Sequential Combination of Unfavorable Histology followed by Clinical Stage M Defines High-Risk Neuroblastoma: A Report from the Children’s Oncology Group

Abstract

Purpose:

Historically, neuroblastoma risk stratification has been performed with clinical stage as the starting point and successively adding other prognostic factors thereafter. This study takes an alternative approach to define neuroblastoma patient risk groups by starting with the International Neuroblastoma Pathology Classification (INPC).

Experimental Design:

The neuroblastoma patient cohort previously used for developing the Children’s Oncology Group Revised Neuroblastoma Risk Classification (RNRC) system was reanalyzed by survival-tree regression analysis starting with the INPC distinguishing favorable histology and unfavorable histology categories. The resultant two branches were further divided first by the International Neuroblastoma Risk Group Staging System (INRGSS) and successively by other prognostic factors.

Results:

This new stratification system, the INPC-Risk Grouping (INPC-RG), is simpler than the RNRC system, eliminating unnecessary decision trees and distinguishes four risk groups (Groups I to IV). Using only INPC (Unfavorable Histology) and INRGSS (Stage M), INPC-RG defines patients with highly aggressive Group IV tumors, whose 5-year event-free survival (EFS) was worse than that of the RNRC high-risk group. Additionally, it identifies Group III patients whose 5-year EFS spanned 50–80%, which was not identified by the RNRC.

Conclusions:

The benefits of using this new INPC-RG system are four-fold: (1) it allows for the rapid identification of Group IV patients; (2) it lays the foundation for further refinement of Group III; (3) it can stratify patients when the amount of tumor tissue is limited; (4) it allows patients in resource limited areas to be appropriately stratified, potentially improving the world-wide treatment of patients with neuroblastoma.

INTRODUCTION

Peripheral neuroblastic tumors, including neuroblastoma, ganglioneuroblastoma, and ganglioneuroma, are the most common extracranial solid neoplasms in children, with more than 650 cases diagnosed annually in the United States (1). Since tumors in this group demonstrate a wide range of clinical behaviors, such as spontaneous regression, tumor maturation, and fatal progression refractory to treatment, various risk classification systems have been proposed for predicting patient clinical outcomes. Among them, two systems, the Children’s Oncology Group (COG) risk classification system, distinguishing low-, intermediate-, and high-risk groups (2,3), and the International Neuroblastoma Risk Group (INRG), distinguishing very low-, low-, intermediate-, and high-risk groups (4), are widely accepted. Both are based on a combination of multiple prognostic factors, including clinical stage, patient age at diagnosis, histopathology, and molecular/genomic properties. The INRG, developed with data from patients diagnosed between 1992 and 2002, was historically designed to facilitate the comparison of risk-based clinical trials conducted in different regions and countries; while the COG risk classification system, developed by using data from patients diagnosed after 2006, was designed for patient stratification and protocol assignment in COG clinical trials conducted in North America, Australia, and New Zealand.

In 2021, we published a Revised Neuroblastoma Risk Classification (RNRC) system (5) utilizing data from patients diagnosed and enrolled on COG studies between 2007–2017 after replacing the International Neuroblastoma Staging System (INSS; post-surgical staging system) (6) with the International Neuroblastoma Risk Group Staging System (INRGSS; pre-surgical, imaging-based staging system) (7) and incorporating segmental chromosomal aberrations (SCAs) as a prognostic factor. The RNRC system was established by survival-tree regression analysis starting with clinical stage according to the INRGSS (5) by analyzing a total of 4,832 patients. In the overall cohort, 5-year event-free survival (EFS) and 5-year overall survival (OS) [+/− standard error (SE)] were 72.0% (+/− 0.8%) and 81.5% (+/− 0.7%), respectively. According to this system, EFS and OS (+/− SE) for patients in the 3 risk groups were 90.7% (+/− 1.0%) and 97.9% (+/− 0.5%) for the low-risk group; 85.1% (+/− 1.4%) and 95.8% (+/− 0.8%) for the intermediate-risk group; and 51.2% (+/− 1.4%) and 62.5% (+/− 1.3%) for the high-risk group, respectively.

While establishing the RNRC system, we observed EFS and OS outcomes and hazard ratios by individual prognostic factors all had significant prognostic effects individually (5). It was also noted that both the age-linked International Neuroblastoma Pathology Classification [INPC (distinguishing favorable histology vs. unfavorable histology)] (8–10) and the INRGSS [Stage L1, L2, or MS vs. M* (see the note below)] had very high survival outcome hazard ratios compared to the other prognostic factors: EFS hazard ratio [95% confidence interval (CI)] of 4.41 (3.86, 5.03) by the INPC and 4.43 (3.90, 5.02) by the INRGSS; and OS hazard ratio (95% CI) of 10.50 (8.44, 13.04) by the INPC and 9.55 (7.85, 11.62) by the INRGSS, respectively. The results prompted us to conduct a new survival-tree regression analysis starting with the INPC, subsequently being split by the INRGSS, then followed by other prognostic factors using the same cohort of patients as those studied in the establishment of the RNRC system.

Note that Stage M* in this study included INRG Stage M and Stage M/MS-indeterminant cases. The Stage M/MS-indeterminant cases indicated metastatic disease in patients between 12 and 18 months, for whom collected clinical data could not define whether the pattern of metastases would meet the criteria for Stage M or MS designation. Patients with Stage M/MS-indeterminant comprised 4.7% of all cases in this series.

MATERIALS AND METHODS

Patient Cohort

Among a total of 4,832 patients enrolled on the COG ANBL00B1 Biology Study between October 1, 2007 and January 19, 2017 (5), 4,569 patients were eligible and included in this study. Central IRB or local-site IRB approval was obtained by the participating institutions, and written informed consents were obtained from the parents/guardians of patients to be enrolled. This study was conducted in accordance with recognized ethical guidelines (e.g., Declaration of Helsinki, CIOMS, Belmont Report, U.S. Common Rule). Eligible patients had newly diagnosed Neuroblastoma, Ganglioneuroblastoma, and Ganglioneuroma (maturing subtype) with the results of histopathology classification according to the INPC as determined by the central review pathologist (HS) at the COG Neuroblastoma Pathology Reference Laboratory. Patients with limited tumor specimens or only bone marrow samples were excluded. The detailed process of INPC classification has been described in previous studies (8–10) and is shown briefly below (see also Supplemental Table 1).

The definition of INPC (Supplemental Table 1)

The INPC system utilizes the degree of Schwannian stromal development for distinguishing Neuroblastoma (Schwannian stroma-poor), Ganglioneuroblastoma, intermixed (Schwannian stroma-rich), and Ganglioneuroma (Schwannian stroma-dominant) categories. Ganglioneuroblastoma, nodular (composite, Schwannian stoma-rich/stroma-dominant and stroma-poor) is a composite tumor with distinct tumor clones of either Ganglioneuroblastoma, intermixed or Ganglioneuroma and Neuroblastoma. Ganglioneuroblastoma, intermixed and Ganglioneuroma are classified into the favorable histology group, and the prognosis of patients is excellent. As for tumors in the Neuroblastoma category, age-dependent evaluation (<18 months; ≥18 months and <60 months; ≥60 months), grade of neuroblastic differentiation (undifferentiated; poorly differentiated; differentiating), and mitosis-karyorrhexis index (low; intermediate; high) are characteristics used to classify individual tumors as favorable histology or unfavorable histology. Tumors in the Ganglioneuroblastoma, nodular category are also classified as favorable histology or unfavorable histology by applying the same criteria to the Neuroblastoma component. In summary, the INPC is a compound prognostic factor consisting of histopathologic characteristics and age at diagnosis.

Prognostic Factors

In addition to the INPC results, clinical data included age at diagnosis (<12 months vs. ≥12 months; <18 months vs. ≥18 months) and staging according to the INRGSS, which is also a compound prognostic factor (7). INRGSS defines locoregional and metastatic diseases. Locoregional disease is divided into Stage L1 [localized disease with no image-defined risk factors (IDRFs)] and Stage L2 (localized disease with IDRFs). Stage M is defined as metastatic disease with noncontiguous disease and distant lymph node involvement. Stage MS is defined as disease with distant metastases confined to skin, liver, and/or bone marrow (<10% tumor involvement in marrow) in children <18 months, and the primary tumor could be INRGSS Stage L1 or L2. For the purposes of this study, Stage M* includes INGRSS M and M/MS-indeterminant cases.

Results of centrally tested ancillary modalities at the Steve and Cindy Rasmussen Institute for Genomic Medicine Clinical Laboratory Core Facility at Nationwide Children’s Hospital (RRID:SCR_017840) including MYCN oncogene status (non-amplified vs. amplified) by fluorescence in situ hybridization (FISH) and ploidy (hyperdiploid vs. diploid) by flow cytometry were also available for this analysis. Segmental Chromosomal Aberrations (SCAs) were assessed in a subset of patient tumors for 1p and 11q loss of heterozygosity (LOH, absent vs. present) using a PCR-based microsatellite marker assay before 2015 and, thereafter, using a SNP-array or whole exome array Comparative Genomic Hybridization.

Statistical Analyses

Details of the statistical approach performed at the COG Statistics and Data Center to establish the RNRC system were published previously (5,11). Time to event was calculated from diagnosis until the first occurrence of relapse, progression, secondary malignancy, or death for event-free survival (EFS); and time until death for overall survival (OS). For patients without an event, data was censored at last contact. Survival curves were generated per Kaplan-Meier (12), with EFS and OS values reflecting 5-year estimates with standard error per Peto (13), and groups compared via the log-rank test.

In cases where the survival curves intersected, to test that the hazards were proportional, a prognostic factor group by survival-time interaction time-varying covariate term was included in a Cox proportional hazards (PH) model for EFS and/or OS. If the PH assumption was not upheld, the p-value from the time-dependent, covariate-adjusted Cox model is presented instead of from the log-rank test.

Development of the Risk Classifier

In this study, cases were initially divided into favorable histology or unfavorable histology according to the INPC. Then, in each decision tree, they were classified into one of the INRGSS Stages (L1, L2, MS, or M*). Further recursive partitioning was performed based on the remaining prognostic factor with the largest statistically significant (p<0.05) hazard ratio from a univariate Cox PH model for EFS, until no statistically significant factors remained, to create a “survival tree”.

Data Availability

The data used in this study were previously published by the authors in Irwin MS, et al. “Revised Neuroblastoma Risk Classification System: A Report from the Children’s Oncology Group” (5). Additionally, the data (including raw data behind figures and graphs) is available upon request to the corresponding author.

RESULTS

The starting cohort for the study

As shown in Figure 1, among a total of 4,569 patients (5-year EFS: 69.8 +/− 1.2%, 5-year OS: 79.1 +/− 1.1%) in this analysis, 2,260 tumors were favorable histology (5-year EFS: 87.3 +/−1.3%; 5-year OS: 96.0 +/−0.7%) and 2,096 tumors were unfavorable histology (5-year EFS: 51.1 +/− 1.9%; 5-year OS: 61.5 +/− 1.8%) per the INPC at the starting point of the survival tree. A total of 263 (5.4%) cases were excluded from this study due to: (1) limited tumor specimens or only bone marrow samples available for pathology review; and (2) L1 or L2 tumors in the Ganglioneuroblastoma, intermixed and Ganglioneuroma category according to INPC; these were classified into the favorable group, Group I, regardless of other factors. Of note, for the same reason, they were not included in the RNRC system (Supplement Figure 1) since they were always classified into the low-risk group (5).

Figure 1. INPC-Risk Grouping System.

Risk Classifier Algorithm for patients starting with INPC (Favorable Histology Tumors vs. Unfavorable Histology Tumors) and then split by INRGSS.

Five-year EFS for the 4 prognostic groups: Group I (blue boxes) >90%; Group II (yellow boxes) >80 to ≤90%; Group III (grey boxes) ≥50 to ≤80%; and Group IV (red boxes) <50%. Quoted values in each box are 5-year EFS and OS survival rates.

In the tree for patients with favorable histology tumors, MYCN oncogene status (non-amplified vs. amplified) came before the split by INRGSS. L1 or L2 tumors with histopathology diagnostic categories of Ganglioneuroma or Ganglioneuroblastoma-intermixed are classified as favorable histology regardless of other factors and are not included in this figure.

Abbreviations: INPC, International Neuroblastoma Pathology Classification; INRGSS, International Neuroblastoma Risk Group Staging System; LOH: loss of heterozygosity; INRG L1: Stage L1 (Locoregional disease without image-defined risk factors); INRG L2: Stage L2 (Locoregional disease with image-defined risk factors); INRG MS: Stage MS; INRG Stage M*: Stage M and Stage M/MS-indeterminant (metastatic disease but their collected clinical data could not determine whether the patients had Stage M or Stage MS);

Survival-tree regression analysis in this study (Figures 1, 2, and 3)

Figure 2. Proportion of Patients in Different Prognostic Groups Based on INPC-Risk Grouping System.

Prognostic groups: Group I (blue), Group II (orange), Group III (grey), and Group IV (red).

Top, all patients (n=3,912) in this study; Bottom left, patients with favorable histology tumors (n=1,851): the majority were in either Group I or Group II and no patients were in Group IV; Bottom right, patients with unfavorable histology tumors (n=2,061): the majority were in Group IV and no patients were in Group I.

Figure 3. Survival Analyses for Risk Groups defined by INPC-Risk Grouping System.

5-year Event-Free Survival (left) and Overall Survival (right) for Group I, Group II, Group III, and Group IV according to the INPC-Risk Grouping System are shown. Log-rank tests were used to compare survival distributions. The patient cohorts, eligibility details, and missing data were described in the publication by Irwin et al. (J Clin Oncol. 2021 39(29):3229–3241). COG, Children’s Oncology Group; EFS=event-free survival; OS=Overall survival.

As shown in Figure 1 and 2, a total of 3,912 patients were classified into 4 different prognostic groups at the endpoints of the tree based on the other available prognostic factors. Five-year EFS for the 4 Groups used in this study were: Group I [n=1,298 (33.2%), blue boxes] >90%; Group II [n=789 (20.2%), orange boxes] >80 to ≤90%; Group III [n=268 (6.9%), grey boxes] ≥50 to ≤80%; and Group IV [n=1,557 (39.8%), red boxes] <50% (Figures 1 and 2). Age at diagnosis was not used as a branching point in this survival-tree regression analysis since it is confounded with both INPC and INRGSS. Figure 3 shows survival curves for Groups I, II, III, and IV: 5-year EFS and OS for Group I patients were 92.0 +/− 1.3% and 98.5 +/− 0.6%; Group II: 83.7 +/− 2.3% and 95.7 +/− 1.3%; Group III: 66.1 +/− 4.7% and 77.8 +/− 4.0%; and Group IV: 42.1 +/− 2.2% and 52.9 +/− 2.2%, respectively. Supplemental Table 2 summarizes the results of detailed analyses on the relevant patient groups. Detailed characteristics of patients with either favorable histology or unfavorable histology tumors in terms of MYCN status, ploidy, and SCA status are described below.

Among patients in Group IV, 169 (10.9%) were <18 months at diagnosis (5-year EFS: 46.6 +/− 6.7%, 5-year OS: 50.6 +/− 6.6%) and 1,388 (89.1%) were ≥18 months at diagnosis (5-year EFS: 41.6 +/− 2.4%, 5-year OS: 53.3 +/− 2.3%), and both age groups had similarly low survival rates. A significant difference in EFS (Cox model p<0.0001) and OS (Cox model p<0.0001) was found between age-groups after adjustment for non-PH. Further analysis of Group IV demonstrated that a small number of patients <18 months with MYCN non-amplified tumor [n=27 (1.7%)] had 5-year EFS of 71.4 +/− 15.6% and 5-year OS of 69.4 +/− 15.7%. Whereas the other remaining Group IV patients [n=1,530 (98.3%); <18 months of age having MYCN-amplified tumor and ≥18 months of age having tumors with or without MYCN amplification] had 5-year EFS of 41.5 +/− 2.2% and 5-year OS of 52.6 +/− 2.2% (EFS log-rank p-value=0.0475, OS log-rank p-value=0.2752).

1). Patients with favorable histology tumors:

Before making a forced split by INRGSS, favorable histology tumors were classified as either MYCN non-amplified (n=2,096) or MYCN amplified (n=60). A total of 1,851 patients were included in the final groups (end nodes) of the survival-tree regression analysis, with each group representing patients who share similar survival outcomes based on the other prognostic factors. As shown in Figures 1 and 2, the majority of patients (n=1,298, 70.1%) were found in Group I (blue boxes): tumors with favorable histology, MYCN non-amplified, and INRG Stage L1 (n=997); favorable histology, MYCN non-amplified, INRG Stage L2, and diploid (n=78); favorable histology, MYCN non-amplified, INRG Stage MS, and no 1p/11q LOH (n=96); or favorable histology, MYCN non-amplified, INRG Stage M* [As mentioned previously, Stage M* included Stage M and Stage M/MS-indeterminant (metastatic disease but their collected clinical data could not determine whether the patients had Stage M or Stage MS)], hyperdiploid, and no 1p/11qLOH (n=127). There were 359 (19.4%) patients in Group II (orange box): tumors with favorable histology, MYCN non-amplified, INRG Stage L2, and hyperdiploid. There were 194 (10.5%) patients in Group III (grey boxes): tumors with favorable histology and MYCN amplified (n=60); favorable histology, MYCN non-amplified, INRG Stage MS, and 1p/11q LOH (n=20); favorable histology, MYCN non-amplified, INRG Stage M*, hyperdiploid, and 1p/11q LOH (n=43); or favorable histology, MYCN non-amplified, INRG Stage M*, and diploid (n=71). No patients were found in Group IV in the favorable histology branch.

2). Patients with unfavorable histology tumors:

A total of 2,061 patients were found at the end nodes of survival-tree regression analysis. As summarized in Figures 1 and 2, the majority of cases (n=1,557, 75.5%) were found in Group IV (red boxes): tumors with unfavorable histology and INRG Stage M* regardless of MYCN status. There were 74 (3.6%) patients in Group III (grey boxes): tumors with unfavorable histology, INRG Stage L2, and diploid (n=54) or unfavorable histology and INRG Stage MS (n=20). There were 430 (20.9%) patients in Group II (orange boxes): tumors with unfavorable histology and INRG Stage L1 (n=318) or unfavorable histology, INRG Stage L2, and hyperdiploid (n=112). No unfavorable histology patients were found in Group I. Of note, for INRG Stage L1 and MS tumors in the unfavorable histology category that were in Group II and Group III respectively, there were no further breakpoints, either by MYCN status, DNA index or 1p/11q LOH, because these indicators were not prognostic.

DISCUSSION

The currently available histopathologic classification system (INPC) was established in 1999 with minor modifications of the Shimada histology approach (8–10). Prior to this, other prognostic factors were also identified, including MYCN amplification (14,15), ploidy (16), and LOH at 1p and 11q (17). INRG and RNRC took the strategy of survival-tree regression analysis using these prognostic factors, which used clinical stage as the starting point with other factors added successively thereafter. As mentioned, during construction of the RNRC system, it became evident that INPC and INRGSS as compound prognostic factors were predictive of disease outcome with similarly high hazard ratios (5). Therefore, it was of great interest to begin constructing survival-tree regression analysis starting with INPC. This was a historic and distinctive departure from the two previous risk stratification schemes.

Survival-tree analysis for the RNRC system, as previously reported (5), started with each clinical stage according to the INRGSS and had 27 endpoints and 3 prognostic groups (Supplemental Figure 1: low-risk in blue and green boxes, intermediate-risk in orange boxes, and high-risk in red boxes). In contrast, the newly proposed INPC-RG system starting with INPC followed by INRGSS has 14 endpoints and 4 prognostic groups (Figure 1 and 2: Group I in blue boxes, Group II in orange boxes, Group III in grey boxes, and Group IV in red boxes). Although both INPC (favorable histology vs. unfavorable histology) and INRGSS (Stage L1, L2, or MS vs. M*) had similarly high hazard ratios, there were other prognostic factors, such as MYCN oncogene status (non-amplified vs. amplified), ploidy (hyperdiploid vs. diploid), and absence or presence of the SCAs 1p/11q LOH, contributing to further prognostic distinctions in the decision tree. Age at diagnosis (18) was excluded as an additional branching point in the algorithm of this survival-tree. This was based on the fact that (1) age was already included in both INPC and INRGSS and (2) favorable histology vs. unfavorable histology distinction according to INPC was reported to add independent prognostic information beyond the prognostic contribution of age (19). However, it should be noted that age at diagnosis is one of the most commonly and easily accessible clinical data points and contributes to a prognostic prediction by itself with the historic notion of the younger the patient, the better the outcome (20).

Most patients (89.5%) in the favorable histology branch were classified into Group I or II, and their clinical outcomes were comparable to the low-/intermediate-risk cases in the RNRC system. In contrast, the majority of patients (75.5%) in the unfavorable histology branch were found in Group IV, and their outcomes were lower than those in the high-risk cases of the RNRC system. Additionally, the OS trend of Group IV patients showed a gradual decline over time (Figure 3), reflecting the tendency of these patients to relapse/die despite the intense high-risk therapy. Furthermore, no patients with favorable histology tumors were classified into Group IV and no patients with unfavorable histology tumors were classified into Group I in this survival-tree regression analysis. Additionally, cases in Group III (6.9% in the total cohort: 10.5% in the favorable histology branch and 3.6% in the unfavorable histology branch) had 5-year EFS (≥50% to ≤80%), which is between Group I/II and Group IV. Because the number of patients in this group is relatively small, the RNRC could have placed some patients into the intermediate-risk group and others into the high-risk group without significantly influencing survival rates of the intermediate-risk and high-risk groups. As indicated in Figure 1, Group III tumors likely represented a heterogenous population. Therefore, the identification of additional prognostic/biologic factors that could further segregate them into either higher-risk or lower-risk categories will be needed. In addition, close clinical follow-up may be required to establish new treatment protocols for these patients in Group III.

It was also noted that patients with tumors characterized as favorable histology and MYCN amplified were in Group III (n=60) corresponding to INRG Stage L1 (n=14), L2 (n=6), MS (n=9), M (n=16), and MS-Ind (n=15) disease. Five-year EFS of patients in this group was 66.6 +/− 11.6%); their survival rate was better than those in Group IV having unfavorable histology, INRG Stage M*, and MYCN amplified (n=571, 5-year EFS: 40.0+/−3.5%). Tumors in this subgroup were rare and have been reported as genotype-phenotype discordant. Patients with these genotype-phenotype discordant tumors generally have better survival rates than those with genotype-phenotype concordant tumors (unfavorable histology and MYCN amplified). This seems primarily due to an error in transcription/translation preventing amplified MYCN genes from resulting in MYCN protein overexpression (21–23). An additional contributing factor is that the majority of these patients were historically treated using high-risk protocols.

Identifying high-risk patients in the RNRC system or Group IV in the new INPC-RG system is critical, since their survival rates are the lowest and progress in improving their clinical outcomes has been challenging, despite clinical trials with aggressive multi-modal therapies. As shown in Figure 1, among patients with unfavorable histology tumors, 2,096 had information available for INRGSS. The majority (n=1,557, 74.3%) had Stage M* and were exclusively classified into Group IV. Furthermore, among patients in Group IV, 1,398 had data available regarding MYCN oncogene status, and 827 (59.2%) had MYCN non-amplified tumors while 571 (40.8%) had MYCN amplified tumors. In essence, Group IV patients had an aggressive clinical course regardless of MYCN oncogene status.

Both the RNRC and the newly proposed INPC-RG systems can identify patients with highly aggressive neuroblastoma requiring multi-modal and intensive therapy. Unfortunately, half of these patients will die of their disease (5). Therefore, it is important to expeditiously identify those who would likely not respond to the current high-risk therapy. Notably, this new INPC-RG system can define those patients in Group IV faster than the RNRC and other historical systems. As mentioned above, Group IV patients comprise a critical cohort for investigating molecular alterations associated with aggressive clinical behaviors in addition to MYCN oncogene amplification. To this end, efforts have been made to identify molecular alterations associated with Group IV tumors to define important subgroups and to find potential therapeutic targets. These molecular alterations, including MYC family (MYCN or MYC) protein overexpression (23), telomere maintenance abnormalities (24–26), and ALK protein activation (27), can potentially be targeted by drugs to neutralize their activities (28).

COG no longer provides ploidy results to treating institutions through the APEC (Project: EveryChild)/MCI (Molecular Characterization Initiative) program (29,30). In the risk classifier algorithm of RNRC system, ploidy results could alter risk group assignments for 12~18-month-old patients having Stage M* and MYCN non-amplified disease; these patients could otherwise be placed into either the intermediate- or high-risk groups. In contrast, in the new INPC-RG system, ploidy results do not significantly influence the distinction between Group IV, defined by unfavorable histology (INPC) and Stage M*, and other groups (Group I, II, III). Therefore, the lack of ploidy information would have minimal impact on high-risk assignments by INPC-RG.

SCAs (1p and/or 11q) were only prognostic for the favorable histology and MYCN non-amplified branch of the INPC-RG statistical regression analysis tree (Figure 1). This branch is further separated, based on 1p/11q SCA status, which places Stage MS cases into Groups I and III, and Stage M/MS-indeterminant and hyperdiploid cases into Groups I and III (Figure 1). Although SCAs are merely “prognostic markers” without specifically defined biological bases, a smaller deletion at 1p36 is associated with MYCN non-amplified tumors (31,32) and CHD5 loss may confer most of the MYCN-independent tumor suppressor effects of 1p36 LOH (33). Thus, a quantitative analysis of CHD5 protein expression by immunohistochemistry may be required to address the prognostic significance of CHD5 expression for the cases of the favorable histology and MYCN non-amplified branch. On the other hand, 11q LOH is inversely correlated with MYCN amplification and significantly associated with the expression of MYC signature genes, suggesting the elevated expression of MYC protein in 11q LOH cases (34,35). Therefore, from the histopathologic standpoint, when CHD5 and MYC family protein expression data become available, the power of SCAs at 1p and 11q as prognostic indicators may diminish.

Finally, survival analyses of INPC-RG Groups reveal that Group IV patients fare worse (5-year EFS: 42.1 +/− 2.2%; OS: 52.9 +/− 2.2%) than the RNRC-defined high-risk group (5-year EFS: 51.2 +/− 1.4%; OS: 62.5 +/− 1.3%) (5), suggesting that the combination of unfavorable histology and Stage M identify patients least likely to respond to maximally intensive therapy (such as tandem myeloablative chemotherapy, etc.) (36,37). To improve the survival of INPC-RG Group IV patients, we must rethink treatment of these patients and pursue innovative new therapeutic approaches that not only target their intrinsic biology (28) but also augment the patient’s immune defense mechanisms (38–41).

CONCLUSION

We conducted a survival-tree regression analysis for risk classification of neuroblastoma patients starting with INPC (favorable histology vs. unfavorable histology) followed by INRGSS (Stage L1, L2, MS, and M*) and other prognostic factors. Two important observations were made. First, the INPC-RG system identifies unique “non-high-risk” Group III patients, whose presence was not recognized by previous risk grouping systems. This subset of patients seems to be a mixture of intrinsically favorable patients who would not require the cytotoxic chemotherapy intended for the COG intermediate-risk group patients and those who would. Prospective analysis of additional prognostic indicators, such as CHD5 and MYC protein expression, as discussed above, will be needed for further exploration (33–35).

Second, in this new INPC-RG system, unfavorable histology combined with clinical Stage M* alone identifies highly aggressive Group IV neuroblastomas. More precisely, INPC-RG identifies a group of “highest-risk” patients with an approximately 10% lower 5-year EFS than that identified by RNRC. Those patients, who are in INPC-RG Group IV but not in the RCRC high-risk group, would require therapy strategies different from the current COG high-risk therapy. As we have mentioned in our report of the extremely unfavorable histology neuroblastoma (28,42), Group IV tumors are likely heterogeneous, and molecular mechanisms other than MYCN amplification, such as MYC family protein overexpression and telomere maintenance abnormalities, should be explored as potential indicators of aggressive clinical behavior and, possibly, therapeutic targets. In addition, frontend immunotherapy, such as anti-GD2 antibody therapy, combined with conventional chemotherapy may prove useful.

On balance, these observations highlight advantages of the new INPC-RG system over the current RNRG risk stratification system. Our ultimate goal is to smoothly transition patient stratification from the stage-based systems (RNRG and INRG) to the INPC-based system that we have developed in this study. Clearly this effort will require consensus building among pediatric pathology and oncology communities worldwide, construction of a new therapy stratification scheme, and prospective verification of the INPC-RG system in diverse cohorts.

There has been recent renewed recognition of INPC as an important prognostic factor predictive of disease outcome. The WHO has adopted INPC as one of the clinical indicators for neuroblastoma prognosis. Additionally, INPC is incorporated into the Cancer Pathology Datasets for Neuroblastoma of the International Collaboration on Cancer Reporting (ICCR), whose activities have been supported by multiple professional organizations across six continents (https://www.iccrcancer.org/datasets/published-datasets/paediatrics/neuroblastoma/).

Finally, future directions of exploration will include the use of artificial intelligence to assist pathologists with the expeditious and consistent determination of the INPC classification of peripheral neuroblastic tumors. The utilization of advanced assistive technologies to determine INPC classification coupled with this new simplified stratification system allows for streamlined clinical decision-making, especially when the amount of tumor tissue is limited and when ancillary testing modalities such as molecular/genomic testing are not feasible.

TRANSLATIONAL RELEVANCE.

Historically, neuroblastoma risk classifications were performed with clinical stage as the starting point and successively adding other prognostic factors thereafter. When the COG neuroblastoma risk classification system was revised (RNRC), it became evident that the International Neuroblastoma Pathology Classification (INPC) and the International Neuroblastoma Risk Grouping Staging System (INRGSS) were equally predictive of disease outcome, prompting us to reanalyze the same patient cohort starting with INPC. This new approach, the INPC-Risk Grouping (INPC-RG), streamlines the identification of Groups I, II, III, and IV patients. Importantly, only INPC unfavorable histology and INRGSS stage M were needed to identify Group IV patients, whose survival was worse than those of the RNRC high-risk group. INPC-RG also identified Group III, which was undetectable by RNRC. INPC-RG Group I/II and RNRC low/intermediate risk groups exhibited similar survival. INPC-RG allows for prompt clinical decision making, especially when tumor tissue is limited and ancillary testing modalities are not feasible.