Predicting Hepatocellular Carcinoma Risk in Patients with Chronic HCV Infection and a Sustained Virological Response to Direct-Acting Antivirals

Introduction

Hepatocellular carcinoma (HCC) is currently the fourth cause of liver-related death worldwide,^1,2 and accounts for one of the most frequent indications for liver transplantation. In patients with chronic hepatitis C (CHC), the achievement of a sustained virological response (SVR) to antiviral treatment was demonstrated to reduce the incidence of HCC, already in the Interferon (IFN) era,^3–6 with a more pronounced benefit in those with advanced fibrosis or cirrhosis. After direct-acting antivirals (DAA) approval, a first pivotal study suggested a time-related association between DAA treatment and HCC recurrence,⁷ this finding being initially supported by others.^8,9 Similarly, some authors also reported an increased incidence and biological aggressiveness of de-novo HCC arising in cirrhotics successfully treated with DAA.^9–11 Next, evidences eventually raised against a definite role of oral anti-hepatitis C virus (HCV) treatments as HCC promoter.^12–19 Different crude incidences of HCC in IFN vs DAA-treated cirrhotics mostly rely on differences in patient population, as DAA allow treatment of patients with more advanced liver diseases. Following an SVR to DAA-based regimens, reported rates of de-novo HCC are estimated nearly 2–2.5%^17,20–22 vs 20–30% per year^18,19,23 of recurrent HCC, being definitively higher than that historically reported in the setting of IFN.⁴ Moreover, HCC risk has been demonstrated to persist up to 10 years from treatment completion.¹⁷

Taking together, these data still justify the need for long-life surveillance,^1,24 resulting in intensive follow-up of large cohorts of cured HCV patients. Therefore, current literature efforts aim at deeply investigating predictors of HCC in HCV patients cured through SVR, with the ultimate goal of personalized risk stratification and individualized surveillance policies.

Therefore, in this review, we report data from published study analyzing the risk of HCC development, either de-novo or recurrent, following DAA-based treatments. Particularly, we focused on those studies reporting not only full patients’ characteristics but also information on HCC rates and predictors.

Predictors of de-novo HCC

According to published studies, up to 14% of patients without history of previous liver cancer may develop de-novo HCC after HCV eradication, although data vary according to patient population, follow-up duration and severity of liver disease (Table 1).

Table 1 Assessment of Liver Fibrosis Severity According to Studies’ Designs

Studies Enrolling Patients with Cirrhosis or Advanced Fibrosis

Among 14 studies that reported data in cirrhotic patients,^{8,10,20–22,25–37} (Table 2), most were able to identify HCC predictors (Table 3). In addition, 5 studies enrolling patients with CHC and any fibrosis stage also found-out factors associated with de-novo HCC in the subset of patients with cirrhosis^38–42 although unable to provide their clinical features (Tables 4 and 5). Finally, 5 authors reported data on patients with advanced fibrosis, defined through histology (F3-F4), non-invasive tests or criteria for chronic advanced liver disease (cACLD)^{11,36,43–46} (Table 2). Eight out of these studies enrolled only patients with an SVR,^{10,21,30,33,37,44–46} whilst in one study data could be extrapolated.³⁴ In studies including also non-SVR patients, rates of treatment failure ranged between 1.9% and 10% (Table 2). Follow-up duration varied according to study designs as reported in Table 2. Overall, studies reported de-novo HCC rates of 1.8–13.6% in cirrhotics, and of 1.4–4.6% in patients with advanced fibrosis. Cumulative incidences (CumI) for each study are reported in Table 3.

Table 2 Characteristics of Studies Reporting Data on HCC Occurrence (de-novo HCC) in Patients with Cirrhosis or Advanced Fibrosis

Table 3 Incidence and Risk Factors of de-novo HCC in Patients with Cirrhosis or Advanced Fibrosis

Table 4 Characteristics of Studies Reporting Data on HCC Occurrence (de-novo HCC) in Patients with Chronic Hepatitis C (Any Fibrosis Stage)

Table 5 Incidence of de-novo HCC and Factors Associated with HCC Occurrence (de-novo HCC) in Patients with Chronic Hepatitis C (Any Fibrosis Stage) from 21 Studies

Severity of Liver Disease

In the setting of cirrhosis, severity of liver disease was identified among the most important predictors of de-novo HCC, and was assessed either through non-invasive tests for fibrosis staging or clinically.

Several studies reported an association between de-novo HCC and LSM, mainly assessed by transient elastography (TE). LSM was correlated with HCC occurrence either when analysed at a single time-point [mostly pre-DAA (baseline)], or as a dynamic variable, by evaluating changes in LSM values between pre- and post-treatment. Thresholds able to discriminate patients at different risk of HCC development varied according to studies (Table 6). For example, Degasperi et al found that the 3-year probability of HCC significantly increased in cirrhotics with baseline LSM values >30 kPa, while Rinaldi et al used the 20 kPa and 30 kPa thresholds^22,27 (Table 6). In addition, Ravaioli et al reported an increased risk of de-novo HCC in patients with a <30% decrease in LSM values, between baseline and the end of treatment (EOT).²⁵

Table 6 Cumulative Incidence (CumI) of HCC According to Values of Clinically Significant Variables

In F3-F4 patients, one European study reported that high baseline TE values as well as changes in LSM (ΔLSM) one-year after EOT were associated with an increased risk of de-novo HCC. Pre-treatment LSM >17.3 kPa and ΔLSM >25.5% were finally included in a predictive model (see below)⁴⁵ (Table 7). These results were not confirmed by Pons et al, reporting that the risk of de-novo HCC in cACLD patients was independent of LSM improvement, either when using the 30% or 20% decline cut-offs. Conversely, the risk of de-novo HCC was increased by LSM values >10 kPa one-year after EOT.⁴⁶

Table 7 Studies Reporting Predictive Scores for de-novo HCC

Among NITs, Fibrosis-4 Index (FIB-4) was the most used to assess fibrosis severity. Baseline FIB-4 emerged as an independent risk factor for de-novo HCC in some studies analysing cirrhotic patients,^22,26,40 although different cut-offs were identified (Table 6). Degasperi et al reported a significantly higher 3-year de-novo HCC incidence in patients with baseline FIB-4 >9, while Ide et al identified the alternative 4.6 cut-off, that was therefore incorporated in a composite predictive score (see below)⁴⁰ (Table 7). In F3-F4 patients, Alonso Lopez et al found that both baseline FIB-4 >3.7 and FIB-4 >3.3 one year after treatment were associated with de-novo HCC⁴⁵ (Table 7). Other NITs had been investigated as predictors of HCC in several studies: the albumin-bilirubin (ALBI) score grade 2–3^29,37,42 and AST to platelet (PLT) ratio index (APRI) >2.5¹¹ emerged as independent risk factors for de-novo HCC in patients with cirrhosis and advanced fibrosis, respectively.

The risk of de-novo HCC was also increased in cirrhotic patients with clinical features of more advanced liver disease, irrespective of LSM and/or NIT values. The presence of portal hypertension (PH) was an independent predictor of HCC in several studies, although definition of PH was heterogeneous. Ogawa et al defined PH by either LSM values (≥20 kPa) or hepatic venous pressure gradient (HVPG; >10 mmHg), or by imaging.³⁸ Thus, among indirect markers of PH were both biochemical tests and clinical features. For example, albumin^{20,31,37,44,45} and PLT^20,21,27,29 were independently associated with HCC occurrence, either as single predictors or when included in predictive scores (Tables 6 and 7). The presence of esophageal varices (EV)²⁸ or ascites was associated with an increased the risk of de-novo HCC.³²

Clinical scores incorporating these parameters, such as Child-Pugh-Turcotte (CPT)^25,36 and Model for End Stage Liver Disease (MELD)⁴¹ or ALBI scores^29,37,42 were independently associated with de-novo HCC in cirrhotics (Table 7). Some studies provided different CumI of de-novo HCC according to the combination of one or more variables associated with portal hypertension (see below).

Patient-Related Factors

Several patients’ features, either modifiable or not modifiable, have been shown to increase the risk of de-novo HCC after the achievement of an SVR. Age was independently associated with HCC occurrence in most studies^{21,27,28,40–42,44} as well as male gender^{22,26,31,36,40,44} (Tables 3 and 5). In US cohorts, a role of non-African American ethnicity has been suggested,^34,41 although this association deserves further confirmation. Among comorbidities, diabetes mellitus (DM) has been associated with an increased risk of HCC in several cohorts;^{22,26,31,36,37} Abe et al incorporated DM status in a multivariable HCC risk score (see below) (Table 7). Other factors such as alcohol consumption,^21,41 and viral co-infections¹¹ are likely to influence post-SVR HCC risk, although these patients were systematically excluded from most clinical trials.

Genetic Predictors

A single-center study conducted in a large cohort of DAA-treated cirrhotic patients found that a genetic risk score combining 4 single nucleotide polymorphisms (SNPs) [PNPLA3, TM6SF2, MBOAT7 and GCKR] was an independent predictor of de-novo HCC, together with other clinical predictors (DM, male gender, albumin values)³¹ (Tables 3 and 6). The same authors found that the tolloid-like 1 (TLL1) gene, which had been previously associated with HCC occurrence in Japanese CHC patients,⁴⁷ did not predict de-novo HCC in 348 European cirrhotics.²⁶

Virus-Related Factors

Two studies, only, reported that HCV genotype might influence the risk of HCC during follow-up. Nahon et al found that HCV-1 patients were at increased risk of HCC development,²¹ whilst genotype 3 was independently associated with de-novo HCC in a large retrospective study from US⁴¹ (Table 3).

Alpha-Fetoprotein (αFP)

Although not universally recommended for HCC surveillance by international guidelines due to its low sensitivity and specificity, broad application of αFP in routine clinical practice has led many authors to investigate its potential for de-novo HCC prediction. αFP was independently associated with HCC occurrence, either in patients with cirrhosis^32,38,42 or advanced fibrosis.⁴⁴ Some studies evaluated the predictive ability of αFP assessed at baseline,^32,42,44 while others analysed the EOT time-point.³⁸ Most studies tried to identify a predictive αFP cut-off: overall, the proposed cut-offs resulted always higher than the reference standard 7 ng/mL (ie, >9 or ≥10 ng/mL, >20 mg/mL)^38,42,44 (Tables 3 and 6).

Undefined Nodules

Sangiovanni et al found that the presence of undefined/non-malignant nodules at baseline was an independent predictor of HCC occurrence in cirrhotic patients.³² Partially in line with this finding is what reported by Tamaki et al, as they found that Li-Rads 3/4 nodules were independently associated with HCC occurrence in CHC patients (6.9% cirrhotics).⁴⁸ To avoid biases related to inclusion of patients carrying nodules at risk of HCC transformation, presence of undefined nodules was declared to be an exclusion criterion in some studies^{10,22,26,31,36} (Table 3).

Lack of a Sustained Virological Response

Some studies including large cohorts of treated patients did not allow separate analysis of those achieving an SVR, thus leading to include non-SVR among potential predictors of de-novo HCC. Although the statistical power when analysing the influence of non-SVR status on HCC risk, some authors reported that the lack of an SVR was associated to increased HCC occurrence^{20,28,34,35,39} (Table 3).

Combined Predictors and HCC Risk Scores

The risk of de-novo HCC increased when two or more independent predictors identified at multivariable analysis were combined. Not surprisingly, in all cases HCC cumulative incidences (CumI) proportionally increased according to the number of risk factors considered.^{11,20,22,28,46} These studies mostly included parameters associated with liver disease severity (LSM, APRI, CPT score, PLT, albumin), DM and SVR status. Conversely, other studies evaluated composite HCC risk scores, which were based on combinations of multiple variables, to stratify patients into different HCC-risk classes. Four studies focused only on patients with advanced fibrosis or cirrhosis, by proposing a combination of patient-related (age, gender, presence of DM) and biochemical variables (albumin, γGT, PLT, αFP) together with data related to liver disease severity.^33,37,44,45 The aMAP score failed in predicting de-novo HCC in 2085 F3-F4 patients with HCV-4,⁴⁹ and, similarly, GES score performance was suboptimal in a Caucasian cohort.³⁰ Cumulative incidences of de-novo HCC according to different risk classes are reported in Table 7.

Studies Enrolling Patients with Chronic Hepatitis C (Any Fibrosis Stage)

Twenty-three studies reported data on HCC occurrence in CHC patients with any stage of liver fibrosis (Table 3). Almost all these studies included cirrhotic patients; rates of cirrhotics ranged between 6% and 73%, although some authors did not provide this information. Most studies included only patients with an SVR, whilst data could be extrapolated from three studies.^34,42,50 In studies including non-SVR patients, treatment failure accounted for 3.0–6.7% of DAA treatment responses (Table 4).

Overall, 0.9% to 6.9% of CHC patients developed de-novo HCC during follow-up, although only few authors reported the prevalence of cirrhosis in CHC cohorts. When reported, rates of cirrhosis were between 38% and 100% in CHC patients developing HCC^{35,39,40,51–53} (Table 4), and overall CumI of de-novo HCC were lower than that reported in cirrhotic cohorts, at each time-point (Tables 2 and 4).

Severity of Liver Disease

Due to the inclusion of cirrhotic patients in CHC cohorts, liver disease severity was independently associated with HCC occurrence in most studies. Only 5 studies were able to identify HCC predictors in non-cirrhotic F0-F3 patients,^{34,37,38,40,42} and three of them included indirect markers of fibrosis, either biochemical tests or NITs (see below) (Table 5).

Whatever defined, cirrhosis and advanced fibrosis were independently associated with HCC development in several studies.^{34,38,39,54,55} In these studies, cirrhosis was differently defined (Table 1), and ranged between 18% and 56% of the overall population. Particularly, the CHC cohort described by Pinero et al included 399 (29%) patients with clinically significant portal hypertension (CSPH) (Table 4).

Baseline LSM obtained by Acoustic Radiation Force Impulse (ARFI) or TE, was associated with an increased risk of post-treatment HCC (Table 5) in three studies analysing SVR patients.^53,56,57 Tachi et al identified the 1.73 m/s threshold as the optimal cut-off to stratify CHC patients according to their de-novo HCC risk⁵⁶ (Table 5). In the other two studies, baseline LSM ≥20 kPa and ≥17.5 kPa were associated with HCC occurrence in 398 and 773 CHC patients from Japan and Denmark, respectively. In both studies, the prevalence of cirrhosis was not reported.^53,57 Ogasawara et al reported that also SVR24-LSM (ie, LSM performed 24 weeks after EOT) ≥10 kPa was independently associated with de-novo HCC.⁵⁷ This finding was in line with two other Japanese studies, identifying in LSM ≥3.75 kPa obtained through Magnetic Resonance Elastography (MRE) 12 weeks after treatment completion (SVR12-MRE) an independent predictor of HCC occurrence.^48,58

Most studies reported that either baseline or post-treatment FIB-4 values were associated with the risk of de-novo HCC in CHC patients. In 5 studies, baseline FIB-4 was reported to be an independent risk factor for HCC occurrence.^{34,40,47,50,59} Risk thresholds varied according to each study: Kanwal et al used the standard 3.25 cut-off,³⁴ whereas Iio et al used the 2.67 cut-off.⁴⁷ In the study by Watanabe et al, baseline FIB-4 predicted HCC in females, only.⁵⁹ In addition, three Japanese studies reported that post-SVR FIB-4 (at EOT and at SVR12) and changes in FIB-4 independently predicted de-novo HCC.^50,55,60 Among investigated serological biomarkers of fibrosis was Wisteria floribunda agglutinin positive Mac-2 (WFA*M2BP), which was tested in the study by Nagata et al, reporting that WFA*M2BP assessed 24 weeks after EOT independently predicted de-novo HCC⁵⁵ (Tables 5 and 6).

Albumin and PLT were independently associated with HCC occurrence in two studies reporting data on F0-F3 patients.^37,38 Also, Watanabe et al found that low pre-treatment albumin values (<3.8 g/dl) increased the risk of HCC⁴⁹ (Tables 5–7).

Patient-Related Factors

Male gender and age independently predicted HCC occurrence also in CHC cohorts. Male gender was associated with de-novo HCC in 5^{40,50,55,59,60} studies, and age in 7^{40,41,48,52,53,55,58} (Table 5). Although age was analysed as a continuous variable in multivariate analysis, different cut-offs were associated with increased risks of HCC (>60, >62, >75 years). Race still emerged as independent predictor of de-novo HCC in the large US cohorts,^34,41 but was not confirmed by other studies. Co-morbidities influenced HCC development also in CHC cohorts: Kanwal et al reported that the presence of DM was independently associated with HCC occurrence,³⁴ whereas other studies found that altered γGT and ALT values predicted post-SVR HCC,^40,55 likely mirroring the presence of underlying metabolic disorders (Tables 5 and 6).

Genetic Predictors

In CHC patients, two genetic factors were independently associated with HCC occurrence. Nagata et al found that IL28B rs8099917 polymorphism (non-TT) was associated with an increased risk of HCC in a large cohort of 752 patients followed-up for 1.8 years.⁵⁴ Conversely, the Japanese study by Iio et al reported that patients carrying the TLL1 rs17047200 AT/TT genotypes had significantly higher CumI of HCC, although T allele was associated with lower PLT and higher FIB-4 values.⁴⁷ In 348 F0-F3 patients from Italy, TLL1 genotype did influence HCC risk²⁶ (Table 5).

Alpha-Fetoprotein

Baseline αFP was independently associated with de-novo HCC in 4 studies, which however identified different cut-offs: >4.6 ng/mL ≥8 ng/mL and ≥10 ng/mL^42,47,55,57 (Table 5). In addition, some authors investigated the predictive values of post-treatment αFP (Table 5): values at both EOT,^50–52,59 and SVR12^48,58,60 time-points were associated with de-novo HCC. At SVR12, the following cut-offs were identified: >5 ng/mL, >6.5 ng/mL, ≥6 ng/mL. Interestingly, Watanabe et al proposed two different cut-offs (ROC analysis) for post-treatment αFP according to patient gender: >6.0 ng/mL in females and >3.5 ng/mL in males, respectively⁵⁹ (Tables 5 and 6).

Virus-Related Factors

The only study reporting a role of virus-related factors is the one by Kanwal et al, finding an association between HCV genotype 3 and de-novo HCC.⁴¹

Combined Scores

In the setting of CHC patients, 4 studies developed scores based on multiple variables to predict de-novo HCC, mostly assessed at EOT or SVR time-points. Hiraoka et al proposed the ADRES score, based on the combination of gender, FIB-4 and αFP assessed at SVR24, while Tani et al incorporated EOT-αFP (>6 ng/mL) and age (>75 years)^52,60 (Table 7). Iio et al combined SVR24- αFP and FIB-4 with the TLL1 genotype,⁴⁷ while Watanabe et al proposed two different models, either pre-DAA (including FIB-4, albumin and gender) or post-DAA (incorporating EOT- FIB-4 and αFP values)⁵⁰ (Table 7).

Predictors of Recurrent HCC

The risk of HCC following antiviral treatment was strongly influenced by previous HCC history. Not only rates of recurrent HCC were significantly higher than those of de-novo HCC (Tables 2, 4, 5 and 8), but previous HCC history was the strongest predictor of HCC development in cohorts analysing cumulative data from patients with and without pre-DAA liver cancer. Rates of HCC recurrence following DAA were similar^18,19,23,61 or even lower⁶² than those reported in untreated patients, and most authors reported that oral antivirals did not enhance the risk of recurrence.^{12,15,18,19,23,61}

Table 8 Characteristics of Studies Reporting Data on HCC Recurrence

Rates of HCC recurrence ranged between 3.2% and 49%, with one study only⁶³ reporting no recurrence however among 8 patients (7 with cirrhosis) (Table 8). Almost all studies tried to identify clinical predictors of HCC recurrence during variable follow-up, despite the inclusion of patients with different characteristics, including tumour burden (Tables 8 and 9). In two retrospective French studies, the authors found HCC predictors different from DAA use when comparing untreated vs DAA-treated patients.^23,61 Similarly, Ogawa et al were not able to identify predictive factors of HCC recurrence in 62 F0-F3 patients with an SVR, whilst other studies on CHC patients did not focused on the sub-group of non-cirrhotics.³⁸ In most cases, data were obtained from cohorts including both SVR and non-SVR patients; 4 studies enrolled only cured patients, whilst this information was lacking in other 4. When reported, rates of treatment failure ranged between 2.5% and 25% (Table 8). In 12 studies all patients had a diagnosis of cirrhosis, and most of them (n=9) included also decompensated (CPT-B) patients (Table 8). In studies enrolling CHC patients, rates of cirrhosis ranged between 33% and 95% (n=9) or were not reported (n=4), and only few authors (n=4) reported information on fibrosis stage in patients with a complete response (CR) to previous HCC who subsequently developed HCC recurrence (Table 8).

Table 9 Incidence and Factors Associated with HCC Recurrence

Severity of Liver Disease

Despite the inclusion of cirrhotic patients in CHC studies, only one of them was able to identify cirrhosis as an independent risk factor for HCC recurrence.³⁸ Although not reported in most cases, expected high rates of cirrhosis in patients developing recurrent HCC might have attenuated the weight of this variable. However, further reinforcing the strength of liver disease severity as HCC predictor, some authors found that indirect markers of fibrosis were independently associated with HCC recurrence. For example, Conti et al reported that baseline LSM independently predicted HCC in 59 cirrhotics followed-up for 24 weeks,⁸ whereas Nagata et al found that WFA*M2BP assessed at SVR24 predicted HCC recurrence in 83 CHC patients⁵⁴ (Table 9).

Patient-Related Factors

According to published studies, patients’ characteristics had low impact on HCC recurrence, as only few authors found that they were independently associated with recurrent HCC following an SVR to DAA. However, both age^8,64 and, in cirrhotic patients, comorbidities such as DM^22,31 and alcohol³² seemed to play a role in influencing HCC risk. Moreover, Degasperi et al reported that ethnicity (ie, Egyptian vs Italian) was an independent risk factor for HCC recurrence in their European cohort²² (Tables 6 and 9).

Tumour Burden

Rates of HCC recurrence were strongly influenced by tumour burden in most studies analysing either F4 or CHC cohorts (Table 9). One of the most important predictors of HCC recurrence was history of HCC recurrence before DAA^8,13,32 together with the number of HCC treatments finally leading to CR achievement before anti-HCV therapies.^62,65,66 In addition, time elapsing between prior HCC treatment and DAA start was significantly associated with an increased risk of recurrent HCC in several studies,^38,64,67,68 where patients treated for HCC less than one year prior to DAA exhibited an increased risk of tumour recurrence.^38,68 Lastly, some authors reported that also prior HCC size,¹³ number of nodules³⁸ and type of HCC treatment (ie, palliative vs curative)^38,67 were independently associated with HCC recurrence, although these data were not confirmed by others^8,22,65 (Table 6). However, these results should be cautiously interpreted, as they are strongly influenced by study design and patients enrollment; recently, an individual patient-data meta-analysis pooling data of 977 patients from 21 studies have further enhanced the importance of pre-DAA HCC history and tumour burden.¹⁹

HCC Biomarkers

Four studies found that higher baseline (DAA start) values of αFP were independently associated with HCC recurrence.^19,28,54,66 Casadei-Gardini and others found that aspartate aminotransferase to lymphocyte ratio (ALRI), which had been previously proposed for inclusion in HCC surveillance algorithms,⁶⁹ independently predicted HCC recurrence in 98 cirrhotic patients (73% CPT-A) treated with DAA 8.5 months after CR²⁹ (Tables 6 and 9).

Conclusions

Despite the expected decrease in HCC burden,⁷⁰ the widespread use of DAA to cure HCV infection will finally lead large cohorts of SVR patients to be maintained under surveillance. In fact, the number of patients requiring HCC surveillance due to pre-treatment advanced fibrosis is expected to increase over time, as a consequence of worldwide diffusion of HCV screening and treatment programs.⁷¹ Therefore, we are going to face with larger, ageing population still at risk of HCC, although HCC risk is lower than that reported in active HCV infection. As a consequence, the investigation of HCC predictors is of paramount importance in order to better optimize surveillance strategies, with the ultimate goal of personalized follow-up algorithms. While advanced fibrosis and cirrhosis represent strong predictors of HCC development, either de-novo or recurrent, literature data suggest that many co-factors may contribute to the oncogenic risk. While some of these factors are modifiable or can be potentially improved by successful antiviral treatments (fibrosis, portal hypertension), others are only partially modifiable (metabolic syndrome) or not modifiable at all (aging, HCC history). Due to the complex interactions and competing risks resulting from these variables, combination analyses or composite scores are those expected to better improve prediction capability, with all the challenges related to large-scale applicability in heterogeneous patient populations. Therefore, in most cases prospective validation in larger cohorts is still needed.