- © 2005 by American Society of Clinical Oncology
Early Responses to Chemotherapy of Normal and Malignant Hematologic Cells Are Prognostic in Children With Acute Lymphoblastic Leukemia
- Stephen J. Laughton,
- Lesley J. Ashton,
- Edward Kwan,
- Murray D. Norris,
- Michelle Haber and
- Glenn M. Marshall
- From the Centre for Children's Cancer and Blood Disorders, Sydney Children's Hospital, Randwick; and Children's Cancer Institute Australia for Medical Research, Sydney, Australia
- Address reprint requests to Glenn M. Marshall, MD, Centre for Children's Cancer and Blood Disorders, Sydney Children's Hospital, High St, Randwick, Sydney, NSW 2031, Australia; e-mail: g.marshall{at}unsw.edu.au
Abstract
Purpose Improved cure rates for children with acute lymphoblastic leukemia (ALL) have resulted from better relapse prediction, using clinical and laboratory features at diagnosis, and more intensive therapy in high-risk patients. More recently, measurements of the variation in the response of malignant lymphoblasts to chemotherapy in vivo have further improved relapse prediction. It is unknown whether the variation in the response of nonmalignant hematologic cells after chemotherapy correlates with the response of lymphoblasts or risk of relapse.
Patients and Methods We retrospectively evaluated myelosuppression during induction and consolidation chemotherapy in 227 children uniformly treated for ALL on consecutive Australian and New Zealand Children's Cancer Study Group protocols. The early response to treatment was assessed in a representative subset (n = 62) by determining minimal residual disease (MRD) level by molecular techniques on the end-of-induction bone marrow sample.
Results We found that a slow rate of myeloid recovery at the end of induction chemotherapy, reflected in a low absolute neutrophil count (ANC), was highly predictive of relapse (P < .0001). Additionally, patients with a high end-of-induction MRD level had a high risk of relapse (P = .001). Multivariate analysis confirmed the independent prognostic significance of MRD and ANC at the end of induction chemotherapy (P < .05). There was no significant association between other measures of myelotoxicity and MRD or relapse.
Conclusion We conclude that the responses of normal myeloid cells and malignant lymphoblasts to chemotherapy predict outcome by distinct mechanisms. While these results are promising, their use in the clinical setting needs to be examined in a future randomized controlled trial.
INTRODUCTION
Cure rates for children with acute lymphoblastic leukemia (ALL) have improved dramatically with the use of multiagent chemotherapy.1,2 A large number of clinical and biologic prognostic factors have been determined during the last 30 years and have been used to stratify patients to risk-directed therapy. Clinical characteristics have included such factors as sex, age, presence of lymphomatous features, and WBC at diagnosis.3–5 Biologic features, including immunophenotype, cytogenetics, and lymphoblast morphology, have also been used to predict the likelihood of relapse.6–11 However, with improved chemotherapy regimens, many of these traditional prognostic factors have lost clinical significance.2,4,12 More recent studies have emphasized the significance of response to initial therapy as measured by the reduction in peripheral circulating lymphoblast count, early bone marrow response, and detection of minimal residual disease (MRD) at the end of induction.13–20 The biologic response of the malignant lymphoblast population to chemotherapy is clearly important in achieving a sustained complete remission and cure.
The administered doses of cytotoxic chemotherapy are calculated based on a child's surface area or weight. Children receiving cytotoxic chemotherapy for ALL have significant variation in the incidence and severity of both hematologic and nonhematologic side effects, often causing delays in scheduled therapy and reduced dose intensity. The variation in side effect profiles in individual patients is largely unexplained, except for the rare instances of inherited dysfunction of key drug metabolizing enzymes (DMEs) involved in cytotoxic drug metabolism, such as Thiopurine S-Methyltransferase (TPMT).21 It is unknown whether the variation in the individual response of nonmalignant hematologic cells to chemotherapy correlates with the risk of relapse or the response of malignant lymphoblasts to chemotherapy in children with ALL.
To address these issues, we examined a number of clinical measures of myelotoxicity during the first 3 months of therapy in a large cohort of children treated on consecutive Australian and New Zealand Children's Cancer Study Group (ANZCCSG) ALL protocols. Our results indicate that markers of chemotherapy-induced myelotoxicity provide additional prognostic information that could be used to further refine therapy.
PATIENTS AND METHODS
Patients
A retrospective analysis was conducted on a cohort of 227 children with newly diagnosed ALL, consecutively enrolled and treated on two successive ANZCCSG protocols for ALL. The patients were aged 1 to 18 years and treated at Sydney Children's Hospital (SCH) during the 12-year period of the ANZCCSG studies. They met the following additional eligibility criteria for participation in this retrospective study: remission achieved after a 5-week induction regimen, and appropriate medical records available. ANZCCSG ALL Study-V22 enrolled children prospectively from April 1986 to January 1992, and ANZCCSG ALL Study-VI, from February 1992 until March 1998. Of the 112 children treated at SCH on Study-V, 95 were eligible for analysis (87%). One child did not achieve remission after induction therapy, and there were two induction deaths. Fourteen children were excluded because medical records could not be obtained. Of 107 children treated on Study-VI at SCH, 98 were included in the analysis (93%). One child failed induction, and there was one induction death. Medical records were unavailable for seven. Overall, the three induction deaths (two due to sepsis, one from hemorrhage) and two failed inductions gave a remission rate of 97%.
To evaluate the association of end-of-induction MRD level with prognosis, the overall cohort was expanded by the addition of 34 children treated on ANZCCSG ALL Study-VI treated at the Children's Hospital at Westmead (n = 16) and The Women's and Children's Hospital, Adelaide (n = 18), who had achieved remission at the end of induction and for whom both the medical records and an end-of-induction bone marrow sample were available.
Treatment Regimens
The 5-week induction phase was identical in both Study-V and Study-VI.22 The 6-week CNS prophylaxis-consolidation phase was also identical in both studies, with the exception of low-risk patients as defined in Study-V.22 The 15 low-risk Study-V patients were thus included in the assessment of myelotoxicity in induction but were excluded from the assays of myelotoxicity during consolidation. No granulocyte colony-stimulating factor was used in any of the patients during the induction or consolidation phases of treatment.
Induction Phase
Induction treatment comprised vincristine 1.5 mg/m2 and daunorubicin 25 mg/m2 intravenously days 1, 8, 15, and 22; prednisolone 40 mg/m2/d orally days 1 to 28, then tapered over 7 days; L-asparaginase 6,000 U/m2 intramuscularly three times weekly for nine doses; and age-appropriate intrathecal methotrexate on days 2 and 22 (Fig 1).
Common induction and consolidation chemotherapy schema for Australian and New Zealand Children's Cancer Study Group Acute Lymphoblastic Leukemia Study-V and Study-VI. ITMTX, intrathecal methotrexate; VCR, vincristine; DNR, daunorubicin; L-Asp, L-asparaginase; CPA, cyclophosphamide; Ara-C, cytosine arabinoside; XRT, Irradiation; Gy, Gray; HR, high risk (see text for definition); IV, intravenously, IM, intramuscularly; PO, orally; SC, subcutaneously.
CNS Prophylaxis-Consolidation Phase
The CNS prophylaxis-consolidation phase commenced after M1 remission was confirmed on the day-35 bone marrow aspirate (BMA); there was no severe infection, and hematological recovery had occurred. Therapy comprised vincristine 1.5 mg/m2 intravenously days 1, 8, 22, and 29; cyclophosphamide 1,000 mg/m2 intravenously days 1 and 22; and cytosine arabinoside 75 mg/m2/d intravenously or subcutaneously on days 1 to 4, 8 to 11, 22 to 25, and 29 to 32. Age-related doses of intrathecal methotrexate were administered on days 1, 8, 15, and 22 independent of blood count. Prophylactic cranial irradiation (18 Gy in 12 fractions) was administered to patients with National Cancer Institute (NCI) high-risk category features, or leukemia/lymphoma.
Study-V
Patients were then randomized to one of two arms: arm A (asparaginase intensification) or arm C (cyclic therapy). Arm A consisted of L-asparaginase, methotrexate, and 6-mercaptopurine, followed by maintenance chemotherapy (vincristine, methotrexate, and 6-mercaptopurine). Arm C commenced with interval maintenance (methotrexate and 6-mercaptopurine), followed by reinduction (vincristine, doxorubicin, L-asparaginase, and dexamethasone), then reconsolidation (cyclophosphamide, cytosine arabinoside, and 6-thioguanine). Maintenance chemotherapy (daunorubicin or methotrexate, cyclophosphamide, vincristine, 6-mercaptopurine, prednisolone, and methotrexate) followed.
Study-VI
Following induction and consolidation-CNS prophylaxis, all patients received L-asparaginase intensification as detailed earlier for arm A, Study-V. Interval maintenance followed (vincristine, methotrexate, and 6-mercaptopurine), before randomization to receive either ongoing maintenance (arm A), or reinduction-reconsolidation (as for Study-V, with L-asparaginase being omitted) followed by maintenance (arm B).
Therapy was discontinued 2 years from the end of induction in both Study-V and Study-VI.
Data Collection
A standardized data abstraction form was designed to capture appropriate information from individual patient medical records, including CBC results during consolidation treatment. Delay in therapy was measured at three early time points: at the end of induction (day 35), midconsolidation (day 56), and end of consolidation (day 77). The absolute neutrophil count (ANC) represents the sum of the number of mature neutrophils and band forms and is expressed in cells per liter. If the ANC was less than 0.5 × 109/L or the platelet count was less than 100 × 109/L at the end of induction or midconsolidation, therapy was delayed until these parameters were satisfied. An ANC of 1.0 × 109/L was required for therapy to proceed at the end of consolidation.
All available medical records were reviewed and data were collected by one of the author's (S.J.L.). Where medical records were incomplete, a search for missing data was performed on the institution's computerized laboratory database. Where follow-up was limited, attempts were made to gain the latest known contact with the medical system by contacting either the last known family practitioner or pediatrician.
MRD Testing
MRD testing was performed on day 35 BMA samples in a subset of the overall cohort from Study-VI (n = 62). These patients formed part of a larger cohort reported elsewhere.23 Polymerase chain reaction (PCR) primers specific for clonal antigen receptor gene rearrangements of the IgH or TCRγ loci were used in a real-time PCR method.24 Patients whose end-of-induction BMA sample had evidence of leukemic lymphoblasts at a level greater than 10−3 on real-time PCR were categorized as having high-level MRD, while levels ≤ 10−3 were classified as having low-level MRD.
Statistics
The Cox proportional hazards regression model was used for both univariate and multivariate analyses. The relationships between various factors and outcomes were assessed by event-free survival (EFS) probabilities, estimated by Kaplan-Meier and Cox regression analysis. An event was defined as a leukemic relapse or death. Children without an event were censored at the date of last follow-up. All statistical analyses were carried out using STATA Version 8 (Stata Corp, College Station, TX). Significance was determined at a P value level of less than .05 (two-sided).
Ethical Clearance
Patients were enrolled prospectively on clinical trials approved by institutional human research ethics committees, following written informed consent from a parent or guardian of each patient.
RESULTS
The clinical and laboratory features at diagnosis and the clinical outcome characteristics for the MRD-tested (n = 62) and MRD-untested (n = 165) patients are presented in Table 1. The MRD subset seems to be representative of the total cohort from which it was drawn. Immunophenotype was unavailable in a significant number of children enrolled on Study V. The baseline characteristics and clinical outcome data seem to be representative of the much larger Australasian-wide cohort of children treated on the same two ANZCCSG protocols.22,23 There were no major differences in baseline characteristics at diagnosis or outcome in the Study-V (n = 95) or Study-VI (n = 132) patients. In addition, the clinical characteristics of the children contributed by the Children's Hospital at Westmead and the Women's and Children's Hospital, Adelaide (n = 34), seem similar to that of the children from SCH (n = 193). Since the induction and CNS prophylaxis-consolidation phases of these two protocols were identical, we have grouped all patients together for the following myelotoxicity analyses.
Clinical Characteristics and Outcome for the Total Patient Cohort
Normal Hematologic Cell Recovery During Induction and Consolidation Therapy
To examine the variation in the response of normal hematopoeitic cells to induction and consolidation chemotherapy, we first evaluated individual patient CBCs at the end of induction and each week during consolidation. We next divided the patient cohort around the median values for hemoglobin, platelet, WBC, ANC, monocyte (AMC), and lymphocyte (ALC) counts, respectively, for each of the three time points, and then evaluated the prognostic significance of fast or slow hematologic recovery (Table 2). We found a significant association between risk of relapse and both low WBC and ANC at the end of induction (Tables 2 and 3). A low WBC at the completion of induction was associated with an increased risk of relapse (P = .001). A low ANC at this same time was also associated with relapse (P < .0001), with a 5-year EFS of 55.9% compared with 80.2% in the higher-ANC group (Table 4). Moreover, when the ANC was divided into tertiles (intertertile divisions 1.87 × 109/L and 3.78 × 109/L), a “dose-response” relationship was observed (Table 4). The lowest ANC tertile hazard ratio (HR) was 3.15 (95% CI, 1.66 to 5.98; P < .0001), while the intermediate ANC tertile had a HR of 2.14 (95% CI, 1.09 to 4.18; P = .026), when compared with the highest tertile (test for trend P = .0004). Additional measures of myelosuppression at the end of induction, including hemoglobin, platelet count, AMC, and ALC, did not have a significant association with outcome in our study cohort. There were no differences in patient characteristics at initial diagnosis such as age or WBC, for those patients with low ANC at end of induction. Furthermore, hemoglobin, platelet count, WBC, ANC, AMC, or ALC at the end of early consolidation or late consolidation were not significantly associated with relapse in this cohort (Table 2).
End of Induction, Early Consolidation, and Late Consolidation Blood Parameters and Relapse Risk
Univariate and Multivariate Analysis of Relapse Risk (N = 227)
Event-Free Survival According to (A) ANC and MRD Level at the End of Induction Chemotherapy and (B) Predictive Models Combining ANC and NCI Risk Category or MRD Level
In multivariate analyses, ANC and NCI risk category remained significantly associated with an increased risk of relapse after adjusting for sex, end-of-induction hemoglobin, platelet count, WBC, AMC, and ALC (P ≤ .015; Table 3). Sex and WBC were no longer significantly associated with EFS (P ≥ .079).
We hypothesized that slow hematologic recovery would correlate with delays in completing induction and consolidation and that patients with delayed treatment or reduced dose intensity would have a poorer prognosis. Therefore, we evaluated the median duration of both induction and each phase of consolidation in 189 of 227 study patients for whom complete data were available. No significant association was found between delay at any measured time point and the risk of subsequent relapse (P > .05; data not shown).
Level of Chemoresponsiveness of ALL Blasts in Vivo
We next examined the rate of disappearance, or chemoresponsiveness, of ALL blasts in peripheral blood and bone marrow in the subset of 62 patients, in whom samples were evaluated for MRD level. We then looked for associations between chemoresponsiveness of ALL blasts and our measures of myelotoxicity. Those children who took longer than the median of 2 days (range, 0 to 11 days) to achieve a peripheral blood lymphoblast count of less than 1.0 × 109/L had an increased risk of relapse (HR, 3.38; 95% CI, 1.21 to 9.39; P = .02) in univariate analysis. Time, greater than the median of 4 days (range, 0 to 23 days), to disappearance of the peripheral lymphoblasts was also associated with an increased risk of relapse (HR, 2.56; 95% CI, 1.04 to 6.32; P = .04).
Results from univariate analysis in the MRD tested cohort are shown in Table 5. Higher MRD level (HR, 4.52; 95% CI, 1.86 to 10.99; P = .001), ANC (HR, 9.84; 95% CI, 2.87 to 33.75; P < .0005), and WBC (HR, 6.59; 95% CI, 2.20 to 19.78; P = .001) at the end of induction remained significantly associated with an increased risk of relapse during follow-up. The early peripheral blast response was not associated with either the end-of-induction MRD level or ANC result. A multivariate model adjusting for end-of-induction WBC and ANC (normal cell chemosensitivity), end-of-induction MRD (malignant lymphoblast chemosensitivity), and NCI risk category was performed. Both higher ANC (HR, 5.60; 95% CI, 1.44 to 21.71; P = .013) and MRD (HR, 3.03; 95% CI, 1.20 to 7.65; P = .019) levels at end of induction remained independently associated with an increased risk of relapse. NCI risk category also remained significantly associated with risk of relapse in this cohort, whereas WBC at the end of induction did not. We also observed that the relationship between end-of-induction ANC and EFS was seen in both the NCI standard- and high-risk groups (Table 4). Lastly, we stratified the 62 patients into four groups, divided around the median end-of-induction ANC and MRD level of 10−3 (Table 4), and showed a significant relationship (test for trend P < .00005). Patients with an end-of-induction low ANC and high MRD, had an extremely poor prognosis (5-year EFS, 23.1%), compared with those patients with a high ANC and low MRD, who had a very low risk of relapse (5-year EFS, 92.9%).
Univariate and Multivariate Analysis of Prognostic Markers and Relapse Risk in the MRD-Tested Cohort (n = 62)
DISCUSSION
The major improvements in survival for children with ALL have resulted from strategies aimed at intensifying therapy for high-risk patients. Here we have shown that, among a range of measures of normal hematologic cell recovery in the early phases of chemotherapy in children with ALL, faster ANC recovery at the end of induction was significantly associated with decreased risk of leukemic relapse. The prognostic significance of the end-of-induction ANC was independent of MRD level at the same time point. Moreover, patients could be grouped on the basis of end-of-induction ANC and MRD levels, into categories that accurately predicted their subsequent risk of relapse, suggesting a model for further individualizing the chemotherapy program for both high- and low-risk patients. Our results indicate that end-of-induction ANC may complement MRD level in predicting later relapse, and suggests that variation in the response of normal hematologic cells at the end of induction might be used in a future trial design.
The mechanism whereby faster recovery of the end-of-induction ANC correlated with a reduced risk of later relapse is unclear. One possibility is that end-of-induction ANC is a general marker of the patient's capacity for normal hematologic cell recovery after chemotherapy. However, we did not find a significant relationship between high ANC at the end of induction, and hematologic recovery or delays during consolidation phase therapy. The two protocols used low ANC and platelet count at the end of induction to determine whether a child had a delay in their therapy at this point. However, only 5 of 227 children in our cohort had delay (range, 2 to 7 days) in starting consolidation that was attributable to low ANC or platelet count. Hence, we believe our findings are not due to confounding by these variables. The association of multiple DME genotypes with markers of myelotoxicity needs to be explored in an adequately powered, prospective study to determine whether such a relationship exists and could explain our clinical finding. The association of end-of-induction ANC may have related to variations between patients in the prednisone responsiveness of normal and malignant cells. Indeed, monotherapy with prednisone in the first week after diagnosis is used by many groups to identify one category of high-risk ALL patients with poor prednisone response.25,26
Our results provide the basis for a hypothetical model for individualizing therapy based on the end-of-induction ANC and MRD level (Table 4). The ultimate success of anticancer chemotherapy in an individual relies on the principle that the malignant cell is much more sensitive to cytotoxic therapy than normal cells. The difference between these two parameters is known as the therapeutic index. A narrow therapeutic index results in treatment failure when there are high levels of intrinsic or acquired drug resistance in the malignant cells, and excessive sensitivity in normal cells. Conversely, a wide therapeutic index may allow reduction in the dose and side effects of the cytotoxic therapy. Thus, patients with high ANC and low MRD at the end of induction may be the subject of a study aimed at reducing treatment intensity, whereas, patients with low ANC and high MRD could be considered for experimental therapy.
In summary, we have shown that measures of chemosensitivity of normal and malignant cells in ALL patients early in the course of treatment enhance relapse prediction and provide a rational basis for future trial design. Our observations suggest several testable hypotheses to explain the mechanism by which end-of-induction ANC affects the subsequent risk of relapse. The clinical utility of these results needs to be confirmed in a prospective, randomized, controlled trial before any alteration in current clinical practice.
Authors' Disclosures of Potential Conflicts of Interest
The authors indicated no potential conflicts of interest.
Acknowledgments
We thank Drs Luciano Dalla-Pozza, Frank Alvaro, and Heather Tapp for facilitating data collection at their institutions.
Footnotes
-
Supported by research grants from the National Health and Medical Research Council of Australia, New South Wales State Cancer Council, and the Sydney Children's Hospital Foundation, Joshua Holland Leukaemia Fund (S.J.L.).
Presented in part at the 45th Annual Meeting of the American Society of Hematology, San Diego, CA, December 8, 2003.
Children's Cancer Institute Australia for Medical Research is affiliated with the University of New South Wales and Sydney Children's Hospital, Randwick.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
- Received April 2, 2004.
- Accepted December 29, 2004.
