Scoring system for diagnosis and pretreatment risk assessment of neuroblastoma using urinary biomarker combinations

Abstract

The urinary catecholamine metabolites, homovanillic acid (HVA) and vanillylmandelic acid (VMA), are used for the adjunctive diagnosis of neuroblastomas. We aimed to develop a scoring system for the diagnosis and pretreatment risk assessment of neuroblastoma, incorporating age and other urinary catecholamine metabolite combinations. Urine samples from 227 controls (227 samples) and 68 patients with neuroblastoma (228 samples) were evaluated. First, the catecholamine metabolites vanillactic acid (VLA) and 3-methoxytyramine sulfate (MTS) were identified as urinary marker candidates through comprehensive analysis using liquid chromatography–mass spectrometry. The concentrations of these marker candidates and conventional markers were then compared among controls, patients, and numerous risk groups to develop a scoring system. Participants were classified into four groups: control, low risk, intermediate risk, and high risk, and the proportional odds model was fitted using the L2-penalized maximum likelihood method, incorporating age on a monthly scale for adjustment. This scoring model using the novel urine catecholamine metabolite combinations, VLA and MTS, had greater area under the curve values than the model using HVA and VMA for diagnosis (0.978 vs. 0.964), pretreatment risk assessment (low and intermediate risk vs. high risk: 0.866 vs. 0.724; low risk vs. intermediate and high risk: 0.871 vs. 0.680), and prognostic factors (MYCN status: 0.741 vs. 0.369, histology: 0.932 vs. 0.747). The new system also had greater accuracy in detecting missing high-risk neuroblastomas, and in predicting the pretreatment risk at the time of screening. The new scoring system employing VLA and MTS has the potential to replace the conventional adjunctive diagnostic method using HVA and VMA.

Abbreviations
• ANCOVA
• analysis of covariance
• CT
• computed tomography
• HVA
• homovanillic acid
• INPC
• International Neuroblastoma Pathology Classification
• INRG
• International Neuroblastoma Risk Group
• MTS
• 3-methoxytyramine sulfate
• VLA
• vanillactic acid
• VMA
• vanillylmandelic acid
• WMW
• Wilcoxon–Mann–Whitney

INTRODUCTION

Neuroblastoma is a neoplasm of the sympathetic nervous system and the most common extracranial solid tumor in children.¹ Since neuroblastoma is a catecholamine-secreting tumor, the measurement of urinary concentrations of the catecholamine metabolites, homovanillic acid (HVA) and vanillylmandelic acid (VMA), has been widely used for adjunct diagnosis.^2,3 However, nearly 10% of patients may still be missed.^4–9 Moreover, patients exhibit a vast diversity of clinical presentations, ranging from spontaneous regression to rapid fatal progression, and therefore should be treated according to their risk categories based on risk factors.^10–12 Nonetheless, a consensus for the prognostic values of HVA and VMA has not been reached.^9,11,13–20 Therefore, a novel predictive approach incorporating noninvasive urinary markers is sought for a more accurate diagnostic and prognostic assessment.

Advances in mass spectrometry (MS) have facilitated the detection of urinary metabolites present at trace amounts; thus, the latest MS techniques may detect other markers more accurately than conventional markers. We previously identified urinary diagnostic marker candidates through comprehensive analysis utilizing liquid chromatography–MS (LC/MS).²¹

Therefore, this study aimed to establish an algorithm for developing a scoring system for both diagnosing and assessing the pretreatment risk for neuroblastoma using age and multiple markers since the excretion of catecholamine metabolites is age-dependent. Additionally, we investigated the efficacy and feasibility of novel urinary catecholamine metabolite combinations, such as vanillactic acid (VLA) and 3-methoxytyramine sulfate (MTS), which, by comprehensive analysis using LC/MS, were found to contribute to neuroblastoma etiology.

MATERIALS AND METHODS Patients and urine sample collection

Pediatric patients with histologically confirmed neuroblastoma who were treated at five institutions in Japan (Nagoya University Hospital; National Center for Child Health and Development; University Hospital, Kyoto Prefectural University of Medicine; Hiroshima University Hospital; and Saitama Children's Medical Center) between January 2016 and August 2022 were enrolled. Additionally, pediatric patients with no known cancer, admitted to Nagoya University Hospital for inguinal hernia treatment between June 2016 and August 2021 were included as the control group.

Spot urine samples from patients with neuroblastoma were collected once before treatment and then at different times during the treatment course such as after surgery, chemotherapy, and hematopoietic stem cell transplantation. For controls, spot urine samples were collected only once before the treatment. Information on age at sample collection, sex, stage, tumor location, MYCN amplification, prognostic groups according to the International Neuroblastoma Pathology Classification (INPC), International Neuroblastoma Risk Group (INRG) stage, and INRG pretreatment risk classification was retrieved from patient medical records.^12,22

Comprehensive LC/MS analysis in the qualitative nontargeted screening of urinary metabolites

To identify urinary marker candidates for neuroblastoma, a comprehensive analysis of the qualitative nontargeted screening of urine samples from control participants and patients with primary neuroblastoma before treatment was conducted using a sensitive and high-resolution Q-Exactive Orbitrap Mass spectrometer (Thermo Fisher Scientific), as previously described.^21,23 Comprehensive LC/MS analysis of the samples was outsourced to Metabolon, Inc.

To select marker candidates useful for the diagnosis of neuroblastoma, the increase or decrease of individual metabolites between controls and patients was evaluated using the Wilcoxon–Mann–Whitney (WMW) test. Subsequently, among the molecules with p < 0.05 in the WMW test, random forest classification was performed to evaluate the importance of these metabolites for diagnosis; unknown substances were excluded from the candidates. Finally, highly relevant metabolites, other than HVA and VMA, were selected as urinary marker candidates.

Consequently, VLA and MTS were identified as urinary marker candidates for neuroblastoma, and subsequent studies were performed focusing on these candidates together with the conventional markers in clinical use, HVA and VMA.

Development of the quantitative LC/MS methods for urinary metabolites

First, a quantitative measurement method was designed for HVA, VMA, VLA, and MTS, as quantitative measurement with high accuracy in a short time is a requirement for practical applications. Pretreatment procedures for urine sample target compounds and LC/MS conditions for quantitative measurements are described in Appendix S1. Calibration curve samples and mimicked samples with known concentrations were prepared to assess the adequacy of the developed measurement method. Consequently, the concentrations calculated using the developed measurement method demonstrated reasonable accuracy and variability.

Quantitative LC/MS analysis of urinary metabolites Quantitative validation of urinary metabolites

The concentrations of the four metabolites in patients' and control participants' urine samples were measured using the quantitative measurement method described in Appendix S1. If the measured values exceeded the upper limit of the calibration range, the concentrations were extrapolated. The creatinine concentrations were measured using the enzymatic method, and the data were analyzed using the concentration of metabolites standardized by the creatinine concentration. This experiment was outsourced to CMIC Pharma Science Co., Ltd. and Sumika Chemical Analysis Service, Ltd.

The four urinary metabolite levels were compared between the following groups: control participants versus patients with primary neuroblastoma before treatment, control participants versus patients with residual tumor after treatment, patients with low- and intermediate-risk neuroblastomas versus patients with high-risk neuroblastomas, and patients with low-risk neuroblastomas versus patients with intermediate- and high-risk neuroblastomas.

Age-related effects of urinary metabolites

Logistic regression analysis was used to investigate age-related effects of the four urinary metabolites. The model was fitted to age and concentrations of the individual metabolites, and scatter plots with age-dependent determination boundaries were plotted.

Development of the scoring system for neuroblastoma

To develop a scoring system that is accurate for diagnosis and associated with the pretreatment risk of neuroblastoma, the target variable was defined as a combination of diagnosis and the INRG pretreatment risk group. Participants were classified into four groups, and the proportional odds model was fitted using the L2-penalized maximum likelihood method,²⁴ which incorporated age on a monthly scale for adjustment. The proportional odds model is a natural expansion of the logistic regression model and provides probabilities for each level of an ordinal outcome. Since the fitted model takes real numbers and is not easy to handle in clinical practice, it was approximated to develop a discrete scoring system following the procedure of the Framingham Study risk score functions.²⁵ The model evaluation measures were estimated by bootstrapping to avoid overfitting. Additional detailed descriptions are provided in Appendix S1.

Moreover, for pretreatment risk assessment prior to invasive biopsy, prediction models for INRG pretreatment risk assessment were also developed incorporating three types of variables: age, urinary markers, and INRG stage evaluated by imaging, as shown in Appendix S1.

Ethical declaration

This study was approved by the Ethics Committee of Nagoya University Hospital (approval no: 2016-0303). The legal guardians of all the participants provided written informed consent. All the study procedures were performed in accordance with the guidelines of the Declaration of Helsinki.

Statistical analyses

Patient characteristics are presented as the mean ± standard deviation. For data on patient characteristics, univariate analyses were performed using Fisher's exact test for sex and the WMW test for age. Differences between the urinary metabolite levels in the two groups were also evaluated using the WMW test and analysis of covariance (ANCOVA) with age adjustment. Statistical significance was set at p < 0.05. The concentration of each metabolite in the quantitative analysis was transformed into a logarithmic scale for ANCOVA and the analysis in the sections “Age-related effects of urinary metabolites” and “Development of the scoring system for neuroblastoma” for better fitting.

Some data used in the quantitative analysis and model development were provided by the same participants in the comprehensive analysis. Because the candidate markers were found based on statistical association in the comprehensive analysis, it is possible that these candidates were overly associated with the target variables in the quantitative analysis of data containing overlapping participants. Therefore, for all the above analyses, we also analyzed the data, excluding participants who were included in the comprehensive analysis, as shown in Appendix S1, Tables S1–S3, and Figures S1–S3.

RESULTS Participants' characteristics

In total, 295 participants were included in this study: 227 control participants and 68 pediatric patients with neuroblastoma. Comprehensive analysis was performed on 110 urine samples from 110 control participants and on 18 urine samples from 18 pediatric patients with primary neuroblastoma before treatment. Of the 110 samples from control participants, 39 samples overlapped with those in our previous study, and of the 18 samples from patients with neuroblastoma 15 samples overlapped.²¹ Quantitative analysis was performed on 155 urine samples from 155 control participants and on 281 urine samples from 68 pediatric patients with neuroblastoma. Of the 281 samples, 41 samples were obtained from 41 patients with primary neuroblastoma before treatment, and 240 urine samples were obtained from 54 patients with residual tumors after treatment (surgery, chemotherapy, hematopoietic stem cell transplantation, etc.).

The control participants and patients used for the comprehensive and quantitative analyses are presented in Tables 1 and 2. There were significant differences between the control participants and patients with neuroblastoma in both the comprehensive and quantitative analyses in terms of age at sample collection.

TABLE 1 Characteristics of control participants and patients with neuroblastoma based on comprehensive analysis.

Variable	Control participants	Patients with neuroblastoma	p-Value
Patient, n	110	18	(−)
Age at sample collection, months (mean ± SD)	59.8 ± 45.6	29.7 ± 21.9	0.004
Sex, male/female	57/53	8/10	0.372
Primary tumor localization
Adrenal gland, n	(−)	9	(−)
Mediastinum, n	(−)	4	(−)
Retroperitoneum, n	(−)	4	(−)
Cervical sympathetic trunk, n	(−)	1	(−)
MYCN status
Amplified/not amplified	(−)	2/16	(−)
Histologic prognostic group (INPC)
Favorable/unfavorable	(−)	8/8	(−)
INRG stage
L1	(−)	3	(−)
L2	(−)	2	(−)
M	(−)	12	(−)
MS	(−)	1	(−)
INRG pretreatment risk classification
Low	(−)	5 (27.8%)	(−)
Intermediate	(−)	5 (27.8%)	(−)
High	(−)	8 (44.4%)	(−)

Note: p-Values from Fisher's exact test for sex and Wilcoxon–Mann–Whitney test for age.

Abbreviations: (−), not applicable; INPC, International Neuroblastoma Pathology Classification; INRG, International Neuroblastoma Risk Group; SD, standard deviation.

TABLE 2 Characteristics of control participants and patients with neuroblastoma based on quantitative analysis.

Variable	Control participants	Patients with primary neuroblastoma before treatment	p-Value (vs. control participants)	Patients with residual tumor after treatment	p-value (vs. control participants)
Patient, n	155	41	(−)	54	(−)
Age at sample collection, months (mean ± SD)	47.9 ± 36.9	36.4 ± 32.8	0.036	53.6 ± 27.1	0.006
Sex, male/female	81/74	21/20	0.522	30/24	0.398
Primary tumor localization
Adrenal gland, n	(−)	20	(−)	32	(−)
Mediastinum, n	(−)	8	(−)	14	(−)
Retroperitoneum, n	(−)	12	(−)	6	(−)
Cervical sympathetic trunk, n	(−)	1	(−)	1	(−)
Unknown	(−)	0	(−)	1	(−)
MYCN status
Amplified/Not amplified	(−)	7/32	(−)	12/40	(−)
Unknown	(−)	2	(−)	2	(−)
Histologic prognostic group (INPC)
Favorable/Unfavorable	(−)	9/20	(−)	7/29	(−)
Unknown	(−)	12	(−)	18	(−)
INRG stage
L1	(−)	5	(−)	2	(−)
L2	(−)	11	(−)	9	(−)
M	(−)	21	(−)	38	(−)
MS	(−)	4	(−)	4	(−)
Unknown	(−)	0	(−)	1	(−)
INRG pretreatment risk classification
Low	(−)	10	(−)	6	(−)
Intermediate	(−)	9	(−)	12	(−)
High	(−)	22	(−)	3	(−)

Note: p-Values from Fisher's exact test for sex and Wilcoxon–Mann–Whitney test for age.

Abbreviations: (−), not applicable; INPC, International Neuroblastoma Pathology Classification; INRG, International Neuroblastoma Risk Group; SD, standard deviation.

Comprehensive LC/MS analysis in the qualitative nontargeted screening of urinary metabolites

Approximately 1500 different metabolites were detected in the urine samples of control participants and patients with neuroblastomas. Based on the random forest classification for the screening of neuroblastoma, the rank order of the catecholamine metabolites according to their importance among all metabolites was HVA (first place), MTS (second place), VMA (third place), and VLA (fourth place). Therefore, VLA and MTS were selected as novel urinary marker candidates for neuroblastoma.

Quantitative LC/MS analysis of urinary metabolites Quantitative validation of urinary metabolites

The levels of all four metabolites were significantly higher in 41 urine samples from 41 patients with primary neuroblastoma before treatment compared with 155 urine samples from 155 control participants (HVA, VMA, VLA, and MTS: p < 0.001 [WMW, ANCOVA]), as shown in Figure 1A. The four metabolite levels were significantly higher in 240 urine samples from 54 patients with residual tumors after treatment compared with 155 urine samples from 155 control participants (HVA, VMA, VLA, and MTS: p < 0.001 [WMW, ANCOVA]), as shown in Figure 1B.

FIGURE 1. Quantitative validation of urinary metabolites. The levels of metabolites (median, interquartile range, minimum, and maximum values) in urine from control participants, (A) patients with primary neuroblastoma before treatment, and (B) patients with residual tumors. (C–E) The levels of metabolites in urine from control participants and patients with low-, intermediate-, and high-risk neuroblastomas. HVA, homovanillic acid; NB, neuroblastoma; VMA, vanillylmandelic acid; VLA, vanillactic acid; MTS, 3-methoxytyramine sulfate. ¶ Wilcoxon–Mann–Whitney test. † Analysis of covariance.

The four metabolite levels in 155 urine samples from 155 control participants and 41 urine samples from patients with low-risk (n = 10), intermediate-risk (n = 9), and high-risk (n = 22) neuroblastomas (all primary neuroblastomas before treatment) are presented in Figure 1C. Significant differences were observed only in VLA and MTS levels between low- and intermediate-risk neuroblastomas and high-risk neuroblastomas using the WMW test (HVA: p = 0.148; VMA: p = 0.120; VLA: p = 0.001; MTS: p = 0.007 [WMW]). Significant differences were observed in all urinary metabolite levels between the two groups using ANCOVA, as illustrated in Figure 1D (HVA: p = 0.027; VMA: p = 0.047; VLA: p < 0.001; MTS: p < 0.001 [ANCOVA]). Significant differences in VLA and MTS levels between low-risk neuroblastomas and intermediate- and high-risk neuroblastomas were observed using both the WMW test and ANCOVA, as shown in Figure 1E (HVA: p = 0.093, 0.064; VMA: p = 0.141, 0.112; VLA: p < 0.001, < 0.001; MTS: p = 0.010, 0.007 [WMW, ANCOVA]).

Age-related effects of urinary metabolites

Scatter plots of the four metabolite levels in urine samples from control participants and patients with primary neuroblastomas before treatment are presented in Figure 2. For all metabolites, the decision boundary was age-dependent.

FIGURE 2. Variation in urinary metabolite levels with age. Decision boundaries from logistic regression. The solid and dashed lines indicate the decision boundaries with a 50% and 10% probability of being positive, respectively. Note that the probability of each boundary is not important because it depends on the positive/control ratio of the participants in this study. HVA, homovanillic acid; NB, neuroblastoma; VLA, vanillactic acid; VMA, vanillylmandelic acid; MTS, 3-methoxytyramine sulfate; NB, neuroblastoma.

Development of the scoring system for neuroblastoma

The accuracy of the model using age and urinary markers is shown in Table 3. The model using the novel urinary markers (VLA and MTS) had higher area under the curve (AUC) values than the model using the conventional urinary markers (HVA and VMA) in terms of diagnosis. The receiver-operating characteristic (ROC) curves of these two models are presented in Figure 3A,B. Note that these ROC curves were drawn using all data. The threshold was set such that the sensitivities were greater than 95%. The model with VLA and MTS had five false-positive and two false-negative cases, and the model with HVA and VMA had 30 false-positive and two false-negative cases, respectively. While the false-negative cases using VLA and MTS were in the low- and intermediate-risk groups in terms of INRG pretreatment risk, the false-negative cases of the model with HVA and VMA were both in the high-risk group. Moreover, the model with VLA and MTS had higher AUC values than the model with HVA and VMA in terms of pretreatment risk assessment (low- and intermediate-risk neuroblastomas vs. high-risk neuroblastomas, low-risk neuroblastomas vs. intermediate- and high-risk neuroblastomas), and prognostic factors (MYCN status and histologic prognostic groups according to INPC), as shown in Table 3.

TABLE 3 Accuracy of the scoring system for neuroblastoma.

	Diagnosis	Low- and intermediate-risk NB vs. high-risk NB	Low-risk NB vs. intermediate-and high-risk NB	MYCN status	Histologic prognostic groups (INPC)
	AUC (95% CI)	AUC (95% CI)	AUC (95% CI)	AUC (95% CI)	AUC (95% CI)
HVA	0.962 (0.911–0.998)	0.706 (0.460–0.929)	0.713 (0.424–0.958)	0.459 (0.115–0.814)	0.805 (0.572–0.986)
VMA	0.958 (0.902–0.998)	0.683 (0.429–0.918)	0.654 (0.344–0.926)	0.273 (0.037–0.560)	0.697 (0.391–0.953)
VLA	0.943 (0.878–0.995)	0.842 (0.664–0.985)	0.888 (0.740–0.995)	0.766 (0.492–0.970)	0.860 (0.674–0.996)
MTS	0.968 (0.921–0.998)	0.814 (0.613–0.976)	0.791 (0.539–0.988)	0.667 (0.322–0.948)	0.960 (0.884–0.998)
HVA + VMA	0.964 (0.916–0.998)	0.724 (0.480–0.943)	0.680 (0.382–0.94)	0.369 (0.058–0.713)	0.747 (0.475–0.968)
VLA + MTS	0.978 (0.948–0.998)	0.866 (0.696–0.992)	0.871 (0.688–0.998)	0.741 (0.448–0.966)	0.932 (0.818–0.998)

Abbreviations: AUC, area under the curve; CI, confidence interval; HVA, homovanillic acid; INPC, International Neuroblastoma Pathology Classification; MTS, 3-methoxytyramine sulfate; NB, neuroblastoma; VLA, vanillactic acid; VMA, vanillylmandelic acid.

FIGURE 3. Accuracy of the scoring system for neuroblastoma. ROC of the models using age and urinary markers of the raw score using HVA and VMA (A), raw score using VLA and MTS (B), and score for the sheet shown in Table 4 using VLA and MTS (C). AUC, area under the curve; HVA, homovanillic acid; MTS, 3-methoxytyramine sulfate; ROC, receiver operating characteristic; VLA, vanillactic acid; VMA, vanillylmandelic acid.

Thus, the new model's formula with age and a better combination of urinary markers (VLA and MTS) was 0.913 × age (months) + 1.36 × log₁₀ (VLA) + 1.51 × log₁₀ (MTS), where the predictors were standardized. This model was converted into a scoring system, and the score sheet was developed as presented in Table 4. In the scoring system, a one-point increase corresponds approximately to an odds ratio of 1.13. Based on the ROC curves, neuroblastoma can be diagnosed with a sensitivity of 0.951 and a specificity of 0.948 when the cutoff value was set to 28 points, as shown in Figure 3C. This score was compared between the diagnosis (control participants/patients with primary neuroblastoma before treatment), INRG pretreatment risk classification (low-/intermediate-/high-risk neuroblastomas), MYCN status (amplified neuroblastomas/nonamplified neuroblastomas), and histologic prognostic groups according to INPC (favorable neuroblastomas/unfavorable neuroblastomas), as illustrated in Figure 4. The scores tended to be higher in patients with primary neuroblastoma before treatment than in control participants. There was a significant difference between the score of the model using VLA/MTS and INPC (p = 0.015), whereas there was no significant difference between that and MYCN amplification (p = 0.53). There was a significant difference (p = 0.043) in the model score using VLA/MTS in MYCN-amplified stage M (n = 3) and MYCN-nonamplified stage M (n = 12).

TABLE 4 The scoring system for neuroblastoma, combining age and the novel urinary markers, vanillactic acid (VLA) and 3-methoxytyramine sulfate (MTS).

Variable	Level	Score
Age (months)	0–8	0
	9–17	2
	18–26	4
	27–35	6
	36–44	8
	45–53	10
	54+	12
VLA	<0.28 (mean + 0.5 SD of control)	0
	0.28 < =, <0.40 (mean + SD of control)	7
	0.40 < =, <0.58 (mean + 1.5 SD of control)	11
	0.58 < =, <0.83 (mean + 2 SD of control)	13
	0.83 < =, <1.20 (mean + 2.5 SD of control)	15
	1.20 < = (mean + 2.5 SD of control)	29
MTS	<0.46 (mean + 0.5 SD of control)	0
	0.46 < =, <0.63 (mean + SD of control)	9
	0.63 < =, <0.86 (mean + 1.5 SD of control)	12
	0.86 < =, <1.18 (mean + 2 SD of control)	15
	1.18 < =, <1.62 (mean + 2.5 SD of control)	18
	1.62 < = (mean + 2.5 SD of control)	34

Abbreviation: SD, standard deviation.

FIGURE 4. Comparison of the model scores. Comparison of the model scores using VLA and MTS among diagnosis (A), INRG pretreatment risk classification (B), MYCN status (C), and histologic prognostic groups according to INPC (D). INPC, International Neuroblastoma Pathology Classification; INRG, International Neuroblastoma Risk Group; MTS, 3-methoxytyramine sulfate; NB, neuroblastoma; VLA, vanillactic acid.

DISCUSSION

To the best of our knowledge, this is the first study to propose a scoring system for the diagnosis and pretreatment risk assessment of neuroblastoma that incorporates age and urinary catecholamine metabolite combinations.

Several studies have evaluated the efficacy of urinary catecholamine metabolites as diagnostic or prognostic biomarkers. Although a high level of urinary VLA has previously been reported to indicate a poor prognosis,^15,26–28 this is the first study to develop a quantitative method for LC/MS with high accuracy and validate its utility. 3-methoxytyramine has also been reported as a good marker for diagnosis and prognosis, reflecting increased MYCN activity in the tumor, and for monitoring therapy response.^{19,20,29–31} However, the only report on MTS, which is a sulfate conjugate of 3-methoxytyramine, is our previous study showing that urinary MTS is a good diagnostic marker.²¹

In this study, all four metabolite concentrations were significantly higher in patients with neuroblastoma compared with control participants. Additionally, age-adjusted ANCOVA showed that all four metabolites were useful for distinguishing high-risk neuroblastomas, whereas only VLA and MTS were useful for distinguishing low-risk neuroblastomas. Therefore, VLA and MTS might be useful for both diagnosis and pretreatment risk assessment.

We subsequently developed a scoring system for neuroblastoma based on age, considering that metabolite excretion is age-dependent. Compared with the model using HVA and VMA, the scoring model using VLA and MTS had higher accuracy in terms of diagnosis, had fewer false positives, and did not miss patients with high-risk neuroblastomas. Thus, this scoring system can simplify diagnosis based on complex statistical models that consider age and multiple markers in a simple table, without the need for calculators or computers, and is easy to use in clinical practice.²⁵ Moreover, the model using VLA and MTS had higher accuracy in terms of pretreatment risk assessment than that using HVA and VMA. The scores tended to be higher in patients with primary neuroblastoma before treatment than in control participants. Similarly, the scores of the model using VLA and MTS are also considered to be associated with poor prognostic factors such as MYCN amplification and histologic classification. Although no statistically significant difference was shown between the scores of the model using VLA and MTS and MYCN amplification in this study (p = 0.52), the distribution of score was skewed upward in the amplification group, as shown in Figure 4C. This aspect should be considered in future studies. Therefore, the new scoring model based on VLA and MTS was able to detect patients with greater accuracy while not missing high-risk neuroblastomas, and it aided in selecting treatment intensity based on the pretreatment risk assessment at the time of screening using only information on age and noninvasive urinary examination, without the need of computed tomography (CT) or biopsy. Thus, the new scoring system may overcome the limitations of the previous mass screening using HVA and VMA.

Most neuroblastomas detected by screening using HVA and VMA have favorable biological features that are associated with a high rate of spontaneous regression or maturation into benign ganglioneuromas, and only a few have unfavorable biological features.^32,33 Two prospective nonrandomized controlled trials comparing neuroblastoma-related mortality in nonscreened and screened areas reported that mass screening did not reduce the incidence of disseminated neuroblastoma and related mortality; conversely, they suggested the possibility of overdiagnosis. Thus, many children with neuroblastoma diagnosed by mass screening may undergo unnecessary treatment for a tumor that would otherwise spontaneously regress.^7,34 However, as our scoring system with VLA and MTS was able to detect patients with greater accuracy without missing high-risk neuroblastomas, it has the potential to reduce neuroblastoma-related mortality. In addition, it would be useful for supporting a noninvasive treatment plan, whether to “wait and see” or to treat immediately and intensively, if it could predict cases suitable for the “wait and see” strategy. The Children's Oncology Group (COG) study reported that small adrenal masses in infants younger than 6 months may safely be observed without biopsy and surgical intervention (the “wait and see” strategy).³⁵ Additional studies, including an expansion of criteria allowing observation without surgery, are underway. Moreover, we aim to investigate the scores and prognoses of these cases in the future.

Nonetheless, this study had some limitations. Although patients with neuroblastoma treated at five high-volume institutions for children in Japan were included in this study, the number of participants was insufficient due to the rarity of the disease. Although random data splitting is often used for model validation, data splitting has an unignorable loss due to the reduction of sample size, and its inefficiency has been pointed out in the commentary of the TRIPOD Statement.³⁶ Considering that the sample size is not necessarily abundant, the bootstrap method, as recommended in the commentary, was used to ensure internal validity instead of data splitting. Furthermore, we confirmed that comparable results were achieved even when the data used in the comprehensive LC/MS analysis were excluded. Thirty-eight control participants and 11 patients with neuroblastoma underwent both comprehensive and quantitative analysis. Owing to the small number of cases, quantitative validation and score development were performed, including participants who underwent both the analyses. The results of the quantitative validation and score development excluding participants who underwent both analyses were similar to those including those participants, as shown in Appendix S1. We believe that we have exhausted all possibilities with the data that are now accessible, and that further validation—including external validation—should be carried out in future research. Future prospective clinical studies need to be conducted to develop an optimal scoring system for both diagnosis and pretreatment risk assessment based on both the INRG pretreatment classification and the COG ver. 2 risk classifier and be validated in a larger population.³⁷ Future studies should also evaluate the usefulness of the monitoring system, including the response to treatment and likelihood of relapse.

In conclusion, compared with the system using HVA and VMA, the new scoring system incorporating age and novel urinary catecholamine metabolite combinations—VLA and MTS—might detect patients with neuroblastoma with greater accuracy, without missing high-risk neuroblastomas, and might predict the pretreatment risk easily and noninvasively, without the need of biopsy or CT during screening. Therefore, this scoring system may replace the conventional adjunctive diagnostic methods. However, future validation studies are warranted.

AUTHOR CONTRIBUTIONS

Hizuru Amano: Conceptualization; methodology; writing – original draft; writing – review and editing. Hiroo Uchida: Conceptualization; funding acquisition; methodology; writing – review and editing. Kazuharu Harada: Formal analysis; writing – review and editing. Atsushi Narita: Data curation; writing – review and editing. Shigehisa Fumino: Data curation; writing – review and editing. Yuji Yamada: Data curation; writing – review and editing. Shun Kumano: Writing – review and editing. Mayumi Abe: Writing – review and editing. Takashi Ishigaki: Writing – review and editing. Minoru Sakairi: Writing – review and editing. Chiyoe Shirota: Data curation; writing – review and editing. Takahisa Tainaka: Data curation; writing – review and editing. Wataru Sumida: Data curation; writing – review and editing. Kazuki Yokota: Data curation; writing – review and editing. Satoshi Makita: Data curation; writing – review and editing. Shuhei Karakawa: Data curation; writing – review and editing. Yuichi Mitani: Data curation; writing – review and editing. Shojiro Matsumoto: Data curation; writing – review and editing. Yutaka Tomioka: Writing – review and editing. Hideki Muramatsu: Data curation; writing – review and editing. Nobuhiro Nishio: Data curation; writing – review and editing. Tsuyoshi Osawa: Writing – review and editing. Masataka Taguri: Formal analysis; writing – review and editing. Katsuyoshi Koh: Data curation; writing – review and editing. Tatsuro Tajiri: Data curation; writing – review and editing. Motohiro Kato: Data curation; writing – review and editing. Kimikazu Matsumoto: Data curation; writing – review and editing. Yoshiyuki Takahashi: Data curation; writing – review and editing. Akinari Hinoki: Conceptualization; methodology; project administration; writing – review and editing.

ACKNOWLEDGMENTS

This study was a joint research project with Hitachi, Ltd. (Grant number 2617Dm-08b) and was supported by AMED under Grant Number JP22cm0106481. We are grateful to Dr. Yoshiharu Hayashi at CMIC Pharma Science Co., Ltd. for the development of the quantitative LC/MS methods. We would also like to acknowledge all the participants and their families in this study.

FUNDING INFORMATION

Hiroo Uchida received joint research funding from Hitachi, Ltd. (Grant number 2617Dm-08b) and a grant from AMED (Grant Number JP22cm0106481). Hiroo Uchida, Akinari Hinoki, Minoru Sakairi, and Mayumi Abe have a patent “Method, Kit and Apparatus for Cancer Detections Using Urinary Tumor Markers”; U.S. Patent.

CONFLICT OF INTEREST STATEMENT

Hiroo Uchida received joint research funding from Hitachi, Ltd. (Grant number 2617Dm-08b) and a grant from AMED (Grant Number JP22cm0106481). Hiroo Uchida, Akinari Hinoki, Minoru Sakairi, and Mayumi Abe have a patent “Method, Kit and Apparatus for Cancer Detections Using Urinary Tumor Markers”; U.S. Patent 11,415,583 B2. Shun Kumano and Mayumi Abe are employees of the Research & Development Group, Hitachi, Ltd. No disclosures have been reported by the other authors.

DATA AVAILABILITY STATEMENT

The datasets generated and/or analyzed during the study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT

Approval of the research protocol by an Institutional Reviewer Board: This study was approved by the Ethics Committee of Nagoya University Hospital (approval no: 2016-0303).

Informed Consent: The legal guardians of all the participants provided written informed consent.

Registry and the Registration No. of the study/trial: N/A.

Animal Studies: N/A.