survival (hazard ratio, 0.515; 95% CI, 0.316-0.838) for the 50.5% of patients with detectable MRD pre- or post-alloSCT, compared to those without detectable MRD. The effect of gilteritinib on OS in patients with detectable MRD has not been reported yet. More research is needed to assess the clinical application of MRD in this subgroup of patients.

The current evidence for accurate MRD monitoring at follow-up is only strong enough for NPM1 and CBF molecular aberrations.⁸  To further establish the role of MFC-MRD for disease monitoring, this is often still measured every 3 months for the first 2 years after consolidation or alloSCT, or when there are clinical indications of disease recurrence. Collecting these data in clinical trials is warranted as MFC-MRD monitoring after consolidation is not yet common clinical practice but may be beneficial for early intervention options and early relapse detection (Table 2). There are several considerations for using MRD for monitoring. First, the type of patient material matters. For MFC-MRD, PB is not well investigated, and MFC-MRD studies suggested that MRD in PB has higher specificity but 10-fold less sensitivity than BM.^32,33  Molecular MRD can be very sensitive, so for this technique, PB is considered suitable. Second, the threshold for MFC-MRD positivity at follow-up and subsequent potential clinical intervention still needs to be established. The cutoff of 0.1% may clearly indicate a high risk for relapse, but lower levels may also already indicate recurrent disease. It is also suggested that kinetics, measured as a certain level of increase in MRD, point to disease recurrence. For molecular techniques, a 1-log increase is considered a useful parameter,⁸  but for MFC-MRD, sufficient data are lacking. Third, the duration of monitoring needs to be clarified. Previous reports suggest that most relapses occur within 2 years. PB-MRD can be analyzed more frequently (every 4-6 weeks) than BM (every 3 months). Because some relapses occur very fast when a small clone has gained a growth advantage, MRD monitoring would preferably be done in PB. Our patient case showed early positivity after alloSCT, which has been described in 7.5% of patients monitored by MFC-MRD shortly after alloSCT.³⁴ 

MRD relapse is defined by the ELN recommendations independent of the technique that is used: “MRD relapse is defined as either (1) conversion of MRD negativity to MRD positivity independent of the MRD technique or (2) increase of MRD ≥1 log10 between any 2 positive samples measured in the same tissue (PB or BM) in patients with low-level MRD. Conversion from negative to positive MRD in PB or BM should be confirmed within 4 weeks, in a second consecutive sample, preferably with a BM sample.”⁸  After MRD relapse, alloSCT can be beneficial for curation.²²  The most suitable technique, best time points, duration of MRD monitoring after alloSCT, and for which subsets of patients are still under investigation. An early indication that NGS-MRD is also prognostic after relapse was recently published for patients treated with fludarabine, high dose cytarabine (Ara-C), idarubicin and granulocyte-colony stimulating factor (G-CSF) and venetoclax, where MRD-negative patients had an excellent survival.³⁵ 

Since many new drugs are currently available, MRD as intermediate end point would be worthwhile to fasten the process of drug approval.³⁶  The association between MRD and prognostic value has been very well established for intensively treated patients.³⁷  However, limited positive trials have included the measurement of MRD to show that the treatment effects are associated with a corresponding decline in MRD or the number of MRD-negative patients.³⁸  Thus, for the regulatory agencies to accept MRD as intermediate end point, additional data are required.³⁹ 

MRD monitoring in AML is quickly evolving from risk stratification to clinical decision-making to guide therapy, monitor the disease, detect MRD relapse early, and improve the drug development process by using MRD as an intermediate end point for clinical trials. For any MRD-including trial design, it is instrumental to have the exact information on when and what clinical decisions are made on MRD. Some clinical aspects may change the treatment decision, and the prognostic value of MRD is lost when treatment is guided based on MRD.²¹  In addition, it still needs to be assessed whether quality of life improves with omitting or delaying alloSCT, since these data are currently lacking.

With the new technologies and insights that are currently under investigation, we may improve patient management in the future. For MFC-MRD, one of the advancements would be detecting the leukemia stem cell (LSC) load, measured as a percentage of CD34⁺/CD38^⁻/LSCmarker⁺ cells per CD45-expressing cells.⁴⁰  Several studies have shown that a high LSC load was associated with a relatively poor outcome at diagnosis and after induction.^41-44  Although promising, more evidence is needed to incorporate LSC load into treatment decisions.

The NGS-based MRD detection is promising but requires standardization. The clinical utility of NGS has been reported for FLT3/ITD and NPM1 mutations. To move this field further, consensus should be accomplished in various aspects such as selection of the most relevant molecular markers, sequencing approaches, sampling tissue (BM or PB), cutoffs, and timing of sampling.

In conclusion, MRD is an important prognostic factor to inform consolidation treatment after CR in intermediate-risk patients. Based on recent MRD data from clinical trials and real-world clinical practice, more evidence is being gathered for the broader use of MRD for the management of AML (Table 2). Still, clearly more data are necessary to determine how MRD (cutoffs, kinetics, patient material, risk group, etc) can best be implemented in clinical practice. The relevance of MRD using current MRD techniques also needs additional studies when novel (targeted) or nonintensive treatments are used.^45-47  Improved standardization of MRD measurements should enable the combination of data from different trials and is essential in the light of the in vitro diagnostics regulation. When harmonized MRD data become available, additional MRD applications may be implemented in newly updated guidelines in the coming years.

The authors thank Prof. G. J. Ossenkoppele for critically reading the manuscript.

Jacqueline Cloos serves an advisory role for Novartis; has received research grants for her institution from Novartis, Merus, Takeda, Genentech, and BD Biosciences; and has received a royalty/license from Navigate and BD Biosciences.

Lok Lam Ngai: no competing financial interests to declare.

Michael Heuser serves an advisory role for Abbvie, BMS, Glycostem, Servier, PinotBio, Amgen, Pfizer, and LabDelbert; has received honoraria from Certara, Jazz Pharmaceuticals, Janssen, Novartis, Pfizer, and Sobi; and has received research funding to his institution from Abbive, Agios, Astellas, BergenBio, BMS, Glycostem, Jazz Pharmaceuticals, Karyopharm, Loxo Oncology, and PinotBio.

Jacqueline Cloos: nothing to disclose.

Lok Lam Ngai: nothing to disclose.

Michael Heuser: Use of venetoclax in relapsed/refractory AML patients and in combination with FLAG-IDA is neither approved by FDA nor EMA. Giltertinib maintenance after alloSCT in FLT3/ ITD-mutated patients is neither approved by FDA nor EMA.