Validation of 3 high-risk resectable hepatocellular carcinoma models

Introduction

Primary liver cancer is the sixth most common cancer worldwide and the third leading cause of cancer mortality, accounting for an estimated 757,048 deaths in 2022. Most of these cases (75–85%) are due to hepatocellular carcinoma (HCC) (1). The majority of these cases are in Asia and sub-Saharan Africa. However, there were about 41,000 new cases of HCC in the US. The overall 5-year survival is estimated at 21.7%, but it is higher for those patients with localized disease at about 37.3% (2). Unfortunately, only 44% of patients present with localized disease, and even those with localized disease may not be candidates for curative surgical resection or liver transplantation.

Effective locoregional therapies such as transarterial embolization, Yttrium-90 (Y-90) radioembolization, and stereotactic body radiotherapy (SBRT) have established benefit. More recently, immunotherapy has also been used to prolong survival in patients with HCC who are not surgical candidates (3-7). These therapies are now changing the paradigm of HCC treatment. There is a race to determine which of these therapies can best downstage HCC to allow for surgical intervention. However, to initiate clinical trials, we need to first determine how to optimally select candidates for downstaging protocols, before investigating whether these protocols can impact recurrence and survival.

Strategies of sequential therapy to improve outcomes have been implemented in other cancers. Specifically, there are defined resection criteria for pancreatic cancer based on tumor location within the pancreas, proximity to vessels, and ability to reconstruct vessels at the time of resection (8,9). Neoadjuvant therapy in high-risk cases may increase the chance of complete resection and long-term survival (10-13). However, it is exceedingly more difficult to define “high-risk” resectable in HCC. HCC resectability is determined not only by tumor location, size, proximity to vessels, and vascular invasion, but also by the underlying liver function. Furthermore, patient comorbidities, performance status, and individual surgeon’s experience may all contribute to the decision for resection.

To our knowledge, only three definitions of borderline resectable HCC have been proposed in the literature (Table 1) (14-16). Henceforth, we refer to these models respectively as “Technical Risk model”, “Integrated Risk model”, and “Simplified Integrated Risk model”, reflecting their predominant characteristics. Given the increasing complexity of multidisciplinary decision-making and the absence of standard criteria, it is imperative to validate a “high-risk resectable” definition. Using a large, prospectively collected dataset of patients who have undergone upfront liver resection, this study aims to assess the predictive utility of these three models for recurrence and survival. We present this article in accordance with the TRIPOD reporting checklist (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-105/rc).

Methods

We retrospectively applied the three published definitions of high-risk resectable HCC to patients who underwent upfront liver resection for HCC at Hawaii’s only liver disease and transplant center. From an overall cohort of 1,779 patients diagnosed with HCC over a 31-year period [1993–2023], 290 patients fit the inclusion criteria. This center served as a tertiary referral center for the United States Affiliated Pacific Islands, including the territories of Guam, American Samoa, the Commonwealth of the Northern Mariana Islands, Saipan, the Republic of the Marshall Islands, the Republic of Palau, the Kingdom of Tonga, and the Federated States of Micronesia. Foreign nationals from Asian countries who sought medical care in the United States were also included. The liver transplant and disease center was affiliated with two medical centers in Honolulu: Saint Francis Medical Center from 1993 to 2012, and The Queen’s Medical Center after 2012. Around 60–70% of the HCC cases in the State of Hawaii were cared for by these institutions. The University of Hawaii Institutional Review Board has approved this cohort of HCC patients (IRB CHS#18278), which included all patients referred to these centers during this timeframe. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Individual consent for this retrospective analysis was waived.

HCC was diagnosed histologically by surgery, percutaneous biopsy, or imaging criteria. In the first decade, patients without histologic confirmation of HCC were diagnosed based on the following criteria: a history of chronic liver disease, a mass of at least 2 cm seen on two imaging studies [computed tomography (CT) scan, ultrasound, or magnetic resonance imaging (MRI)], and one of the following: (I) vascular blush evident on CT scan or MRI; (II) alpha-fetoprotein (AFP) >200 ng/mL; or (III) arteriogram confirming the tumor (17). Since 2011, the diagnosis of HCC was determined via contrast-enhanced study (triple phase CT or MRI) revealed LI-RADS 5 (Liver Imaging Reporting and Data System) lesions (18).

Data collected

Demographics, medical history, laboratory data, tumor characteristics, treatment, and survival were collected. Demographic data specifically included age, biologic sex, birthplace, and the patient’s ethnicity characterized by “Asian”, “White”, “Hispanic or Latino”, “Native Hawaiian/Other Pacific Islander” (NHOPI) or “Other”.

Medical comorbidities included hyperlipidemia, diabetes mellitus, prior malignancy, and smoking. Risk factors for HCC were also recorded including viral hepatitis B or C, significant alcohol use (more than two alcoholic beverages daily for ≥10 years), and other chronic liver diseases. We collected information on body mass index (BMI). Obesity was defined as BMI 30 kg/m² or higher. metabolic dysfunction-associated steatotic liver disease (MASLD)/metabolic-associated steatohepatitis (MASH) was defined as a potential risk factor if patients were diagnosed with comorbid diabetes, hyperlipidemia, or obesity, without other risk factors for HCC including viral hepatitis B or C, significant alcohol use, or other chronic liver disease.

The collected laboratory data included albumin, bilirubin, creatinine, prothrombin time with international normalized ratio (INR), aspartate aminotransferase (AST), alanine aminotransferase (ALT), AFP level, and platelet count. Hepatitis B and C serologies were obtained if otherwise unavailable. For hepatitis B, we also noted patients who had only hepatitis B core antibody positivity in the absence of hepatitis B surface antigen. We categorized AFP as normal if the level was 15 ng/mL or lower.

Patients were categorized as cirrhotic based on biopsy or imaging findings. We identified the size of the largest tumor, whether the tumor was unilobular or multinodular, and whether there was macrovascular invasion or extrahepatic metastasis. We calculated Model for End Stage Liver Disease (MELD) score, Childs-Turcotte-Pugh (CTP) score with classification into Childs A (CTP 5–6), B (CTP 7–8), or C (CTP 9 or higher). We also calculated AST to platelet ratio index (APRI) and neutrophil-to-lymphocyte ratio (NLR).

Patient selection

We compared the 3 models described in the literature to determine patients who would be considered as high-risk resectable. Of the cohort of 1,779 patients, we selected 290 patients who underwent upfront liver resection. Patients who had preoperative therapy before liver resection were excluded from the analysis. We also excluded those patients with ruptured HCC who had arterial embolization to control bleeding prior to liver resection. High-risk resectable patients were determined based on the criteria for each model as listed in Table 1. Although not specifically stated in these criteria, we excluded all patients who had extrahepatic spread or hepatic vein/inferior vena cava (IVC) invasion.

Models

Table 1 summarizes the models included in this study. These models represent existing definitions in the literature, incorporating combinations of tumor size, number, macrovascular invasion, serum AFP, and technical considerations for assessing future liver remnant adequacy. The objective was not to redefine technical resectability, but to determine if these existing high-risk criteria could effectively stratify patients at higher biological risk for poor outcomes, such as recurrence and mortality.

Two of the models were derived from prospective high-risk resectable HCC downstaging studies, while the third model was based on a retrospective cohort and externally validated.

The Technical Risk model classified HCC as high-risk resectable based on tumor size, number, and vascular involvement (14). Specifically, it included patients with a single tumor greater than 10 cm, 2 to 3 tumors with any greater than 3 cm, or 4 or more tumors that did not exceed half of the liver. Additionally, it included patients with portal vein tumor thrombus of grade ≤ Cheng’s classification I/II and no extrahepatic metastasis. This model was developed from a prospective phase Ib/II clinical trial conducted in China, where patients were treated with a combination of transarterial chemoembolization (TACE), lenvatinib, and camrelizumab.

The Integrated Risk model defined high-risk resectable HCC based on several key tumor features, including a solitary tumor greater than 5 cm, unilobar multifocal disease with 3 or more tumors or one tumor greater than 3 cm, or technically resectable bilobar disease with an adequate future liver remnant (15). This model also included high-risk features such as macrovascular invasion or an AFP ≥400 ng/mL. This model was derived from a phase I clinical trial conducted in the United States, where patients received a combination of cabozantinib and nivolumab prior to resection.

The Simplified Integrated Risk model, on the other hand, categorized HCC as high-risk resectable based on the presence of AFP ≥400 ng/mL, a tumor larger than 5 cm, ≥3 tumors, or macrovascular invasion (16). It was developed from a retrospective cohort study in Japan and validated with an external cohort. This model is similar to the Integrated risk Model in that it includes both technical and prognostic risk parameters, but the requirements are less granular. Unlike the other models, this third model relied on surgery alone to identify prognostic predictors of worse outcomes.

Each of these models provided a unique framework for determining high-risk resectability in HCC, with varying degrees of sensitivity and specificity, and their predictive accuracy was assessed in this study.

Statistical analysis

We determined the recurrence rate, 1-, and 3-year survival for the patients who underwent liver resection for each of the models. Recurrence was defined as radiographic evidence of local tumor recurrence, with time to recurrence defined from the date of surgery to the date last follow-up or recurrence. Overall survival (OS) was defined from the date of surgery to the date of last follow-up or death. We calculated the sensitivity and specificity for each model in predicting recurrence, 1-, and 3-year survival. We then calculated the C-statistic, or area under the receiver operating characteristic curve (AUROC), to measure each model’s predictive accuracy. AUROC values were calculated with 95% confidence intervals using bootstrap resampling (1,000 iterations). All analyses were performed using complete case analysis; no imputation was performed for missing data.

We then performed binary logistic regression for each model to predict recurrence, 1-, and 3-year survival. The binary logistic regression models were adjusted for the following covariates: age, sex, ethnicity, BMI, hepatitis B, hepatitis C, MASH, diabetes, hyperlipidemia, and hypertension.

Results

Of the overall cohort of 1,779 patients with HCC, 290 underwent an upfront resection. Further clinicodemographic characteristics of the patients who underwent a liver resection are described in Table 2. The mean age was 64 years (SD 11), with 85 patients (29%) aged 70 years or older. There were more males 204 (70%) – than females. The racial/ethnic distribution was as follows: Asian 203 (70%), White 39 (13%), NHOPI 31 (11%), Hispanic 2 (0.7%), and Other 10 (3.4%). Risk factors for HCC included hepatitis B surface antigen positivity 89 (31.0%), hepatitis B core antibody positive only 37 (13%), hepatitis C antibody positive 87 (30%), significant alcohol use 86 (30%), and MASH/MASLD 48 (16.3%). Other comorbidities included diabetes 94 (32%), hyperlipidemia 97 (34%), hypertension 165 (62%), smoking 160 (56%), and obesity (BMI ≥30 kg/m²) 47 (17.0%).

Of the 290 patients who underwent upfront liver resection without preoperative therapies or ruptured HCC, 109 experienced recurrences. Patients were classified as high-risk resectable according to each model’s criteria. Although all models significantly predicted recurrence and survival, the Integrated Risk model demonstrated the highest sensitivity for predicting recurrence (63.3%), 1-year (79.6%), and 3-year (70.1%) survival. The Technical Risk and Simplified Integrated Risk models had lower sensitivity but higher specificity, indicating that they were more precise in identifying patients likely to experience recurrence or reduced survival, albeit identifying fewer patients overall.

Table 2 describes the clinical and demographic characteristics of patients that underwent resection and met the criteria for “high-risk resectable”. Of the patients who underwent resection, 60, 148, and 40 patients were deemed “high-risk resectable”, as defined by the Technical Risk, Integrated Risk, and Simplified Integrated Risk models, respectively. Roughly 95% of high-risk resectable patients had CTP A disease irrespective of the model. Roughly 35% and 40% of patients who fit the high-risk resectable definition in the Technical Risk and Integrated Risk models had a normal AFP <15 ng/mL, while no high-risk resectable patients in the Simplified Integrated Risk model had an AFP <15 ng/mL. About 53%, 83%, and 87.5% of high-risk resectable patients had a single tumor, and 3.4%, 3.4%, and 12.5% of the high-risk resectable patients had macrovascular invasion in the Technical Risk, Integrated Risk, and Simplified Integrated Risk models, respectively.

Table 3 presents a detailed breakdown of recurrence, 1-, and 3-year OS by model. Notably, the rates of recurrence were similar across all three models, with 52%, 47%, and 53% of patients experiencing recurrence in the Technical Risk, Integrated Risk, and Simplified Integrated Risk models, respectively. With respect to 1-year OS, high-risk resectable patients had a 1-year OS of 63%, 73%, and 54% and a 3-year OS of 46%, 56%, and 29% in the Technical Risk, Integrated Risk, and Simplified Integrated Risk models, respectively. All three models were predictive of local recurrence, 1-year OS, and 3-year OS on both univariable and multivariable analysis when adjusting for ethnicity, hepatitis B, hepatitis C, BMI, diabetes mellitus, hyperlipidemia, and hypertension. We further examined the patterns of recurrence with respect to early to late recurrences (Table S1). Early and late recurrence was defined as recurrence ≤2 years from resection or greater than 2 years, respectively. Over 25% of patients experienced early recurrence when disease was felt to be high-risk by each model, respectively. Of note, early recurrence was more prevalent than late recurrence.

When examining AUROC, the Technical Risk, Integrated Risk, and Simplified Integrated Risk models had an AUROC >0.6 for recurrence and >0.75 for 1-year OS (Table 4, Figure 1). For local recurrence, the Technical Risk, Integrated Risk, and Simplified Integrated Risk models had sensitivities of 29%, 63.3%, and 19%, respectively, and specificities of 84.5%, 56.4%, and 89.5%, respectively. For 1-year OS, the Technical Risk, Integrated Risk, and Simplified Integrated Risk models had sensitivities of 44.9%, 79.6%, and 36.7%, and specificities of 84.5%, 55.9%, and 91.2%, respectively. For 3-year OS, the Technical Risk, Integrated Risk, and Simplified Risk models had sensitivities of 34.5%, 70.1%, and 31.0%, and specificities of 85.2%, 56.3%, and 93.8%, respectively.

Figure 1 AUROC comparing Models 1, 2, & 3 with respect to (A) local recurrence, (B) 1-year overall survival, and (C) 3-year overall survival. Model 1 represents the “Technical Risk model” (14). Model 2 represents the “Integrated Risk model” (15). Model 3 represents the “Simplified Integrated Risk model” (16). AUROC, area under the receiver operating characteristic curve; ROC, receiver operating characteristic.

Discussion

Comparative analysis of models

The results of this study validate the utility of the Technical Risk model, Integrated Risk model, and Simplified Integrated Risk model, in predicting both local recurrence and OS for patients with high-risk resectable HCC. Each model demonstrated acceptable to excellent discrimination as measured by AUROC. However, the Integrated Risk Model emerged as the most sensitive in predicting recurrence, 1-year survival, and 3-year survival, with sensitivities of 63.3%, 79.6%, and 70.1%, respectively.

This higher sensitivity is crucial for identifying patients at risk for poor outcomes, allowing clinicians to consider escalated treatment strategies, which may improve outcomes. On the other hand, the Technical Risk model and Simplified Integrated Risk model showed higher specificity, suggesting that they were more accurate in identifying patients with recurrence or poor survival outcomes; the higher specificity may identify patients where upfront surgery may not be a preferred option.

Distinguishing technical from prognostic resectability

Technical resectability primarily involves anatomical considerations such as tumor size, number, and proximity to critical vasculature. Biological or prognostic resectability encompasses factors that indicate aggressive tumor behavior, such as elevated AFP, multifocality, and macrovascular invasion. Clearly distinguishing these two aspects is vital for effective patient stratification. The Technical Risk model emphasizes primarily anatomical complexities, while the Integrated Risk model comprehensively incorporates both anatomical and biological criteria. The Simplified Integrated Risk model captures these elements with fewer criteria, providing a streamlined yet prognostically effective approach. Although all patients in our study were technically resectable, the varying predictive capabilities of these models underscore the distinction between technical resectability and biological resectability.

Clinical implications and utility

The Integrated Risk Model utilizes clinically relevant factors including tumor size, multifocal disease, macrovascular invasion, and AFP. This comprehensive approach aligns with the evolving understanding of HCC biology and the multifactorial nature of resectability. The Integrated Risk model provides a robust framework for identifying patients with more aggressive tumor biology where neoadjuvant therapies prior to surgical resection may prove beneficial. While there is a tradeoff between escalating therapy and overtreatment when considering a more inclusive eligibility, the high rate of local recurrence (>40%) for patients that met criteria for “high-risk resectable” suggests that escalation is worthy of consideration; recurrence thresholds as low as 15–30% are often used in other cancer types for consideration of locoregional therapy (19,20). While repeat resection may be a potential option for patients with a local recurrence with a delayed recurrence, patients with an early recurrence are less likely to be candidates for re-resection. Roughly 75% of recurrences occurred within 2 years of resection for patients classified as “high risk” by any of the 3 models, and recurrence was associated with worse prognosis (Table S1) further supporting the role of escalated therapy.

The Integrated Risk model’s inclusivity is particularly valuable in multidisciplinary treatment planning. Its ability to identify a broader cohort of patients offers a practical tool for tumor boards, enhancing collaborative decision-making among surgeons, oncologists, interventional radiologists, and radiation oncologists. This aligns with established practices in other cancers, such as pancreatic adenocarcinoma, where clearly defined high-risk resectable criteria have improved outcomes through tailored neoadjuvant strategies (9,11). For example, patients meeting the high-risk criteria of the Integrated Risk model may be prioritized for either locally directed therapies, such as transarterial radioembolization (TARE) or SBRT, prior to resection, or for systemic therapies aimed at reducing tumor burden and enhancing resectability (15,21-25). Recent data suggest that multimodal strategies combining systemic or locoregional therapies with surgery can improve outcomes in biologically aggressive, resectable HCC (21,26). Identification of patients at high risk of recurrence or poor outcome is essential to correctly identify patients for future protocol enrollment.

Importance in the context of liver transplantation

Although liver transplantation remains the gold standard treatment for patients with cirrhosis and early-stage HCC, its applicability is limited by organ availability, stringent transplant eligibility criteria, and patient comorbidities (17). The shortage of donor livers necessitates careful patient selection to optimize outcomes, prioritizing those who stand to benefit most from transplantation (18). Consequently, surgical resection remains critical as a therapeutic option for many patients who either do not meet transplantation criteria or for whom transplant waiting times are prohibitive (27). Thus, refining definitions for high-risk resectable HCC becomes increasingly important to effectively manage and stratify patients who could benefit from surgical resection coupled with adjunctive therapies, ultimately preserving transplantation for patients who are most suitable for it.

Racial/ethnic and etiologic considerations

Each of these models was developed in a different country with differences in ethnicity and HCC etiologies (e.g., hepatitis B, hepatitis C, and MASLD). This may potentially impact the performance of each model if it were applied to datasets with differing clinicodemographic features. For example, models based on a population with predominantly hepatitis B will likely include a larger proportion of younger patients and those with non-cirrhotic hepatitis B. Conversely, cohorts with more alcohol-related liver disease may include patients with worse liver function, fewer HCC detected with surveillance, and more advanced disease (28). Our cohort had a similar composition for hepatitis B (24% vs. 20%), hepatitis C (38.7% vs. 29%), and patients classified as high-risk resectable (14% vs. 19%) when compared to the cohort in the study by Haruki et al. (16). This further supports the validity of our results. Additionally, the composition of our cohort appeared to be similar to that of the study by Ho et al., although the sample size in that study was more limited (15).

AFP has been shown to have lower sensitivity in detecting tumors associated with MASH/MASLD, as these patients may not exhibit elevated AFP levels despite having significant liver pathology. Given the rising incidence of MASH/MASLD-related HCC, declining rates of hepatitis B over time, and lower sensitivity of AFP in detecting MASH-related HCC, this may eventually favor a model that is less reliant on AFP (29). Our cohort is predominantly composed of patients from Hawaii and the United States Affiliated Pacific Islands, and they included a broad mix of risk factors: Hepatitis B surface Ag (24.3%), hepatitis B core Ab positive only (11.0%), hepatitis C Ab positive (38.7%), significant alcohol use (42.9%), MASH/MASLD (16.8%), and diabetes (38%). The diverse etiologies and ethnic backgrounds in our cohort support the broad applicability of these models.

Strengths and limitations

This study is limited in that it is retrospective and based on a diverse but primarily Asian/Pacific Islander population, which may limit its generalizability. This cohort spans three decades, and survival rates may have changed over time due to advancements in surgical management and more effective strategies for post-surgical recurrences. While the evidence supporting adjuvant strategies may be evolving, it was nearly absent during much of the study period. Therefore, while survival outcomes may have varied over time, the risk of recurrence is less affected by temporal changes due to the limited availability of adjuvant therapies during the earlier years of the cohort.

Despite these limitations, this analysis was based on a large prospectively collected database with highly granular data and thorough follow-up. Future efforts should focus on validating the Integrated Risk model in multi-institutional and prospective cohorts to confirm its predictive accuracy and clinical utility. The incorporation of advanced imaging techniques, such as radiomics, and emerging biomarkers, such as circulating tumor DNA, may further enhance patient stratification (30,31). Additionally, incorporating the Integrated Risk model into clinical trials evaluating neoadjuvant therapies, such as TARE, SBRT, immunotherapy, or combination therapies, may provide a more uniform approach to optimizing treatment strategies (15).

The Integrated Risk model also has the potential to address disparities in HCC care. By providing standardized criteria for identifying high-risk resectable patients, this may reduce variability in care delivery and improve access to advanced therapies, particularly in underserved populations (32).

Conclusions

Our results validate previously published definitions of high-risk resectable HCC. While all 3 models had reasonable predictive capabilities for clinical outcomes, the Integrated Risk model represents the most clinically applicable definition of high-risk resectable HCC due to its superior sensitivity, inclusivity, and balanced predictive performance. Using the Integrated Risk model to define high-risk resectable HCC may be a promising strategy to identify patients for escalated sequential therapy strategies.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-105/rc

Data Sharing Statement: Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-105/dss

Peer Review File: Available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-105/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tgh.amegroups.com/article/view/10.21037/tgh-25-105/coif). J.D.A. served on an advisory board for GlaxoSmithKline and reports receiving material support (drugs, equipment, or services) from the company. L.L.W. reports honoraria for speaking engagements with Astra Zeneca. The other authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the University of Hawaii Institutional Review Board (IRB CHS#18278) and individual consent for this retrospective analysis was waived.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J Clin 2024;74:229-63. [Crossref] [PubMed]
2. National Cancer Institute, Surveillance, Epidemiology and End Results Program, Cancer Stat Facts; Liver and Intrahepatic Duct. Cancer Stat Facts: Liver and Intrahepatic Bile Duct Cancer. Available online: https://seer.cancer.gov/statfacts/html/livibd.html
3. Salem R, Johnson GE, Kim E, et al. Yttrium-90 Radioembolization for the Treatment of Solitary, Unresectable HCC: The LEGACY Study. Hepatology 2021;74:2342-52. [Crossref] [PubMed]
4. Bush DA, Volk M, Smith JC, et al. Proton beam radiotherapy versus transarterial chemoembolization for hepatocellular carcinoma: Results of a randomized clinical trial. Cancer 2023;129:3554-63. [Crossref] [PubMed]
5. Abou-Alfa GK, Lau G, Kudo M, et al. Tremelimumab plus Durvalumab in Unresectable Hepatocellular Carcinoma. NEJM Evid 2022;1:EVIDoa2100070.
6. Acoba JD, Rho Y, Fukaya E. Phase II study of cobolimab in combination with dostarlimab for the treatment of advanced hepatocellular carcinoma. J Clin Oncol 2023;41:580.
7. Finn RS, Qin S, Ikeda M, et al. Atezolizumab plus Bevacizumab in Unresectable Hepatocellular Carcinoma. N Engl J Med 2020;382:1894-905. [Crossref] [PubMed]
8. National Comprehensive Cancer Network. Pancreatic Adenocarcinoma (Version 1.2025). Available online: https://www.nccn.org/professionals/physician_gls/pdf/pancreatic.pdf
9. Callery MP, Chang KJ, Fishman EK, et al. Pretreatment assessment of resectable and borderline resectable pancreatic cancer: expert consensus statement. Ann Surg Oncol 2009;16:1727-33. [Crossref] [PubMed]
10. Smaglo BG. Role for Neoadjuvant Systemic Therapy for Potentially Resectable Pancreatic Cancer. Cancers (Basel) 2023;15:2377. [Crossref] [PubMed]
11. Versteijne E, van Dam JL, Suker M, et al. Neoadjuvant Chemoradiotherapy Versus Upfront Surgery for Resectable and Borderline Resectable Pancreatic Cancer: Long-Term Results of the Dutch Randomized PREOPANC Trial. J Clin Oncol 2022;40:1220-30. [Crossref] [PubMed]
12. Bahadoer RR, Dijkstra EA, van Etten B, et al. Short-course radiotherapy followed by chemotherapy before total mesorectal excision (TME) versus preoperative chemoradiotherapy, TME, and optional adjuvant chemotherapy in locally advanced rectal cancer (RAPIDO): a randomised, open-label, phase 3 trial. Lancet Oncol 2021;22:29-42. [Crossref] [PubMed]
13. Conroy T, Bosset JF, Etienne PL, et al. Neoadjuvant chemotherapy with FOLFIRINOX and preoperative chemoradiotherapy for patients with locally advanced rectal cancer (UNICANCER-PRODIGE 23): a multicentre, randomised, open-label, phase 3 trial. Lancet Oncol 2021;22:702-15. [Crossref] [PubMed]
14. Liao M, Yang Y, Jiang H, et al. 966P Borderline resectable hepatocellular carcinoma: Definitions, efficacy and safety results from a prospective phase Ib/II study evaluating camrelizumab plus lenvatinib combined with TACE as preoperative therapy (BRHCC). Annals of Oncology 2024;35:S664-S5.
15. Ho WJ, Zhu Q, Durham J, et al. Neoadjuvant Cabozantinib and Nivolumab Converts Locally Advanced HCC into Resectable Disease with Enhanced Antitumor Immunity. Nat Cancer 2021;2:891-903. [Crossref] [PubMed]
16. Haruki K, Harimoto N, Furukawa K, et al. Proposal for Prognosis-Oriented Definition of Borderline Resectable Hepatocellular Carcinoma. J Am Coll Surg 2024;238:1137-47. [Crossref] [PubMed]
17. Mazzaferro V, Regalia E, Doci R, et al. Liver transplantation for the treatment of small hepatocellular carcinomas in patients with cirrhosis. N Engl J Med 1996;334:693-9. [Crossref] [PubMed]
18. Lucey MR, Furuya KN, Foley DP. Liver Transplantation. N Engl J Med 2023;389:1888-900. [Crossref] [PubMed]
19. Cedermark B, Dahlberg M, et al. Improved survival with preoperative radiotherapy in resectable rectal cancer. N Engl J Med 1997;336:980-7. [Crossref] [PubMed]
20. Clarke M, Collins R, Darby S, et al. Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005;366:2087-106. [Crossref] [PubMed]
21. Sangro B, Kudo M, Erinjeri JP, et al. Durvalumab with or without bevacizumab with transarterial chemoembolisation in hepatocellular carcinoma (EMERALD-1): a multiregional, randomised, double-blind, placebo-controlled, phase 3 study. Lancet 2025;405:216-32. [Crossref] [PubMed]
22. Chiang CL, Chan KSK, Chiu KWH, et al. Complete Response to Locoregional Therapy Plus Immunotherapy for Hepatocellular Carcinoma. JAMA Oncol 2024;10:1548-53. [Crossref] [PubMed]
23. Li Z, Liu J, Zhang B, et al. Neoadjuvant tislelizumab plus stereotactic body radiotherapy and adjuvant tislelizumab in early-stage resectable hepatocellular carcinoma: the Notable-HCC phase 1b trial. Nat Commun 2024;15:3260. [Crossref] [PubMed]
24. Braat MN, Samim M, van den Bosch MA, et al. The role of (90)Y-radioembolization in downstaging primary and secondary hepatic malignancies: a systematic review. Clin Transl Imaging 2016;4:283-95. [Crossref] [PubMed]
25. Llovet J, Finn RS, Ren Z, et al. LBA3 Transarterial chemoembolization (TACE) with or without lenvatinib (len) + pembrolizumab (pembro) for intermediate-stage hepatocellular carcinoma (HCC): Phase III LEAP-012 study. Annals of Oncology 2024;35:S1229.
26. Wang Z, Fan J, Zhou S, et al. Perioperative camrelizumab plus rivoceranib versus surgery alone in patients with resectable hepatocellular carcinoma at intermediate or high risk of recurrence (CARES-009): a randomised phase 2/3 trial. Lancet 2025;406:2089-99. [Crossref] [PubMed]
27. Singal AG, Llovet JM, Yarchoan M, et al. AASLD Practice Guidance on prevention, diagnosis, and treatment of hepatocellular carcinoma. Hepatology 2023;78:1922-65. [Crossref] [PubMed]
28. Zeng RW, Ong CEY, Ong EYH, et al. Global Prevalence, Clinical Characteristics, Surveillance, Treatment Allocation, and Outcomes of Alcohol-Associated Hepatocellular Carcinoma. Clin Gastroenterol Hepatol 2024;22:2394-2402.e15. [Crossref] [PubMed]
29. Wong LL, Kim CJ, Kwee SA, et al. Alpha-fetoprotein testing for hepatocellular carcinoma may not be helpful in nonalcoholic steatohepatitis. Open Journal of Gastroenterology 2013;3:49-54.
30. Korpics MC, Polley MY, Bhave SR, et al. A Validated T Cell Radiomics Score Is Associated With Clinical Outcomes Following Multisite SBRT and Pembrolizumab. Int J Radiat Oncol Biol Phys 2020;108:189-95. [Crossref] [PubMed]
31. Ignatiadis M, Sledge GW, Jeffrey SS. Liquid biopsy enters the clinic - implementation issues and future challenges. Nat Rev Clin Oncol 2021;18:297-312. [Crossref] [PubMed]
32. Singal AG, Lampertico P, Nahon P. Epidemiology and surveillance for hepatocellular carcinoma: New trends. J Hepatol 2020;72:250-61. [Crossref] [PubMed]