Measurable residual disease in chronic myeloid leukemia

Abstract

Chronic myeloid leukemia is characterized by a single genetic abnormality resulting in a fusion gene whose mRNA product is easily detected and quantified by reverse-transcriptase polymerase chain reaction analysis. Measuring residual disease was originally introduced to identify patients relapsing after allogeneic stem cell transplantation but rapidly adopted to quantify responses to tyrosine kinase inhibitors. Real-time quantitative polymerase chain reaction is now an essential tool for the management of patients and is used to influence treatment decisions. In this review we track this development including the international collaboration to standardize results, discuss the integration of molecular monitoring with other factors that affect patients’ management, and describe emerging technology. Four case histories describe varying scenarios in which the accurate measurement of residual disease identified patients at risk of disease progression and allowed appropriate investigations and timely clinical intervention.

Introduction

Monitoring residual disease has been integral to the management of chronic myeloid leukemia (CML) for more than 30 years, and has paved the way for the introduction of similar methodology for assessing measurable residual disease (MRD) in other malignancies.

Patients with CML have an ideal marker to directly measure therapy response: the BCR::ABL1 fusion oncogene. BCR::ABL1 is the product of the t(9:22) chromosomal translocation, which in >95% of patients can be visualized in karyotyping as a shortened chromosome 22, termed the Philadelphia chromosome. The genetic breakpoints occur in well-defined regions leading to common RNA transcript types, termed e13a2 and e14a2, in approximately 98% of patients (Figure 1).¹ These transcripts only differ by 75 base pairs and can be measured in a single assay.

Identifying residual leukemic cells, by cytogenetic and later, molecular technology, was originally employed to recognize disease recurrence after allogeneic stem cell transplantation (SCT).^2-5 The ability to identify early relapse became particularly important after the observation that the infusion of additional donor lymphocytes was capable of restoring durable remissions.⁶ Further work confirmed that donor lymphocyte infusions were more likely to be effective if delivered at the point of low disease burden, and necessitated the development of more sensitive methodology to identify and later quantify residual or emerging leukemia.⁷

As a consequence, reverse transcriptase polymerase chain reaction (RT-PCR) became the molecular monitoring workhorse and over time, the methods advanced from qualitative to quantitative BCR::ABL1 detection when it became apparent that a positive signal after allogeneic SCT had limited predictive value for relapse.⁸ Serial analysis of quantitative BCR::ABL1 mRNA levels provided more information and identified patients at risk of relapse to allow timely therapeutic intervention.⁹ This early work in the 1990s heralded the era of quantitative measurement of BCR::ABL1 transcripts, which became highly relevant after the introduction of tyrosine kinase inhibitor (TKI) therapy when most patients rapidly achieved BCR::ABL1 levels that could only be measured using sensitive molecular analysis. Techniques advanced from competitive PCR to real-time quantitative PCR (RT-qPCR) and, more recently, to digital PCR.^10-16 Recent reviews have comprehensively discussed the standardization of PCR methods,^17,18 and future molecular technology for monitoring patients with CML.¹⁹

Molecular monitoring of CML is now well established, widely used, and is the recommended monitoring strategy in international guidelines.^20,21 Evidence-based, milestone-driven molecular results define levels of response, guide therapeutic decisions and direct BCR::ABL1 kinase domain mutation analysis to assess for drug resistance. The current recommendations from the European LeukemiaNet (ELN)²⁰ and the National Comprehensive Cancer Network (NCCN)²¹ are summarized in Tables 1 and 2. These have evolved over time but maintain a focus on early molecular response in the first 3-12 months of therapy. The initial degree of BCR::ABL1 reduction is a powerful predictor of response.^22-29 Patients who achieve a major molecular response (MMR, BCR::ABL1 ratio ≤0.1% on an international scale [IS]) are highly unlikely to experience disease progression. Deep molecular responses (DMR, BCR::ABL1 ratio ≤0.01% IS), sustained for 1-2 years, are a prerequisite to trial treatment discontinuation and possible treatment-free remission (TFR).

Figure 1.Schematic of BCR::ABL1 transcripts. (A) BCR and ABL1 genes showing the general location of breakpoints. The red arrows are breakpoint regions that generate rare BCR::ABL1 transcripts. (B) Size differences of the typical transcripts and some of the rare transcripts. Note that the direct fusion of BCR exon 8 and ABL1 exon 2 does not generate an in-frame protein. The e8a2 BCR::ABL1 transcript requires an inserted sequence or a genomic break within an exon to generate a constitutively activated protein. (C) Characterization of the BCR::ABL1 transcript type is essential at the time of diagnosis. Multiplex reverse transcriptase polymerase chain reaction techniques can simultaneously detect various transcript types. Gel image courtesy of Professor Andreas Hochhaus, Universitätsklinikum Jena, Germany. PCR: polymerase chain reaction; CML: chronic myeloid leukemia.

Table 1.European LeukemiaNet²⁰ and National Comprehensive Cancer Network²¹ treatment response milestones for chronic myeloid leukemia expressed as BCR::ABL1 on the International Scale. Bold text indicates where the recommendations differ.

Despite the widespread use and clinical applicability of monitoring BCR::ABL1 ratios, the molecular methodology is not perfect. It is not always easy to maintain consistency in the results and clinicians should be aware of the pitfalls. As a consequence, it is important to consider trends in BCR::ABL1 ratios, and avoid making management decisions on the basis of a single result. The use of an internal control gene is essential to maintain reliability and reproducibility as it determines the quality of individual RNA samples and compensates for differences in the BCR::ABL1 transcript level due to sample degradation.³⁰ Appropriate control genes are ABL1, GUSB and BCR and molecular values are reported as the percentage ratio of BCR::ABL1 transcripts to the control gene transcripts on the IS. ABL1 is the most widely used control gene. The effective measurement range on the IS is for BCR::ABL1 ratios of ≤10% due to potential methodological inaccuracies at higher levels related to the control genes.^31,32 In the laboratory, vigilance is required to monitor and detect any shift in the ratios that might occur with a myriad of factors, such as something as simple as a new lot of reagents. Enrolment in quality assurance programs and regular use of quality control material to identify and mitigate these trends are essential. A change of methodology may require re-calculation of the IS conversion factor. Unfortunately, despite being recommended from the early days of the international effort to harmonize methods, it is often unclear how rigorously these quality controls are used in daily practice.^30,31

Molecular monitoring, ongoing for very many years will be essential in the management of most patients. In most patients the BCR::ABL1 decline is rapid upon initiation of TKI treatment but the trend and dynamics of BCR::ABL1 ratios over time provide information to guide clinical decisions.^33,34 The dynamics of an initial BCR::ABL1 decline can be measured as the BCR::ABL1 halving time. A number of studies using various control genes to measure BCR::ABL1 (ABL1, GUSB or BCR) have reported an association between the halving time and molecular response.^32,34-37 Similarly, BCR::ABL1 doubling times provide prognostic information for the disease phase at loss of response, using BCR or ABL1 control genes.^35,38 At this stage, BCR::ABL1 halving and doubling times are non-standardized metrics and as such, are not included in guidelines for routine monitoring of CML.

In this review we present examples of long-term molecular results, their clinical interpretation and guidance on therapeutic decisions for individual patients diagnosed in chronic phase (CP). Our aim is to provide advice that may ultimately enhance patients’ management and outcome.

Transcript types: relevance for molecular monitoring and treatment outcome

By 2013, there was a substantial body of evidence demonstrating the importance of standardized molecular monitoring for the prediction of response for TKI-treated patients.^22-29 At this point the ELN recommended that BCR::ABL1 ratios at specific timepoints should be used to guide therapy decisions.³⁹ This was a decade after the association between achievement of MMR and a reduced risk of progression had been reported.⁴⁰ All of this work was performed for patients with the most frequent BCR::ABL1 transcript types, e14a2 and e13a2. Note that e14a2 is the primary transcript for patients who co-express e13a2 and e14a2. In these patients e13a2 is expressed due to alternative splicing.⁴¹ The transcript type must be characterized at diagnosis to ensure an appropriate method is used to monitor the remaining 2% of patients with atypical BCR::ABL1 transcripts. Standardized molecular methods for BCR::ABL1 monitoring are not designed for patients with atypical transcripts and if used will generate false negative results.^20,42

Different proteins are translated from each of the BCR::ABL1 transcript types that can theoretically influence the biological properties of the disease and potentially affect response to therapy. In the pre-TKI era some studies reported an inferior outcome for patients with the e14a2 transcript, including shorter duration of CP and shorter time to the onset of blast phase (BP)^43,44 but these findings were not always corroborated.^45-48 Fast forward to the TKI era and several studies have assessed the influence of transcript type on outcome.^49-54 Although the findings were occasionally contradictory^55,56 the body of evidence suggested that patients with the e13a2 transcript reached milestones, such as complete cytogenetic response (CCyR), MMR and DMR, more slowly than those expressing e14a2. Two studies found a difference in overall survival, but one favored the e14a2 transcript⁵⁰ and the other the e13a2 transcript.⁵⁴ Overall, the data do not currently support a specific upfront therapy recommendation based on transcript type.^20,21

Table 2.European LeukemiaNet²⁰ and National Comprehensive Cancer Network²¹ treatment response and testing recommendations.

It is also possible that methodological differences could be responsible for some of the associations observed between the BCR::ABL1 transcript type and molecular responses.^57-60 Most real-time PCR methods amplify the e14a2 and the shorter e13a2 transcript in a single reaction using a calibration standard that contains the e14a2 fusion junction. Theoretically, the efficiency of PCR amplification could be enhanced for the shorter e13a2 transcript and generate artificially higher BCR::ABL1 values. A bias in the reported BCR::ABL1 values has indeed been demonstrated between the transcript types for a number of methods.^57,59,60 The degree of bias varied and may or may not alter the interpretation of the reported value and influence treatment decisions.^57,60

Technical differences in the efficiency of PCR amplification do not easily explain recent observations that the BCR::ABL1 transcript type might influence the achievement of TFR. The molecular relapse rate after stopping therapy was twice as high for patients with the e13a2 transcript compared with e14a2 in the Destiny study of TKI de-escalation prior to treatment cessation.⁶¹ Similar results from smaller studies were reported in patients who stopped TKI therapy after achieving sustained DMR.^62-64

Role of digital polymerase chain reaction in the assessment of rare transcripts and prediction of successful treatment-free remission

Digital PCR provides absolute quantification and should not be subject to bias associated with differences in amplification efficiency (Figure 2). These methods allow replicate analysis to improve the detection of rare transcripts. The sample is divided into thousands of individual replicate PCR using nanofluidic technology and the reagents and workflows are similar to those of real-time PCR using hydrolysis probes. Digital droplet PCR (ddPCR) uses a water-oil emulsion droplet system and the sample is partitioned into droplets. PCR amplification occurs in each of the droplets and at the end of the reaction each droplet is assessed to establish the fraction of positive droplets. The high number of replicates enhances the precision of target detection.

Digital PCR has demonstrated higher sensitivity for the detection of BCR::ABL1 transcripts¹⁰ and has been used to detect residual disease at the time of stopping TKI in patients who attempted a trial of TFR.^11,12,14 More sensitive detection of BCR::ABL1 was associated with a higher rate of molecular relapse after TKI discontinuation. In two studies a cut-off BCR::ABL1 positivity value was established in order to predict relapse.^12,14 Nicolini and colleagues in their 2019 study¹⁴ stressed that ddPCR was not ready to be incorporated into the criteria for eligibility of a clinical trial of TKI cessation. The assay requires careful calibration of signal-to-noise ratio and standardization across laboratories. Furthermore, excluding patients on the basis of a ddPCR value at the time of considering TKI cessation would exclude a proportion of patients who would maintain TFR. A recent multicenter study has demonstrated the feasibility of using ddPCR to monitor treatment response.⁶⁵ Broader use of digital PCR would require a thorough demonstration of comparable results across laboratories and clinical applicability if it were to be used to direct treatment decisions. Toward this goal, the performance characteristics of the first US Food and Drug Administration approved ddPCR assay for BCR::ABL1 monitoring have been published.⁶⁶ The study demonstrated reproducibility of results across laboratories. Attempts to increase the sensitivity of the methodology for better prediction of TFR included BCR::ABL1 DNA PCR.⁶⁷ The method could not reliably predict TFR and is not currently recommended.

Figure 2.Measurement of residual disease during treatment. Measurable residual disease in chronic myeloid leukemia can be measured using real-time quantitative polymerase chain reaction or digital polymerase chain reaction: analog versus digital measurement. Slide courtesy of Dr Jerry Radich and Dr Daniel Chiu. PCR: polymerase chain reaction; MRD: measurable residual disease; IS: International Scale.

Molecular monitoring in resource-poor regions

Standardized molecular monitoring is not available in all regions due to economic constraints. Automated systems for measuring BCR::ABL1, such as the Cepheid GeneXpert,⁶⁸ relieve the burden of resource-intensive in-house method development, optimization, IS standardization and validation. The Cepheid technique incorporates a stand-alone microfluidic system in which all processes necessary to generate a standardized BCR::ABL1 ratio occur within a disposable cartridge. Results are generated rapidly and the method requires minimal training. The system may be a viable option for monitoring patients with CML in resource-poor regions. Shipment of dried blood spots by regular mail has also been demonstrated as a means of extracting viable RNA and for generating reliable standardized BCR::ABL1 ratios.⁶⁹ This process is a cost-effective alternative to shipment of samples to a central laboratory for testing.

The impact of genomic heterogeneity at the time of diagnosis for treatment response

Evidence has accumulated using various next-generation sequencing techniques (Figure 3) that mutation of cancerrelated genes is associated with treatment failure and drug resistance in CML.^70-76 In patients selected for genomic analysis at diagnosis on the basis of their known response to therapy, mutation of cancer genes was associated with poor outcome.^71,73,74 Research on unselected cohorts of consecutively treated patients is required to establish the true predictive value of these mutations if present at diagnosis. However, it is known that some patients with cancer gene mutations at diagnosis can achieve optimal responses. For example, the most frequently mutated gene at diagnosis of CML is ASXL1⁷² and some patients with mutated ASXL1 can rapidly achieve MMR.⁷¹ The long-term outcome for patients with additional mutations at diagnosis is unknown. However, a recent small study suggested that somatic mutation of epigenetic modifier genes within the leukemic clone at CML diagnosis may impact the chance of TFR.⁷⁷

Importance of treatment adherence

Adherence to therapy is known to be a critical factor for achieving and maintaining response.^78-80 One study found that as few as 14% of patients were completely adherent in taking all TKI doses and a third of patients were classified as non-adherent.⁷⁸ In a UK population, 26% of patients had less than 90% adherence.⁷⁹ Factors associated with better adherence were older age, male sex, appropriate management of side effects, taking only one tablet per day, and feeling well informed and supported by the clinician.^81,82 Patients were also less adherent when more than 2 years from diagnosis.⁸² Long-term, regular molecular monitoring can help to identify patients who are less adherent. Features of non-compliance include unexpected variations in BCR::ABL1 ratios (usually in patients who have previously achieved at least MMR) and plateauing of response after prior steady declines in transcript levels.

Enhanced detection of BCR::ABL1 kinase domain mutations

BCR::ABL1 kinase domain mutations are the best-recognized mechanism of acquired resistance and signal treatment failure.^20,21 Early detection can allow timely therapeutic intervention. Since each leukemic cell has one copy of the BCR::ABL1 gene fusion and one copy of normal ABL1, the mutated allele can be specifically isolated by positioning PCR amplification primers within BCR, just before the fusion junction, and in ABL1, immediately after the kinase domain sequence. This allows exquisite sensitivity to detect mutations in patients with MRD and kinase domain mutations can be detected using Sanger sequencing in patients with BCR::ABL1 ratios <0.1%, prior to relapse.³⁰ However, the sensitivity of Sanger sequencing is limited to 10-20%.^83,84 The relevance of sensitive BCR::ABL1 mutation detection using next-generation sequencing has been demonstrated for patients with a nonoptimal molecular response, when the actionable threshold of mutant detection was 3%.^83,84 In a prospective study that compared Sanger sequencing and next-generation sequencing, low-level TKI-resistant mutants detectable using next-generation sequencing invariably expanded over time if the patient was not switched to an appropriate alternate TKI.⁸⁴ Early detection of mutants at levels as low as 3% could warrant treatment intervention to curtail clonal expansion of the resistant clone and loss of response. The clinical relevance of TKI-resistant mutations using next-generation sequencing that are below the current threshold of 3% has not been established. Furthermore, the introduction of next-generation sequencing for routine clinical monitoring requires appropriate validation according to international standards for diagnostic testing.⁸³ Importantly, there is no clinical need for BCR::ABL1 mutation analysis at the time of diagnosis in CP using Sanger sequencing or sensitive next-generation sequencing.

Figure 3.Next-generation sequencing for the detection of various mutation types. New technology can detect BCR::ABL1 transcripts and kinase domain mutations, plus other clinically relevant variants. Some of these techniques enhance the detection of low level variants. (A) RNA-sequencing detection of e14a2 BCR::ABL1 fusion transcripts. Sequence reads are mapped to the hg19 reference genome and visualized in the Integrative Genomics Viewer. ‘View mate region in split screen’ is selected. The multicolored region in the left panel at the junction of ABL1 exon 2 indicates that the reads are derived from a different genomic location. The dark green color specifies that the reads are derived from chromosome 22. The multicolored region in the right panel at the junction of BCR exon 14 and the light green solid color indicate the reads are derived from chromosome 9 and map to ABL1. (B) Whole exome sequencing or targeted gene sequencing can detect clinically relevant variants in cancer-related genes at the time of diagnosis of chronic myeloid leukemia or at drug resistance. In this case a common ASXL1 23 base pair deletion was detected. (C) Next-generation sequencing of the BCR::ABL1 kinase domain provides enhanced sensitivity. A low level T315I mutation was below the level of detection using Sanger sequencing but was retrospectively detected at 2.6% in a drug-resistant patient prior to commencing dasatinib (top panel). The dasatinib-resistant mutant rapidly expanded to 23% after commencing dasatinib (bottom panel). (D) Whole genome sequencing is not suitable for the detection of residual disease but can simultaneously detect multiple different mutation types, with the exception of fusion transcripts. A circos plot provides a snapshot of rearrangements and other variants. In this case, the standard 9;22 translocation was detected and indicated by the arc between chromosomes 9 and 22, plus other rearrangements. Chromosomes are indicated in the outermost track. Circos plot courtesy of Dr Philippi May, Imperial College Healthcare NHS Trust, in conjunction with the Genomics England 100,000 Genomes project. CML: chronic myeloid leukemia.

Treatment decisions for patients with long-term measurable residual disease

A major goal for many patients is TFR but not all patients achieve the strict criteria for a trial of stopping therapy. Most patients face life-long TKI therapy, which can cause debilitating side effects. With the introduction of increasingly potent TKI, clinicians are faced with therapy-related dilemmas for patients on treatment for many years. Should a patient who is tolerating TKI therapy and has minimal or no side-effects, switch to a more potent TKI to achieve a DMR in order to qualify for a trial of drug cessation? Should prior TKI resistance influence decisions? What are the long-term vascular risks for patients treated with potent TKI over many years and should this influence treatment decisions? How should patients be managed when BCR::ABL1 ratios remain relatively and stubbornly high without meeting the molecular criteria for treatment failure? Are these patients at risk of disease progression? We present theoretical cases in which these treatment dilemmas may arise and discuss the pros and cons of treatment options.

Case scenarios

Patient 1, a case of sudden blast phase

A 48-year-old male was diagnosed with CP CML: he had a high-risk EUTOS long-term survival (ELTS) score, expressed the e13a2 transcript, had no additional chromosome abnormalities and no mutations in cancer-related genes. He commenced imatinib 600 mg OD in the context of a clinical trial. Figure 4 shows the BCR::ABL1 ratios measured over time. The BCR::ABL1 ratio was 160% at diagnosis and 12% IS at 3 months. Early research suggested a BCR::ABL1 ratio >10% at 3 months was sufficient to identify patients destined to fare poorly, thereby allowing early treatment intervention.²⁷ However, subsequent studies highlighted the importance of the trend of initial BCR::ABL1 decline using multiple data points.^33,34 This strategy is recommended when interpreting the 3-month ratio.^20,21 There was a substantial BCR::ABL1 decline at 3 months for this patient although the response fell into the warning/possible TKI resistance categories. A good mantra would be to try to avoid making decisions on any single result and always confirm an unexpected result. It is also important to consider treatment compliance and discuss with the patient the presence of side effects or any other reason why they might have missed any medication.^20,21 Assuming good compliance, continuing the initial therapy would be a perfectly reasonable treatment decision. In this case the trial protocol mandated a rapid treatment switch to nilotinib because of failure to achieve time-dependent molecular milestones at 3, 6 or 12 months Treatment was therefore changed to nilotinib 400 mg BD at 5 months and a 12-month BCR::ABL1 ratio compatible with CCyR was achieved. As can be seen from Figure 4, the MMR achieved at 2 years was not stable, perhaps again raising issues of compliance, which in turn may reflect the presence of troublesome side effects and/or problems associated with twice daily dosing. Consideration was given to performing allogeneic SCT at this time, and HLA-typing of the patient and siblings was requested. In the meantime treatment was changed to dasatinib. A stable MMR was maintained for a number of years but a DMR was never achieved.

A rapid rise in the BCR::ABL1 ratio occurred at 7 years. This triggered a kinase domain mutation analysis, which was negative. Prior studies had determined that the average rate of a BCR::ABL1 rise after stopping TKI corresponds to a BCR::ABL1 doubling time of 8-9 days.^38,85 The BCR::ABL1 doubling time for Patient 1 was 11 days, which is rapid and consistent with complete lack of kinase inhibition. This could indicate complete non-adherence to TKI therapy or could portend a more dangerous scenario for the patient: progression to BP. In a study of 12 CP patients with BCR::ABL1 <10% IS who relapsed into BP, the median BCR::ABL1 doubling time was 9 days.³⁸ In contrast, the BCR::ABL1 doubling time was significantly longer (median 48 days) for 30 patients who acquired BCR::ABL1 mutations but maintained CP.³⁸

Figure 4.A rare case of sudden blast phase. Measurement of BCR::ABL1 ratios using realtime quantitative polymerase chain reaction revealed that the patient never achieved a deep molecular response despite switching to second-generation tyrosine kinase inhibitors. The very rapid BCR::ABL1 rise at 7 years after diagnosis heralded blast phase. IS: International Scale; ELTS: EUTOS long-term survival score.

Issues of compliance were addressed and the patient denied missing his drugs. Unfortunately the patient had also lost complete hematologic response and subsequently progressed to BP within a month. He was treated with two courses of AML-like chemotherapy and achieved a second CP. He has recently undergone a sibling allogeneic SCT and his RT-qPCR confirms undetectable disease. Could we have predicted this tragic turn of events? Progression to advanced phase after the establishment of a durable MMR is unusual but has been described.^86,87 Although the achievement of MMR has been termed a ‘safe haven’, Claudiani and colleagues showed that attainment of DMR, sustained for at least 12 months, was associated with a remarkably low probability of losing MMR in the absence of other events such as a trial of treatment discontinuation, lack of compliance, or reduced drug dosing.⁸⁸ The mechanism of treatment failure in Patient 1 is unknown. Additional chromosome abnormalities were not detected but broader genomic analysis would likely identify mutated cancer-related genes at the time of BP.

Patient 2, a case of non-adherence and late acquisition of BCR::ABL1 mutations

A 28-year-old male was diagnosed in CP in 2003: he had a low risk ELTS score, expressed e14a2 and had no additional chromosome abnormalities. He was treated with imatinib 400 mg OD. Figure 5 shows his BCR::ABL1 ratios measured over time. The patient met the current criteria for treatment failure at 6 and 12 months, but maintained CP. A slow BCR::ABL1 rise occurred in the third year of imatinib treatment, when the doubling time was slow at 53 days. A BCR::ABL1 kinase domain mutation was not detected. The rise was attributed to non-adherence to therapy and the dynamics of the rise were consistent with intermittent imatinib dosing.³⁸ MMR was achieved at 6 years but was somewhat unstable, again attributed to non-compliance which was admitted by the patient. This did not seem to be related to any particular side-effect that might have been best treated by a switch of TKI. Unfortunately the patient experienced a significant BCR::ABL1 rise at year 8. At that time two imatinib-resistant BCR::ABL1 mutations were detected using Sanger sequencing: E275K and E459K. BCR::ABL1 mutations are mostly acquired within the first few years of first-line TKI therapy in resistant patients so this was a late occurrence. Whether the persistent, relatively high levels of BCR::ABL1 contributed to an environment conducive to DNA damage and the acquisition of mutations is unknown. Both of the mutations are sensitive to second-generation TKI and a switch to dasatinib reinstated and indeed deepened the response such that the patient is now in stable DMR.

Figure 5.A case of non-adherence and late acquisition of BCR::ABL1 kinase domain mutations. Increases in BCR::ABL1 ratio over the first 3 years after diagnosis for this patient were associated with non-adherence to therapy. A deep molecular response on imatinib was never attained and a slow BCR::ABL1 rise at 8 years was accompanied by the detection of BCR::ABL1 mutations and failure of imatinib. This case demonstrates the importance of regular and sustained molecular monitoring. The rise prompted BCR::ABL1 mutation analysis, which confirmed imatinib resistance rather than non-adherence and allowed timely therapeutic intervention. IS: International Scale; ELTS: EUTOS long-term survival score.

What next for this patient who has been on dasatinib for >10 years and maintained a DMR for 3 years? Does the patient qualify for a trial of treatment cessation? Could residual leukemic cells carry the BCR::ABL1 mutations and does this influence decisions regarding treatment cessation? BCR::ABL1 mutations can be selected and deselected.^89,90 In some cases the mutants persist at undetectable levels for many years and even reappear at molecular relapse upon treatment cessation.⁹¹ Some BCR::ABL1 mutants may have a proliferative advantage.⁹² Prior TKI resistance is among the current ELN exclusion criteria for a treatment-cessation trial.²⁰ However, recent versions of the NCCN guidelines for treatment cessation no longer exclude prior detection of BCR::ABL1 mutations in patients who maintained CP.²¹ Claudiani and colleagues assessed ten patients with previous BCR::ABL1 mutations who stopped TKI due to intolerance.⁹³ All had maintained MR4 for at least 1 year prior to stopping TKI (median 6.3 years) and the median duration of TKI therapy before stopping was 13 years. The molecular relapse-free survival for the ten patients was 50% with 1 to 4.7 years of follow-up. Two of the patients who maintained TFR had prior T315I mutations. The rate of TFR for the ten patients with prior resistance is consistent with the rate reported in clinical trials of TKI cessation that excluded patients with prior resistance.⁹⁴ The authors speculated that if a patient with a BCR::ABL1 mutation promptly receives an effective alternative TKI and a DMR is achieved and maintained, the adverse outcomes associated with BCR::ABL1 mutations can be overcome.⁹³

Patient 3, a case of early treatment failure

A 37-year-old female was diagnosed with CML in 2013 and received nilotinib as first-line therapy in a clinical trial. The transcript was e14a2/e13a2, the ELTS score was intermediate and there were no additional chromosome abnormalities at diagnosis. Exploratory genomic analysis at diagnosis revealed an ASXL1 nonsense mutation. Figure 6 shows the BCR::ABL1 ratios measured over time. The pa tient rapidly developed a nilotinib-resistant mutation, F359V, despite a good initial response. A switch to dasatinib was swiftly followed by the acquisition of the T315I mutation, which is resistant to imatinib and the second-generation TKI nilotinib, dasatinib and bosutinib. T315I was acquired in an independent clone, which was indicated by its clonal dominance and disappearance of the F359V clone. T315I is sensitive to ponatinib and the patient rapidly achieved and maintained a DMR after commencing treatment with ponatinib. The dose of ponatinib at the start of treatment was 45 mg OD and this was reduced to 30 mg OD within 1 month. Recent results from the OPTIC study, in which patients were randomized to one of three doses of ponatinib (45, 30 or 15 mg) and instructed to dose reduce to, or continue on, 15 mg, once the RT-qPCR fell below 1% IS, would suggest that she could now be safely reduced to 15 mg daily. Probably given the length of time she has been on 30 mg, she is not at high risk of arterial thrombotic events but minimizing the dose while maintaining response is a reasonable goal for all patients.⁹⁵ Monitoring general health is recommended and interventions should be made where necessary. Asciminib is a BCR::ABL1 inhibitor recently approved by the US Food and Drug Administration for patients in whom prior TKI therapy has failed.^96,97 Early data suggest that asciminib, at a higher dose of 200 mg twice daily, has efficacy against T315I and may be better tolerated than ponatinib. However, the follow-up was short. The other possibility is a trial of treatment discontinuation but as discussed above, data are sparse as to the safety of this approach in patients with kinase domain mutations, particularly T315I.

Figure 6.A case of early treatment failure. Despite an initial rapid response to first-line nilotinib the patient failed to benefit from nilotinib and subsequent dasatinib therapy, as indicated by the acquisition of successive BCR::ABL1 mutations. T315I detected at the second BCR::ABL1 rise prompted a switch to the third-generation inhibitor, ponatinib. The rapid achievement of a deep molecular response, sustained for many years, indicates that the T315I-mutant leukemic cells were highly sensitive to ponatinib. IS: International Scale; ELTS: EUTOS long-term survival score.

Biomarkers at diagnosis cannot predict the early acquisition of TKI-resistant BCR::ABL1 kinase domain mutations. Why did this patient acquire a resistant mutation within months of commencing treatment, whereas Patient 2 only acquired resistant mutations after 8 years? Patient 3 had an ASXL1 mutation at diagnosis, whereas the mutation status of Patient 2 at diagnosis was unknown. Ongoing genomic studies of cohorts of unselected patients may provide further evidence for enhanced risk stratification on the basis of a cancer gene mutation at diagnosis. Mutated ASXL1 is not only the most frequently detected mutation at diagnosis of CML, but is also among the most frequently observed in BP CML.⁷² In the largest study of genomic heterogeneity in BP CML, ASXL1 mutations were associated with a poorer outcome, even in this very poor risk setting.⁷⁶

Figure 7.A case of two failed atempts to achieve treatment-free remission.

The patient achieved the optimal milestone BCR::ABL1 ratios and sustained a deep molecular response for more than 5 years before imatinib discontinuation in an attempt to achieve treatment-free remission. Molecular relapse was rapid at both cessation attempts. It is not known why some patients are unsuccessful in multiple attempts to sustain treatment-free remission and the reasons for molecular relapse could be multifactorial. Life-long tyrosine kinase inhibitor therapy may be required for this patient. Regular molecular monitoring could be critical to monitor for potential episodes over time of non-adherence to therapy. A rapid BCR::ABL1 rise associated with non-adherence could potentially lead to loss of complete hematologic response, unless detected promptly by the clinician through molecular monitoring. IS: International Scale; ELTS: EUTOS long-term survival score.

Patient 4, a case of treatment-free remission attempts

A 53-year-old male was diagnosed in CP in 2010 and commenced imatinib 400 mg OD. The transcript was e13a2, the ELTS score was low and there were no additional chromosome abnormalities at diagnosis. Exploratory genomic analysis at diagnosis revealed an ASXL1 frameshift mutation. Figure 7 shows the BCR::ABL1 ratios over time. All BCR::ABL1 optimal milestones were achieved. MR4.5 was maintained for 4.5 years before imatinib was discontinued for a trial of TFR. A rapid rise that commenced at 1 month after cessation prompted imatinib restart and MR4.5 was rapidly regained. Imatinib was ceased for a second attempt at TFR after a further 3.5 years of DMR, but relapse was again rapid.

The chance of TFR is approximately 50% for patients who attempt TFR. Longer treatment and DMR durations were associated with an increased probability of maintaining TFR at 6 months in the EURO-SKI study, which was the largest TKI cessation trial.⁹⁴ The optimal cut-offs were 5.8 years on therapy and 3.1 years of DMR. Patient 4 was on imatinib for 5 years before attempting TFR and the chances of success may have increased with longer time on imatinib. However, the EURO-SKI study determined that the duration of DMR was the most important factor affecting the probability of TFR. Patients with e13a2, as in this case, may have an inferior probability of TFR.^62,64

TFR is achievable after a second TKI cessation attempt for some patients.⁹⁸ The French RE-STIM study of 70 patients reported a TFR rate of 42% at 24 months after cessation. The relapse pattern at the first cessation attempt was the only factor significantly associated with TFR at the second attempt. Patients who relapsed after 3 months had a significantly higher rate of TFR at the second attempt: 72% versus 36% at 24 months. Patient 4 had a very rapid relapse at the first attempt. This patient had mutated ASXL1 at diagnosis which was not detectable in remission and a recent small study has found an association between mutations in epigenetic modifier genes at diagnosis and a lower rate of TFR.⁷⁷

What next for this patient? He is now 12 years after diagnosis and has received only imatinib to which he has responded deeply and durably. He has tolerated the imatinib well and could remain on the drug life-long. After two unsuccessful attempts at treatment discontinuation of imatinib it seems unlikely that further treatment cessation will achieve a better result. If TFR is an important goal for this patient then re-starting treatment using a more potent TKI would be an entirely reasonable approach but the chance of successful discontinuation must be balanced against the increased risk of side effects with a new drug. This is an excellent example of the need for honest and transparent dialogue between patient and physician.

Conclusion

In 2022 it is virtually impossible to imagine managing any patient with CML without accurate molecular monitoring. The technology accurately identifies patients who are responding well and who might be future candidates for treatment discontinuation. Conversely patients with primary and secondary resistance can be recognized promptly and treatment switched in an attempt to induce response and prolong survival. If the change in therapy is unsuccessful the patient can be referred for allogeneic SCT while still in CP and thereby maximize their chance of a good outcome. But accurate monitoring can also highlight issues of compliance, which can then be addressed and the patient supported to adhere to treatment and deepen their response. The methodology continues to evolve and can be adapted to suit most clinical situations and resources. There is little doubt that the efficacy of the TKI in CML has been complemented by the ability to accurately measure residual disease and modify treatment accordingly to optimize outcome.

Footnotes

• Received May 30, 2022
• Accepted September 9, 2022

Correspondence

S. Branford susan.branford@sa.gov.au

Disclosures

SB is a member of advisory boards for Qiagen, Novartis and Cepheid; has received honoraria from Qiagen, Novartis, Bristol-Myers Squibb and Cepheid, and has received research support from Novartis and Cepheid. JFA is a member of advisory boards for Incyte and Novartis; has received honoraria from Incyte, Novartis and Pfizer; and has received research support from Incyte, Novartis and Pfizer.

Contributions

SB conducted the literature review, prepared the figures, wrote and reviewed the manuscript. JFA conducted the literature review, designed the original layout, wrote and reviewed the manuscript. The authors approved the final version of the manuscript.