Minimal residual disease in acute myelogenous leukemia

Abstract

Treatment of acute myelogenous leukemia (AML) over the past four decades remains mostly unchanged and the prognosis for the majority of patients remains poor. Most of the significant advances that have been observed are in defining cytogenetic abnormalities, as well as the genetic and epigenetic profiles of AML patients. While new cytogenetic and genetic aberrations such as the FLT3-ITD and NPM1 mutations are able to guide prognosis for the majority of patients with AML, outcomes are still dismal and relapse rates remain high. It is thought that relapse in AML is in part driven by minimal residual disease (MRD) that remains in the patient following treatment. Thus, there is a need for sensitive and objective methodology for MRD detection. Methodologies such as multiparameter flow cytometry (MFC), quantitative real-time polymerase chain reaction (RQ-PCR), digital PCR (dPCR), or next-generation sequencing (NGS) are being employed to evaluate their utility in MRD assessment. In this review, we will provide an overview of AML and the clinical utility of MRD measurement. We will discuss optimal timing to MRD measurement, the different approaches that are available, and efforts in the standardization across laboratories.

1 | INTRODUCTION

Acute myelogenous leukemia (AML) is characterized by infiltration of the bone marrow (BM), blood and other tissues by abnormally differentiated myeloid cells.¹ It is estimated that approximately 4.1 in 100 000 adults are diagnosed with AML every year with an average age at diagnosis of 67 years.² Remission is achieved in 35%–40% of adult patients who are 60 years or younger but only in 5%–15% of patients who are older than 60 years of age.¹ As the population ages, the frequency of AML is expected to increase. In addition to age, other contributing factors to development of AML include treatment for prior malignancies with chemotherapy or radiation, antecedent hematologic disease (ie, myelodysplastic or myeloproliferative syndromes) and benzene exposure. Nevertheless, in a majority of individuals the causative factor for development of leukemia is unknown.

The overall 5-year survival for adults with AML in the United States is a dismal 26.6%, and in older patients unable to tolerate or receive intensive chemotherapy, the median survival is only 5–10 months. However, that number can vary based on patient characteristics including complex cytogenetics, age, comorbidities, and performance status.^1,2 Advances in high-throughput sequencing have revealed a multitude of recurring genetic abnormalities in AML including FLT3 internal tandem duplication (FLT3-ITD), FLT3 tyrosine kinase domain (FLT3-TKD), NPM1 insertions, TP53, DNMT3A, IDH1, IDH2, CEBPA, KIT, RUNX1, NRAS, KRAS, and MLL partial tandem duplication (MLL-PTD).³ Of the above-mentioned mutations, NPM1, CEBPA, FLT3-ITD, DNMT3A, IDH1, or IDH2 is associated with a cytogenetically normal karyotype (CN-AML).⁴ The association of molecular lesions with outcome has informed risk stratification and is the basis for the World Health Organization (WHO) Classification of Tumors of Hematopoietic and Lymphoid Tissues. As per the WHO 2016 Guidelines, AML is classified into four major categories: (i) AML with recurrent genetic abnormalities, which includes the following entities “AML with NPM1 Mutation,” “AML with CEBPA mutation,” “AML with RUNX1 mutation,” and “AML with BCR-ABL1”; (ii) AML with myelodysplasia-related changes; (iii) therapy-related AML; (iv) and AML not otherwise specified. As per the European LeukemiaNet (ELN) recommendations, these groups are further stratified as Favorable, Intermediate, or Adverse Risk based on the presence of molecular genetic and cytogenetic alterations (see Table 1).

TABLE 1.

2017 European LeukemiaNet AML risk stratification by genetics

Risk category	Genetic abnormalities
Favorable	t(8;21)(q22;q22.1); RUNX1-RUNX1T1
	inv(16)(p13.1q22) or t(16;16)(p13.1;q22); CBFB-MYH11
	Mutated NPM1 without FLT3-ITD or with FLT3-ITDlow
	Biallelic mutated CEBPA
Intermediate	Mutated NPM1 and FLT3-ITDhigh allelic ratio
	Wild-type NPM1 without FLT3-ITD or with FLT3-ITDlow allelic ratio (w/o adverse-risk genetic lesions)
	t(9;11)(p21.3;q23.3); MLLT3-KMT2a
	Cytogenetic abnormalities not classified as favorable or adverse
Adverse	t(6;9)(p23;q34.1); DEK-NUP214
	t(v;11q23.3); KMT2Arearranged
	t(9;22)(q34.1;q11.2); BCR-ABL1
	inv(3)(q21.3q26.2) or t(3;3)(q21.3;q26.2); GATA2,MECOM(EVI1)
	−5 or del(5q); −7; −17/abn(17p)
	Complex karyotypeb, monosomal karyotypec
	Wild-type NPM1 and FLT3-ITD high
	Mutated RUNX1d
	Mutated ASXL1d
	Mutated TP53e

Adapted from reference 6: Dohner et al., Diagnosis and management of AML in adults: 2017 ELN recommendations from an international expert panel.

^aThe presence of t(9;11)(p21.3;q23.3) takes precedence over rare, concurrent adverse-risk gene mutations.

^bThree or more unrelated chromosome abnormalities in the absence of one of the World Health Organization-designated recurring translocations or inversions, that is, t(8;21), inv(16) or t(16;16), t(9;11), t(v;11)(v;q23.3), t(6;9), inv(3) or t(3;3); AML with BCR-ABL1.

^cDefined by the presence of one single monosomy (excluding loss of X or Y) in association with at least one additional monosomy or structural chromosome abnormality (excluding core-binding factor AML).

^dThese markers should not be used as an adverse prognostic marker if they co-occur with favorable-risk AML subtypes.

^eTP53 mutations are significantly associated with AML with complex and monosomal karyotype.

Despite the aggressive multi-agent chemotherapy and allogeneic stem cell transplant (allo-SCT) regimens used in the treatment of AML, relapse is frequent even following typically occurring within 1–3 years after complete remission.¹ Even in favorable-risk AML, the relapse rate is still 30%–40%.^5,6 The formal hematologic definition of remission for AML allows for the persistence of significant AML burden⁷: (i) less than 5% blast cells without detectable Auer rods in a BM sample displaying more than 20% cellularity with maturation of all hemopoietic cell lines in the BM aspirate; (ii) absence of extramedullary leukemia; (iii) absence of leukemic blasts in PB; and (iv) PB neutrophil count greater than 1.5×10⁹/L and platelet count greater than 100×10⁹/L.⁸ It is the presence of the residual chemotherapy-resistant cells during remission that drive the leukemia relapse and make the cure AML challenging. Therefore, early and accurate detection of the remaining leukemia cells or MRD has emerged as a topic of interest in the past decade. Detection of MRD can impact clinical decision-making by allowing clinicians to predict survival and to identify patients at high risk of relapse.

2 | NEED FOR MINIMAL RESIDUAL DISEASE (MRD) DETECTION ASSAYS

There is increasing evidence that early detection of subclinical levels of residual disease or MRD provides additional prognostic information.⁹ Thus, MRD follow-up is considered crucial for the identification of AML patients at high risk of relapse. Nevertheless, the best method and timing of MRD detection still needs to be standardized, so it can impact clinical decision-making. In the following sections, we will discuss the different tools for MRD assessment and their associated sensitivities and limitations. We will also discuss what markers should be used in the different patient groups, what is the MRD level above which relapse is more likely to occur, a potential clinical decisions that can be made based on MRD quantitation.

3 | MRD MEASUREMENT BY REAL-TIME QUANTITATIVE PCR

Real-time quantitative PCR (RQ-PCR) is used to identify chimeric fusion genes (such as PML-RARA, AML1-ETO, or CBFB-MYH11), gene rearrangements, genetic alterations, and overexpressed genes. Chimeric fusion genes represent only 25% of AML cases and are well-established markers for MRD assessment because they are disease specific, extremely stable between diagnosis and relapse, and allow MRD detection by RQ-PCR assays with high sensitivity. However, due to their low incidence, this approach is limited to a small subgroup of patients.¹⁰ A larger subgroup of patients that would benefit from RQ-PCR assessment is the CN-AML subgroup. To date, the most clinically relevant marker for molecular monitoring in CN-AML is in the nucleophosmin 1 gene (NPM1) gene because it is present in 53% of CN-AML patients. NPM1 mutations (NPM1mut) are typically 4-nucleotide frameshift insertions in exon 12, 90% of which are reliably detectable by three RQ-PCR assays. The patient-to-patient consistency and relative stability through diagnosis and successive relapse in each patient has rendered NPM1mut an ideal and useful molecular MRD marker. Other mutated genes present in CN-AML such as FLT3-ITD (31% patients) and CEBPA (13% patients)¹¹ can be assayed similarly but demonstrate greater patient-to-patient variability in the specific type and position of mutation, thereby necessitating highly personalized RQ-PCR-based tests for which deployment is more challenging. Thus, NPM1mut remains the best-studied molecular MRD marker in CN-AML.

The practical utility of molecular MRD measurements in predicting outcomes and thereby possibly informing clinical decisions is exemplified by a recent report from Ivey and colleagues.⁵ Indeed, the study revealed increased risk of relapse in patients who demonstrate persistent NPM1-mutated transcripts in peripheral blood after the second chemotherapy cycle. Serial monitoring of paired BM and PB samples showed that analysis of marrow increases MRD detection rate, affording a median 1-log increment in sensitivity. Although they noted that mutations associated with preleukemic events remained detectable during remission after chemotherapy, they found that NPM1 mutations were detected in the majority of patients at the time of relapse concluding that NPM1 is a reliable marker for monitoring disease status. Other studies have also found NPM1 monitoring of prognostic value^12–15 with no universally accepted cutoffs for NPM1mut MRD levels or the kinetics of rise in mutated transcripts. However, a recent study analyzed NPM1 risk prognostication potential and showed that after induction therapy, patients who did not achieve a 4-log reduction in NPM1mut peripheral blood MRD (PB-MRD) had a higher cumulative incidence of relapse and a shorter overall survival. In this same study, a multivariable analysis demonstrated that an abnormal karyotype, the presence of FLT3-ITD, and a <4-log reduction in PB-MRD were significantly associated with a higher relapse incidence and shorter OS.¹⁶

The presence of the FLT3-ITD mutation is associated with worse outcomes, and therefore, FLT3 inhibitors are being studied as a potential target in multiple clinical trials.¹⁰ Evaluation of FLT3-ITD mutation by RQ-PCR has a reported sensitivity of 10⁻³ to 10^−4. In a study by Abdelhamid et al.,¹⁰ high levels of FLT3-ITD after induction chemotherapy were strongly associated with a shorter CR. In 17 of 20 evaluated patients, FLT3-ITD levels correlated with clinical evolution, suggesting the possibility of FLT3-ITD as a marker for MRD detection. However, the study also showed high discrepancy with other markers such as NPM1 and WT1 as three patients that relapsed had FLT3-ITD-negative disease. They attributed these findings to the loss of FLT3-ITD during disease evolution. In this study, detection of FLT3-ITD MRD was slightly more sensitive when measured in BM samples when compared to peripheral blood (PB) samples. These findings provide support for the validity of FLT3-ITD as a relevant molecular marker in CN-AML patients, in particular those treated by FLT3 tyrosine kinase inhibitors and those with no other molecular marker. Nevertheless, despite its strong association with relapse, the potential loss of FLT3-ITD at relapse should alert clinicians to use this marker with caution.¹⁰

As the majority of AML patients do not present an informative mutation for MRD monitoring, there has been interest in evaluating the utility of WT1 as an MRD marker by RQ-PCR. WT1 transcripts are overexpressed in 80%–90% of AML cases, and are detectable at low levels in normal cells; thus, WT1 represents a potentially more universal MRD marker for AML. Unfortunately, its utility for MRD assessment remains questionable due to its lack of specificity to leukemia cells.¹⁷ Additionally, the need for 50-or 100-fold upregulation to obtain sufficient sensitivity results is a restriction in the number of patients who can be evaluated.⁹ Lastly, as analyzed by Smith et al,¹⁸ maximum reproducible sensitivity of the CEBPA mutation by RQ-PCR was 10⁻⁴ (8.0 x 10⁻⁵; 10 in 125 000) and therefore was of the required order of sensitivity, >10⁻⁴ for an MRD assay.

As summarized in this section, diagnosis of MRD by RQ-PCR assessment is useful and highly sensitive for the two subgroups of patients described above: the 25% of patients that exhibit chimeric fusion genes PML-RARA, AML1-ETO, or CBFB-MYH11 and the 45% of CN-AML patients that have mutations in NPM1, CEBPA, and FLT3-ITD.

4 | DIGITAL PCR (DPCR)

Digital PCR is another technique that recently has emerged as a potentially powerful technique for MRD monitoring by PCR. A single PCR is partitioned into hundreds to millions of droplets or wells (depending on the technological platform) each containing single or few copies of the target template (Figure 1). The partitioned reaction is then submitted to thermocycling with each of these partitions constituting an individual reaction during the cycling process. Fluorescent signal is measured after an end-point PCR amplification for each partition individually. Thus, the number of positive molecules is determined from counting the number of successfully amplified fluorescent partitions rather than reaction kinetics as is the case with RQ-PCR. In addition, through absolute counting, the technique obviates the requirement for plasmid standards as is required for RQ-PCR. This technique has demonstrated promising results in the monitoring of residual disease in hematological malignancies using both RNA-based and DNA-based methods.^19,20 The sensitivity of dPCR is comparable to RQ-PCR²⁰ and has demonstrated special promise for detecting single nucleotide variations (SNVs) due to the greater capability to differentiate the mutant versus normal allele in the absence of competing normal allele in each partition.²¹

FIGURE 1.

Diagram illustrating the quantification of mutant NPM1 transcripts using real-time or droplet digital PCR for molecular MRD assessment. For RQ-PCR a standard curve is generated and NPM1 copies are determined by matching the Ct values to the curve. Digital PCR counts the absolute number of fluorescent partitions following both partitioning of the bulk PCR reaction and PCR amplification. [Colour figure can be viewed at wileyonlinelibrary.com]

Reliance of dPCR upon reaction end-points rather than reaction kinetics confers an ability to remain quantitative even under highly multiplexed assay conditions. Indeed, this attribute was reportedly effective in accurately quantifying hundreds of NPM1mut in a single assay covering the >95% of patients present with 4-nucleotide insertions in exon 12 at coding position 863.^21,22 NPM1mut assessment and schematic of dPCR technology are shown (Figure 1B). Outside of the AML, recent data also indicate similar results for multiplex detection of point mutations including EGFR mutations.²³

5 | MRD MEASUREMENT BY MULTIPARAMETER FLOW CYTOMETRY (MFC)

MFC techniques are based on antigen expression patterns that characterize the diverse lineages of normal hematopoietic cells. AML blasts have aberrant antigen expression patterns that are not detectable on the surface of bone marrow cells from healthy donors.²⁴ These distinct immunophenotypic patterns are called leukemia-associated immunophenotype (LAIP) and are present in 80%–100% of AML patients.²⁵ To date, using LAIP assessment by MFC has proven to be of prognostic impact.^24–26

The different LAIP expression patterns can be divided into four different types (Table 2): (i) overexpression—when normally expressed antigens are expressed to a higher degree on each individual cell (CD33, CD34, CD45, CD123); (ii) lack of expression—when normal expression antigens are lacking (HLA-DR); (iii) cross-lineage expression—when T-cell (CD2, CD3, CD5, CD7), B-cell (CD19), or natural killer cell (CD56) markers are expressed on the myeloid blasts; and (iv) asynchronous expression—when immature antigens (CD34, CD117) and mature myeloid antigens (CD13, CD38, CD15, CD11c, CD14, or CD65) are expressed together in an aberrant manner. For example, co-expression of CD11b, CD14, CD65, or CD56 antigens is not found in normal CD34+ BM cells.^9,24,27

TABLE 2.

LAIP examples and incidence

LAIP phenotype	Incidence	Examples
Overexpression	20%–30%	CD4, CD33, CD34, CD45, CD123, CD64
Lack of expression	20%–30%	HLA-DR
Cross-lineage expression	30%–40%	T-cell marker CD2, CD3, CD5, CD7
		B-cell marker CD19
		Natural killer marker CD56
Asynchronous expression	60%–70%	Immature myeloid markers CD34, CD117
		Mature myeloid markers CD11c, CD13, CD14, CD15, CD33, CD38, or CD65
Overall	90%–95%

Adapted from reference 23–25.

As in the case of mutations, LAIP expression in AML is very heterogenous, meaning that different LAIPs can be observed in a single patient. As a result of this heterogeneity, individual LAIPs are found in <5% of all AML patients. However, the use of comprehensive antibody panels allows for the definition of one or several LAIPs in practically all AML patients. When different LAIPs are present, the LAIP with the largest logarithmic difference to normal bone marrow (considering frequency of positive cells) should be chosen for MRD monitoring.²⁴

The sensitivity for MFC differs from other molecular approaches, such as RQ-PCR as shown recently by Petterson et al.¹² They found that in NPM1 mutant patients for MRD detection, RQ-PCR analysis of NPM1 was more sensitive (10⁻⁴) than MFC analysis (10⁻³) of common LAIPs which is concordant with sensitivities reported in other studies. Despite having a lower sensitivity for MRD detection when compared to RQ-PCR, MFC assessment has been the gold standard for evaluation of MRD as it is applicable to most patients. With the further development of multicolor flow cytometric platforms in recent years, up to ten-color antibody panels can be used, allowing for the analysis of more cell surface antigens in a single tube. However, as the number of parameters evaluated by MCF in a single tube increases, the complexity of the analysis increases therefore requiring an experienced pathologist. This method has shown the capacity of detecting 0.1% to 0.01% residual leukemic cells.^12,24,26 In an effort to increase the sensitivity of the MFC technique, Kern et al.²⁴ suggests the utilization of the pan-leukocyte antigen, CD45 to facilitate the identification of blasts. Myeloid precursors are separated from all other cells by weak CD45 antigen expression and low side scatter (SSC) signal. When CD45 and SSC characteristics are combined, separation of all relevant cell populations is improved, and the blast population covers a separate area (CD45low/SSClow) without overlapping with other populations. He suggests that the incorporation of CD45 gating in MRD diagnostics is a highly effective approach increasing sensitivity and specificity that should be investigated further.

To date, several clinical studies have used MFC to evaluate MRD and the ability to predict outcome. Studies have shown that a threshold of 3.5×10⁻⁴ cells is able to separate groups into good and poor risk subsets.^28,29 A study performed in 126 AML patients after induction chemotherapy demonstrated that postinduction MRD levels are able to separate patients into smaller different risk groups: very low risk (<10⁻⁴ cells), low risk (10⁻⁴ to 10⁻³ cells), intermediate risk (>10⁻³ to 10⁻² cells) with a 50% relapse rate, and high risk (>10⁻² cells) with an 84% relapse rate.³⁰ At least two other studies advocate for early assessment of MRD during the first 14–16 days after induction chemotherapy.^24,31 In 2010, Kern et al.³² demonstrated that a logarithmic difference >2.11 at day 16 after induction chemotherapy (LD16) was able to produce two different prognostic groups. Those with LD16>2.11 had a higher CR rate, improved 2-year EFS, and improved OS when compared to patients with lower LD. Therefore, patients with LD16 2.11 should be considered at high risk of relapse after induction chemotherapy. Gianfaldoni et al.³¹ determined that patients who achieved a complete remission on day +14 showed faster reduction of the MRD load by MFC within the first 8 days of treatment when compared to those who failed to achieve a remission on day +14. Despite all these data, adequate time to study remains controversial, as other groups have suggested that postconsolidation MRD levels have stronger prognostic power than postinduction levels (day 14–16).^28,29

As this technique becomes part of standard clinical laboratory procedures, efforts increase for methodology standardization across laboratories. Factors important for standardization procedures include sample processing, instrument configuration, selection of fluorochromes, antigen selection (including manufacturer and clone number), number of events to be acquired, the timing of evaluation (during induction, end of induction, end of consolidation), and the sample type (ie, peripheral blood vs. bone marrow). The benefits of MCF for MRD measurement are its applicability to most patients with AML (>80%) and a relative rapid turnaround. Drawbacks with this approach include: (i) need for standardization of antibody panels (which is being addressed by the ELN) and instrument configuration; (ii) pathologist trained in over eight-color interpretation of the flow cytometric data generated; (iii) difficulty collecting enough events to evaluate the rare population that contribute to relapse (which is estimated to necessitate at least 10⁸ cells); (iv) decreased sensitivity of at least 1 log below that of RQ-PCR assays; and lastly (v) in some patients, LIAPs at the time of diagnosis are different from the LIAPs found at the time of relapse, thus making it difficult to assess MRD in those patients that present a different LAIP as they relapse.³³

Another proposed approach to evaluate MRD using MFC is the evaluation and monitoring of leukemia stem cells (LSCs), a rare population of cells responsible for initiating and maintaining disease. LSCs can be defined as CD34+CD38− for the majority of AML patients, an immunophenotype that is shared with HSCs. Several studies have shown that AML patient samples with higher proportions of LSCs have worse clinical outcomes.⁷ When comparing LSCs to their normal counterparts, several cell surface markers have been reported to be aberrantly expressed in LSCs, and thus, they represent putative markers to isolate and identify LSCs. Such markers include CD47, CD96, CD44, CD32, CD25, CD133, CD123, TIM3, CLL-1, and IL1RAP.⁷ The use of LSC markers has also demonstrated prognostic relevance. For example, the study by Terwijn et al.,³⁴ used CD34+CD38− CLL-1+ as a marker to detect LSCs. Using this as a marker, the authors found that high levels of LSCs after the first cycle of chemotherapy predicted poor survival and that LSC-negative results had the better prognosis. However, as with LAIPs, LSC markers differ between patients and within patients. To overcome this challenge and to facilitate a more accurate assessment and monitoring of LSCs in AML patients with limited amounts of bone marrow (BM) cells, Zeijlemaker W and colleagues designed a single flow cytometric tube for LSC assessment.³⁵ The design of the LSC tube anticipates the potential emergence of new LSC populations during treatment and/or disease progression. Currently, validation experiments are ongoing for an improved standardized LSC detection and standardization of the assay.

6 | NEXT-GENERATION SEQUENCING

The use of single target assays for molecular MRD detection remains dominant in the field. However, given well-known patterns of clonal evolution evident in cancers and the need for highly personalized assays in patients with no recurrent or common lesion, there is a need for economical alternatives that enable more patients to be assessed for MRD in a manner robust to clonal evolution (molecular MRD) or phenotype switching (flow cytometry MRD). Going forward, next-generation sequencing (NGS) represents a potentially powerful alternative.

Patients with no established molecular marker represent 42%, 38%, and 68% of patients with ages <15, 15–60, and >60 years old, respectively.²⁰ Furthermore, due to the heterogeneity of the mutation repertoire in AML and the lack of hot spots in important frequently mutated genes (ie, RUNX1), it seems unfeasible to develop standardized assays on a per-patient basis. In addition, as it has been observed when tracking a single mutation, some AML patients will relapse with a different mutant clone, and in such case, the marker being monitored was not informative to predict relapse. With this technique, there is no need for patient-specific assays as practically all mutations are detected. Some recent studies have reported the use of NGS alone or in parallel with RQ-PCR for MRD detection.^36,37

One caveat of NGS that has limited its use for MRD assessment in the past few years is the sequencing error rate and its impact on the sensitivity of this technique compared with previously discussed methods.^9,20 Recently, the introduction of error-corrected read technologies has helped to overcome this limitation greatly improving its sensitivity (lower than 0.03%)³⁸ as well as the use of Bayesian analytical techniques for mutation calling informed by site-specific error rates and prior clinical data regarding mutation frequencies.³⁹ However, even the gains of error-corrected reads are likely to be eroded in certain genes by mappability limitations and errors arising from potential factors including gene paralogs.⁴⁰ In addition to error-corrected reads, further improvements in read length and read mapping algorithms coupled to statistically principled variant calling techniques are likely to extend the sensitivity, specificity, and thus overall utility of NGS in MRD monitoring.

Overall, NGS-based methods have the potential to detect subclinical disease in AML samples that do not qualify for the established leukemia-specific RQ-PCR or dPCR assays. In addition, NGS provides the ability to detect new emerging therapy-related leukemia that would otherwise be missed. Like MFC, NGS provides accurate information about leukemic tumor burden, an important parameter to measure treatment response. Validation and standardization in large clinical AML trials will be necessary.²⁰ However, NGS opens the possibility to measure MRD to a large cohort of patients.

7 | CONCLUSION

In adult AML, universally accepted standards for the measurement of MRD itself and how it should alter clinical decision-making are lacking. One possibility is that MRD monitoring will be most useful in guiding postconsolidation maintenance therapy in older patients for whom allogeneic SCT is not appropriate because of frailty or comorbidities. Before MRD for AML can be used as a standard for making treatment decisions for adults with AML, methods for MRD monitoring will need to be standardized and made readily available to all locations in which these patients are treated. This can best be accomplished by co-operative group trials like the MRC AML17 and AML19 studies that are currently investigating the beneficial role of MRD in disease surveillance in a large randomized manner to provide definitive evidence for this transition.

Immunophenotyping by MFC provides an excellent option for MRD monitoring applicable to most AML patients. Aberrant cells are detectable by MFC with sensitivities ranging from 10⁻² to 10⁻⁴ (1 abnormal cell in 100–10 000 normal cells). Even more sensitive for follow-up monitoring is the RQ-PCR or dPCR methodologies, which can achieve sensitivities of 10⁻⁴ to 10⁻⁶, representative of the detection of one abnormal cell in 10 000–1 000 000 normal cells.²⁴ It is important to note that sensitivities vary as a function of the technology, amount of input material, and target-specific features. Ranges of sensitivity are summarized in Table 3. The downside to RQ-PCR is that in contrast to MFC which is applicable to most patients, RQ-PCR monitoring is only clinically relevant for a subgroup of AML patients, those patients carrying either distinct reciprocal gene fusions (CBFB-MYH11, RUNX1-RUNX1T1, PML-RARA fusion genes) or an NPM1 mutation. Perhaps for the sub-group of patients that have targetable mutations by RQ-PCR (eg, NPM1 mutation or fusion genes), RQ-PCR should be the method of choice for MRD measurement. For patients with no targetable abnormalities, MFC should be the method of choice. Prognostic data are available for both RQ-PCR and MFC, and in the next years, we should expect prognostic data with assessment of MRD by NGS and digital PCR. Most studies validating timing of MRD have determined that it should be done early within the first 14–16 days after induction chemotherapy for risk stratification, and postconsolidation if more accurate prognosis is needed.

TABLE 3.

Sensitivity rages and standardization factors

Method	Sensitivity	Optimal timing	Factors that need to be standardized	References
RQ-PCR	10−4–10−6	Unclear	• Sample type (BM or PB)	Abdelhamid (2012)10
			• Sample processing	Smith (2006)18
			• Heparin vs EDTA in tubes for collecting samples	Hindson (2013)21
			• Analytical cutoffs	Kern (2010)24
				Pettersson (2016)12
dPCR	10−4–10−6	Unclear	• Sample type (BM or PB)	Schlenk (2008)11
			• Sample processing	Kern (2010)24
			• Heparin vs EDTA in tubes for collecting samples	Pettersson (2016)12
			• Analytical cutoffs
Multiparameter flow cytometry (MFC)	10−3–10−5	Postinduction (day 14–day 16) or Postconsolidation	• Sample processing	Pettersson (2016)12
			• Instrument configuration	Buccisano (2006)28
			• Selection of fluorochromes	Venditti (2000)29
			• Antigen selection	Gianfaldoni (2006)31
			• Number of events to be acquired	Oelschlagel (2000)33
			• Timing of evaluation	Kern (2010)24
			• Sample type (BM or PB)	Jaso (2014)26
			• Training of analyst
Next-generation Sequencing (NGS)	10−2 or 10−4 with error read correction	Unclear	• Sequencing platform	Young (2016)38
			• Library preparation protocol	Grimwade (2014)20
			• Clinically relevant genes to be targeted
			• Depth of coverage required
			• Data analysis pipeline and variant calling algorithms

The best method for MRD detection still needs to be determined, so it can impact clinical decision-making. Currently, approaches such as cytogenetics, RQ-PCR, and MFC are being utilized for MRD assessment, and newer technologies such as dPCR and NGS may provide a way to monitor several mutations at once to potentially more accurately predict the relapse clone and deal with evolution and genetic heterogeneity (Figure 2). The ability to develop assays covering a range of mutations or immunophenotypes is especially important in carefully regulated settings such as New York State where it is challenging currently to develop and obtain regulatory approval for a patient-tailored test in the time frame of an AML relapse (eg, custom RQ-PCR test for a rare NPM1 mutation).

FIGURE 2.

Current MRD assessment techniques include: cytogenetics, RQ-PCR, and flow cytometry. Newer techniques such as digital PCR and next generation sequencing are still under evaluation. [Colour figure can be viewed at wileyonlinelibrary.com]