Minimal Residual Disease to Monitor Therapy in Acute Leukemia

Treatment response in patients with acute leukemia is traditionally measured by searching for leukemic cells that persist during chemotherapy. This is done by screening smears of bone marrow aspirates and peripheral blood samples using light microscopy, striving to identify leukemic cells by their morphologic features. However, leukemic cells in most cases closely resemble normal bone marrow and blood cells: acute lymphoblastic leukemia (ALL) lymphoblasts can be confused with normal immature lymphoid cells and activated lymphocytes; acute myeloid leukemia (AML) myeloblasts are similar to normal immature myeloid cells. Hence, the identification of leukemic cells among normal cells is a difficult task even for an experienced hematopathologist, and the cutoff to define leukemia remission is therefore rather lax; at most centers it is 5%. However, patients with <5% leukemic blasts in the bone marrow and in complete remission by this criterion may still have a large total number of residual leukemic cells (potentially up to 10¹⁰).

Measuring treatment response more accurately is important because the degree of initial leukemia cytoreduction is strongly associated with risk of subsequent relapse. It is well known that patients who have a poor response to remission induction therapy and fail to achieve morphologic remission have a high risk of relapse. However, with contemporary therapy, the majority of patients with ALL and AML do achieve complete remission as defined by the above criterion, yet a substantial number subsequently relapse. The hypothesis underlying studies of minimal residual disease (MRD), which denotes leukemia not detectable by conventional morphologic analysis, is that among “good responders” there may be substantial heterogeneity in residual leukemia burden that is unappreciated by morphologic analysis. Studies reported during the past 2 decades are in line with this hypothesis, and the prognostic importance of MRD has been demonstrated unequivocally.^1,2 Therefore, the more stringent definition of remission provided by MRD assays is progressively replacing that defined by morphologic analysis.

Methods to Measure MRD
Leukemic cells have several distinguishing features that can be used to track them during chemotherapy. ALL cells have unique clonal rearrangement of the genes encoding immunoglobulin (Ig) and T-cell receptor (TCR) proteins that can be identified in the majority of cases.³ The leukemia-specific rearrangements are elucidated in each patient at diagnosis, and specific primers are synthesized and then incorporated into a polymerase chain reaction (PCR) assay. After determining the optimal conditions of the assay, these are applied to follow-up samples to monitor MRD. PCR analysis of Ig and TCR genes allows for the routine detection of 1 leukemic cell in 10,000 to 100,000 normal cells.³ In a study by the AIEOP-BFM ALL group, only 305 of the 3341 (9.1%) diagnostic samples examined either lacked suitable gene rearrangements for PCR analysis or had rearrangements with sequences not sufficiently distinct to reach a sensitivity of 0.01%.⁴ In a study performed at St. Jude Children’s Research Hospital, 475 of 539 (88.1%) bone marrow samples from patients with newly diagnosed B-lineage ALL had clonal antigen receptor gene rearrangements suitable for PCR monitoring of MRD with a sensitivity of at least 0.001%.⁵ Ig and TCR genes are rarely rearranged in AML and therefore their contribution to the range of potential targets is negligible.³ Another feature that can be used to distinguish leukemic from normal cells consists of chromosomal abnormalities and their resulting gene fusions.³ Fusion transcripts, such as BCR-ABL1, MLL-AFF1, TCF3-PBX1, and ETV6-RUNX1 in ALL, and RUNX1-RUNX1T1, CBFB-MYH11, and PML-RARA in AML can be used as targets for amplification.^2,3 Real-time PCR provides the most accurate way to measure their levels. Overall, up to one-third of patients with ALL or AML have leukemic cells with genetic abnormalities that currently can be performed as routine assays in molecular pathology laboratories, allowing the detection of 1 leukemic cell in 1000 to 100,000 normal bone marrow cells.^2,3 Abnormal fusion transcripts may persist long into clinical remission. Therefore, positive results, particularly at low levels, are difficult to interpret for some of the fusion transcripts such as ETV6-RUNX1, RUNX1-RUNX1T1, and CBFB-MYH11 as they often do not indicate a higher risk of relapse, but sequential monitoring may be informative.^6-8 In addition to gene fusions, NPM1 gene mutations, when present, can be used as targets for MRD studies.⁹

Leukemic cells can also be identified by the expression of unique combinations of cell markers that are best detected with monoclonal antibodies and flow cytometry (Figure).¹⁰ The most current flow cytometers allow the detection of 8 or more markers. Leukemia-associated immunophenotypes are identified at diagnosis in each patient by comparing the cell marker profile of leukemic blasts to that of reference bone marrow samples from healthy donors and from patients during and after chemotherapy or transplant.¹⁰ If a sufficient number of cells are analyzed (eg, 100,000 mononuclear cells per antibody combination), flow cytometry has a routine sensitivity of detection of 0.01% in ALL.¹⁰ Among 2143 patients with B-lineage ALL enrolled on 9900 series treatment protocols of the Children’s Oncology Group, a test with the sensitivity of at least 0.01% on day 29 could be performed in 92% of patients.¹¹ In the St. Jude Total XV study, MRD could be monitored by flow cytometry with a 0.01% sensitivity in 482 of 492 patients (98%).¹² In AML, distinctive markers can also be identified in most patients, although in approximately 40% of patients the routine sensitivity that can be achieved is not higher than 1 in 1000, a consequence of partial overlap between the phenotype of leukemic cells and those of normal hematopoietic cells.¹³ In the studies performed in the St. Jude AML02 trial, we found immunophenotypes suitable for MRD studies with a sensitivity of at least 0.1% in 200 of 210 (95%) patients.¹³ The number of antibody combinations used to identify leukemic cells, the stability of the markers targeted, and the skill of the investigator interpreting the results are key to the reliability of MRD monitoring by flow cytometry. 

Prognostic Significance of MRD in ALL

Many studies have demonstrated the clinical significance of MRD monitoring in children and adolescents with ALL. MRD tests performed during the early phases of treatment appear to be particularly informative and often outweigh the prognostic impact of clinical and biological features previously associated with treatment outcome (reviewed in Campana¹). For example, a recently reported study of 3184 B-lineage ALL patients enrolled in the AIEOP-BFM ALL 2000 protocol showed that MRD measurements by PCR amplification of antigen-receptor gene rearrangements on days 33 and 78 of treatment were superior to previous stratification criteria based on leukocyte count, age, early response to prednisone, and genetic subtype.¹⁴ Other recent studies also pointed to the predictive value of MRD for otherwise homogeneous subgroups of patients, such as infants,¹⁵ those with T-ALL,16and those with specific genetic abnormalities.¹⁷ Another study from the same group measured MRD by flow cytometry on day 15 of treatment in 815 patients and found a strong correlation between higher MRD levels and subsequent relapse.¹⁸ A high level of MRD (ie, ≥1%) at the end of remission induction predicts a particularly poor outcome, and it remained an unfavorable predictor of outcome in the St. Jude Total XV study (the only independent adverse predictor together with CNS3 status), despite MRD-based treatment intensification.¹² Conversely, patients who attain MRD <0.01% early during therapy have excellent chances to remain in continuous complete remission with contemporary postremission therapy. In our early study with patients enrolled in the Total XIII study, those with early leukemia cell clearance (ie, MRD <0.01% on day 19 of treatment; 46%) had a 5-year cumulative incidence of relapse of 6.0% ± 3.4%.¹⁹ Likewise, in the AIEOP-BFM ALL 2000 study, patients with MRD <0.01% on day 33 (42%) had a 5-year event-free survival of 92.3% ± 0.9%.¹⁴

It is also clinically useful to monitor MRD in children and adolescents with first-relapse ALL who achieve a second remission,²⁰ and in those undergoing allogeneic hematopoietic stem cell transplantation, as the level of MRD before transplant is associated with risk of relapse posttransplant.^21,22 Finally, convincing evidence is emerging that MRD is clinically important in adult patients with ALL. In studies by the German Multicenter Study Group for Adult Acute Lymphoblastic Leukemia, MRD on days 11 and 24 of treatment predicted outcome in adult patients with “standard-risk” ALL,²³ and sequential studies of MRD were also informative.²⁴ Other studies also demonstrated the predictive value of MRD in different subgroups of adult ALL (reviewed in Campana¹).

Prognostic Significance of MRD in AML

In an early study, Children’s Oncology Group investigators showed that children with AML in remission who were MRD positive had a 4.8 times higher risk of relapse than those who were MRD negative and that MRD was the strongest prognostic factor among the variables considered.²⁵ Likewise, a study performed in the St. Jude Children’s Research Hospital AML97 cohort showed that the mean 2-year survival estimates for patients with MRD positivity (≥0.1%) after induction therapy was 33%, compared with 72% for those with lower levels or no detectable MRD; MRD was also the strongest predictor of outcome in this cohort.²⁶ Subsequent studies in children with AML consolidated these early findings.^27,28 In the St. Jude AML02 study, we used MRD to guide treatment schedule and intensity. Patients with MRD ≥1% at the end of remission induction therapy were classified as high risk and offered transplant; patients with MRD 0.1% to <1% received gemtuzumab ozogamicin and, if MRD persisted, became candidates for transplant. MRD remained a predictive prognostic factor, although patients with lower levels of MRD (0.1%-<1%) appeared to benefit from this strategy.¹³

Early studies in adult patients with AML showed that MRD detected in the first bone marrow obtained after induction treatment was highly predictive of relapse.²⁹ Others reported that the rate of decrease in cells expressing aberrant immunophenotypes during treatment was significantly and independently related to treatment outcome.³⁰ More recently, Maurillo and colleagues³¹ reported that levels of MRD after consolidation therapy were particularly informative: MRD-negative patients had a significantly better outcome, regardless of whether they underwent autologous or allogeneic stem cell transplantation. 

Practical Considerations for the Clinical Application of MRD 

In patients with ALL, MRD results obtained by PCR amplification of Ig and TCR genes or by flow cytometry are generally concordant when MRD is present at levels of ≥0.01%.^32,33 Each method has unique strengths (Table).¹⁰ Flow cytometry is more likely to be readily available as it is routinely used for leukemia profiling for diagnostic purposes. For studies at early time points during therapy, such as on day 15, flow cytometry has an edge over PCR because the development of a PCR assay currently requires more than 2 weeks. By contrast, the higher sensitivity of PCR can be advantageous for studies at the end of therapy or after transplant, where it may uncover MRD undetectable by flow cytometry. PCR has often been said to be more expensive, but many variables must be factored into this calculation including equipment costs, consumables, and staff support. In our experience, the costs of the 2 methods are similar. Both assays must be performed in laboratories with proven expertise; therefore the type of expert laboratory available to a cancer center or cooperative group may ultimately be the decisive factor in selecting the method to be used to monitor MRD in ALL. Flow cytometry is the only method that can study MRD in most patients with AML. Studies of MRD by PCR amplification of fusion transcripts are also an option, but they are obviously limited to specific leukemic subtypes and are generally used on an ad hoc basis. MRD is most commonly tested in bone marrow aspirates, as the bone marrow is the site where leukemic cells grow and are typically found in higher proportions. Some studies examined the potential for using peripheral blood instead. It was found that in patients with T-lineage ALL, MRD levels in peripheral blood were similar to those in bone marrow.^34,35 Interestingly, this was also the case in patients with T-cell lymphoblastic lymphoma.³⁶ Thus, in patients with T-cell malignancies, sequential MRD testing can be performed in blood. In patients with B-lineage ALL, early MRD measurements in peripheral blood (eg, on day 8) may be clinically informative.¹¹

The clinical significance of MRD and its use for treatment stratification depends on protocol design. Therefore there is considerable variability in the time points during treatment at which MRD studies are performed and in the levels of MRD considered to be most informative. The current COG AALL08B1 protocol for children with B-lineage ALL includes MRD measurements in peripheral blood on day 8 of treatment and in bone marrow at the end of remission induction therapy (day 29), and the cutoff levels used to assign risk are 1% and 0.01%, respectively. In the Dana-Farber Cancer Institute study 0501 for childhood ALL, the MRD threshold of 0.001% at the end of remission induction therapy (day 28) is used for risk assignment. In the St. Jude Children’s Research Hospital Total XVI study, MRD levels on day 15 and day 42 are used for treatment assignment. Patients with MRD of ≥1% on day 15 receive intensified remission induction therapy; further intensification is reserved for patients with ≥5% leukemic cells. Patients with MRD <0.01% on day 15 receive a slightly less intensive reinduction therapy and lower cumulative doses of  anthracycline. Patients with standard-risk ALL who have MRD of ≥0.01% on day 42 are reclassified as high-risk; patients with MRD ≥1% are eligible for allogeneic transplant in first remission. For patients with AML, MRD is typically measured after each block of chemotherapy, as well as before and after transplant. In the current St. Jude AML08 study for children and adolescents with AML, 0.1% is the threshold used to define MRD positivity, while levels of MRD ≥1% indicate high-risk AML. 

The prevalence of MRD varies among different genetic subtypes of ALL and AML.^{1,12,13,37,38} Other presenting features that have been shown to be associated with MRD are gene profiles of leukemic cells,^39,40 and germline- or leukemia-associated gene polymorphisms.⁴¹ These associations raise the question whether MRD studies can be replaced by these presenting features. However, in virtually all correlative studies performed to date, MRD was an independent prognostic factor, and it could dissect subgroups of patients with different outcomes among those with apparently identical leukemia subtype. 

In addition to measuring initial response to chemotherapy, MRD assays have many other uses in the management of patients with acute leukemia. For patients who achieve MRD-negative status after remission induction therapy, conversion to MRD positivity alerts of the possibility of relapse; a subsequent increase in MRD levels usually indicates impending relapse. Therefore, such a finding can give a head start in the preparation for salvage therapy and/or transplant. Because levels of MRD before transplant are strongly related to the success of the procedure, it is increasingly common to attempt to reduce them with chemotherapy before initiating the transplant conditioning, although evidence that this practice is effective is not yet available. MRD measurements can also be used to guide intervention post-transplant, such as reduction of immunosuppression, administration of donor lymphocyte infusions, or preparation for a second transplant. 

The Future of MRD Studies 

Evidence for the prognostic importance of MRD in both ALL and AML is solid. One could argue that the collective data supporting the clinical use of MRD are considerably more compelling than those provided in the past in support of established parameters widely used for risk stratification. Moreover, as discussed in this article, MRD testing has multiple applications during the course of treatment that go beyond prognostication. Clinical studies have shown that MRD results can be applied to most patients with ALL and AML and can be delivered in a timely fashion.^12,13 Some believe that, with time, MRD studies may become obsolete in the context of the ever-increasing knowledge of the biology and genetics of leukemic cells, but we think it unlikely as MRD measurements reflect the composite influence of multiple factors, such as drug dosage, drug interaction, pharmacokinetic and pharmacogenomic variable, and compliance with the scheduled treatment plan, which cannot be captured by examining the leukemic cells alone. 

Methods to study MRD are continually being refined.⁴² MRD studies can be relatively expensive in relation to other routine diagnostic laboratory assays but, overall, are not significantly more expensive than high-complexity tests, such as cytogenetics or molecular pathology. Moreover, when one considers the potential benefits of avoiding undertreatment and relapse, or overtr eatment and its sequelae, the cost of MRD testing is fully justified as it can ultimately reduce the cost of patient care. Regardless, current MRD assays remain too expensive and complex for many centers to implement. Thus, the potential benefits of careful leukemia monitoring are limited to a minority of patients. The approach of having reference centers performing the tests in the context of multicenter studies can help in this regard, as it saves costly and time-consuming setup and standardization efforts in peripheral centers while ensuring an expert sample processing and analysis. Ultimately, however, it would be desirable to make MRD testing simpler and less expensive. To this end, we developed a simplified flow cytometric assay to study MRD in patients with B-lineage ALL during remission induction therapy.⁴³ When applied to samples collected on day 19 of treatment, the results of the simplified assay correlated well with those of the more complex flow cytometric assay and those of PCR amplification of antigen-receptor genes. This assay is currently being used in international protocols to identify patients with good initial response to therapy who are then offered a low-intensity, low-toxicity treatment regimen, an approach that has produced encouraging preliminary results. 

It is sometimes disputed that only a few leukemic cells have long-term repopulating capacity (“leukemia stem cells”) and that current MRD assays cannot determine whether the signals detected originate from these or more differentiated cells incapable of driving relapse. This is a rather controversial topic. Evidence that such "stem cells,” defined as leukemic subsets that preferentially grow in immunodeficient mice, also have a growth advantage in humans is lacking. Data indicating that such subsets are more resistant to chemotherapy are less than compelling. In any case, the strong correlation between MRD levels and relapse indicates that MRD positivity should be interpreted as evidence for the persistence of leukemic cells that are resistant to further chemotherapy and are capable of driving the recurrence of clinically overt disease. 

Besides their use in clinical management, MRD studies are likely to be used more often as a surrogate marker to measure the effectiveness of novel antileukemic therapies. For example, MRD levels during and at the end of remission induction therapy can be used to develop a stopping rule when comparing the efficacy of a new remission induction regimen in relation to that of its predecessor. In addition, MRD testing could be used to quickly determine whether a new agent exerts significant antileukemic effects in vivo, thus sparing the need for lengthy trials with ineffective agents.