Impact of tenofovir alafenamide vs. entecavir on hepatocellular carcinoma risk in patients with chronic hepatitis B

Whether entecavir (ETV) or tenofovir alafenamide (TAF) is better at preventing hepatocellular carcinoma (HCC) development among patients with chronic hepatitis B (CHB) remains unclear. The present study was conducted to explore the ability of these two antivirals to prevent HCC. From 2012 to 2019, treatment-naïve CHB patients undergoing ETV or TAF therapy were recruited at three academic teaching hospitals. The TAF group comprised patients starting TAF as first-line antiviral and those switching antivirals from tenofovir disoproxil fumarate to TAF. Patients with decompensated cirrhosis or HCC at enrollment were excluded from the analysis. Cumulative probabilities of HCC were assessed using the Kaplan–Meier method. In total, 1810 patients (1525 and 285 in ETV and TAF groups, respectively) were recruited. The annual HCC incidence was statistically not different between the ETV and TAF groups (1.67 vs. 1.19 per 100 person-years, respectively) with an adjusted hazard ratio (HR) of 0.681 (p = 0.255), as determined by multivariate analysis. Male, hypertension, liver cirrhosis, FIB-4 index, and albumin were independent prognostic factors for HCC development. Propensity score-matched and inverse probability of treatment weighting analyses yielded similar results, with non-statistically different HCC incidence between the ETV and TAF groups (1.07 vs. 1.19 per 100 person-years (HR = 0.973; p = 0.953) and 1.67 vs. 1.89 per 100 person-years, respectively (HR = 0.949; p = 0.743). These findings suggest that ETV- and TAF-treated CHB patients have similar risk of developing HCC. Further studies with the larger sample size and longer follow-up are needed to validate these results.


Introduction
Chronic hepatitis B virus (HBV) infection affects approximately 350 million people worldwide, and chronic hepatitis B (CHB) is endemic to East Asia [1][2][3][4][5]. Given the persistent intrahepatic replication status of HBV-DNA, HBV infection itself is significantly associated with increased risk of liverdisease progression to cirrhosis and/or hepatocellular carcinoma (HCC) [6,7]. Therefore, replication-suppressing antiviral therapy with potent nucleos(t)ide analogues (NUCs) and high genetic barrier to resistance is recommended to patients with chronic HBV infection to prevent liver-disease progression [2]. Nevertheless, as the HBV-DNA integrates into the host hepatocyte genome, the virus is rarely eradicated through long-term antiviral therapy, and most patients with CHB additionally require periodic HCC surveillance [8,9].
Recently, along with entecavir (ETV) and tenofovir disoproxil fumarate (TDF), tenofovir alafenamide (TAF) was accepted as first-line NUC for the treatment of older populations or patients with co-morbidities for renal or bone disease. This approval was based on the similar short-to intermediate-term antiviral effects of these three agents in treatment-naïve CHB patients [10,11]. Furthermore, effective rescue regimens may offset the potential hazard by suboptimal virological response or genotypic resistance even in a very small proportion (approximately 1%) of patients treated with ETV [12,13]. Accordingly, the long-term clinical efficacy for preventing the risk of liver-disease progression to cirrhosis and/or HCC is expected to be similar among the regimens. Nevertheless, since Choi et al. [14] reported that TDF is associated with a significantly lower risk of HCC (hazard ratio [HR] = 0.61) and all-cause mortality or orthotopic liver transplant (HR = 0.77) than ETV, several studies were conducted to validate such phenomena. However, this issue remains controversial due to somewhat contradictory results among studies, including similar efficacy between patients receiving two antivirals, overall favorable outcomes among those treated with TDF, or discrepant results according to presence of cirrhosis or follow-up duration [15][16][17][18][19][20]. In addition, a more recent study based on data from two phase III clinical trials [21,22] showed that patients treated with TAF showed a tendency for lower risk of HCC development, even though not statistically significant (p = 0.14), compared to those treated with TDF [23].
The present large-scale, multi-center cohort study was conducted in three academic teaching hospitals in the Republic of Korea aiming to further explore the efficacy of ETV-and TAF-based treatment in treatment-naïve CHB patients, regarding the risk of HCC development.

Subjects
Treatment-naïve CHB patients who underwent antiviral therapy with either ETV 0.5 mg/day (ETV group) or TAF 25 mg/day-based regimen (TAF group) from 2012 to 2019 in three academic teaching hospitals (Yonsei University Severance Hospital, Kyungpook National University Hospital, and Chung Ang University Hospital) were consecutively screened for eligibility. TAF group comprised patients starting TAF as a first-line antiviral regimen as well as those who switched NUCs from the TDF to the TAF regimen. The inclusion criteria were as follows: (1) age ≥ 19 years, (2) well-preserved liver function, and (3) follow-up duration of at least 6 months. The exclusion criteria were as follows: (1) history of HCC at enrollment, (2) decompensated cirrhosis at enrollment, (3) change of antiviral from ETV to TDF or TAF, (4) change of antiviral from TDF or TAF to ETV, (5) co-infection with other hepatitis virus, (6) history of organ transplantation, (7) development of clinical events (HCC, death, or orthotopic liver transplant) within 6 months of enrollment, and (8) other significant medical illnesses such as any kind of malignancy with advanced stage. Owing to the homogenous nature of the study population, data on race/ ethnicity were not collected.
In the Republic of Korea, the reimbursement criteria for ETV, TDF, or TAF are identical (Supplementary Table 1). If histologic information was not available, compensated cirrhosis was clinically defined according to the following criteria: (1) platelet count < 150,000/μL and ultrasonographic findings suggestive of compensated cirrhosis, including a blunted, nodular liver surface accompanied by splenomegaly (> 12 cm); or (2) esophageal or gastric varices.

Clinical evaluation, follow-up, and outcomes
During follow-up, all patients underwent routine bloodchemistry testing, and serum HBV-DNA levels and of other viral markers were assessed every 3-6 months. Patients also evaluated by ultrasonography and the serum levels of alphafetoprotein (AFP) were determined every 6 months to screen for HCC and cirrhotic complications [24][25][26].
The primary outcome of the study was HCC development, as diagnosed based on histological evidence or dynamic computed tomography, and/or magnetic resonance imaging findings (nodule > 1 cm with arterial hypervascularity and portal/delayed-phase washout) [27][28][29][30]. The index date was the date of the first antiviral prescription and the time to HCC development was considered as the period between the index date and the date of HCC diagnosis or the end of follow-up in the absence of HCC development.

Statistical analysis
Data are expressed as means ± standard deviation or as numbers (%). Differences among continuous and categorical variables were examined for statistical significance using the Student's t test (or the Mann-Whitney test, if appropriate) and the chi-squared test (or Fisher's exact test, if appropriate). The cumulative risk of HCC was calculated using the Kaplan-Meier method and was compared using the log-rank test. To find independent predictors of HCC development, univariate Cox regression analysis was first performed for each variable, and then, only significant univariate predictors were entered into subsequent multivariate Cox regression analysis.
To reduce selection bias and the effect of potential confounders, propensity scores (PS) were calculated by logistic regression based on age, gender, diabetes, hypertension, compensated cirrhosis, hepatitis B e-antigen (HBeAg), aspartate aminotransferase (AST), alanine aminotransferase (ALT), total bilirubin, albumin, and platelet count. Differences between the ETV and TAF groups were balanced by a 1:1 PS-matched and inverse probability of treatment weighting (IPTW) analyses.

Prognostic factors affecting HCC development
Next, these identified potential risk predictors were further assessed by multivariate analysis, which revealed that male (   Using two statistical methods to test the proportional hazard assumption for Cox regression model, we confirmed that TAF group (vs. ETV group) did not satisfy such an assumption; (1) proportionality test using timedependent Cox regression model (p = 0.019) and (2) Schoenfeld residual method (p = 0.002; Supplementary  Fig. 3). Thus, stratified Cox regression analysis was additionally performed according to stratification by other major covariates such as male, diabetes, hypertension, liver cirrhosis, FIB-4 index > 3.25, and albumin ≤ 3.5 g/dL. Supplementary Table 3 consistently shows that the TAF group (vs. ETV group) was not associated with the HCC risk (all p > 0.05).

Clinical outcomes after adjustment by PS matching
The 1:1 PS-matched analysis generated 285 pairs, of which the standardized mean differences of all variables converged upon almost 0.1 (Supplementary Fig. 1), suggesting the appropriate balancing of the variables between the ETV and TAF groups. The baseline characteristics of the two groups are described in Table 3. No significant difference in any variables was observed between the groups (all p > 0.05), except for AST and ALT levels and platelet count. The cumulative risk of HCC development at 1, 3, and 5 years was of 0.7%, 2.6%, and 5.8% (annual incidence: 1.07 per 100 person-years) in the ETV group and was 0.0%, 1.3%, and 10.6% (annual incidence: 1.19 per 100 person-years) in the TAF group ( Fig. 2; p =

Clinical outcomes after adjustment by IPTW
The standardized mean differences of all variables after IPTW also converged upon almost 0.1 ( Supplementary  Fig. 2), suggesting the appropriate balancing of the variables between the ETV and TAF groups. The baseline characteristics of the two groups were described in Table 4. No significant differences were observed for most variables between the ETV and TAF groups (all p > 0.05), except for age (p < 0.001), liver cirrhosis (p = 0.034), and the levels of AST (p < 0.001), ALT (p < 0.001), total bilirubin (p = 0.040), and albumin (p = 0.001). The cumulative risk of HCC development at 1, 3, and 5 years was of 0.8%, 4.3%, and 10.6% (annual incidence: 1.67 per 100 person-years) in the ETV group and 0.0%, 2.7%, and 17.2% (annual incidence: 1.89 per 100 person-years) in the TAF group ( Fig. 3;

Discussion
In line with the ongoing controversy about which is better between ETV vs. TDF for preventing HCC development, the most recent study showed a trend that TAF has a more favorable preventive effect than TDF [23]. Given the poor prognosis of HCC, determining the treatment of choice for patients with CHB could become a scientifically, socio-economically, and ethically important matter. Hence, in our independent, large-scale, multi-center cohort study, this issue was addressed. The present study showed that prescribing TAF as a first-line NUCs or switching NUCs from TDF to TAF did not provide a statistically significant benefit over long-term ETV as a first-line treatment, as consistently demonstrated by the comparable clinical outcomes regarding HCC development in all the analyses performed (including not only unadjusted analysis but also multivariate, PS-matched, and IPTW analyses; all p > 0.05). This study had several strengths. First, the large sample of approximately 1800 patients from three independent academic teaching hospitals enhanced the reliability of the results. Considering that TAF has been officially reimbursed by the National Health Insurance Service of the Republic of Korea since November 2017, the relatively sufficient number of HCC cases (4.3%) with a mean follow-up period of 37.0 months in the present study contrast with the 1.3% HCC incidence reported by Lim et al. [23], thereby providing adequate statistical reliability to address this issue [29]. Furthermore, from the statistical viewpoint, the HCC-free probabilities up to 70 months were 0.882 (95% CI 0.849-0.915) for the ETV group and 0.851 (95% CI 0.761-0.941) for the TAF group. Since the width of the 95% CI for the TAF group completely contains that for the ETV group, we can cautiously speculate that the lack of statistical difference between two groups was genuine, not a type II error. Second, to minimize the confounding effects through the change of medication from ETV to TDF/TAF, or vice versa, owing to poor compliance from adverse events or sub-optimal virological response, we excluded such patients. Indeed, Lim et al. [23] did not show also the statistical significance between the two groups (p = 0.14). Nevertheless, it is noteworthy that their Kaplan-Meier curves of the two groups did not cross each other during the whole follow-up period [23], suggesting a possibility of positive results in a future study with a larger sample size and longer follow-up. Therefore, further clinical studies or trials should be required to solve such a controversial issue. Finally, the major novelty of this study as a point of difference compared to Choi et al.'s study [31] is that we aimed to explore whether TAF-based regimen might reduce HCC incidence compared with ETV, on the assumption that ETV vs. TDF showed the similar level of preventive effect on HCC risk and that TAF had theoretically better anticancer profile than TDF in terms of reducing necro-inflammation and other metabolic complications. Although Choi et al. [31] concluded that TDF was associated with a significantly lower risk of HCC compared with ETV in a population-based cohort of adults with CHB, it was not reproduced through further studies from the Republic of Korea [16,20,32]. Likewise, expectably, in the present study, there was no significant difference in HCC risk between ETV group and those who experienced TDF treatment (p = 0.651). Most decisively, Choi et al. [33], based on the similar kind of a nationwide cohort used by Choi et al. [31], recently indicated that significant difference between two groups before adjustment became quite offset, after the "timing of starting medications" between two groups was adjusted to "between 2013 and 2017", because TDF had been available only since Dec. 2012 in the Republic of Korea. Hence, from a series of such studies, we should also recognize the importance of time-related bias for an appropriate interpretation of the study results.
One of the most plausible hypotheses explaining such a trend of the favorable outcome in patients treated with TAF is that they are more likely to achieve the normalization of ALT than those receiving ETV or TDF [10,21,22], given that on-treatment ALT normalization is associated with a lower risk of HCC development [34][35][36]. However, since the upper limit of normal value varies among studies and/or local laboratories and the ALT normalization itself is closely associated with various factors such as age, gender, steatosis, metabolic syndrome, alcohol use, and other medications that can potentially affect the long-term prognosis, further studies are required to draw a thorough conclusion. The carcinogenic potential of ETV and the induction of interferon (IFN)-λ3 production by tenofovir might in part explain the favorable outcome by TAF in comparison with ETV. However, such hypotheses are also still problematic. First, ETV was reported to increase the incidence of lung adenomas and carcinomas, HCC, and vascular tumors in mice at 4 mg/kg; and of HCC, brain microglial tumors, and skin fibroma in rats at 1.4-2.6 mg/kg [37]. However, these dosages are at least 100-fold higher than those used in humans. In contrast, two recent large-scale real-life studies reported that longterm ETV therapy does not increase the risk of cancer [38,39]. Moreover, in the long-term follow-up study by Kim et al. [40], the incidence of HCC was statistically not different during and after the first 5 years of ETV treatment (2.29% vs. 1.66%, p = 0.22). If long-term ETV maintenance retained significant pro-carcinogenic effects for humans, the HCC incidence would have progressed rapidly with time. In addition, Murata et al. [41] reported that the level of serum IFN-λ3 which might have anticarcinogenic and antiviral effect was higher in patients treated with tenofovir than in those treated with ETV. However, since conflicting data have also been reported [28,[42][43][44][45] and IFN-λ assays have not been standardized, the causality of the relationship between higher IFN-λ3 levels and lower incidence of HCC should be confirmed in further studies.
In the present study, arterial hypertension was an independent risk factor for the development of HCC, in contrast to diabetes. To date, the direct causality between hypertension and hepato-carcinogenesis had not been proved yet. However, there have been some observations that diabetes is associated with the risk of HCC development [45,46]. We also observed that diabetes had a marginal effect on HCC development, but hypertension is more closely associated with HCC risk. The one of the most possible explanations is multi-collinearity among hypertension and other risk factors, e.g., steatosis and obesity. Actually, patients with hypertension were more likely to have steatosis (14.5% vs. 10.7%; p = 0.036) and obesity (37.5% vs. 26.9%; p < 0.001) than those without. In the similar context, according to the recent study [46], the metabolic syndrome which comprehensively covers hypertension, diabetes, obesity, and dyslipidemia is closely associated with HCC development. Therefore, to evaluate the causality between the metabolic diseases and the risk of HCC development appropriately, we should recognize various kinds of metabolic diseases as a spectrum of "specific" syndrome, rather than each single disease entity.
This study also had some limitations. First, since TAF has been officially reimbursed by the National Health Insurance Service in the Republic of Korea since November 2017, a relatively small proportion of patients treated with TAF was available, and their follow-up data were not adequate to observe a sufficient number of liver-related events, which may introduce selection bias, particularly regarding treatment allocation. To overcome this, various statistical adjustments including PS-matched and IPTW analyses were performed, which confirmed the reproducibility of the results. Nevertheless, a prospective cohort study of the association between antiviral type and HCC risk with long-term follow-up is needed. Second, most of our study population (> 98%) was infected with HBV genotype C through vertical transmission, which was significantly associated with the increased probability of HCC occurrence [47][48][49][50]. Thus, further studies for external validation are required. Third, the use of biomarkers (e.g., serum quantitative HBsAg, serum hepatitis B core antigen [HBcAg], serum HBV-RNA, specific HBV mutants, or prothrombin Induced by vitamin K absence-II [PIVKA-II]) that could reflect the clinical course of CHB would have allowed more detailed analyses [51][52][53][54][55][56]. However, in South Korea, HBcAg is currently unavailable in the routine practice and the reimbursement of quantitative HBsAg at baseline for patients starting AVT is not also approved according to the guideline by the National Health Insurance Service. Likewise, a measurement of PIVKA-II level without evidence of HCC is not reimbursed. Further studies are required using the new biomarkers at baseline and on-treatment. Finally, many HCC cases were diagnosed using imaging tests without confirmatory biopsy. Nevertheless, the non-invasive diagnosis of HCC based on imaging modalities is generally accepted in the routine practice [27,28,57] and liver biopsy was performed for equivocal cases in our study.
In conclusion, the overall prognosis regarding HCC development was not statistically different between patients treated with ETV or TAF. Because prevention of liver-disease progression by appropriate antiviral therapy is a very important medical and socio-economical issue, further studies with long-term follow-up are needed to validate these results.