Incidence and risk factors of hepatocellular carcinoma in patients with hepatitis C in China and the United States

Hepatitis C virus (HCV) infection is the main cause of hepatocellular carcinoma (HCC) in the United States (US) and an increasingly common cause of HCC in China. We aimed to evaluate the incidence and risk factors of HCC in HCV patients in the US and China. 795 HCV RNA + patients without HCC from University of Michigan Health System (UMHS) in the US and 854 from Peking University Health Sciences Center (PUHSC) in China were prospectively followed for a median of 3.2 and 4.0 years, respectively. 45.4% UMHS and 16.2% PUHSC patients had cirrhosis. 57.6% UMHS and 52.0% PUHSC patients achieved SVR. 45 UMHS and 13 PUHSC patients developed HCC. Cumulative incidence of HCC at 5 years was 7.6% in UMHS and 1.8% in PUHSC cohort (P < 0.001). Ten patients not diagnosed with cirrhosis at enrollment but median APRI ≥ 2.0 developed HCC. Multivariate analysis showed age, gender, cirrhosis and APRI were predictors of HCC while study site and SVR were not. In this study of HCV patients, HCC incidence in the PUHSC cohort was lower than in the UMHS cohort, due to lower proportion of PUHSC patients with cirrhosis. APRI can identify risk of HCC among patients not diagnosed to have cirrhosis.


Results
A total of 1957 patients were enrolled: 1000 at UMHS, and 957 at PUHSC. Of these, 167 were excluded for having HCC (141 UMHS and 26 PUHSC) at enrollment, and 141 (64 UMHS and 77 PUHSC) were excluded due to lack of follow-up. The remaining 1649 patients: 795 at UMHS and 854 at PUHSC were included in this analysis. Figure 1A,B show the outcomes of patients and occurrence of HCC stratified by baseline liver disease stage.
To further examine the impact of SVR on HCC incidence, we combined the two cohorts and stratified the patients by cirrhosis status at baseline and analyzed SVR as a time-dependent variable ( Table 2). The incidence of HCC was 2.6 vs. 3.1 per 100 person-years (HR 0.9, 95% CI 0.4-1.9) in cirrhosis patients with vs. those without SVR; and 0 vs. 0.3 per 100 person-years in non-cirrhosis patients with vs. those without SVR. When SVR was analyzed as time-fixed variable (Supplementary Table 1), incidence of HCC was 0.8 vs. 6.2 per 100 person-years (HR 0.2, 95% CI 0.1-0.4) in cirrhosis patients with vs. those without SVR; and 0 vs. 0.6 per 100 person-years in non-cirrhosis patients with vs. those without SVR.
Factors associated with HCC development. In both cohorts, patients with HCC development were significantly older, more likely to have cirrhosis and had higher APRI and FIB-4 at enrollment and less likely to have achieved SVR during follow-up. However, there were no differences in BMI, diabetes, use of alcohol, tobacco or coffee, HCV genotype and prevalence of anti-HBc compared to those without HCC ( Table 3). Results of univariate competing risk analysis are shown in Supplementary Table 2.
Multivariate Cox regression analysis showed that older age, male sex, higher APRI, and baseline cirrhosis were associated with a higher risk of HCC in the combined cohort, while study site, diabetes and SVR were not (Fig. 3A). Analysis of risk factors for HCC identified older age and male sex but not SVR in the subgroup with compensated cirrhosis (Fig. 3B) and older age, and SVR in the subgroup with decompensated cirrhosis (Fig. 3C). When the two cohorts were separately analyzed, older age, higher APRI and baseline cirrhosis were associated with a higher risk of HCC in the UMHS cohort ( Supplementary Fig. 1), while no risk factor of HCC was found in the PUHSC cohort ( Supplementary Fig. 2). HCC in non-cirrhosis patients. Ten patients (8 UMHS and 2 PUHSC) not diagnosed to have cirrhosis at enrollment developed HCC during follow-up. Median APRI at baseline was 2.0 and FIB-4 was 4.5 in these 10 patients, compared to 0.6 and 1.7, respectively in those who did not develop HCC. Of note, none of 509 patients with baseline APRI ≤ 0.5 and none of 437 patients with baseline FIB-4 < 1.45 developed HCC during follow-up, while 5 of 86 (5.8%) patients with baseline APRI > 2.0 and 8 of 161 (5.0%) with baseline FIB-4 > 3.25 developed HCC. Of the 10 patients who developed HCC, 3 met protocol criteria for cirrhosis at the time of HCC diagnosis, and an additional 3 had APRI > 2.0 and FIB-4 > 3.25 but did not meet protocol criteria for cirrhosis at the time of HCC diagnosis.
Due to the small number of HCC outcome, analysis of risk factors of HCC in the group with no cirrhosis at baseline is exploratory. Multivariate analysis showed that older age, diabetes, APRI > 2.0, and decompensation during follow-up were independently associated with a higher risk of HCC, while achieving SVR was independently associated with a lower risk of HCC in the combined cohort of UMHS and PUHSC patients with no cirrhosis at baseline (Fig. 3D).

Discussion
Chronic HCV infection is the leading cause of HCC in the US and an increasingly common cause of HCC in China. Our study sought to compare the incidence of HCC in UMHS and PUHSC HCV cohorts and to identify the risk factors of HCC development. We found that HCC incidence in the PUHSC cohort was lower than in the UMHS cohort, mainly due to a lower proportion of PUHSC patients with cirrhosis.

Scientific Reports
| (2020) 10:20922 | https://doi.org/10.1038/s41598-020-77515-y www.nature.com/scientificreports/ Other differences in the two cohorts may also contribute to the higher incidence of HCC in the UMHS cohort, e.g., more men, higher BMI, more diabetes and more patients with history of regular alcohol or tobacco use. Male sex had been shown in many studies to be a predictor of HCC 16 , which was confirmed in the combined cohort and UMHS cohort. Previous studies demonstrated that obesity, diabetes, alcohol or tobacco use increased risk of HCC in HCV patients [17][18][19][20] . In our study, diabetes and tobacco use were associated with a higher risk of HCC on univariate analysis but not on multivariate analysis in the combined cohort. We were not able to confirm the associations of higher BMI and alcohol use with HCC possibly due to the small number of patients with HCC.
Twenty patients in the PUHSC cohort and none in the UMHS cohort were HBsAg positive but none of them developed HCC during follow-up. There was a lower prevalence of anti-HBc and more frequent coffee consumption in the UMHS cohort. We found no difference in prevalence of anti-HBc between patients with and those without HCC development in the combined cohort and also when the two cohorts were separately analyzed. Reports on the role of anti-HBc positivity and occult HBV infection in HCC development among patients with chronic HCV infection showed conflicting results 21,22 . Coffee consumption had been demonstrated to have a protective effect against HCC 23 . In our study, coffee consumption was associated with lower HCC incidence on univariate analysis but not on multivariate analysis in the combined cohort.
In our study, the most consistent predictor of HCC was cirrhosis. The higher proportion of UMHS patients with cirrhosis at enrollment resulted in a significantly higher incidence of HCC in the UMHS than the PUHSC cohort. When patients were stratified for cirrhosis at enrollment, there was no difference in incidence of HCC in the two cohorts among the patients with cirrhosis, overall and when stratified for presence or absence of  24 . We found that APRI and FIB-4 can stratify the risk of HCC in patients who did not have a diagnosis of cirrhosis. Among the patients with no cirrhosis at enrollment, none of the patients with baseline APRI ≤ 0.5 or FIB-4 < 1.45 developed HCC during follow-up. Validation of our finding will help to identify HCV patients with advanced fibrosis or undiagnosed cirrhosis who might benefit from HCC surveillance. Most studies showing a significant reduction in incidence of HCC after achieving SVR have used SVR as a time-fixed variable 10,11,25 . The results of these studies can be potentially biased even when stratified for cirrhosis status at enrollment because patients with more advanced liver disease are more likely to develop HCC and at the same time less likely to achieve SVR. In a French cohort study of 1270 HCV-associated cirrhosis patients (336 treated with DAAs and 934 treated with interferon) in which SVR was used as a time-dependent variable, SVR was independently associated with a lower risk of HCC (HR 0.44; 0.29-0.69) 26 .
In our study, we used SVR as a time-fixed variable as well as a time-dependent variable. Using SVR as a time-fixed variable, we found that SVR was associated with a significant reduction in incidence of HCC. When we used SVR as a time-dependent variable, SVR was not associated with a statistically significant reduction in incidence of HCC overall, possibly because of the short duration of post-SVR follow-up, median (IQR) 12.0 (4.6-19.4) months in UMHS cohort and 0 (0-22.5) months in PUHSC cohort. However, SVR was associated with a statistically significant reduction in incidence of HCC in patients with no cirrhosis and those with decompensated cirrhosis at baseline. We also found that the incidence of HCC after SVR was still high in cirrhosis patients while no HCC occurred after SVR in non-cirrhosis patients. These data highlight the importance of continued HCC surveillance in patients with cirrhosis prior to achieving SVR, and the need to diagnose and treat patients early, prior to cirrhosis.
Our study has notable strengths. The parallel study design and the prospective nature of our study allow us to compare the rate of incident HCC in two cohorts of patients with chronic HCV infection in the US and in China. The use of standardized protocol, manual of procedures, and criteria for diagnosis of cirrhosis and HCC, and source documents to verify these diagnoses provide confidence in our results. There are several limitations in our study. First, while this study enrolled more than 1600 patients, the number is too small to be representative of patients with chronic HCV infection in the US and in China. Second, our follow-up duration, in particular the duration of post-SVR follow-up, was short, limiting our ability to study the impact of SVR on long-term risk of HCC. Third, the number of HCC events was small limiting the power to detect risk factors within as well as between cohorts. Fourth, there was ascertainment bias in our study because if a patient was lost to follow-up and developed HCC then we would have no way to record this HCC event.
In summary, our study found that HCC incidence in the PUHSC cohort was significantly lower than in the UMHS cohort. This difference is mainly due to less advanced liver disease in the PUHSC cohort. Incidence of HCC was decreased but not eliminated in cirrhosis patients who achieved SVR. By contrast, no HCC was observed in non-cirrhosis patients who achieved SVR. Among patients with no cirrhosis, high APRI identified those who were at risk of HCC. Our data support that patients with chronic HCV infection should be treated prior to development of cirrhosis to prevent liver-related complications, including HCC.  A description of the study design had been previously reported 12 . In brief, consecutive adults age ≥ 18 years with chronic HCV infection (HCV RNA positive) were enrolled between September 2011 and July 2015. Patients who had undergone liver transplantation, known coinfection with HIV, life expectancy < 12 months due to extra-hepatic illnesses, or were receiving HCV treatment at enrollment, were excluded. Patients enrolled at both centers were evaluated using an identical protocol. All patients except those with HCC at enrollment were prospectively followed. Patients with cirrhosis were evaluated every 6 months while those without cirrhosis were evaluated every 12 months. Follow-up continued until patient died, developed HCC, received liver transplantation, withdrew informed consent, or when the study was closed in December 2017. Patients who did not have any follow-up visit after enrollment were excluded from this analysis. The research protocol was approved by the Institutional Review Boards of both the University of Michigan and Peking University Health Science Center and conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all subjects. Protocol, surveys, and data forms were developed in English and then translated into Chinese, and data were entered into a web-based database. Alcohol, tobacco and coffee consumption were assessed using a standardized questionnaire. Regular alcohol use was defined as at least 1 drink/day, regular tobacco use was at least one cigarette/day, and regular coffee consumption at least 1 cup/day. For American patients, overweight was defined as body mass index (BMI) ≥ 25 but < 30, and obesity as BMI ≥ 30 kg/m 2 27 . For Chinese patients, overweight was defined as BMI ≥ 24 but < 28, and obesity as BMI ≥ 28 kg/ m 2 28 . Diabetes mellitus (DM) was defined by medical history or use of medications for treatment of diabetes; and for those with no history of diabetes by fasting blood glucose ≥ 126 mg/dL or random blood glucose ≥ 200 mg/dL. Assessment of liver disease. Diagnosis of cirrhosis was based on histology when available. In the absence of histology, diagnosis of cirrhosis was based on evidence of clinical decompensation or 2 of the following 4 criteria: radiological imaging showing features of cirrhosis (nodular liver, varices or splenomegaly), platelet count < 100 × 10 9 /L in the absence of other explanations, liver stiffness measurement by vibration controlled transient elastography (Fibroscan) > 13 kPa, and gastro-esophageal varices on endoscopy. HCC was diagnosed by histology or by triple phase CT or MRI per the American Association for the Study of Liver Diseases guidelines 29 . Source documents supporting the diagnosis of cirrhosis and HCC were collected and investigators from each country audited the documents from the other country to confirm these diagnoses. www.nature.com/scientificreports/ Statistical analyses. Data were analyzed using SAS 9.4 (SAS Institute, Cary, NC). Categorical data are presented as number (percent) and continuous data as median (interquartile range, IQR). T-test or Mann-Whitney test was used to compare continuous variables according to the data distribution and chi-square test for comparison of categorical data. Kaplan-Meier analysis was used to estimate cumulative incidence of HCC. Univariate and multivariate competing risk models were used to identify risk factors of HCC. Death or liver transplantation before HCC diagnosis was considered as competing risk event, and SVR was included as a time-dependent variable. Due to the short duration of post-SVR follow-up, analysis was also performed with SVR as time-fixed variable. Demographic, clinical and environmental factors with P < 0.1 on univariate analysis were incorporated in the full model. P values < 0.05 were considered statistically significant.