Plasma Adipokines and Risk of Hepatocellular Carcinoma in Chronic Hepatitis B Virus–Infected Carriers: A Prospective Study in Taiwan

Adipose tissue was traditionally considered a tissue only for triglyceride storage, and gradually was accepted as a genuine endocrine organ (1). In addition to energy homeostasis effects, adipocytokine are shown to have profound influences on many aspects of human physiology. Accumulating evidence shows that lower adiponectin level is associated with an increased risk of colorectal cancer (2), prostate cancer incidence and progression (3, 4), breast (5–7), endometrial (8), and gastric cancers (9). In humans, blood leptin level rises with the increase of adiposity (10) and has been implicated in the modulation of immunity (11), reproduction (12), and even the ageing process (13). Leptin has been shown to have a positive association with breast, prostate, colorectal, pancreatic, and endometrial cancers (14, 15). Visfatin is known as pre-B-cell colony-enhancing factor (16), an intracellular transcription factor shown in vitro to be related to apoptosis (17), and it may have carcinogenetic effect as studies have linked visfatin to colorectal (18, 19) and endometrial (20) cancers.

Several clinical studies showed a significant inverse correlation between circulating adiponectin levels and liver functions tests, including serum aspartate aminotransferase (AST), alanine aminotransferase (ALT), and γ-GT (21, 22). However, cross-sectional studies showed the opposite results of significantly elevated circulating adiponectin levels in cirrhosis patients of different etiologies when compared with healthy controls (23, 24). Furthermore, adiponectin was found to be significantly higher in patients with more advanced chronic liver disease, leading to the speculation that adiponectin might be an indicator of liver disease severity (24). Nevertheless, the cross-sectional features of these studies hinder the temporal sequences of the relationship. Because liver may be a major site of adiponectin extraction, it is not suitable to evaluate the role of adiponectin on the risk of liver disease in cross-sectional or case–control study designs. To date, there are three prospective studies reporting higher adiponectin levels associated with a higher hepatocellular carcinoma (HCC) risk either among patients with chronic hepatitis C (25), among patients with hepatitis B or C (26), or among all participants (27). We therefore conducted a nested case–control study among chronic hepatitis B carriers from a community-based cohort in Taiwan to investigate prospectively the association between plasma adipocytokine levels, including adiponectin, leptin, and visfatin, and the risk of HCC development.

Detailed descriptions of the recruitment procedures of the cohort from whom the case and control subjects arise have been previously published (28, 29). In brief, a total of 89,293 residents ages 30 to 65 years living in seven townships in Taiwan were invited to participate in a follow-up study. During 1991 to 1992, 23,820 (11,973 males and 11,847 females) agreed to participate in questionnaire interview, health examination, blood collection and follow-up of health status through health examination, medical record review, and data linkage with national health profiles. All participants gave informed consent to participate in this study and the data collection procedures were reviewed and approved by the Institutional Review Board of the National Taiwan University College of Public Health and College of Medicine (Taipei, Taiwan).

Newly developed primary HCCs were detected among surface antigen of hepatitis B virus (HBsAg; +) antibody against hepatitis C virus (anti-HCV; −) participants by follow-up health examination or by computerized data linkage with the profiles of National Cancer Registry in Taiwan from January 1, 1991 through December 31, 2004. To ensure complete ascertainment, we also carried out data linkage with the National Death Certification profiles to identify cases unregistered in cancer registry system and the death certificate only was found to be 1.5% for HCC in this analysis. The nation-wide cancer registry system was implemented in 1978 in Taiwan with updated, accurate, and complete information. The ascertainment of newly developed HCC was complete with supplementary data linkage and accurate with their medical records reviewed by gastroenterologists (30). All study participants were without HCC or other types of cancer when first joined the study.

Eligible controls were those who were HBsAg (+) anti-HCV(−) at enrollment and alive and free of cancer during the month of the case patients' diagnoses. Two controls frequency matched to a case by age at diagnosis (5 year intervals), sex, and residence areas (Taiwan main island vs. Panhu islet) at enrollment were randomly selected from the cohort members. There were 374 controls identified for current analysis.

Personal interviews using a structured questionnaire were administered by well-trained public health nurses at recruitment. Information collected included sociodemographic characteristics, dietary habits, habits of cigarette smoking, alcohol drinking, medical and surgical history, and family history of HCC and liver cirrhosis. Using standard sterile techniques, a 10-mL blood sample was collected at cohort entry and follow-up examinations, fractionated on the day of collection, and stored in a deep freezer (−70°C) until examination. Serologic tests of HBsAg, anti-HCV, and serum levels of AST, ALT on all cohort participants, HBV e antigen (HBeAg), HBV viral loads, and genotype on those tested positive in HBsAg were performed using commercial kits and detailed laboratory procedures were described previously (28, 29). The reference dates for the cases were their HCC diagnosis dates and June 30 of the same diagnosis year was assigned to their frequency-matched controls. Plasma adiponectin, leptin, and visfatin levels were measured using the samples collected and stored at baseline by commercially available ELISA kits (B-Birdge International, Inc; for adiponectin and leptin; Phoenix Pharmaceuticals, Inc.; for visfatin, respectively). Laboratory testing followed standard quality control procedure to ensure the reliability of the results. All cases and controls were tested in the same batch with 90 samples per testing and the technician was blinded to the case–control status. The sensitivity is 0.02 ng/mL for adiponectin, 0.5 ng/mL for letpin, and 1.85 ng/mL for visfatin assay, respectively. The intra-assay and inter-assay coefficients of variation (CV) were <6% and <8% for adiponectin and <6% and <9% for leptin, respectively. The corresponding numbers for visfatin were <10% and <12%.

For those who tested positive in HBsAg at baseline, periodic (6–12 months) health examinations, including blood drawing and ultrasound, were provided. For the analysis of adiponectin and HBV-related factors, the cases and controls were mixed and 535 with adiponectin data were included. Among them, 484 (90%) had follow-up ultrasound exams to determine incident liver cirrhosis and fatty liver and 467 (87%) with at least one follow-up information to determine their HBsAg seroclearance status. To determine the HBV DNA change and ALT change, more than two follow-up examinations were required, and only those with higher HBV DNA at baseline (≥2 × 10³ IU/mL) were checked for their follow-up HBV DNA. As a result, there were 190 (68% among those with high viral load) with information on the HBV DNA changing pattern and 391 (73%) with ALT changing pattern. The HBsAg seroclearance was defined as attaining seronegative in any of the follow-up tests. The long-term changes in serum HBV DNA and ALT levels were also evaluated using the same definitions as described previously (31). Newly diagnosed liver cirrhosis (32) via ultrasound during follow-up were also identified to clarify whether the relationship between adipocytokine and HCC were mediated through cirrhosis. To ensure the complete ascertainment of cirrhosis cases, the data linkage to the National Health Insurance profiles in Taiwan was conducted. In addition, the medical records of these cases were reviewed by gastroenterologists (30).

All risk factors collected were analyzed by comparing the mean and SDs for continuous variables and frequency distribution for categorical variables between cases and controls. Student t test or ANOVA were used to test the difference of adiponectin levels in different HBV infection and long-term patterns of HBV DNA and ALT. Adipocytokine levels were further categorized into quintiles based on the distribution of the controls. Conditional logistic regression models stratifying on matching variables and unconditional logistic regression were conducted. Because the major results were similar, we decided to use unconditional logistic regression to estimate the ORs and 95% confidence intervals (CI) for the association between three adipocytokine levels and the risk of HCC. Linear trends across quintiles of the three adipocytokine levels were assessed by testing the statistical significance of a single trend variable.

In addition to the matching factors, established HCC risk factors and metabolic factors that were available were considered as potential confounders, this included age at reference, gender, cigarette smoking, habitual alcohol consumption and betel nuts chewing at enrollment, liver function measurement, and history of liver cirrhosis and nonalcoholic fatty liver disease (NAFLD). The metabolic factors included body mass index (BMI; <23 kg/m², 23–<25 kg/m², 25–<30 kg/m², and ≥30 kg/m²), central obesity (waist circumferences >90 cm for men and >80 cm for women), waist-to-hip ratio (>0.9 for men and >0.7 for women was defined as obese), history of diabetes and hypertension, hypercholesterolemia (total cholesterol ≥ 240 mg/dL) as well as hypertriglyceridemia (triglyceride ≥150 mg/dL). A total of five multiple logistic regression models were conducted to adjust for the effect of these factors. Model 1 adjusted for matching factors, demographic (education level), and lifestyle factors (habitual cigarette smoking, alcohol consumption, and betel nuts chewing). In model 2, history of diabetes or hypertension, hypertriglyceridemia and hypercholesterolemia as well as BMI was further added as adjustment variables. In addition to model 2, model 3 and 4 added the most important HCC risk factors (liver function in model 3, liver function plus HBV DNA in model 4) to determine whether adiponectin was an independent risk factor. Model 5 included all covariates in model 4 and the long-term ALT change patterns (four categories).

The relationship between adiponectin levels and HCC in certain subgroups of the population was modeled by separate analyses within each subgroup of interest. These potential effect modifiers included metabolic factors (BMI; central obesity; ultrasonographic fatty liver) and HBV-related factors (genotype; HBV DNA; ALT). The presence of effect modification was tested by use of an interaction term between the adiponectin trend variable and the group variables. All analyses were two sided and performed by Stata (Version 12.1, Stata Corp.).

Among 187 cases and 374 controls, three reported with history of cirrhosis (two cases and one control) and were excluded from the analysis to rule out the possibility of prevalent cirrhosis changing adipocytokines levels due to wasting. The mean duration between enrollment and HCC diagnosis were 6.66 (SD = 3.56) years. The distribution and matching factors (age, gender, and residential areas) adjusted ORs (95% CI) of selected demographic characteristics and HCC risk factors were shown in Table 1. As expected, because of the frequency matching, mean age at enrollment, gender, and residential area distributions for HCC cases and controls were similar. Compared with controls, cases were more likely to have habitual alcohol consumption (OR = 2.00; 95% CI, 1.28–3.13) and betel nuts chewing (OR = 2.08; 95% CI, 1.11–3.90). For metabolic factors, cases had higher BMI and waist circumferences, and a higher frequency of diabetes history. Those with elevated AST (OR = 5.31; 95% CI, 2.89–9.76), ALT (OR = 3.38; 95% CI, 1.93–5.94) levels, higher titer of HBV DNA (P_trend < 0.001), and genotype C (OR = 2.14; 95% CI, 1.33–3.43) were at an increased risk of HCC. The matching factor–adjusted ORs of HCC in relation to adiponectin, leptin, and visfatin levels were modeled as continuous and quintile-categorical variables. Compared with controls, cases had higher mean levels of adiponectin (13.05/11.76) and leptin (8.69/7.89), and lower visfatin levels (11.10/12.30) at baseline (Table 2). The trends were positive and significant for increasing HCC risk per one unit increase of adiponectin (OR = 1.04; 95% CI, 1.01–1.07) and showed no association for leptin (OR = 1.02; 95% CI, 0.99–1.04) or visfatin (OR = 0.995; 95% CI, 0.98–1.01). When using the lowest quintile as the reference, HCC risk was significant at the third and the highest quintile of adiponectin level (OR = 1.08, 95% CI, 0.55–2.14; OR = 2.07, 95% CI, 1.12–3.82; OR = 1.84, 95% CI, 0.98–3.44; OR = 2.29, 95% CI, 1.22–4.28, respectively; P_trend = 0.003). No apparent association was found for leptin and visfatin.

Subsequent analyses were restricted to adiponectin because it was the only adipocytokines associated with HCC risk (Table 3). Adiponectin levels were associated with an increased risk of HCC after the adjustment of demographic and lifestyle risk factors. The risk of HCC was doubled from the third quintile of adiponectin levels (OR = 1.08, 95% CI, 0.53–2.19; OR = 2.28, 95% CI, 1.20–4.34; OR = 2.05, 95% CI, 1.07–3.93; OR = 2.47, 95% CI, 1.28–4.75, with each additional quintiles, respectively) with a significant dose–response trend (P = 0.002). Model 2 showed that a higher adiponectin level was independently associated with a higher risk of HCC (OR = 0.99, 95% CI, 0.47–2.06; OR = 2.34, 95% CI, 1.21–4.55; OR = 2.08, 95% CI, 1.05–4.12; OR = 2.66, 95% CI, 1.31–5.36, with each additional quintiles, respectively; P_trend = 0.001), after additional adjustment for metabolic factors and BMI. Adiponectin remained as an independent HCC risk factor with an approximately 2-fold increased risk and a significant dose–response trend (P = 0.009 for model 3 and P = 0.02 for model 4) after further adjustment of liver function (model 3) and HBV DNA (model 4). Additional adjustment of long-term ALT changing patterns revealed even higher ORs from the third quintile and up to 4-fold (model 5: OR = 0.51, 95% CI, 0.12–2.11; OR = 4.88, 95% CI, 1.46–16.3; OR = 3.79, 95% CI, 1.10–13.0; OR = 4.13, 95% CI, 1.13–15.1, with each additional quintiles, respectively; P_trend = 0.003).

The association between adiponectin and HCC was further stratified by obesity-related and liver-related factors (Table 4). Because HCC risk started to elevate at the third quintile, these subgroup analyses were based on adiponectin levels divided into two categories using the third quintile as cutoff. The OR estimates were similar by obesity status, either defined by BMI or waist circumferences. HCC risk associated with higher adiponectin level was higher in HBV carriers with ultrasonographic fatty liver (OR = 3.85; 95% CI, 1.24–11.9), with genotype C infection (OR = 3.09; 95% CI, 1.06–9.03), with higher viral load (OR = 3.29; 95% CI, = 1.78–7.32), and with ALT more than upper limit of normal (OR = 26.1; 95% CI, 1.93–351.8). However, none of the interaction terms reached statistical significance. In addition, HBV carriers with genotype C infection or elevated ALT combining with higher adiponectin levels were associated with a 5-fold increased risk (OR = 5.31; 95% CI, 2.09–13.5 for genotype C; OR = 5.80; 95% CI, 1.79–18.7 for elevated ALT, respectively). Those with higher adiponectin and high HBV DNA (≥2 × 10⁴ IU/mL) were at more than 10-fold increased HCC risk (HCC = 15.1; 95% CI, 5.10–40.7) comparing with those with lower levels of HBV DNA and adiponectin.

We further examined the relationship between adiponectin levels and several HBV infection characteristics at baseline and the changes during follow-up (Table 5). Older age, female, and people with lower education levels had higher adiponectin levels, whereas participants who were obese, with dislipidemia or fatty liver, or who had habitual cigarette smoking, alcohol consumption, and betel nut chewing had lower adiponectin. Adiponectin levels were significantly higher among HBV carriers with elevated AST, ALT, and higher HBV DNA at baseline, as well as those tested positive for HBeAg and with genotype C. Those who achieved HBsAg seroclearance during follow-up had significantly lower adiponectin at baseline. HBV carriers whose HBV DNA decreased to <2,000 IU/mL during follow-up had the lowest adiponectin levels, whereas those with persistently high HBV DNA (>2 × 10⁴ IU/mL) had higher adiponectin levels. The ALT changing patterns during follow-up was not associated with significantly different adiponectin. Furthermore, incident cirrhosis diagnosed during follow-up was treated as an endpoint to further investigate the independent effect of adiponectin (Table 6). HBV carriers who were newly diagnosed with liver cirrhosis also had significantly higher adiponectin levels at baseline. After the adjustment of all other risk factors, including HBV DNA, a higher adiponectin level was associated with higher cirrhosis risk (OR = 1.65, 95% CI, 0.62–4.39; OR = 3.85, 95% CI, 1.47–10.1; OR = 2.56, 95% CI, 0.96–6.84; OR = 3.76, 95% CI, 1.33–10.7 for the second, third, fourth, and fifth quintiles, respectively; P_trend = 0.017).

To the best of our knowledge, this is the first case–control study nested within a prospective cohort study investigating adipocytokines levels and HCC risk among HBV carriers. Our results showed that elevated adiponectin levels at baseline were associated with an increased risk of HCC over time, while no overall association was found for leptin and visfatin. Even after the adjustment of liver function, HBV DNA, ALT changes during follow-up, the increased HCC risk associated with higher adiponectin levels persisted with a significant dose–response trend, indicating an independent effect. We also observed higher adiponectin levels among chronic HBV carriers with HBeAg positivity, HBV genotype C infection, and higher HBV viral load. Longitudinally, those with higher adiponectin levels at baseline were less likely to achieve HBsAg seroclearance and more likely to have persistently higher HBV DNA levels. Eventually, these people were more likely to develop liver cirrhosis before HCC occurrence, even after the adjustment of all other important HCC risk factors, including HBV DNA. Therefore, we demonstrated that elevated adiponectin levels could predict the severity of HBV-related liver diseases, independent of other metabolic-related parameters.

To clarify the possibility that those who had high adiponectin levels at baseline was the consequence of end-stage liver diseases that caused wasting thus lowering their BMI, we performed several sensitivity analyses. First, we limited the multiple regression models to those whose HCC were diagnosed more than one year after the enrollment and found similar risk estimates and dose–response trend (P_trend = 0.008). Similar results were also found when restricting to 2, 3, or even 4 years after enrollment. Among 484 with ultrasound exams, there were 55 participants who were diagnosed with liver cirrhosis at their first follow-up exam and 50 (90.9%) of them eventually developed HCC. Because they also had higher adiponectin levels (13.39 vs. 11.83), excluding them from the analysis resulted in lower risk estimates for the higher adiponectin categories and less statistical power. However, the possibility of wasting in them can be ruled out because their BMIs were very similar to those without cirrhosis at first exam (24.39 vs. 24.08, respectively). Furthermore, our analysis on newly diagnosed liver cirrhosis during follow-up revealed similar results, indicating that high adiponectin levels predispose to the development of these end-stage liver diseases and can predict their occurrence.

Obesity and diabetes are two metabolic components found to be associated with increased HCC risk. Adiponectin was reported to be inversely associated with BMI and other metabolic components (33), including type 2 diabetes, and weight reduction can increase its level (34). On the basis of these observations and since adiponectin was long considered as possessing anti-inflammatory and insulin-sensitizing properties, the association between adiponectin and HCC would be expected to be an inverse one. However, the obesity HCC association is mostly found in western countries, mainly of nonviral causes. In addition, our previous study showed only slight and statistically nonsignificant HCC risk (RR = 1.36; 95% CI, 0.64–2.89) in relation to extreme obesity (BMI ≥30 kg/m²) among HBV carriers. On the basis of several cross-sectional studies, the distribution of adiponectin in relation to liver diseases probably needs to be separated by their etiologies (metabolic or viral). Adiponectin was consistently shown to be lower in metabolic liver diseases, including NAFLD and NASH (35), but elevated in advanced liver disease like cirrhosis (24, 36, 37) and increases along with the severity of disease (23). However, blood adiponectin concentration was similar between patients with NASH-related cirrhosis and cirrhosis from other causes (24, 37), indicating that the stage of the liver disease should also be taken into account when examining adiponectin in liver diseases, as liver could be one of the sites for adiponectin degradation. It was also found that adiponectin was positively correlated with surrogate markers of hepatic fibrosis, including transient elastography, fasting serum bile acids, and hyaluronate (37). A recent cross-sectional study showed that HBV-related HCC patients had higher adiponectin levels in comparison with healthy controls, patients with chronic hepatitis B, and cirrhosis patients (36). Moreover, prospective studies all showed that higher adiponectin levels were associated with a higher risk of developing HCC, both in patients with hepatitis B (this study) and hepatitis C (25), as well as in participants with hepatitis B or C (26), or mostly without hepatitis (27).

Studies from animals and humans indicate that circulating adiponectin had some hepatoprotective effects on various stages of liver injuries, and most of them were found in NAFLD (38, 39). It has been demonstrated that the increase of circulating adiponectin was associated with the activation of AMP-activated protein kinase, PPAR-α and PGC1-α, which in turn led to markedly increased rates of fatty acid oxidation, thus preventing hepatic steatosis and alleviating liver enzyme elevation (40–43). It is well known that obesity and insulin resistance are two common causes shared by low plasma adiponectin and hepatic steatosis, the initiating hit of NAFLDs (44). It is generally believed that metabolic factors are the major driving forces for disease progression in NAFLD, whereas low plasma adiponectin secondary to obesity and insulin resistance only serves as an accomplice. On the other hand, serum adiponectin was found to be elevated in patients with chronic hepatitis C in comparison with healthy subjects (25), and our analysis as well as another study (45) showed that adiponectin was positively associated with hepatitis B viral load, indicating that adiponectin may play different roles in viral liver diseases. In viral hepatitis, the main driving force for disease progression includes virus-induced cell death and inflammation (46). As liver disease progresses, plasma adiponectin may rise as the result of reduced degradation by the liver. It was reported that even in healthy HBV carriers, the plasma adiponectin levels were higher than noncarrier controls after adjustment for age, gender, and BMI (47). Therefore, the rise in adiponectin may serve as a biomarker for compromised liver function and liver disease progression. However, we also showed that higher adiponectin levels could predict HCC risk even after adjustment for all the metabolic variables as well as the liver function-related variables, and most importantly the HBV DNA. Furthermore, it was also documented that adiponectin may have proinflammatory and proproliferative effects in various cell types via the activation of the NF-κB, AP-1, and MAPK pathways (48–50). Therefore, the independent procancerous effect of adiponectin cannot be excluded. Because high HBV DNA viral load and elevated ALT level indicate liver disease progression, it is possible that higher adiponectin levels are the result of less capacity of degradation from the liver. One study reported positive correlation of adiponectin with HBV viral load among 266 HBV carriers (45). Another study using HepG2-HBV stable cells found that by lowering adiponectin level inhibited the HBV replication, suggesting a direct stimulatory effect of adiponectin on viral replication (51).

The nested case–control study design minimized the possibility that the association was the results of cirrhosis or HCC altering adiponectin levels because all blood samples were collected before the onset of disease. The other advantage is that in addition to obesity-related measurements, we are able to take into account not only HBV viral load and genotype, but also the long-term changing patterns of ALT levels during follow-up. However, we were only able to measure adipocytokines levels at baseline and cannot rule out the possibility of fluctuation of these measurements during follow-up periods. However, there was no reason to believe that this possible misclassification was differential and therefore, the nondifferential misclassification would lead to underestimates of the true association. Even with the possibility of underestimation, the association between adiponectin and HCC remained strong. Another limitation is that not all participants had follow-up information on ultrasound exams, repeated measurements of ALT and HBV DNA, and the results from analyses using these parameters should be interpreted with caution. In addition, the small sample size rendered wide CIs of some results and may limit their interpretation. Further analysis using all REVEAL-HBV cohort participants to investigate the relationship between adiponectin levels and longitudinally changing HBV infection status is warranted.

In summary, elevated plasma adiponectin level is independently associated with an increased HCC risk in HBV carriers. It is possible that adiponectin may play different roles in the pathogenesis of virus-induced and metabolic-related liver diseases. The exact mechanisms remain unknown, and the fact that liver is the major organ for energy homeostasis and protein degradation makes the interpretation of the relationship between HBV virus, adiponectin, and HCC more complicated. Whether elevated adiponectin level represents liver disease progression and worsening liver function status or it has independent tumorigenic effects, or both, remains unclear and needs to be further elucidated.

No potential conflicts of interest were disclosed.

Conception and design: C.-L. Chen, W.-S. Yang, L.-Y. Wang, C.-J. Chen

Development of methodology: C.-L. Chen, W.-S. Yang, C.-F. Chen, L.-Y. Wang, C.-J. Chen

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.-L. Chen, W.-S. Yang, H.-I. Yang, S.-L. You, L.-Y. Wang, S.-N. Lu, C.-J. Liu, P.-J. Chen, C.-J. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C.-L. Chen, C.-F. Chen, P.-J. Chen, D.-S. Chen, C.-J. Chen

Writing, review, and/or revision of the manuscript: C.-L. Chen, W.-S. Yang, H.-I. Yang, L.-Y. Wang, J.-H. Kao

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C.-L. Chen, W.-S. Yang, H.-I. Yang, C.-J. Chen

Study supervision: C.-L. Chen, D.-S. Chen, C.-J. Chen

This work was supported by grants NSC-94-2314-B002-268 from the National Science Council (to C.-L. Chen, W.-S. Yang, and C.-J. Chen), Department of Health, Executive Yuan (to C.-J. Chen), and Academia Sainica (to C.-J. Chen), Taipei, Taiwan.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.