Malignant Hepatocellular Neoplasm, Not Otherwise Specified, Displays Poorer Chemoresponsiveness and Postoperative Prognosis Than Hepatoblastoma.

Introduction
Primary liver malignancies account for 1% to 2% of all pediatric solid tumors, with most being malignant hepatocellular neoplasm [1]. Malignant hepatocellular neoplasm has been classically classified into hepatoblastoma (HB; 70%-80%) and pediatric hepatocellular carcinoma (p-HCC; 20%-30%) [2]. HB primarily occurs in children aged between 6 months and 5 years without an underlying liver disease [3]. It is generally chemo-responsive and successfully treated with a combination of chemotherapy and surgical resection [4]. Pathology HB specimens consist of epithelial or both epithelial and mesenchymal components with a mixture of different histologic patterns [5]. In contrast, p-HCC affects patients aged 10-14 years and is frequently associated with chronic liver diseases, such as biliary atresia and hepatitis B [3]. It is often chemo-resistant and associated with poor prognosis [6]. p-HCC can exhibit conventional histology similar to adult HCC, fibrolamellar histology, or clear-cell histology [7].
However, a small group of patients exist who cannot be clearly classified into HB or p-HCC. The term of malignant hepatocellular neoplasm, not otherwise specified (HCN-NOS) was recently introduced in the World Health Organization classification of tumors 5th edition as a provisional entity for such patients [8]. It was initially reported by Prokurat et al. [9] in 2002 and is also known as transitional liver cell tumor, HB with HCC-like progenies, HB with carcinoma features, between HB, and HB demonstrating focal atypia or pleomorphism [10-12]. It is treated as a high-risk HB in the current Paediatric Hepatic International Tumor Trial [13,14]. Clinical outcomes of HCN-NOS are debated, with some studies reporting poorer outcomes than HB [9,11,12,15], whereas others reported similar or better outcomes [16,17].
As one of the largest cohort studies of rare pediatric liver cancer conducted at a single pediatric cancer center, we reviewed and reclassified surgically resected pediatric hepatocellular neoplasms into HB, HCN-NOS, and p-HCC using our specimen data. We re-evaluated the clinicopathologic characteristics and treatment outcomes of HCN-NOS, and compared them with HB and p-HCC.

Materials and Methods

1. Patients and samples
Records of 191 patients, who were < 19 years of age and underwent surgery at Asan Medical Center for HB, p-HCC, HCN-NOS, or hepatocellular adenoma between 1990 and June 2023, were retrieved from the institutional database. Thirty-one patients who did not receive surgical resection and three patients whose specimen slides were unavailable for review were excluded (S1 Table). For the remaining 157 patients, histopathologic features and clinical information were retrospectively reviewed. After a comprehensive review of medical records, specimen gross images, and hematoxylin and eosin (H&E)–stained slides, five patients with benign hepatocellular adenoma, three patients with no viable tumor cells after preoperative therapy, and one patient with fibrolamellar HCC were additionally excluded. Finally, 148 hepatocellular malignancies were included in the study (Fig. 1).

2. Patient treatment
The treatment plan for HB followed the routine protocol at our center [18]. The initial plan was established based on the resectability at the time of admission. Surgical resection was performed for patients whose resectability was confirmed by computed tomography or magnetic resonance image. For patients deemed unresectable, neoadjuvant chemotherapy proceeded following tissue diagnosis via percutaneous needle biopsy or incisional biopsy. Subsequent surgical resection or liver transplantation (LT) was performed based on resectability.
The treatment of p-HCC followed the same protocols as that for adult HCC. Treatment strategies aligned with the Barcelona Clinic Liver Cancer guidelines [19] and included surgical resection, local therapy followed by resection, systemic chemotherapy, and LT based on risk stratification. Local therapies, such as transarterial chemoembolization, were used in cases of non-resectable tumors, with systemic therapies and LT considered for more advanced or unresectable cases.

3. Histological evaluations for classification of HB, HCN-NOS, and p-HCC
All available H&E-stained and immunohistochemistry (IHC) slides prepared during diagnosis were reviewed for classification into either HB, p-HCC, or HCN-NOS following the consensus of three board-certified pathologists. HB was diagnosed in tumors that exhibited histologic components described in the International Pediatric Liver Tumors Consensus Classification, particularly mesenchymal component [5], whereas p-HCC was diagnosed in malignant tumors that were exclusively composed of neoplastic cells demonstrating hepatocellular differentiation and exhibiting loss of normal architecture, high cellularity, high nuclear-cytoplasmic ratio, and nuclear pleomorphism [20]. HCN-NOS was defined as tumors displaying biphasic or intermediate histologic features between those of HB and p-HCC [9,17]. In cases where the decisions of the three pathologists did not align, the diagnosis agreed upon by two of the pathologists was adopted. After histologic classification, a pediatric oncologist and a pediatric surgeon reviewed the results with clinical data and agreed with all the classifications.

4. IHC and molecular analysis
IHC staining for β-catenin was conducted in 115 patients with available formalin-fixed paraffin-embedded tumor tissue blocks. Any nuclear expression in the tumor cell was counted as positive.
During the diagnosis and treatment courses, targeted next-generation sequencing (NGS) was conducted in 14 patients using the OncoPanel AMC v4.5, as previously described [21-23]. Briefly, the OncoPanel AMC v4.5 is a DNA-based hybrid-capture gene panel targeting 343 genes. The gene list and detailed information about this panel are provided in S2-S4 Tables. In addition, Sanger sequencing for TERT promoter mutation was conducted in nine patients with available tumor tissue material [24].

5. Statistical analysis
The association between different clinicopathological variables was determined by the Wilcoxon rank-sum, Kruskal-Wallis test, chi-square test, and Fisher’s exact test, as appropriate. For the survival analysis, the Cox proportional hazards regression model and backward stepwise selection with Akaike information criterion values were used as appropriate. A p-value < 0.05 was considered statistically significant. All statistical analyses were performed using R ver. 4.2.3 (R Foundation for Statistical Computing). Methods are further detailed in the Supplementary Materials.

Results

1. Histologic features of HCN-NOS
Examples of histologic images of HCN-NOS from three patients are demonstrated in Fig. 2. Patient 1 was an 8-year-old girl with no background liver disease. The resected tumor consisted of malignant hepatocellular cells exhibiting intermediate morphology between that of HB and HCC and focally expressed nuclear β-catenin (Fig. 2A-D). Patient 2 was a 10-year-old girl with no background liver disease. The resected tumor exhibited both HCC-like areas and HB-like areas along with nuclear β-catenin expression (Fig. 2E-H). Patient 3 was a 9-year-old boy with no background liver disease. The resected tumor consisted of diverse morphological patterns with extremely atypical primitive neoplastic cells, which did not fit into HB or HCC (Fig. 2I-L).

2. Clinicopathologic characteristics of patients with HB, HCN-NOS, and p-HCC
The clinicopathologic characteristics of each group are summarized in Table 1. A total of 148 patients were profiled as 109 (73.6%) with HB, 21 (14.2%) with HCN-NOS, and 18 (12.2%) with p-HCC. The Sankey diagram of the original diagnosis and reclassification is demonstrated in S5 Fig. The median age of patients with HCN-NOS was 8.6 years (range, 1.9 to 17.5 years), higher than HB (median, 1.2 years; range, 0.0 to 11.7; p < 0.001) but similar to p-HCC (median, 7.9 years; range, 0.1 to 17.8 years). Compared with p-HCC, patients with HCN-NOS less frequently had underlying liver diseases (p < 0.001) and higher serum α-fetoprotein (AFP) concentration at diagnosis (p < 0.001). Among the patients with p-HCC included, 33% had hepatocellular adenoma as the background, whereas none had HB or HCN-NOS background adenoma. Tumor size at diagnosis was significantly larger in patients with HCN-NOS than in patients with p-HCC (p=0.003). Furthermore, the PRETEXT stage was higher in HCN-NOS than in HB and p-HCC (p=0.033). Notably, a higher proportion of patients with HCN-NOS presented with metastasis at diagnosis than HB and p-HCC (p < 0.001). Additionally, 19 patients (90.5%) with HCN-NOS were classified under intermediate to high-risk categories in CHIS-HS risk classification, significantly higher than that in HB and p-HCC (p < 0.001). Lymphovascular invasion (LVI) was more frequently observed in patients with HCN-NOS compared to those with HB and p-HCC (p < 0.001).

3. Molecular profiles of HB, HCN-NOS, and p-HCC
The results of molecular studies, including IHC, NGS, and Sanger sequencing, are summarized in Table 2. The expression of nuclear β-catenin was detected in 80.2% (73/91) of HB, 50.0% (9/18) of HCN-NOS, 100% (4/4) of p-HCC in β-catenin-activated hepatocellular adenoma (β-HCA), and 0% (0/2) of conventional p-HCCs. Mutations of the Wnt/β-catenin signalling pathway-associated genes, such as APC, AXIN1, and CTNNB1 were detected in 75.0% (6/8) of HB and 50.0% (2/4) of HCN-NOS. Six HB harboured the CTNNB1 mutation, one HCN-NOS harboured both APC and CTNNB1 mutations and the other HCN-NOS harboured the AXIN1 mutation. The TERT promoter mutation was detected in 20.0% (1/5) of HCN-NOS and 33.3% (1/3) of p-HCC in β-HCA; however, it was absent in all eight available HBs and one p-HCC.
Arm-level copy-number alterations and tumor mutational burden were evaluated in eight HBs, four HCN-NOSs, and two p-HCCs in β-HCA. HB frequently exhibited gain of 1q (37.5%), 2 (37.5%), and 20 (50.0%), whereas HCN-NOS frequently demonstrated gain of 1q (75.0%), 4p (50.0%), 6p (50.0%), and 8q (50.0%) and loss of 1p (75.0%) and 4q (75.0%). The median tumor mutational burden was 14.85 mutations/Mb in HCN-NOS, 9.4 mutations/Mb in HB, and 7 mutations/Mb in p-HCC in β-HCA. A statistical comparison could not be performed due to the limited number of available cases. No other clinically significant genetic alterations were detected beyond those described above. The complete results of NGS and TERT promoter sequencing are demonstrated in S6 Table.

4. Response to preoperative chemotherapy in patients with HB and HCN-NOS
The response to neoadjuvant chemotherapy was compared between HB and HCN-NOS (Table 3). We excluded p-HCC from this analysis because only one patient with p-HCC in our study received neoadjuvant chemotherapy. The number of patients requiring more than four cycles of neoadjuvant chemotherapy, indicating a poorer response and the need for extended treatment, was significantly higher in the HCN-NOS group than in the HB group (p=0.043). The majority (88.7%) of patients with HB achieved a ≥ 90% decrease in the serum AFP level after neoadjuvant chemotherapy, whereas 58.8% of patients with HCN-NOS achieved that level (p=0.004). The tumor size was evaluated by calculating the maximum tumor diameter; we found that 81.4% of the patients with HB achieved a ≥ 30% decrease, whereas 58.8% of patients with HCN-NOS patients did (p=0.044). Portal vein invasion status and metastasis remained unchanged in most HBs and HCN-NOSs. The difference between POSTTEXT and PRETEXT stages was not significantly different between the two groups. In the surgical specimens, 70.6% of HCN-NOS exhibited viable tumor cells in ≥ 50% of the entire tumor area, whereas 50% of HB did, although statistically not significant (p=0.205).

5. Prognostic significance of HCN-NOS
Event-free survival (EFS) and overall survival (OS) in the entire cohort of resected pediatric hepatocellular malignancies were compared based on different factors using the Kaplan-Meier survival analysis (Fig. 3) and univariable Cox proportional hazard regression model (S7 Table). Age ≥ 8 years, advanced PRETEXT stage, the presence of portal vein invasion, metastasis at diagnosis, presence of LVI, and diagnosis of HCN-NOS were significantly associated with poorer prognoses. Regarding the treatment method, patients who received LT reported the longest EFS and OS, followed by those receiving delayed resection and upfront resection; however, the p-values were insignificant in the entire cohort. Sex, the presence of underlying liver disease, and serum AFP level at diagnosis did not affect EFS or OS.
The subsequent multivariable regression analysis (Fig. 4) revealed that the diagnosis of HCN-NOS and p-HCC, PRETEXT stage IV, and the presence of LVI were significant negative predictors for EFS. For OS, the presence of underlying liver disease, presence of portal vein invasion, and LVI were significant negative predictors, whereas the serum AFP level > 1,000 ng/mL was a good predictor for OS.
Finally, a survival analysis of treatment methods was conducted in HB and HCN-NOS groups (Fig. 5). LT displayed the longest EFS and OS in both HB and HCN-NOS groups. However, delayed resection after neoadjuvant chemotherapy exhibited longer EFS and OS than upfront resection in only the HB group but not in the HCN-NOS group, although all p-values were insignificant probably due to the insufficient number of patients.

Discussion
We identified 21 patients with HCN-NOS, a recently emerging disease entity, by comprehensively reviewing a consecutive cohort of patients with pediatric hepatocellular neoplasms. To date, to the best of our knowledge, this is the largest cohort study to investigate the clinicopathologic characteristics of HCN-NOS. Our study emphasizes the importance of a multidisciplinary approach to diagnosing HCN-NOS and provides new insights into its progression and distinct pathological features.
Before patient classification, we excluded 31 non-resected patients due to insufficient availability for histologic review (S1 Table). These patients were generally old with a median age of 7.5 years and presented with advanced disease. Approximately half of these cases had underlying liver disease, most frequently hepatitis B virus (HBV). These findings indicate that this group likely represents advanced p-HCC or, in fewer cases, advanced HCN-NOS. Most patients died within a few months after diagnosis due to disease progression, suggesting extremely poor prognosis in advanced, non-resected p-HCC and HCN-NOS.
Reclassification of surgical cases showed that approximately 14% were classified into HCN-NOS, similar to the 10.4% reported by Zhou et al. [25] but higher than the 3.7% by Nagae et al. [11] and 1.2% by Cho et al. [26] The histologic, demographic, and molecular characteristics of each group generally align with previous reports [14]. However, careful interpretation of demographic data for our p-HCC group is necessary because this includes early-stage and resectable HCCs only. In particular, the incidence of HBV-related HCC has steeply decreased in Korea since the 1983 HBV vaccination program [27]. Some recent p-HCCs were unexpectedly diagnosed from resected hepatocellular adenomas or explanted cirrhotic livers, explaining their lower stage, less frequent preoperative treatment, higher LT rates, and better outcomes. Meanwhile, the proportion of HCN-NOS in older pediatric patients has increased accordingly, suggesting the need for better recognition and strategies for diagnosis and treatment.
HCN-NOS is primarily a histological diagnosis and defined as lesions with either a histology intermediate between that of HB and HCC or mixed [8]. However, distinguishing among HB, HCN-NOS, and HCC can be challenging. For instance, it is difficult to differentiate HBs with predominantly macrotrabecular or pleomorphic patterns from HCN-NOS. Similarly, p-HCC, without an underlying liver disease and HCN-NOS, intricately resemble each other. An analysis of molecular features revealed that HCN-NOS shared CTNNB1 mutations with HB. However, it exhibited certain features similar to HCC, including higher mutation rates, chromosomal instability, and TERT promoter mutation, whereas HB was genetically simple [10-12,15,17,25,28,29]. In practice, pathologists may consider using β-catenin IHC, reported as positive in approximately 80% of HB and HCN-NOS but positive in 20%-30% of p-HCC, and TERT promoter mutation testing, reported as never positive in HB but positive in approximately 40% of HCN-NOS and HCC. Although HCN-NOS was initially considered fatal [9], recent studies have reported that LT might improve its prognosis to levels similar to HB, although these studies included only dozens of cases [12,17].
We conducted a comprehensive review of a consecutive cohort of 148 cases by a team including three pathologists, a pediatric hemato-oncologist, and a pediatric surgeon, to ensure diagnostic accuracy. A comprehensive statistical analysis identified HCN-NOS as an independent risk factor for shorter EFS in pediatric hepatocellular malignancy. In addition, HCN-NOS was associated with significantly higher rates of metastasis at diagnosis and the presence of LVI in resected specimens. Notably, LVI was an independent poor prognostic factor of both EFS and OS in pediatric hepatocellular malignancies. Therefore, we suggest that HCN-NOS harbours a different interaction with surrounding vascular structures than HB or p-HCC, which may contribute to its bad prognosis. The mechanisms of LVI and metastasis require further investigation.
Patients with HCN-NOS required longer cycles of neoadjuvant chemotherapy but showed poorer chemoresponsiveness than those with HB, which can be another reason for its bad prognosis. This poor chemoresponsiveness may be, at least partially, explained by the aforementioned HCC-like molecular features [10,12]. The current treatment of HCN-NOS is the same as that for high-risk HB, with cisplatin-based chemotherapy [13,14]. However, our study showed poorer chemoresponsiveness and postoperative prognosis in HCN-NOS than HB, which could be improved by LT. The optimal timing and choice of surgery, whether tumor resection or LT, should be carefully discussed in a multidisciplinary setting for patients with HCN-NOS. In addition, atezolizumab and bevacizumab have significantly improved the outcome of adult patients with HCC [30]. These new treatments may be effective in HCN-NOS as well as p-HCC, especially when considering high mutation rates of HCN-NOS. Therefore, a proper diagnosis of HCN-NOS is necessary not only to predict patients’ prognosis more accurately but also to establish the best treatment strategy.
The strengths of this study are as follows: As a tertiary center with extensive clinical experience with HB in Korea, we could collect a relatively large sample size for a single-center study. The multidisciplinary approach and involvement of experienced clinicians in HB allowed us to comprehensively diagnose HCN-NOS. However, the study had certain limitations due to its retrospective design and single-institution nature, such as the lack of robust molecular analysis data and its reliance on surgical specimens. In addition, data from patients lost to follow-up or who died without surgery were excluded from the analysis, which could have influenced the findings. Despite these limitations, this study revealed that HCN-NOS exhibits distinct manifestations than those of HB and p-HCC.
The possibility of HCN-NOS should be considered in pediatric patients with liver cancer, especially those > 5 years of age with distant metastasis at diagnosis and no background liver disease. Although patients with HCN-NOS exhibit histologic features intermediate of those between HB and p-HCC, they show poorer response to chemotherapy, more frequent LVI, and shorter EFS than those with HB and p-HCC. Therefore, recognition and proper diagnosis of this disease are highly crucial in the management of pediatric patients with hepatic cancer. Particularly, LT should be strongly considered in patients with HCN-NOS.