Batia Stark, MD, The Centre of Paediatric Haematology/Oncology and the Cancer Cytogenetic Laboratory, Schneider Children’s Medical Centre of Israel, Petah Tiqwa 49202, Israel. E-mail: email@example.com
Owing to the increased central nervous system (CNS) relapse risk in T-cell acute lymphoblastic leukaemia (ALL), it is unclear whether preventive cranial radiation (pCRT) can be safely omitted. In this study, pCRT was replaced by extended triple intrathecal therapy (TIT) in prednisone good early responders – medium-risk (MR) group, accounting for 76% of T-ALL patients. From 1989 to 2003, 143 T-ALL patients aged 1–18 years were enrolled in the Israel National Studies (INS) 89 (n = 84) and INS 98 (n = 59) trials, based on ALL-Berlin–Frankfurt–Munster (BFM) 86/90 and ALL-BFM 95 protocols, respectively. Five-year event-free survival (EFS) of the MR group in the INS 89 (n = 60) was 70 ± 5·9% and the INS 98 (n = 43), 83·7 ± 5·6% (P = 0·12); the cumulative incidence (CI) of any CNS relapse was 5·0 ± 2·8% and 2·3 ± 2·3% (P = 0·50), respectively. There was no difference in outcome between MR patients with a white blood cell count (WBC) ≥100 × 109/l treated with extended TIT (n = 17) or pCRT (n = 10). For all T-ALL patients, 5-year EFS was 61·9 ± 5·3% in INS 89 and 72·9 ± 5·8% in INS 98, (P = 0·21); the CI of any CNS relapse was 7·1 ± 2·8% and 1·7 ± 1·7% (P = 0·142), respectively. Outcome of T-ALL MR patients given extended TIT in the context of BFM-based protocols with long-term follow-up appeared to be comparable to studies in which a larger proportion of patients was irradiated, and was associated with low risk of CNS relapse, regardless of the WBC.
In patients with T-cell leukaemia, who are at higher risk of CNS recurrence, most investigators were more cautious and limited the elimination of pCRT to a much lower proportion of patients. In a collaborative group overview of randomized trials of CNS-directed therapy, uncertainty remained regarding T-ALL because of the small number of patients (Clarke et al, 2003).
The present study evaluated the outcome of patients with T-ALL treated with the Berlin–Frankfurt–Munster (BFM)-based protocols modified by the substitution of pCRT with extended TIT in the group of good early prednisone responders (medium risk group, MR), which accounted for the majority of T-ALL patients.
Patients and methods
Two consecutive, non-randomized, T-ALL Israel National Studies (INS) were reviewed. The studies had recruited 143 patients with T-ALL, comprising 22·7% of the total 630 newly diagnosed ALL patients aged 1–18 years, between May 1989 and 2003. Of these, 84 were enrolled in the ALL-INS 89 (May 1989–January 1998) and 59 in the ALL-INS 98 (February 1998–May 2003).
Informed consent was obtained from the guardians of all patients, and the protocols were approved by the hospital ethics committees.
The diagnosis of T-ALL was based on morphological French–American–British criteria, negative staining for myeloperoxidase or Sudan black and immunophenotypical expression of T-cell surface antigens CD7 or CD2, often with CD5, CD3, CD4, or CD8 in more than 20% of blasts and nuclear deoxynucleotide transferase (TdT) or intracytoplasmic CD3 in more than 10%. Unequivocal cases were reviewed centrally. CNS involvement at diagnosis and relapse was defined as more than five mononuclear cells/mm3 on chamber count and presence of blasts in a cytospin preparation.
Patient stratification and treatment
The T-ALL patients were assigned to MR and high-risk (HR) groups by early response to treatment (as in the ALL-BFM 90; Schrappe et al, 2000a,b). HR criteria were prednisone-poor response (PPR) after 7 d of monotherapy with prednisone and one dose of TIT, with the presence of ≥1 × 109 blasts/l in peripheral blood (PB) on day 8 and/or no bone marrow (BM) remission (M2/M3) after induction phase 1A (day 33 or day 40). The remaining patients were considered MR.
The ALL-INS 89 and ALL-INS 98 protocols for T-cell ALL were based on the ALL-BFM 86/90 (Reiter et al, 1994; Schrappe et al, 2000a,b) and ALL-BFM 95 (Möricke et al, 2008), respectively. Some modifications in systemic therapy and, mostly, in CNS preventive therapy were made in the MR group only (Table I): In the INS 89, pCRT was replaced by extended TIT (×18) in all MR patients, and in the INS 98, prompted by the results reported by the Associazione Italiana Ematologia Oncologia Pediatrica (AIEOP) (Conter et al, 1997), pCRT (12 Gy) was reindicated for patients with white blood cell count (WBC) ≥100 × 109/l. CNS involvement was treated with cranial radiation at a dose of 18 or 24 Gy. Within the MR group, 14 patients actually received more intensified protocols because of a very high WBC (n = 8) and/or high leukemic cell mass (n = 5) and/or CNS involvement (n = 4). As their outcome did not differ from the other MR patients, and because of the relatively limited total number of patients, they were included in the study.
Table I. Treatment schedule for medium risk (MR) group T-ALL INS 89 and INS 98*.
*In T-ALL INS 98 MR modification from previous INS 89 included: systemically, the omission of etoposide (VP-16) i.v., and during maintenance, the addition of six cycles every 10 weeks with vincristine 1·5 mg/m2 IV days 1, 8 and dexamethasone 6 mg/m2 PO days 1–8.
The duration of event-free survival (EFS) was defined as the time from diagnosis to the date of failure: induction failure, death, relapse, or the development of a second malignancy or until the date of last contact in complete remission. Patients who did not attain complete remission were considered failures at time zero. Overall survival (OS) was calculated from diagnosis to death (from any cause). Disease-free survival (DFS) was calculated from complete remission to the first event or the last follow-up date. For all analyses, patients lost to follow-up were censored at the time of their withdrawal. Follow-up data were updated as of December 2008. All analyses were performed on the basis of ‘intention to treat’. The Kaplan–Meier method was used to estimate survival rates; differences were compared with the log-rank test. Cumulative incidence (CI) functions were constructed by the method of Kalbfleisch and Prentice. DFS and CI rates were compared using Gray’s test (Gray, 1988) with the appropriate competing risks. Cox proportional hazards model was used for multivariate analysis. Differences in the distribution of individual parameters among patient subgroups were analysed with Fisher’s exact test. P values < 0·05 were considered significant. The R-2.6.2 package (The R Foundation for Statistical Computing) was used for analysis of the data.
The features of the 143 patients with T-ALL in the ALL-INS 89 and ALL-INS 98 trials are presented in Table II. Median age was 8·28 years. The initial median WBC was 68 × 109/l, and CNS involvement was present in 7·7%. Thirty-two of the 135 (23·7%) stratifiable patients were assigned to the HR group because of a PPR on day 8 (n = 31) or lack of BM remission on day 33 (n = 1); 103 were from the MR group. Eight patients were not stratifiable because the PB blast count on day 8 was missing (all in the ALL-INS 89 trial). Their characteristics and outcome did not differ from the patients who were stratified (results not shown).
Table II. T-ALL INS 89/98 (combined): patient characteristics, prognostic factors and, EFS for all and within risk groups.
*National Cancer Institute (NCI) risk group criteria: standard risk (SR) group: age 1 < 10 years and WBC <50 × 109/l. High risk (HR) group: all other PBd8 blasts: peripheral blood blast count on day 8.
1 < 10
Initial CNS leukaemia
<1 × 109/l
≥1 × 109/l
Patient characteristics and outcome in the ALL-INS 89 compared to the ALL-INS 98
Comparison of the patient characteristics between the two trials showed that the ALL-INS 89 had a higher incidence of CNS involvement (11·9% vs. 1·7%, P = 0·03) and of a high (≥1·7) BFM risk factor (BFM-RF), an estimate of leukemic cell mass calculated by the WBC and liver and spleen size (25% vs. 11·9%, P = 0·06) (Schrappe et al, 2000a,b; see Table SI for equation).
The treatment results of the ALL-INS 89 (median follow-up 15·2 years, range 11–19·7) and ALL-INS 98 (median follow-up 8·1 years, range 5·7–11) were similar (Table III): 5-year EFS was 61·9 ± 5·3% and 72·8 ± 5·8% respectively (P = 0·217), and OS, 69·0 ± 5·0% and 74·6 ± 5·7%, respectively (P =0·358). ALL-INS 98 was associated with a significant decrease in the CI of all relapses (10·2 ± 5% vs. 29·8 ± 5% in the ALL-INS 89, P = 0·005) (Table III), but a higher cumulative death rate in induction (3·4 ± 2·4% vs. 0%, P = 0·091) and complete remission (8·5 ± 3·7% vs. 3·6 ± 2·0%, P = 0·214) (Table III).
Table III. Comparison of treatment results between ALL INS 98 to INS 89 protocols.
CI, cumulative incidence; SE, standard error; CR, complete remission; EFS, event-free survival; OS, overall survival; CNS, central nervous system.
Death in Induction
Death in first CR
MR group [patients]
Death in induction
Death in first CR
5-year EFS [n events]
5-year OS [n events]
HR group [patients]
Death in induction
Death in first CR
Within the MR group, the incidence of CNS involvement in the ALL-INS 89 was higher than in the ALL-INS 98: 13·3% vs. 2·3% (P = 0·08). Compared with the ALL-INS 89, the ALL-INS 98 MR was associated with a trend of improved 5-year EFS, of 83·7 ± 5·6% vs. 70 ± 5·9%, (P = 0·12) (Table III, Fig 1) and a lower cumulative relapse incidence (9·3 ± 4·5% vs. 26·7 ± 5·8%, P = 0·029) but an increased cumulative death rate in induction (4·7 ± 3·2% vs. 0%, P = 0·093). CNS relapse rate did not differ [2·3 ± 2·3% vs. 5·0 ± 2·8% (P = 0·5)].
In the HR group, no significant differences in outcome were noted between the protocols (Table III, Fig 1).
Treatment results of INS 89/98 combined
The 5-year EFS of all T-ALL patients in both studies (n = 143) was 66·4 ± 3·9%, and the OS, 71·3 ± 3·8% (Table II, Table SI). Relapses occurred in 31 patients (5-year CI 21·7 ± 3·5%), all within 5 years, including three isolated CNS relapses (CI 2·1 ± 1·2%) and four combined CNS relapses (CI 2·8 ± 1·4%) (Table SI).
On univariate analysis, the adverse prognostic factors for EFS (Table II) were younger age (1 to <10 years) (P = 0·01), initial CNS involvement (P = 0·029), and most prominently, PB early response to prednisone (P < 0·001) (Table II).
On multivariate Cox regression analysis, the significant predictors of poor EFS were BFM HR group [hazard ratio 5·97, 95% confidence interval (3·05, 11·67), P < 0·01], CNS involvement [hazard ratio 5·25, 95% confidence interval (2·18, 12·63), P < 0·001] and age younger than 10 years [hazard ratio 2·59, 95% confidence interval (1·28, 5·25), P = 0·008]. Findings were not significant for WBC ≥100 × 109/l or ≥200 × 109/l.
For CNS relapses, CNS involvement at diagnosis was a significant univariate predictor (CI 27·3 ± 14·5% vs. 3 ± 1·5%, P < 0·001), whereas age, BFM-RF, BFM-risk group, and WBC ≥100 × 109/l were not.
MR group INS 89/98: impact of extended TIT replacing pCRT
Treatment modification of the ALL-BFM protocol and omission of pCRT were applied in the MR group, which accounted for 76% of the T-ALL population. The characteristics of the 103 MR patients are shown in Table II. The 5-year EFS was 75·7 ± 4·2% and the OS, 82·5 ± 3·7% (Table SI). The CI of all relapses was 19·4 ± 3·9%, and of any CNS relapses (isolated + combined), 3·8 ± 1·9%. In patients without CNS involvement at diagnosis, the CI of CNS relapses was 2·1 ± 1·5% (Table SI).
To study the impact of extended TIT or pCRT in the MR group, we analysed the outcome results for the 91 patients without initial CNS involvement who were in first remission for at least 6 months at the time pCRT was scheduled in the original ALL-BFM 86/90/95 protocols (Table IV). Twelve of them were actually irradiated, and their 5-year DFS and CI of any CNS relapse were similar to the 79 non-irradiated patients (Table IV). Most of the patients with WBC <100 × 109/l (n = 62/64) were not irradiated, and their 5-year DFS was 79·0 ± 5·2%, and CI of all relapses, 16·1 ± 4·7%, without any CNS relapse. Twenty-seven patients had WBC ≥100 × 109/l, of whom 10 were actually irradiated (12 Gy in eight, 18 Gy in two) and 17 received extended TIT. Their characteristics were similar (Table SII). Outcome was the same in the irradiated (pCRT) and non-irradiated patients (Table IV): DFS, 80·0 ± 12·6% and 82·4 ± 9·2% respectively (P = 0·79); OS (10-year), 80 ± 12·6% and 85·6 ± 9·5% (P = 0·554); cumulative any-relapse incidence, 20 ± 13·4% and 17·6 ± 9·6% (P = 0·79); any CNS relapse, 10 ± 9·5% and 5·9 ± 5·7% (one patient in each group) (P = 0·64).
Table IV. Comparison of outcome between pCRT versus extended TIT of T-ALL MR group INS 89/98 without initial CNS leukaemia and in CCR >6 months.
In the HR group, pCRT was indicated in all patients. The 5-year EFS and OS was 37·5 ± 8·6%, and the 5-year CI of all relapses and CNS relapses was 28·1 ± 8·2% and 6·3 ± 4·4%, respectively (Table SI). Seven patients in the HR group who were alive at 6 months in first remission did not actually receive pCRT, although it was indicated in all. Their 5-year DFS was lower than that of the irradiated patients (n = 14): 28·6 ± 17·1% vs. 71·4 ± 12·1% (P = 0·049).
T-cell leukaemia accounts for 10–15% (in our study, 23%) of childhood ALL. It is associated with distinct biological characteristics (Winter et al, 2007; Aifantis et al, 2008), clinical high risk features, and a historically worse prognosis and higher risk of CNS relapse than B-lineage leukaemia (Uckun et al, 1998; Pui, 2006; Pui et al, 2008). The allocation of patients with T-ALL to distinct or HR protocols, including intensive presymptomatic CNS treatment, has led to an improvement in their outcome comparable to B-lineage ALL. Although the indications for cranial irradiation have become more restricted, it is still used in a relatively high proportion of T-ALL patients. In the present studies, pCRT was omitted from large proportion (76%) of T-ALL patients, namely, the good early responders to steroids, relying on our previous favourable results for non-T ALL (Stark et al, 2000) and the introduction of intensified systemic treatment (modified ALL-BFM protocols), including high dose MTX (HD-MTX). Compared to other trials, it seems that replacing pCRT with extended TIT did not jeopardize systemic and CNS control: 5-year EFS was 61·9 ± 5·3% in the INS 89 and 72·9 ± 5·8% in the INS 98 (P = 0·21), and CI of any CNS relapse was 7·1 ± 2·8% and 1·7 ± 1·7% (P = 0·142) respectively; specifically for the non-irradiated MR group, 5-year EFS was 70 ± 5·9% in the INS 89 and 83·7 ± 5·6% in the INS 98 and CNS relapse rate was 5·0 ± 2·8% and 2·3 ± 2·3% (P = 0·5), respectively. These results (for the whole cohort of T-ALL patients and the MR group) are similar to those in the original ALL-BFM 90 (Schrappe et al, 2000a,b) and ALL-BFM 95 (Möricke et al, 2008; M. Schrappe, Department of Paediatrics, University Hospital Schleswig-Holstein, Kiel, Germany, personal communication), in which all T-ALL patients were irradiated (Table V).
Table V. CNS-preventive therapy and treatment outcome from clinical trials in the 1990s in childhood T-ALL.
CNS Invl. (%)
Indicated pCRT (%) [Gy]
5-year EFS, % (SE)
CI of any CNS Rel. % (SE)
Pts, patients; WBC, white blood count; CNS, central nervous system; Invl, involved at diagnosis; pCRT, preventive cranial radiotherapy; IT, intrathecal; HD, high dose; MTX, methotrexate; EFS, event-free survival; CI, cumulative incidence; Rel, relapse; INS, Israel National Study group; BFM, Berlin–Frankfurt–Münster; AIEOP, Associazione Italiana di Ematologia ed Oncologia Pediatrica; EORTC, European Organization for Research and Treatment of Cancer; DCLSG, Dutch Childhood Leukemia Study Group; CCG, Children’s Cancer Group; NOPHO, Nordic Society of Pediatric Hematology and Oncology; SJCRH, St Jude Children’s Research Hospital; POG, Pediatric Oncology Group; UKALL, United Kingdom Medical Research Council Working Party on Childhood Leukaemia; DFCI, Dana-Farber Cancer Institute consortium; NCI, National Cancer Institute; SR, standard risk; MR, medium risk; IR, intermediate risk; HR, high-risk; VHR, very high risk; TIT, triple intrathecal therapy; Dexa, dexamethasone; Rnd, randomization; Asparg, asparaginase; ARA, cytarabine; V, Vincristine; PGR, prednisone-good response/responder; RF, Risk Factor; vs., versus; i.v., intravenous; VANDA, Dexamethasone, Cytarabine, Mitoxantrone, Etoposide, Asparaginase and Methotrexate (IT); 6MP, 6-mercaptopurine; (i), isolated CNS relapse in patient without CNS involvement; Pred Res, prednisone response; Cont, continuation.
Using the ALL-BFM 90 protocol, three different studies omitted pCRT in T-ALL patients (Table V). In the Italian ALL-AIEOP 91 trial (Conter et al, 1997), substituting extended TIT (17 doses) for pCRT in good prednisone responders (i.e. intermediate risk group) yielded a similar EFS to the irradiated BFM 90 patients, but only in those with WBC count <100 × 109/l. Among patients with WBC ≥100 × 109/l, those who were not irradiated (n = 14) fared much worse than the irradiated patients (n = 24). By contrast, our 17 nonirradiated good prednisone responders with WBC ≥100 × 109/l had the same outcome as the 10 irradiated patients (and, indeed, as the rest of the good responders with a WBC <100 × 109/l) in terms of systemic and CNS leukaemia control. This discrepancy was hard to explain, and the number of patients was too small to reach definitive conclusions. Nevertheless, it is noteworthy that the AIEOP trial used the Erwinia product, which has proven to be less effective for ALL than E-coli asparaginase (Vilmer et al, 2000; Moghrabi et al, 2007). In the other two studies using the ALL-BFM 90, the Children’s Leukemia Cooperative Group-European Organization for Research and Treatment of Cancer (CLCG-EORTEC) 58881 (Vilmer et al, 2000) and the Dutch Childhood Leukemia Study Group (DLCSG)ALL-8 (Kamps et al, 2002), pCRT was eliminated in all T-ALL patients. The cure rate was similar to the BFM 90 (Table V), although the incidence of CNS relapse was relatively high in the EORTEC study. The results of both trials were not influenced by the WBC (</>100 × 100/l). Applying an increased-intensity augmented ‘BFM’ regimen, the Children’s Cancer Group (CCG) 1961 (Seibel et al, 2008) eliminated pCRT in the T-ALL National Cancer Institute (NCI)-HR group rapid early (BM day 7) responders (RER), and achieved better results than for their standard modified BFM (Table V). Comparable results (5-year EFS, 81·3 ± 6·9%) were noted in the INS 98 when analysed by the criteria of NCI-HR (Smith et al, 1996) in prednisone RER (comprising 73% of our MR group). The Nordic Society for Pediatric Hematology and Oncology (NOPHO) 92 (Saarinen-Pihkala et al, 2004) and Saint Jude Children’s Research Hospital (SJCRH) XIIIB (Pui et al, 2004; Pui & Howard, 2008) trials omitted pCRT from smaller proportions (around 30%) of T-ALL patients, who had a low WBC of <50 × 109/l (NOPHO 92) or <100 × 109/l (SJCRH XIIIB). Their results were in line with those of the present study and the BFM 90 and 95, respectively (Table V). In the Pediatric Oncology Group (POG) studies (9296/7/8), when extended TIT was intensified by incorporating intermediate-dose MTX, high-dose cytarabine and intensive PEG-asparaginase (Amylon et al, 1999; Laver et al, 2000; Winter et al, 2006), the EFS achieved in patients with T-ALL and WBC >50 × 109/l was similar to that of irradiated patients from historical POG studies (Pullen et al, 1999; Laver et al, 2000; Maloney et al, 2000) (Table V). In the only recent prospective randomized study, the Medical Research Council (MRC) UKALL XI (Hill et al, 2004), in which the efficacy of pCRT was compared to IT MTX with HD-MTX in T-ALL patients with WBC ≥50 × 109/l, the cure rate was the same in both arms, and did not differ from the nonirradiated patients with lower WBC (Table V). The best results (5-year EFS, 85 ± 5%) were achieved in the small group of the Dana Farber Cancer Institute (DFCI) 95-01 study (Moghrabi et al, 2007), wherein all T-ALL patients received pCRT. However, given that the findings were also better than for other protocols using pCRT in all patients (Pui et al, 2004; Möricke et al, 2008) and that the study was not randomized, we cannot differentiate the relative contribution of systemic therapy from that of pCRT (Table V).
Our trials demonstrated that, for the 75% of children with T-ALL – good early prednisone responders – regardless of the WBC, extended TIT in the context of the intensive BFM-based protocols did not compromise outcome. Given that these results were comparable to studies in which preventive irradiation was administered to a wider proportion of patients, it seems reasonable to exclude pCRT in non-HR T-ALL patients.
This manuscript is dedicated to the memory of Prof Rina Zaizov Marx, who laid the foundation for Pediatric Oncology in Israel. We thank Prof D. Steinberg for helping with the statistical analysis, and Ms Vardit Shai, Dina Kugel and Bella Lagun for processing the data.