Risk Factors for Hepatocellular Carcinoma in a Japanese Population: A Nested Case-Control Study

Hepatocellular carcinoma is one of the most common cancers worldwide. Chronic infections with hepatitis B virus (HBV) or with hepatitis C virus (HCV) are recognized as critically important risk factors for hepatocellular carcinoma. In addition, a large number of epidemiologic studies have shown that environmental factors such as dietary aflatoxin, smoking, alcohol consumption, and oral contraceptive intake are associated with increased risk for hepatocellular carcinoma (1, 2). It is generally considered that effects of these environmental factors are modified by gender, age, and race of patients (2-4).

Obesity and diabetes mellitus have recently received increased attention as risk factors for hepatocellular carcinoma (5-9). A large number of epidemiologic studies have shown that obesity and diabetes mellitus increase risks of a variety of cancers, including colon, renal, prostate, postmenopausal breast, and ovarian, in Asian and Western countries (7, 10, 11). Several recent epidemiologic studies indicated that obesity might be associated with an increased risk for hepatocellular carcinoma, but few cohort studies have incorporated HBV and HCV infection status in a strict and in-depth manner. A recent study of liver cirrhosis showed that, although obesity [body mass index (BMI), >30 kg/m²] is an independent risk factor for hepatocellular carcinoma among patients with alcoholic cirrhosis or cryptogenic cirrhosis, it is not a significant risk factor for hepatocellular carcinoma in patients with chronic HBV and/or HCV infections (12).

Compared with viral etiologic factors, alcohol consumption, smoking, obesity, and diabetes mellitus may have less effect on hepatocellular carcinoma occurrence (13, 14); however, most epidemiologic studies have indicated that such factors promote development from chronic hepatitis to hepatocellular carcinoma (6, 8). Alcohol consumption, obesity, and diabetes mellitus have been shown to be involved in the progression of liver fibrosis; it is possible that liver fibrosis results from advanced oxidative stress due to hepatic steatosis and iron overload (15-17). Liver cirrhosis characterized by severe liver fibrosis may underlie the occurrence of hepatocellular carcinoma, specifically in the presence of chronic hepatitis C, nonalcoholic steatohepatitis, and alcoholic liver diseases (3, 8). On the other hand, several recent large-scale studies have indicated that coffee drinking suppressed the progression of liver fibrosis and inhibited the development of hepatocellular carcinoma (18, 19).

The fact that liver cirrhosis is not a necessary condition for hepatocellular carcinoma occurrence was already known, not only from clinical findings but also from genetic findings. Among hepatocellular carcinoma cases with HBV, a part of the HBV genome has been shown to be integrated into the host's intracellular DNA, thereby causing hepatocellular carcinoma (20). Among hepatocellular carcinoma cases with HCV, the HCV core protein seems to directly contribute to the mechanism of carcinogenesis by elevating oxidative stress (21). In light of the aforementioned findings, for the purpose of determining independent risk factors for hepatocellular carcinoma, careful analyses are needed controlling for severity of liver fibrosis, as well as for viral etiologic factors.

With the aim of determining whether HBV or HCV infections, alcohol consumption, smoking, coffee drinking, BMI, and diabetes mellitus are independent risk factors for hepatocellular carcinoma, and how the effects of these factors might change after adjusting for severity of liver fibrosis, we conducted a nested case-control study among the Adult Health Study longitudinal cohort using stored sera. We also evaluated whether viral etiology and increase of BMI exert synergistic effects on the risk for hepatocellular carcinoma.

The Atomic Bomb Casualty Commission and its successor, the Radiation Effects Research Foundation, established the Adult Health Study longitudinal cohort in 1958, in which 20,000 age-, gender-, and city-matched proximal and distal atomic bomb survivors and persons not present in the cities at the time of bombings have been examined biennially in outpatient clinics in Hiroshima and Nagasaki.

Serum samples obtained from the study participants on each occasion of visiting outpatient clinics have been collected and stored systematically since 1969 (22). Incident cancer cases were identified through the Hiroshima Tumor and Tissue Registry and Nagasaki Cancer Registry, supplemented by additional cases detected via pathologic review of related diseases (23). There were 359 primary hepatocellular carcinoma cases among Adult Health Study participants diagnosed between 1970 and 2002, who visited our outpatient clinics before their diagnosis. Of these, 130 cases were excluded because of nonavailability of stored serum or having only one stored sample. The other 229 cases had serum samples obtained within 6 years before hepatocellular carcinoma diagnosis. After excluding five cases with inadequate stored serum, 224 cases remained for our study. For each case, three controls were selected from the cohort in nested case-control fashion. Nested control selection was random among those who matched the case on gender, age (±2 years), city, time of serum storage (±2 years), and method of serum storage, and countermatched on radiation exposure (24). Although the total number of potential matched control serum samples is 672, because of occasional lack of subjects with stored sera who met the matching and countermatching criteria, the total number of control serum samples actually used was 644.

HBV surface antigen and antibody to hepatitis B core antigen were measured by enzyme immunoassay, and anti-HCV antibody was measured by second-generation enzyme immunoassay as previously described (22, 25). Qualitative detection of HCV RNA among anti-HCV–positive samples was done using a thermocycler (Whatman Biometra) with two sets of PCR primers corresponding to the 5′-untranslated region, as previously described (25). Qualitative detection of HCV RNA was conducted at least twice. HBV infection (HBV+) status was defined as positive for HBV surface antigen or having a high titer of the antibody to hepatitis B core antigen. HCV infection (HCV+) status was defined as positive for HCV RNA (25). Hyaluronic acid and type IV collagen as liver fibrosis markers were measured using an autoanalyzer (Hitachi 7180, Hitachi, Ltd.) and latex agglutination–turbidimetric immunoassay (Fujirebio, Inc., Daiichi Pure Chemicals Co. Ltd.). Ferritin was measured using an autoanalyzer (Hitachi 7180, Hitachi) and colloidal gold immunoassay (Alfresa Pharma Corporation). Platelet count was measured using an automatic blood cell counter at the time of serum storage.

Self-administered questionnaires on various lifestyle factors were given to participants in 1965 during attendance at the Adult Health Study examination and in 1978 by mail survey. Information from the 1978 survey was obtained before hepatocellular carcinoma diagnosis for all but 19 (15%) of the cases. Information on alcohol consumption was obtained from the 1965 questionnaire when available, with missing data complemented using the 1978 survey. Alcohol consumption per volume of each type of alcoholic beverage was quantified as previously described (26), and mean ethanol amounts were calculated as grams per day. Information on smoking habits was obtained from the 1965 questionnaire; subjects were divided into the following categories: never, prior, and current smoker. Information on coffee drinking was obtained from the 1978 survey; subjects were divided into the following categories of frequency of coffee consumption: never, 1 day per week, 2 to 4 days per week, and almost daily. Disease diagnoses were based on the International Classification of Diseases (ICD) codes: diabetes mellitus was defined by ICD-7 code 260, ICD-8 code 250, ICD-9 code 250, and ICD-10 codes E10 through E14. BMI (kg/m²) was calculated from height and weight measured at the Adult Health Study examination.

Subjects were classified based on BMI quintiles with cut points of 19.5, 21.2, 22.9, and 25.0. The number of hepatocellular carcinoma cases with BMI of >30.0 kg/m² was too small to be analyzed in detail. Following the recommendations for Asian people by the WHO, the International Association for the Study of Obesity, and the International Obesity Task Force (27), 21.3 to 22.9 kg/m² was considered as normal, 23 to 25 kg/m² as overweight, and >25.0 kg/m² as obese in the present study. We used information on diabetes mellitus and BMI obtained 10 years before the time of hepatocellular carcinoma diagnosis or control matching because these conditions are subject to change because of disease progression in the later stages before diagnosis of hepatocellular carcinoma. Atomic bomb radiation dose was estimated for each subject according to the Dosimetry System DS02 (28).

This nested case-control study was based on RERF Research Protocol 1-04 and approved by the Human Investigation Committee of Radiation Effects Research Foundation.

The nested case-control design is analyzed using a partial likelihood method analogous to that used for cohort follow-up studies (29), which is, in practice, the same as the conditional binary data likelihood for matched case-control studies (30) except that the subjects (cases and controls) in the study are not completely independent because of the possibility of repeated selection. All factors other than radiation were analyzed using relative risks estimated by a log-linear model. The population attributable fraction was estimated for individual factors that increased the risk for hepatocellular carcinoma in the present study. Population attributable fraction was calculated as pd × [(mRR − 1) / mRR], where mRR is the multivariate adjusted relative risk for the covariates and pd is the proportion of cases exposed to the risk factor. Statistical interaction between viral infection and BMI was tested by adding the product of the two factors to the log-linear model, which tests departure from a multiplicative relationship. Reported P values and confidence limits are based on Wald statistics. Although radiation exposure could have been adjusted by matching on radiation dose as an additional matching factor in the control selection (31), in addition to assessing effects of lifestyle factors and viral hepatitis, another purpose of the present study was to examine effects of radiation exposure after adjustment for possible confounding and interaction by these factors, so matching on radiation, which prevents analysis of radiation risk, was not desirable; rather, we countermatched on radiation (29, 32). Radiation risk was analyzed by using an excess relative risk model as has been done previously (33).

Characteristics of the 224 hepatocellular carcinoma cases and 644 comparison subjects are shown in Table 1. The mean age of the cases was 67.6 years, and 61% were men. Cases and controls were comparable with respect to gender, age, city, time of serum storage, and method of serum storage by design. Virological and biochemical assays were done on 211 case and 640 control sera because 13 case samples and 4 control samples had insufficient stored sera for these assays. Hepatocellular carcinoma case sera evidenced a higher prevalence of HBV or HCV infection status, higher values of fibrosis markers and ferritin, and lower platelet counts compared with control sera. Greater proportions of hepatocellular carcinoma cases had a history of alcohol consumption of ≥40 g of ethanol per day, were current smokers, were obese, had diabetes mellitus, and received high radiation doses compared with the controls. In addition, hepatocellular carcinoma cases were less likely than controls to be daily coffee drinkers. There were no important differences in characteristics such as gender, age at hepatocellular carcinoma diagnosis, city, or BMI between hepatocellular carcinoma cases excluded because of nonavailability of stored serum and those included in this study.

Table 2 shows the results of univariate and multivariate analyses using HBV and HCV infection status, alcohol consumption, smoking habit, coffee drinking, BMI, diabetes mellitus, and radiation dose. Strong association was found between hepatocellular carcinoma and hepatitis virus infection, resulting in unadjusted relative risks of 33.7 [95% confidence interval (95% CI), 12.7-89.6] for HBV+/HCV− status and 64.5 (95% CI, 29.1-143) for HBV−/HCV+ status. As expected, the risk for hepatocellular carcinoma for alcohol consumption was significant, with an unadjusted relative risk of 1.34 (95% CI, 1.12-1.60) per 20 g of ethanol per day using continuous alcohol consumption and 2.66 (95% CI, 1.55-4.55) at ≥40 g of ethanol per day using grouped alcohol consumption. Although the grouped results suggest that a simple log-linear model in continuous alcohol consumption may not be adequate, a quadratic term did not significantly improve the model (data not shown). Current smoking was significantly associated with hepatocellular carcinoma risk, with an unadjusted relative risk of 1.87 (95% CI, 1.14-3.07). Daily coffee drinking was associated with decreased risk for hepatocellular carcinoma, with an unadjusted relative risk of 0.51 (95% CI, 0.29-0.90). The presence of obesity and diabetes mellitus 10 years before diagnosis were statistically associated with increased risk for hepatocellular carcinoma, resulting in unadjusted relative risks of 1.88 (95% CI, 1.13-3.13) and 1.88 (95% CI, 1.01-3.50), respectively. The relative risk for a 1-unit difference in BMI was 1.04 (95% CI, 0.99-1.09). Radiation exposure was marginally significantly associated with increased risk for hepatocellular carcinoma (P = 0.055).

The risks for viral infection in multivariate analysis did not meaningfully differ from those obtained in the univariate analysis. Alcohol consumption of ≥40 g of ethanol per day and obesity remained significant risk factors for hepatocellular carcinoma even after adjusting for viral infection status and the other factors, whereas the effects of current smoking and diabetes mellitus became nonsignificant after adjustment. Daily coffee drinking was marginally significantly associated with decreased risk for hepatocellular carcinoma after adjustment for viral infection and the other factors. The adjusted relative risk for a one unit difference in BMI, 1.12 (95% CI, 1.03-1.22), was statistically significant, but a quadratic term was not significant.

Table 3 shows the estimated population attributable fraction based on the multivariate adjusted relative risks in the present study. The proportion of hepatocellular carcinoma cases that is attributable to HBV+/HCV−, HBV−/HCV+, HBV+/HCV+, alcohol consumption of ≥40 g of ethanol per day, and obesity were 13.4%, 62.0%, 2.4%, 17.4%, and 20.1%, respectively. These values are not mutually exclusive because some cases were exposed to more than one risk factor.

Table 4 shows results for univariate analyses incorporating biomarkers associated with progression of liver fibrosis, such as hyaluronic acid and type IV collagen of fibrosis markers, platelet count, and ferritin. Large statistically significant differences in the mean values of these variables were observed between hepatocellular carcinoma cases and controls. Figure 1 shows a comparison of multivariate analysis results with or without adjustment for ln(type IV collagen) and platelet count using HBV and HCV infection status, alcohol consumption, smoking habit, coffee drinking, BMI, diabetes mellitus, and radiation dose as adjustment variables. We evaluated type IV collagen and platelet count as surrogate markers associated with severity of liver fibrosis. Hepatocellular carcinoma risk for hepatitis virus infection status after adjusting for liver fibrosis meaningfully decreased compared with the results indicated in the previous multivariate analysis, with relative risks of 20.8 (95% CI, 4.8-90.3) and 37.8 (95% CI, 12.4-115) for HBV+/HCV− status and HBV−/HCV+ status, respectively (Fig. 1A). Effects of ≥40 g of ethanol per day and daily coffee drinking decreased and disappeared, respectively, so that adjustment for liver fibrosis decreased the effect of these factors on risk for hepatocellular carcinoma. Current smoking became marginally significantly associated with increased risk for hepatocellular carcinoma after adjusting for liver fibrosis. Obesity remained a significant risk factor independent of adjustment for severity of liver fibrosis, and the relative risk for diabetes mellitus did not meaningfully differ from that without such adjustment (Fig. 1B).

Table 5 shows the joint effects of hepatitis virus infection status and BMI, with adjustment for alcohol consumption, smoking habit, coffee drinking, diabetes mellitus, and radiation dose. Although being obese was clearly a risk factor for hepatocellular carcinoma subjects with adjustment for viral factors, it was not a significant risk factor in those with HBV−/HCV− status. However, despite the appearance of a trend with BMI, only 15 hepatocellular carcinoma cases were identified among HBV−/HCV− individuals with obesity. Among hepatocellular carcinoma subjects with HBV−/HCV+ status, the relative risk increased dramatically with increasing BMI. Linear (P = 0.003) and quadratic (P = 0.013) terms in continuous BMI were significant among HBV−/HCV+ individuals. Among hepatocellular carcinoma subjects with HBV+/HCV− status, the relative risk for hepatocellular carcinoma did not show evidence of an increase with increased BMI, although the examination of a joint effect of HBV infection and BMI was based on only one hepatocellular carcinoma case out of three subjects who were HBV+/HCV− and obese. The reason for the relatively small unadjusted relative risk for obesity (Table 2) might have been due to the small number of cases and controls with HBV+/HCV− status, which apparently offset the increase observed in HBV−/HCV+ status individuals.

This nested case-control study indicated that HBV and HCV infection, alcohol consumption of ≥40 g of ethanol per day, and obesity 10 years before hepatocellular carcinoma diagnosis were independent risk factors for hepatocellular carcinoma, and that obesity as well as hepatitis virus infection remained independent risk factors for hepatocellular carcinoma after taking into account the severity of liver fibrosis. Furthermore, significant multiplicative interaction in hepatocellular carcinoma risk between viral etiology and increased BMI was observed in HCV-infected individuals. The population attributable fraction of 62.0% for hepatocellular carcinoma cases with HCV infection was highest, and hepatocellular carcinoma cases with HBV infection, alcohol consumption of ≥40 g of ethanol per day, or obesity had population attributable fractions in the range of 13.4% to 20.1%. These are only approximate estimates of the potential for reducing hepatocellular carcinoma occurrence, as we do not know what effect removal of one risk factor would have on the distribution of the other risk factors.

Multivariate analysis after adjusting for severity of liver fibrosis indicated that hepatocellular carcinoma risk for HBV and HCV infections significantly decreased, which is consistent with the existing notion that hepatocellular carcinoma risk increases with progression from chronic hepatitis B and C to liver cirrhosis. A large-scale meta-analysis (34) and a case-control study (35) showed a combined effect of HBV and HCV infections on hepatocellular carcinoma risk, whereas our study did not detect similar effects among those with HBV+/HCV+ status. This difference may be partly attributable to the extremely limited number of coinfected subjects with HBV and HCV among our study population. It may be also partly because most past epidemiologic studies have defined chronic HCV infection by either anti-HCV antibody positivity or by HCV RNA positivity in serum (34, 35).

Several epidemiologic studies and clinical trials revealed an association between obesity and hepatocellular carcinoma risk (9-12), but few population-based cohort studies have been conducted with precise adjustment for HBV and HCV infection status, the major risk factors for hepatocellular carcinoma. Obesity was recently found to be one of the etiologic factors for nonalcoholic steatohepatitis, which is considered a non-B, non-C liver disease, and it has been shown to be a risk factor for hepatocellular carcinoma (12, 16). Although many clinical studies showed that, among chronic hepatitis C patients, obesity was associated with progression of inflammation, insulin resistance, hepatic steatosis, and liver fibrosis (17, 36), a study by Nair et al. (12) reported that obesity was not an independent risk factor for hepatocellular carcinoma among liver cirrhosis patients with HBV and HCV. On the other hand, a recent Western cohort study showed that being overweight (BMI, 25 to <30 kg/m²) or obese (BMI, ≥30 kg/m²) was an independent risk factor for hepatocellular carcinoma (37).

In the present study, we adjusted for potentially confounding factors including hepatitis virus infection and also found that being obese 10 years before hepatocellular carcinoma diagnosis was associated with a 4.57-fold increase in hepatocellular carcinoma risk. Furthermore, we observed a statistically significant, positive, multiplicative interaction between HCV infection and increased BMI on the risk for hepatocellular carcinoma, which indicates decisively that the joint effect of the two factors is greater than additive.

Obesity contributes to a high rate of visceral fat storage, accelerating production of tumor necrosis factor-α, interleukin 6, resistin, and leptin, and decreasing production of adiponectin (16). These cytokines presumably foster insulin resistance (16), cause hepatic steatosis and oxidative stress, and eventually promote hepatocellular carcinoma occurrence. A large number of studies pointed out association between progression of liver fibrosis and insulin resistance or hepatic steatosis (15-17), but authenticity of any connection is now being questioned (8, 36). Interestingly, in this study, obesity remained an independent risk factor for hepatocellular carcinoma even after adjusting for all confounding factors including severity of liver fibrosis. The following are the possible reasons why obesity increases hepatocellular carcinoma risk irrespective of severity of liver fibrosis: Several animal experiments showed that liver tumors were not always accompanied by advanced fibrosis among a variety of genetically engineered mouse models with steatohepatitis (38), and some reports indicated several nonalcoholic steatohepatitis–derived human cancer cases without significant liver fibrosis (39). The findings suggest that significant liver fibrosis is not essential for the carcinogenic process, but that steatohepatitis itself is a state conferring a risk for high carcinogenicity. With regard to the proven relationship between obesity and such malignant tumors as colon, breast, and ovarian cancers (10), the cell proliferation activity of insulin due to hyperinsulinemia is believed to play a role in a common carcinogenic mechanism (5).

It is well documented that obesity induces insulin resistance, with a tendency to cause diabetes mellitus. In the case of hepatic cirrhosis accompanied by highly advanced liver fibrosis, glucose intolerance tends to lead to diabetes mellitus. A recent animal experiment showed that HCV contributed to progression of insulin resistance, resulting in diabetes mellitus (40). The present study failed to show that diabetes mellitus 10 years before hepatocellular carcinoma diagnosis was an independent risk factor for hepatocellular carcinoma, but an adjustment for all factors, except alcohol consumption and BMI, brought about a 30% increase in the effect of diabetes on hepatocellular carcinoma risk (data not shown). Such findings suggest a relationship between diabetes mellitus and alcohol consumption, as well as BMI. Therefore, by taking into account the proven association between alcohol consumption, obesity, and increased risk for hepatocellular carcinoma, our results will not likely refute an association between diabetes mellitus and hepatocellular carcinoma risk.

A large number of epidemiologic studies showed that heavy alcohol consumption was an independent risk factor for hepatocellular carcinoma and that there was correlation between increased risk for hepatocellular carcinoma and amount of alcohol consumed (3, 9, 13, 14). In addition, in some case-control studies of hepatocellular carcinoma risk, synergistic interactions between alcohol consumption and hepatitis virus infection, or between obesity and diabetes mellitus, have been observed (9, 13, 14). In the present study, after adjusting for other factors such as hepatitis virus infection and BMI, alcohol consumption of ≥40 g of ethanol produced a 4.36-fold increase in hepatocellular carcinoma risk. A few recent case-control studies suggested that ethanol consumption of <50 to 60 g/d (41, 42) or alcohol exposure <1,500 gram-years (9) had protective effects on the progression of liver fibrosis and risk for developing hepatocellular carcinoma. Reasons for such discrepancy between our result and former reports are unclear, but factors such as gender, age, race (43), hereditary predisposition, and etiology of liver disease presumably affect the severity of alcohol-related liver diseases. Our study also showed that effects of alcohol consumption of ≥40 g of ethanol per day on hepatocellular carcinoma risk were reduced after adjusting for all confounding factors including severity of liver fibrosis. The finding suggests that alcohol consumption may contribute to hepatic carcinogenesis by enhancing oxidative stress and aggravating liver fibrosis.

As a result of recent assessments by the IARC, hepatocellular carcinoma has been positioned as a smoking-related malignant disease (44). However, it has yet to be determined whether smoking itself has direct hepatic carcinogenic effects or whether smoking contributes to hepatic carcinogenesis by way of progression of liver fibrosis. A case-control study showed that 4-aminobiphenyl DNA adducts contained in tobacco smoke are a liver carcinogen (45). In the present study, we adjusted for potential confounding factors including hepatitis virus infection and failed to detect significant smoking effects on hepatocellular carcinoma risk; however, a multivariate analysis that excluded hepatitis virus infection showed significant effects of smoking (data not shown). With adjustment for all factors including severity of liver fibrosis, effects of smoking on hepatocellular carcinoma risk were found to be marginally significant. These findings suggest the possibility that smoking, in conjunction with hepatitis virus infection, further enhances the risk for hepatocellular carcinoma and might directly contribute to the mechanism of liver carcinogenesis.

Several epidemiologic studies indicated the involvement of coffee in decreased alanine aminotransferase activity and γ-glutamyltransferase level, suppression of progression to liver cirrhosis, and inhibited development of hepatocellular carcinoma (18, 19). Such oxidation inhibitors as caffeine, coffee diterpenes, and chlorogenic acid are among candidate substances in coffee that potentially reduce the risk for hepatocellular carcinoma, and several animal experiments have shown that such substances have direct inhibitory effects on hepatic carcinogenesis (46). Adjusting for all potential confounding factors including hepatitis virus infection rendered the effects of coffee drinking on hepatocellular carcinoma risk marginally significant, whereas adjusting for all factors, except hepatitis virus infection, revealed significant effects of coffee drinking (data not shown). Furthermore, adjusting for all factors including severity of liver fibrosis erased the effects of coffee drinking on hepatocellular carcinoma risk. These findings suggest that coffee drinking may somehow suppress liver fibrosis and thereby indirectly reduce hepatocellular carcinoma risk.

The main strengths of our study are its prospective cohort-based, nested case-control design, which minimized selection bias and provided for the use of stored sera and a wealth of epidemiologic information obtained before hepatocellular carcinoma diagnosis. Indeed, the distributions of HBV and HCV infection status among hepatocellular carcinoma cases and controls and mean age at diagnosis among hepatocellular carcinoma cases were similar to those in previous reports on Japanese populations (2, 4). Another major strength of our study is that it incorporated, in a strict and in-depth manner, HBV and HCV infection status and showed the interrelationship between these and numerous other epidemiologic factors. It is difficult and expensive to perform full cohort serum analyses, whereas the nested case-control design used here can provide substantial reductions in cost and effort with little loss of statistical efficiency.

The main limitation of our study is that the severity of liver fibrosis could not be classified into fibrosis stage of F0 to F4 based on liver specimens. We used platelet counts and type IV collagen concentrations as surrogate, but independent, markers of liver fibrosis. Previous reports showed a strong correlation between platelet count and fibrosis stage in the presence of chronic hepatitis C (47) and a close association between levels of type IV collagen, a basic component of the hepatic basal membrane, and severity of liver fibrosis. Another limitation of our study is the usage of sera that had been stored for long periods of time. Proteins and HCV RNA trend to degrade during prolonged storage of either frozen or freeze-dried sera. However, we minimized this degradative effect by the selection of matched controls relative to time and method of serum storage. Furthermore, we have previously shown that the freeze-dried sera are interchangeable with frozen sera in serologic and molecular biological detection of HBV and HCV (22, 25). Finally, some hepatocellular carcinoma cases had to be excluded because of nonavailability of stored sera. We did not detect any differences between included and excluded cases in terms of demographic variables or BMI.

In conclusion, HBV and HCV infection and obesity were independent risk factors for hepatocellular carcinoma, even after taking into account the severity of liver fibrosis. Moreover, the combination of HCV infection and increased BMI exerted a synergistic effect on the risk for hepatocellular carcinoma. Alcohol consumption of ≥40 g of ethanol per day was also an independent risk factor for hepatocellular carcinoma, likely contributing to the development of hepatocellular carcinoma through liver fibrosis. The radiation effect on hepatocellular carcinoma risk was shown to be marginally significant in univariate analysis; whether the radiation effect is confounded with other factors will be closely examined in a separate report. A precise understanding of the mechanism by which obesity contributes to development of hepatocellular carcinoma should lead to better therapeutic strategies, public health policies, and cost-effectiveness.

We thank Naomi Masunari and Sachiko Teranishi for the collection and processing of the data; all members of division of clinical laboratories for their excellent assistance in experiments; and Michiko Yamada, Yoshimi Tatsukawa, and Kyoji Furukawa for helpful discussion. The Radiation Effects Research Foundation, Hiroshima and Nagasaki, Japan, is a private, nonprofit foundation funded by the Japanese Ministry of Health, Labour and Welfare and the U.S. Department of Energy, the latter through the National Academy of Sciences.