The philadelphia chromosome (Ph1) is detectable in 2% to 5% of children with acute lymphoblastic leukemia (ALL) 12. The detection of a philadelphia chromosome remains a major prognostic factor of induction failure. Despite the steady improvement in the management of ALL in children, Ph1-ALL is associated with high rates of relapse or resistance to treatment 345. This disease is heterogeneous in terms of clinical parameters such as leukocyte count, age at diagnosis, and initial steroid response 46. A slow early response to conventional therapy has also been reported as indicative of a poor prognosis 2.

Recent gene expression studies have identified a heterogeneous pattern of expression associated with BCR-ABL status, which may be useful for developing novel prognostic markers and future patient stratification procedures 789. New therapeutic agents such as tyrosine kinase inhibitors (imatinib and dasatinib) have been developed and yield good results in adults with Ph1-ALL 101112. Little information on the use of these drugs in children has been reported; the identification of predictors of responsiveness to early conventional treatment may thus be beneficial for the accurate stratification of children and for improving outcome 1314.

We studied the impact of the National Cancer Institute (NCI) risk factors and steroid and early chemotherapy responses in 36 children with untreated Ph1-ALL enrolled in the FRALLE 93 trial between 1993 and 1999.

The FRALLE 93 trial was open to children aged 0 to 20 years with untreated ALL, not including those with L3 ALL or Down's syndrome. Between June 1, 1993, and December 31, 1999, 1395 children were enrolled onto the FRALLE 93 trial in 18 French pediatric centers and one Belgian pediatric center. This study was approved by the ethics committee of the hôpital Saint Louis, France (accepted April 29, 1993). All patients, or their parents, provided informed consent in accordance with the Declaration of Helsinki. The diagnosis of ALL was based on morphological, immunophenotypic and cytogenetic analyses of bone marrow samples. From 1994, children were systematically screened for four fusion transcripts (TEL-AML1, BCR-ABL, E2A-PBX1, MLL-AF4).

Ph1 was detected by conventional cytogenetic analysis (t(9;22) (q34;q11)) or a BCR-ABL rearrangement was detected by reverse transcriptase-polymerase chain reaction (RT-PCR) in 3% of the B-lineage ALL. Of the 36 Ph1-ALL children, 30 had detectable t(9;22) and six were identified solely by the presence of the BCR-ABL fusion transcript. Thus, fusion transcript analysis was able to detect nearly 17% more Ph1 patients than conventional karyotype. The m-BCR breakpoint was detected in 30 patients and the M-BCR breakpoint in three others (3 patients were identified solely by the detection of t(9;22)). At baseline, median age was eight years [range: 0.70–19.5; Q1–Q3: 4.35 to 12.35], male-to-female ratio was 1.1, median leukocyte count was 29.1 × 10⁹/l [Q1–Q3: 12.5–137] and median hemoglobin was 9 g/dl [2.5–15]; four patients displayed central nervous system defects (11.1%).

Complete remission was observed in 26 children (72%), after a median of 40 days [range 32–51] consistent with previous findings 24. We did not find any differences between the effects of including patients at two different time periods (before or after July 1996) and the addition of one dose of daunorubicin and two doses of asparaginase (but the numbers of children within each group are small) (Table 2).

Ten patients did not achieve CR after induction therapy, but CR was observed in seven of them after one further course of chemotherapy. The dexamethasone, cytarabine, cyclophosphamide, etoposide, and idarubicin or daunorubicin (CAZED) scheme was recommended as salvage treatment in the FRALLE 93 protocol and thus was used in five of these seven patients (amsacrine and cytarabine for the 2 others) 5. All seven children were then transplanted (2 matched related transplants, 4 matched unrelated transplants and one autologous transplantation); outcomes are described in Figure 1. Of the other children with CR observed after induction, five of the six children with an HLA-matched sibling (MRD) received alternate courses of R3 and COPADM treatment before allogeneic bone marrow transplantation (one toxic death occurred before graft). Eight children received an autologous transplant, five a mismatched related transplant and two others a matched unrelated transplant (MUD) (Figure 1). Five other CR patients received a different post-induction regimen with intensified chemotherapy, including one consolidation, a double delayed intensification, and an interim phase over 36 weeks before 18 months of maintenance therapy.

Patient follow-up data were updated in February, 2007. In an intention-to-treat analysis, 10 children remained in CR1, giving a five-year disease-free survival (DFS) rate of 42.3% ± 9.7% and a five-year OS rate of 47.2% ± 8.3. Leukemic relapse was the most common cause of adverse events. Eleven children relapsed (no late relapse after 5 years) and five toxic deaths occurred (including one late death from infection). No secondary malignancy was observed.

Table 2 summarizes the main endpoints – CR, EFS and survival – overall and as a function of the main patient characteristics. Based on univariate analyses using a 5% significance level for all endpoints, we identified two predictors of poor outcome: age ≥ 10 and white blood cells (WBC) count >100,000/mm³. Response to chemotherapy based on day 21 bone marrow was only predictive of induction response (p = 0.0004). When incorporated simultaneously in multivariate regression models, only one predictive factor retained significant prognostic value for a particular endpoint: M1 for CR, WBC count >100,000/mm³ for EFS and age ≥ 10 for OS. Combining the information of these three binary variables defined two groups differing widely in terms of outcome: the 14 children with age<10, WBC<100,000/mm³ and M1 defined a group with a favorable prognosis (100% CR, 5-yr EFS: 57%, 5-yr OS: 79%), whereas the remaining 22 children had poorer outcomes (55% CR, 5-yr EFS: 18%, 5-yr OS: 27%) (p = 0.003, 0.002, and 0.003, respectively) (Figure 2). These prognostic factors remain important in the transplanted patients (equally divided in both groups), even if significance is not reached due to the small number of patients (5-yr DFS: 83 ± 14% for patients with age<10, WBC<100,000/mm³ and M1 and 33 ± 15% for the remaining patients) (p = 0.14).

This long-term study, carried out before the introduction of imatinib mesylate, confirmed the prognostic value of two major clinical factors (age, WBC count at diagnosis) and identified the D21 marrow response as a powerful complementary tool. We did not find prednisone response to have statistically significant predictive value for any endpoint studied, in contrast with earlier reports 46. This could be explained by the small sample size, and the small number of poor prednisone responders (5 of 36; 14%). However, in our protocol the combination of a three-drug intrathecal injection (methotrexate, cytarabin and depomedrol) on day 1 with the 7 day prednisone prophase prevents the comparison of our steroid-resistance data to data in other studies using intrathecal methotrexate.

Based on these three easily available indicators (age, WBC at diagnosis and D21 response to chemotherapy), a predictive model can be built to define two subsets of children that differ widely in terms of outcome. Such a model should be further investigated in larger samples and in ongoing pediatric trials integrating tyrosine-kinase inhibitors. The improved early responses observed with imatinib or dasatinib in adult studies, and similar, but very preliminarily, results obtained in one pediatric study with imatinib, now question the appropriate use of allogeneic stem-cell transplantation 1418. Indeed, the COG AALL0031 study showed that continuous administration of imatinib given in combination with intensive chemotherapy backbone resulted in a significant improvement in early EFS. More specifically cohort 5, who received 340 mg/m2 of imatinib for 280 days with chemotherapy only, had a similar two years EFS as compared to the cohort of patients who underwent stem cell transplantation (32 patients) either according to the protocol (MRD-21 patients) or from MUD (11 patients)14. The combination of available tools, including minimal residual disease assessment, with these easily measured predictive features could be useful for refining the indications for bone marrow transplantation 19202122. Thus, if the long term follow-up of the AALL0031 non randomized study was confirmed, good risk patients could be spared by transplantation.

The authors declare that they have no competing interests.

All the authors have substantially contributed to the conception and design or acquisition of data or analysis and interpretation of the data in this multicentric study, and participated to drafting and revising the article. SC and MFA performed statistical analysis. All authors read and approved the final manuscript.