Potential conflict of interest: Nothing to report.
Recent studies suggest that diabetes mellitus increases the risk of developing hepatocellular carcinoma (HCC). The aim of this study is to quantify the risk of HCC among patients with both diabetes mellitus and hepatitis C in a large cohort of patients with chronic hepatitis C and advanced fibrosis. We included 541 patients of whom 85 (16%) had diabetes mellitus. The median age at inclusion was 50 years. The prevalence of diabetes mellitus was 10.5% for patients with Ishak fibrosis score 4, 12.5% for Ishak score 5, and 19.1% for Ishak score 6. Multiple logistic regression analysis showed an increased risk of diabetes mellitus for patients with an elevated body mass index (BMI) (odds ratio [OR], 1.05; 95% confidence interval [CI], 1.00-1.11; P = 0.060) and a decreased risk of diabetes mellitus for patients with higher serum albumin levels (OR, 0.81; 95% CI, 0.63-1.04; P = 0.095). During a median follow-up of 4.0 years (interquartile range, 2.0-6.7), 11 patients (13%) with diabetes mellitus versus 27 patients (5.9%) without diabetes mellitus developed HCC, the 5-year occurrence of HCC being 11.4% (95% CI, 3.0-19.8) and 5.0% (95% CI, 2.2-7.8), respectively (P = 0.013). Multivariate Cox regression analysis of patients with Ishak 6 cirrhosis showed that diabetes mellitus was independently associated with the development of HCC (hazard ratio, 3.28; 95% CI, 1.35-7.97; P = 0.009). Conclusion: For patients with chronic hepatitis C and advanced cirrhosis, diabetes mellitus increases the risk of developing HCC. (HEPATOLOGY 2008;47:1856–1862.)
Recent epidemiological studies suggest that the presence of diabetes mellitus increases the risk of hepatocellular carcinoma (HCC).1, 2 An explanation for this association may be that diabetes often occurs as part of the metabolic syndrome, which increases the risk of nonalcoholic steatohepatitis (NASH), and that HCC can be a late complication of NASH.3
Diabetes mellitus is more prevalent among patients with chronic hepatitis C than in the general population.4 Liver disease contributes to insulin resistance because it leads to a progressive impairment of insulin secretion and it induces hepatic insulin resistance.5 Studies in transgenic mouse models that harbored the hepatitis C core gene have shown that hepatic insulin resistance may be caused by elevated levels of tumor necrosis factor-alpha, which disturbs the tyrosine phosphorylation of insulin receptor substrate-1.6
Chronic hepatitis C virus (HCV) infection itself also increases the risk of HCC. It leads to chronic inflammation of the liver, to liver fibrosis, and it may eventually progress to cirrhosis. For patients with hepatitis C cirrhosis the risk for development of HCC is 0.54 to 2.0% per year.7, 8
Therefore, among patients with both chronic hepatitis C and diabetes mellitus there are two potential ways in which HCC may develop: by the metabolic pathway and by the carcinogenic effect of the HCV.
The aim of this study was to quantify the risk of HCC among patients with both diabetes mellitus and chronic hepatitis C in a large cohort of patients with chronic hepatitis C and advanced fibrosis.
anti-HBc, anti-hepatitis B core antigen; BMI, body mass index; CI, confidence interval; HCC, hepatocellular carcinoma; HCV, hepatitis C virus; HR, hazard ratio; IQR, interquartile range; NASH, nonalcoholic steatohepatitis; OR, odds ratio.
Patients and Methods
We included in the study all consecutive patients with chronic hepatitis C and biopsy-proven advanced fibrosis or cirrhosis (Ishak score 4 to 6) treated between 1990 and 2003 in five large hepatology units in Europe and Canada. Patients were not eligible if they were infected with hepatitis B (hepatitis B surface antigen positive) or human immunodeficiency virus.
We obtained approval from the ethics committees of each participating center. Local investigators identified all eligible patients and the principal investigator (B.J.V.) then visited each center to enter the individual patient chart review data in a central database in a standardized and predefined way.9
We obtained data on patient demographics (gender, age), on body weight and height and on the presence of diabetes. We calculated BMI as weight divided by height squared. We defined overweight as BMI 25 to 30 kg/m2, defined obesity as BMI 30 to 35 kg/m2, and defined morbid obesity as BMI >35 kg/m2. For patients of Asian descent we used an adapted scale of BMI 23 to 27.5 kg/m2 for overweight, BMI 27.5 to 32.5 kg/m2 for obesity, and BMI >32.5 kg/m2 for morbid obesity.10
Patients with recent alcohol abuse were not eligible for treatment and therefore were not included in the study.
We recorded whether patients had any diagnosis of diabetes mellitus at baseline; either by elevated fasting serum glucose (>6.1 mmol/L) or a positive glucose tolerance test, or if patients used antidiabetic medication. We measured biochemical data (bilirubin, albumin), hematological data (platelet count, prothrombin time) and virological data (HCV genotype and viral load, anti hepatitis B core antigen [HBc] positivity) in the certified laboratories of participating hospitals and we corrected them centrally for local normal values. We assessed fibrosis in liver biopsies scored by local pathologists, who all had a large experience in scoring liver biopsies from patients with viral hepatitis. Finally, we recorded details of hepatitis C treatment (duration of treatment, interferon or pegylated interferon, monotherapy or combination treatment with ribavirin).
All participating centers had a protocol for follow-up of patients with advanced fibrosis or cirrhosis. We followed up patients with advanced fibrosis with yearly visits, including physical examination, serum alanine aminotransferase, HCV ribonucleic acid (RNA), and alpha fetoprotein testing, and abdominal ultrasound, computed tomography, or magnetic resonance imaging. We evaluated patients with cirrhosis (Ishak score 5 or 6) every 6 months. The protocol was similar for patients with and without diabetes mellitus. We considered patients to have HCC if the diagnosis was histologically confirmed, or if two coincident imaging techniques (ultrasound, computed tomography, or magnetic resonance imaging) showed a new focal lesion >2 cm with arterial hypervascularization, or if one imaging technique showed a new focal lesion >2 cm with arterial hypervascularization in the presence of an alpha-fetoprotein level >400 ng/mL. For the occurrence of HCC, we used the date when the diagnosis was confirmed by histology or radiography.
If the follow-up was incomplete, the treating physician contacted the patient. In case the patient could not be reached, the treating physician contacted the patient's general practitioner in order to complete the follow-up.
To identify factors associated with the presence of diabetes mellitus and to evaluate whether the presence of diabetes mellitus affects the response to treatment of chronic HCV infection, we used univariate and multivariate logistic regression analyses. We checked all variables for interactions. We expressed results as odds ratios (ORs) and their 95% confidence intervals (CIs).
We used the Kaplan-Meier method to estimate the occurrence of HCC over time. Entry into the study started at the initiation of treatment. We applied Cox regression analysis to determine which factors were independently associated with the development of HCC. We considered the following baseline factors: age, fibrosis stage, diabetes mellitus, genotype, gender, BMI, anti-HBc positivity, bilirubin, albumin, and platelet count.
Anti-HBc positivity is a risk factor for HCC; however, the hazard of HCC was not proportional over time. Therefore, we stratified the analysis for risk of development of HCC for serum anti-HBc positivity. For covariates other than anti-HBc positivity, the assumption of proportionality was not violated. We observed multiple colinearity between fibrosis stage and the covariates bilirubin, platelet count, and albumin. Both the fibrosis stage and these laboratory values reflect the severity of liver disease. Therefore, we developed two separate models for either inclusion of fibrosis in the model or inclusion of the covariates bilirubin, platelet count and albumin. The model including the covariates bilirubin, platelet count, and albumin had the lowest Akaike information criterion score, which is a measure of the goodness of fit of a statistical model. The reported hazard ratios (HRs) are the relative increases in hazard associated with increases of 1 μmol/L bilirubin, 1 g/L albumin, 1 × 109 cells/L platelets, and 1 mmol/L glucose.
To assess whether the effect of diabetes mellitus on HCC could be confounded by response to interferon-based treatment, we performed an additional analysis in which response to treatment was modeled as a time-dependent covariate to represent the ability of patients undergoing more than one treatment course to change their status from nonresponders to responders in sequential courses. Since the definition of sustained virological response is undetectable serum HCV-RNA by sensitive molecular tests at 24 weeks after the end-of-treatment, we used this time point as time 0 for classifying response versus nonresponse.
We reported results as relative hazards with 95% CIs.
Since either baseline bilirubin, albumin, or platelet count were missing in 28.5% of cases, we used multiple imputation to impute missing values:11, 12 we applied the Markov Chain Monte Carlo method with single chain (PROC MI, SAS; SAS Institute Inc., Cary, NC) to construct 10 complete datasets. We entered all baseline factors and time related factors and events related to the analyses into the imputation procedure and if necessary transformed to conform to the multivariate normality assumption. We then run Cox analysis (PROC PHREG in SAS) or logistic regression analysis (PROC LOGISTIC in SAS) on each dataset and the results and inference combined using PROC MIANALYZE in SAS. We performed all statistical analyses using SAS (version 9.1.3).
A total of 541 patients were eligible and were included in the study, 85 (16%) of them had a diagnosis of diabetes mellitus at the time of inclusion.
The median age at inclusion was 50 years, the patients with diabetes mellitus being older than those without diabetes (median 51 versus 49 years) (Table 1).
Table 1. Baseline Characteristics
P Value (Mann-Whitney U/χ2)
Median (interquartile range).
Genotype was missing in 12% of the patients.
Viral load was measured by local hybridization or polymerase chain reaction assays and could be retrieved in 68% of the patients.
Patients with diabetes mellitus had more severe fibrosis, lower mean albumin levels, and lower mean platelet counts. The prevalence of diabetes mellitus was 10.5% among patients with an Ishak fibrosis score of 4, 12.5% for Ishak fibrosis score 5, and 19.1% for Ishak fibrosis score 6. Of the 85 patients with a diagnosis of diabetes mellitus, 20 used subcutaneous insulin, 34 used oral antidiabetic medication, and 31 had dietary measures only. The mean fasting glucose was 8.3 mmol/L (standard deviation 3.1) for patients on treatment (oral or insulin) and 7.2 mmol/L (standard deviation 2.4) for patients on dietary measures. The prevalence of diabetes mellitus was similar among the European and the Canadian patients enrolled in the study (15.9% versus 15.5%, P = 0.91). A total of 72% of the patients had a complete follow-up until January 1, 2005, 6 months prior to the data acquisition.
Effects of BMI and Previous Alcohol Use.
A total of 144 patients without diabetes mellitus (39%) and 32 patients with diabetes mellitus (44%) were overweight, 58 (16%) versus 11 (15%) were obese, and 22 (6%) versus 10 (14%) had morbid obesity. Patients with diabetes mellitus tended to have higher median BMIs than patients without diabetes mellitus (28 kg/m2 versus 26 kg/m2; P = 0.25; Table 1). Multiple logistic regression analysis showed a strong trend toward a higher risk of diabetes mellitus among patients with an elevated BMI (OR, 1.05; 95% CI, 0.99-1.11; P = 0.060), and a trend toward a lower risk of diabetes mellitus for patients with higher serum albumin levels (OR, 0.81; 95% CI, 0.63-1.04; P = 0.095) and higher platelet count (OR, 0.95; 95% CI, 0.88-1.02; P = 0.181). The effects of age, gender, fibrosis stage, and bilirubin levels were not significant.
A total of 20 patients had a history of previous alcohol abuse and one of them developed HCC. A total of three patients both had a history of alcohol abuse and had diabetes mellitus, but none of these three patients developed HCC. Past alcohol abuse was not associated with development of HCC in univariate Cox regression analysis (HR, 0.77; CI, 0.11-5.58; P = 0.79).
Diabetes Mellitus and Development of HCC.
HCC developed more frequently among patients with diabetes mellitus. During a median follow-up of 4.0 years (interquartile range [IQR], 2.0-6.7), 11 patients (13%) with diabetes developed HCC, versus 27 patients (5.9%) without diabetes mellitus. A total of 24 cases (65%) had a histological diagnosis.
The median time interval between the last imaging without tumor (ultrasound, computed tomography, or magnetic resonance imaging) and the confirmation of the diagnosis of HCC was similar for patients with and without diabetes mellitus (0.9 years [IQR, 0.1-2.1] versus 0.9 years [IQR, 0.5-1.7]), respectively (P = 0.68). The 5-year occurrence of HCC was 11.4% (95% CI, 3.0-19.8%) for chronic hepatitis C patients with diabetes mellitus and 5.0% (95% CI, 2.2-7.8%) for patients without diabetes mellitus (P = 0.013). Univariate Cox regression analysis showed that there were no statistically significant differences in the risk of developing HCC between patients using insulin or oral antidiabetic medication and patients on dietary measures (HR, 1.11; CI, 0.32-3.79; P = 0.87). Among patients with diabetes mellitus, there was a trend toward a higher risk of HCC as fasting glucose levels increased (HR, 1.22; CI, 0.98-1.52; P = 0.082).
Multivariate Cox regression analysis of the overall study population showed that male gender (HR, 2.97; 95% CI, 1.20-7.39; P = 0.019) and older age (HR, 1.07; 95% CI, 1.03-1.11; P = 0.001) were significantly associated with an elevated risk of developing HCC. In addition, there was a strong trend toward a higher incidence of HCC among patients with diabetes mellitus (HR, 2.07; 95% CI, 0.95-4.47; P = 0.066). The associations with BMI, platelet count, bilirubin, and albumin levels were not statistically significant.
All 11 diabetic patients who developed HCC had Ishak fibrosis score 6 (Fig. 1). We performed a subgroup analysis on patients with Ishak fibrosis score 6 (n = 303), in which diabetes mellitus was significantly associated with the development of HCC (HR, 3.28; 95% CI, 1.35-7.97; P = 0.0087). Other factors predictive of HCC among patients with Ishak fibrosis score 6 were male sex (HR, 2.91; 95% CI, 1.03-8.26; P = 0.044) and older age (HR, 1.07; 95% CI, 1.02-1.12; P = 0.007; Table 2).
Table 2. Multivariate Cox Regression Analysis Evaluating Factors Associated with Development of Hepatocellular Carcinoma (HCC) in the Overall Cohort and Among Patients with Ishak Fibrosis Score 6 Only
Overall Cohort (n = 541)
Patients with Ishak Fibrosis Score 6 (n = 303)
Hazard Ratio (95% CI)
Hazard Ratio (95% CI)
Body mass index
Effect of HCV Treatment
All patients were treated for chronic hepatitis C with (peg)interferon with or without ribavirin. A total of 20 patients with diabetes mellitus (24%) and 139 patients without diabetes mellitus (30%) reached sustained virological response after one or more treatment courses. Logistic regression analysis showed no statistically significant effect of the presence of diabetes mellitus on the chance of responding to HCV treatment, in this population with advanced liver disease (OR, 0.70; 95% CI, 0.41-1.20; P = 0.20).
When response to treatment was entered as a time-dependent covariate in the Cox regression model, it had no statistically significant effect on development of HCC (HR, 0.54; CI, 0.16-1.88; P = 0.33). Moreover, after adjusting for response to treatment, the effect of diabetes mellitus on HCC development remained essentially unchanged: HR, 1.77; CI, 0.77-4.09; P = 0.182 for the overall study population; and HR, 3.35; CI, 1.28-8.74; P = 0.014 for patients with Ishak fibrosis score 6.
This large cohort study gives, for the first time, a quantification of the risk of developing HCC over time for patients with chronic hepatitis C, advanced fibrosis, and diabetes mellitus. The 5-year risk of developing HCC is 11.4% for patients with both diabetes mellitus and chronic hepatitis C with advanced fibrosis. Patients without diabetes mellitus included in our study had a lower risk of HCC with occurrence of HCC in 5.0% after 5 years, which is consistent with the 2.7%-10% that has been described in the literature for patients with cirrhosis due to hepatitis C.7, 8
A Japanese study of 279 chronic hepatitis C patients without cirrhosis showed that the 5-year incidence of HCC was 9.5% for diabetics and 2.0% for nondiabetics.2 Although in a Western population development of HCC is very infrequent among patients without cirrhosis, the rate of diabetics versus nondiabetics developing HCC is similar.
All patients in our cohort have been treated with an interferon-based regimen. A previous study in the same cohort of patients showed that two of three sustained virological responders who developed HCC during follow-up had diabetes mellitus and this study failed to show a statistically significant effect of treatment response on the risk of developing HCC.9
In the present study, advanced liver disease itself turned out to be one of the factors determining the risk of diabetes mellitus. This concurs with previous reports, showing that advanced liver disease can lead to insulin resistance, independent of the cause of cirrhosis.13
In subgroup analysis of patients with Ishak fibrosis score 6, diabetes mellitus was an independent predictor of HCC. Although there probably is a range for true severity of cirrhosis within fibrosis stage 6, diabetes mellitus was still an independent predictor of HCC if we corrected for other markers of advanced liver disease such as platelet count, bilirubin, and albumin levels. This suggests that the presence of diabetes mellitus is more than just a marker for more severe liver disease with a greater propensity to progress to HCC.14 Moreover, a large population-based case-control study in the United States has shown that diabetes mellitus was independently associated with a three-fold increase of the risk of HCC even for subjects without concomitant disease such as chronic hepatitis C.15
The effect of diabetes mellitus on HCC development may be partly explained by behavioral factors such as alcohol use, tobacco smoking, and a high-fat diet.16 It is well established that high BMI is a risk factor for diabetes mellitus and steatohepatitis. In addition, tobacco smokers have a 1.5-fold increased risk of developing diabetes mellitus17, 18 and although moderate alcohol consumption may have a protective effect on the development of diabetes mellitus, it clearly is an additional risk factor for steatohepatitis.19
Alcoholic steatohepatitis or NASH superimposed on chronic inflammation due to hepatitis C may lead to an elevated risk of carcinogenesis. A cohort study in an Asian population of patients with hepatitis C suggested that steatosis is an independent risk factor for development of HCC.20 A recent Australian case-control study found similar amounts of steatosis in patients with chronic hepatitis C who subsequently developed HCC and matched controls. However, the size of this study was limited.21 Among chronic hepatitis C patients, genotype 3 is most commonly associated with liver steatosis.22 In our cohort, we found no effect of hepatitis C genotype 3 on the risk of HCC (Table 1). Furthermore, the prevalence of obesity was high, both among patients with and without diabetes mellitus, suggesting that obesity is frequent among all chronic hepatitis C patients with advanced fibrosis.
Unfortunately, we did not have data about tobacco use in our study population. Our cohort includes patients treated for HCV and because recent alcohol abuse was a contraindication for treatment, no patients with active alcohol abuse were included. We found no increased risk of HCC among patients with a past history of alcohol abuse.
Interestingly, among patients with diabetes mellitus, there was a trend toward a higher risk of HCC as fasting glucose levels increased. Higher fasting glucose levels are associated with higher compensatory insulin levels in patients with diabetes mellitus23 and because previous in vitro studies have shown that hyperinsulinemia enhances the proliferation of human hepatocellular cancer cells,24, 25 we hypothesize that the presence of hyperinsulinemia might further explain the increased risk of HCC among patients with diabetes mellitus.
In conclusion, for patients with chronic hepatitis C and cirrhosis, diabetes mellitus increases the risk of developing HCC.