Patients who were diagnosed with NB and received chemotherapy between March 2000 and December 2018 were screened. Patients who underwent the first cycle chemotherapy with a cisplatin+etoposide+doxorubicin+cyclophosphamide (CEDC) regimen and had ANC data available from the first cycle chemotherapy were included in this study. In total, 313 patients were enrolled in the present study. The patients’ medical records were reviewed for detailed clinical and biological data, including clinical features at diagnosis, tumor biology (including MYCN amplification), tumor histology according to the International Neuroblastoma Pathology Classification (INPC), treatment, survival, and treatment toxicity. The following equation was calculated to measure tumor volume: (π/6)×depth×width×height [8]. In the GWAS, the analytic cohort comprised 269 patients whose peripheral blood samples were cryopreserved at the Samsung Medical Center Biobank.

Patients were classified into three risk groups according to age at diagnosis, stage according to the International Neuroblastoma Staging System (INSS), and MYCN amplification. In brief, stage 1 or 2 tumors without MYCN amplification were classified as low-risk tumors. Stage 4 tumors in patients aged > 12 months (before 2015) or > 18 months (from 2015) and MYCN-amplified tumors were classified as high-risk tumors. All other tumors were classified as intermediate-risk tumors. For chemotherapy, CEDC and ifosfamide+carboplatin+etoposide (ICE) regimens were used alternatively. Detailed information on the chemotherapy regimen has been described in a previous report [9]. In general, low-risk patients with stage 2 tumors received six cycles of chemotherapy; intermediate-risk patients received nine cycles of chemotherapy plus differentiation therapy (13-cis-retinoic acid) with or without local radiotherapy to the primary site; and high-risk patients received nine cycles of induction chemotherapy, followed by tandem high-dose chemotherapy (HDCT), local radiotherapy, and differentiation therapy with or without immunotherapy with interleukin 2 [9,10]. Therefore, all patients received the same chemotherapy regimen for the first six cycles. A smaller chemotherapy dose was used in patients aged < 24 months as follows: a body weight-based dose in patients aged < 6 months, a 70% dose based on body surface area in patients aged 6–11 months, an 80% dose in patients aged 12–17 months, and a 90% dose in patients aged 18–23 months. From the second cycle, the chemotherapy dose was reduced by 20–30% if the patients suffered from severe sepsis during the previous cycle or if the hematologic recovery took more than 42 days. Granulocyte colony-stimulating factor was used from the day when the ANC < 500/μL until the time when ANC > 1,000/μL.

Genomic DNA was extracted from the patients’ peripheral blood lymphocytes using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI) according to the manufacturer’s protocol. Approximately 250 ng of genomic DNA was used to genotype each sample using Illumina’s Global Screening Array BeadChip (Illumina, San Diego, CA). The samples were then processed according to the Illumina Infinium assay manual. The quality of the sample was checked based on the sample call rate (> 95%). The clustering quality of each marker was measured based on the GenTrain score. Markers with a score of > 0.7 were used in the study. We performed imputation using the Michigan Imputation Server (https://imputationserver.sph.umich.edu). Markers with low imputation quality (defined as a call rate < 98%), minor allele frequency (MAF) < 1%, a p-value of Hardy-Weinberg equilibrium < 1e–5, duplicated markers, and ambiguous strand markers were excluded from the association analysis. Low-quality samples with a call rate < 95% were used for quality control.

Maximally selected log-rank statistics were tested to obtain the optimal ANC cutoff values that categorized patients into the good and poor prognosis subgroups in terms of progression and treatment-related mortality (TRM) [11]. Cumulative incidences of progression and TRM were estimated using competing risk methods by considering progression and TRM as competing risks [12]. The cumulative incidences of progression and TRM were compared using the Gray’s test. The Fine and Gray competing risk regression model was used to evaluate the effect of the ANC and clinical covariates on the cumulative incidences of progression and TRM. In this analysis, progression and TRM were defined as events, and event-free survival (EFS) was calculated from the date of diagnosis until the date of progression or TRM, whichever occurred first. Overall survival (OS) was calculated from the date of diagnosis until death from any cause. The EFS and OS rates were estimated using the Kaplan-Meier method, and differences in survival curves were compared using the log-rank test. Clinical characteristics were compared between the two groups using the Pearson chi-square test or Fisher exact test for categorical variables and the t test or Kruskal-Wallis rank-sum test for continuous variables. All statistical analyses were performed using R ver. 4.0.0 (R Foundation for Statistical Computing, Vienna, Austria), and p < 0.05 were considered statistically significant.

To investigate significant markers, numeric regression analyses were performed on the ANC with genotypes using HelixTree software (Golden Helix Inc., Bozeman, MT). To assess multiple corrections, the false discovery rate was used. The DAVID functional annotation tool (https://david.ncifcrf.gov/) was used for the analysis. Gene pathway analysis was performed to detect SNPs significantly associated with the ANC after chemotherapy. To visualize significant regional markers, Locus zoom plotting was performed using LocusZoom [13].

Altogether, 313 patients were enrolled in the current study. The median age at diagnosis was 2.4 years (range, 0.0 to 19.2 years). One hundred ninety-one patients (61.0%) had metastatic disease and 110 (35.1%) patients had bone marrow metastasis. Tumor histology according to the INPC was favorable in 137 patients (43.8%), unfavorable in 154 patients (49.2%), and unknown in 22 patients (7.0%). Cytogenetic analysis was only performed in some patients; 68 of 310 patients (21.9%) had MYCN amplification, 17 of 124 patients (13.7%) had 1p deletion, 35 of 125 patients (28.0%) had 11q deletion, and 33 of 123 patients (26.8%) had 17q gain. A total of 173 patients (55.1%) had high-risk tumors.

During the study period, 59 patients showed progression and 25 experienced TRM; the 5-year EFS and OS rates were 73.1%±2.6% and 80.7%±2.3%, respectively. In high-risk patients, progression and TRM occurred in 51 and 21 patients, respectively, and the 5-year EFS and OS rates were 58.6%±3.9% and 69.0%±3.7%, respectively.

First, we analyzed whether the degree of neutropenia after the first cycle of chemotherapy was associated with the cumulative incidence of progression. Maximally selected log-rank statistics determined that an ANC of 32.5/μL was an optimal cutoff point to categorize patients into the good and poor prognosis subgroups in terms of progression. The patient characteristics in the nadir ANC subgroups are presented in Table 1. The frequency of bone marrow involvement did not differ between the two groups. Patients with a high nadir ANC had a higher cumulative incidence of progression than those with a low nadir ANC (p < 0.001). The 5-year cumulative incidences of progression were 12.6%±0.1% and 30.3%±0.2% in the low and high nadir ANC groups, respectively. However, the EFS and OS did not differ between the two groups (p=0.084 and p=0.428, respectively) because TRM showed opposite outcome (high TRM rate in those with a low nadir ANC) (Fig. 1A). Multivariate analysis using the Fine and Gray competing risk regression showed that high nadir ANC, age, bone marrow involvement, and unfavorable tumor histology were poor prognostic factors for the cumulative incidence of progression (Table 2).

Because age was significantly different between the two groups and patients aged < 2 years received chemotherapy at a reduced dose, patients aged > 2 years were included in subgroup analysis. In subgroup analysis, patients with a low nadir ANC had a better outcome in terms of cumulative incidence of progression than those with a high nadir ANC (p=0.004), but the EFS and OS did not differ among the groups (p=0.051 and p=0.304, respectively) (Fig. 1B). The age distribution was different between the two groups (median, 3.3 years in the low nadir ANC group vs. 4.2 years in the high nadir ANC group; p=0.022). Since the ANC might have served as a confounding factor in the analysis of age and outcome, multivariate analysis was conducted using both age and the nadir ANC as covariates. Only high nadir ANC had a significant effect on the cumulative incidence of progression (Table 2).

In patients with high-risk NB, the difference in the cumulative incidence of progression and EFS became more prominent (Fig. 1C), and multivariate analysis showed that only high nadir ANC was significantly associated with the cumulative incidence of progression (Table 2). In the analysis of only non–high-risk patients, there were no significant differences in the cumulative incidence of progression or survival among the nadir ANC groups.

The primary tumor volume was measured at diagnosis and at the first response evaluation after three cycles of induction chemotherapy. The percentage tumor volume at the first response evaluation compared with the initial tumor volume was calculated. Tumor volume was only evaluated in 164 patients who had undergone three-dimensional computed tomography or magnetic resonance imaging at diagnosis and had not undergone front-line surgery at diagnosis. Patients with a low nadir ANC showed greater tumor volume reduction than those with a high nadir ANC (median residual volume %: 17.5% vs. 47.8%, respectively; p=0.004) (Fig. 2A). This difference was more prominent in high-risk patients (median residual volume %, 16.3% vs. 49.7%; p=0.001) (Fig. 2B).

Subgroup analysis only including undifferentiated or poorly differentiated NB was performed because histological tumor differentiation, which can affect the tumor response, was different between the two groups (Table 1). When the analysis was confined to patients with undifferentiated or poorly differentiated NB, those with a low nadir ANC still showed greater tumor volume reduction than those with a high nadir ANC (median, 14.4% vs. 27.3%; p=0.003) (Fig. 2C).

TRM occurred in 25 patients—three patients died during induction chemotherapy, 11 patients showed acute TRM after HDCT, nine patients had late TRM after HDCT, and two patients died due to secondary malignancy. Maximally selected log-rank statistics showed that an ANC of 51.0/μL was the optimal cutoff point to categorize patients based on TRM. Patients with a low nadir ANC (≤ 51.0/μL) showed a higher cumulative incidence of TRM than those with a high nadir ANC (p=0.010). The 5-year cumulative incidences of TRM were 9.43%±0.04% and 1.25%±0.02% in the low and high nadir ANC groups, respectively (Fig. 3A). The EFS and OS did not differ between the two groups. In the analysis of high-risk patients, the cumulative incidence of TRM was higher in patients with a high nadir ANC than in those with a low nadir ANC (Fig. 3B).

A quantile–quantile plot of the association test using the ANC showed a significant deviation of measures at the tail (S1 Fig.), indicating potentially true associations between SNPs and the ANC. The GWAS of common SNPs (MAF > 0.05), which were associated with the ANC, are represented in a Manhattan plot (Fig. 4). The top 30 SNP loci associated with the ANC are listed in S2 Table. A SNP at rs4285743 in the intron of LPHN2 showed the most strong association with the ANC (p=2.9e–09 and P^corr=0.002); a SNP at rs143699161 in the intron of CRHR1 was also significantly associated with the ANC (p=6.1e–09 and P^corr=0.004). To examine the biological function of these SNPs, we performed gene ontology (GO) analysis using DAVID, and the results were presented in S3 Table. Eleven GO terms were observed to have significantly corrected p < 0.05. A number of GO terms related to neural development and synaptic signaling were isolated.