Should Polymerase Chain Reaction Analysis to Detect Minimal Residual Disease in Patients With Chronic Myelogenous Leukemia Be Used in Clinical Decision Making?

CHRONIC MYELOGENOUS leukemia (CML) is a unique myeloproliferative disorder usually associated with a distinctive cytogenetic abnormality—the Philadelphia chromosome (Ph)—that leads to leukemogenesis.1 Ph is a shortened chromosome 22 resulting from a reciprocal translocation of the long arms of chromosomes 9 and 22 that transposes the 3′ segment of theABL gene from 9q34 to the 5′ segment of the BCR gene on 22q11. The resulting BCR-ABL gene is transcribed into a chimeric mRNA and then translated into fusion proteins of varying size (p190^bcr-abl, p210^bcr-abl, and p230^bcr-abl) according to the breakpoint location of the genes involved.2 Ph is observed by cytogenetic analysis in more than 90% of patients with CML, almost all of whom express the chimeric protein p210 ^bcr-abl, but much less frequently in patients with other leukemias in which expression of the smaller fusion protein p190^bcr-abl occurs more frequently.2,3 Preclinical in vitro and in vivo models of tumor development suggest a crucial role for Ph translocation-associated molecular events to initiate and perpetuate CML disease.4 5 Thus, the fusion products of theBCR-ABL gene are thought to play a central role in CML as mediators of myeloid proliferation and transformation.

The availability of information from cytogenetic and molecular studies lent itself to a more precise assessment of low volumes of residual disease in CML. It has become apparent that conventional therapy with hydroxyurea and busulfan produces hematologic remissions (ie, normalization of peripheral blood counts and absence of signs and symptoms of disease) in 50% to 80% of patients but only rare and short-lived cytogenetic remissions (ie, suppression of the Ph⁺ clone). As a result, patients will invariably progress to a blastic phase and die from its complications.6 

Two treatment modalities introduced in the 1980s have become especially useful tools for inducing cytogenetic remissions in CML patients. Allogeneic matched sibling stem cell transplantation (SCT) in CML has a curative potential arising from the elimination of Ph⁺hematopoietic progenitors and allows disease-free survival (DFS) rates of 30% to 70%.7,8 Interferon-α (IFN-α) was identified as the first biological agent capable of inducing cytogenetic remission in patients with CML.9 Over the past decade, numerous studies of IFN-α as treatment for CML have shown complete hematologic remissions (CHR) in 80% to 90% of patients and major cytogenetic remissions (ie, suppression of Ph⁺ cells below 35% in metaphase spreads) in 30% to 40%, some of which are durable.10,11 The significance of achieving a cytogenetic response as an independent variable for survival prolongation was confirmed by multivariate analysis, further supporting the therapeutic efforts aimed at suppressing Ph⁺ clones.10 

Therefore, detection of the Ph translocation is not only a diagnostic tool but also at the core of a strategy to assess the response of individual patients to therapeutic interventions, namely SCT and IFN-α, and evaluate treatment efficacy by monitoring residual disease. Indeed, arguments in favor of monitoring residual disease are, at first sight, compelling. Typically, cytogenetic relapse precedes hematological relapse and, provided disease recurrence is detected early enough, effective salvage therapy (such as donor lymphocyte infusions) can produce a response in up to 80% of patients whose disease relapses after allogeneic SCT.12 For IFN-α–treated patients who attain a complete cytogenetic remission, the duration of treatment can be individually tailored according to the presence or absence of the BCR-ABL transcription product in blood or bone marrow samples.

Cytogenetic analysis is essential to diagnose and follow the course of CML and for identifying karyotypic changes in addition to the Ph chromosome. However, its sensitivity is low (1% to 5%) due to the limited number of adequate metaphases examined (usually 20 to 25).13 By the time disease is detectable by cytogenetic studies, clinical relapse may be inevitable. It may also miss some cases of BCR-ABL⁺ CML in which Ph is not detected.14 Therefore, more sensitive methods for the molecular diagnosis and follow-up of residual disease are now being used. Fluorescence in situ hybridization (FISH) and its variants, interphase (i-FISH) and hypermetaphase FISH (h-FISH), allow analysis of a larger number of cells (>500) in a timely and efficient manner, with the added benefit that some (i-FISH) can be applied to nondividing cells, a fraction of cells with a low proliferative rate (as may be the case in residual disease).15 However, in many centers, polymerase chain reaction (PCR) assays are becoming the techniques of choice for detecting residual disease on a molecular level. PCR allows identification of a single Ph-bearing cell among 10⁴ to 10⁸ normal cells, a sensitivity unparalleled by any other available method.16 

In applying PCR techniques clinically, what might be learned about the kinetics of the disappearance and reappearance of residual disease and the significance of detecting residual disease in CML patients treated with either allogeneic SCT or IFN-α?

Many investigators have reported, with varying results, on the detection of minimal residual disease in patients with CML who have undergone allogeneic SCT (Table1).17-22 Using nested PCR, Roth et al20 analyzed 64 CML patients after allogeneic SCT and detected BCR-ABL transcripts at one time point in 37 of the patients. Of those 37, 13 eventually had a disease relapse, with a median time to relapse of 5 months. No relapses were observed in patients with negative PCR results. Roth et al thus concluded that nested PCR could define subgroups of patients in apparent clinical remission but with an increased risk of disease recurrence. In contrast, Miyamura et al,19 also using nested PCR, detected no association between PCR positivity and subsequent relapse in their series of 64 patients with CML in remission after allogeneic SCT. In five cases, the persistence of detectable BCR-ABL transcript for up to 2 years postremission did not result in disease recurrence. Hughes et al,17 on the basis of nested PCR results in 37 CML patients in remission after allogeneic SCT, concluded that PCR positivity within 6 months after transplantation did not predict a worse outcome, whereas PCR positivity later than 6 months after transplantation did. Radich et al18 presented a comprehensive multivariate analysis of 346 patients who had undergone allogeneic transplants for CML and then been analyzed by PCR for the presence of BCR-ABL transcripts. The multivariate analysis identified PCR-positivity at 6 to 12 months posttransplantation as one independent variable for influencing subsequent relapse. The significance of the presence of BCR-ABL transcripts in predicting disease recurrence was, however, lost in patients who tested positive more than 36 months post-SCT. A different approach was chosen by Lin et al,22 who used a competitive PCR method to quantify BCR-ABL transcripts rather than analyzing qualitative PCR data. In 98 patients with CML post-SCT, the probability of relapse was significantly increased in patients with higher versus lower levels of BCR-ABL. Gaiger et al21 reported that only serial PCR measurements identified patients at high risk (ie,BCR-ABL positivity in serial samples) and low risk of relapse (ie, transient positivity). Different techniques, various levels of PCR sensitivity, and short follow-up durations make it difficult to interpret these studies.

The discordance among these studies raises two key questions: (1) Why are some results so discordant and how can they be interpreted? (2) Is persistent residual disease predictive of relapse?

First, data from molecular studies, especially those involving PCR, should be appraised critically. PCR is a powerful tool, but has a number of shortcomings that can lead to both false-negative and false-positive results. Technical pitfalls such as sample contamination, inadequate sample volume, inappropriate sample source, loss of sensitivity of junctional probes, and degradation of target molecules are a major cause of inaccurate results. Furthermore, shifts in disease markers as reflected in the findings of oligoclonality, subclone formation, clonal evolution, and incomplete target sequence rearrangements in the neoplastic cell population must be considered if one intends to correlate PCR negativity with absence of disease.13 In CML in particular, attention should be paid to rare molecular fusions (such as e6a2, e19a2, fusions lacking exon a2) and evolution of subclones expressing different fusion products than the original clone.3,23-25 This biologic diversity may give rise to false-negative results and pose problems in diagnosis and monitoring of residual disease in CML. A major drawback of current PCR studies is the lack of quantification of PCR data. A mere positive or negative PCR result is not enough information to draw meaningful conclusions. Given the degree of variability in the sensitivity of the PCR assays used in the studies cited (Table 1), it is difficult to interpret and compare results reported by different investigators. What has also become apparent from these studies is the shift in the population of patients with the increasing sensitivity of PCR assays. Only a few patients are left that qualify as PCR⁻, thus rendering PCR “too sensitive” to serve a prognostic purpose. But even if quantitative PCR is used and the amount of residual disease monitored, can data from such studies be informative? If a threshold of residual disease above which a patient is likely to relapse or, conversely, below which remission is sustained is found, then residual disease monitoring using quantitative PCR may be useful. Hochhaus et al26 compared cytogenetic responses to treatment with IFN-α to molecular responses and defined cut-off points for the ratio of BCR-ABL to ABL in four response groups. However, the cut-off points do not establish a threshold for relapse versus remission, and no other study to date has provided convincing evidence of one. In fact, the behavior of residual disease may be just as individual as the patients treated, with different kinetics, distinct host-disease interactions, and varying responses to diverse treatment regimens.

Taking these considerations into account, what conclusions can be drawn from identifying persistent residual disease by a positive PCR reaction? The BCR-ABL rearrangement may be neither the only nor the first molecular abnormality pathogenetically linked to the leukemic process in CML, even though BCR-ABL may be considered a “window” though which antecedent clonal cells must progress to result in CML. However, secondary or epigenetic events may be required for full leukemogenesis. Fialkow et al27 and later Raskind et al28 were the first to show, by using glucose-6-phosphate dehydrogenase as a marker for clonality, that some hematopoietic cells of the neoplastic clone were Ph⁻. PCR primers can only detect a sequence that one is looking for, and cannot be used as a screening tool for other unknown but nevertheless equally important sequences. By the same token, it is well documented that patients can express the BCR-ABL transcripts and stay in clinical remission for many years.29 Are the residual cells detected in these patients indeed leukemic or just “BCR-ABL⁺” and lacking some major attributes of truly leukemic cells? Using a highly sensitive reverse transcriptase (RT)-PCR assay capable of detecting one BCR-ABL mRNA-expressing cell among 10⁸ white blood cells, Biernaux et al30 found BCR-ABL transcripts in healthy individuals in an age-dependent manner. The transcripts were detected most often in adults (22 of 73), much less often in children (1 of 22), and not at all in umbilical cord blood (0 of 22). This observation has been confirmed by Bose et al,31 who found p210 and p190BCR-ABL transcripts in 4 of 16 and 11 of 16 normal individuals, respectively, using a similar RT-PCR approach. Although further validation of these findings is needed, they indicate the presence of leukemia-specific fusion genes in some hematopoietic cells without the accompanying clinical syndrome of leukemia. Such genes may constitute abnormalities that by themselves are not able to generate the leukemic phenotype without the interaction with as yet unidentified leukemogenic events. Alternatively, leukemic fusion genes may be expressed in hematopoietic cells that have entered an apoptotic pathway before acquiring a characteristic leukemic karyotype that may already have lost its relevance. Similarly, clinically nonrelevant clonal disease has been observed in otherwise cured tumors.32 33 Two questions then follow: What does a positive PCR reaction mean in a patient who is in clinical and cytogenetic remission? Does it justify a therapeutic intervention using chemotherapy or allogeneic SCT?

Finally, consideration should be given to some important host factors.BCR-ABL⁺ cells in patients with CML who maintain long-term remission may be dormant. A tumor is considered dormant if the malignant cells present in an organism are kept under growth control by certain mechanisms for a prolonged period but retain their neoplastic potential.34 

For example, in patients with CML who are in cytogenetic remission after treatment with IFN-α, PCR techniques can still commonly identify BCR-ABL transcripts up to a median of 22+ months postremission (Table 2).35,36Hochhaus et al37 measured levels of BCR-ABLexpression in 20 such patients using a quantitative PCR method. In all of them, a median of 750 transcripts/μg RNA was found. They suggested that treatment with IFN-α was not capable of eradicating residual disease, thus establishing an impediment to a cure with IFN-α therapy, unlike with allogeneic SCT. Recently Kurzrock et al38 showed that patients in complete cytogenetic response after IFN-α treatment may become PCR⁻ forBCR-ABL if followed long enough. In particular, 10 of 18 patients in complete cytogenetic response tested negative forBCR-ABL by PCR; the median duration of cytogenetic responses was longer in patients who were PCR⁻ than in those who were PCR⁺ (42 v 12 months; P < .01). However, the association of PCR negativity with long-term event-free survival was not clear because dormant progenitor cells below the threshold of PCR detection may still have been present. Using RT-PCR, Talpaz et al39 analyzed seven patients who attained complete cytogenetic remission after IFN-α treatment and who were also PCR⁻ for BCR-ABL, and found that myeloid and erythroid colonies from blood and marrow samples in these patients still expressed BCR-ABL transcripts. In addition, one patient in this study group has been in a complete cytogenetic remission for 3.5 years without relapse. Furthermore, no cytogenetic relapse was observed in two patients with positive colonies in the subsequent 4.5 years of a maintained remission after this publication (M. Talpaz, personal communication, December 1998). These results were confirmed by a study by Pasternak and Pasternak,40 who showed the persistence of BCR-ABL mRNA-expressing cells in Dexter-type long-term cultures derived from bone marrow and blood samples from CML patients in cytogenetic remission.

That the immune system plays an important role in keeping the malignant cell population in CML dormant is confirmed by several other observations: (1) 60% to 80% of patients whose disease relapses after allogeneic SCT attain cytogenetic remissions with donor lymphocyte infusions12; (2) there is a positive correlation between the presence of graft-versus-host disease and a reduced risk of relapse after SCT and, conversely, an increased risk of relapse after the transplantation of T-cell–depleted donor marrows41; (3) there is a positive correlation between cytogenetic response and grade of IFN-α–associated autoimmune phenomena42; and (4)BCR-ABL transcripts can be present in healthy individuals in the absence of CML.30 31 

As the findings discussed demonstrate, sensitive techniques such as PCR can detect evidence of residual disease in many patients with CML who have achieved complete cytogenetic responses with either allogeneic SCT or IFN-α therapy. A positive PCR reaction does not always equal relapse, and a negative PCR reaction does not equal cure. Therefore, “cure” should be understood as a functional process (“functional cure”) rather than the absence of all evidence of disease (“molecular cure”), which is probably not possible, or even relevant, in any case. This concept has important implications in understanding residual disease, its kinetics, tumor dormancy, and the role of immunomodulation in residual disease. With increasing sensitivity of PCR assays, most patients will test positive, making qualitative PCR results less useful and raising the issue of a quantitative threshold of residual disease as prognostic marker. However, no such threshold has been established in the studies to date that would be clinically useful.

PCR has provided us with a wealth of exciting and interesting data that has already contributed significantly to our understanding of residual disease in CML and other leukemias. However, to conclude that “molecular cures” are the only possible pathways to long-term event-free survival is premature. The PCR technique, although considered a valid clinical testing procedure, should be used cautiously as a laboratory test until sufficient data are available to show that it meets acceptable criteria of sensitivity, specificity, and positive and negative predictive values. How new PCR technologies such as “real-time” PCR quantification will solve these concerns and become a reliable tool for the clinician merits further investigation. Until then, clinicians should exercise caution in basing clinical decision making on such studies, given the significant morbidity and mortality associated with aggressive therapeutic interventions aimed at molecular disease eradication in patients who might just do as well without.