Neuroblastoma (NBL) is a malignant tumor of incompletely understood etiology, originating from neuroblasts. The peak incidence of NBL occurs during the second year of life, with 90% of cases diagnosed in children under 5 years of age (1, 2). The location of the primary tumor varies depending on the origin and migratory pathways of neuroblasts during fetal development. Most commonly, NBL arise in the retroperitoneal area of the abdominal cavity, posterior mediastinum, neck, or pelvis (3). Clinical symptoms largely depend on the location of the primary lesion. Symptoms of relapsed neuroblastoma can include among others an abdominal mass, enlarged lymph nodes in the neck, swelling and bruising of the area around the eyes, unexplained fevers, bone pain, or limping, weakness or paralysis, weight loss or poor appetite, and uncontrolled eye or leg movements (4). Additionally, symptoms may be associated with excessive secretion of catecholamines or intestinal vasoactive hormones (2). Neuroblastoma is often associated with nonspecific symptoms that may delay diagnosis, such as somnolence, lack of appetite, weight loss, abdominal pain, fever, lymphadenopathy, pallor, easy bruising, weakness, or irritability (2, 5). Relapsed disease does not need to manifest clinically as symptomatic tumor progression (6). Typically, new lesions are detected in bone or bone marrow during follow-up scanning after completion of upfront therapy with earlier detection allowing for earlier treatment initiation aiming to improve outcomes (7).

In high-risk NBL (HR-NBL), defined by metastatic disease in children older than 12 or 18 months, or by MYCN amplification at any age, approximately half of those who achieve remission will experience relapse (8). The median time to relapse is 13.2 months (range, 1 day to 11.4 years), and the 5-year overall survival (OS) from the time of first relapse is 20% (9). This indicates that relapse negatively affects survival, with earlier relapse associated with a higher risk of death (9). Relapsed NBL requires intensive treatment that may involve a combination of chemotherapy, immunotherapy, and other targeted therapies. The most important goal of treatment in patients with relapsed/refractory NBL is to control disease and delay progression in the short-term, with the aim of prolonging long-term disease-free survival and potentially providing a route to cure with increasing emphasis on quality of life (10).

The presence of GD2, a disialoganglioside expressed on the surface of almost all NBL cells, was identified as a target for immune recognition by anti-GD2 monoclonal antibodies (11), forming the basis for effective immunotherapy and chemoimmunotherapy. Dinutuximab beta (dB, ch14.18/CHO) is a chimeric monoclonal IgG1 antibody specifically directed against the carbohydrate moiety of GD2, which is overexpressed on NBL cells (12). In vitro studies have demonstrated that dB binds to GD2-expressing NBL cell lines and induces both complement-dependent and antibody-dependent cell-mediated cytotoxicity (13, 14). In the presence of human effector cells, dB was found to mediate the lysis of human NBL and melanoma cell lines in a dose-dependent manner (14). Unlike other mouse-human chimeric monoclonal IgG1 antibodies, dB is produced in Chinese hamster ovary (CHO) cells. These cells are virus free and are considered the current standard for the production of recombinant proteins used in clinical trials (15). Additionally, CHO-derived production has favorable glycosylation pattern (less amount of immunogenic Neu5Gc forms of sialic acid) (16, 17), while anti-GD2 monoclonal antibody produced in SP2/0 cells (such as dinutuximab (Unituxin^®), contains Galactose-alpha-1,3-Galactose epitopes which have the potential to cause allergic/hypersensitivity reactions, including anaphylaxis (18–22). No Galactose-alpha-1,3-Galactose-epitopes are present on dB and the reported difference in the risk serious anaphylactic reactions is approximately two-fold in favor of dB (18), albeit head-to-head study data have not been identified. At present, dB is the only anti-GD2 treatment that has been approved by the European Medicines Agency for the treatment of NBL since May 2017 (14), and is the only anti-GD2 antibody approved for maintenance therapy both in patients with newly diagnosed and relapsed disease.

Maintenance therapy with anti-GD2 antibodies, such as dinutuximab and dB, has demonstrated efficacy in patients with HR-NBL in the first-line setting, with significant improvements in event-free survival and OS, as shown in COG and SIOPEN trials (p<0.05) (8, 11, 23). Subsequent research investigating the addition of interleukin-2 (IL-2) revealed no further benefit (8, 24, 25), leading to the recommendation that dB therapy be administered without IL-2 (26).

Furthermore, anti-GD2 immunotherapies have been established as a standard of care for the treatment of relapsed/refractory NBL when administered in combination with chemotherapy (27). Following disease control, anti-GD2 monotherapy is subsequently used as maintenance therapy, with the aim of sustaining remission and preventing further relapse (25, 28).

In a retrospective Spanish study by de las Heras et al. (2025) (29), which included patients with refractory and relapsed/progressive HR-NBL, anti-GD2 immunotherapy (administered in combination with chemotherapy, as consolidation, or as maintenance) was associated with significantly improved OS (48.4% vs. 19.4%, p=0.012), compared to no immunotherapy (29). Previously, a comparison of retrospective data to historical controls from Italian Neuroblastoma Registry (INBR) demonstrated improved overall survival on dB (p=0.002) (30). While the impact of dB on OS in patients with relapsed/refractory NBL is acknowledged, the design limitations of available studies (such as cross-over effects in prospective trials, lack of separate analyses for dB in maintenance therapy, and small sample sizes) (31, 32) preclude a reliable estimation of the effect size. No trials are currently available that directly assess dB used as maintenance therapy, either in combination with IL-2 or as a single agent, as recommended by SIOPEN, compared to no dB (i.e., no immunotherapy) in the relapsed NBL setting. Given the high unmet need among patients with relapsed NBL and the lack of alternative therapies at the maintenance stage, it is essential to conduct a reliable indirect comparison of survival outcomes. Indeed, it is likely the only viable approach to generate comparative evidence, as a parallel prospective study of dB versus non-immunotherapy control, may not be ethically feasible.

Therefore, the aim of this study was to compare OS in patients with relapsed NBL treated with dB as maintenance therapy (either in combination with IL-2 or as a single agent, as recommended by SIOPEN) with that of historical control patients who did not receive anti-GD2 immunotherapy.

Despite the growing body of clinical and real-world evidence, head-to-head comparisons between dB (with or without IL-2) and no dB treatment in the maintenance setting for patients with relapsed NBL are lacking. This absence is primarily due to ethical constraints, as a randomized controlled trial by Yu et al. (2010) (23) demonstrated that the addition of an anti-GD2 antibody (dinutuximab), combined with granulocyte-macrophage colony-stimulating factor and IL-2, to isotretinoin following intensive multimodal therapy significantly improves survival outcomes in patients with HR-NBL. Thus, the objective of our analysis was to conduct an indirect comparison of the efficacy of dB in maintenance treatment versus no anti-GD2 treatment in the population of patients with relapsed NBL. The only feasible analysis was an unanchored indirect comparison using available IPD from dB studies (APN311-202 (18), APN311-304 (37) and APN311-303 (18)) and from historical cohorts of untreated patients, including the INBR control from Garaventa et al. (2009) (33) and the R1 control from the HR-NBL-1/SIOPEN study (34).

Newly available clinical study reports enabled the pooling of data from 144 patients treated with dB, with or without IL-2, and 82 historical control patients who did not receive immunotherapy, a uniquely large sample with extended follow-up, given the rarity of NBL (44). Considering that this population consists exclusively of patients with relapsed disease, recruiting large cohorts for prospective trials poses substantial challenges. In other studies investigating dB use in relapsed/refractory NBL, the number of enrolled patients did not exceed 70 (25). Moreover, previously published indirect comparisons of dB versus no dB treatment in the relapsed setting, including those originally conducted for the European Medicines Agency (18) (as reported in the European Public Assessment Report) encompassed significantly fewer patients. These comparisons were limited to studies such as APN311–303 and APN311-202 (N = 48 in both) (18), or the analysis by Mueller et al. (2018) (30), which included 29 patients on dB treatment versus 27 control patients from the AIEOP database.

In our study, harmonized inclusion criteria were applied across all study populations, enhancing the likelihood of selecting patients with comparable known and unknown characteristics from different data sources. A two-stage adjustment process (i.e., selection of appropriate patients based on the harmonized criteria, followed by adjustment for key baseline characteristics) enabled a reliable comparison. Significant differences were found in key baseline characteristics (i.e., age, age at NBL diagnosis, year of diagnosis, time from diagnosis to the starting point, MYCN status, and INSS stage at diagnosis) between patients from the dB arm and historical controls. However, these differences were well balanced after PS reweighing. The only baseline parameter that remained insufficiently balanced was year of diagnosis. This imbalance was expected, as the studies on dB use and those involving historical control cohorts were conducted using different designs and during different time periods, making alignment on year of diagnosis unfeasible. Almost half of dB patients (n=71, 49%) were diagnosed after 2009, while control patients were diagnosed between 1999 and 2009. Over the years, the standard of care may have improved (apart from better availability of anti-GD2 treatments) for patients diagnosed later. This advancement may be partially responsible for the observed improved OS in dB-treated patients, which constitutes a limitation of our analysis. However, the resulting bias is unlikely to be clinically meaningful, given the lack of effective treatments in the setting of multiple relapses. While chemoimmunotherapy with dB has been shown to improve survival outcomes (45–49), its wider adoption occurred after the completion of the studies included in this analysis.

Another limitation of this study is the difference in response to induction therapy between patients in the R1 control group and those in the INBR control group. In the R1 group, only patients who experienced relapse after achieving an initial complete response were included in this analysis (based on IPD availability), whereas the INBR group included all relapsed patients with any prior response (both partial and complete responses). In the dB studies patients with both complete and partial response were included. It is well established that individuals with a prior partial response generally have a poorer prognosis compared to those who achieved a complete response. Therefore, the inclusion of only those R1 patients, who were required to have achieved a CR, in the pooled analysis may have led to an underestimation of the survival benefit associated with dB. Furthermore, the limited availability of information on historical control patients prevented full adjustment for baseline differences between groups, such as prior treatment history, number of previous relapses or progressions, and subsequent treatments. Therefore, residual confounding cannot be excluded. We adjusted for the main post-relapse prognostic factors, but other clinically important variables (end-of-induction response, detailed histology, ALK status, tandem transplant was not done in that era at all type of transplant, number of frontline anti-GD2 cycles, chemotherapy backbone at relapse) were not available in a harmonized form. For several of these, robust evidence for an independent ‘carry-over’ effect on OS after relapse, beyond time to relapse, age, INSS and MYCN, is lacking, yet some residual confounding is likely.

The only assessed endpoint was OS, which is considered the gold standard for evaluating the clinical benefit of cancer therapy, particularly in the absence of cross-over. Overall survival is straightforward to measure and is not subject to investigator bias. However, treatment efficacy may be influenced by subsequent therapeutic interventions, specifically those administered after relapse, which can vary across clinical settings. Consequently, data necessary to balance patient populations based on post-progression treatments were not available (50).

In dB studies, OS was defined as the time from the initiation of dB treatment to death. On the other hand, for historical control patients, OS was originally calculated from the date of the last relapse. To correct for differences between dB studies and historical control groups in the base-case analysis, the starting point was estimated by adding 269 days (i.e., the median time between the first relapse and the start of dB and/or IL-2 treatment for patients from the APN311–202 study) to the date of relapse in the historical control groups. A limitation of this analysis is that it cannot be ruled out that some historical control patients may have received an anti-GD2 agent as part of other studies. If anti-GD2 therapy positively influences OS, then the inclusion of such patients in the control group would likely attenuate the observed difference in OS between the treatment and control groups.

Progression-free survival could not be defined consistently because dates and criteria for progression after relapse were incompletely and non-uniformly recorded in the historical cohorts. Consequently, overall survival from first relapse was used as the primary comparative endpoint.

The results of both unadjusted and adjusted indirect treatment comparisons revealed a significant improvement in OS with dB therapy, with or without IL-2, compared to historical controls not treated with dB. This risk of death was reduced by 57% in the unadjusted comparison and by 47% in the adjusted analysis. Additionally, 24 unadjusted and 3 adjusted sensitivity analyses were consistent with the primary findings and confirmed a significant survival benefit of dB over no immunotherapy. Sensitivity analyses examining the concomitant use of IL-2 indicated that the survival benefit of dB without IL-2 was comparable to that of dB with or without IL-2.

The results obtained in this analysis are consistent with previous indirect comparisons of dB versus historical controls, as presented in the European Public Assessment Report for dB (18). In that analysis, the proportion of patients remaining alive at 3 years was nearly twice as high in the dB-treated group (from the APN311–202 and APN311–303 studies) compared to historical controls (R1 (34) and Garaventa et al. (33)) who did not receive immunotherapy in the relapsed NBL population. Similar results were reported by Mueller et al. (2018) (30) for the APN311–303 study alone, in which the 3-year survival rate among patients with relapsed NBL treated with dB was twice as high as that of patients not treated with dB in the historical control cohort from the AEIOP database (54% vs 26%). As reported in the analysis by Wex et al. (2023), the comparison of dB plus IL-2 with the control cohort from the AEIOP database based on earlier data yielded a HR for OS of 0.43 (95% CI: 0.24–0.78); p=0.00542), while the HR of 0.56 (95% CI: 0.32–0.97; p=0.0376) was reported for comparison with the R1 control group (51).

A similarly beneficial effect on OS was demonstrated in primary studies evaluating anti-GD2 therapy in other subpopulations of patients with NBL, where immunotherapy was the primary intervention in the treatment group. In a meta-analysis by Khan et al. (2024) (52), dinutuximab was associated with a significant reduction in overall mortality, with a pooled relative risk of 0.41 (95% CI, 0.22–0.75; p = 0.004), and improved 5-year event-free survival among patients diagnosed with HR-NBL compared to alternative treatment modalities. Similarly, in the study by Ladenstein et al. (2020) (8), the 5-year event-free survival and OS rates for patients with HR-NBL who did not receive immunotherapy were 42% and 50%, respectively, whereas for those treated with dB, the rates were 57% and 64%, respectively (p<0.001) with the corresponding OS HR = 0.70 (95% CI: 0.54-0.91); p=0.0071 (53). Also, a retrospective observational study from Saudi Arabia (54) demonstrated that dB significantly improves survival outcomes in children with HR-NBL as compared with a control cohort, with a median OS of 46.88 months vs 15.91 months (p=0.001). Real World Evidence from the Netherlands Cancer Registry similarly confirmed efficacy of anti-GD2 therapy relative to non-immunotherapy control, adjusting for potential change in clinical practice, with HR = 0.37 (95% CI: 0.19-0.72); p <0.01 (55).

A factor that can potentially impact the magnitude of the treatment effect of dB in relapsed patients is prior exposure to dB or other anti-GD2 immunotherapies. Such an effect could be mediated through the presence of anti-drug antibodies or a reduced expression of GD2 antigen after relapse. Currently no clinical evidence is available allowing to confirm this effect, but anti-anti-GD2 antibodies were detected in patients treated with anti-GD2 immunotherapy with unknown clinical effect (14, 54), while high GD2 expression was reported to persist after relapse (56, 57). Complete or partial loss of GD2 expression was found in only 10-12% of neuroblastoma cells from bone marrow, with no clear relationship between GD2 expression and ADCC, suggesting that other mechanisms play more important role in developing resistance to treatment (58, 59). While it can be assumed that on multiple exposure, efficacy of anti-GD2 diminishes in some patients, our analysis demonstrates a high effect size, unlikely to be markedly reduced on re-treatment. Limited evidence suggests continued efficacy or anti-GD2 immunotherapies after prior exposure (48, 60, 61).

We lacked sufficiently detailed data to quantify prior anti-GD2 exposure in the DB cohort (some APN311 studies permitted enrolment of patients with prior anti-GD2 therapy provided no anti-DB antibodies were present). If prior anti-GD2 reduces the effectiveness of subsequent anti-GD2-based therapy at relapse, any such unmeasured exposure would be expected to bias our comparison toward the null, so the observed OS benefit associated with DB is likely conservative.

It should be noted that in all studies included in the indirect comparison, patients received dB via long-term infusion (LTI), administered over 10 days. LTI has been shown to provide a significantly greater survival benefit compared to short-term infusion, which is administered over 5 days, in patients with HR-NBL (37, 62). In addition, LTI is associated with a more favorable safety profile, including lower pain scores, reduced intravenous morphine use, and a lower incidence of grade ≥3 adverse events, which supports its feasibility in outpatient settings (30). These findings further support the use of LTI of dB, as recommended by SIOPEN, including in patients with relapsed/refractory NBL.

Although indirect comparisons do not represent the highest level of clinical evidence, they are commonly used in the assessment of treatments for reimbursement decisions. Regulatory authorities also accept evidence from single-arm studies using historical controls, as demonstrated by the recent approval of eflornithine by the U.S. Food and Drug Administration for first-line post-maintenance therapy in NBL (63). As confirmatory randomized studies comparing anti-GD2 therapies to treatment without immunotherapy are no longer considered for ethical reasons, this comparison constitutes best available comparative evidence informing clinical and reimbursement decisions.

Indirect comparisons inherently carry certain unavoidable methodological limitations due to differences in clinical experiences, which may affect the reliability of the results. Despite these limitations, and the inability to fully balance baseline characteristics across all potential prognostic factors, such as prior treatments or therapies received after progression, this study provides a robust comparison of OS in a context where conducting a parallel prospective trial of dB with a placebo control would not be ethically feasible. In this analysis, patients with relapsed NBL treated with dB were shown to have almost half lower risk of death compared to historical control patients. The magnitude of this treatment effect supports the efficacy of dB as maintenance therapy of NBL patients after relapse, even in the presence of potential biases. Unadjusted comparisons and multiple sensitivity analyses further confirmed these findings, reinforcing the strength of the conclusion.