FOR SUCCESSFUL treatment of childhood acute lymphoblastic

Transcription

1 Diagnostic Cerebrospinal Fluid Examination in Children With Acute Lymphoblastic Leukemia: Significance of Low Leukocyte Counts With Blasts or Traumatic Lumbar Puncture By Britta Bürger, Martin Zimmermann, Georg Mann, Joachim Kühl, Lutz Löning, Hansjörg Riehm, Alfred Reiter, and Martin Schrappe FOR SUCCESSFUL treatment of childhood acute lymphoblastic leukemia (ALL), it is mandatory to give sufficient therapy directed to the central nervous system (CNS) to treat subclinical or overt CNS leukemia. Without CNS-directed therapy, relapses originating from the CNS in up to 75% of cases can be expected. 1 With the introduction of effective prophylactic CNS treatment, such as intrathecal chemotherapy and cranial irradiation, up to 80% of patients in current studies are finally regarded as cured and only about 5% relapse with involvement of the CNS. 2 Patients with CNS leukemia, however, which is commonly defined as more than five leukocytes per microliter CSF in the presence of lymphoblasts after cytocentrifugation, suffer from more relapses with CNS involvement and have a significantly poorer outcome compared with CNS-negative patients. 3-5 At a time when the intensity of CNS-directed therapy is being reduced, it is even more important to identify patients with meningeal leukemia at the time of initial diagnosis, as they require more CNS-directed therapy. 6,7 From the Department of Pediatric Hematology/Oncology, Hannover Medical School, Hannover; University Children s Hospital, Würzburg; Children s Hospital, Klinikum Oldenburg; Department of Pediatric Hematology/Oncology, University of Giessen, Germany; and St. Anna Children s Hospital, Vienna, Austria. Submitted April 15, 2002; accepted October 5, Supported by grants from the Deutsche Krebshilfe. Address reprint requests to Britta Bürger, MD, Department of Pediatric Hematology/Oncology, Hannover Medical School, Carl-Neuberg-Str 1, Hannover, Germany; buerger.britta@mh-hannover.de by American Society of Clinical Oncology X/03/ /$20.00 Purpose: To determine the significance of leukemic blasts or traumatic lumbar puncture (TLP) in diagnostic CSF of children enrolled in the Berlin-Frankfurt-Münster (BFM) Acute Lymphoblastic Leukemia BFM-95 trial. Patients and Methods: A total of 2,021 patients were retrospectively evaluated according to initial central nervous system (CNS) status. Patients were classified as follows: CNS1 (CNS negative, n 1,605), CNS2 (< 5 WBC/ L CSF with blasts, n 103), CNS3 (CNS positive, n 58), (TLP with blasts, n 135), or (TLP without blasts, n 111). Patients with CNS2 and status were eligible for two additional doses of intrathecal (IT) methotrexate (MTX). CNS3 patients received additional IT MTX and cranial irradiation (18 Gy). Results: CNS2, CNS3, and groups contained a higher percentage of patients with unfavorable characteristics. Cox regression analysis identified and CNS3 status as prognostically significant (CNS3): risk ratio (RR) 2.3; 95% confidence interval [CI], 1.4 to 3.6; P.0005; :RR 1.5; 95% CI, 1.02 to 2.2; P.04. Overall 5-year event-free survival (EFS) is 79%, for CNS1 it is 80%, and for it is 83%. CNS2 patients have an EFS of 80%, but the cumulative incidence of relapses with CNS involvement is higher compared with CNS1 patients (0.10 v 0.04). patients have a significantly reduced EFS (73%, P.003) because of an increased incidence of CNS relapses. CNS3 patients suffer from more systemic and CNS relapses (EFS 50%). Conclusion: CNS2 patients have the same prognosis as patients with CNS1 status, whereas the EFS of patients is inferior to CNS1 but superior to CNS3 patients (P.001). Both subgroups may have benefitted from additional IT MTX. J Clin Oncol 21: by American Society of Clinical Oncology. There has been controversy about the significance of blasts detected in CSF without pleocytosis. Investigators from the Children s Cancer Group reported that patients with blasts in CSF preparations, in the presence of five or fewer WBCs per microliter CSF, are not at greater risk for CNS or other relapse compared with CNS-negative patients. 8,9 In contrast, investigators from St. Jude Children s Research Hospital have shown that CNS2 status ( 5 WBC/ L CSF with blasts) resulted in a higher risk of relapse and would, thus, require more intensive intrathecal therapy. 10,11 In addition, most studies looking at outcome within CNS status subgroups, to date had not evaluated that particular subset of patients with traumatic lumbar punctures at diagnosis. Gajjar et al 11 recently published data demonstrating that initial traumatic lumbar punctures (TLPs) combined with the presence of blasts negatively affect treatment outcome. In this study, we wanted to elucidate the prognostic significance of blast cells in CSF without pleocytosis and of TLPs at the time of initial diagnosis for event-free survival (EFS) and types of relapse. PATIENTS AND METHODS A total of 2,021 patients (from 85 centers in Germany, Austria, and Switzerland) with newly diagnosed ALL were enrolled in trial Acute Lymphoblastic Leukemia Berlin-Frankfurt-Münster-95 (ALL-BFM 95) of the Berlin-Frankfurt-Münster Study group between April 1995 and June Informed consent was obtained from the guardians of all patients, and the protocol was approved by local and central ethical committees. Only patients with a minimum follow-up of 2 years were included in this retrospective analysis. Nine patients had to be excluded from the analysis because of missing CNS data. According to protocol, patients underwent diagnostic lumbar puncture and received their first intrathecal (IT) methotrexate (MTX) on day 1 of induction therapy. CSF evaluation was done at the treating hospitals. A total of 974 patients (48%) had centrally reviewed CSF cytocentrifuge preparations. CNS status was defined as follows: CNS1 184 Journal of Clinical Oncology, Vol 21, No 2 (January 15), 2003: pp DOI: /JCO

2 CNS INVOLVEMENT IN CHILDREN WITH ALL 185 Table 1. Patient Characteristics According to CNS Status in the ALL-BFM 95 Trial All CNS1 CNS2 CNS3 All 2,012 (100) 1,605 (80) 103 (5) 58 (3) 135 (7) 111 (6) Sex Male 1,139 (57) 890 (56) 59 (57) 40 (69)* 79 (59) 71 (64) Female 873 (43) 715 (45) 44 (43) 18 (31)* 56 (42) 40 (36) Age 1 year 40 (2) 20 (1) 6 (6) 6 (10) 6 (4) 2 (2) 1-9 years 1,575 (78) 1,286 (80) 74 (72) 38 (66) 89 (66)* 88 (79) 10 years 397 (20) 299 (19) 23 (22) 14 (24) 40 (30)* 21 (19) WBC (per L) 20,000 1,286 (64) 1,081 (67) 44 (42)* 19 (33)* 56 (42)* 86 (78) 20, , (25) 377 (24) 37 (36)* 13 (22)* 57 (42)* 18 (16) 100, (11) 147 (9) 22 (21)* 26 (45)* 22 (16)* 7 (6) Immunology pro-b 74 (4) 51 (3) 12 (12)* 4 (7) 5 (4) 2 (2) pre-b/common 1,586 (79) 1,299 (81) 67 (65) 26 (45) 104 (77) 90 (81) T-ALL 256 (13) 178 (11) 22 (21)* 28 (48)* 20 (15) 8 (7) Predicted response Good 1,807 (90) 1,453 (91) 89 (86) 40 (69)* 117 (87) 108 (97) Poor 170 (8) 123 (8) 13 (13) 15 (26)* 16 (12) 3 (3) Risk Group SR 694 (35) 592 (37) 22 (21)* 8 (14)* 24 (18)* 48 (43) MR 1,073 (53) 833 (52) 63 (61)* 31 (53)* 92 (68)* 54 (49) HR 245 (12) 180 (11) 18 (18)* 19 (33)* 19 (14)* 9 (8) NOTE. Percentages are in parentheses. In the first column, where percentages refer to the whole group, all other percentages are calculated within the CNS status groups. Statistically significant differences, compared with the CNS1 group, are marked with an asterisk ( 2 test for categorized variables, Wilcoxon rank sum test for continuous variables; eg, WBC count). Abbreviations: TLP, traumatic lumbar puncture; SR, standard risk; MR, medium risk; HR, high risk. (puncture nontraumatic without leukemic blasts after cytocentrifugation), CNS2 (puncture nontraumatic, 5 WBC/ L CSF with identifiable blasts), CNS3 (puncture nontraumatic, 5 WBC/ L CSF with identifiable blasts), (TLP with blasts), and (TLP without blasts). CNS3 definition was derived from Mastrangelo et al. 12 A TLP was defined as 10 or more erythrocytes per microliter CSF or macroscopically contaminated CSF. In addition to the CNS3 group as defined above, patients with a cerebral mass or patients with cranial nerve palsy in combination with blasts after cytocentrifugation were regarded as having CNS3 disease. Treatment Patients were treated according to the protocols of the ALL-BFM 95 trial. In contrast to the ALL-BFM 90 trial, which has been described previously, 13 a new stratification system was introduced. In ALL-BFM 95, patients were stratified according to age, initial WBC count, day 8 response to prednisone, T immunology, and molecular rearrangements such as t(9;22) and t(4;11) into standardrisk (SR), medium-risk (MR), and high-risk (HR) groups. SR and MR therapy consisted of an eight-drug induction, consolidation with four times high-dose (HD) MTX, and an eight-drug reintensification, followed by maintenance therapy. HR patients were treated with a five-drug induction, followed by six intensive multiagent blocks and the identical reintensification as SR and MR patients. CNS-directed therapy consisted of 11 doses of IT MTX for both SR and MR, whereas HR patients received five doses of IT MTX and six doses of triple intrathecal therapy. Prophylactic cranial irradiation was given after reintensification only to patients with T immunology or to HR patients (12 Gy). Patients with CNS2 and disease received two additional doses of IT MTX during induction. Some patients (n 7) were treated as having CNS3 disease although CSF findings were not consistent with the CNS3 definition described previously. Patients with CNS3 status received four (five for HR patients) additional doses of IT therapy and therapeutic cranial irradiation with 18 Gy. Maintenance therapy was without IT therapy. Statistical Analysis The duration of EFS is defined as the time from diagnosis until the date of the first adverse event (relapse, death for any reason, or the development of a second malignancy) or, if no such event occurred, until the date of last contact. Patients who did not attain a complete remission were considered failures at time zero. Duration of disease-free survival for patients who achieve remission is defined as the time from attainment of a complete remission until the date of an adverse event (relapse, death, second malignancy) or, if no such event occurred, until the date of last contact. Distributions of EFS and continuous complete remission were estimated by the methods of Kaplan and Meier with SE according to Greenwood, 14 and were compared using the log-rank test. 14,15 All analyses were performed on the basis of intent-to-treat. Cumulative incidence functions of CNS relapse were constructed by the method of Kalbfleisch and Prentice 16 for patients who achieved a complete remission. 16 Incidence functions for all competing factors of failure were also calculated. Functions were compared with Gray s test. 17 A CNS relapse is defined as an isolated CNS relapse or a CNS relapse in combination with another type of relapse or failure. Differences in the distribution of variables among patient subsets were analyzed using the 2 test for categorized variables and the Wilcoxon rank sum test for continuous variables. The prognostic relevance of the different CNS groups compared with CNS-negative patients was examined by Cox regression analysis with known prognostic factors as covariables. 18 The database for all analyses was frozen, meaning that no data about patients who became known to the study center were entered, on August 1, RESULTS The number of patients classified into each of the different CNS status groups is shown in Table 1. CNS1 status was diagnosed in 79.77% of patients; 2.88% were found to have CNS leukemia (CNS3). The remaining 17.34% of patients are divided among the other groups: CNS2 (5.12%), (6.71%), and (5.51%). Among the 58 patients diagnosed with CNS3 status, six were defined as having CNS leukemia by demonstration of a cerebral/meningeal mass only, and three could be defined as having CNS3 status by presence of blasts without CSF pleocytosis in combination with cranial nerve palsy. Evaluation of the patient characteristics among the different CNS status groups revealed that CNS2, CNS3, and

3 186 BÜRGER ET AL Table 2. CNS Treatment According to CNS Status All (n 2,012) CNS1 (n 1,605) CNS2 (n 103) CNS3 (n 58) (n 135) (n 111) Treatment, CNS negative 1,716 1, Additional IT MTX in protocol I Treatment, CNS positive Unknown Abbreviations: TLP, traumatic lumbar puncture; IT MTX, intrathecal methotrexate; CNS, central nervous system. groups include a higher percentage of unfavorable characteristics, as shown in Table 1. Compared with CNS1 patients, CNS2, CNS3, and groups had significantly higher WBC counts at diagnosis (Wilcoxon test, P.0001) and significantly fewer SR patients and more HR patients ( 2 test, P.01). A statistically significant difference in platelet count could not be identified (median platelet count in patients: 46,500/ L; median platelet count in CNS1 patients: 55,000/ L; Wilcoxon test, P.13). Regarding immunology, in the CNS2 group, there were significantly more patients with T- and pro-b-all. T- lineage ALL was significantly more frequent in CNS3 patients ( 2 test, P.01). The CNS-directed therapy in the different groups is shown in Table 2. The 5-year EFS estimates ( SE) for patients in each group were as follows: CNS1, 80% ( 1%); CNS2, 80% ( 4%); CNS3, 50% ( 8%);, 73% ( 4%); and, 83% ( 4%; Fig 1). The EFS for CNS1, CNS2, and groups were almost identical at 80%. Compared with CNS1 patients, the prognosis of the group was significantly worse (73%, P.003). CNS3 status had the worst prognosis, with a 5-year EFS estimate of 50% (P.0001). If analyzed by the Cox regression analysis (covariables: age 10 years, sex, WBC count 100,000, response to prednisone, and T immunology versus B-precursor immunology), and CNS3 status maintain prognostic significance (: RR 1.5; 95% confidence interval, 1.02 to 2.2; P.04; CNS3: RR 2.3; 95% confidence interval, 1.4 to 3.6; P.0005). Subset analysis for patients with T-ALL showed a result similar to that for the overall CNS status groups, but because of small numbers, these differences were not statistically significant. Regarding CNS-directed therapy, we compared the patients (B-precursor and T-ALL combined) treated as CNS negative (n 36) with those who received two additional doses of IT MTX (n 88). No statistically significant difference could be demonstrated, although there was a tendency toward higher EFS with additional IT MTX (77% [SE 5%] v 66% [SE 9%], P.69). The same was also observed for the CNS2 group: 73% EFS (SE 10%) for patients treated as CNS negative (n 26) versus 82% EFS (SE 5%) for patients treated with two additional doses of IT MTX (n 76, P.36). The distribution of relapses within the different CNS status groups is shown in Table 3. From 2,012 patients, 20 were excluded from the analysis because of death before complete remission (n 15) or nonresponse (n 5). For CNS1 patients, the cumulative incidence (CI) for any relapse is 0.17, of which the CI for an isolated or combined CNS relapse is 0.04 and for other relapses (bone marrow, testes, other) it is This distribution is also found among patients. For CNS2 patients, the overall CI for relapses is identical to that of the CNS1 group. However, among the CNS2 group, there is a higher proportion of relapses with CNS involvement (CI 0.10) and a reduction in CI regarding other relapses (0.07). For the group, an increase in CI for all relapses is observed (0.20), the CI for relapses with CNS involvement (0.08) being comparable to the CNS2 group, whereas the CI for other relapses (0.12) is closer to the CNS1 group. As expected, the CNS3 status patients experience the highest overall relapse rate (CI 0.35), relapses without CNS involvement being the major problem in this subgroup (CI 0.22). DISCUSSION Fig 1. Event-free survival (EFS) for the five CNS status groups is shown. Only patients enrolled in the ALL-BFM 95 trial with a minimum follow-up of 2 years were evaluated. The EFS of patients is significantly inferior to CNS1 patients (P.003), but superior to CNS3 patients (P.001). In this study, we have shown that patients with five or fewer WBCs per microliter in initial diagnostic CSF with blasts (CNS2) have the same prognosis regarding EFS as CNSnegative patients if treated with two additional doses of IT MTX (EFS for both groups, 80%; SE 1% and 4%, respectively). patients had the same overall outcome (EFS 83%) as CNS1 and CNS2 groups. patients, however, were found to have an EFS of 73%, which is inferior to that of CNS1 patients (P.003) but superior to CNS3 patients (EFS 50%, P.001). The distribution of patients within CNS status groups is similar to the distribution reported by Gajjar et al, 11 although group sizes of CNS2,, and patients are smaller in our study (5%, 6%, and 7% v 15%, 10%, and 11%). This might be the effect of our study being a multicenter study with 85 participating institutions in Germany, Austria, and Switzerland,

4 CNS INVOLVEMENT IN CHILDREN WITH ALL 187 Table 3. Distribution of Relapses According to CNS Status Groups in ALL-BFM 95 Trial All (n 2,012) CNS1 (n 1,605) CNS2 (n 103) CNS3 (n 58) (n 135) (n 111) All relapses 269 (0.17) 203 (0.17) 15 (0.17) 16 (0.35) 23 (0.20) 12 (0.12) Relapses with CNS involvement 71 (0.043) 44 (0.035) 8 (0.10)* 7 (0.13)* 10 (0.08)* 2 (0.02) Isolated CNS relapse BM CNS relapse Other relapses 198 (0.131) 159 (0.134) 7 (0.07) 9 (0.22) 13 (0.12) 10 (0.10) Other events (death during CCR, second malignancy, LFU) 61 (0.03) 42 (0.03) 2 (0.02) 5 (0.10) 6 (0.05) 6 (0.06) NOTE. Cumulative incidences are given in parentheses. Statistically significant differences compared with CNS1 patients are marked with an asterisk. Abbreviations: BM, bone marrow; CCR, continuous complete remission; LFU, lost to follow up in CCR; CNS, central nervous system. in which CSF samples have not always been centrally evaluated. Another explanation could lie in the fact that thrombocytopenic patients in most participating institutions of the ALL-BFM trials usually receive platelet transfusions before lumbar puncture, especially if platelet counts are below 50,000/ L, although serious complications are an extremely rare event. 19 However, the size of the CNS3 group (3%) is identical in both studies and matches that of most other reported trials, indicating that a multicenter approach allows a valid data collection. The CNS2 group contains higher percentages of patients with unfavorable prognostic parameters. In this group, there are fewer SR and more HR patients, more T and pro-b immunology, and more patients with higher WBC counts at diagnosis. The CSF findings can be interpreted as an expression of minimal meningeal leukemia, which might correlate with a poorer prognosis. However, with additional IT therapy and a total of four courses HD MTX, these patients have a good prognosis, as previously described by the Children s Cancer Group and van den Berg et al. 8,9,20 Our result of an EFS of 80% is strikingly different than the experience from St. Jude s Hospital, where CNS2 patients with regular CNS therapy had a survival of only 55% 11 or 53%, 10 respectively. This result is remarkable, as the number of IT chemotherapy applications (nine to 13 times for lower-risk and 13 to 20 times for higher-risk patients) is comparable to trial ALL-BFM 95, and the percentage of patients receiving cranial radiotherapy (RT) is higher in the St. Jude trials (21% in BFM-ALL 95 v 63% in study XI and 30% in study XII). 5 One explanation for this difference might be the fact that, in the St. Jude trials, IT therapy was generally not given at the time of the diagnostic lumbar puncture, but 24 to 48 hours later, and the second IT therapy was not given until 3 weeks later, whereas in our study, the second IT MTX was always instilled on day 12. One could also argue that systemic treatment has improved with time; thus, a comparison between trials from 1984 up to 1991 and one starting in 1995 is not appropriate. However, in the ALL-BFM 90 trial (1990 to 1995), we found that survival of patients with CNS2 status was identical to the EFS in CNS1 and groups (80%, 80%, and 79% respectively; CNS2 group n 30; M. Schrappe, unpublished data). In a subsequent study of 165 patients from 1991 to 1994, St. Jude investigators reported that early intensification of IT therapy resulted in a reduced cumulative risk for all relapses with CNS involvement of 4.4%. 21 This result, which is similar to our overall cumulative incidence for relapses with CNS involvement (CI 0.043), was achieved, first, by intensifying IT therapy early in induction (weekly for CNS2, CNS3, and patients) and, second, by increasing the total number of IT chemotherapy applications from 21 to 26. Another difference between the ALL-BFM 95 trial and the St. Jude studies XI and XII is the number of courses with HD MTX. In the BFM trial, patients are exposed to HD MTX (5 g/m 2 ) four times, whereas patients received two courses HD MTX (2 g/m 2 ) in study XI and five courses of MTX (1.5 g/m 2 ) in study XII. 10,22 Compared with CNS1 patients, of whom the majority are suffering systemic relapses, the relapses among CNS2 patients seem to be more equally divided between those with CNS involvement and all other relapses. This distribution (50% relapses with CNS involvement, 50% bone marrow and other relapses) was also found by Children s Cancer Group investigators. 9 As overall outcome is identical when compared with the CNS1 group, a modification of the treatment strategy for CNS2 patients does not seem mandatory. It could be of interest, however, to examine whether the percentage of blasts in CSF differential cell counts can identify a subgroup among CNS2 patients with an inferior prognosis compared with the overall CNS2 group. The outcome for the group also differs from the results reported by Gajjar et al, 11 although the variation is not as pronounced as for the CNS2 group. Reasons for this may be the time interval between diagnostic and TLP and the interval to the second instillation of IT therapy. One has to ask, however, why patients have an inferior outcome, especially when compared with CNS2 patients. One explanation could be that this difference is the result of an iatrogenic introduction of blasts into the CSF. But it is not easy to understand why a single, iatrogenic introduction of blasts should cause a difference in overall outcome, when patients with minimal meningeal leukemia, as in the CNS2 group, can be handled quite easily with two additional applications of IT MTX. Another hypothesis to explain the inferior outcome in patients could be more advanced disease, for example, with more perivenular or parameningeal leukemic infiltrates, or a biologically different disease, permitting migration of leukemic blasts more easily. Bleyer 23 reported that certain types of leukemia penetrate the CNS with a higher rate (eg, T-cell leukemias) and that other factors such as blast and platelet count, patient s age, or maturity of the blood-brain barrier are factors that influence ingress of leukemic cells into the CNS. We did not find significantly more T-ALL or lower platelet counts among the patients, but we did find a significant number of patients had higher initial WBC counts compared with CNS1 patients. But by Cox regres-

5 188 BÜRGER ET AL sion analysis, status maintains prognostic significance, indicating that the higher incidence of unfavorable patient characteristics cannot account for the observed difference in EFS. Also, when comparing patients with those in the CNS2 group, which has a good prognosis, patients do not seem to have more advanced disease. On the contrary, the CNS2 group comprises more patients with initial hyperleukocytosis, T immunophenotype, and prednisone poor-response or high-risk disease than does the group. A third explanation for the inferior EFS in the group could be that, among patients, there are hidden CNS3 patients who are simply not identified at the time of initial diagnosis. Because the other possible explanations for the inferior outcome in patients seem unlikely, we favor the latter possibility unidentified CNS3 patients as the most likely explanation for the observed difference in outcome. How can we identify true CNS3 patients within the group? First, one could compare the CSF with the peripheral blood differential cell count. If the percentage of blasts in the CSF is significantly higher than in the peripheral blood, CNS3 disease is likely. One could also compare the ratio of erythrocytes to leukocytes in the CSF and in peripheral blood. If this ratio is smaller in the CSF compared with peripheral blood, CNS3 status might also be assumed. In our study, seven patients in the group were treated as having CNS disease because of their CSF findings. Among those, two suffered a relapse. Although these numbers are small, one could expect more relapses if this group consisted of true CNS3 patients. What can we do for patients? An introduction of cranial RT for all patients with B-precursor ALL in this subgroup still remains an option in trying to improve outcome, as the increased incidence of CNS relapses is the major problem among patients. However, cranial RT might lead to significant toxicity or second malignancies in patients who otherwise have a good prognosis. There was also no statistically significant difference in patients treated as CNS negative (EFS 66%) versus those treated with additional IT MTX (EFS 77%). However, we will continue to treat CNS2 and patients with two additional doses of IT MTX. In conclusion, this study has shown that CNS2 patients have a good prognosis (EFS 80%) that does not require any intensification of CNS-directed therapy. patients have an inferior prognosis compared with the CNS1 group (73% v 80%) but a superior EFS compared with the CNS3 group (50%). In our opinion, further intensification of CNS-directed therapy is not mandatory at this time if the treatment schedule is comparable to the ALL-BFM strategy. Evaluation of the ongoing ALL-BFM 2000 trial for CNS status and survival will further clarify the prognostic significance of CNS2 and findings. ACKNOWLEDGMENT We thank E. Odenwald, B. Puttkamer, and T. Büchner for expert cytology; the data managers N. Götz, U. Meyer, I. Krämer, and K. Mischke for precise data management; and the nurses and doctors of all participating hospitals. 1. Evans AE, Gilbert ES, Zandstra R: The increasing incidence of central nervous system leukemia in children. Cancer 26: , Reiter A, Schrappe M, Ludwig W-D, et al: Chemotherapy in 998 unselected childhood acute lymphoblastic leukemia patients. Results and conclusions of the multicenter trial ALL-BFM 86. Blood 84: , Schrappe M, Reiter A, Zimmermann M, et al: Long-term results of four consecutive trials in childhood ALL performed by the ALL-BFM study group from 1981 to Leukemia 14: , Silverman LB, Declerck L, Gelber RD, et al: Results of Dana-Farber Cancer Institute Consortium protocols for children with newly diagnosed acute lymphoblastic leukemia ( ). Leukemia 14: , Pui CH, Boyett JM, Rivera GK, et al: Long-term results of Total Therapy studies 11, 12 and 13A for childhood acute lymphoblastic leukemia at St Jude Children s Research Hospital. Leukemia 14: , Schrappe M, Reiter A, Henze G, et al: Prevention of CNS recurrence in childhood ALL: Results with reduced radiotherapy combined with CNS-directed chemotherapy in four consecutive ALL-BFM trials. Klin Paediatr 210: , Schrappe M, Reiter A, Riehm H: Prophylaxis and treatment of neoplastic meningeosis in childhood acute lymphoblastic leukemia. J Neurooncol 38: , Gilchrist G, Tubergen D, Sather H, et al: Low numbers of CSF blasts at diagnosis do not predict for the development of CNS leukemia in children with intermediate-risk acute lymphoblastic leukemia: A Children s Cancer Group report. J Clin Oncol 12: , Tubergen D, Cullen J, Boyett J, et al: Blasts in CSF with a normal cell count do not justify alteration of therapy for acute lymphoblastic leukemia in remission: A Children s Cancer Group study. J Clin Oncol 12: , Mahmoud H, Rivera G, Hancock M, et al: Low leukocyte counts with blast cells in cerebrospinal fluid of children with newly diagnosed acute lymphoblastic leukemia. N Engl J Med 329: , Gajjar A, Harrison P, Dandlund J, et al: Traumatic lumbar puncture at diagnosis adversely affects outcome in childhood acute lymphoblastic leukemia. Blood 96: , 2000 REFERENCES 12. Mastrangelo R, Poplack D, Bleyer A, et al: Report and recommendations of the Rome workshop concerning poor-prognosis acute lymphoblastic leukemia in children: Biologic bases for staging, stratification, and treatment. Med Pediatr Oncol 14: , Schrappe M, Reiter A, Ludwig W-D, et al: Improved outcome in childhood ALL despite reduced use of anthracyclines and of cranial radiotherapy: Results of trial ALL-BFM 90. Blood 95: , Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53: , Mantel N: Evaluation of survival data and two new rank order statistics arising in its consideration. Cancer Chemother Rep 50: , Kalbfleisch JD, Prentice RL: The Statistical Analysis of Failure Time Data. New York, John Wiley, 1980, pp Gray RJ: A class of K-sample tests for comparing the cumulative incidence of a competing risk. Ann Stat 16: , Cox DR: Regression models and life tables. J R Stat Soc 34:187, Howard SC, Gajjar A, Ribeiro RC, et al: Safety of lumbar puncture for children with acute lymphoblastic leukemia and thrombocytopenia. JAMA 284: , Van den Berg H, Vet R, Den Ouden E, et al: Significance of lymphoblasts in cerebrospinal fluid in newly diagnosed pediatric acute lymphoblastic malignancies with bone marrow involvement: Possible benefit of dexamethasone. Med Pediatr Oncol 25:22-27, Pui CH, Mahmoud HH, Rivera GK, et al: Early intensification of intrathecal chemotherapy virtually eliminates central nervous system relapse in children with acute lymphoblastic leukemia. Blood 92: , Evans WE, Relling M, Rodman J, et al: Conventional compared with individualized chemotherapy for childhood acute lymphoblastic leukemia. N Engl J Med 338: , Bleyer WA: Biology and pathogenesis of CNS leukemia. Am J Pediatr Hematol Oncol 11:57-63, 1989