Abstract
This study aimed to investigate the association of fasting insulin and glucose levels with hepatocellular carcinoma (HCC) risk in a case–cohort study within a cohort (1989–2006) of 2903 male government employees chronically infected with hepatitis B virus (HBV) in Taiwan. Insulin, glucose and HBV-related factors were assayed in baseline plasma among 124 HCC cases and a random subcohort of 1084 of the total cohort. After adjustment for demographics and HBV-related factors, including viral load and genotype, the HCC risk was higher for the highest [>6.10 μU/ml, hazard ratio (HR) = 2.36, 95% confidence interval (CI): 1.43–3.90] and lowest (<2.75 μU/ml, HR = 1.57, 95% CI: 0.96–2.58) categories of insulin, compared with insulin of 2.75–4.10 μU/ml. The dose–response relationship between insulin and HCC varied by follow-up time, with stronger association for the HCC cases that occurred ≥8 years after baseline (P for trend <0.0001). The effect of higher insulin on HCC risk remained after adjustment for other metabolic factors, and was fairly consistent across strata of age, body mass index, and HBV genotypic variants. However, it was more profound among those with viral load <4.39 log10 copies/ml at recruitment (>6.10 μU/ml, HR = 6.15, 95% CI: 2.48–15.22). Higher insulin was also associated with an increased risk for cirrhosis diagnosed by ultrasonography and elevated alanine aminotransferase. No association with either cirrhosis or HCC was noted for glucose or diabetes after adjusting for insulin. In conclusion, elevated insulin levels are an independent risk factor for HCC among HBV carriers, especially for those with lower viral load.
Introduction
Worldwide, chronic hepatitis B virus (HBV) infection affects >350 million people who are at increased risk for hepatocellular carcinoma (HCC) (1). In Asia-Pacific endemic regions of hepatitis B, HBV-related factors, including prolonged HBV e antigen (HBeAg)-positive phase, high viral load, HBV genotype C infection and the basal core promoter (BCP) double mutations, have been documented as important predictors of HCC (2–8). Except for alcohol abuse (9), however, little is known about the association between modifiable lifestyle factors and the risk of HCC in hepatitis B.
Emerging data indicate that the prevalence of obesity is rising globally (10). Excess body weight increases the risk of non-alcoholic fatty liver disease (NAFLD) (11), and has been associated with risks for HCC and liver cirrhosis of different etiology, including hepatitis B, by multiple epidemiological studies (12–14). Obesity is closely associated with a cluster of metabolic disorders, including glucose intolerance, insulin resistance (IR) and diabetes (11). While IR risk has recently been demonstrated to be increased as a result of chronic hepatitis C virus (HCV) infection with IR itself associated with increased hepatic steatosis and advanced fibrosis risk in hepatitis C patients (15,16), the effect of IR on hepatitis B progression remains unclear. Furthermore, although diabetes or elevated fasting glucose has been associated with HCC incidence or mortality, this association is found primarily in developed countries where HCV is the major cause of HCC (17–21). In areas endemic for hepatitis B, inconsistent results have been reported and previous studies did not incorporate viral factors in the risk analysis of diabetes (12,22–24).
In the present study, we assessed the relative contribution of metabolic factors, including excess body mass index (BMI), fasting levels of insulin and glucose and diabetes, to HCC risk, with a particular focus on the role of insulin and glucose, in a prospective cohort of male HBV carriers. Because information on various viral factors is available, we were able to explore the impact of viral factors on the metabolic factors-related HCC risk.
Subjects and methods
Study design and participants
This study was conducted within an ongoing prospective cohort study comprising 2903 unrelated, originally healthy, male HBV surface antigen (HBsAg)-positive government employees aged 30–65 years, who were enrolled during routine free physical examination between August 1989 and June 1992 (12). At baseline, participants completed questionnaires regarding lifestyle habits and medical history. They also underwent a physical examination by physicians, an anthropometric evaluation, and a blood pressure measurement and provided a blood sample following an overnight fast of at least 8 h for serological assays. BMI was calculated using weight divided by height squared (kg/m2).
Participants were followed by medical examinations approximately every year and by computerized data linkage with the profiles on the national registry of cancer and death in Taiwan. HCC was diagnosed by either a histological finding or elevated serum α-fetoprotein level (≥400 ng/ml) combined with at least one positive image on angiography, sonography, and/or computed tomography. All subjects gave written informed consent, and the research ethics committee at the College of Public Health, National Taiwan University approved this study.
In the present study, subjects were recruited from an existing case–cohort study that designed to evaluate the role of longitudinal HBV DNA levels in the development of HCC (5). This case–cohort study was established in 2004, when we chose a random subcohort (n = 1084; including 53 incident HCC cases) from all cohort members meeting the following criteria: (i) antibodies to HCV (anti-HCV) negative, (ii) no prior history of antiviral therapy and (iii) at least two blood samples collected after study entry. All cases (n = 59) not already included who developed HCC by 2005 and met the above criteria were added to complete the case–cohort population (n = 1143). The present study extended the follow-up period to 31 December 2006 and identified 13 additional HCC cases from the subcohort. We excluded one case due to insufficient sample; remaining were 124 cases and 1018 non-cases.
Biomarkers
All subjects were tested for HBsAg, alanine aminotransferase (ALT) and α-fetoprotein at baseline. Their frozen baseline serum samples were retrieved from blood bank to assay for anti-HCV, and frozen baseline plasma samples were retrieved to assay for glucose, insulin, and HBV-related factors that included HBeAg, viral load, genotype and the BCP double mutations.
Follow-up assessment included ultrasonography and biochemical assay. High-resolution real-time ultrasonography was performed routinely since 1993, serum ALT and aspartate aminotransferase were examined routinely after August 1994, and γ-glutamyltransferase (GGT), cholesterol and triglycerides after August 1996. We retrospectively measured HBV viral load in all the plasma samples (2–12 samples per subject) taken from each subject during follow-up until the date of the case's diagnosis (if before 31 December 2004) or 31 July 2005, as described in our previous study (5). To identify change in diabetes status, plasma samples collected during 2–5 years after entry into the cohort from participants were retrieved to assay for glucose.
HBsAg, anti-HCV and HBeAg were assayed using commercially available immunoassay kits, whereas plasma HBV DNA, viral genotype and the BCP double mutations were assayed by polymerase chain reaction-based methods as described previously (5–7). Plasma insulin concentrations were determined by use of the Siemens Coat-A-Count Insulin radioimmunoassay (Siemens Medical Solutions Diagnostics, Los Angeles, CA). Glucose concentrations were determined using an automatic dry-chemical analyzer (VITROS 5,1 FS; Johnson & Johnson, New Brunswick, NJ). We used the homeostasis model assessment-insulin resistance (HOMA-IR) index to estimate IR [HOMA-IR index = fasting insulin (μU/ml) X fasting glucose (mmol/l)/22.5] (25).
Statistical analysis
Plasma insulin and IR were categorized according to previous studies (26,27) or in quartiles based on the distribution of subjects. For cross-tabulated data, a Mantel–Haenszel extension of the χ2 test for trend was used to examine the dose–response relationship across the four insulin groups. Continuous data were compared using analysis of variance or the Kruskal–Wallis test. The Cox proportional hazards model, adapted for the case–cohort design (28), was used to estimate the hazard ratios (HRs) for incident HCC. We estimated HRs according to insulin groups for two categories of time (<8 and ≥8 years) with the cutoff that approximately corresponds to the median of follow-up time of cases. Unconditional logistic regression was used to determine associations with liver abnormalities diagnosed by ultrasonography or biochemical tests; these results are reported as odds ratios and 95% confidence intervals (CIs). In multivariate models, factors known or suspected to influence HCC incidence were considered as covariates. Trend tests were conducted using the category medians as the scores. Effect modification was assessed by stratified analyses and by including cross product terms of the primary factor and the variable of interest in multivariable models. Tests for interaction to identify heterogeneity of the insulin effect were computed based on likelihood ratio tests. A two-sided P value of <0.05 was considered to indicate statistical significance. All analyses were conducted with SAS software 9.1.3 (SAS Institute, Cary, NC).
Results
Baseline characteristics
Only 31 (2.7%) of the 1142 subjects reported a history of diabetes. After adjusting for age, HCC cases were more probably to have a higher prediagnostic value of insulin, HOMA-IR and HBV viral load than those free from HCC during follow-up (all P < 0.0001), whereas the fasting glucose levels did not significantly differ between cases and non-cases. Cases also had a higher frequency of BMI ≥ 25, elevated ALT, HBeAg positivity, HBV genotype C infection, BCP double mutations and a first-degree family history of HCC than non-cases (all P < 0.02) (Table I). With increasing fasting levels of insulin, weight, BMI, blood pressure, fasting levels of glucose, HOMA-IR, the proportion of subjects reporting a history of diabetes and the prevalence of elevated ALT all increased (all P for trend <0.02). However, insulin levels were not associated with any hepatitis B viral factors (data not shown).
Table I.
Baseline characteristics of HCC cases and non-cases
| Characteristic | HCC (n = 124) |
Non-cases (n = 1018) |
Pa | ||
| Demographics | |||||
| Age, years, No. (%) | <0.0001 | ||||
| 30–39 | 15 | (12.1) | 369 | (36.3) | |
| 40–49 | 44 | (35.5) | 394 | (38.7) | |
| 50–59 | 31 | (25.0) | 159 | (15.6) | |
| ≥60 | 34 | (27.4) | 96 | (9.4) | |
| College education or above, No./total (%) | 92/124 | (74.2) | 845/1017 | (83.1) | 0.0753 |
| Alcohol consumption, No. (%) | 29 | (23.4) | 210 | (20.6) | 0.7078 |
| Alcohol consumption >140 g/week, No./total (%) | 11/122 | (9.0) | 84/1009 | (8.3) | 0.5868 |
| Smoking status, No. (%) | 0.4211b | ||||
| Never | 74 | (59.7) | 712 | (69.9) | |
| Former | 19 | (15.3) | 89 | (8.7) | |
| Current | 31 | (25.0) | 217 | (21.3) | |
| First-degree family history of HCC, No. (%) | 22 | (17.7) | 64 | (6.3) | <0.0001 |
| Metabolic factors | |||||
| History of diabetes mellitus, No. (%) | 8 | (6.5) | 23 | (2.3) | 0.0792 |
| Height, mean (SD), cm | 167.6 | (5.2) | 168.1 | (5.8) | 0.5107 |
| Weight, mean (SD), kg | 67.8 | (9.4) | 67.1 | (8.6) | 0.2330 |
| BMI, mean (SD), kg/m2 | 24.1 | (2.9) | 23.7 | (2.6) | 0.0677 |
| BMI≥25 kg/m2, No. (%) | 49 | (39.5) | 302 | (29.7) | 0.0145 |
| Fasting plasma glucose, mean (SD), mg/dL | 93.5 | (16.0) | 88.7 | (17.0) | 0.9325 |
| Fasting plasma glucose≥126 mg/dL, No. (%) | 5 | (4.0) | 20 | (2.0) | 0.4362 |
| Fasting plasma insulin, mean (SD), μU/ml | 5.68 | (5.91) | 4.33 | (4.32) | <0.0001 |
| IR (HOMA-IR), mean (SD) | 1.35 | (1.53) | 0.98 | (1.09) | <0.0001 |
| Clinical and viral features | |||||
| ALT >40 U/l, No./total (%) | 33/123 | (26.8) | 64/1015 | (6.3) | <0.0001 |
| Plasma HBV DNA, mean (SD), log10 copies/ml | 5.93 | (2.02) | 4.37 | (1.89) | <0.0001 |
| HBeAg positivity, No./total (%) | 20/122 | (16.4) | 89/993 | (9.0) | <0.0001 |
| HBV genotype, total no. | 123 | 1002 | <0.0001b | ||
| B, No. (%) | 54 | (43.9) | 818 | (81.6) | |
| B + C, No. (%) | 6 | (4.9) | 45 | (4.5) | |
| C, No. (%) | 63 | (51.2) | 139 | (13.9) | |
| BCP double mutations, No./total (%) | 74/118 | (62.7) | 295/983 | (30.0) | <0.0001 |
Except for age, significant test was performed by using Cox regression adjusted for age (continuous).
P value was calculated based on the likelihood ratio test.
Metabolic factors and follow-up ultrasonographic and biochemical features
In total, 1142 subjects contributed 8052 person-examinations (median, eight per person). The number of visits was not significantly different between insulin groups. Increasing levels of insulin were associated with elevated levels of fasting glucose, triglycerides and the occurrence of diabetes (P for trend <0.005). After adjustment for the number of visits, demographics and viral factors, there was a monotonic rise in the risk for elevated levels of ALT during follow-up with increasing insulin (P for trend = 0.0006). But a J-shaped relation was observed between insulin and the risk of liver cirrhosis detected by ultrasonography, such that both high and low levels of insulin were found to be associated with increased risk (Table II). When the number of visits, demographics and other metabolic factors, including glucose, BMI and a history of diabetes, were included as covariates to adjust their effects, the associations for insulin with elevated ALT (HR for >6.10 versus 2.75–4.10 μU/ml = 1.64, 95% CI: 1.09–2.46, P for trend = 0.0202) and liver cirrhosis (HR for >6.10 versus 2.75–4.10 μU/ml = 2.84, 95% CI: 1.45–5.57, P for trend = 0.0843) did not materially change. But the associations for insulin levels with the risks of GGT elevation and fatty liver on ultrasonography were greatly attenuated, and the association between insulin and GGT became statistically non-significant (data not shown).
Table II.
Ultrasonographic and biochemical features assayed during follow-up by baseline fasting plasma insulin levels
| Characteristic | Fasting plasma insulin levels (μU/ml) |
||||
| <2.75 (n = 395) | 2.75–4.10 (n = 265) | 4.11–6.10 (n = 259) | >6.10 (n = 223) | P for trend | |
| No. of visits, median (interquartile range) | 8 (5–9) | 8 (5–9) | 7 (5–9) | 7 (5–9) | 0.4081a |
| Self-reported diabetes, No./total (%)b | 5/372 (1.3) | 6/250 (2.4) | 6/240 (2.5) | 23/209 (11.0) | <0.0001 |
| Plasma glucose, mean (SD), mg/dlc | 87.2 (20.4) | 87.8 (28.1) | 87.2 (27.7) | 99.1 (57.4) | 0.0002d |
| Plasma glucose≥126 mg/dl, No./total (%)c | 12/375 (3.2) | 7/253 (2.8) | 7/241 (2.9) | 19/209 (9.1) | 0.0044 |
| Newly diagnosed diabetes, No./total (%)e | 9/368 (2.5) | 6/246 (2.4) | 7/236 (3.0) | 15/194 (7.7) | 0.0048 |
| Total cholesterol >200 mg/dl, No./total (%) | 187/343 (54.5) | 135/221 (61.1) | 103/213 (48.4) | 107/185 (57.8) | 0.8763 |
| Triglycerides >200 mg/dl, No./total (%) | 43/343 (12.5) | 26/221 (11.8) | 32/213 (15.0) | 46/185 (24.9) | 0.0005 |
| Liver biochemistry | |||||
| ALT, total No. | 358 | 242 | 225 | 200 | |
| Single elevation, No. (%) | 68 (19.0) | 44 (18.2) | 45 (20.0) | 44 (22.0) | |
| Multiple elevations, No. (%) | 68 (19.0) | 44 (18.2) | 52 (23.1) | 57 (28.5) | |
| ALT elevation ≥1 visit | |||||
| Multivariate OR (95% CI)f | 0.84 (0.57–1.23) | 1 (Ref) | 1.13 (0.74–1.72) | 1.68 (1.09–2.58) | 0.0006 |
| Aspartate aminotransferase, total No. | 358 | 242 | 225 | 200 | |
| Single elevation, No. (%) | 59 (16.5) | 37 (15.3) | 41 (18.2) | 36 (18.0) | |
| Multiple elevations, No. (%) | 47 (13.1) | 24 (9.9) | 39 (17.3) | 33 (16.5) | |
| Aspartate aminotransferase elevation≥1 visit | |||||
| Multivariate OR (95% CI)f | 1.01 (0.67–1.53) | 1 (Ref) | 1.39 (0.89–2.17) | 1.32 (0.84–2.10) | 0.1215 |
| GGT, total No. | 342 | 221 | 213 | 185 | |
| Single elevation, No. (%) | 35 (10.2) | 25 (11.3) | 23 (10.8) | 24 (13.0) | |
| Multiple elevations, No. (%) | 29 (8.5) | 12 (5.4) | 18 (8.5) | 25 (13.5) | |
| GGT elevation≥1 visit | |||||
| Multivariate OR (95% CI)f | 0.94 (0.58–1.53) | 1 (Ref) | 0.96 (0.56–1.65) | 1.64 (0.97–2.77) | 0.0241 |
| Ultrasound scanning, total No. | 372 | 253 | 236 | 214 | |
| Fatty liver, No. (%) | 204 (54.8) | 182 (71.9) | 163 (69.1) | 169 (79.0) | |
| Multivariate OR (95% CI)f | 0.45 (0.31–0.66) | 1 (Ref) | 0.87 (0.57–1.32) | 1.79 (1.12–2.87) | <0.0001 |
| Liver cirrhosis, No. (%) | 41 (11.0) | 15 (5.9) | 34 (14.4) | 34 (15.9) | |
| Multivariate OR (95% CI)f | 1.78 (0.88–3.57) | 1 (Ref) | 2.49 (1.21–5.15) | 2.55 (1.22–5.33) | 0.0886 |
Total number of subjects may vary because of missing values due to non-participation in the follow-up examinations.
Statistical significance was examined by the Kruskal–Wallis test.
Self-reported history of diabetes obtained by follow-up questionnaire in 1071 subjects.
Fasting glucose levels were tested in plasma samples collected during 2–5 years after study entry from 1078 subjects.
Significant test by one-way analysis of variance.
Defined by the criteria of self-reported history of diabetes or fasting plasma glucose level of at least 126 mg/dl.
Odds ratios (ORs) and 95% CIs were adjusted for number of visits, age (continuous), smoking, alcohol consumption, a first-degree family history of HCC and baseline viral factors, including HBV viral load (continuous), genotype (C versus non-C), HBeAg status and the BCP double mutations.
In subjects with a BMI ≥25, the risks for ALT (HR = 1.48, 95% CI = 1.10–2.00) or GGT (HR = 1.66, 95% CI = 1.14–2.39) elevation and fatty liver (HR = 2.46, 95% CI = 1.75–3.47) were significantly higher than in those with a BMI <25 after adjustment for the number of visits, demographics and other metabolic factors including insulin, confirming our previous observation from using the entire cohort (12). However, the risk for cirrhosis related to a BMI ≥25 was greatly attenuated and became statistically non-significant after adjustment for insulin. After adjustment for insulin, we found no association with any liver abnormality for diabetes or fasting levels of glucose determined in plasma samples collected at baseline or during 2–5 years after study entry (data not shown).
Metabolic factors and HCC
We also found a J-shaped relation between insulin or the HOMA-IR index and incident HCC, irrespective of using cutpoints as described previously (26,27) or quartile cutpoints. In a model using insulin of 2.75–4.10 μU/ml as the referent group adjusted for differences among the four insulin groups including the number of visits, demographic characteristics and viral factors, the HR of HCC was 1.57 (95% CI: 0.96–2.58) for the lowest category of insulin (<2.75 μU/ml) and 2.36 (95% CI: 1.43–3.90) for the highest category of insulin (>6.10 μU/ml) (Table III). This association was largely unaltered by adjustment for demographics and other metabolic factors (>6.10 versus 2.75–4.10 μU/ml, HR = 2.24, 95% CI = 1.37–3.67). Exclusion of men reporting a history of diabetes at baseline did not substantially change the results. In contrast, after adjustment for demographics and other metabolic factors including insulin, the association between a BMI ≥25 and HCC became statistically non-significant (HR = 1.35, 95% CI: 0.98–1.86, P = 0.0648). Also, neither fasting glucose nor a history of diabetes measured at baseline or follow-up was associated with HCC (data not shown).
Table III.
HRs for incident HCC by metabolic factors assayed at baseline
| Metabolic factors | HCC (n=124) |
Non-cases (n=1018) |
HRa | 95% CI | P for trend | ||
| No. | (%) | No. | (%) | ||||
| Fasting insulin levels, μU/ml | |||||||
| Category cutpoints | 0.0140 | ||||||
| <2.75 | 41 | (33.1) | 354 | (34.8) | 1.57 | (0.96–2.58) | |
| 2.75–4.10 | 16 | (12.9) | 249 | (24.5) | 1 | (Ref) | |
| 4.11–6.10 | 29 | (23.4) | 230 | (22.6) | 1.28 | (0.76–2.15) | |
| >6.10 | 38 | (30.6) | 185 | (18.2) | 2.36 | (1.43–3.90) | |
| Quartile cutpoints | 0.0556 | ||||||
| <2.27 | 32 | (25.8) | 250 | (24.6) | 1.43 | (0.89–2.30) | |
| 2.27–3.60 | 21 | (16.9) | 266 | (26.1) | 1 | (Ref) | |
| 3.61–5.36 | 26 | (21.0) | 263 | (25.8) | 0.95 | (0.58–1.56) | |
| >5.36 | 45 | (36.3) | 239 | (23.5) | 1.78 | (1.15–2.75) | |
| Fasting glucose levels, mg/dl | 0.0161 | ||||||
| <110 | 108 | (87.1) | 955 | (93.8) | 1 | (Ref) | |
| 110–125 | 11 | (8.9) | 43 | (4.2) | 1.40 | (0.80–2.45) | |
| ≥126 | 5 | (4.0) | 20 | (2.0) | 2.37 | (1.12–5.04) | |
| History of diabetes by questionnaire | |||||||
| No | 116 | (93.5) | 995 | (97.7) | 1 | (Ref) | |
| Yes | 8 | (6.5) | 23 | (2.3) | 2.11 | (1.13–3.94) | 0.0192b |
| IR (HOMA-IR) | |||||||
| Category cutpoints | 0.0382 | ||||||
| <0.70 | 48 | (38.7) | 459 | (45.1) | 1.30 | (0.82–2.08) | |
| 0.70–1.02 | 19 | (15.3) | 229 | (22.5) | 1 | (Ref) | |
| 1.03–1.53 | 26 | (21.0) | 190 | (18.7) | 1.20 | (0.73–1.98) | |
| >1.53 | 31 | (25.0) | 140 | (13.8) | 1.97 | (1.20–3.22) | |
| Quartile cutpoints | 0.0175 | ||||||
| <0.46 | 31 | (25.0) | 250 | (24.6) | 1.48 | (0.90–2.44) | |
| 0.46–0.77 | 20 | (16.1) | 270 | (26.5) | 1 | (Ref) | |
| 0.78–1.22 | 23 | (18.5) | 262 | (25.7) | 0.92 | (0.54–1.55) | |
| >1.22 | 50 | (40.3) | 236 | (23.2) | 1.96 | (1.23–3.10) | |
| BMI ≥25 kg/m2 | |||||||
| No | 75 | (60.5) | 716 | (70.3) | 1 | (Ref) | |
| Yes | 49 | (39.5) | 302 | (29.7) | 1.58 | (1.16–2.17) | 0.0041b |
HR and 95% CIs were adjusted for number of visits, age (continuous), smoking, alcohol consumption, a first-degree family history of HCC and baseline viral factors, including HBV viral load (continuous), genotype (C versus non-C), HBeAg status and the BCP double mutations.
P values were obtained by the Wald test.
Insulin and HCC in selected subgroups
The median lag time measured as the interval from blood draw to HCC diagnosis was 7.5 (range, 0.6–16.3) years. High insulin level (≥4.11 μU/ml) was more strongly associated with increased HCC risk among cases with a lag time of at least 8 years, whereas the lowest category of insulin was more strongly associated with HCC among cases with a lag time of <8 years (Table IV).
Table IV.
Association of insulin with incident HCC stratified by time and selected risk factors
| Fasting plasma insulin levels (μU/ml) |
P for trend | |||||
| Factor, group | <2.75 | 2.75–4.10 | 4.11–6.10 | >6.10 | ||
| Follow-up, years | ||||||
| <8 (n = 1082) | HR (95% CI)a | 2.11 (1.17–3.82) | 1 (Ref) | 1.34 (0.69–2.61) | 1.79 (0.91–3.49) | 0.5785 |
| ≥8 (n = 1061) | HR (95% CI)a | 1.33 (0.62–2.88) | 1 (Ref) | 2.32 (1.11–4.85) | 4.15 (2.05–8.39) | <0.0001 |
| Age, years | ||||||
| <50 (n = 806) | HR (95% CI)a | 1.56 (0.83–2.93) | 1 (Ref) | 0.93 (0.45–1.95) | 2.24 (1.17–4.28) | 0.1400 |
| ≥50 (n = 336) | HR (95% CI)a | 2.25 (1.11–4.53) | 1 (Ref) | 2.82 (1.40–5.66) | 3.13 (1.55–6.30) | 0.0436 |
| BMI, kg/m2 | ||||||
| <25 (n = 791) | HR (95% CI)a | 2.07 (1.19–3.59) | 1 (Ref) | 1.95 (1.06–3.60) | 2.22 (1.17–4.23) | 0.6184 |
| ≥25 (n = 351) | HR (95% CI)a | 1.43 (0.55–3.71) | 1 (Ref) | 1.02 (0.44–2.37) | 2.12 (0.97–4.64)b | 0.0306 |
| <23c (n = 455) | HR (95% CI)a | 1.63 (0.83–3.24) | 1 (Ref) | 1.48 (0.65–3.38) | 2.70 (1.17–6.23) | 0.2269 |
| ≥23 (n = 687) | HR (95% CI)a | 2.20 (1.15–4.22) | 1 (Ref) | 1.83 (0.98–3.43) | 2.77 (1.51–5.07) | 0.0394 |
| HBV genotype | ||||||
| C (n = 202) | HR (95% CI)a | 1.28 (0.67–2.45) | 1 (Ref) | 1.10 (0.56–2.19) | 2.28 (1.20–4.34) | 0.0223 |
| B or B + C (n = 923) | HR (95% CI)a | 2.46 (1.20–5.02) | 1 (Ref) | 2.13 (1.01–4.49) | 3.38 (1.63–6.98) | 0.0747 |
| Baseline HBV viral load, log10 copies/mld | ||||||
| <4.39 (n = 666) | HR (95% CI)a | 1.68 (0.63–4.46) | 1 (Ref) | 1.85 (0.65–5.25) | 6.15 (2.48–15.22) | <0.0001 |
| ≥4.39 (n = 476) | HR (95% CI)a | 1.71 (1.00–2.94) | 1 (Ref) | 1.43 (0.82–2.51) | 1.49 (0.84–2.66) | 0.7928 |
| Time trend for HBV viral loade | ||||||
| Sustained | ||||||
| Low (n = 686) | HR (95% CI)a | 3.31 (1.13–9.71) | 1 (Ref) | 1.90 (0.57–6.34) | 5.80 (1.98–17.00) | 0.0337 |
| Sustained | ||||||
| High (n = 366) | HR (95% CI)a | 2.10 (1.13–3.89) | 1 (Ref) | 1.74 (0.93–3.28) | 1.93 (1.02–3.63) | 0.7497 |
| Extremely | ||||||
| High to low (n = 90) | HR (95% CI)a | 0.43 (0.14–1.37) | 1 (Ref) | 0.87 (0.27–2.86) | 0.84 (0.26–2.75) | 0.2363 |
| BCP double mutations | ||||||
| Absence (n = 732) | HR (95% CI)a | 1.74 (0.71–4.27) | 1 (Ref) | 2.85 (1.20–6.78) | 4.12 (1.77–9.60) | 0.0007 |
| Presence (n = 369) | HR (95% CI)a | 1.97 (1.10–3.52) | 1 (Ref) | 1.57 (0.83–2.95) | 2.31 (1.25–4.28) | 0.3359 |
HRs and 95% CIs were adjusted for number of visits, age (continuous), smoking, alcohol consumption and a first-degree family history of HCC.
P = 0.0607.
Asian BMI criterion of overweight (29).
The cutpoint is the threshold of viral load associated with increased risk for HCC in previous studies (3,5). P for heterogeneity of HRs by HBV viral load = 0.0021.
Three patterns of time trend for viral load (‘sustained low’: having a level of 3–4 log10 copies/ml over time; ‘sustained high’: having a level of 5–6 log10 copies/ml over time and ‘extremely high to low’: having a level of 8–9 log10 copies/ml at study entry that declined linearly over time) were defined according to our previous longitudinal viral-load study (6), in which trajectory analysis, a type of latent class analysis which identifies homogeneous groups within a heterogeneous population assumed to contain multiple latent trajectories, was performed with the SAS procedure PROC TRAJ to estimate trajectories of viral load across 16 years of follow-up. P for heterogeneity of HRs by time trends for HBV viral load = 0.0629.
We also performed analyses stratified according to established risk factors for HCC (3–8,12,13) (Table IV). The effect of an insulin level >6.10 μU/ml (versus 2.75–4.10 μU/ml) in increasing HCC risk was consistent across subgroups of age, BMI, HBV genotype and the presence or absence of the BCP double mutations, but its effect varied by baseline viral load of HBV (P for heterogeneity = 0.0021), with relative risk estimate being larger among men with viral load <4.39 log10 copies/ml, the threshold of viral load associated with increased risk for HCC in previous studies (3,5) (HR: 6.15, 95% CI: 2.48–15.22), than among men with a higher viral load (HR: 1.49, 95% CI: 0.84–2.66). According to our previous study (6), there are three trajectory classes for the 16-year time trends of HBV viral load. The positive association of insulin with HCC was also stronger among men with sustained low viral load in the levels of 3–4 log10 copies/ml than among those with sustained high viral load in the levels of 5–6 log10 copies/ml or those characterized by extremely high viral load in the levels of 8–9 log10 copies/ml at recruitment that declined linearly over time (P for heterogeneity = 0.0629).
Discussion
Fatty liver is common in middle-aged HBV carriers with a prevalence of 30–90%, depending on BMI (12,30,31). Simple fatty liver without inflammation was not a predictor for the risk of HCC among male HBV carriers (12). Inasmuch, it remains of interest to elucidate the interaction between NAFLD and HBV viral factors in the progression of hepatitis B.
Overweight and obesity have been associated with an increased risk to develop HCC worldwide (12,13). IR, which leads to the accumulation of fat within hepatocytes, is linked to both excess BMI and NAFLD. Hepatic fat accumulation produces oxidative stress, resulting in inflammation and fibrosis (11,32). In experimental models, insulin stimulates cellular proliferation in cancer cell lines and promotes tumor growth in mice (33–35). Recent researches have demonstrated higher risk of several malignancies among individuals with higher insulin levels, a consequence of IR (26,27,36,37). However, the role of excess body weight compared with that of higher insulin in inducing HCC and its related conditions remains unknown.
To our knowledge, this is the first study to demonstrate a prospective association between fasting insulin levels and IR with the risk of HCC. After a maximum follow-up of 17 years, higher insulin was found to be strongly related to increased risk of incident HCC and related conditions, including elevated liver enzymes and liver cirrhosis detected by ultrasonography, among male HBV carriers. Overall, there was a J-shaped relation between insulin and HCC. A similar pattern was seen with the effect of the HOMA-IR index, which is highly dependent on insulin (25). The highest insulin levels (>6.10 μU/ml) were seen to be associated with an ∼2.4-fold risk of HCC when compared with insulin 2.75–4.10 μU/ml. However, the lowest insulin levels (<2.75 μU/ml) also conferred a 1.6-fold risk for HCC. The dose–response effect of insulin appeared to be evident for an increased risk of HCC that occurred at least 8 years after the baseline blood draw.
There was also a J-shaped association between insulin and the risk of liver cirrhosis detected by ultrasonography, which is an important premalignant condition for HCC (12,38), whereas the rise in the risk for ALT elevation with increasing insulin was more monotonic. These findings raise the possibility that liver destruction in certain hepatitis B cirrhotic patients, even in the absence of overt clinical signs, may lead to diminished insulin levels many years before HCC diagnosis. Consistent with this hypothesis, a previous study indicated that liver damage from hepatic tumor is associated with a reduction in circulating levels of insulin-like growth factor-I, a hormone related to insulin (39). In our data, excluding cases diagnosed with HCC during the first 8 years of follow-up revealed that the increased HCC risk was only attenuated among men at the low end of insulin, but, contrary to this finding, the risk among those in the highest category was even greater. Therefore, the possibility of the elevated risk in the lowest category of insulin due to antecedent disease cannot be excluded, and high insulin may exert its effect on the early stage of HCC development.
In endemic area of hepatitis B, literature on the association of HCC with diabetes is contradictory (12,22–24). A Korean cohort study followed 1.3 million people for up to 10 years and indicated a 1.5-fold increased HCC risk with glucose-defined diabetes (≥126 mg/dl) in men and, to a lesser degree for women (24). However, the study did not include the assessment of HBsAg or anti-HCV. Whether the findings of the Korean study can be generalizable to HBV carriers is unclear. Besides our previous cohort study (12), other two Taiwanese cohort studies have examined the association between diabetes and HCC, and the findings did not suggest an effect of diabetes in HBV carriers (22,23). Most of the HCC cases in this study were non-diabetic at baseline or during follow-up. We observed that adjustment for BMI, glucose and a history of diabetes only slightly attenuated the positive association between higher insulin and HCC. In contrast, despite a significant association observed between HCC and a history of diabetes or glucose-defined diabetes after adjustment for various viral factors, neither diabetes nor glucose was statistically associated with HCC risk after adjusting for insulin. The results suggest a more direct role of elevated fasting insulin than diabetes in the development of HCC in chronic hepatitis B.
The present study extended our previous work that demonstrated overweight as a risk factor for HBV-related HCC (12). However, because a BMI ≥25 was no longer associated with elevated HCC risk after adjusting for insulin, the previously reported association may probably be largely explained by the high levels of insulin in overweight subjects. Our finding of a relationship between insulin and HCC across BMI groups again strengthens the relative importance of high insulin level versus excess body weight in determining HCC. We showed that BMI was associated with fatty liver and elevation of liver enzymes but not with more advanced liver disease in terms of cirrhosis and HCC after adjusting for insulin, indicating that excess body weight may have a greater role in causing hepatic steatosis and inflammation than following advanced illness.
Among the strengths of the present study is the characterization of various viral factors, which enabled evaluation of potential confounding and effect modification by these important risk factors for HBV-related HCC (2–8). We observed that the elevated HCC risk associated with higher insulin was largely unaltered when controlling various viral factors, and there existed an effect of higher insulin across subgroups categorized by HBV genotype and the presence or absence of the BCP double mutations. However, the HR for HCC associated with higher insulin was greater in male HBV carriers with low viral load than in those with high viral load. Currently, the guidelines for antiviral treatment in hepatitis B are based on HBV viral load >105 copies/ml and ALT levels more than two times the upper limit of normal (40). However, HBV carriers with normal ALT and viral load <104 copies/ml are still at a substantial risk of HCC (41). In light of our findings, the further reduction of the HCC burden would be achievable by insulin-lowering approaches, either through lifestyle changes or via pharmacologic approaches.
In conclusion, among male HBV carriers, higher-fasting insulin levels are a risk factor for the development of HCC and its related conditions, including cirrhosis and hepatitis. The analysis suggested little confounding by other metabolic factors, including BMI, fasting glucose and diabetes, as well as various viral factors, but the observed association of HCC risk with higher insulin levels was stronger in low than in high viral load group.
Funding
National Science Council, Taiwan (NSC 97-3112-B-002-028, NSC 98-3112-B-002-021 and NSC 98-2628-B-002-089-MY3).
Acknowledgments
Conflict of Interest Statement: None declared.
Glossary
Abbreviations
- ALT
alanine aminotransferase
- BCP
basal core promoter
- BMI
body mass index
- CI
confidence interval
- GGT
γ-glutamyltransferase
- HBeAg
HBV e antigen
- HBV
hepatitis B virus
- HBsAg
HBV surface antigen
- HCC
hepatocellular carcinoma
- HCV
hepatitis C virus
- HR
hazard ratio
- HOMA-IR
homeostasis model assessment-insulin resistance
- IR
insulin resistance
- NAFLD
non-alcoholic fatty liver disease
References
- 1.Kao JH, et al. Global control of hepatitis B virus infection. Lancet Infect. Dis. 2002;2:395–403. doi: 10.1016/s1473-3099(02)00315-8. [DOI] [PubMed] [Google Scholar]
- 2.Chen YC, et al. Age-specific prognosis following spontaneous hepatitis B e antigen seroconversion in chronic hepatitis B. Hepatology. 2010;51:435–444. doi: 10.1002/hep.23348. [DOI] [PubMed] [Google Scholar]
- 3.Yu MW, et al. Hepatitis B virus genotype and DNA level and hepatocellular carcinoma: a prospective study in men. J. Natl Cancer Inst. 2005;97:265–272. doi: 10.1093/jnci/dji043. [DOI] [PubMed] [Google Scholar]
- 4.Chan HL, et al. High viral load and hepatitis B virus subgenotype ce are associated with increased risk of hepatocellular carcinoma. J. Clin. Oncol. 2008;26:177–182. doi: 10.1200/JCO.2007.13.2043. [DOI] [PubMed] [Google Scholar]
- 5.Wu CF, et al. Long-term tracking of hepatitis B viral load and the relationship with risk for hepatocellular carcinoma in men. Carcinogenesis. 2008;29:106–112. doi: 10.1093/carcin/bgm252. [DOI] [PubMed] [Google Scholar]
- 6.Sung FY, et al. Hepatitis B virus core variants modify natural course of viral infection and hepatocellular carcinoma progression. Gastroenterology. 2009;137:1687–1697. doi: 10.1053/j.gastro.2009.07.063. [DOI] [PubMed] [Google Scholar]
- 7.Chou YC, et al. Temporal relationship between hepatitis B virus enhancer II/basal core promoter sequence variation and risk of hepatocellular carcinoma. Gut. 2008;57:91–97. doi: 10.1136/gut.2006.114066. [DOI] [PubMed] [Google Scholar]
- 8.Yuen MF, et al. Independent risk factors and predictive score for the development of hepatocellular carcinoma in chronic hepatitis B. J. Hepatol. 2009;50:80–88. doi: 10.1016/j.jhep.2008.07.023. [DOI] [PubMed] [Google Scholar]
- 9.Becker U, et al. Prediction of risk of liver disease by alcohol intake, sex, and age: a prospective population study. Hepatology. 1996;23:1025–1029. doi: 10.1002/hep.510230513. [DOI] [PubMed] [Google Scholar]
- 10.Hossain P, et al. Obesity and diabetes in the developing world-a growing challenge. N. Engl. J. Med. 2007;356:213–215. doi: 10.1056/NEJMp068177. [DOI] [PubMed] [Google Scholar]
- 11.Farrell GC, et al. Nonalcoholic fatty liver disease: from steatosis to cirrhosis. Hepatology. 2006;43(2 suppl. 1):S99–S112. doi: 10.1002/hep.20973. [DOI] [PubMed] [Google Scholar]
- 12.Yu MW, et al. Body-mass index and progression of hepatitis B: a population-based cohort study in men. J. Clin. Oncol. 2008;26:5576–5582. doi: 10.1200/JCO.2008.16.1075. [DOI] [PubMed] [Google Scholar]
- 13.Larsson SC, et al. Overweight, obesity and risk of liver cancer: a meta-analysis of cohort studies. Br. J. Cancer. 2007;97:1005–1008. doi: 10.1038/sj.bjc.6603932. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Wong GL, et al. Metabolic syndrome increases the risk of liver cirrhosis in chronic hepatitis B. Gut. 2009;58:111–117. doi: 10.1136/gut.2008.157735. [DOI] [PubMed] [Google Scholar]
- 15.Moucari R, et al. Insulin resistance in chronic hepatitis C: association with genotypes 1 and 4, serum HCV RNA level, and liver fibrosis. Gastroenterology. 2008;134:416–423. doi: 10.1053/j.gastro.2007.11.010. [DOI] [PubMed] [Google Scholar]
- 16.Cua IH, et al. Genotype-specific interactions of insulin resistance, steatosis, and fibrosis in chronic hepatitis C. Hepatology. 2008;48:723–731. doi: 10.1002/hep.22392. [DOI] [PubMed] [Google Scholar]
- 17.Ogunleye AA, et al. A cohort study of the risk of cancer associated with type 2 diabetes. Br. J. Cancer. 2009;101:1199–1201. doi: 10.1038/sj.bjc.6605240. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.El-Serag HB, et al. Diabetes increases the risk of chronic liver disease and hepatocellular carcinoma. Gastroenterology. 2004;126:460–468. doi: 10.1053/j.gastro.2003.10.065. [DOI] [PubMed] [Google Scholar]
- 19.Adami HO, et al. Excess risk of primary liver cancer in patients with diabetes mellitus. J. Natl Cancer Inst. 1996;88:1472–1477. doi: 10.1093/jnci/88.20.1472. [DOI] [PubMed] [Google Scholar]
- 20.Wideroff L, et al. Cancer incidence in a population-based cohort of patients hospitalized with diabetes mellitus in Denmark. J. Natl Cancer Inst. 1997;89:1360–1365. doi: 10.1093/jnci/89.18.1360. [DOI] [PubMed] [Google Scholar]
- 21.Stocks T, et al. Blood glucose and risk of incident and fatal cancer in the metabolic syndrome and cancer project (Me-Can): analysis of six prospective cohorts. PLoS Med. 2009;6:e1000201. doi: 10.1371/journal.pmed.1000201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Wang CS, et al. The impact of type 2 diabetes on the development of hepatocellular carcinoma in different viral hepatitis statuses. Cancer Epidemiol. Biomarkers Prev. 2009;18:2054–2060. doi: 10.1158/1055-9965.EPI-08-1131. [DOI] [PubMed] [Google Scholar]
- 23.Lai MS, et al. Type 2 diabetes and hepatocellular carcinoma: a cohort study in high prevalence area of hepatitis virus infection. Hepatology. 2006;43:1295–1302. doi: 10.1002/hep.21208. [DOI] [PubMed] [Google Scholar]
- 24.Jee SH, et al. Fasting serum glucose level and cancer risk in Korean men and women. JAMA. 2005;293:194–202. doi: 10.1001/jama.293.2.194. [DOI] [PubMed] [Google Scholar]
- 25.Matthews DR, et al. Homeostasis model assessment: insulin resistance and beta-cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia. 1985;28:412–419. doi: 10.1007/BF00280883. [DOI] [PubMed] [Google Scholar]
- 26.Stolzenberg-Solomon RZ, et al. Insulin, glucose, insulin resistance, and pancreatic cancer in male smokers. JAMA. 2005;294:2872–2878. doi: 10.1001/jama.294.22.2872. [DOI] [PubMed] [Google Scholar]
- 27.Albanes D, et al. Serum insulin, glucose, indices of insulin resistance, and risk of prostate cancer. J. Natl Cancer Inst. 2009;101:1272–1279. doi: 10.1093/jnci/djp260. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Therneau TM, et al. Computing the Cox model for case cohort designs. Lifetime Data Anal. 1999;5:99–112. doi: 10.1023/a:1009691327335. [DOI] [PubMed] [Google Scholar]
- 29.Haslam DW, et al. Obesity. Lancet. 2005;366:1197–1209. doi: 10.1016/S0140-6736(05)67483-1. [DOI] [PubMed] [Google Scholar]
- 30.Mellen JS, et al. The clinical presentation of chronic hepatitis B virus (HBV) infection in Asian Americans: a single center retrospective study. J. Clin. Gastroenterol. 2010;44:364–370. doi: 10.1097/MCG.0b013e3181b5c7a8. [DOI] [PubMed] [Google Scholar]
- 31.Peng D, et al. Hepatic steatosis in chronic hepatitis B patients is associated with metabolic factors more than viral factors. J. Gastroenterol. Hepatol. 2008;23:1082–1088. doi: 10.1111/j.1440-1746.2008.05478.x. (7 Pt 1) [DOI] [PubMed] [Google Scholar]
- 32.Angulo P. Nonalcoholic fatty liver disease. N. Engl. J. Med. 2002;346:1221–1231. doi: 10.1056/NEJMra011775. [DOI] [PubMed] [Google Scholar]
- 33.Giovannucci E. Insulin, insulin-like growth factors and colon cancer: a review of the evidence. J. Nutr. 2001;131(suppl. 11):3109S–3120S. doi: 10.1093/jn/131.11.3109S. [DOI] [PubMed] [Google Scholar]
- 34.Chappell J, et al. Effect of insulin on cell cycle progression in MCF-7 breast cancer cells. Direct and potentiating influence. J. Biol. Chem. 2001;276:38023–38028. doi: 10.1074/jbc.M104416200. [DOI] [PubMed] [Google Scholar]
- 35.Shafie SM, et al. Role of hormones in the growth and regression of human breast cancer cells (MCF-7) transplanted into athymic nude mice. J. Natl Cancer Inst. 1981;67:51–56. [PubMed] [Google Scholar]
- 36.Gunter MJ, et al. Insulin, insulin-like growth factor-I, endogenous estradiol, and risk of colorectal cancer in postmenopausal women. Cancer Res. 2008;68:329–337. doi: 10.1158/0008-5472.CAN-07-2946. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Gunter MJ, et al. Insulin, insulin-like growth factor-I, and risk of breast cancer in postmenopausal women. J. Natl. Cancer Inst. 2009;101:48–60. doi: 10.1093/jnci/djn415. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Yu MW, et al. Prospective study of hepatocellular carcinoma and liver cirrhosis in asymptomatic chronic hepatitis B virus carriers. Am. J. Epidemiol. 1997;145:1039–1047. doi: 10.1093/oxfordjournals.aje.a009060. [DOI] [PubMed] [Google Scholar]
- 39.Stuver SO, et al. Insulin-like growth factor 1 in hepatocellular carcinoma and metastatic liver cancer in men. Int. J. Cancer. 2000;87:118–121. doi: 10.1002/1097-0215(20000701)87:1<118::aid-ijc17>3.0.co;2-w. [DOI] [PubMed] [Google Scholar]
- 40.Liaw YF, et al. Asian-Pacific consensus statement on the management of chronic hepatitis B: a 2005 update. Liver Int. 2005;25:472–489. doi: 10.1111/j.1478-3231.2005.01134.x. [DOI] [PubMed] [Google Scholar]
- 41.Kumar M, et al. Virologic and histologic features of chronic hepatitis B virus-infected asymptomatic patients with persistently normal ALT. Gastroenterology. 2008;134:1376–1384. doi: 10.1053/j.gastro.2008.02.075. [DOI] [PubMed] [Google Scholar]