Thioguanine offers no advantage over mercaptopurine in maintenance treatment of childhood ALL: results of the randomized trial COALL-92

The prognosis of children with acute lymphoblastic leukemia (ALL) has markedly improved during the past 30 years. Nevertheless, 25% of the children still suffer a relapse. Current research focuses mainly on identification of further prognostic factors for better risk-adapted treatment strategies. On the other hand, the use of other, not necessarily new, drugs might improve outcome in childhood ALL.

For historical reasons, mercaptopurine (MP) has become the standard drug in maintenance treatment of ALL, whereas thioguanine (TG) has thus far been used only to treat acute myeloid leukemia (AML) and relapsed ALL. The thiopurines TG and MP have to be converted intracellularly to thiopurine nucleotides before expressing cytotoxic activity.^1,2  In TG, conversion into G-thioguanine nucleotides (TGNs) is more direct. Low concentrations of TGN correlated with a higher risk of relapse in childhood ALL.^3,4  Furthermore, in vitro studies on leukemic cells and cell lines showed that with TG the same cytotoxic effect was achieved with a 10- to 20-fold lower concentration and 3-fold shorter exposure time compared with MP.⁵  For these reasons a higher clinical effectiveness of TG in maintenance treatment was postulated. The cooperative study group for childhood acute lymphoblastic leukemia (COALL), part of the German Society for Pediatric Oncology and Hematology, started a randomized multicenter trial to examine this hypothesis. Approval for this study was obtained by the Ethics Commission Hamburg, Germany.

From February 1992 until August 1997, 561 children 1 to 18 years of age with B-precursor B-ALL or T-ALL were enrolled from the 19 participating German hospitals. A total of 40 patients (7.1%) were not eligible for COALL-92 according to the protocol criteria. Of these patients, 24 had undergone major pretreatment, 10 received inadequate treatment because of nonmedical reasons, 4 were diagnosed with biphenotypical leukemia, and 2 died prior to any treatment. Thus, 521 patients were eligible for the study. Informed consent was obtained from all patients or their parents. Diagnosis of ALL was established when at least 25% lymphoblasts was present in the bone marrow (BM) or when blasts were present in cerebrospinal fluid (CSF). BM, blood smears, and CSF cytospin were reviewed centrally in the laboratory of the study in Hamburg, Germany. Central nervous system (CNS) involvement was diagnosed if more than 5 cells per microliter were counted in the CSF with unequivocal lymphoblasts.

Immunophenotypic subgroups were defined according to the criteria provided by the European Group for the Immunological Characterization of Leukemias: common ALL—terminal deoxynucleotide transferase (TdT)⁺, CD19⁺, CD10⁺, cytoplasmic immunoglobulin M (cyIgM^–), surface Ig (sIg)^–; pre–B-ALL—TdT⁺, CD19⁺, CD10^+/–, cyIgM⁺, sIg^–; pro-B-ALL—TdT⁺, CD19⁺, CD10^–, cyIgM^–, sIg^–; and T-ALL—TdT⁺, cyCD3⁺, CD7⁺.⁶  Surface antigens were considered positive if at least 20% of the leukemic cells expressed the antigen. Molecular screening for rearrangement of BCR/ABL, TEL/AML1, and MLL/AF4 was carried out in most patients.

Patients were stratified into a low-risk (LR) or high-risk (HR) group according to age, initial white blood cell (WBC) count, and immunologic subtype. Patients in the LR group had a WBC count of less than 25 × 10⁹/L, were aged 1 to 9 years, and had common or pre–B-ALL. HR patients had a WBC count of 25 × 10⁹/L or higher, or were 10 years or older, or presented with pro-B-ALL or T-ALL. Furthermore, all patients with Philadelphia chromosome–positive (Phi+) ALL and those patients who did not achieve complete remission (CR) until day 29 of the protocol were treated according to the HR protocol. Figure 1 shows an outline of the treatment strategy; the details of the protocol elements are provided in Table 1. In contrast to the prior COALL protocol, besides randomization in maintenance treatment, 2 further treatment modifications were evaluated: limitation of preventive cranial irradiation (CRT) to HR patients with T-ALL or WBC count of 25 × 10⁹/L or higher in order to avoid toxicity and administration of a daunorubicin (DNR) prephase in order to examine the efficacy of prolonged less cardiotoxic anthracycline infusions with bolus treatment.

LR patients without primary CNS involvement received a total of 18 injections of intrathecal methotrexate (MTX) in age-adjusted dosage (12 during intensive treatment and 6 during maintenance therapy). Preventive CNS irradiation at age-adjusted dosage (12-18 Gy) was given only to HR patients with T-ALL or WBC count of 25 × 10⁹/L or higher. During intensive phase, HR patients received a total of 12 doses MTX intrathecally; to patients who did not undergo CRT the additional 6 injections during maintenance treatment were given.

During the study the following amendments were made in April 1994: (1) According to the policy of the previous studies to stepwise increase the leukocyte count above which preventive CRT should be given, patients with B-precursor ALL and WBC count less than 50 × 10⁹/L were excluded from CRT. (2) DNR dose on day 22 was omitted to reduce the total anthracycline dose to that of the previous protocol. (3) Asparaginase dose was halved since measurements showed that asparagine was depleted up to 8 weeks.

Allogenic bone marrow transplantation (BMT) was recommended for Phi+ ALL defined by translocation t(9;22) or BCR/ABL rearrangement t(4;11), or late/nonresponse to induction therapy. (Patients not in remission on day 29 continued the protocol and had another BM examination 2 to 4 weeks later. Those obtaining CR by day 56 were called late responders; if no remission was achieved by this time the patient was considered a nonresponder.).

After completion of intensive treatment, patients were randomized within the low- and high-risk group to receive either MP or TG orally daily at a scheduled dose of 50 mg/m². When thrombocytopenia turned out to be a major problem, the starting dose of TG was reduced to 40 mg/m². In addition, MTX was given orally at a scheduled dose of 20 mg/m². Doses were then adjusted in order to maintain the WBC count between 2 to 3 × 10⁹/L. For all patients, the total duration of therapy was 24 months. WBC count; percentage of lymphocytes and granulocytes; platelets; drug doses of MP, TG and MTX; and reasons and duration of treatment interruptions were recorded for each patient weekly and evaluated centrally.

Duration of event-free survival (EFS) in this study was defined as the time from start of maintenance treatment until the date of failure (death, relapse, second malignancy). In alteration of the usual definition of EFS induction, failures such as death during induction therapy or failure to achieve complete remission by day 56 (nonresponse) were not considered for this investigation because this kind of event meant exclusion from randomization for maintenance treatment. Probability of EFS was estimated as intent-to-treat using the Kaplan-Meier method, and log-rank tests were used to evaluate differences between the EFSs in patient groups.^7,8  Patients without failure were censored at the day of last follow-up on September 1, 2002.

Table 2 summarizes the characteristics of all patients evaluable in trial COALL-92. The median age was 4.8 years (range, 1-17.8 years). The median WBC count at presentation was 10.5 × 10⁹/L (range, 0.4-1258 × 10⁹/L). B-cell–precursor ALL predominated (85%); only 15% of patients had T-ALL. Among patients with B-cell–precursor ALL, common ALL was diagnosed in 66.8%, pre–B-ALL in 15.7%, and pro-B-ALL in 2.5%. The distribution of immunosubtypes was identical to that in former studies. Primary CNS involvement was diagnosed in 3.1% of all patients; 4 boys (1.3%) had testicular involvement. CRT was given to 165 HR patients (60%) and to 2 LR patients (0.8%). Among 407 patients successfully tested with cytogenetic or molecular genetic methods, 7 (1.7%) were positive to have Phi+ ALL.

Of 488 patients eligible for randomization, 474 patients or their guardians accepted randomization: 238 children received MP, and 236 children received TG. According to the choice of the parents, 14 patients were allotted to the MP group. Table 3 gives the initial characteristics in both randomization groups. Distribution regarding sex, age, initial WBC count, immunologic subtypes, and risk groups was comparable in both arms. In the LR group 2 patients received CRT, both randomized to the TG arm. Of the HR patients, 155 received CRT.

Further, 33 patients were not randomized due to the following: early death (2), nonresponse (5), early relapse (8), death in remission (5), BMT in first CCR due to Philadelphia chromosome (7) or late response (5), and loss of follow-up (1).

EFS was estimated as intent-to-treat analysis from the start of maintenance treatment. The probability of EFS for the MP group was 78% ± 3% and 78% ± 3% for the TG group (P = .87) with a median observation time of 7.6 years (Figure 2). In the MP group, 47 relapses occurred, mostly in bone marrow; 45 relapses occurred in the TG group. However, risk for CNS relapse was higher in patients receiving TG during maintenance treatment (P = .053). In both randomization arms these events occurred mainly in patients with high-risk features. According to risk groups, EFS rate was 73% ± 5% (TG) versus 80% ± 4% (MP) for HR patients (P = .73) and 84% ± 3% (TG) versus 77% ± 5% (MP) for the LR group (P = .77) (Table 4). While still on maintenance treatment in the MP group, 1 patient died during remission due to cerebral infarction and 2 developed second malignancy (Morbus Hodgkin, osteosarcoma). The patient developing Morbus Hodgkin 8 years after T-ALL was diagnosed received preventive CRT during primary treatment; the other patient had been treated according to the LR protocol because of pre–B-ALL 7.5 years earlier and did not receive CRT. In the TG group, 1 patient died in complete remission due to viral encephalitis and pneumonia after maintenance treatment; in 3 children second malignancies were diagnosed (AML, glioblastoma, osteosarcoma). These 3 patients initially presented with HR features and had preventive CRT because of T-ALL. It is of importance that AML was diagnosed only 10 months after start of ALL treatment.

Toxicity data of maintenance treatment were available in 257 patients. Toxicity profiles based on a total number of 17 529 weeks of documented maintenance therapy forms differed markedly in both treatment arms. During this time, maintenance treatment had to be interrupted for 255 weeks (2.8%) in the MP group. In the TG group, treatment interruptions occurred at a 1.7-fold higher rate. Therapy had to be stopped for the total amount of 389 weeks (4.7%). The higher frequency of interruptions was due to a 7.5-fold higher incidence of thrombocytopenia below 100 × 10⁹/L without evidence of leukocytopenia below 1 × 10⁹/L. In 70% of these episodes, platelets dropped to values between 50 to 100 × 10⁹/L; in 8% below 25 × 10⁹/L. Severe bleeding did not occur in any of these cases; however, on 10 occasions platelet transfusions were given. Other reasons for treatment interruptions, such as leukocytopenia or infections with or without leukocytopenia, occurred at comparable rates in both groups.

Severe hepatotoxicity, especially venous occlusive disease (VOD), was not reported on the toxicity documentation sheets. Since cases of VOD occurred in a study of the Children's Cancer Study group (P. Gaynon, personal communication, February 2000), a questionnaire was sent to the participating hospital to recheck. Again, no VOD was reported.

In spite of the higher rate of treatment interruptions, the mean WBC count was comparable in both groups (MP, 3.4 × 10⁹/L; TG, 3.6 × 10⁹/L).Also, mean lymphocyte counts were analyzed since lymphocytopenia might better reflect the cytotoxic effect on lymphoblasts. In both groups the mean value was 0.7 × 10⁹/L. Further analysis revealed that most episodes of “isolated” thrombocytopenia were accompanied by prolonged lymphocytopenia but not necessarily low WBC count. An analysis of EFS according to mean lymphocyte count at several cut points (< vs > 0.5 × 10⁹/L; < vs > 1.0 × 10⁹/L; < vs > 1.5 × 10⁹/L) was made showing no statistical difference so far.

Mean drug doses were 25% lower in the TG group (MP, 49 mg/m²; TG, 36 mg/m²). On molar basis the mean thioguanine dose was even 30% lower (MP, 315 μmol/m²; TG, 215 μmol/m²). A detailed analysis regarding hematologic toxicity will be published elsewhere. Analysis of EFS according to the amendments resulted in similar outcome as well.

The probability for EFS of all 521 patients was 74% ± 3% after a median observation time of 6.6 years (range, 0-8.6 years); the estimate of disease-free survival (DFS) was 77% ± 3%, and the estimate of overall survival (OS) was 83% ± 2%.

Trial COALL-92 was the largest of the 5 studies of the COALL study group so far. The combination chemotherapy corresponded to the former protocol. The EFS of 74% represents a constant result compared with the former trial COALL-89.⁹  The incidence of CNS recurrences remained as low as it was in trials COALL-85 and COALL-89 in spite of limitation of CRT to patients with T-ALL or patients with a WBC count of 25 ×10⁹/L or higher (≥ 50 ×10⁹/L) and non–T-ALL. Results of DNR response are reported elsewhere (submitted).

According to the studies of Lilleyman and Lennard⁴  that showed an increased risk of relapse for children with low TGN levels during maintenance therapy with MP, an improvement of EFS with TG was expected because of its differences in metabolism. Actually, measurements of TGN concentrations in a subset of our patients confirmed that during TG treatment TGN levels exceeded 7 times those of the MP group.¹⁰  Nevertheless, in our study, maintenance treatment with TG was not more effective than the standard therapy with MP and was more complicated to handle due to a specific toxicity profile of prolonged myelosuppression with marked thrombocytopenia (submitted).

It could be argued that the frequent treatment interruptions in the TG group might have neutralized the possible superiority of TG. However, when effective treatment on leukemic cells is measured either by total leukocyte count, neutrophil count, or maybe even more appropriately by lymphocyte count, the patients receiving TG did not have less myelosuppression and even had more prolonged lymphocytopenia.

Why then were the high TGN levels of no therapeutic advantage? One explanation could be that the impact of maintenance treatment and therefore also of changes in this part of therapy are less pronounced in a very intensive frontline protocol.

Another argument could be that the TGNs are not the only metabolites important for the cytotoxic effect on leukemic cells. As previously published, in patients receiving TG, methylated TGNs reached about 40% of the concentration of unmethylated TGNs, whereas during MP treatment methylated thioinosine nucleotides were measured at a 26-fold higher concentration than TGNs.¹⁰ 

A role of methylated metabolites in cytotoxicity might explain why patients receiving MP in our study had the same prognosis as children receiving TG. The similar leukocyte count in children with MP and TG in spite of 7-fold higher TGN concentrations in the TG group supports the additional cytotoxic role of the methylated metabolites.

The sample size of our randomized cohort was sufficient only to detect a difference in EFS of 10% between the groups. A larger trial might find an advantage for either drug, especially in low-risk patients with a longer relapse cascade in whom maintenance treatment may play a more important role. There are 2 randomized studies, one by the Childhood Cancer Group and the other by United Kingdom Medical Research Council, presently addressing this question.

In summary, in our study maintenance treatment with TG was as effective as standard treatment with MP but more complicated to handle due to a specific toxicity profile of prolonged myelosuppression with marked thrombocytopenia. Although no cases of venous occlusive disease as reported by others occurred in our TG group, MP should remain the preferred drug for maintenance treatment of ALL unless other studies demonstrate superiority of TG in larger trials or selected patient groups.

Additional members of the COALL Study Group, Germany: Dr James F. Beck, Children's University Hospital, Dept. Hematology and Oncology, Greifswald; Dr Wilburg Streitberger, Hospital Heide, Dept. Pediatrics, Heide; Dr Wenzel Nürnberger, Hospital for BMT and Oncology, IdarOberstein; Dr Christian von Klinggräff, Hospital Kiel, Dept. Hematology and Oncology, Kiel; Dr Peter Thomas, Hospital Krefeld, Dept. Pediatrics, Krefeld; Dr Dieter Körholz, Children`s University Hospital, Dept. Hematology and Oncology, Leipzig; Dr Wolfgang Müller, Hospital Neuwerk, Dept. Pediatrics, Mönchengladbach; Dr Peter Klose, Hospital Harlaching, Dept. Pediatrics, München; Dr Hermann L. Müller, Hospital Oldenburg, Dept. Pediatrics, Oldenburg; Dr Johannes Wolff, Hospital St. Hedwig, Dept. Pediatrics, Regensburg; Dr Joachim Weber, Hospital Dr-Horst-Schmidt, Dept. Pediatrics, Wiesbaden; and Dr Brigitte Dohrn, Hospital Helios, Dept. Pediatrics, Wuppertal.