Hepatocellular carcinoma (HCC) is the most serious and dreaded complication of chronic liver disease and the most frequent cause of death in patients with compensated cirrhosis.1 A significant increase in its incidence has been reported in the United States in recent years, and this may be explained by a substantial increase in hepatitis C virus (HCV)–related HCC.2 Therefore, the identification of factors associated with tumor development is extremely important for the timely referral of patients eligible for curative treatment. Several years ago, we reported early results of a prospective cohort study showing that HCV genotype 1b was an independent risk factor for HCC in patients with cirrhosis.3 Surprisingly, despite a number of confirmatory studies carried out in different populations,4–6 no consensus has emerged yet on this matter, which is still controversial.7 Thus, no specific recommendation on screening for the HCV genotype has been issued in the most recently published American Association for the Study of Liver Diseases practice guidelines.8 Consequently, in an attempt to definitely assess the risk of HCC according to the HCV genotype, we re-examined the same cohort of patients, who were prospectively followed up for up to 17 years.
Hepatocellular carcinoma (HCC) is the most frequent cause of death in patients with hepatitis C virus (HCV)–induced cirrhosis. Despite a number of studies in different populations worldwide suggesting an association between HCV genotype 1 and the risk of HCC, no consensus has emerged yet on this matter, which is still controversial. In an attempt to clarify this issue, a prospective study of 163 consecutive HCV-positive patients with cirrhosis, who were enrolled between January 1989 and December 1990, was carried out. HCC occurrence was detected by ultrasound surveillance every 6 months. Independent predictors of HCC were assessed with a Cox regression analysis. After a median follow-up of 10.7 years, 44 [4.26/100/year, confidence interval (CI) = 3.11–5.68/100/year] of 104 patients infected with genotype 1b developed HCC versus 10 (1.69/100/year, CI = 0.82–3.09/100/year) of 52 patients infected with genotype 2a/c (P = 0.0001). Multivariate analysis showed that HCV genotype 1b was independently associated with HCC development [hazard ratio (HR) = 3.02, 95% CI = 1.40–6.53]. Other predictors of HCC were esophageal varices (HR = 2.15, 95% CI = 1.03–4.47), male gender (HR = 2.12, 95% CI = 1.10–4.11), and age over 60 years (HR = 5.96, 95% CI = 1.23–28.8). Conclusion: HCV genotype 1b is associated with a statistically significant higher risk of developing HCC. Patients with cirrhosis that are infected with this genotype require more intensive surveillance for the early detection and aggressive management of neoplasia. (HEPATOLOGY 2007.)
Patients and Methods
Between January 1989 and December 1990, all patients with cirrhosis (n = 163) attending the outpatient clinic at San Paolo Hospital in Milan, Italy, were enrolled in a prospective study aimed at the early recognition of HCC. Informed consent was obtained from all patients, and the study protocol was approved by the hospital ethics committee. A database including demographic, clinical, laboratory, and endoscopic examinations obtained at entry was set up. In addition, blood samples from all patients were frozen (−80°) for further analysis. Criteria for inclusion and for surveillance follow-up were reported in the original study.3 The diagnosis of HCC was based on a histological assessment obtained by a fine-needle liver biopsy whenever a focal liver lesion was detected by ultrasound and was subsequently modified according to the guidelines established by the Barcelona conference.9
Baseline demographic, clinical, biochemical, and virological features of the patients are reported in Table 1. In our previous report,3 the Child-Pugh class was determined according to the classification, which included the nutritional status and prothrombin time expressed as a percentage. In this report, we classify patients on the basis of the updated Child-Pugh score, using the information collected at the baseline (the international normalized ratio, bilirubin, albumin, and presence of ascites or encephalopathy).
|Baseline Characteristic||Number of Patients||HCV Genotype||P for 1b versus 2a/c||HCC|
|1b||2a/c||Other*||Number||Rate per 100 per year||Univariate HR (95% CI)||Multivariate HR (95% CI)||Multivariate Excluding SVR Patients|
|All subjects||163 (100%)||104||52||7||55||3.27|
|1b||104 (64%)||104||—||—||44||4.26||2.61 (1.31–5.18)||3.02 (1.40–6.53)||2.87 (1.27–6.47)|
|Other||7 (4%)||—||—||7||1||1.72||0.99 (0.13–7.77)||2.97 (0.30–29.3)||—|
|50–59||69 (42%)||40||28||P trend||25||3.46||4.17 (0.99–17.6)||6.62 (1.39–31.6)||5.29(1.17–23.9)|
|60+||73 (45%)||48||22||0.20||28||3.87||4.65 (1.11–19.5)||5.96 (1.23–28.8)||4.73 (1.01–22.1)|
|Male||80 (49%)||52||23||5||0.61||31||3.92||1.45 (0.85–2.47)||2.12 (1.10–4.11)||2.03 (1.02–4.04)|
|B||28 (17%)||12||13||3||0.04||9||5.17||1.95 (0.95–4.04)||1.52 (0.37–6.17)||1.90 (0.43–8.52)|
|Yes||46 (28%)||30||15||2||1.00||18||5.52||2.43 (1.36–4.34)||2.15 (1.03–4.47)||1.96 (0.86–4.51)|
|Yes||20 (13%)||18||1||1||0.004||8||5.33||1.90 (0.89–4.05)||1.04 (0.45–2.45)||1.08 (0.45–2.55)|
|Yes||42 (26%)||25||14||3||0.70||17||4.14||1.40 (0.79–2.49)||0.91 (0.46–1.79)||0.85 (0.40–1.79)|
|≥80,000/mL||121 (74%)||80||35||6||0.25||9||4.25||1.49 (0.72–3.09)||0.97 (0.39–2.43)||1.02 (0.39–2.65)|
|>3.5 mg/dL||21 (13%)||93||41||6||47||3.09||1.00||1.00||1.00|
|≤3.5 mg/dL||140 (86%)||9||11||1||0.04||7||4.73||1.71 (0.77–3.81)||1.12 (0.28–4.38)||1.05 (0.25–4.32)|
|≤80%||79 (48%)||54||23||2||0.31||31||4.16||1.74 (1.01–2.99)||1.27 (0.65–2.50)||1.07 (0.54–2.12)|
|<1.2 mg/dL||111 (68%)||71||36||4||37||3.10||1.00||1.00||1.00|
|≥1.2 mg/dL||50 (31%)||31||16||3||1.00||17||3.56||1.17 (0.66–2.09)||0.80 (0.36–1.77)||0.70 (0.30–1.64)|
|<10 ng/mL||111 (68%)||66||39||6||34||2.81||1.00||1.00||1.00|
|≥10 ng/mL||50 (31%)||37||12||1||0.14||21||4.75||1.76 (1.02–3.04)||1.67 (0.92–3.05)||1.71 (0.92–3.20)|
|Yes||86 (53%)||53||31||2||0.39||29||3.36||1.06 (0.62–1.81)||0.88 (0.45–1.74)||0.87 (0.43–1.75)|
|Nonresponder||73 (45%)||54||16||3||20||2.89||0.61 (0.35–1.08)||0.75 (0.39–1.44)||0.72 (0.37–1.39)|
|SVR||12 (7%)||4||7||1||0.01||3||1.66||0.65 (0.19–2.20)||0.61 (0.16–2.37)||—|
|Yes||7 (4%)||3||2||2||0.75||1||2.27||0.44 (0.06–3.30)||1.02 (0.12–8.92)||1.33 (0.16–10.9)|
HCV RNA was detected by a nested reverse-transcription polymerase chain reaction on frozen sera with conserved primers localized in the 5′ noncoding region of the viral genome.10 In our previous report, genotyping was performed with a modified form of Okomoto's method,10 which used type-specific primers of the core region. As this method allowed us to assess only a limited number of viral genotypes, all cases were later re-evaluated with either frozen-stored or fresh serum samples with INNO-LiPA HCV II (Bayer Corp., Tarrytown, NY) for sequences in the 5′ noncoding region. This report is based on the revised/updated genotyping information. In the patients treated with interferon (IFN) or IFN and ribavirin (RBV), frozen sera collected during and after treatment cessation were tested for HCV RNA by a qualitative polymerase chain reaction assay. The sustained virological response (SVR) was defined as undetectable serum HCV RNA (<50 IU/mL) in patients who had persistently normalized aspartate aminotransferase and alanine aminotransferase values 24 weeks after IFN discontinuation. In patients achieving an SVR, qualitative HCV-RNA determinations were repeated yearly during follow-up until the last visit.
We used Fisher's exact test and the Mantel-Haenszel chi-square test for trends to assess the difference in the baseline characteristics of patients with various HCV genotypes. Because the progression of the disease is slow and mortality (liver-related and non–liver-related) from intercurrent events is important, the cumulative incidence of HCC according to the genotype was determined with consideration of competing risk due to death and compared by the method of Pepe and Mori.11 This method prevents the overestimation of the probability of failure in the presence of competing risk. We also calculated the incidence rates of HCC by dividing the number of patients who developed HCC by the number of person years at risk. We used univariate and multivariate Cox proportional hazards regression to identify baseline characteristics (age, gender, HCV genotype, Child class, presence of varices, past alcohol consumption, diabetes, laboratory values, IFN monotherapy, and IFN-RBV defined as a time-dependent covariable) associated with the development of HCC. Analyses were performed with SAS software (SAS Institute, Inc., Cary, NC). All tests were 2-sided.
One hundred four patients (64%) were infected with genotype 1b, 52 (32%) were infected with genotype 2a/c, 1 was infected with genotype 1a, and 2 were infected with genotype 3a. In 4 patients, the HCV genotype could not be assigned, whereas the new INNO-LiPA test allowed a proper classification of the few patients with mixed or untypeable genotypes, as reported in our previous study. The median age at entry was 58.7 years in patients infected with genotype 1b and 59.5 years in patients infected with genotype 2a/c (P = 0.31). Overall, 38 patients received blood transfusions, but the date of the presumed infection was known for only 32 patients. The median interval between the transfusion and enrollment in the study was 14.5 years in genotype 1b patients (n = 21), 15.5 years in genotype 2a/c patients (n = 9; P = 0.70), and 17.4 years in patients with other genotypes (n = 2).
A higher proportion of patients with genotype 2a/c than patients with genotype 1b had Child-Pugh class B at enrollment (25% versus 12%, P = 0.038) or had a low albumin level (≤3.5 mg/dL; 21% versus 9%, P = 0.042), whereas diabetes was significantly more frequent in patients infected with genotype 1b than in those infected with genotype 2a/c (17% versus 2%, P = 0.004; Table 1). All other baseline characteristics were similar in the two genotype subgroups.
Eighty-two (51%) of the 163 patients (79 in Child class A and 3 in Child class B) were treated with 6 MU of standard IFN-α 3 times a week for a minimum of 6 months before the HCV genotype was determined. Those who showed a complete biochemical response were treated for another 6 months with a dose of 3 MU 3 times a week. The timing of the treatment was similar for patients who eventually did or did not develop HCC. Only a minority (n = 12) of treated patients achieved SVR, with a significant difference between genotypes (7% in genotype 1b versus 30% in genotype 2a/c, P = 0.01; Table 1).
In Italy, RBV became available for expanded access only in 1998, and the Italian Health Service did not allow the retreatment of non-SVR patients with cirrhosis until 2003. After that date, a combined IFN-RBV retreatment was occasionally given because some of our patients were already in Child class B at enrollment, and others developed decompensation or progressed to Child class B during follow-up. Moreover, 45% of the patients at enrollment were older than 60 years, 28% had varices, and 17% had a platelet number lower than 80,000 mm3. As a result, only 7 patients were eligible for retreatment, regardless of the genotype, and received IFN-RBV therapy (Table 1).
After a median follow-up of 10.7 years (range = 0.6–17.2), 55 patients developed HCC at an annual incidence rate of 3.27/100/year, and 85 patients (52%) died (42 after the development of HCC). Eighteen (12%) patients were lost to follow-up, with no difference among genotypes (genotype-1b, 12/104 = 11.5%, and genotype-2a/c, 4/52 = 7.7%, P = 0.58; missing/other genotypes, 2/7 = 28.6%).
Moreover, the proportion of patients lost to follow-up or who died in the absence of HCC was similar in the genotype 1b and genotype 2a/c subgroups (Fig. 1).
The cumulative incidence of HCC was significantly higher in patients with genotype 1b than in patients with genotype 2a/c (P = 0.0001; Fig. 1), with incidence rates of 4.26 and 1.69/100/year, respectively (Table 1). In a univariate analysis, HCV genotype 1b, older age, esophageal varices, a low prothrombin time, and elevated α-fetoprotein were associated with significantly increased risks of HCC, whereas IFN therapy was associated with a decreased risk of HCC attributable to the more favorable clinical conditions of patients who were offered treatment (Table 1). In a multivariate analysis, HCV genotype 1b, older age [hazard ratio (HR) = 5.96, 95% confidence interval (CI) = 1.23–28.8], male gender (HR = 2.12, 95% CI = 1.10–4.11), and esophageal varices (HR = 2.15, 95% CI = 1.03–4.47) remained independent predictors of HCC, regardless of IFN treatment and of the achievement of SVR (Table 1). Compared with patients with genotype 2a/c, patients with genotype 1b carried a greater than 3-fold risk of developing HCC (HR = 3.02, CI = 1.40–6.53; Table 1)
In order to completely remove the potential confounding effect of SVR, a separate multivariate analysis excluding the 12 patients who had achieved SVR was also provided. In this analysis, genotype 1 remained independently associated with HCC development (HR = 2.87, CI = 1.27–6.47; Table 1).
Interestingly, the rate of HCC among patients with genotype 2a/c was similar between SVR (1.9/100/year) and non-SVR (1.4/100/year) patients, whereas among those with genotype 1b, the rate of HCC was higher in non-SVR patients (3.1/100/year), but none of the 4 patients who achieved SVR developed HCC. Overall, the rate of HCC was lower in SVR treated patients (1.66/100/year) than in non-SVR treated patients (2.89/100/year).
Because the development of HCC could have been the reason for dropout in the subset of patients lost to follow-up, we also performed a conservative analysis that assumed all patients infected with genotype 2a/c who were lost to follow-up developed HCC, whereas no one with genotype 1b did. In this analysis, genotype 1b remained significantly associated with HCC development (HR = 1.93, 95% CI = 1.00–3.72) after adjustments for the age, sex, Child-Pugh class, presence of varices at the baseline, alcohol, bilirubin, alpha-fetoprotein, and IFN treatment.
Finally, we performed a separate analysis, considering only the period and the events that occurred after our initial report and the end of the study.3 On January 1, 1996, 127 patients were still alive, free of HCC, and on follow-up. Eighty-three patients (65%) were infected with genotype 1b, 41 (32%) were infected with genotype 2a/c, and 3 were infected with other genotypes. Since then, 33 patients developed HCC. Again, the cumulative incidence of HCC was significantly higher in patients with genotype 1b (n = 25) than in patients with genotype 2a/c (n = 7; P = 0.020; Fig. 2).
This prospective study, which remains unique in the field because of the length of our follow-up and the large number of events available for the analysis, definitively demonstrates that patients with genotype 1b have a substantially higher risk of developing HCC than those with genotype 2a/c. The short enrollment period, the use of standardized criteria to classify events, the extended follow-up time ensuring outcome occurrence, the accuracy in measuring confounding factors, the adequate size of the population studied, and the limited number of patients lost to follow-up represent the strengths of this study, which complies with the criteria of appropriately designed observational studies evaluating disease risk factors.12
In recent years, the role of genotype 1b in increasing the risk of HCC has been questioned, even though several studies have described this association.7 The main concern was that this result could have been attributed to a cohort effect generated by older individuals who had been infected at a time when HCV genotype 1b was most prevalent.13 In this study, patients infected with genotype 2a/c had more advanced liver disease at study entry; moreover, the evidence that age was similar in patients infected with either genotype and that the elapsed time between transfusion and enrollment into the study was identical in patients with a history of blood transfusion suggests that the presumed duration of infection was similar between the 2 groups. These data argue against a cohort effect and lower the possibility that the deterioration of liver function may be advocated as a confounding factor in this analysis. In addition, it is well known that these 2 genotypes were simultaneously introduced into Italy several years ago,14 most likely between World War II and the late sixties, when genotypes 1b and 2a/c were the only prevalent ones. This also explains the lack of other genotypes in our study population. In fact, it has been reported that the epidemiological spread of other genotypes in Italy emerged in the early seventies, and this was associated with intravenous drug use for genotypes 1a and 3a15 and with immigration for genotype 4.16 Only 1 of our patients was reported as an intravenous drug user, whereas immigrants were exceedingly rare in our country during the recruitment period.
Another major concern was that mortality associated with intercurrent events (both liver-related and non–liver-related) may have been higher in genotype 2a/c and, therefore, have affected the reliability of our results. In order to rule out this possibility, we plotted and compared the cumulative incidence of HCC in the various genotype groups, using competing risk analysis methods.
A number of studies conceived to assess the response to IFN rather than to evaluate the outcome of the disease17, 18 had suggested that treatment with IFN was associated with a lower risk of HCC development. In our study, which was not randomized and not specifically designed to address this question, no difference in the incidence of HCC development was seen between treated and untreated subjects (regardless of genotype) when confounding was removed by a multivariate analysis. This confirms that IFN treatment, when failing to lead to an SVR, has no effect on HCC occurrence, as previously suggested.19
In contrast, it is now recognized that SVR patients have a reduced incidence of HCC in comparison with non-SVR patients, at least during the first 8 years following the discontinuation of treatment.20 Moreover, it is known that even at the stage of cirrhosis, patients with genotype 2a/c respond better to IFN therapy.21 Accordingly, in our study, it could be argued that patients with HCV genotype 2a/c either may have been too advanced in their disease to receive treatment or may have been cured from infection and therefore protected against HCC. In this study, IFN therapy was administered before any information regarding genotypes was available. Only a minority of treated patients achieved SVR, and only a few of the non-SVR subjects later received IFN-RBV. Still, in order to minimize the potential confounding effect of SVR, we adjusted the estimates for SVR achievement, and we also provided a separate multivariate analysis that excluded all patients who had achieved SVR. Both analyses confirmed that genotype 1b was independently associated with the development of HCC. Of interest, although the overall rate of HCC was lower in SVR treated patients than in non-SVR treated patients, and this is in agreement with results reported for a large retrospective study in Italy,20 the rate of HCC was similar among SVR and non-SVR genotype 2a/c patients but higher in non-SVR genotype 1b patients than in SVR genotype 1b patients. Although the small number of genotype 1b patients who achieved SVR did not allow a definite conclusion to be drawn, these findings suggest that the reduction of HCC risk in SVR patients might be attributed to the removal of genotype 1 infection. In contrast, they indicate that genotype 2a/c does not constitute an additive risk factor for HCC in cirrhosis, which remains the main determinant associated with the development of the tumor.
Despite the long follow-up period of our study, only 18 patients were lost to follow-up, and they were equally distributed among genotypes. As the development of HCC could have been the reason for dropout for some of these patients, we performed a conservative analysis that assumed all those with genotype 2a/c, but none of the patients infected with genotype 1b, had developed HCC. This analysis ruled out the possibility that dropout could have affected the study results.
Despite the lack of in vitro and in vivo evidence of different oncogenic properties for different HCV genotypes, our study suggests that as for certain human papilloma virus genotypes,22 at least 1 HCV genotype may have higher oncogenic potential. We therefore speculate that the weak and focused host's immune selection of viral quasispecies in patients with genotype 1 with respect to those with genotype 223, 24 could eventually be responsible for the conservation of potentially oncogenic viral variants.
An additional piece of information emerging from this study is that the presence of small varices at enrollment, a surrogate marker of portal hypertension, was associated with the risk of HCC. Although the linkage between the severity of portal hypertension and the risk of decompensation (particularly bleeding), especially in advanced liver disease, is well established, only one study has described so far the relationship between esophageal varices and the risk of HCC occurrence.25 This may reflect the severity of chronic liver injury but may also result from impaired regional blood flow and local hypoxia, which may stimulate the synthesis of angiogenic factors.26
In conclusion, our results firmly indicate that patients with cirrhosis that are infected with HCV genotype 1b and concurrent esophageal varices have a higher risk of developing HCC, which implies that this subset of subjects should be more aggressively monitored for the occurrence of HCC at an early stage when it can be effectively treated. Besides the obvious benefits in routine clinical practice, this information could have important public health implications and be of major help to policymakers, allowing projections about the future burden of the disease.
We thank Lara Firmo for editorial assistance.