Hepatocellular carcinoma (HCC) is currently the most serious complication of chronic hepatitis B virus (HBV) infection and one of the leading causes of cancer-related mortality worldwide [1]. Multiple HCC risk factors in chronic HBV patients have been described to date, including cirrhosis, older age, male sex, co-existence of alcohol abuse, diabetes or metabolic syndrome, active smoking, positive family history and others [2]. Additionally, HCC has been associated with some features pertaining to HBV infection, such as chronic necro-inflammatory activity, high HBV DNA and/or HBsAg levels, HBV genotype C (versus B) and presence of certain mutations, especially mutations in the basal core promoter region [3] or nonsense mutations in the surface gene (preS1 and preS2 regions).

HBV has high oncogenic potential itself and HBV-related carcinogenesis follows a multifactorial and multi-route process, which involves insertional mutagenesis following HBV DNA integration into host genome, increased genomic instability caused by HBV DNA integration and the direct effect of viral proteins, as well as dysregulation of normal cell functions (i.e. proliferation, apoptosis, DNA repair). This oncogenic activity is further enhanced in case of chronic active inflammation, during which increased oxidative stress and necrosis lead to subsequent regeneration, angiogenesis and cellular senescence, thus promoting mutagenesis and carcinogenesis. HBV DNA integration into the host genome has been shown to occur early in the phase of chronic HBV infection and at early steps of liver carcinogenesis [4]. The early HBV DNA integration may explain why non-cirrhotic chronic hepatitis B (CHB) patients under antiviral treatment can still carry a non-negligible risk of HCC development, which has considerable implications for their long-term monitoring. However, it is well recognized that the majority of HBV infected patients who are diagnosed with HCC have already developed cirrhosis.

The current management of CHB is based on therapy with interferon-alfa (IFNa) or a nucleos(t)ide analogue (NA). As long-standing high viral replication and active necro-inflammation have been associated with increased risk for HCC [5] in CHB patients, antiviral therapy which inhibits HBV replication and improves the necro-inflammatory activity is expected to decrease the HCC incidence. However, several studies suggest that HCC may still develop in treated CHB patients and it is debatable whether the HCC risk is decreased in CHB patients under antiviral therapy, particularly with the current treatment options, pegylated IFNa (peg-IFNa) or one of the high genetic barrier NAs, entecavir and tenofovir. In this review we sought to assess relevant evidence which evaluated the risk of HCC in CHB patients under treatment and determined predictors of HCC in this setting.

In a recent systematic review from Raffetti et al [2] which included 66 studies with 347,859 untreated patients, the summary HCC incidence rates ranged from 0.03 to 0.017 cases per 100 person-years (PYs) in inactive chronic HBV carriers, 0.12 to 0.49 cases per 100 PYs in CHB patients and 2.03 to 3.37 cases per 100 PYs in patients with HBV compensated cirrhosis. Accordingly, the 5-year cumulative HCC risks ranged from 0.1% to 0.3% in inactive carriers, 0.6% to 2.4% in CHB patients and 9.7% to 15.5% in cirrhotics, with the rates being higher in patients from East Asia rather than Europe. Multivariate analysis confirmed previous knowledge showing a significant increase of the HCC risk with more advanced phases of liver disease, older age, male gender, HBV genotype C and increasing levels of HBV DNA and HBsAg.

Although peg-IFNa is practically the only IFNa currently used in the treatment of CHB, almost all studies assessing the HCC in IFNa treated CHB patients have used standard IFNa. Theoretically, IFNa therapy may decrease the HCC risk not only due to its antiviral but also due to its immunomodulatory and antitumoral properties. In agreement with the theoretical background, most published studies have shown a significant reduction of the HCC incidence risk in IFNa treated patients compared to untreated controls (Table 1). The first meta-analysis published in 2001 by Camma et al included 7 studies (2 Oriental, 5 European) with 1,505 cirrhotic patients and suggested that IFNa therapy can achieve 6.4% risk reduction in the incidence of HCC (P<0.001) [6]. The reduction of the HCC incidence resulted mainly from the two Oriental studies, while the 4.8% reduction of HCC incidence in the five European studies did not reach statistical significance. In another meta-analysis published in 2008 by Sung et al [7] including 12 studies with 2,742 patients, IFNa was found to reduce the HCC incidence by 34% over a follow-up of 4.9-8.9 years. More specifically, HCC developed in 4.6% in patients treated with IFNa and 9.0% in untreated controls (relative risk [RR]: 0.66, 95% confidence interval [CI]: 0.48-0.89). IFNa therapy offered no benefit in the low HCC incidence rates in non-cirrhotic patients (0.9% vs. 1.1%, RR: 0.72, 95% CI: 0.16–3.15), but a clear reduction in the HCC incidence in patients with cirrhosis (11.6% vs. 21.5%; RR: 0.53, 955 CI: 0.36-0.78). Similar findings were reported in four subsequent meta-analyses published in 2009-2011 [8-11].

According to the existing data, IFNa and probably peg-IFNa seem to decrease the incidence of HCC, especially in Asian patients with cirrhosis who have high baseline risk for HCC. The HCC risk is lower in IFNa treated patients with sustained off-treatment virological response than non-sustained responders [7], but such a response can be achieved in up to 35% of HBeAg-positive and up to 25% of HBeAg-negative CHB patients who receive peg-IFNa [12].

Studies comparing the effect of IFNa and NAs on the prevention of HCC are generally lacking. In a very recent report by Lianq et al [13], however, treatment with peg-IFNa was associated with a lower HCC incidence compared to NA (P=0.011) or entecavir therapy (P=0.018) in 330 CHB patients, even after baseline matching, despite the more potent suppression of HBV replication by NAs.

Older age, male gender and presence of cirrhosis represent widely accepted risk factors for HCC development in chronic HBV patients, which have been confirmed not only in untreated but in NAs treated patients as well [2,15,19,21].

In particular for NAs treated CHB patients, the role of on-therapy virological remission has been assessed in several studies. In the systematic review by Papatheodoridis et al [15] including mostly the early lamivudine studies with or without adefovir rescue therapy, failure to remain in virological remission under lamivudine was found to be associated with a higher HCC incidence particularly in cirrhotic patients. More recent studies further supported the concept that maintained on-lamivudine virological remission reduces the HCC incidence compared to patients with suboptimal responses or virological breakthroughs [30,31]. However, in a large Greek multicenter study [32] including 818 HBeAg-negative CHB patients who started therapy with lamivudine and received adefovir upon lamivudine resistance, the virological on-lamivudine remission was not found to significantly affect the HCC incidence. Similarly to the lamivudine data, virological remission on-entecavir has also been found to reduce the HCC incidence in recent studies [33,34]. A critical factor that may be responsible for the above discrepancies might be the proportion of patients with lamivudine resistance who received rescue therapy as well as differences in the follow-up of patients under lamivudine therapy and/or in the intervals from lamivudine resistance to rescue therapy.

Given that HBsAg loss is considered a difficult to be achieved but the optimal treatment end-point, the risk of HCC in patients who clear HBsAg under NAs is of particular interest. Recently, Kim et al [35] evaluated the HCC incidence (among other clinical outcomes) in 110 of 5,409 CHB patients who achieved HBsAg seroclearance over a median 6-year therapy with NAs. During 287 PYs following HBsAg seroclearance, only two patients (both with cirrhosis at baseline) developed HCC or died (0.7% annual risk) having a significantly lower rate compared to propensity-matched CHB patients without HBsAg seroclearance (HR 0.09, P<0.01). However, this study also confirmed that a residual HCC risk still exists even after HBsAg seroclearance.

The prognostic role of alpha fetoprotein (AFP) for HCC development has also been assessed in three studies including NAs treated patients. In a study by Wong et al [36] with 57 cases of HCC in 1,531 patients treated with entecavir for a mean of 51 months, baseline serum levels of AFP were predictive of HCC development. In this study, a baseline AFP cut-off value of 20 μg/L offered sensitivity 39% and specificity 99% for subsequent HCC development, while a lower cut-off of 6 μg/L increased the sensitivity to 81% but decreased the specificity of 80%. A study by Shim et al found a cumulative HCC incidence of 9.5% during 3 years of follow-up in 207 patients treated with NAs [37]. Serum AFP levels of 20 ng/dL at 12 months of therapy were strongly associated with HCC development offering positive predictive value of 100%. Finally, in the study by Yanq et al including 244 entecavir treated patients, HCC development was associated with persistently elevated AFP levels for at least 6 months of therapy [38].

Individual HCC risk factors have been identified, but they cannot adequately classify CHB patients according to their HCC risk. Thus, there have been recent efforts for the development of accurate risk scores combining some of these risk factors that can predict the HCC risk in this setting.

HCC risk scores have been initially developed in untreated cohorts of Asian CHB patients. The most well-known such scores are CU-HCC, CAG-HCC and REACH-B (Table 2) [39-42]. The predictability of these scores has also been assessed in Asian CHB patients treated with NAs, mainly entecavir. In particular, Wonq et al. investigated the accuracy of these HCC risk scores in a cohort of 1531 CHB patients (22% with cirrhosis), treated with entecavir and followed for a mean of 42 months [43]. The 5-year cumulative HCC incidence rates were 12.9% in cirrhotics and 2.1% in non-cirrhotics, while the areas under the ROC curves of baseline CU-HCC, GAG-HCC, and REACH-B scores for HCC prediction were 0.80, 0.76, and 0.71, respectively. Moreover, CHB patients were reported to be at no or negligible HCC risk if they had low scores at both baseline and two years of therapy, at intermediate HCC risk if they had high score at baseline and low score at two years of therapy and at high HCC risk if they had high scores at both time points.

In three other studies comparing observed HCC incidence rates with that predicted by REACH-B score, the HCC incidence rates under NAs [44], entecavir [45] or tenofovir [46] were reported to be significantly lower than the expected HCC incidence rates according to the REACH-B score (standardized incidence ratios 0.37-0.46).

Unfortunately, the above HCC risk scores, which have been developed in Asian patients, do not seem to offer acceptable predictability for HCC in Caucasian CHB patients. In a large study by Papatheodoridis [29] including 1666 Caucasian CHB patients treated with entecavir and/or tenofovir, GAG-HCC, CU-HCC and REACH-B were associated with HCC development only in the univariate but not in the multivariate analyses and offered poor to modest predictability for HCC with areas under the ROC curve of 0.63-0.75. Similarly, in another multicenter study from 11 European referral centers including 744 entecavir treated CHB patients (42% Caucasian, 29% Asian, 19% other, 10% unknown), the HCC predictability of CU-HCC, GAG-HCC and REACH-B scores was low offering areas under the ROC curve of 0.54-0.74 [47].

Finally, a HCC risk score for Caucasian CHB patients under antiviral therapy was developed and validated in a recent study by Papatheodoridis et al [42]. They included 1815 adult Caucasian patients treated with entecavir or tenofovir for at least 12 months and developed a simple and accurate risk score for HCC development within 5 years of therapy, called PAGE-B, which includes platelets, age and gender (Table 2).

The risk of HCC merits special attention in the management of CHB due to its relatively high incidence even in patients who achieve long-standing viral suppression under treatment. Well known HBV-related HCC risk factors include older age, male sex, presence of cirrhosis and long-standing HBV replication, among others. The current armamentarium of antiviral drugs used in the treatment of CHB has improved the overall outcome of CHB patients but HCC may still develop. The effect of anti-HBV treatment in HCC prevention in CHB is more evident in patients with higher baseline HCC risk such as those with cirrhosis. Peg-IFNa probably decreases the HCC risk particularly in patients who achieve sustained off-treatment responses. NAs, which are used by the majority of CHB patients, achieve long-standing inhibition of HBV replication and seem to decrease but not eliminate the HCC risk, according to data from matched controlled studies. Thus, prediction of HCC risk remains of particular importance in patients under current therapies. In this effort, several HCC risk scores have been developed and offer good predictability for HCC mostly in Asian CHB patients. Recently, the PAGE-B risk score developed and seem to offer satisfactory performance in Caucasians CHB patients under the current oral agents.