Clinical Risk Factors and Outcomes of Spontaneous Hepatoblastoma Rupture in Children: A 10-Year Propensity Score-Matched Study

1. Introduction

Hepatoblastoma (HB) is the most common primary malignant liver tumor in childhood and accounts for approximately 70-80% of pediatric hepatic malignancies [1]. Outcomes have improved through cisplatin-based chemotherapy, complete surgical resection, and liver transplantation in selected patients, but prognosis remains heterogeneous and is strongly influenced by pretreatment extent of disease, metastasis, vascular involvement, age, alpha-fetoprotein level, and response to therapy [2,3,4].

Tumor rupture is an uncommon but potentially catastrophic presentation. Acute hemoperitoneum may cause severe anemia, coagulopathy, hemorrhagic shock, and early death, while treatment interruption and intraperitoneal tumor spillage may adversely affect subsequent disease control. The 2017 PRETEXT revision standardized radiologic annotation of rupture, yet cooperative groups have differed in the weight assigned to rupture during risk stratification [5,6,7,8]. It therefore remains uncertain whether rupture itself drives prognosis or primarily identifies children with advanced local disease.

The evidence base is limited by heterogeneous definitions. Pondrom et al. described poor event-free survival and peritoneal progression or relapse in a multicenter cohort, whereas other single-center studies have reported more favorable outcomes [6,10]. Chang et al. identified a large primary tumor and vascular invasion as risk factors, but their rupture cohort included spontaneous, traumatic, biopsy-related, and post-chemotherapy events [9]. Data focused exclusively on spontaneous rupture at presentation remains scarce, and the clinical factors associated with spontaneous rupture remain incompletely defined. Clarifying these factors may help clinicians identify vulnerable children earlier, prioritize monitoring, and plan timely intervention.

We therefore conducted a propensity score-matched cohort study to characterize the clinical phenotype, emergency management, associated risk factors, and outcomes of spontaneous HB rupture.

2. Materials and Methods

2.1. Study Design, Setting, and Population

This retrospective single-center cohort study was conducted at West China Hospital, Sichuan University, a tertiary academic referral center in southwest China. Consecutive children with pathologically confirmed HB treated between July 2010 and July 2020 were screened. The report was prepared with reference to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) recommendations [11].

Of 106 children assessed for eligibility, 10 were excluded because clinicopathologic information or follow-up was incomplete. The complete-case cohort therefore comprised 96 children: 14 with spontaneous rupture and 82 without rupture. No statistical imputation was performed. The patient-selection process is shown in Figure 1.

2.2. Definition and Diagnosis of Spontaneous Rupture

Spontaneous rupture was defined as rupture identified at initial presentation before biopsy, chemotherapy, embolization, ablation, or surgery for the index tumor. Children whose rupture occurred after abdominal trauma or after any tumor-directed procedure were not classified as spontaneous rupture cases.

The diagnosis was based on contrast-enhanced imaging and/or intraoperative findings, supported by clinical evidence of acute intra-abdominal hemorrhage when applicable. Imaging features included focal capsular discontinuity, a subcapsular hematoma or fluid collection, and intraperitoneal blood adjacent to the tumor. Representative computed tomography findings are shown in Figure 2.

2.3. Variables and Definitions

Demographic characteristics, presenting symptoms, laboratory measurements, imaging findings, histopathologic subtype, treatment, and follow-up outcomes were extracted from medical records. Laboratory values used in between-group comparisons were obtained at initial presentation before emergency intervention or chemotherapy. Tumor extent was categorized according to PRETEXT, and histopathologic subtype was classified according to the international pediatric liver tumor consensus classification [12].

Macrovascular invasion was defined as radiologic encasement or obstruction of all three hepatic veins or the inferior vena cava, encasement or obstruction of both portal vein branches or the main portal trunk, or tumor thrombus in a major hepatic or portal venous branch [5]. Macrovascular invasion was analyzed separately from PRETEXT stage because vascular involvement may vary within the same PRETEXT category and may plausibly contribute to congestion, tumor friability, and bleeding.

2.4. Emergency and Definitive Treatment Strategy

Children with rupture underwent immediate resuscitation, including rapid intravenous access, blood-product support, correction of coagulopathy, and close hemodynamic monitoring. The method of hemorrhage control was individualized according to circulatory stability, tumor anatomy, resectability, local expertise, and response to initial supportive care. If conservative management failed to stabilize the patient’s condition, prompt emergency intervention was undertaken, including partial hepatectomy, radiofrequency ablation, transcatheter arterial embolization (TAE), or radiofrequency ablation (RFA) [13,14,15,16,17].

After stabilization, subsequent oncologic treatment was determined by a multidisciplinary team and was generally guided by the Expert Consensus for Multidisciplinary Management of Hepatoblastoma (CCCG-HB-2016), tumor extent, treatment response, and resectability [13]. Definitive treatment could include chemotherapy, delayed resection, or liver transplantation when indicated.

2.5. Follow-Up and Outcomes

Follow-up information was obtained from outpatient visits, inpatient records, and telephone interviews. Overall survival (OS) was defined as the interval from diagnosis to death from any cause or last follow-up. Event-free survival (EFS) was defined as the interval from diagnosis to relapse, progression, death, or last follow-up without an event. Complete response, relapse, and progressive disease were determined by multidisciplinary review of imaging, serum alpha-fetoprotein kinetics, operative findings, and pathology when available.

2.6. Propensity Score Matching and Statistical Analysis

Propensity scores were estimated from prespecified baseline demographic variables: age, sex, height, and weight. Nearest-neighbor matching without replacement was performed at a 1:3 ratio, using a caliper width of 0.05 times the standard deviation of the logit of the propensity score. Covariate balance was evaluated using standardized mean differences (SMDs), with values <0.10 considered acceptable.

Continuous variables are reported as median and interquartile range (IQR), except where otherwise stated, and were compared using the Mann-Whitney U test. Categorical variables were compared using the chi-square test or Fisher exact test, as appropriate. Given the limited number of rupture events, tumor location was simplified as unilobar vs bilobar disease and PRETEXT stage as I-II vs III-IV.

Variables with clinical relevance and significant between-group differences were entered into multivariable logistic regression. Because only 14 rupture events were available, a two-predictor L2 ridge-penalized logistic model was fitted with maximum tumor diameter as a continuous variable and macrovascular invasion as a binary variable, percentile-based 95% confidence intervals were estimated from 2000 bootstrap resamples. Model discrimination was quantified by the receiver operating characteristic area under the curve (AUC). Internal validation used 2000 bootstrap resamples to estimate optimism and an optimism-corrected AUC, and calibration was assessed descriptively. The sensitivity and specificity reported for the archived operating point were calculated from the combined model’s predicted probabilities. Overall survival (OS) and event-free survival (EFS) were estimated by the Kaplan-Meier method and compared with the log-rank test. Two-sided p-values <0.05 were considered statistically significant. Analyses were performed using R version 4.4.2 within RStudio and IBM SPSS Statistics version 27.0.

2.7. Ethical Considerations

The study was conducted in accordance with the Declaration of Helsinki and was approved by the Ethics Committee of West China Hospital, Sichuan University (protocol code 2023357). The requirement for informed consent was waived by the ethics committee because of the retrospective design and use of anonymized clinical data.

3. Results

3.1. Patient Selection and Matching

Among 106 children screened, 96 had complete clinicopathologic and follow-up data and were eligible for analysis. Fourteen presented with spontaneous rupture and 82 did not. After 1:3 propensity score matching, 42 controls were retained, yielding a final matched cohort of 56 children. Age, sex, height, and weight were well balanced after matching, with all SMDs below 0.10 (Table S1).

3.2. Clinical Characteristics at Presentation

Compared with matched controls, children with rupture had lower hemoglobin at presentation (80.0 versus 103.0 g/L; p = 0.032), a significantly larger maximum tumor diameter (median, 14.00 versus 10.55 cm; p < 0.001), more frequent bilobar disease (92.9% versus 45.2%; Fisher p = 0.002), more advanced PRETEXT stage (III-IV in 92.9% versus 45.2%; Fisher p = 0.002), and more frequent macrovascular invasion (71.4% versus 23.8%; p = 0.003) (Table 1). Histopathologic subtype and the other measured laboratory variables did not differ materially between groups (Table S2).

3.3. Presentation, Emergency Management, and Oncologic Course

Among the 14 children with rupture, 8 were boys and 6 were girls. The median age at presentation was 76.4 months. Thirteen children presented with sudden abdominal pain, whereas 1 initially presented with an abdominal mass and subsequently ruptured before tumor-directed treatment. All 14 had anemia and four developed hemorrhagic shock. One child had a contained subcapsular rupture and 13 had intraperitoneal rupture. Two children had metastatic disease at presentation.

One child died during the acute hemorrhagic episode before further oncologic therapy. Among the remaining 13 children, nine underwent emergency radical resection, two underwent partial hepatectomy for hemostasis, one achieved hemorrhage control with TAE, and one underwent salvage RFA after unsuccessful TAE. Eight of the 13 children who survived the acute event achieved complete response after first-line treatment, and three of these eight subsequently developed recurrence. By last follow-up, eight of the 14 children had died: one during the acute hemorrhagic episode and seven after documented progression or relapse (Table 2).

3.4. Survival

The median follow-up in the rupture cohort was 54.2 months (IQR 7.9-91.1). The rupture group had worse OS and EFS than the matched control group (Figure 3). The 1-, 3-, and 5-year OS rates were 64.8%, 54.1%, and 37.8% in the rupture group and 90.5%, 78.6%, and 66.7% in controls. Corresponding EFS rates were 62.2%, 46.6%, and 31.0% versus 85.7%, 73.8%, and 59.5%, respectively.

3.5. Exploratory Association and Model Analyses

In multivariable logistic regression, maximum tumor diameter (OR 3.078, 95% CI 1.62-5.55; P<0.001) and macrovascular invasion (OR 13.521, 95% CI 1.16-148.23; P=0.037) remain independently associated with spontaneous rupture, whereas tumor distribution and PRETEXT stage were no longer significant after adjustment (Table 3).

ROC analysis of the combined model incorporating maximum tumor diameter and macrovascular invasion yielded an area under the curve of 0.92 (Figure S1). Bootstrap internal validation demonstrated low AUC optimism, indicating robust discriminative performance of the model, while the calibration curve showed excellent agreement between predicted probabilities and observed outcomes (Figure S2). Sensitivity analysis further confirmed a significant association between vascular invasion and rupture (Tables S3 and S4).

4. Discussion

Tumor rupture may occur secondary to biopsy, trauma, chemotherapy, or may develop spontaneously without any identifiable cause. This propensity score-matched study focused specifically on spontaneous HB rupture. Rupture clustered with lower hemoglobin, a significantly larger maximum tumor diameter, bilobar involvement, advanced PRETEXT stage, and macrovascular invasion. Children with rupture also had lower OS and EFS rates and substantial acute and disease-related mortality.

These findings are broadly consistent with, but more narrowly defined than, previous reports. Chang et al. identified a primary-tumor diameter >13.4 cm and vascular invasion as independent risk factors in a cohort that included spontaneous, traumatic, biopsy-related, and post-chemotherapy rupture [9]. By contrast, our exposure definition excluded rupture after trauma or treatment, reducing etiologic heterogeneity. The present results support large tumor burden and vascular involvement as warning features.

The prognostic literature remains mixed. Pondrom et al. reported 3-year EFS and OS of 49.6% and 68.2%, respectively, with several peritoneal progression or relapse events among children with ruptured HB [6]. Zhang et al., however, reported no deaths among nine children with rupture in a small single-center cohort [10]. Differences in rupture definition, coexisting high-risk features, emergency stabilization, definitive treatment, and follow-up likely explain part of this variation. Our data add a cohort in which spontaneous rupture was accompanied by markedly lower 5-year OS and EFS rates, but they cannot isolate the causal effect of rupture from the aggressive tumor phenotype that predisposes to rupture.

The association with tumor burden is clinically and biologically plausible. Rapid tumor expansion can increase intratumoral pressure, promote ischemia and necrosis, and thin the tumor capsule. In this cohort, the marked difference in median maximum diameter is an important descriptive finding and supports tumor size as a warning feature. The imprecise per-centimeter ridge estimate should be interpreted separately: it reflects uncertainty in the adjusted slope after accounting for macrovascular invasion in a small dataset, not the absence of a between-group size difference. Macrovascular invasion may further indicate locally aggressive growth, venous obstruction, congestion, and tissue friability.

The immediate clinical priority after rupture is hemodynamic stabilization and hemorrhage control. Emergency resection may be necessary in unstable children or when interventional radiology is unavailable or unsuccessful, whereas TAE can provide rapid control in selected anatomically suitable patients [14,15,16]. RFA is not a standard treatment for ruptured HB; its single salvage use in this cohort should be interpreted as an exceptional rescue strategy rather than as evidence of comparative effectiveness [17]. Once bleeding is controlled, care should transition promptly to protocol-based chemotherapy and definitive resection or transplantation when indicated.

The findings may inform monitoring without supporting prophylactic invasive treatment. Children with very large tumors and macrovascular invasion may merit a lower threshold for monitored admission, serial abdominal examination, repeated hemoglobin and coagulation testing, advance preparation of blood products, and early multidisciplinary review involving pediatric oncology, hepatobiliary surgery, anesthesia, critical care, and interventional radiology. These measures aim to reduce delay in recognizing deterioration; they cannot be assumed to prevent the underlying biological event.

The exploratory combined model provides a more coherent interpretation of the archived ROC analysis. Maximum tumor diameter and macrovascular invasion jointly yielded an apparent AUC of 0.930 and an optimism-corrected AUC of 0.924; at the archived predicted-probability operating point, sensitivity was 92.9% and specificity was 81.0%.

This study has limitations. It was retrospective and conducted at a single tertiary referral center. Patients with incomplete records or follow-up were excluded, creating potential selection bias. Portal and hepatic venous involvement were combined because of sparse events, so vessel-specific effects could not be estimated. Treatment evolved over the 10-year period and was individualized, precluding comparison of emergency modalities. In addition, although matching improved comparability, residual confounding cannot be excluded, and the study cannot fully separate the prognostic effect of rupture itself from that of the aggressive tumor phenotype with which rupture is associated. Multicenter validation using standardized rupture definitions and prospectively specified analyses is needed.

5. Conclusions

Spontaneous hepatoblastoma rupture in children was independently associated with large tumor size and macrovascular invasion and was linked to inferior survival. Rupture should be regarded both as an emergency event and as a marker of biologically and clinically advanced diseases. Early recognition of high-risk imaging features, vigilant monitoring, and expedited multidisciplinary management may help reduce acute mortality and improve the chance of subsequent curative treatment.