|Home | About | Journals | Submit | Contact Us | Français|
We assessed the long-term outcome of patients enrolled on CCG-3891, a high-risk neuroblastoma study in which patients were randomly assigned to undergo autologous purged bone marrow transplantation (ABMT) or to receive chemotherapy, and subsequent treatment with 13-cis-retinoic acid (cis-RA).
Patients received the same induction chemotherapy, with random assignment (N = 379) to consolidation with myeloablative chemotherapy, total-body irradiation, and ABMT versus three cycles of intensive chemotherapy. Patients who completed consolidation without disease progression were randomly assigned to receive no further therapy or cis-RA for 6 months.
The event-free survival (EFS) for patients randomly assigned to ABMT was significantly higher than those randomly assigned to chemotherapy; the 5-year EFS (mean ± SE) was 30% ± 4% versus 19% ± 3%, respectively (P = .04). The 5-year EFS (42% ± 5% v 31% ± 5%) from the time of second random assignment was higher for cis-RA than for no further therapy, though it was not significant. Overall survival (OS) was significantly higher for each random assignment by a test of the log(−log(.)) transformation of the survival estimates at 5 years (P < .01). The 5-year OS from the second random assignment of patients who underwent both random assignments and who were assigned to ABMT/cis-RA was 59% ± 8%; for ABMT/no cis-RA, it was 41% ± 7%; for continuing chemotherapy/cis-RA, it was 38% ± 7%; and for chemotherapy/no cis-RA, it was 36% ± 7%.
Myeloablative therapy and autologous hematopoietic cell rescue result in significantly better 5-year EFS and OS than nonmyeloablative chemotherapy; cis-RA given after consolidation independently results in significantly improved OS.
Children with high-risk neuroblastoma have poor long-term survival despite intensive multimodal treatment.(1) Previous studies attributed improvement in outcome to the use of myeloablative cytotoxic therapy with autologous bone marrow transplantation (ABMT). However, because these studies were not prospectively randomized, selection bias may have influenced the results.2-4
From 1991 to 1996, the Children's Cancer Group compared myeloablative chemotherapy and radiotherapy plus purged ABMT with continuing intensive chemotherapy without ABMT in a randomized study. Because of the high risk of relapse from minimal residual disease, each cohort underwent a second random assignment to test the efficacy of 13-cis-retinoic acid (cis-RA), a known differentiating agent for neuroblastoma, to prevent relapse.5-7 The initial outcome, which was reported in 1999, showed a significant difference in 3-year event-free survival (EFS) for the group randomly assigned to myeloablative therapy and ABMT compared with the group randomly assigned to continuing chemotherapy (CC), although overall survival (OS) was not significantly different.8 There was also a statistically significant advantage in EFS for patients randomly assigned to cis-RA, regardless of whether they had undergone ABMT or received CC. This report describes long-term outcome, more than 8 years after completion of accrual, with analysis of prognostic factors.
Of 539 eligible patients who had newly diagnosed high-risk neuroblastoma and who were aged 1 to 18 years, 434 patients had stage 4 neuroblastoma; 72 had stage 3 neuroblastoma with one or more of the following: MYCN oncogene amplification, serum ferritin ≥143 ng/mL, and unfavorable histopathology; one patient (aged older than 1 year) had stage 2 neuroblastoma with MYCN amplification; 13 patients had stage 1 or 2 disease but developed bone metastases after resection only; and 19 patients younger than 1 year had stage 4 disease and MYCN amplification. Parents or guardians provided written informed consent, and the study was approved by local institutional review boards. Enrollment occurred between January 1991 and April 1996.
All patients received five cycles of induction with cisplatin, doxorubicin, etoposide, and cyclophosphamide plus surgery and received radiotherapy for residual local and metastatic disease, as previously described.8 Patients in the ABMT group received carboplatin (1,000 mg/m2) and etoposide (640 mg/m2) as continuous infusions over 96 hours; melphalan (140 mg/m2 on day −7 and 70 mg/m2 on day −6); and total-body irradiation (3.33 Gy daily on days −3, −2, and −1) followed by infusion of immunomagnetically purged bone marrow and granulocyte-macrophage colony-stimulating factor. Patients in the CC group received three cycles of cisplatin (160 mg/m2), etoposide (500 mg/m2), and doxorubicin (40 mg/m2), as continuous infusions over 96 hours with bolus ifosfamide (2,500 mg/m2 on days 0 to 3) and granulocyte colony-stimulating factor. The first random assignment was performed just before the third cycle of induction therapy (median, 60 days after diagnosis) for patients without progressive disease (PD). After ABMT or CC (median, 288 days after diagnosis), patients without PD or histologically confirmed disease were randomly assigned to receive six cycles of cis-RA (160 mg/m2/d in two divided doses for 14 days every 28 days) or no further therapy. Disease extent was assessed by international neuroblastoma response criteria at diagnosis, end of induction, after CC or ABMT, and after cis-RA.9
The study design included two separate sequential random assignments in a quasi-factorial design. Patients with PD before week 8 were ineligible for the first random assignment. Patients with PD or histologically confirmed disease at the completion of ABMT or CC were ineligible for the second random assignment. Patients ineligible for the first random assignment were non–randomly assigned to CC (NRCC). If these patients remained progression free without documented tumor after CC, they were eligible for the second random assignment but were not included in the intention-to-treat analysis of the first random assignment.
A permuted-block design was used for random assignment of approximately equal numbers of patients from each of two strata (stage 3 or 4) to ABMT or CC. The second random assignment was balanced with respect to patients from each group of the first random assignment and non–randomly assigned patients who were ineligible for transplantation.
Similarities between groups were assessed with Fisher's exact test. Outcome analyses used life-table methods according to the Kaplan-Meier method and associated statistics.10 SEs of the life-table estimates of EFS and OS were calculated according to Peto et al.11 The primary end point was EFS calculated from time of random assignment. An event was defined as relapse, disease progression, death, or second neoplasm. A two-sided log-rank test, robust to deviations from proportional hazards, was used to compare survival curves. For situations in which the assumption of proportional hazards was obviously violated, survival curves were compared at a fixed point in time by using a two-sided test of a log(−log(.)) transformation of the survival estimates at the 5-year time point.12 Treatment regimens were compared by intention-to-treat analyses. A Cox proportional hazards model was used to identify prognostic factors. P less than .05 was considered statistically significant except for the two-sided test of the log(−log(.)) transformation of the survival estimates, in which P < .025 was considered statistically significant.
Of 539 eligible patients, 379 participated in the first random assignment, and 258 participated in the second. Patient characteristics of the randomly assigned groups were comparable at baseline (P > .05; Appendix Table A1, online only). Patients in the group that was non–randomly assigned to chemotherapy differed from randomly assigned patients in the higher percentage of stage 3 (21% v 12%; P = .02) and lower median age (2.6 v 3.1 years; P = .01; Appendix Table A1). Detailed characteristics and outcomes for patients with stage 3 disease are reported in a separate manuscript. Outcomes for all randomly assigned high-risk patients are reported below.
Of the 539 eligible patients, 190 were randomly assigned to CC, and 189 were assigned to ABMT. Of those assigned to ABMT, 129 actually received ABMT according to protocol; of those assigned to CC, 150 received chemotherapy according to protocol. The assigned treatment was not received in 100 patients. Fifty-two patients had PD before they could receive assigned therapy (28 in CC group; 24 in ABMT). Two patients died after random assignment but before they could receive the assigned treatment. The remaining 46 patients did not receive assigned therapy because of physician or parent decision. Among patients without PD, compliance with the first random assignment was 86%. All analyses were performed on the basis of the assigned treatment.
A total of 319 patients completed consolidation therapy without PD. Of these, 130 were randomly assigned to cis-RA, and 128 were assigned to no further therapy. Of the remaining 61 patients, 37 had histologically documented residual disease and were nonrandomly assigned to cis-RA. The parents of 24 patients declined the second random assignment, so those patients are not analyzed here. Fifty-two percent of the patients who were randomly assigned to ABMT and 55% of those who were randomly assigned to CC underwent the second random assignment (98% compliance).
The overall 5-year EFS and OS rates ± SE from the time of enrollment for all 539 patients were 26% ± 2% and 36% ± 2%, respectively (Fig 1A). Survival rates for patients with stage 4 disease were slightly lower than the overall cohort (Fig 1B). The median follow-up time of patients alive without an event was 7.7 years (range, 130 days to 12.8 years). The data on which these analyses are based include 91.3% of the expected information (ie, the number of events expected).
The EFS for patients randomly assigned to ABMT was significantly higher than those randomly assigned to CC, as 5-year EFS rates were 30% ± 4% and 19% ± 3%, respectively (P = .0434; Fig 2A). By using the log-rank test to measure the overall difference between the curves, the OS for patients in the ABMT group was not statistically significantly higher than for the patients in the CC group; the 5-year OS rates from the time of random assignment were 39% ± 4% and 30% ± 4%, respectively (P = .3917; Fig 2B). However, in patients who survived more than 3 years, there appears to be a benefit of ABMT. At 5 years, the curves are significantly different (P < .0001) in a test of the log(−log(.)) transformation of the survival estimates. The results were similar within stage 4 patients; EFS and OS rates for ABMT were 26% ± 4% and 37% ± 4%, respectively, versus 16% ± 3% and 28% ± 4%, respectively, for CC, though there were no statistically significant differences when using a log-rank test between any treatment groups.
There was a trend that associated improved EFS for patients who were randomly assigned to cis-RA versus no cis-RA, though it was not statistically significant; 5-year EFS rates from time of random assignment were 42% ± 5% versus 31% ± 5%, respectively (P = .1219; Fig 3A). Both the cis-RA and no cis-RA groups had EFS rates that were significantly higher than patients with documented residual disease who were non–randomly assigned to cis-RA, who had a 5-year EFS rate of 6% ± 4% (P = .0025 and P < .0001, respectively). The trend for a better outcome with cis-RA continued for OS. Although the OS curve for patients who received cis-RA was not statistically significantly higher than that of the patients who received no cis-RA, (5-year OS from the time of random assignment, 50% ± 5% v 39% ± 5%, respectively; P = .1946; Fig 3B), the OS for cis-RA was significantly higher in a test of the log(−log(.)) transformation of the survival estimates at 5 years (P = .0006). Both the cis-RA and no cis-RA group had OS rates that were significantly higher than the patients who were nonrandomly assigned to cis-RA (n = 37) and who had a 5-year OS rate of 21% ± 8% (P = .0052 and P = .0003, respectively). Within stage 4, the 5-year EFS rates for cis-RA versus no cis-RA were 36% ± 5% and 26% ± 4%, respectively (P = .2577), and the 5-year OS rates for cis-RA versus no cis-RA were 44% ± 5% and 36% ± 5%, respectively (P = .4538).
Figure 4displays the EFS for the patient subset that participated in both random assignments and identifies the time to an event from the time of second random assignment. The 5-year EFS rate for patients who were randomly assigned to ABMT and cis-RA (n = 50) was 50% ± 8% and was 20% ± 6% for patients assigned to CC and no cis-RA (n = 53; P = .0038). The 5-year OS rate for patients who were randomly assigned to ABMT and cis-RA (n = 50) was 59% ± 8% and was 41% ± 7% for patients assigned to ABMT without cis-RA (n = 48); the rate was 38% ± 7% for patients assigned to CC and cis-RA (n = 52) and was 36% ± 7% for patients assigned to CC and no cis-RA (n = 53; P = .054 for the OS of ABMT/cis-RA v CC/no cis-RA). There was no evidence of statistical interaction between treatments of the second random assignment and the first random assignment (P = .8239), and the order of the treatment group outcomes makes clinical sense. Treatment with the combination of ABMT and cis-RA appears superior to the other combinations. This should be interpreted with caution, because the study was not designed to identify a superior treatment combination in this fashion.
Of 462 patients who were assessable for induction response, 265 (57%) were complete response (CR)/very good partial response (VGPR), and 197 (43%) were at least partial response (PR). The EFS for the CR/VGPR patients was significantly greater than that of the patients who were at least PR (P < .0001; Table 1). Within each treatment group in the first phase of the study, CR/VGPR patients had statistically significantly higher EFS at P < .0001. However, the CR/VGPR patients did not have statistically significantly better EFS rates than the patients who were at least PR within the treatment groups of the second phase of the study that tested cis-RA (Table 1).
Of the 405 first events, 378 were relapse or PD. Of these 378 patients, 67 patients underwent ABMT after going off protocol therapy and had a 5-year OS of 17% ± 5%.
The potential prognostic factors analyzed included International Neuroblastoma Staging System stage, age, MYCN status, pathology, ferritin, and response to induction chemotherapy (Appendix Table A2, online only). In both the overall patient cohort and the subset of patients with stage 4 disease, factors prognostic of lower EFS and OS rates in univariate analyses included stage 4 disease, MYCN gene amplification, unfavorable tumor histopathology, elevated serum ferritin (> 143 ng/mL), and less than a very good partial response (< VGPR) to induction chemotherapy. Older age (< 547 days [n = 92] v ≥ 547 days [n = 447])13 was prognostic of lower EFS in the overall cohort and in the patients with stage 4 disease, but it was not prognostic of lower OS in either cohort (Appendix Table A2).
Multivariable analyses revealed that response (P < .0001), MYCN status (P < .0001), histology (P = .0064), stage (P = .0016), and ferritin level (P = .0392) were prognostic of EFS in the 247 patients with complete data for all factors, whereas only response (P < .0001), MYCN status (P < .0001), histology (P = .0131), and stage (P = .0232) were prognostic of OS in 269 patients with complete data. For International Neuroblastoma Staging System stage 4, response, MYCN status, and histology were prognostic of both EFS and OS in the 222 patients with complete data. Multivariable analyses of prognostic variables within each treatment arm showed that MYCN amplification was the factor most often prognostic of worse outcome (Table 2). Neither age ≥ 547 days nor unfavorable histology were prognostic in any of the arms.
Treatment-related toxicity data were unchanged from the prior report.8 Second malignant neoplasms occurred in four patients, which included one patient with each of the following: T-cell acute lymphoblastic leukemia 5.0 years after study enrollment (not a first event) in a patient randomly assigned to CC; acute myeloblastic leukemia as a first event 2.7 years after enrollment in a patient randomly assigned to ABMT; clear-cell carcinoma as a first event at 2.5 years in a patient non–randomly assigned to CC; and follicular carcinoma of the thyroid as a first event at 7.3 years in a patient randomly assigned to ABMT. Only the last of these patients received cis-RA.
Of 365 deaths, causes were progressive disease (n = 297), treatment (n = 57), and unknown (n = 11). Fifteen of the treatment-related deaths occurred on protocol therapy (during induction, five; during BMT, 8; during CC, 1; during cis-RA, 1), 8 occurred during early follow-up, and the remainder occurred in patients who had relapsed and who had received nonprotocol therapy. Treatment-related deaths occurred in similar proportions in patients who were actually treated with ABMT (22 of 122) versus with CC (22 of 138; P = .7408), and the proportion of treatment-related deaths also was similar in patients with cis-RA (9 of 70) and without cis-RA (7 of 78; P = .5973).
This long-term analysis with an 8-year median follow-up validates our previously reported results that showed the unequivocal benefit of myeloablative therapy in high-risk neuroblastoma. There is a trend for consistently better outcome for each randomly assigned therapy separately, for all patients and for those with stage 4 disease, and for both EFS and OS, although this is statistically significant by log-rank comparison only for the ABMT versus CC random assignment for EFS. By using a comparison of the log(−log(.)) transformation of the survival estimates at 5 years, OS was significantly higher for each random assignment separately, and OS was higher for the ABMT and the cis-RA groups. Because more than 91% of the expected events are already observed, the results of this analysis are stable. For patients who underwent both random assignments, the EFS and OS were significantly higher for patients assigned to ABMT and cis-RA compared with those assigned to CC and no cis-RA.
Two other randomized trials showed a significant improvement in EFS for patients who received myeloablative therapy compared with chemotherapy or with no further treatment. Neither of these randomized trials showed a significant advantage for OS, despite the comparison to no further therapy in one study and to low-dose oral maintenance therapy in the other study.14,15 The 5-year EFS rate for all patients on CCG-3891 was 30% compared with 38% for the European Neuroblastoma Study Group 1 study, in which the random assignment occurred later and was limited to patients who achieved good response. The improvement in OS was significant in the European Neuroblastoma Study Group 1 study when analysis was restricted to patients with stage 4 disease who were older than 1 year of age.15 The more recent German study that used peripheral blood stem cells and incorporated immunotherapy resulted in a 5-year EFS rate of approximately 38%.14 Although CCG-3891 incorporated total-body irradiation in the conditioning regimen, neither European study used this modality, and there was no apparent detriment to EFS. Because of increased late effects with total-body irradiation,16 it has been deleted in the majority of current high-risk neuroblastoma trials.
Outcome was improved additionally with cis-RA compared with no further therapy, and there was a significant increase in the 5-year OS by the log(−log(.)) transformation. This study likely showed more benefit than the study reported by Kohler et al,17 because the latter used a much lower dose of cis-RA than was shown effective in preclinical studies.18 Although the rate of occurrence of events and deaths appeared similar between the two therapies within the first 2 years after transplantation, cis-RA appeared beneficial in patients who survived, event free, beyond 2 years from transplantation. One possible explanation is that cis-RA is effective only against minimal residual disease and does not prevent recurrence from larger amounts of tumor. Therefore, a test at a fixed point in time is appropriate for the measurement of the late benefit of cis-RA instead of the log-rank test, which equally weights all parts of the survival curves. By using the test of the log(−log(.)) transformation of the survival estimates at 5 years, OS was significantly improved for patients who were randomly assigned to receive 13-cis-RA (P = .0006).
Characteristics that were significant for all patients in univariate analyses of EFS and OS included stage, serum ferritin, MYCN gene copy number, histopathology, and response to induction therapy.8 Prognostic variables for EFS in multivariable analyses for all randomly assigned patients for the first random assignment of ABMT versus CC were response to induction, whereas MCYN amplification and response were significant for OS. The importance of disease response has been reported in other analyses of myeloablative therapy,19 though some suggest intensive consolidation can overcome this difference.20 Retrospective multivariable analyses of 529 patients who were on an European Bone Marrow Transplant registry found that the only significant factors for EFS were presence of disease by MIBG scan or persistent bone marrow disease at the end of induction.21 MYCN status and stage disappear as prognostic variables for EFS in the patients who underwent ABMT in our study, similar to other reports. For the second random assignment that tested cis-RA, the most constant significant variables for both EFS and OS in both arms were MYCN amplification and disease stage 4. Response to induction lost significance, perhaps because eligibility for the cis-RA random assignment excluded documented persistent disease.
In conclusion, long-term follow-up of this randomized trial showed significantly better 5-year EFS and OS rates for myeloablative therapy with purged ABMT than for non–myeloablative chemotherapy, and cis-RA given after intensive therapy resulted in significant improvement in 5-year OS rates, regardless of the type of consolidation.
The Acknowledgment and Appendix are included in the full-text version of this article, available online at www.jco.org. They are not included in the PDF version (via Adobe® Reader®).
|Characteristic||Random Assignment Group by Phase|
|First Phase||Second Phase|
|No. of patients||189||190||118||130||128||37|
|Age at diagnosis, years|
|MYCN gene copy|
NOTE. There were no significant differences between groups with respect to any characteristic after the first or second random assignment. However, the group that was nonrandomly assigned to continuing chemotherapy differed from the randomized cohort only in the higher percentage of patients with stage 3 disease (21% v 12%; P = .02) and the lower median age (2.6 v 3.1 years; P = .01). The cis-RA group was randomly assigned to 13-cis-retinoic acid. The no cis-RA group was randomly assigned to no 13-cis-retinoic acid. The NRRA group was nonrandomly assigned to 13-cis-retinoic acid.
Abbreviations: ABMT, autologous bone marrow transplant; CC, continuing chemotherapy; NRCC, nonrandomly assigned to continuing chemotherapy; NRRA, nonrandomly assigned to 13-cis-retinoic acid; CR, complete response; VGPR, very good partial response; PR, partial response; SD, stable disease; MR, mixed response; PD, progressive disease.
|Prognostic Factor||No. of Patients Assessed||Univariate Analyses by Survival|
|P*||Relative Risk||P*||Relative Risk|
|Age < 547 v ≥ 547 days||539||< .0001†||NA||.1667†||NA|
|Age < 730 v ≥ 730 days||539||< .0001†||NA||.9951†||NA|
|Evan's stage 3 v 4||539||< .0001||2.228||.0001||2.053|
|MYCN single copy v amplified||395||< .0001||2.161||< .0001||2.474|
|Histology favorable v unfavorable||373||.0007||4.635||.0028||3.865|
|Ferritin ≤ 143 v > 143 ng/mL||487||< .0001||1.693||< .0001||1.631|
|Response CR/VGPR v ≤ PR||462||< .0001||2.048||< .0001||1.719|
|Stage 4 only|
|Age < 547 v ≥ 547 days||466||< .0001†||NA||.1882†||NA|
|Age < 730 v ≥ 730 days||467||< .0001†||NA||.0034†||NA|
|MYCN single copy v amplified||334||< .0001||1.990||< .0001||2.315|
|Histology favorable v unfavorable||312||.0026||4.573||.0094||3.712|
|Ferritin ≤ 143 v > 143 ng/mL||417||.0002||1.518||.0021||1.459|
|Response CR/VGPR v ≤ PR||399||< .0001||1.913||.0001||1.577|
Abbreviations: EFS, event-free survival; OS, overall survival; CR, complete response; VGPR, very good partial response; PR, partial response.
Supported by Grants No. CA98413 from the Children's Oncology Group and CA13539 from the Children's Cancer Group, Division of Cancer Treatment, National Cancer Institute, National Institutes of Health, Department of Health and Human Services, and by the Kasle and Tcalcevik Neuroblastoma Research fund (K.K.M.), the Campini Foundation (K.K.M.), and the Conner fund (K.K.M.), and by Grants No. CA82830 and CA81403 from the National Cancer Institute (C.P.R.).
Authors' disclosures of potential conflicts of interest and author contributions are found at the end of this article.
The initial results of this trial were previously published in the New England Journal of Medicine 341:1165-1173, 1999.
The author(s) indicated no potential conflicts of interest.
Conception and design: Katherine K. Matthay, C. Patrick Reynolds, Robert C. Seeger, E. Stanton Adkins, Daphne Haas-Kogan
Provision of study materials or patients: Katherine K. Matthay, Hiroyuki Shimada, Daphne Haas-Kogan, Wendy B. London
Collection and assembly of data: C. Patrick Reynolds, Hiroyuki Shimada, Robert B. Gerbing
Data analysis and interpretation: Katherine K. Matthay, C. Patrick Reynolds, Robert C. Seeger, E. Stanton Adkins, Robert B. Gerbing, Wendy B. London, Judith G. Villablanca
Manuscript writing: Katherine K. Matthay, C. Patrick Reynolds, Judith G. Villablanca
Final approval of manuscript: Katherine K. Matthay, C. Patrick Reynolds, Robert C. Seeger, Hiroyuki Shimada, E. Stanton Adkins, Daphne Haas-Kogan, Robert B. Gerbing, Wendy B. London, Judith G. Villablanca