research paper
Clinical significance of minimal residual disease at day 15 and at the end of therapy in childhood acute lymphoblastic leukaemia
Rosemary Sutton,1 Nicola C. Venn,1 Jonathan Tolisano,1 Anita Y. Bahar,1 Jodie E. Giles,1 Lesley J. Ashton,1 Lochie Teague,2 Gemma Rigutto,3 Keith Waters,3 Glenn M. Marshall,1,4 Michelle Haber,1 Murray D. Norris1 and the Australian and New Zealand Children’s Oncology Group 1
Children’s Cancer Institute Australia for Medical
Research, University of NSW, Sydney, NSW, Australia, 2Starship Hospital Auckland, Auckland, New Zealand, 3Royal Children’s Hospital, Melbourne, Vic., and 4Centre for Children’s Cancer and Blood Disorders, Sydney Children’s Hospital, Sydney, NSW, Australia
Received 23 March 2009; accepted for publication 15 April 2009 Correspondence: Dr Rosemary Sutton, Molecular Diagnostics, Children’s Cancer
Summary Detection of minimal residual disease (MRD) after induction and consolidation therapy is highly predictive of outcome for childhood acute lymphoblastic leukaemia (ALL) and is used to identify patients at high risk of relapse in several current clinical trials. To evaluate the prognostic significance of MRD at other treatment phases, MRD was measured by real-time quantitative polymerase chain reaction on a selected group of 108 patients enrolled on the Australian and New Zealand Children’s Cancer Study Group Study VII including 36 patients with a bone marrow or central nervous system relapse and 72 matched patients in first remission. MRD was prognostic of outcome at all five treatment phases tested: at day 15 (MRD ‡ 5 · 10)2, log rank P < 0Æ0001), day 35 (‡1 · 10)2, P = 0Æ0001), 4 months (‡5 · 10)4, P < 0Æ0001), 12 months (MRD ‡ 1 · 10)4, P = 0Æ006) and 24 months (MRD ‡ 1 · 10)4, P < 0Æ0001). Day 15 was the best early MRD time-point to differentiate between patients with high, intermediate and low risk of relapse. MRD testing at 12 and particularly at 24 months, detected molecular relapses in some patients up to 6 months before clinical relapse. This raised the question of whether a strategy of late monitoring and salvage therapy will improve outcome.
Institute Australia, PO Box 81, Randwick 2031, Sydney, NSW, Australia. E-mail:
[email protected]
Keywords: minimal residual disease, real-time quantitative PCR, ANZCHOG clinical trial, acute lymphoblastic leukaemia, relapse.
Numerous studies have demonstrated that the detection of minimal residual disease (MRD) in the first 3 months of therapy has a high prognostic value in childhood acute lymphoblastic leukaemia (ALL) using either polymerase chain reaction (PCR)-based methodologies or quantitative flow cytometry (Campana, 2008). Consequently, most current clinical trials for paediatric ALL include MRD measurement of early response to therapy to partly determine later treatment (Flohr et al, 2008). In contrast, the prognostic value of MRD at later phases of treatment for paediatric ALL has been less extensively analysed and the few studies that have been carried out measured MRD with either older PCR techniques (van Dongen et al, 1998) or flow cytometry (Coustan-Smith et al, 1998, 2000, 2002; Dworzak et al, 2002). Our own previous study of MRD later in therapy in patients with childhood ALL enrolled on Australian and New Zealand Children’s Cancer Study Group (ANZCCSG) Study VI was performed using a non-quantitative First published online 3 June 2009 doi:10.1111/j.1365-2141.2009.07744.x
nested-PCR technique with patient-specific primers (Marshall et al, 2003). This study clearly showed that the presence of MRD at 12 and 24 months were both predictive of outcome. Furthermore, analysis of MRD at day 35 measured by real-time quantitative (RQ)-PCR in these patients was also highly predictive of their outcome, and in a multivariate analysis including the three MRD time points, MRD at day 35 and 24 months were independently predictive of outcome. This finding led to the hypothesis that an additional group of patients at risk of relapse could be identified by MRD positivity at the end of therapy. However, the unusually high relapse rate in this clinical trial made the generality of this finding unclear. One factor that has contributed to the paucity of information available to independently validate MRD measurement at the end of therapy is that measuring MRD at this time point is difficult. The sensitivities required (0Æ01–0Æ001% cells) are usually below detection limits of quantitative flow cytometric analysis of leukaemic populations and close to the limits for
ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 146, 292–299
Minimal Residual Disease in Childhood ALL RQ-PCR. For instance, Coustan-Smith et al (1998, 2000, 2002) reported negative MRD at the end of therapy in all 65 patients tested. End of therapy MRD positivity is usually rare and there are challenges in distinguishing low ‘background’ levels of signal from true MRD positivity by either method (van der Velden et al, 2008). The more sensitive PCR techniques involving semi-quantitative PCR, dot blots or non-quantitative nested PCR used in earlier MRD studies were very labour intensive. Although MRD measured by RQ-PCR requires less work, a significant number of the patient specific assays are associated with some background amplification (usually 0Æ001% or 10)5 or less). This issue was addressed by the recent development of precise guidelines for both performance and consistent interpretation of RQ-PCR data by the European Study Group on MRD in ALL (ESG-MRD-ALL) (van der Velden et al, 2007). Based on these guidelines, we therefore evaluated the prognostic significance of MRD measured by RQ-PCR at five time points including end of therapy using 388 samples collected from 108 childhood ALL patients enrolled on ANZCCSG Study VII.
Materials and methods Patients All patients (n = 416) were diagnosed with ALL between 1998 and 2002 and enrolled on ANZCCSG Study VII at one of 10 children’s hospitals in Australia or New Zealand. They were stratified at diagnosis into either a standard or high risk group using similar criteria to National Cancer Institute/Rome risk groups. The high risk group included patients with any of the following characteristics at diagnosis: an age of ‡10 years; white cell count (WCC) > 50 · 109/l; presence of translocations t(9,22), t(4;11) or t(1;19); evidence of central nervous system (CNS) or testicular disease; or unresponsive marrow (M3) at day 15. Patients lacking these high risk criteria features were stratified into standard risk. The trial was not randomised. This study was approved by the Human Research Ethics Committee of the South Eastern Sydney Area Health Service. The chemotherapy protocol was a modification of the BFM95 protocol. All patients began with a 5 week induction with four drugs (vincristine, prednisone, L-asparaginase, daunorubicin) and intrathecal methotrexate, followed by a 6 week CNS consolidation phase with intrathecal methotrexate, cytarabine, cyclophosphamide and vincristine. At 11 weeks post-diagnosis, standard risk patients received an 8 week highdose methotrexate intensification phase, followed by an 8 week reinduction/reconsolidation phase, then maintenance therapy with vincristine, dexamethasone, 6-mercaptopurine and methotrexate, to a total of 2 years of chemotherapy. High risk patients instead received 18 weeks of high-dose cytarabine, etoposide and high-dose methotrexate, before also proceeding to the same reinduction/reconsolidation and maintenance therapy. Prophylactic cranial irradiation (18 Gy) was given to all patients categorized as high risk if any of the following
features were present: T cell immunophenotype and a WCC > 50Æ0 · 109/l, a WCC > 50Æ0 · 109/l and an age of ‡10 years, a WCC > 100Æ0 · 109/l, or CNS disease at diagnosis.
Study design Unlike most previous MRD studies in which a variety of risk factors were evaluated in a random selection of patients, this study used a case control design similar to Biondi et al (2000) so that we could stringently assess the value of MRD, particularly at later time points in a small group of patients. Because patient factors, such as older age group, higher WCC and T immunophenotype, were associated with slower treatment response (Fronkova et al, 2008), we controlled for these variables and risk/treatment group. This case control study was nested in the Study VII cohort. The ‘cases’ are all of the patients who relapsed either in the bone marrow or central nervous system (CNS) (n = 36) and each case was matched to two controls who remain in CR1 (n = 72) and were in the same treatment group, had the same immunophenotype and gender and similar age and WCC at diagnosis. The 19 Study VII patients who relapsed, but who could not be included in this study, included nine without a diagnostic sample for MRD; three with isolated testicular relapses; six who had failed to achieve remission; and an older male patient with an isolated CNS relapse who lacked suitable matches. The median time to relapse was 2Æ1 years ranging from 0Æ3 to 3Æ7 years, and all T-ALL relapses occurred on therapy.
Samples This study involved MRD analysis of bone marrow samples corresponding to day 15, end of induction (day 35), end of intensification for standard risk patients (4 months) or after the first set of the double intensification blocks in high risk patients (4 months), during maintenance (12 months) and at the end of therapy (24 months). All relapse and post-relapse samples were excluded from this analysis, which restricted the numbers of 12- and 24-month samples available. In most cases DNA samples were obtained from whole frozen marrow aspirates and isolated using Nucleobond (Machery-Nagel, Duren, Germany) column purification. DNA was checked for quality using RQ-PCR for the ACTB gene (Choi et al, 2007). Bone marrow aspirates from diagnosis were unavailable for 10 patients, so a bone marrow smear dried onto slides was substituted for one patient and peripheral blood samples were used for the remaining nine patients. Purified mononuclear bone marrow cells were used instead of whole samples for 14 patients from two participating hospitals. Of the expected 540 remission bone marrow samples for MRD analysis, 23% were either not collected or had inadequate yields of DNA (especially at day 15). In addition, 18 patients relapsed on therapy, reducing the 4 month samples by one, 12 month samples by five and 24 month samples by 18 in the relapse group. MRD analyses were performed on all the first remission
ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 146, 292–299
293
R. Sutton et al samples available by research assistants who were not aware of the patients’ relapse-free survival status or outcome.
Molecular MRD analysis Minimal residual disease markers were identified by PCR amplification of IGH and IGK rearrangements; the STIL (SILTAL) deletion; and TRG, TRD, TRB and TRD-TRA gene rearrangements (Pongers-Willemse et al, 1999; van Dongen et al, 2003; Szczepanski et al, 2004); followed by heteroduplex analysis and direct sequencing of clonal products (Choi et al, 2007). Pairs of patient-specific primers were designed to enable RQ-PCR detection of the unique junctional regions identified using the National Center for Biotechnology Information (NCBI) IgBlast (http://www.ncbi.nlm.nih.gov/ igblast/) database and hardcopies circulated by the ESGMRD-ALL. RQ-PCR to measure one MRD marker for each patient was performed on an ABI PRISM 7700 platform (Applied Biosystems, Foster City, CA, USA) with the specific primers and an appropriate hydrolysis probe (Kwan et al, 2000; van der Velden et al, 2003; Sutton et al, 2008).
Fig 1. Event-free survival for all patients (n = 416) enrolled on ANZCCSG Study VII, according to their assigned treatment group.
Data analysis All MRD RQ-PCR data were analysed according to the guidelines established by the ESG-MRD-ALL (van der Velden et al, 2007). Standard curves met the minimum standards of a 0Æ98 correlation coefficient and a slope between 3Æ1 and 3Æ9. MRD was scored positive or negative according to the definition established for protocols in which therapy intensification is intended. Thus, to prevent false positive MRD results, samples were considered positive only when the cycle threshold (Ct) was at least 3Æ0 Ct earlier than all the normal peripheral blood and bone marrow control DNA samples. Statistical analyses were performed using statview (SAS Institute, Inc, Cary, NC, USA) and stata version 8.2 (StataCorp, College Station, TX, USA); P values < 0Æ05 were considered significant.
Results Selection of MRD study group The patients studied were all enrolled on the ANZCCSG Study VII clinical trial, which predated the use of MRD stratification and which had an overall event-free survival (EFS) of 83% at 5 years. Patients in the standard risk treatment group had an EFS of 90% compared with an EFS of 74% for the high risk patients (Fig 1). A comparison is shown in Table I of the assigned risk group and clinical features at diagnosis for all patients enrolled in ANZCCSG Study VII (n = 416) and for subsets of patients who relapsed and patients who were MRD tested. Due to the case control study design, one in three of the MRD tested patients had relapsed (compared to 13% of the whole cohort) and consequently, there were more high risk and T-ALL patients. The relapse cases available for MRD testing in this study were representative of relapses from the overall clinical trial, and the patients still in first remission (CR1) tested for MRD were also adequately matched to the relapse cases (Table I).
Table I. Patient characteristics at diagnosis for all Study VII patients, all patients who relapsed and the patients selected for MRD testing.
High risk (n) Standard risk (n) T-ALL (n) Precursor B-ALL (n) Males (n) Females (n) Median age, years (range) Median WCC, ·109/l (range)
All study VII (n = 416)
All relapses (n = 55)
MRD cohort (n = 108)
MRD relapse (n = 36)
MRD CR1 (n = 72)
168 40%* 248 60% 47 11%* 357 86% 237 57% 179 43% 4Æ7 (1Æ1–17Æ8) 9 (0–1000)
34 62% 21 38% 11 20% 44 80% 35 64% 20 36% 6Æ3 (1Æ1–15Æ8) 28 (1Æ3–625)
69 64% 39 36% 24 22% 84 78% 65 60% 43 40% 7Æ2 (1Æ2–15Æ9) 19 (0Æ4–887)
23 64% 13 36% 8 22% 28 78% 22 61% 14 39% 7Æ7 (1Æ2–15Æ8) 36 (2Æ7–628)
46 64% 26 36% 16 22% 56 78% 43 60% 29 40% 6Æ7 (1Æ3–15Æ9) 15 (0Æ4–887)
*The patient subgroups were significantly different from the whole cohort based on numbers of high risk and T-ALL patients P < 0Æ05 (Chi test) but not gender (Chi-test) or age or white cell count (WCC) (P > 0Æ05, Wilcoxon signed-rank test). There were no significant differences between patient subgroups.
294
ª 2009 Blackwell Publishing Ltd, British Journal of Haematology, 146, 292–299
Minimal Residual Disease in Childhood ALL threshold, selected for each time point on the basis of log rank analysis, progressively dropped during early therapy from 5 · 10)2 at day 15, to 1 · 10)2 at day 35, to 5 · 10)4 at 4 months. These relatively high thresholds provided high specificity at the three early time-points with all the patients exhibiting an MRD level above the relevant threshold going onto relapse (Fig 3). The MRD levels at the three early time-points showed the progressive disappearance of the residual disease and MRD < 5 · 10)2 at day 15, MRD < 1 · 10)2 at day 35 and MRD < 5 · 10)4 at 4 months were not independent prognostic factors when tested with a Cox proportional hazard model. While each was significant in a univariate analysis, the 4-month MRD showed colinearity with others and the day 15 MRD alone remained a significant prognostic factor after adjusting for MRD at day 35 (hazard ratio = 5Æ7; 95% confidence interval = 1Æ9–17Æ1, P = 0Æ002). As well as identifying very high risk patients, the day 15 MRD could also predict patients with low, intermediate and high risk of relapse when three different thresholds were applied (Fig 4).
Predictive value of MRD at 12 and 24 months
Fig 2. Comparison of MRD levels in relapsed and matched control patients throughout therapy. The number (n) of samples tested for each group and time-point are shown. There were 18 relapses on therapy, reducing n in the relapse group by one at 4 months, five at 12 months and 18 at 24 months. Numbers in both groups were reduced due to missing samples.
Changes in MRD levels throughout therapy The proportions of patients with different levels of MRD in bone marrow measured by RQ-PCR are shown for both those patients who later relapsed and for those who remained in CR1 (Fig 2). It is clear that at all time points, more patients were MRD negative in the matched CR1 group compared to the relapse group. As expected, the proportion of patients who were MRD negative increased as therapy progressed with the exception of the end of therapy time point in the relapse group, which had more MRD positive samples than at 12 months. Examination of the individual patient data revealed that MRD levels at day 15, day 35 and 4 months dropped successively over the first three time-points in all except two patients, one of whom went on to have an early relapse. Almost all patients achieved MRD negativity by 4 or 12 months and the higher MRD levels at 12 and 24 months in the relapse group represented the reappearance of MRD and molecular relapse.
Predictive value of MRD at early treatment time points Kaplan Meier analyses (Fig 3) compared the relapse-free survival (RFS) for patients based on MRD measured at different stages of therapy. MRD was predictive of outcome at all treatment phases at a range of thresholds. The optimum
While MRD levels at early time-points provided a measure of response to therapy, by 12 months all patients had achieved MRD negativity and the detection of MRD later in therapy was usually due to molecular relapse. The optimium threshold for MRD at 12 months and end of therapy was 1 · 10)4 (Fig 3). When we explored the effect of lowering the threshold (data not shown), the difference between MRD positive and MRD negative patients remained significant (P < 0Æ05 12 months; P < 0Æ0001 24 months), but the test became less specific with several false positives in samples from non-relapsed patients who had non-quantitative positive MRD results (Fig 2). Several possible causes of an MRD negative result prior to relapse were investigated in the patients who relapsed. Firstly, this group included five patients with relapses clinically isolated to the CNS. MRD reappeared at 24 months in one of these patients, but the remaining four patients with isolated CNS relapses had no detectable MRD after day 35. Secondly, because clonal evolution can contribute to false negative results, particularly in studies using only one MRD marker, MRD was tested with a second marker in five of these patients. The second marker was MRD positive (1 · 10)4) for one patient and negative for the other four patients. Finally, a major factor contributing to MRD negative results in relapse patients was a long time interval between the collection of the last time-point tested and relapse. Of the eight patients who had a bone marrow relapse >7 months after the last MRD sample available for testing, only one was MRD positive (5% blasts) at day 14 demonstrated in successive Children Cancer Group trials (Gaynon et al, 1997). Even when this subgroup was excluded, the power of day 15 MRD kinetics in predicting outcome was evident and high MRD (‡5 · 10)3) and moderate levels of MRD (