Validation of Risk Models for Predicting Post-SVR HCC in Real-World Surveillance Across Global Geographic Regions.

Introduction

The use of oral direct‐acting antiviral (DAA) regimens as anti‐hepatitis C virus (HCV) therapy has improved antiviral efficacy and tolerability. Consequently, the proportion of patients who have achieved sustained virologic response (SVR) has markedly increased [1]. Although SVR significantly decreases HCC risk [2], studies have reported persistent HCC risk exceeding 1% per year in patients with cirrhosis [3], underscoring the importance of continued HCC surveillance [4]. However, there is variation in HCC risk among patients with cirrhosis, as well as HCC developing in some patients without cirrhosis. In this context, assessing HCC risk after SVR is important for effective HCC surveillance implementation.

Several clinical risk models have been proposed to stratify HCC risk after SVR, with each showing promising performance when locally validated; however, it is unclear if performance would vary regionally and if these models would work in real‐world clinical practice of the respective region. In addition to differences in demographics and clinical characteristics between cohorts, there is a variation in post‐SVR surveillance practices [5,6,7], including candidate selection, surveillance intensity, and adherence. For example, criteria for identifying post‐SVR patients who warrant HCC surveillance differ across guidelines, particularly the required degree of liver fibrosis [5,6,7]. American Association for the Study of Liver Diseases (AASLD) guidelines and the European Association for the Study of the Liver (EASL) guidelines recommend post‐SVR HCC surveillance for only patients with cirrhosis [5,6]. In contrast, Asian Pacific Association for the Study of the Liver (APASL) guidelines recommend surveillance for all SVR patients regardless of fibrosis [7]. Consequently, tumour burden and treatment receipt in post‐SVR patients who develop HCC widely vary by region/country [8]. These regional differences could influence the performance of risk stratification models, so it is unclear if one prediction model can be applied universally or if region‐specific models are needed. In addition to variation by region, the applicability of these models to real‐world practice for management of post‐SVR patients would also be unclear.

Herein, we leveraged an international collaboration combining several real‐world cohorts to evaluate and compare the discrimination of HCC risk stratification models for predicting HCC in patients with hepatitis C who achieved SVR.

Patients and Methods

Study Patients

We enrolled patients who were confirmed as having achieved SVR by DAA therapy (i.e., absence of serum HCV RNA at least 12 weeks after end of DAA therapy) from 30 high‐volume academic centers in Europe (Italy, Germany, Austria, and Switzerland), North America (USA), South America (Argentine and Brazil), Middle East (Egypt), South and Southeast Asia (India and Thailand), or East Asia (Japan and Taiwan). Baseline demographic and laboratory data prior to the start of DAA therapy and outcomes including the development of HCC were collected. In principle, patients were followed up after SVR according to the guidelines by academic societies of respective region [5,6,7,9,10,11,12]. Patients with a previous history of HCC, patients with presence of HCC or suspicious liver nodules on imaging prior to DAA initiation were excluded. Cirrhosis was determined by non‐invasive markers of fibrosis (e.g., transient elastography or Fibrosure), biopsy or imaging showing a cirrhotic‐appearing liver and signs of portal hypertension (e.g., gastroesophageal varices, collateral veins on imaging, or splenomegaly).

The study protocol was approved by the institutional review board of Ogaki Municipal Hospital and by the IRB of each participating institution, with a waiver of written informed consent. All procedures adhered to the Declaration of Helsinki.

Post‐SVRSurveillance and Diagnosis ofHCC

Study patients were surveyed for HCC after achievement of SVR. In principle, the surveillance was according to the guidelines of the regions of the cohort. Recall policies for missed surveillance intervals were institutionally determined. The diagnosis of HCC was based on guidelines of academic societies for each participating country, typically consisting of histological confirmation or characteristic radiological imaging features (i.e., arterial phase hyperenhancement and delayed phase washout).

Calculations ofHCCRisk Models and Categorisations of Risk Groups

We assessed four reported models for stratifying the risk of HCC: aMAP risk score [13], FIB‐4 index [14], GES score [15], and Toronto HCC risk index (THRI) [16]. All patients were categorised into three risk groups, low‐risk, intermediate‐risk, and high‐risk groups using the predefined threshold for each model (TableS1). These HCC risk stratification models were selected among the many reported HCC risk models [13,14,15,16,17,18,19,20,21,22,23,24,25] given their components are routinely available in clinical practice regardless of regions and the same numbers of risk categories (three categories) across models.

Statistical Analysis

Continuous variables are expressed as medians with interquartile ranges (IQRs) and categorical variables are expressed as numbers and percentages. The discriminative ability of each model was first assessed visually by plotting the Kaplan–Meier estimate of the incidence of HCC for each group, and the results were then compared with the log‐rank test. For HCC incidence, time zero was defined as the date of SVR‐12. Patients who did not develop HCC were censored at the last follow‐up. Discriminative ability was assessed using the concordance index, specifically the standard Harrell's c‐index, for which higher values indicate better discrimination. A c‐index of 0.50 indicates no discrimination, whereas a c‐index of 1.0 indicates perfect discrimination. In addition to discrimination, model calibration was assessed using Brier scores and calibration plots to evaluate agreement between predicted and observed HCC risk. Clinical utility was examined using decision curve analysis (DCA), estimating net benefit across a range of clinically relevant threshold probabilities compared with ‘treat‐all’ and ‘treat‐none’ strategies. These analyses were performed overall and by geographic regions.

Data analysis was performed using JMP statistical software, version 11.0.0 (Macintosh version; SAS Institute, Cary, NC) and EZR (Saitama Medical Center, Jichi Medical University, Saitama, Japan), the latter for which is a graphical user interface for R (The R Foundation for Statistical Computing, Vienna, Austria).

Results

Characteristics of Study Patients by Regions

Table1summarises the background characteristics of the study patients, stratified by region. There were regional differences in patient age, gender, HCV genotype, and proportions of patients with cirrhosis. The median age of patients at SVR was highest in East Asia (70 years) and lowest in South and Southeast Asia (52 years), between 55 and 60 years in other regions. Whereas male gender was predominant in Europe, North America, and South and Southeast Asian, female was predominant in South America and East Asia. The predominant HCV genotype was genotype 1 in Europe, North and South America, and East Asia, genotype 3 in South and Southeast Asia, and genotype 4 in the Middle East. The percentages of patients with cirrhosis in the cohort were between 50% and 65% in all regions except for South America (33.8%) and South and Southeast Asia (88.6%).

Table: Characteristics of patients with post‐SVR de novo HCC by regions.

HCCIncidence AfterSVR, Stratified by Region

Patients were followed up as medians of 21.9–42.7 months. FigureS1demonstrates the cumulative incidence of HCC after SVR, stratified by region. The highest incidence of HCC was observed in South and Southeast Asia over the first 3 years; however, the cumulative incidences were similar at 5 years across cohorts, outside of lower incidence in the European cohort. We also observed variations in the maximum diameter of HCC at diagnosis across regions, from 1.8 cm in East Asia to 4.0 cm in South and Southeast Asia, and in the percentage of early‐stage HCC (BCLC stage 0 or A), from 95.8% in East Asia to 52.5% in Europe. The one‐, three‐, and five‐year incidence were 2.8%, 6.7%, and 9.3% in Europe, 2.3%, 10.0%, and 13.2% in North America, 0.9%, 2.8%, and 16.8% in South America, 0.3%, 3.8%, and 16.2% in the Middle East, 6.6%, 16.5%, and 16.5% in South and Southeast Asia, and 1.9%, 7.0%, and 12.1% in East Asia.

Performance ofHCCRisk Stratification Models AfterSVRby Regions

Table2showed the distribution of patients for each HCC risk stratification model in each cohort, outside of FIB‐4 index in the South American cohort which could not be calculated due to lack of required data. There were large variations in patient distributions by each risk model. For example, aMAP categorised less than 20% of each cohort as low risk, whereas GES categorised over 50% of most cohorts as low risk, outside of North America (37.3%) and South and Southeast Asia (34.3%). Similarly, aMAP and FIB‐4 both categorised over one‐third of each cohort as high risk, whereas THRI classified less than 15% as high risk.

Table: Distributions of patients according to respective risk stratification model by regions.

When focusing on patients in whom all four risk scores were available (all three risk scores in the South American cohort), the distribution patterns of the respective region did not differ markedly, although there was a trend of increases in prevalences of patients categorised to higher risk groups (TableS2).

Table3showed the number and the incidence of HCC in low, intermediate, and high‐risk groups by respective risk model in each cohort. There were trends of increasing HCC incidences from low‐risk to high‐risk groups for each respective model, with some exceptions. FiguresS2–S7show the cumulative incidence of post‐SVR HCC in each cohort, stratified by clinical risk model categories. Whereas HCC incidence of the intermediate risk group for aMAP score was similar to that of the high‐risk group in the North American and South American cohorts (FiguresS3andS4), it was similar to the low‐risk group in the Middle East cohort and South and Southeast Asia cohort (FiguresS5andS6). HCC incidence was not associated with risk groups by GES score in the North American and South American cohorts (FiguresS3andS4). Notably, only one patient (South and Southeast Asia cohort) developed post‐SVR HCC among all patients of the low‐risk group of the aMAP model.

Table: Numbers and incidences of post‐SVR HCC according to regions and risk stratification models.

Table4describes the Harrel's c‐indices of each model across cohorts, which ranged between 0.5 and 0.8; however, discrimination was below 0.70 for most models across most regions. When comparing models within each respective region, all 4 models showed comparable c‐index values but the model with the highest c‐index, usually suggesting the best performance, differed by regions. The models with the highest c‐index for post‐SVR HCC were aMAP score in the European and Middle East cohorts, GES score in the South American, South and Southeast Asian, and East Asian cohorts, and THRI in the North American cohort. However, the differences between HCC risk models were not statistically significant, except for a few select comparisons: aMAP score vs. THRI risk index in Middle East cohort, and aMAP score vs. GES score, FIB‐4 index vs. GES score, and GES score vs. THRI risk index in East Asia cohort (TableS3).

Table: Harrel's C index of respective risk stratification model by regions.

Sensitivity Analyses

Because patients who are recommended for undergoing post‐SVR surveillance by most guidelines are those with cirrhosis, we conducted a sensitivity analysis focusing on patients who had cirrhosis at SVR. Although the models with the highest c‐index changed in some cohorts, they remained comparable within each cohort and ranged between 0.5 and 0.8 and below 0.70 for most models across most regions (TableS4). When excluding patients in whom post‐SVR HCC was detected and diagnosed at advanced stage (BCLC 2–4), for whom effective surveillance had actually failed, the results were similar, but c‐indices slightly increased in comparison to those in the entire study patients (TableS5).

Calibration and Overall Prediction Accuracy

Overall prediction accuracy varied across models and geographic regions as assessed by Brier scores, with lower values indicating better combined calibration and discrimination. The model with the lowest Brier score varied across regions (TableS6). Calibration plots showed generally acceptable agreement between predicted and observed HCC risk across models, although the degree of calibration varied by region and risk range (FigureS8).

Clinical Utility Assessed by Decision Curve Analysis

Decision curve analysis demonstrated that all four HCC risk stratification models provided greater net clinical benefit than ‘treat‐all’ or ‘treat‐none’ strategies across clinically relevant threshold probabilities. At lower threshold probabilities, corresponding to more aggressive surveillance strategies, net benefit was similar across models. At higher thresholds, the GES score showed the greatest net benefit in the overall cohort. In contrast, region‐specific analyses showed heterogeneity in clinical utility of the respective model (FigureS9).

Discussion

In this study, we validated 4 proposed HCC risk stratification models with routinely available components in real‐world practice from six different regions worldwide. The results showed large variations in the performances of these models and the difficulty of using HCC risk models in real clinical settings. The distribution of patients categorised as low, intermediate, or high risk by each model varied across geographic regions. Although the discrimination of the clinical risk scores was comparable in each respective cohort, the model with the highest c‐index varied by region, with no single model that universally demonstrated the best performance. Whereas the low‐risk group by aMAP score showed low HCC incidence regardless of cohorts, no model was able to consistently identify high‐risk patients across all cohorts. These data highlight that clinicians and researchers should carefully consider which model, if any, would be most appropriate for their individual cohort when assessing HCC risk after SVR.

These differences in model performance may partly be related to variation in the characteristics of post‐SVR patients by regions, including patient age, HCV genotype, and percentage of cirrhosis [26]. In association with the variation of the characteristics of SVR patients, HCC cumulative incidence after SVR varied, with higher incidence rates noted in South/Southeast Asia shortly after SVR. However, cumulative incidence rates became similar across cohorts with longer follow‐up. This may have been partly due to the higher percentage of patients with cirrhosis in the cohort but also due to the potential failure to detect minute HCC nodules prior to the commencement of DAA therapy.

Regional variations are also likely related to differences in surveillance practices for post‐SVR patients. Surveillance guidelines after SVR differ; AASLD guidelines recommending surveillance using semi‐annual ultrasound plus AFP and restricting only to those with cirrhosis at SVR [6], whereas EASL guidelines recommend post‐SVR surveillance using ultrasound alone for patients with cirrhosis at SVR [5], and APASL guidelines recommend performing surveillance for all SVR patients using ultrasound and multiple tumour markers (des‐gamma‐carboxy prothrombin and AFP‐L3 fraction in addition to AFP) [7]. Similarly, recall policies for those who missed visits or the threshold for considering diagnostic evaluation also likely differ by region. Thus, differences in surveillance intensity and adherence may partly explain variation in model performance and tumour stage at diagnosis. Indeed, in association with the differences in surveillance practice, tumour burden or post‐SVR HCC varies by regions [8]. However, there remained variations in c‐index of respective models between regions even focusing on patients with cirrhosis or excluding patients with post‐SVR advanced‐stage HCC at diagnosis, that is, HCC that were missed for early detection. This indicated that the regional variations in risk‐stratification model performances were not only due to the regional differences in surveillance practice.

Beyond geographic variation, it was noteworthy that model discrimination was suboptimal, with the c‐statistic below 0.70 for each model across most regions in real‐world settings. This reflects the inherent complexity of post‐SVR HCC risk prediction and the limitations of models based solely on routine clinical variables. Each predictive model had been developed on the basis of regional patient populations. For example, the GES score has been developed based on Egyptian patients with SVR, whose HCV genotype was uniquely 4, which may limit the transportability of respective models across diverse populations. In addition, the FIB‐4 index originally reflects liver fibrosis, although liver fibrosis is one of the strongest factors associated with the development of HCC. Similarly, calibration showed no model was perfectly able to identify high and low‐risk patients. These data indicated the difficulty of applying the reported models in real‐world practice with the same performances as those in the original reports. All cohorts were based on real‐world clinical settings, and all patients may not have always been surveyed strictly according to the surveillance guidelines of the respective region or adhered to surveillance recommendations. Therefore, the suboptimal risk stratification performance with regional variations of the respective model was not always due to differences in the surveillance guidelines of post‐SVR patients of regions directly. The results highlight the need for improved risk stratification prior to implementation in clinical practice. The performance of included models may be low given methodological issues in model derivation [27], as well as each model only included readily available clinical variables, and more nuanced variables such as liver stiffness or blood‐based biomarkers may be of assistance [28,29].

The extended performance analyses highlighted important regional differences in calibration, clinical utility, and reclassification performance of post‐SVR HCC risk models. Decision curve analysis demonstrated that all models provided meaningful clinical benefit compared with uniform surveillance strategies; however, the magnitude of benefit of using models differed by region. These findings further suggested that while risk‐stratified surveillance can improve efficiency in many regions, a uniform approach to post‐SVR risk prediction is unlikely to be optimal globally, and region‐adapted or population‐specific models may be required.

It is noteworthy that only one patient developed post‐SVR HCC in the low‐risk group of aMAP score; no patients developed HCC in all cohorts except for the cohort of South/Southeast Asia. This model may delineate a subgroup with negligible HCC risk and may help identify patients at very low risk, warranting further prospective evaluation for potential surveillance optimisation.

There were several limitations to this study. We validated only 4 risk‐stratification models for predicting post‐SVR HCC because the components for calculating scores of these models were feasible with routinely available laboratory tests and because the number of risk categories was the same. Other variables such as liver stiffness by transient elastography, the data of which were not available from all cohorts, may have improved model performance and could have been useful universally across regions. Second, the number of patients in the cohorts varied largely, from 757 patients in the South/Southeast Asian cohort to 3026 patients in the East Asian cohort, which may have influenced the performances of risk stratification models. Third, the definition of cirrhosis was not consistent across cohorts. Fourth, the competing risk of death during the study period was not considered in the analysis of HCC incidence due to the lack of information about death. Finally, there may have been a small variation in the way of data collection by cohorts due to the retrospective nature of data collection despite the fact that all cohorts were being followed up prospectively. However, all participating institutions of this study were high‐volume academic centers, and the influence of the variations in surveillance practice or data collection was minimal.

In conclusion, our results underscore the difficulty of applying risk stratification models for surveillance of post‐SVR HCC in real‐world clinical practice. The performance of risk stratification models to predict HCC development was suboptimal, and the ‘optimal model’ differed by region. Future studies should focus on improving the diagnostic performance of risk stratification models, as well as generalisability to broad international populations.

Author Contributions

Concept and study design: Hidenori Toyoda and Amit G. Singal. Data acquisition: All authors. Data analysis: Hidenori Toyoda, Amit G. Singal., and Philip J. Johnson. Preparation of the manuscript: Hidenori Toyoda and Amit G. Singal. Data interpretation, review and/or revision of the manuscript: All authors.

Funding

The authors have nothing to report. Dr. Singal's research is supported by NIH U01 CA283935, P50 CA295495, and U01 CA271887.

Ethics Statement

The study protocol was approved by the institutional review board of Ogaki Municipal Hospital and by IRB of each participating institution, with waiver of written informed consent. All procedures adhered to the Declaration of Helsinki.

Conflicts of Interest

The authors declare no conflicts of interest.

References

1. A. G. Singal, F. Kanwal, andJ. M. Llovet, “Global Trends in Hepatocellular Carcinoma Epidemiology: Implications for Screening, Prevention, and Therapy, ”Nature Reviews. Clinical Oncology20(2023): 864–884. doi.org/10.1038/s41571-023-00825-3
2. F. Kanwal, J. Kramer, S. M. Asch, M. Chayanupatkul, Y. Cao, andH. B. El‐Serag, “Risk of Hepatocellular Carcinoma in HCV Patients Treated With Direct‐Acting Antiviral Agents, ”Gastroenterology153(2017): 996–1005. doi.org/10.1053/j.gastro.2017.06.012
3. S. Aleman, N. Rahbin, O. Weiland, et al. , “A Risk for Hepatocellular Carcinoma Persists Long‐Term After Sustained Virologic Response in Patients With Hepatitis C‐Associated Liver Cirrhosis, ”Clinical Infectious Diseases57(2013): 230–236. doi.org/10.1093/cid/cit234
4. G. Ioannou, L. A. Beste, P. K. Green, et al. , “Increased Risk for Hepatocellular Carcinoma Persists up to 10 Years After HCV Eradication in Patients With Baseline Cirrhosis or High FIB‐4 Scores, ”Gastroenterology157(2019): 1264–1278. doi.org/10.1053/j.gastro.2019.07.033
5. European Association for the Study of the Liver, “EASL Clinical Practice Guidelines on the Management of Hepatocellular Carcinoma, ”Journal of Hepatology82(2025): 315–374. doi.org/10.1016/j.jhep.2024.08.028
6. A. G. Singal, J. M. Llovet, M. Yarchoan, et al. , “AASLD Practice Guidance on Prevention, Diagnosis, and Treatment of Hepatocellular Carcinoma, ”Hepatology78(2023): 1922–1965. doi.org/10.1097/HEP.0000000000000466
7. T. Kanda, G. K. K. Lau, L. Wei, et al. , “APASL HCV Guidelines of Virus‐Eradicated Patients by DAA on How to Monitor HCC Occurrence and HBV Reactivation, ”Hepatology International13(2019): 649–661. doi.org/10.1007/s12072-019-09988-7
8. H. Toyoda, M. Kanneganti, J. Melendez‐Torres, et al. , “Regional Differences in Clinical Presentation and Prognosis of Patients With Post‐Sustained Virologic Response Hepatocellular Carcinoma, ”Clinical Gastroenterology and Hepatology22(2024): 72–80. doi.org/10.1016/j.cgh.2023.06.026
9. F. Piñero, M. Tanno, G. Aballay Soteras, et al. , “Argentinian Clinical Practice Guideline for Surveillance, Diagnosis, Staging and Treatment of Hepatocellular Carcinoma, ”Annals of Hepatology19(2020): 546–565. doi.org/10.1016/j.aohep.2020.06.003
10. A. Omar, A. Kaseb, T. Elbaz, et al. , “Egyptian Society of Liver Cancer Recommendation Guidelines for the Management of Hepatocellular Carcinoma, ”Journal of Hepatocellular Carcinoma10(2023): 1547–1571. doi.org/10.2147/JHC.S404424
11. K. Hasegawa, N. Takemura, T. Yamashita, et al. , “Clinical Practice Guidelines for Hepatocellular Carcinoma: The Japan Society of Hepatology 2021 Version (5th JSH‐HCC Guidelines), ”Hepatology Research53(2023): 383–390. doi.org/10.1111/hepr.13892
12. Surveillance group, Diagnosis group, Staging group, et al. , “Management Consensus Guideline for Hepatocellular Carcinoma: 2016 Updated by the Taiwan Liver Cancer Association and the Gastroenterological Society of Taiwan, ”Journal of the Formosan Medical Association117(2018): 381–403. doi.org/10.1016/j.jfma.2017.09.007
13. R. Fan, G. Papatheodoridis, J. Sun, et al. , “aMAP Risk Score Predicts Hepatocellular Carcinoma Development in Chronic Hepatitis Patients: An International Cohort Collaboration, ”Journal of Hepatology73(2020): 1368–1378. doi.org/10.1016/j.jhep.2020.07.025
14. R. K. Sterling, E. Lissen, N. Clumeck, et al. , “Development of a Simple Noninvasive Index to Predict Significant Fibrosis in Patients With HIV/HCV Coinfection, ”Hepatology43(2006): 1317–1325. doi.org/10.1002/hep.21178
15. G. Shiha, I. Waked, R. Soliman, et al. , “GES: A Validated Simple Score to Predict the Risk of HCC in Patients With HCV‐GT4‐Associated Advanced Liver Fibrosis After Oral Antivirals, ”Liver International40(2020): 2828–2833. doi.org/10.1111/liv.14666
16. S. A. Sharma, M. Kowgier, B. E. Hansen, et al. , “Toronto HCC Risk Index: A Validated Scoring System to Predict 10‐Year Risk of HCC in Patients With Cirrhosis, ”Journal of Hepatology68(2018): 92–99. doi.org/10.1016/j.jhep.2017.07.033
17. K. Abe, H. Wakabayashi, H. Nakayama, et al. , “Factors Associated With Hepatocellular Carcinoma Occurrence After HCV Eradication in Patients Without Cirrhosis or With Compensated Cirrhosis, ”PLoS One15(2020): e0243473. doi.org/10.1371/journal.pone.0243473
18. S. Alonso López, M. L. Manzano, F. Gea, et al. , “A Model Based on Non‐Invasive Markers Predicts Very Low Hepatocellular Carcinoma Risk After Viral Response in Hepatitis C Virus‐Advanced Fibrosis, ”Hepatology72(2020): 1924–1934. doi.org/10.1002/hep.31588
19. H. S. Chun, B. K. Kim, J. Y. Park, et al. , “Design and Validation of Risk Prediction Model for Hepatocellular Carcinoma Development After Sustained Virological Response in Patients With Chronic Hepatitis C, ”European Journal of Gastroenterology & Hepatology32(2020): 378–385. doi.org/10.1097/MEG.0000000000001512
20. A. Hiraoka, T. Kumada, C. Ogawa, et al. , “Proposed a Simple Score for Recommendation of Scheduled Ultrasonography Surveillance for Hepatocellular Carcinoma After Direct Acting Antivirals: Multicenter Analysis, ”Journal of Gastroenterology and Hepatology34(2019): 436–441. doi.org/10.1111/jgh.14378
21. C. C. Hu, C. H. Weng, M. C. Hua, et al. , “New Scoring Method to Predict Risk of Hepatocellular Carcinoma in Patients With Chronic Hepatitis C After Pegylated Interferon and Ribavirin Therapy, ”Journal of Interferon & Cytokine Research40(2020): 82–91. doi.org/10.1089/jir.2019.0103
22. Y. Tahata, R. Sakamori, R. Yamada, et al. , “Prediction Model for Hepatocellular Carcinoma Occurrence in Patients With Hepatitis C in the Era of Direct‐Acting Anti‐Virals, ”Alimentary Pharmacology & Therapeutics54(2021): 1340–1349. doi.org/10.1111/apt.16632
23. N. Tamaki, M. Kurosaki, Y. Yasui, et al. , “Hepatocellular Carcinoma Risk Assessment for Patients With Advanced Fibrosis After Eradication of Hepatitis C Virus, ”Hepatology Communications6(2022): 461–472. doi.org/10.1002/hep4.1833
24. J. Tani, A. Morishita, T. Sakamoto, et al. , “Simple Scoring System for Prediction of Hepatocellular Carcinoma Occurrence After Hepatitis C Virus Eradication by Directacting Antiviral Treatment: All Kagawa Liver Disease Group Study, ”Oncology Letters19(2020): 2205–2212. doi.org/10.3892/ol.2020.11341
25. T. Watanabe, Y. Tokumoto, K. Joko, et al. , “Predictors of Hepatocellular Carcinoma Occurrence After Direct‐Acting Antiviral Therapy in Patients With Hepatitis C Virus Infection, ”Hepatology Research49(2019): 136–146. doi.org/10.1111/hepr.13278
26. F. Kanwal, Z. Feng, Y. Hoshida, et al. , “Development and validation of risk stratification models for hepatocellular carcinoma: A framework from the Translational Liver Cancer Consortium, ”Hepatology(2026), . doi.org/10.1097/HEP.0000000000001745
27. H. InnesandP. Nahon, “Statistical Perspectives on Using Hepatocellular Carcinoma Risk Models to Inform Surveillance Decisions, ”Journal of Hepatology79(2023): 1332–1337. doi.org/10.1016/j.jhep.2023.05.005
28. N. Fujiwara, M. Kobayashi, A. J. Fober, et al. , “A Blood‐Based Prognostic Liver Secretome Signature and Long‐Term Hepatocellular Carcinoma Risk in Advanced Liver Fibrosis, ”Medicus2(2021): 836–850. doi.org/10.1016/j.medj.2021.03.017
29. N. Fujiwara, C. Lopez, T. L. Marsh, et al. , “Phase 3 Validation of PAaM for Hepatocellular Carcinoma Risk Stratification in Patients With Cirrhosis, ”Gastroenterology168(2025): 556–567. doi.org/10.1053/j.gastro.2024.10.035

---

Republished from the open web under CC-BY. Authors: Toyoda H, Hoshida Y, Parikh ND, Jalal PK, Piñero F, Mendizabal M, Ridruejo E, Cheinquer H, Casadei-Gardini A, Kanneganti M, Weinmann A, Peck-Radosavljevic M, Dufour JF, Radu P, Shiha G, Soliman R, Sarin SK, Kumar M, Wang JH, Tangkijvanich P, Sukeepaisarnjaroen W, Atsukawa M, Uojima H, Nozaki A, Nakamuta M, Takaguchi K, Hiraoka A, Abe H, Matsuura K, Watanabe T, Shimada N, Tsuji K, Ishikawa T, Mikami S, Itobayashi E, Johnson PJ, Singal AG. Read the original.