Current salvage therapy for recurrent high-risk neuroblastoma is rarely curative. Assessment of the effectiveness of new, primarily cytostatic agents requires the redefinition of study endpoints to reflect disease stabilization rather than tumor response or regression. The intervals of disease control in the patients in the current study with recurrent neuroblastoma were characterized to provide comparison criteria for exploratory studies of new agents.
Disease control intervals, disease-free survival, postrecurrence survival, and median time to treatment failure were estimated in 90 patients with high-risk neuroblastoma treated between January 1991 and June 2002 on 3 St. Jude neuroblastoma protocols.
The estimated median time to disease recurrence was 18.3 months (95% confidence interval [95% CI], 15.9–22.4 months) for the first recurrence, 8.7 months (95% CI, 5.0–12.2 months) for the second recurrence, and 3.8 months (95% CI, 2.5–5.4 months) for the third recurrence. The 5-year estimate of survival after the first disease recurrence was 11% ± 4%. Patients with longer initial disease control had a postrecurrence survival advantage:the 5-year estimated postrecurrence survival was 15.3% ± 6.3% for patients with initial disease control ≥16 months and 8.1% ± 5.5% for others (P = .006). The median disease control interval was approximately halved after each disease recurrence.
Neuroblastoma is advanced at the time of presentation in approximately 50% of patients, and long-term survival rates are suboptimal in these cases.1, 2 Multimodal therapy incorporating intensive cytoreductive chemotherapy, surgery, irradiation, and subsequent treatment for minimal residual disease was found to improve event-free survival in 2 recent large cooperative group studies.3–5 Retreatment with standard therapies after disease recurrence can produce brief second or subsequent remissions, but many patients have a poor response. Approximately one-third of patients have no meaningful response to second-line therapy.6 The majority of patients die due to recurrent neuroblastoma, and therefore new approaches to treatment are needed.7, 8
Several new agents that target specific signaling pathways or cellular processes offer potential alternative postrecurrence therapies (either as single agents or in combination with standard cytotoxic agents). However, many of these new agents are primarily cytostatic (likely to produce disease stabilization rather than regression), and therefore tumor response (regression) is not a useful criterion of their efficacy. The accurate assessment of the activity of these agents requires study endpoints based on intervals of disease progression (or control). We reviewed the records of patients with high-risk neuroblastoma who were treated on 3 recent institutional protocols to characterize their disease-free survival and postrecurrence survival intervals. These intervals can serve as reference points for the evaluation of new agents.
MATERIALS AND METHODS
Patients and Treatment
We examined the records of all 91 patients with newly diagnosed high-risk neuroblastoma (defined on the basis of age, stage, and biologic features) treated at St. Jude Children's Research Hospital between January 1991 and June 2002 on 3 institutional frontline protocols (NB8814, NB91, and NB97). Details of these protocols have been reported previously.9–12 Briefly, patients received induction chemotherapy with cyclophosphamide, doxorubicin, cisplatin, ifosfamide, and etoposide (pharmacokinetically guided dosing of topotecan began in the NB97 protocol), surgical resection of the primary tumor after remission induction, locoregional radiotherapy as appropriate, and continuation chemotherapy (NB88 protocol) or consolidation with autologous stem cell transplantation and maintenance with interferon-α (NB91 protocol) or cis-retinoic acid (NB97 protocol). Twenty-three patients were enrolled on the NB8814 protocol, 38 on the NB91 protocol, and 30 on the NB97 protocol. One patient was excluded from this analysis because of discrepancy in stage assignment. Thus, a total of 90 patients are reported here. Patients were staged at the time of diagnosis and at the time of disease recurrence according to the guidelines of the International Neuroblastoma Staging System (INSS). Disease recurrence was confirmed by histologic findings (biopsy of bone marrow or lesion[s]) or by unequivocal meta-iodobenzylguanidine (MIBG) scan findings. Clinical characteristics (age, sex, race, primary tumor site, stage, and tumor MYCN status) and outcome data were obtained.
The majority of the disease recurrences in this patient population were systemic (bone marrow, skeletal, lymph node, central nervous system, or hepatic) and thus salvage attempts used primarily chemotherapy. Twenty-eight patients in first disease recurrence were treated with a regimen combining ifosfamide, carboplatin, and etoposide (ICE); 19 were treated with a topotecan-containing or irinotecan-containing regimen; 4 were treated with a taxane-containing regimen; 2 received cyclophosphamide and etoposide; and the remaining patients received phase 1 agents (oxaliplatin or fenretinide).
Nineteen patients were treated for second disease recurrence with a camptothecin, 11 patients were treated with an alkylator with or without an anthracycline or platinating agent, 3 patients were treated with an ICE regimen, and the remaining patients were treated with various phase 1 agents (tumor vaccines, hydroxyurea, interferons, etc) or allogeneic- or -matched unrelated stem cell transplantation.
All protocols and this analysis were approved by the St. Jude Children's Research Hospital Institutional Review Board (IRB) and informed consent was obtained from patients, parents, or guardians for the therapeutic protocols. The IRB waived consent requirements for the retrospective review of the data and for our analysis.
Disease-free survival (DFS) was defined as the time interval between frontline study enrollment and first disease recurrence, progression, or most recent contact. Postrecurrence survival (PRS) was defined as the time interval between first disease recurrence and death or most recent contact. The DFS and PRS were estimated by the method of Kaplan and Meier13; standard errors were calculated using the method of Peto and Pike.14 The median time to disease recurrence and 95% confidence intervals (95% CI) were estimated by the Kaplan-Meier method. The observed median time to disease recurrence was also calculated for patients who experienced disease progression or recurrence, although the primary analyses were those based on time-to-event methods. The exact log-rank test was used to compare PRS distributions by time to first recurrence. An event was defined as disease recurrence, disease progression, second malignancy, or death.
Of the 90 patients, 66 (73%) had experienced disease recurrence or disease progression at the time of last follow-up (78%, 68%, and 77% of patients enrolled on the NB8814, NB91, and NB97 protocols, respectively). Table 1 summarizes the characteristics of the patient group overall and according to disease recurrence/progression status. MYCN status was available for 85% of patients.
Table 1. Characteristics of 90 Cases of High-risk Neuroblastoma
Disease recurrence (n = 66)
No disease recurrence (n = 24)
All patients (n = 90)
INSS indicates International Neuroblastoma Staging System.
Median age at diagnosis (range)
3.3 y (0.5 mo–19.4 y)
2.7 y (0.8 y–7.1 y)
3.1 y (0.5 mo–19.4 y)
Only 5 (8%) of the 66 patients who had experienced recurrent or progressive disease remained alive and free of disease 13.7 years, 8.5 years, 6.0 years, 5.5 years, and 5.3 years, respectively, after frontline study enrollment. All 24 patients who had not experienced disease recurrence or progression were alive at a median of 10.0 years (range, 4.0–16.4 years) after frontline study enrollment. Of the 29 survivors, 24 (83%) had been seen or contacted within the past year and 28 (97%) had been seen or contacted within the past 3 years.
In the patient group as a whole, the first event was disease recurrence or progression in 66 patients (73%). The remaining 24 patients had no adverse events. The 5-year DFS estimate for all patients was 26.7% ± 4.7%. The median time to first disease recurrence after frontline study enrollment was 18.3 months (95% CI, 15.9–22.4 months).
For the 66 patients who had developed disease recurrence or progressive disease, the 5-year estimate of PRS was 11.1% ± 4.3% (Fig. 1). Longer initial disease control was significantly associated with longer survival after disease recurrence. As depicted in Figure 2, the 5-year estimate of PRS was 15.3% ± 6.3% for patients whose disease recurrence occurred ≥16 months after study enrollment and 8.1% ± 5.5% for patients whose recurrence occurred sooner (P = .006). We used 16 months because it was the lower boundary of the 95% CI for the median time to recurrence after frontline therapy. Results were similar when disease recurrence before versus after 18 months was analyzed.
Of the 5 patients who remained alive after disease recurrence/progression, 4 had INSS stage 4 disease and 1 had INSS stage 3 disease. MYCN status was available for 4 of these patients, none of whom had amplified MYCN. The interval between study enrollment and first disease recurrence for these 5 patients was 6.7 to 36.1 months (median, 18.6 months). At the time of last follow-up, all 5 patients were alive with no evidence of disease, >5 years after first recurrence.
Disease Control Intervals
In the study group overall, the median time between frontline protocol enrollment and first disease recurrence was 18.3 months (95% CI, 15.9–22.4 months).
Of the 66 patients who experienced disease recurrence or progression, 58 had a second recurrence, a median of 7.2 months (range, 0.4–29.0 months) after the first recurrence; 4 patients had not experienced second disease recurrence and were alive 11.8 years, 6.9 years, 4.8 years, and 3.0 years, respectively, after first recurrence; and 3 patients had died the day of (1 patient) or the day after (2 patients) first recurrence. The remaining patient died of neuroblastoma 4.6 years after first disease recurrence, but information regarding second or subsequent disease recurrences was unavailable because the patient did not return to St. Jude for treatment. This patient was excluded from the disease control analysis. The 3 patients who died at the time of first disease recurrence were censored at their times of death.
Among the 65 patients analyzed who had at least 1 recurrence, the estimated median time from first to second disease recurrence was 8.7 months (95% CI, 5.0–12.2 months). Of the 58 patients who had at least 2 recurrences, 44 experienced a third recurrence at a median of 3.5 months (range, 0.3–18.7 months) after the second disease recurrence, 1 had a second malignancy after the second disease recurrence, 6 died at the time of the second recurrence, 2 died within 2 weeks of the second recurrence (3 days and 12 days afterward, respectively), and 1 was alive at the time of last follow-up, 4.0 years after second disease recurrence (5.5 years after frontline study enrollment). The remaining 4 patients died 4.7 months, 5.6 months, 9.2 months, and 13.2 months, respectively, after second disease recurrence (information regarding a third recurrence was not available). Patients who died before a third disease recurrence were censored at the date of death. The patient who had a second malignancy after second disease recurrence was censored at the date of the diagnosis of a second malignancy. The estimated median time from second to third recurrence was 3.8 months (range, 2.5–5.4 months).
Table 2 shows estimates of the median time to first, second, and third disease recurrence (15.7 months, 7.2 months, and 3.5 months, respectively). The disease control interval decreased by approximately half after each successive disease recurrence, as illustrated by Figure 3.
Table 2. Summary of Disease Control Intervals (in Months) in Patients With Recurrent/Progressive Neuroblastoma
One patient who died nearly 5 years after first recurrence was excluded because of insufficient data regarding treatment and additional recurrences of disease.
Estimated using the time-to-event method of Kaplan and Meier.
Study enrollment to first recurrence
First to second recurrence
Second to third recurrence
All patients that died immediately before or after first or second disease recurrence had rapidly progressive systemic tumor (epidural, hepatic, peritoneal, renal, central nervous system) and the parents chose no further interventions.
Herein, we report the disease progression intervals in a large cohort of children with advanced stage neuroblastoma who were treated on successive frontline protocols at a single institution. These patients face a very poor prognosis; only 5 of 66 patients in the current series experienced durable long-term remission after disease recurrence. The 5-year estimate of survival after recurrence was 11% ± 4%. However, the distribution of time-to-event profiles suggests that the disease behaves differently in subsets of patients. Recurrence of the original tumor is the leading cause of late mortality in neuroblastoma patients, and 38% of the deaths occurred within 1 year after the first disease recurrence. However, 16 of 66 patients (24%) were alive 3 years after the first recurrence. Similarly, patients with a longer first disease remission (≥16 months) survived longer than others after disease recurrence. The duration of first disease remission appears to be a powerful predictor of outcome in patients with acute lymphoblastic leukemia as well.
Five (8%) patients in the current series remained alive >5 years after disease recurrence. These patients may have been cured by salvage therapy or may represent variant cases of ‘chronic neuroblastoma.’15 Such patients characteristically demonstrate improvement of symptoms and may have prolonged mixed responses to therapy, but the disease eventually progresses. A similar course typically follows subsequent salvage attempts. To our knowledge, the clinical features and biologic tumor characteristics that uniquely identify this subset of patients currently are unknown. However, investigators designing future salvage protocols should note that a subset of patients will be ‘chronic’ survivors despite previous treatment failures.
Disease progression data are of value in planning the testing of new therapy agents for neuroblastoma. Patients must be selected for these trials on the basis of clinical characteristics at the time of disease recurrence and the anticipated subsequent disease control or progression intervals. This information can be derived from historic data such as that presented herein, allowing studies to be stratified on the basis of the number of previous disease recurrences and by the estimated disease control intervals. In the absence of these considerations, phase 1 and 2 studies of different agents might be conducted in study populations that are not comparable, and agents could be inappropriately discarded or considered effective when the results are confounded by the natural course of the clinical disease.
Many of the newer molecularly targeted agents are cytostatic and are unlikely to cause tumor regression (the current standard of response in pediatric oncology drug development). Furthermore, they may not show their effect until after the first 2 courses of therapy (the conventional evaluation period in many pediatric oncology phase 2 trials). Therefore, use of the traditional phase 1/2 study response endpoints may obscure the true activity of new antineuroblastoma agents. Recent data from trials of new agents in adults suggest that progression-free survival may be the most appropriate outcome for comparison of new versus standard chemotherapeutic agents. For example, sorafenib was found to significantly improved progression-free survival in adults with advanced renal cell carcinoma, despite having only a 2% partial response rate (and no complete responses) by standard criteria.16 Sunitinib significantly reduced time to progression of advanced gastrointestinal stromal tumors yet had only a 7% response rate (no complete responses).17 In addition, neuro-oncologists have recently advocated progression-free survival at 6 months as a significant endpoint for assessing a new intervention for brainstem glioma. The historic disease control (progression) distributions must be defined for patients who have had multiple recurrences of neuroblastoma so that appropriate boundaries can be established for exploratory studies of new agents.
Many barriers stand in the way of neuroblastoma clinical trials that incorporate new agents.18 It is difficult to choose the best agent to carry forward, the population of patients available for study is small, the profitability of clinical development is likely to be low, and the endpoints for use in regulatory approval are not well defined for this disease.19 Furthermore, at most, only 1 large phase 3 neuroblastoma trial can be conducted every 5 to 7 years; therefore, it is imperative that phase 2 trials be designed optimally for selection of the best agent (or therapy modality) to bring forward. A better understanding of the natural course of neuroblastoma in patients undergoing salvage therapy and the use of historic data to inform the design of exploratory phase 2 trials will help to serve this purpose. In parallel, new validated imaging and laboratory tools, such as positron emission tomography scans with metabolic markers, may help to advance our understanding of the effects of a given treatment on the tumor and of whether the imaging changes reflect a specific cellular response.20, 21
In summary, we recommend that in designing exploratory trials of new agents for neuroblastoma, investigators and sponsors incorporate the historic disease control intervals for the population and use progression-free survival as a primary endpoint. These steps may allow us to better distinguish the clinical benefit of such agents and facilitate the regulatory approval process.
We thank Valerie McPherson for data management support and Sharon Naron for editorial assistance.