Antitumor Activity and Tolerability of hu14.18-IL2 with GMCSF and Isotretinoin in Recurrent or Refractory Neuroblastoma: A Children's Oncology Group Phase II Study Free

This article is featured in Highlights of This Issue, p. 6013

Neuroblastoma is a disease characterized by its clinical and biologic heterogeneity with half of all patients presenting with high-risk disease features including widespread dissemination and/or adverse tumor-specific biology (1). While roughly 70% of children with high-risk neuroblastoma respond initially to intensive multimodal induction and consolidation therapy, half will eventually die from recurrent disease, most often due to residual, refractory disease that is undetectable, but present at the end of first-line therapy (2). The differentiating agent isotretinoin has historically been used to treat this minimal residual disease (MRD; refs. 3, 4). More recently an immunotherapy strategy has been studied, using a mAb to target the disialonganglioside GD2, which is highly expressed by neuroblastomas, and has restricted distribution in normal tissue (5–7). A randomized trial conducted by the Children's Oncology Group demonstrated superior outcome for those who received treatment with the chimeric monoclonal anti-GD2 antibody ch14.18 (dinutuximab) combined with alternating cycles of GM-CSF and IL2 in addition to isotretinoin therapy compared with treatment with isotretinoin alone (8). This combination is now the standard of care for treatment of MRD in patients with high-risk neuroblastoma in North America (9).

Hu14.18-IL2 is an immunocytokine consisting of the humanized 14.18 anti-GD2 mAb linked to IL2 (10). Preclinical data suggest activity of hu14.18-IL2 is mediated by activation of antibody-dependent cellular cytotoxicity (ADCC) and complement-dependent cytotoxicity via the binding of hu14.18-IL2 to GD2 on the tumor cell surface, followed by binding to Fc receptors on effector cells along with activation of NK and T cells via IL2 receptor binding (10–12). The immunocytokine has activity in neuroblastoma mouse models (13) and has been shown in mice to have superior activity compared with a mixture of separate mAb and IL2 (14, 15). Phase I testing of hu14.18-IL2 in pediatric patients demonstrated biologic activity and clinical tolerability at a MTD of 12 mg/m²/day for 3 days (16). Dose-limiting toxicities included hypotension and allergic reactions. Phase II testing of Hu14.18-IL2 showed tolerability and promising activity in the MRD setting (17).

Given the preclinical and clinical activity of hu14.18-IL2, we postulated that substituting this immunocytokine in place of ch.14.18 and IL2 in the now standard MRD regimen might improve high-risk neuroblastoma patient outcomes. Hu14.18-IL2 has only been tested previously as a single agent and therefore the primary objectives of this study were to evaluate the safety, tolerability, and antitumor activity of hu14.18-IL2 given in combination with GM-CSF and isotretinoin in a schedule similar to current, standard MRD therapy (8).

The NCI was the sponsor of this phase II single-arm trial (Clinicaltrials.gov trial number: NCT01334515) that ran from September 26, 2011 to August 17, 2012. Hu14.18-IL2 was supplied collaboratively by the National Cancer Institute Biological Resources Branch (NCI-BRB, Frederick, MD) via a cooperative research and development agreement, currently held by Apeiron Biologics (Vienna, Austria). The study was conducted by the Children's Oncology Group (COG) following ethical principles of the Declaration of Helsinki. Each COG site's Institutional Review Board approved the study. Patients and/or legal guardians provided written informed consent with assent obtained as appropriate prior to patient enrollment.

Patients with recurrent or refractory neuroblastoma (age 12 months–30 years) were eligible. For patients with persistent tumor after first-line therapy (primary refractory disease), a biopsy demonstrating viable tumor was required. Patients were required to have tumor imaging and a bone marrow evaluation done within 21 days of enrollment and were required to have either measurable tumor by CT or MRI, defined as a minimum of 20 mm in at least 1 dimension, a metaiodobenzylguanidine (MIBG) scan with uptake in a minimum of one site or a bone marrow aspirate or biopsy with tumor cells seen on routine morphology in at least one sample. Eligibility required performance status, organ function, recovery from prior therapy, and life expectancy the same as required in the prior hu14.18-IL2 phase II study (17). Patients who had previously received therapeutic anti-GD2 antibody were eligible provided that they did not have progressive disease or a severe allergic reaction during their previous treatment. Exclusion criteria included patients with central nervous system tumor involvement, patients requiring chronic immunosuppression, and patients with a prior history of ventilator support related to lung injury, due to toxicities seen in the phase II trial. Patients were also excluded if they had symptomatic pleural effusions or ascites, had major surgery within 2 weeks of enrollment or had organ allografts (including bone marrow).

A treatment course was 28 days in length (Fig. 1). GM-CSF was given subcutaneously at a dose of 250 mg/m² on days 1 through 2 and days 8 through 14. Hu14.18-IL2 (12 mg/m²/dose) was administered on an inpatient basis as a 4-hour intravenous infusion on days 4 through 6. Patients received indomethacin (0.5 mg/kg/dose) every 6 hours while receiving the immunocytokine. Isotretinoin was given at standard dosing (160 mg/m²/day, divided twice daily) for 14 days starting on day 11. GM-CSF was not given concurrently with Hu14.18-IL2 due to concern of neurotoxicity seen in a prior study when GM-CSF and IL2 were given concurrently (but not seen when IL2 and GM-CSF were given sequentially; refs. 18, 19).

Disease evaluations were done every two courses and response was determined using the International Neuroblastoma Response Criteria (20). Treatment could be continued for a response of stable disease (SD) or better for up to 4 courses. After 4 courses of treatment, patients with at least a partial response (PR) could continue treatment for up to a total of 10 courses, provided that they did not have disease progression or drug intolerance.

Toxicities were graded by the NCI Common Toxicity Criteria (v4.0). A DLT for hu14.18-IL2 was defined as any ≥grade 3 toxicity, with certain reversible exceptions identified in the phase I study (16). Hu14.18-IL2 treatment was held for a DLT and restarted at 50% the previous dose once toxicity resolved. DLTs for GM-CSF and isotretinoin were defined standardly. GM-CSF was held if the total white blood cell count was >50,000/μL but this was not considered a toxicity.

All patients who completed ≥2 courses of combination therapy or who had an event (relapse or progressive disease) at any time in the study were evaluable for inclusion in the response analysis. All responses were confirmed by independent radiology and pathology review.

Response for patients with measurable disease (stratum 1) was determined using the Response Evaluation Criteria in Solid Tumors (RECIST). Response for patients with evaluable disease (MIBG avid and/or bone marrow disease only, stratum 2) was determined as follows:

Patients graded locally with complete response (CR) or partial response (PR) by MIBG were scored by central review using Curie scoring (21–24). CR was defined by complete resolution of all MIBG avid lesions. PR was defined as a reduction in MIBG avid lesions so that the relative Curie score (Curie score at response evaluation/baseline Curie score) was between 0.2 and 0.5.

For patients who entered with bone marrow disease (neuroblastoma identified in the bone marrow aspirate and/or biopsy by the local pathologist using standard histology), CR was defined as no tumor cells detectable by morphology and immunocytologic analysis on two subsequent bilateral bone marrow aspirates/biopsies done ≥3 weeks apart. Progressive disease (PD) was defined as ≥25% tumor in the marrow and a doubling in the percentage of tumor. Stable disease (SD) was defined as persistence of disease that does not meet criteria for CR or PR.

A secondary aim of the study was to describe the disease burden of the stratum 2 subjects. This was done through centralized Curie scoring of each patient's baseline MIBG scan. In addition, treating institutions described baseline bone marrow disease following bilateral bone marrow aspirates and biopsies based on the following scale: score “0″ if no tumor seen in any of the enrollment specimens; score “1″ if less than 10% tumor cells seen in all specimens but some tumor cells seen in at least one enrollment specimen; score “2″ if greater than or equal to 10% tumor cells seen in at least one enrollment specimen.

Peripheral blood samples were obtained prior to the first dose of the hu14.18-IL2 immunocytokine and at 3 subsequent timepoints, for each cycle of hu14.18-IL2 treatment from all patients providing consent. Serum ELISA assays determined the peak hu14.18-IL2 serum levels achieved with each intravenous infusion during the 3 courses of hu14.18-IL2, as described previously (23–25). Serum obtained following each course of therapy was also assayed by bridging ELISA for the generation of an antibody response to the hu14.18-IL2 immunocytokine, as described previously (16). The peak anti-immunocytokine antibody was graded as negative, intermediate, or positive based on optical density (O.D. units). Serum soluble IL2 receptor (sIL2R) alpha levels were determined using the Human CD25/IL2 R alpha DuoSet ELISA Kit (R&D Systems, catalog number DY223), following manufacturer's instructions.

Activity of hu14.18-IL2 combined with GM-CSF and isotretinoin was evaluated separately within each patient stratum. Safety and tolerability were evaluated in the overall cohort using a two-stage design as described below.

The primary objectives of this study were to evaluate safety and tolerability as well as efficacy. The primary endpoints, respectively, were the occurrence of an unacceptable DLT and occurrence of a best overall response of CR, very good partial response (VGPR), or PR. A two-stage stopping rule was used to monitor for an excessive number of unacceptable DLTs. “Unacceptable” was defined as a requirement for pressor and/or ventilator support due to acute vascular leak syndrome or grade 3 or 4 neurotoxicity with MRI evidence of new CNS thrombi, infarction, or bleeding. The combination therapy was considered intolerable if there was ≥2 unacceptable toxicities in the first 10 patients or ≥7 unacceptable toxicities in the first 30 patients. This rule had a 91% power and type I error of 0.098 under the null hypothesis that the probability of having an unacceptable DLT is <5% and the alternative is >28%.

Responders were defined as evaluable patients who demonstrated a best overall response of CR, VGPR, or PR. Response was analyzed separately within stratum 1 and stratum 2 patients using two separate Simon two-stage rules. For stratum 1, if 1 or more patient responded out of the first 12 evaluable patients, 9 more evaluable patients were enrolled. If 2 or more patients out of 21 total evaluable patients responded, the regimen was considered effective. The rule for stratum 1 has 90% power to detect a true response rate of 20% (under a null hypothesis of 1%) with a type 1 error rate of 0.017. For stratum 2, if 2 or more patients responded out of the first 15 evaluable patients, 15 more evaluable patients were enrolled. If 5 or more patients out of 30 total evaluable stratum 2 patients responded, the regimen was considered effective. The rule for stratum 2 has 92% power to detect a true response rate of 28% (under a null hypothesis of 7%) with a type 1 error rate of 0.048.

Median baseline disease burden between responders and nonresponders was compared using a Wilcoxon rank sum test. Mixed models with a compound symmetry covariance structure were built with data from courses 1 and 2 to test for differences [by stratum, responder status (among stratum 2 patients), and prior anti-GD2 therapy] in peak serum hu14.18-IL2 levels on days 4 and 6 postinfusion. The dependent variable was serum hu14.18-IL2 level and the independent variables included time point (categorical variable with four levels, one for each of the two peak time points across two courses, ordered chronologically), variable of interest, and an interaction term. Similarly, mixed models were constructed for serum IL2 receptor levels at various time points [prior to hu14.18-IL2 and 2, 3, and 7 days following the first day (day 4) of immunocytokine treatment] across courses to assess in vivo immune activation, in comparison with baseline (prior to hu14.18-IL2, using contrasts) and by responder status (stratum 2 patients only). Fisher exact test was used to evaluate the association between response and development of anti-hu14.18-IL2 antibody among stratum 2 patients.

Fifty-two eligible patients were enrolled, 16 in stratum 1 and 36 in stratum 2. Characteristics of these patients are shown in Table 1. The 16 patients in stratum 1 received a total of 43 treatment courses (median, 3 courses) and the 36 patients in stratum 2 received a total of 120 courses (median, 2 courses).

Fifty-one patients were evaluable for toxicity. The one inevaluable patient withdrew consent prior to receiving treatment. Table 2 shows the grade 3 or greater toxicities experienced over all treatment courses.

The combination therapy met the protocol definition for tolerability with ≤7 unacceptable toxicities in the first 30 evaluable patients. In total there were four unacceptable DLTs. One patient developed hypoxia and respiratory distress due to capillary leak during course 2 and required mechanical ventilation for 24 hours. Two patients had grade 4 hypotension secondary to capillary leak, one during course 1 and one during course 2, and both required pressor support for 24 hours. A fourth patient developed grade 5 pulmonary edema during course 2 that was attributed definitely to disease progression and possibly due to treatment. no Grade 3 or 4 neurologic toxicities were reported.

The best overall responses for all evaluable stratum 1 and stratum 2 patients are shown in Table 3. Two of the 16 patients in stratum 1 were not evaluable for response. One patient withdrew consent prior to treatment and one had an unacceptable DLT during course 2 and was taken off protocol therapy. Of the 14 evaluable patients in stratum 1, there were no responders. Two subjects were enrolled beyond the first stage of the trial due to reported responses not centrally reviewed in real time and ultimately not confirmed.

Five of the 36 patients in stratum 2 were not evaluable for response. One patient came off protocol therapy after course 1 due to refusal of further therapy. Four patients did not complete at least two courses of therapy due to unacceptable or other grade 4 DLTs. The protocol planned to enroll 30 patients evaluable for response in stratum 2; a 31st patient was enrolled into stratum 2 before it was recognized that a 30th patient had met evaluability criteria. Five patients out of the first 30 evaluable patients responded to treatment as confirmed by central review, meeting protocol-defined activity criteria, for an overall response rate of 16.1% (95% Wilson confidence interval; 7.1%–32.6%). The statistical criterion for activity required at least 5 responders in stratum 2, and this boundary was met.

The characteristics of the 5 responders in stratum 2 are described in Table 4. Two patients (4 and 37) had prolonged responses and were alive with no evidence of disease 5 years from the completion of therapy. Both of these patients ended protocol therapy with a partial response and received radiation to residual MIBG avid sites. Patient 48 had a complete response after 2 courses of treatment and received a total of 10 courses of treatment. He remained in a complete response with no further therapy until relapse was detected on a surveillance MIBG scan done 5 years after completion of therapy. Of the two other responders, one (patient 24) had a complete response after 2 courses of therapy but then had progressive disease at the end of 8 courses. The other (patient 50) received a total of 10 courses of therapy and was in a complete remission at the end of treatment but relapsed 3 months later. In addition to these 5 responders confirmed by central review, there was a sixth patient scored as a complete responder by the treating institution. This patient (52) was reported to have bone marrow involvement only at study entry. The patient received 4 courses of protocol therapy with no further treatment and remained in remission 5 years from the completion of therapy. This patient did not have baseline bone marrow samples available for central review and thus could not be considered a responder.

There were no responses seen in the 11 patients who received prior anti-GD2 antibody, although the patient mentioned above whom was judged a CR by the treating institution, but not confirmed by central review did receive prior dinutuximab.

To better understand which stratum 2 patients responded to therapy, baseline disease burden was described using the baseline Curie score and a qualitative bone marrow score. Responders had a lower baseline median Curie score (4, range 0–9), compared with nonresponders (10, range 0–25) but this difference was not statistically significant (P = 0.1395). The median baseline bone marrow score was 2 for responders versus 1 for nonresponders (P = 0.1243).

There was no significant difference between peak hu14.18-IL2 serum levels on days 4 and 6 between stratum 1 and stratum 2 patients, between responders and nonresponders among stratum 2 patients, or between patients who received prior anti-GD2 antibody and those who did not. Serum IL2 receptor levels were significantly increased compared with baseline (course 1 day 4, P < 0.0001) at all time points except course 2 day 4 and course 3 day 4. Serum IL2 receptor concentration did not correlate with response.

The generation of an anti-immunocytokine antibody level was positive or intermediate in 12 of the 26 stratum 2 patients evaluable for response and with serum available for testing. There was no association of anti-immunocytokine antibody response with prior mAb treatment. Three of the 5 responders were within the positive or intermediate anti-IC group. The data in Table 5 suggest there is no correlation of anti-IC with response. A single patient (a nonresponder) developed a strong anti-immunocytokine antibody following course 1, which effectively neutralized detection of circulating hu14.18-IL2 (levels <10% of predicted) in the three subsequent courses.

This study demonstrates antitumor activity of hu14.18-IL2 in combination with GM-CSF and isotretinoin in patients with relapsed/refractory neuroblastoma with disease evaluable by MIBG or bone marrow examination. Five of 31 evaluable patients had a confirmed objective response, three of which were durable 5 years or more after the completion of therapy. These results are similar to what was seen in the single-agent phase II trial (5 complete responders out of 23 evaluable patients; ref. 17) and confirm the antitumor activity of Hu14.18-IL2. The results also suggest that the addition of GM-CSF and isotretinoin was neither synergistic nor antagonistic in regards to activity, although the study was not designed to make this direct comparison.

The previous phase II results suggested that hu14.18-IL2 was more effective in patients with a smaller disease burden, as no responses were seen in patients with measurable disease (17). This was also seen in the current study with no confirmed responders seen in stratum 1. To further address the question of the correlation between response and disease burden, this study used baseline Curie and bone marrow scores as surrogate measures of disease burden. Neither statistically impacted response, although there was a suggestion that patients with a lower disease burden by Curie score (i.e., those with fewer MIBG avid sites) were more likely to respond to therapy than those with a larger MIBG-avid disease burden. The median degree of bone marrow involvement was actually greater for responders than nonresponders. This may not be a true trend but may be related to the wide cut-off values set to semiquantitatively measure bone marrow disease (all patients with greater than or equal to 10% tumor cells seen in at least one enrollment specimen were considered to have the same amount of marrow disease in our analysis).

The primary objective of this study was to evaluate the tolerability of combining hu14.18-IL2 with GM-CSF and isotretinoin in a schedule similar to the way dinutuximab is now given to high-risk neuroblastoma patients for minimal residual disease maintenance therapy as per the completed Children's Oncology Group Phase III trial ANBL0032 (8).We knew there may be risk of increased neurologic toxicity because of DLT seen in a phase Ib/II study where GM-CSF and IL2 were given concurrently to adults with renal cell cancer (18, 19). For this reason, similar to what was tolerable in the renal cell carcinoma study, we dosed GM-CSF and the immunocytokine sequentially and did not see any grade 3 or 4 neurologic toxicities in this neuroblastoma trial. There were 4 “unacceptable” toxicities observed in this study, one of which was more likely due to progressive disease than treatment. The other 3 unacceptable toxicities resolved 24 hours after stopping immunocytokine and were similar to what was described in the single-agent phase II trial (17). The other grade 3 and 4 toxicities observed were consistent with those previously reported for hu14.18-IL2 and for anti-GD2 mAb plus IL2. These results suggest that the addition of GM-CSF and isotretinoin did not change the toxicity profile of hu14.18-IL2.

Evidence for immune activation was seen as treatment induced increases in serum sIL2R levels, but the magnitude of these was not correlated with tumor response. None of the responders received prior antibody but it is not clear whether this is a direct correlation or the result of small sample size. One patient previously treated with dinutuximab, with bone marrow–only disease at study entry, who could not be considered a responder due to lack of material for central review, did have a reported (by local institutional histologic review of bone marrow aspirates and biopsies) durable complete response in the bone marrow after receiving two courses of treatment in this study. This observation is consistent with, but does not prove the speculation that prior dinutuximab exposure did not prevent activity, but this question needs to be studied on a larger scale. Clearly, this is an important consideration for future studies given the widespread use of anti-GD2 antibodies in patients with newly diagnosed neuroblastoma. Data about prior antibody therapy were not collected on the prior hu14.18-IL2 trials so this cannot be used to help draw conclusions.

Despite the antitumor activity and relative tolerability of the combination of hu-14.18-IL2 with GM-CSF and isotretinoin in this study, the overall response rate in this study was less than 20% and substantial toxicities were still observed. Even though the chimeric version of this immunocytokine (ch14.18-IL2) shows far superior antitumor activity in neuroblastoma-bearing mice over the combination of equivalent doses of ch14.18 plus IL2 as separate molecules (14), it is difficult to compare the clinical results from this phase II study with published phase II trials of ch14.18 mAb, with or without GM-CSF or IL2 (26) and be convinced that any one regimen is superior to the other, in part, due to differences in patient population, eligibility criteria, and other variables that prevent direct comparisons. Given the durability of responses seen in some stratum 2 relapsed/refractory patients, and the smaller baseline tumor burden in patients that have responded, hu14.18-IL2 might also be considered as possible MRD treatment of relapsed patients who have achieved a second objective response. Other anti-GD2–based approaches are also being pursued. The recent reports from COG of what appears to be greater efficacy seen when dinutuximab is combined with irinotecan, temozolomide, and GM-CSF (27), and from St. Jude with a similar approach combining chemotherapy with a separate humanized anti-GD2 mAb (hu14.18K322A; ref. 28), suggest this approach deserves additional focus, in other combinations and with other anti-GD2–based agents. Finally, recent preclinical data have demonstrated potent in vivo antitumor efficacy, in mice with large GD2 tumors, when intratumoral hu14.18-IL2 is combined with local low-dose radiotherapy (29) and even greater potency when anti-CTLA4 checkpoint blockade is added (30). This regimen induces both innate and adaptive antitumor immune activity in these tumor-bearing mice. Clinical testing of intratumoral hu14.18-IL2 with local low-dose radiotherapy and checkpoint blockade is being initiated in the treatment of GD2-expressing melanomas; this approach with hu14.18-IL2 may be worthy of clinical investigation in GD2-expressing pediatric tumors. As the incorporation of anti-GD2 immunotherapy into combination therapy approaches proceeds it will be helpful to be evaluating correlative biomarkers to potentially help predict which patients may be most likely to benefit from which therapeutic approach (31).

A. Naranjo is a consultant/advisory board member for Novartis. W.B. London is a consultant/advisory board member for United Therapeutics Corp. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Shusterman, W.B. London, J.M. Maris, J.R. Park, P.M. Sondel

Development of methodology: S. Shusterman, J.A. Hank, J. Gan, J.R. Park, P.M. Sondel

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Shusterman, J.A. Hank, S. Servaes, W.B. London, H. Shimada, J.M. Maris, P.M. Sondel

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Shusterman, A. Naranjo, C. Van Ryn, J.A. Hank, M.T. Parisi, B.L. Shulkin, S. Servaes, W.B. London, J. Gan, J.M. Maris, J.R. Park, P.M. Sondel

Writing, review, and/or revision of the manuscript: S. Shusterman, A. Naranjo, J.A. Hank, M.T. Parisi, B.L. Shulkin, S. Servaes, W.B. London, H. Shimada, J. Gan, S.D. Gillies, J.M. Maris, J.R. Park, P.M. Sondel

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Shusterman, C. Van Ryn, J.A. Hank, W.B. London, P.M. Sondel

Study supervision: S. Shusterman, P.M. Sondel

Other (acquisition of data via running immune monitoring data and organizing data and contributing to the analysis of these data): J.A. Hank

Other (invented and developed the agent the study is based on): S.D. Gillies

This work was supported by Alex's Lemonade Stand Foundation (to P.M. Sondel, J.A Hank, and J.M. Maris), The St. Baldrick's Foundation (to P.M. Sondel and J.A. Hank), and grants CA032685 (to P.M. Sondel, J.A. Hank, and J. Gan), CA87025 (to P.M. Sondel, J.A. Hank, and J. Gan), CA166105 (to P.M. Sondel, J.A. Hank, and J. Gan), CA 180886 (to J.R. Park), CA180899 (to C. Van Ryn and A. Naranjo), CA197078 (to P.M. Sondel and J.A. Hank), CA220500 (to J.M. Maris), UL1TR000427 (to P.M. Sondel and J.A. Hank), and 1TL1RR025013-01 (to P.M. Sondel and J.A. Hank). This work was also supported by The Stand Up To Cancer–St. Baldrick's Pediatric Dream Team Translational Research Grant (SU2C-AACR-DT1113). Stand Up to Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the Scientific Partner of SU2C (to P.M. Sondel, J.A. Hank, J.M. Maris, and J.R. Park).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.