Tumor size predicts metastatic disease at diagnosis and prognosis in unilateral Wilms tumor: a population based study with external validation

Abstract

Background

The impact of tumor size on prognosis, metastasis, and treatment response in Wilms tumor (WT) remains incompletely understood. This study aimed to elucidate the influence of tumor size on survival, metastatic risk, and the efficacy of radiotherapy and chemotherapy in WT patients.

Methods

We analyzed WT patients in the Surveillance, Epidemiology, and End Results (SEER) program (2000–2022) and from our institution’s electronic medical records (2013–2024). Survival analysis was performed using Kaplan-Meier curves and log-rank tests. Prognostic factors were identified via univariate and multivariate Cox proportional hazards models, and propensity score matching (PSM) was used to balance baseline characteristics.

Results

A total of 1916 patients from the SEER cohort and 237 patients from our institutional cohort with WT were included in this study. In the SEER cohort, patients with tumor size > 10.4 cm had significantly worse survival compared to those with tumors ≤ 10.4 cm. Restricted cubic spline (RCS) analysis revealed linear associations between tumor size and patient outcomes for overall survival (OS) (p overall = 0.011, p nonlinear = 0.906) and cancer-specific survival (CSS) (p overall = 0.010, p nonlinear = 0.998). Multivariate analysis confirmed that larger tumor size was an independent risk factor for poorer OS (HR = 1.79, p = 0.011). The metastasis rate was significantly lower in the small tumor group than in the large tumor group (87/762 vs. 187/880, p < 0.001). Among all patients who received chemotherapy, those with larger tumors had worse prognosis, particularly in the distant metastasis stage, showing a 9.8% reduction in 5-year OS. These findings were consistently validated using our institutional dataset.

Conclusion

Larger tumor size is associated with higher metastatic risk and unfavorable prognosis in WT. This highlights its potential value as a practical supplementary indicator for risk stratification and clinical decision-making, particularly in resource-limited settings.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12672-026-05239-y.

Introduction

Wilms tumor (WT) is the most prevalent malignant kidney tumor in children, representing around 7% of all pediatric malignancies [1, 2]. The current approach to treatment involves surgical resection, followed by adjuvant chemotherapy for nearly all patients, and radiation therapy when indicated. Prognostic risk stratification—guiding personalized treatment—incorporates factors such as tumor stage, age at diagnosis, histology, and loss of heterozygosity at 1p and 16q. As a key clinical-pathological feature of solid tumors, tumor size holds potential as a prognostic indicator and may influence therapeutic decision-making, warranting further investigation.

Tumor size is defined as the maximum tumor diameter [3]. Previous studies have demonstrated its prognostic importance in T1a renal cell carcinoma [4], adrenocortical carcinoma [5], colorectal cancer [6], and neuroblastoma [7]. The Children’s Oncology Group (COG) used resected tumor weight to determine which patients may be treated with nephrectomy alone [8]. The International Society of Paediatric Oncology (SIOP) integrates tumor volume following preoperative chemotherapy into its risk stratification [9]. However, neither approach has addressed maximum tumor diameter—a simple and reproducible parameter. Furthermore, the relationship between tumor size and metastasis, its impact on sensitivity to chemotherapy and radiotherapy remain unclear.

This study aims to investigate the effect of tumor size on prognosis and treatment decisions in WT using data from the Surveillance, Epidemiology, and End Results (SEER) database and a national-level regional medical center.

Methods

Data source and collection

The study population was identified using data from the SEER Program. SEER*Stat software (version 9.4.41) was utilized to extract the relevant data. The cohort was selected from the SEER Research Data, 17 Registries, Nov 2024 Sub (2000–2022) dataset.

Additionally, clinical data for WT patients were collected from the electronic medical records of the Children’s Hospital of Chongqing Medical University from 2013 to 2024. This retrospective study was approved by the Medical Research Ethics Committee of the Affiliated Children’s Hospital of Chongqing Medical University (Approval No. 1324, 2024), which waived the need for parental or guardian informed consent in accordance with the Declaration of Helsinki. The collected data included demographic characteristics (age, sex, race), tumor characteristics (stage, tumor size, laterality, metastatic status), treatment details (type of surgery, chemotherapy, radiotherapy), and follow-up information (overall and cancer-specific survival status, survival months). Eligible participants were individuals under 18 years of age diagnosed with WT, corresponding to ICD-O-3 code 8960. Exclusion criteria comprised unknown tumor stage or size, age over 18 years, bilateral tumors or side unspecified, unknown race, unknown metastatic status, unknown radiotherapy, chemotherapy, or surgical approach, incomplete survival data, and less than one month of follow-up. The specific study workflow and inclusion/exclusion criteria are detailed in Fig. 1. Tumor size was defined as the maximum diameter measured on the specimen after surgical removal. For the patients who received preoperative chemotherapy, tumor size was measured on pre-treatment cross-sectional imaging (CT) prior to the initiation of chemotherapy, as pathological measurement after neoadjuvant therapy may not accurately reflect the original tumor burden. Overall survival (OS) was calculated from the time of surgery to the date of last contact or death due to any cause. Cancer-specific survival (CSS) was calculated from the time of surgery to the date of last contact or death due to cancer, whereas deaths due to other causes were censored. Local organ invasion was defined as the direct extension of the primary tumor into adjacent structures (e.g., adrenal gland, liver, bowel, or major vasculature). The assessment of local invasion was fundamentally established by histopathological evaluation of the resected specimen following nephrectomy.

Fig. 1.

Flow chart of the study

Treatment

Patients in the SEER database are treated according to COG protocols, which generally favor upfront nephrectomy followed by adjuvant chemotherapy and/or radiotherapy based on surgical-pathological staging. At our institution, prior to initiating treatment, all patients undergo chest CT and contrast-enhanced abdominal CT scans. Neoadjuvant chemotherapy is indicated for the following conditions: inferior vena cava tumor thrombus, invasion of adjacent organs (such as the spleen, pancreas, or colon), large tumor size precluding complete resection, presence of distant metastases, or bilateral Wilms tumor. The chemotherapy regimen follows the SIOP-WT 2001 protocol, utilizing a combination of two or three agents (vincristine and actinomycin D, with doxorubicin added in cases with metastasis). Patients undergoing upfront surgery are staged according to the COG staging system and receive corresponding postoperative treatment.

Statistical analysis

Data processing and analysis were performed using R software (version 4.4.1). Categorical variables are reported as absolute and relative frequencies and were compared using the χ² test or Fisher’s exact test, as appropriate. Continuous variables are expressed as medians with interquartile ranges (IQR) and were analyzed using the t tests or nonparametric Wilcoxon rank sum test.

The optimal cutoff values for tumor size were determined using X-tile software (https://medicine.yale.edu/lab/rimm/research/software/), which evaluates all possible divisions by exhaustive log-rank χ² testing and identifies the cut-point associated with the smallest p-value [10]. Survival outcomes between groups were compared using the log-rank test, and Kaplan–Meier curves were generated to visualize survival. Both univariate and multivariate Cox proportional hazards models were applied to identify variables associated with survival; results are reported as hazard ratios (HRs) with 95% confidence intervals (CIs). To mitigate baseline imbalances and potential selection bias, propensity score matching (PSM) was performed using the R package MatchIt (version 4.7.0) with a 1:1 nearest neighborhood matching ratio and a caliper width of 0.2. A two-sided p-value < 0.05 was considered statistically significant.

Result

Patient demographics

According to the inclusion and exclusion criteria, a total of 1,916 WT patients from the SEER database were included in this study (Fig. 1). Based on X-tile analysis of the SEER cohort, a cutoff value of 10.4 cm was identified (Fig. 2 A–C). Tumors were subsequently categorized into “Small” and “Large” groups for further analysis. The demographic and clinical characteristics of the cohort are summarized in Table 1. Among these, 46.4% (n = 889) were male and 74.11% (n = 438) were White. The median age at diagnosis was 3 years (IQR: 4 years). Tumor stage distribution was as follows: 23.02% (n = 441) were classified as Distant, 33.87% (n = 649) as Regional, and 43.11% (n = 826) as Localized. The median tumor size was 11.0 cm (IQR: 5.30 cm), and the median follow-up time was 91 months (IQR: 119 months).

Fig. 2.

Determination of the optimal prognostic cut-off value for tumor size using X-tile analysis in the SEER cohort. A X-tile plot; B Patient distribution histogram; C Kaplan-Meier curves stratified by the optimal cut-off value

Table 1.

Characteristics of WT patients from SEER database and CHCMU

Variables	SEER cohort	CHCMU cohort
	(N = 1916)	(N = 237)
Sex, n (%)
Male	889 (46.40%)	113 (47.68%)
Female	1027 (53.60%)	124 (52.32%)
Race, n (%)
White	1446 (75.47%)	-
Black	338 (17.64%)	-
Others	132 (6.89%)	-
Tumor stage, n (%)
Localized	826 (43.11%)	120 (50.63%)
Regional	649 (33.87%)	84 (35.44%)
Distant	441 (23.02%)	33 (13.92%)
Laterality, n (%)
Left	989 (51.62%)	118 (49.79%)
Right	927 (48.38%)	119 (50.21%)
Type of surgery, n (%)
NSS	93 (4.85%)	25 (10.55%)
RN	1823 (95.15%)	212 (89.45%)
Histology type, n (%)
FH	-	188 (79.32%)
uFH	-	49 (20.68%)
Chemotherapy, n (%)
Yes	1771 (92.43%)	191 (80.59%)
No	145 (7.57%)	46 (19.41%)
Radiation, n (%)
Yes	971 (50.68%)	41 (17.30%)
No	945 (49.32%)	196 (82.70%)
Neoadjuvant chemotherapy
Yes	-	18(7.6%)
No	-	219(92.4%)
Metastasis, n (%)
Yes	274 (14.30%)	32 (13.50%)
No	1642 (85.70%)	205 (86.50%)
Age, M(IQR), year	3.00 (4.00)	2.00 (2.00)
Tumor volume, M(IQR), mL	-	828.58 (746.85)
Tumor size, M(IQR), cm	11.00 (5.30)	11.35 (4.80)

M median, IQR interquartile range, CHCMU Children’s Hospital of Chongqing Medical University, NSS nephron sparing surgery, RN radical nephrectomy, FH Favorable Histology, uFH Unfavorable Histology

Additionally, data from 237 pediatric WT patients were collected from our institution. The median age of this cohort was 2 years (IQR: 2 years), with 47.68% (n = 113) being male. Histologically, 79.32% (n = 188) were favorable histology (FH) and 20.68% (n = 49) were unfavorable histology (uFH). Tumor staging included 50.63% (n = 120) Localized, 35.44% (n = 84) Regional, and 13.92% (n = 33) Distant cases. The median follow-up time was 53 months (IQR: 55 months).

The impact of tumor size on prognosis

Due to significant differences in baseline characteristics among the different tumor size groups, we performed PSM to minimize baseline imbalance and potential selection bias. After matching, all baseline characteristics were well balanced, and their distributions are summarized in Table S1. The balance assessment illustrated in Figure S1 confirms a substantial improvement in covariate comparability between the groups, with all standardized mean differences (SMDs) after matching falling within acceptable ranges.

Kaplan–Meier survival analysis with the log-rank test revealed statistically significant differences in cancer-specific survival CSS and OS between the small and large tumor groups in the SEER cohort before matching (p < 0.001; Fig. 3A and B). After PSM, patients in the large tumor group still exhibited significantly better OS and CSS than those in the small tumor group (p < 0.05; Fig. 3C and D).

Fig. 3.

Kaplan-Meier survival curves stratified by tumor size before and after propensity score matching (PSM). A Overall survival (OS) before PSM; B cancer-specific survival (CSS) before PSM; C OS after PSM; D CSS after PSM

Restricted cubic spline (RCS) models revealed a significant linear association between tumor size and patient prognosis. Specifically, larger tumor size was linearly correlated with an increased risk of mortality for both OS (p for overall association = 0.011; p for nonlinearity = 0.906) and CSS (p for overall association = 0.010; p for nonlinear = 0.998; Fig. 4A and B).

Fig. 4.

Association between tumor size and survival outcomes using restricted cubic spline models. A Overall survival (OS); B Cancer-specific survival (CSS)

Univariate and multivariate Cox regression analyses were conducted to identify clinical factors associated with prognosis in patients with WT (Tables S2 and S3). Multivariate analysis demonstrated that higher tumor stage (HR = 3.76, p = 0.015), absence of chemotherapy (HR = 2.39, p = 0.034), and large tumor size (HR = 1.79, p = 0.011) were independent risk factors for poorer OS (Fig. 5A). Similarly, higher tumor stage (HR = 4.32, p = 0.017), no chemotherapy (HR = 3.01, p = 0.009), and large tumor size (HR = 1.78, p = 0.018) were identified as independent predictors for worse CSS (Fig. 5B).

Fig. 5.

Multivariable Cox regression analyses of prognostic factors for A overall survival (OS) and B cancer-specific survival (CSS) in WT patients

The association between tumor size and metastasis

Patients in the SEER cohort were divided into metastasis and no-metastasis groups. Tumor size was significantly larger in the metastasis group compared to the no-metastasis group (p < 0.001), as shown in Fig. 6A. Subsequently, the cohort was divided based on tumor size to compare metastasis rates. The metastasis rate was significantly lower in the group with small tumor than in the group with large tumor (87/762 vs. 187/880, p < 0.001), as shown in Fig. 6B. Visualization of tumor size distribution between metastasis and no-metastasis patients (Fig. 6C) revealed that tumor size in the metastasis group was predominantly distributed within the larger range, whereas the no-metastasis group exhibited a concentration of smaller tumors.

Fig. 6.

Association between tumor size and metastasis. A Comparison of tumor size between the metastasis group and no metastasis group; B Percentage of metastasis by tumor size category; C Distribution of metastasis across tumor sizes; D Association of tumor size with metastasis risk (RCS model)

The association between tumor size and metastasis risk was evaluated using RCS analysis. A statistically significant nonlinear relationship was identified (p for overall < 0.001; p for nonlinear = 0.009). The analysis revealed a threshold tumor size of 11.38 cm. Below this size, the log odds of metastasis remained relatively stable, whereas beyond it, a marked upward trend was observed, indicating a substantially increased risk of metastasis with further increases in tumor size (Fig. 6D).

The influence of tumor size on chemotherapy and radiation efficacy

The Kaplan-Meier survival curves revealed a significant difference in survival between the large tumor group and the small tumor group among patients receiving chemotherapy (p = 0.005, Fig. 7A). The 5-year OS rate of the small tumor group was as high as 96.4%, higher than that of the large tumor group (93.3%). Subsequent stratified analysis based on tumor stage indicated that tumor size did not significantly affect the prognosis of chemotherapy patients with Localized (p = 0.050) or Regional (p = 0.154) disease, as shown in Fig. 7B and C. Conversely, among those with Distant-stage disease, tumor size significantly influenced prognosis for those treated with chemotherapy (p = 0.036, Fig. 7D), with the small tumor group exhibiting a 5-year survival rate of 90.7% compared to 80.9% in the large tumor group, an absolute difference of 9.8%. In patients who underwent radiation, tumor size did not significantly impact on patient survival in the overall cohort (p > 0.05, Fig. 7E). Further stratified analysis confirmed that tumor size did not significantly influence the efficacy of radiation across any of the stages—Localized, Regional, or Distant (all p > 0.05, Fig. 7F–H).

Fig. 7.

Overall survival stratified by tumor size in patients. A All patients receiving chemotherapy; B Localized-stage patients receiving chemotherapy; C Regional-stage patients receiving chemotherapy; D Distant-stage patients receiving chemotherapy; E All patients receiving radiation; F Localized-stage patients receiving radiation; G Regional-stage patients receiving radiation; H Distant-stage patients receiving radiation

Validation of our institution data

We first analyzed the relationship between tumor size and prognosis using institutional data. Kaplan–Meier analysis indicated significantly improved survival in patients with small tumors compared to those with large tumors (p = 0.035; Fig. 8A), which is consistent with trends observed in the SEER cohort. However, RCS models did not detect a significant linear relationship between tumor size and prognosis (p for overall association = 0.289; Fig. 8B). Multivariate analysis identified large tumor size as an independent risk factor for OS (HR = 3.40, p = 0.012; Fig. 8C). External validation confirmed the association between large tumor size and adverse outcomes.

Fig. 8.

A Kaplan-Meier of overall survival (OS) curves stratified by tumor size; B Association between tumor size and OS outcomes using restricted cubic spline (RCS) models; C Multivariable Cox regression analyses of prognostic factors for OS; D Comparison of tumor size between the metastasis group and no metastasis group; E Percentage of metastasis by tumor size category; F Association of tumor size with metastasis risk using RCS model

Comparison between metastasis and non-metastasis patient groups revealed significantly larger tumors in the metastasis group (p = 0.004; Fig. 8D). Subsequent analysis showed a higher proportion of metastasis cases among patients with large tumors group (p < 0.01; Fig. 8E). RCS model further supported an association between increasing tumor size and elevated metastasis risk (p for overall association = 0.013; Fig. 8F). These findings corroborate observations from the SEER cohort that larger tumors are associated with a greater propensity for metastasis.

Due to the limited number of radiotherapy patients in our center, we only assessed the effect of tumor size on chemotherapy efficacy. Among all chemotherapy-treated patients, those with small tumors exhibited significantly better prognosis than those with large tumors (p = 0.046; Figure S2A). In the small tumor group, the 5-year OS was 90.2%, whereas in the large tumor group was 78.6%, representing a reduction in survival rate of 11.6%. Stratified analysis by tumor stage indicated consistent—though not statistically significant—survival advantages for small tumors in both localized and distant-stage patients, with 5-year OS differences of 2.9% and 13.7%, respectively. However, in regional-stage, a significant survival difference was observed: the 5-year OS was 75.1% for the small tumor group compared to 61.4% for the large tumor group, representing a 13.7% difference (Figure S2 B–D).

In addition, we analyzed the relationship between tumor size and tumor volume. The results showed a significant positive correlation (R = 0.558, p < 0.001, Figure S3A), indicating that tumor size can reflect tumor burden to a certain extent. To link tumor size with biological characteristics, we further compared the distribution of pathological subtypes (uFH vs. FH) stratified by tumor size. The analysis revealed a significant association between tumor size and pathological subtype (p < 0.001). As shown in Figure S3B, the proportion of uFH subtype is notably higher in the large tumor group compared to the small tumor group.

Discussion

The COG and SIOP have reported overall survival rates exceeding 90% for WT [11]. Due to the clinical heterogeneity of Wilms tumor, treatment strategies are stratified according to different risk groups. COG stratifies childhood renal malignancies into very low, low, standard, and high risk groups based on stage, age, histology (favorable vs. anaplastic), and LOH at 1p/16q [8]. In contrast, SIOP classifies WT into three risk groups based on histology assessed after preoperative chemotherapy: low risk (completely necrotic or cystic partially differentiated WT), intermediate risk (regressive, epithelial, stromal, mixed, or focal anaplastic types), and high risk (blastemal or diffuse anaplastic types) [12]. Notably, while both the COG and SIOP groups have incorporated measures of tumor size into their risk stratification systems, the metrics they use differ fundamentally from the focus of our study. The earlier protocols of COG suggested that patients with stage I FH WT could omit chemotherapy if the tumor weight was less than 550 g [8]. In contrast, the SIOP approach uses the tumor volume following postoperative chemotherapy—if it exceeds 500 ml—to determine the further treatment plan [9]. Our study, however, specifically evaluates maximum tumor diameter—a simple, reproducible parameter routinely obtained from diagnostic imaging. Unlike weight or volume, which are influenced by surgical technique or chemotherapy-induced changes, diameter offers a practical and standardized metric for preoperative risk assessment, particularly in real-world settings where detailed pathological or volumetric data may be unavailable.

In recent years, tumor size in WT has attracted considerable attention from numerous researchers. Han et al. reported that the probability of a unilateral renal mass being a WT increases with tumor size and may help identify cases suitable for nephron-sparing surgery [13]. Huang et al. observed a significantly elevated risk of metastasis when the tumor size > 12.5 cm [14], a finding that aligns with our results. However, our study identified a threshold of 10.4 cm. Furthermore, we demonstrated a linear correlation between increasing tumor size and metastatic risk. Provenzi et al. indicated that post-chemotherapy tumor size could serve as a prognostic indicator for WT patients [15]. Additionally, Nanteza et al. identified large tumor size as a risk factor for treatment abandonment [16]. In two prognostic models for WT, tumor size was incorporated as a predictive variable in nomograms [17, 18]. Our study further supports that tumor size aids in predicting patient outcomes and metastatic risk, and also reflects response to chemotherapy. These findings collectively suggest that tumor size has potential as a reference indicator for risk stratification and treatment planning.

As tumors enlarge, core hypoxia and necrosis develop, upregulating pro-angiogenic factors such as VEGF. However, the resulting blood vessels are structurally abnormal and leaky, facilitating tumor cell intravasation [19]. Hypoxia and metabolic stress also selectively promote tumor cell subpopulations with enhanced migratory, invasive, and survival capabilities. In neuroendocrine tumors, micro vessel density positively correlates with tumor diameter, supporting a link between tumor growth and angiogenic or metastatic potential [20]. Additionally, larger tumors are more prone to mechanical invasion, breaching anatomical barriers, and can exert systemic effects by releasing more circulating tumor cells, exosomes, and immunosuppressive factors [21].

In the study by Renfro et al. [9], tumor size was not found to be a prognostic factor for stage I or II FH WT after adjusting for other risk factors, including age, stage, histology, and molecular biomarkers such as 1p/16q LOH and 1q gain. This finding implies that in well-resourced settings with comprehensive molecular profiling, tumor size may be supplanted by more potent biological indicators. However, in routine clinical practice—particularly in low- and middle-income countries or hospitals lacking advanced molecular diagnostics—tumor size remains a practical and valuable tool for early risk stratification.

Our study has several limitations that warrant consideration. First, the SEER database lacks critical clinical and biological information, such as specific genomic alterations (e.g., 1p/16q LOH, 1q gain), proliferative indices, and intraoperative tumor spill, which precluded a more granular analysis and may limit the generalizability of our findings. Second, the absence of detailed recording regarding specific metastatic sites in many SEER cases restricted further subgroup investigations. Third, as the SEER database aggregates data across the United States while our external validation cohort originated from China, a degree of population and geographic bias is inevitable.

In this study, we found that larger tumor size in WT is associated with greater biological aggressiveness and may indicate an unfavorable prognosis. Moreover, metastatic risk increases with tumor size. Tumor size also shows potential in guiding postoperative chemotherapy strategies, which may help improve long-term survival. Therefore, further research is warranted to explore the integration of tumor size into risk stratification and treatment protocols.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

WT

Wilms tumor

COG

Children’s Oncology Group

SIOP

International Society of Paediatric Oncology

SEER

Surveillance, Epidemiology, and End Results

PSM

Propensity score matching

CSS

Cancer-specific survival

OS

Overall survival

RCS

Restricted cubic spline

Author contributions

K.C.: Conceptualization, Investigation, Writing–original draft. Z.H.: Visualization. P.H.: Data curation, Visualization. T.L.: Methodology. Z.Z.: Project administration. X.T.: Methodology. Q.S.: Conceptualization, Investigation, Writing–review and editing. G.W.: Conceptualization, Validation.

Funding

This study was supported by the National Natural Science Foundation of China (No. 82302948).

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of the Children’s Hospital of Chongqing Medical University (Date: 2024/No: 1324). Due to the retrospective nature of the study and the anonymization of data collection, the Institutional Review Board waived the requirement for parental or guardian informed consent. The SEER database analysis of this study did not require ethics approval, as they utilized publicly available de-identified data and previously published literature, respectively.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.