Current trends in the risk prediction for hepatitis B virus-related hepatocellular carcinoma

A variety of risk factors have been identified for the development of HCC, which can be divided into patient factors and viral factors (Table 1). Cirrhosis is the single most important risk factor for HCC in patients with CHB even though HCC can develop in its absence. In both East Asian and Western populations, patients with CHB and cirrhosis have a more than a 5-fold increased incidence of HCC compared with CHB patients without cirrhosis [4]. As liver injury and fibrosis in patients with CHB accrues over time, the incidence of HCC also increases with age [5]. Other nonmodifiable risk factors include male sex [1] and a family history of HCC [6]. Additional liver insults from coinfection (e.g. hepatitis C, hepatitis delta) or concomitant liver disease (e.g. alcohol abuse, nonalcoholic fatty liver disease) can accelerate fibrosis progression and development of HCC [7]. Similarly, liver inflammation as evidenced either biochemically (high alanine aminotransferase) or histologically (necroinflammation) has been associated with increased risk of HCC [8].

Because HBV is a direct carcinogen to the liver, several viral factors augment the risk of HCC. Patients with active viral replication as demonstrated by hepatitis B e antigen (HBeAg) seropositivity [9] or high serum HBV DNA levels [10] are at increased risk of HCC compared with so-called “inactive carriers” (HBeAg negative, normal alanine aminotransferase [ALT], low level HBV DNA). The REVEAL-HBV study demonstrated a dose-response relationship between the incidence of HCC and baseline HBV DNA level with elevated HBV DNA levels ≥10,000 copies/mL (i.e. approximately 2,000 IU/mL) being a strong independent predictor of HCC [10]. Although a weaker predictor than HBV DNA, the clinical significance of quantitative hepatitis B surface antigen (HBsAg) has also become increasingly recognized. HBsAg levels can complement HBV DNA level in predicting HCC development, especially when HBV DNA <2,000 IU/ mL, with high levels (≥1000 IU/mL) conferring greater risk than lower levels [11]. Patients with genotype C or certain mutations (e.g. A1762T/G1764A basal core promoter or pre-S deletion) are also contributors to HCC risk [12].

Treatment of CHB with antiviral therapy has been shown to reduce, but not completely eliminate the risk of HCC [13]. While overall HCC risk scores can still be accurately applied to patients on treatment [14], some risk factors such as HBeAg status, HBV DNA, ALT, and cirrhosis may need to be refined or even omitted, because the natural history of the disease has been modified. For example, in treated patients, baseline HBV DNA is no longer a predictor of HCC, whereas failure to achieve and maintain viral suppression is associated with a higher risk of HCC [15-17].

A challenge with HCC is that it can be rapidly progressive and present at an advanced stage in the absence of symptoms. Hence, there is a need to identify early tumors, which may have better prognosis and be amenable to effective treatment. However, only modest evidence supporting the efficacy of HCC surveillance in high-risk groups is found. Although observational and uncontrolled studies have demonstrated earlier detection (stage migration) and improved survival with surveillance [18, 19], they have been subject to lead-time bias and may not reflect a true reduction in cancer-specific mortality [20]. There is one large randomized controlled trial of 18,816 Chinese HBV positive patients that showed surveillance with 6-monthly ultrasound scan (US) and alpha-fetoprotein (AFP) was associated with a 37% reduction in mortality because of HCC [21]. Nonetheless, surveillance for patients at high risk of HCC is recommended in practical guidelines issued by American Association of the Study of Liver Diseases (AASLD), Asian Pacific Association for the Study of the Liver (APASL), and European Association for the Study of the Liver (EASL) (Table 2) [22-24].

All three key international guidelines support surveillance in cirrhotic patients, but differ on their definition of the at-risk population in noncirrhotic CHB patients. APASL acknowledges the need for further (randomized) studies in order to make recommendations in noncirrhotic patients [23]. AASLD guidelines target populations where the incidence of HCC exceeds 0.2% per year (the cutoff for cost-effectiveness in CHB) based on age, sex, ethnicity, and family history [22], while EASL recommends surveillance in HBV carriers with active hepatitis and patients with a family history of HCC [24]. All three guidelines support 6-monthly HCC surveillance with abdominal ultrasonography. The APASL recommends the addition of AFP to US for surveillance; however, this has not been adopted by EASL or AASLD because of its suboptimal performance as a serological test in this setting.

US, ultrasound scan; AFP, alpha-fetoprotein; AASLD, American Association for the Study of Liver Diseases; APASL, Asian Pacific Association for the Study of the Liver; EASL, European Association for the Study of the Liver; EORTC, European Organisation for Research and Treatment of Cancer

HCC surveillance in resource-limited, high HBV-burden settings remains a challenge. While HCC surveillance has been demonstrated to be cost-effective in some countries (Egypt) [28], it has not in others (India) [29]. These differences reflect the variation in HCC prevalence and economic situation in each region, with greater cost-effectiveness of surveillance programs seen in countries with high prevalence [30]. Effective surveillance programs also require a means for implementing treatment for early HCC. Access to curative treatments such as liver transplantation or resection is limited, even in high-income countries. Therefore evaluation of low-cost treatment strategies including alcohol injection for small HCCs in low- and middle-income countries is a current research gap [31]. Furthermore, preventive measures that have been shown to reduce HCC incidence such as HBV vaccination [32] and antiviral therapy [13] have been hampered by poor access in developing countries with high endemicity [33].

Several published scoring systems have been developed to predict the risk of HCC in patients with CHB (Table 3). The development of all HCC risk scores follows a similar methodology. First, a collection of independent factors associated with HCC is derived from a (training) cohort of patients with CHB through multivariable analysis by Cox proportional hazards regression. The relative weighting assigned to each risk factor in the final score is determined by its regression coefficient. A scoring system is then formulated and must be validated with an independent (validation) cohort [5]. If an external validation cohort is not available then a leave-one-out cross validation is performed [34].

The three commonly applied HCC risk scores were derived from data on prospective Asian cohorts primarily untreated with antiviral therapy.

While HCC risk scores need to be re-evaluated in cohorts of European ancestry and refined in patients on antiviral therapy, another frontier for improvement is in the diagnosis of cirrhosis. Presence of cirrhosis has been assigned heavy weightings in the CU-HCC [35] and GAG-HCC [34] scores and was used to exclude patients in the study of the REACH-B score [5]. In the PAGE-B score [15], cirrhosis was strongly associated with HCC, but was less predictive than platelet count, a nonspecific marker. However, the diagnosis of cirrhosis in these studies was assessed using APRI, hypoalbuminemia, and presence of nodularity or portal hypertension on abdominal ultrasound. As these tests are insensitive markers of early cirrhosis [43, 44], a substantial number of patients may have been misclassified as having no cirrhosis. Despite being the most important risk factor for HBV-related HCC, cirrhosis has been diagnosed using relatively crude measures in major risk score studies thus far.

The cohorts used in HCC risk score studies existed before the widespread use of transient elastography, a noninvasive test to evaluate liver stiffness measurements (LSM). Not only have LSM been widely validated to accurately diagnose cirrhosis in patients with CHB [45], it has also been shown to predict HBV-related HCC in a dose-dependent manner [46]. Therefore, it should be advantageous to integrate LSM into existing HCC risk scores. This was studied on a prospective cohort of 1,555 Chinese patients (1,035 in a training cohort, 520 in a validation cohort) with CHB referred for transient elastography [47]. Roughly a third of patients underwent antiviral therapy. Patients were assessed using the CU-HCC score with the original clinical diagnosis of cirrhosis (using ultrasound features) substituted by LSM, thus creating a new LSM-HCC score. LSM cut-off values of 8.0 kPa and 12.0 kPa were chosen to stratify patients into three risk categories for HCC. Serum bilirubin from the CU-HCC score was no longer discriminatory after integration of LSM. The final components of the LSM-HCC score were age, albumin, HBV DNA, and LSM, which summate to a total score ranging from 0 to 30. The LSM-HCC score appeared superior to the CU-HCC score with higher AUROCs at 3 years (0.89 vs 0.71) and 5 years (0.83 vs 0.75), although this difference was not statistically significant. By applying a cut-off of score 11, the LSM-HCC was able to identify a low risk group with a 3- and 5- year HCC-free survival of 100% and 99.4%, respectively.

A small Korean study of 192 CHB patients with undetectable serum HBV DNA after treatment with entecavir evaluated a modified version of the REACH-B score (mREACH-B) on predicting HCC development [17]. In this model, LSM was incorporated into the REACH-B score in place of serum HBV DNA level, which proved to have no prognostic value in this treated cohort. The mREACH-B was found to have a better predictive performance for HCC at 3 years compared with the conventional score (AUROC 0.805-0.814 vs 0.629). The authors recently proceeded to validate the mREACH-B score in a larger cohort of 1,308 patients (17.8% cirrhosis, 65% commenced or continued antiviral therapy) and assessed its accuracy against other conventional HCC prediction models (CU-HCC, GAG-HCC, REACH-B, and LSM-HCC scores) [16]. The mREACH-B had significantly higher AUROCs for the prediction of HCC development at 3 and 5 years compared with other models, with the improvement seen primarily among patients receiving antiviral therapy. Interestingly, the exclusion of serum HBV DNA levels from the REACH-B score alone (without the addition of LSM) led to a paradoxical improvement in its predictive performance. This suggests the superior performance of the mREACH-B score is, in part, because of removal of HBV DNA level from the original score, rather than wholly from the addition of LSM. While results from studies of the LSM-HCC and mREACH-B scores are very promising, clearly more study on the role of incorporating LSM into HCC risk scores is warranted.

It is likely that other measurements of liver stiffness such as acoustic radiation force impulse, shear wave elastography and magnetic resonance elastography have predictive capabilities for HCC. However, they have not yet been extensively evaluated to allow for use in HCC risk scores.

Direct biomarkers reflect the turnover of extracellular matrix within the liver. These include glycoproteins (hyaluronate, laminin, YKL-40), collagens (procollagen III N-peptide, type IV collagen), collagenases, and their inhibitors (matrix metalloproteases, tissue inhibitor of metalloprotease-1). The Enhanced Liver Fibrosis (ELF) test is a panel consisting of three direct biomarkers: hyaluronic acid, N-terminal propeptide of collagen type-III, and tissue inhibitor of metalloproteinase-1. Although first studied in hepatitis C patients, it has also proved to be useful in predicting ≥F2 fibrosis in Asian CHB patients [49]. The same authors then studied 170 Korean CHB patients who underwent liver biopsy, transient elastography and ELF testing and demonstrated that the ELF test was a good predictor of liver related events (hepatic decompensation, HCC and death) [50]. The ELF test consistently outperformed LSM and liver histology in predicting all liver related events with higher AUROC values, although the differences were not significant. In particular, for predicting the 31 HCCs in the study, ELF tests had the highest AUROC of 0.746, followed by LSM (0.708) and liver histology (0.686). The authors also showed by using cut-offs values of 8.10 and 10.40, the ELF test can risk stratifying patients into low-, intermediate-, and high-risk groups for HCC. These results suggest the prognostic utility of the ELF test (and potentially other direct biomarkers) is encouraging. However, further study is needed before it can be incorporated in future HCC risk scores.

Indirect markers are nonspecific measurements in blood, which reflect alterations in hepatic function. Examples include platelet count, aspartate to alanine aminotransferase ratio (AST/ALT), age-spleen-platelet ratio (ASPRI) and APRI. Biomarkers can also be combined into larger panels and used as a diagnostic tool (e.g. Fibrotest). Practical advantages of using serum biomarkers to measure fibrosis include their high applicability, reproducibility, widespread availability, and low cost. In the setting of CHB, indirect biomarkers have been shown to predict significant fibrosis, cirrhosis, and even 5-year survival [45]. However, there is currently only limited data to suggest a correlation between these biomarkers and HCC. The aforementioned Korean study by Kim et al. revealed the performance of biomarkers such as ASPRI were suboptimal and inferior to LSM and ELF test for predicting liver related events, including HCC [50]. The role of indirect biomarkers in HCC prediction needs to be explored.