Fax: (323) 865-0136
Synergism of alcohol, diabetes, and viral hepatitis on the risk of hepatocellular carcinoma in blacks and whites in the U.S.
Article first published online: 7 JUL 2004
Copyright © 2004 American Cancer Society
Volume 101, Issue 5, pages 1009–1017, 1 September 2004
How to Cite
Yuan, J.-M., Govindarajan, S., Arakawa, K. and Yu, M. C. (2004), Synergism of alcohol, diabetes, and viral hepatitis on the risk of hepatocellular carcinoma in blacks and whites in the U.S. Cancer, 101: 1009–1017. doi: 10.1002/cncr.20427
- Issue published online: 18 AUG 2004
- Article first published online: 7 JUL 2004
- Manuscript Accepted: 10 MAY 2004
- Manuscript Revised: 4 MAY 2004
- Manuscript Received: 11 FEB 2004
- National Institute of Cancer. Grant Numbers: R35 CA53890, R01 CA80205
- Research Center for Alcoholic Liver and Pancreatic Diseases. Grant Number: P50 AA11999
- National Institute on Alcohol Abuse and Alcoholism
- hepatocellular carcinoma (HCC);
- viral hepatitis;
Heavy alcohol consumption, viral hepatitis, and diabetes are risk factors for hepatocellular carcinoma (HCC). However, to the authors' knowledge, the information concerning their interaction effect in patients with risk of HCC is sparse.
A population-based, case–control study of HCC was conducted during 1984–2002. The study involved 295 HCC cases and 435 age-, gender-, and race-matched control subjects among Hispanic and non-Hispanic whites and blacks in Los Angeles County, California. Lifestyle risk factors were ascertained through in-person interviews. Infections with the hepatitis B and C (HCV) viruses were determined using their serologic markers.
Fourteen HCC cases but no control subjects tested positive for the hepatitis B surface antigen. Seropositivity for antibodies to HCV was associated with an odds ratio (OR) of 125 (95% confidence interval [95% CI], 17–909) for HCC, whereas seropositivity for antibodies to the hepatitis B core antigen was related to an OR of 2.9 (95% CI, 1.7–5.0). Heavy alcohol consumption and cigarette smoking were found to be independently associated with a statistically significant two to threefold increase in risk of HCC after adjustment for hepatitis B and C serology. Subjects with a history of diabetes had an OR of 2.7 (95% CI, 1.6–4.3) for HCC compared with nondiabetic subjects. A synergistic interaction on HCC risk was observed between heavy alcohol consumption and diabetes (OR = 4.2; 95% CI, 2.6–5.8), heavy alcohol consumption and viral hepatitis (OR = 5.5; 95% CI, 3.9–7.0), or between diabetes and viral hepatitis (OR = 4.8; 95% CI, 2.7–6.9).
Heavy alcohol consumption, diabetes, and viral hepatitis were found to exert independent and synergistic effects on risk of HCC in U.S. blacks and whites. Cancer 2004. © 2004 American Cancer Society.
Hepatocellular carcinoma (HCC) is a rare malignancy among non-Asians in the U.S. In Los Angeles County, California, the average annual age-standardized (world) incidence rates for primary liver carcinoma from 1992 to 1997 in Hispanic white, black, and non-Hispanic white men were 10.6, 6.5, and 3.9 per 100,000, respectively. The corresponding rates in women were 3.8, 2.0, and 1.6 per 100,000.1 Chronic infections with the hepatitis B virus (HBV) and hepatitis C virus (HCV) are two well established causal factors in the development of HCC in humans.2–5 Among populations at high risk for HCC, such as Chinese in southern China, virtually all patients with HCC were chronic HBV carriers.2, 4 In contrast, approximately one-half of HCC cases among black and white residents in Los Angeles County are HCV and/or HBV related.5 Heavy alcohol use, cigarette smoking, and diabetes have been suggested as risk factors for HCC.6–10 The incidence of HCC in the U.S. has reportedly doubled during the past two decades,11 but to our knowledge, the reasons for such an increase are unknown. This large, population-based, case–control study of HCC examined the independent and synergistic effects of viral and nonviral risk factors on the risk of HCC in U.S. blacks and Hispanic and non-Hispanic whites in Los Angeles County.
MATERIALS AND METHODS
The study design was described previously.12 In brief, we studied histologically confirmed cases of newly diagnosed HCC in black, Hispanic, and non-Hispanic white residents in Los Angeles County who were 18–74 years of age at diagnosis. Cases were identified through the Los Angeles County Cancer Surveillance Program, a population-based cancer registry that records all incident cancers diagnosed in Los Angeles County. Cases identified between January 1984 and December 2001 were eligible for inclusion in the study, which comprised a personal interview, and if consent was obtained, collection of a venous blood specimen. Because of the rapidly fatal nature of HCC (the median time interval between diagnosis and death is approximately 3 months),13 84% of eligible cases died before we were able to contact them. Of the 473 cases we contacted, 34 (7%) were too ill to be interviewed, and 320 (73%) of the remaining 439 were interviewed. Histologic confirmation was sought from an expert hepatopathologist, who excluded 25 interviewed cases judged to have either metastatic liver carcinoma or liver adenoma. A total of 295 HCC cases were included in the current study. We compared age, gender, and race/ethnicity between HCC cases included and not included in the current study. The mean age (i.e., 60 years) was identical for the 2 groups, but there were significantly more women (35% vs. 26%) and fewer non-Hispanic whites (58% vs. 64%) in the included than excluded cases. Blood specimens were collected from 245 (83%) of the 295 interviewed cases. There were no statistically significant differences between HCC cases with and without blood specimens in terms of age (60.4 years vs. 61.8 years), gender (65% vs. 64% men), education (48% vs. 48% with at least some college education), cigarette smoking (68% vs. 64% of ever smokers), and alcohol consumption (70% vs. 66% regular drinkers). Non-Hispanic white cases (87%) were more likely to donate blood than Hispanic white (82%) and black cases (70% P = 0.03). The prevalence of diabetes among HCC cases with blood specimens was lower than those without blood specimens (19% vs. 32%; P = 0.04).
For each interviewed case, we sought to recruit up to two control subjects from the neighborhoods in which patients with HCC resided at the time of diagnosis. Using the house of each case as a reference point and proceeding in a systematic and invariable sequence, we canvassed ≤ 150 residential units to identify 2 control subjects who could be matched to the index case by gender, age (within 5 years), and race (Hispanic white, non-Hispanic white, black). We relaxed the race-matching criteria if we failed to recruit 2 control subjects within 150 housing units we canvassed. A total of 435 neighborhood control subjects were recruited to the study. Most (88%) were the first (77%) or second (11%) eligible neighbors. Three hundred fifty-nine controls were matched to 195 HCC cases on all 3 matching factors.
All study subjects were interviewed by an experienced interviewer using a structured questionnaire that requested demographic information, history of tobacco and alcohol use, medical history of liver disease, diabetes and other upper gastrointestinal diseases, history of surgery and blood transfusion, and, for women only, menstrual and reproductive history and the use of hormones including oral contraceptives. We only included information up to 2 years before the diagnosis of HCC for cases and 2 years before the cancer diagnosis of the index case for the matched controls.
Our initial study protocol did not call for blood donation from control subjects and therefore no blood was taken from the first 111 control subjects recruited to the study. We began to request blood donation from control subjects in January 1992. Among those requested (n = 264), 198 (75%) consented. We also retrospectively asked for blood donation from control subjects interviewed from September 1989 to December 1991 (n = 60), and collected blood specimens from 22 (37%) of them. We compared demographic and lifestyle characteristics between control subjects with and without blood samples. The distributions between the 2 control groups were similar by age (mean, 60.1 years vs. 60.0 years), gender (60% vs. 64% men), cigarette smoking (67% vs. 61% ever smokers), alcohol consumption (67% vs. 65% of regular drinkers), and history of adult-onset diabetes (8% vs. 10%), but were different by race (81% vs. 67% of non-Hispanic whites; P = 0.002) and education (69% vs. 49% with at least some college education; P < 0.001).
Signed informed consent forms were obtained from all study subjects. Separate informed consent forms were used for tumor tissue retrieval, personal interview, and blood donation. The study protocol was approved by the institutional review board of the University of Southern California Keck School of Medicine. All subjects recruited during 1984–2002 were included in the current study.
Blood samples from cases and controls were processed and stored (−20 °C) in an identical manner. The assays used for testing serologic markers of HBV and HCV infections have been described.12 Briefly, we tested all study samples for the presence of hepatitis B surface antigen (HbsAg; AUSRIA; Abbott Laboratories, North Chicago, IL), and negative samples were further tested for the presence of antibodies to the hepatitis B core antigen (anti-HBc; Corab; Abbott Laboratories). All samples also were tested for the presence of anti-HCV, most of them (91%) using the Version 2.0 enzyme-linked immunoadsorbent assay (ELISA) (Ortho Diagnostic Systems, Raritan, NJ) with confirmation of positive samples by Version 2.0 recombinant immunoblot assay (RIBA) (Chiron, Emeryville, CA). We had reported that the 43 HCC cases tested for anti-HCV using Version 1.0 ELISA demonstrated rates of positivity that were similar to the remaining patients tested using the second-generation kits.12
One alcoholic drink was defined as 1 can of beer (360 mL), 1 glass of wine (103 mL), or 1 shot of spirits (45 mL). Cumulative drink-years over lifetime was defined as the number of drinks per day multiplied by the number of years of drinking. Similarly, cumulative pack-years of cigarettes over lifetime was defined as the number of packs (20 cigarettes per pack) of cigarettes smoked per day multiplied by the number of years of smoking.
We used unconditional logistical regression methods14 to analyze the entire study data comprising 295 patients with HCC and 435 control subjects. The strength of the exposure–HCC association was measured by odds ratios (ORs) and their 95% confidence intervals (95% CIs) and P values. The matching factors—age, gender, and race—were included as covariates in all models. We also used conditional logistical regression methods14 to examine the 195 matched case–control sets. The two sets of results were similar. All statistics were based on the analysis of 295 cases and 435 control subjects. All analyses were performed for males and females separately, and for both genders combined. We did not detect any difference in exposure–disease associations between the two genders. Therefore, only results based on both genders combined were reported. We also used multivariate logistic regression models to examine the relation between a specific type of alcoholic beverage (beer, wine, or spirits) and HCC risk while controlling for the consumption of the other two types. Analyses pertaining to serologic markers of HBV and HCV were restricted to subjects with known HBV and HCV markers only (245 cases and 220 controls).
We examined the potential interaction effects between any two independent risk factors on HCC risk. When examining whether the combined effect of two factors was greater than the multiplicative product of their individual effects, we used multivariate logistic regression models with the two main effects and their product term as covariates. When assessing if the combined effect of two factors on HCC was greater than the sum of the individual effects, we used the method described by Rothman.15 Under the null hypothesis of additivity, the synergy index (S) proposed by Rothman15 would take on the value 1. A lower 95% CI of S that is > 1 is indicative of a statistically significant synergistic effect between the two factors. All P values quoted are two sided.
The mean ages (± standard deviation [SD]) of HCC cases and control subjects were 60.6 (± 10.9) years and 60.1 (± 11.4) years, respectively. Sixty-five percent of cases and 62% of control subjects were males. Blacks, Hispanics, and non-Hispanic whites accounted for 15%, 28%, and 57% of cases, respectively, and 11%, 15%, and 75%, respectively, of controls (P < 0.001). Approximately 27% of cases and 17% of controls did not graduate from high school whereas 18% cases and 29% of controls graduated from college (P < 0.001). Therefore, all ORs presented were adjusted for level of education in addition to age, gender, and race.
The current study confirmed that HBV and HCV infections were independent risk factors for HCC. HCV infection demonstrated a stronger effect on HCC risk than HBV infection (Table 1).
|HBV/HCV markers||Cases (n = 245)||Controls (n = 220)||OR (95% CI)a|
|HBV positivec||86||25||3.6 (2.1–6.0)|
|Anti-HBc positive only||72||25||2.9 (1.7–5.0)|
|HBsAg positive||14||0||—b (5.0–b)|
|HCV positive||99||1||124.8 (17.1–908.6)|
|Both HCV and HBV negative||109||195||1.0|
|HCV negative and HBV positive||37||24||2.6 (1.5–4.7)|
|HCV positive and HBV negative||50||0||—b (28.1–b)|
|Both HCV and HBV positive||49||1||63.9 (8.6–475.3)|
Compared with nonsmokers/long-term ex-smokers (those who quit smoking ≥ 10 years ago), current smokers/recent ex-smokers (those who quit smoking within the last 10 years) had a statistically significant, 60% increase in risk of HCC after adjustment for potential confounders. Intensity of smoking and cumulative exposure to smoking were only modestly related to HCC risk. Results were similar when only subjects negative for both HBV and HCV serology were included in the analysis (Table 2).
|Characteristics||Total subjects||HBV and HCV-negative subjects onlya|
|Cases (n = 295)||Controls (n = 435)||OR (95% CI)b||Cases (n = 109)||Controls (n = 195)||OR (95% CI)b|
|Non or long-term ex-smokersc||160||300||1.0||62||139||1.0|
|Current or recent ex-smokersc||135||135||1.6 (1.1–2.2)||47||56||1.7 (1.0–3.0)|
|< 20 cigarettes per day||53||53||1.5 (1.0–2.5)||15||22||1.3 (0.6–2.9)|
|≥ 20 cigarettes per day||82||82||1.6 (1.1–2.4)||32||34||2.0 (1.1–3.7)|
|P for trend||0.02||0.03|
|Cumulative cigarettes smoked over lifetime (pack-year)d|
|20 to < 40||64||76||1.2 (0.8–1.9)||19||35||1.3 (0.6–2.7)|
|≥ 40||73||82||1.2 (0.8–1.9)||32||36||1.7 (0.9–3.4)|
|P for trend||0.32||0.12|
|> 0 to 2 drinks per day||66||183||0.6 (0.4–0.9)||23||86||0.5 (0.3–1.0)|
|> 2 to 4 drinks per day||43||57||1.4 (0.8–2.4)||20||24||1.5 (0.7–3.3)|
|> 4 drinks per day||95||47||3.2 (1.9–5.3)||29||22||2.1 (0.9–4.7)|
|P for trend||< 0.001||0.03|
|Cumulative alcohol consumed over lifetime (drink-year)e|
|1 to < 30||52||140||0.7 (0.4–1.0)||18||66||0.5 (0.3–1.0)|
|30 to < 60||35||55||1.1 (0.6–1.9)||7||26||0.5 (0.2–1.4)|
|60 to < 90||21||31||1.2 (0.6–2.3)||8||12||1.4 (0.5–4.0)|
|≥ 90||96||61||2.5 (1.5–4.1)||39||28||2.4 (1.1–5.1)|
|P for trend||< 0.001||0.007|
|History of diabetes|
|Yes||62||38||2.7 (1.6–4.3)||26||16||3.0 (1.5–6.2)|
|No treatment||7||3||3.4 (0.8–14.4)||2||1||1.9 (0.1–26.4)|
|Received treatment||55||35||2.6 (1.6–4.3)||24||15||3.1 (1.5–6.6)|
|Diagnosed within 10 years||28||20||2.4 (1.3–4.5)||9||11||1.5 (0.6–4.1)|
|Diagnosed ≥ 10 yrs ago||34||18||3.0 (1.6–5.7)||17||5||6.3 (2.1–18.8)|
Compared with subjects who never consumed alcohol on a weekly basis, those who consumed ≤ 2 drinks per day had an approximately 40% reduction in risk of HCC (OR = 0.6; 95% CI, 0.4–0.9) whereas those who drank > 4 drinks a day had a 3-fold increase in risk of HCC after adjustment for potential confounders. The increased risk of HCC associated with heavy alcohol consumption was primarily attributed to beer and spirits. The OR for each 10-g increment of ethanol was 1.12 (95% CI, 1.05–1.18) for beer and 1.10 (95% CI, 1.04–1.16) for spirits (both Ps for trend ≤ 0.001). Conversely, the corresponding figures were 1.07 (95% CI, 0.97–1.17) for wine (P for trend = 0.17). Subjects with < 30 drink-years had a reduced risk of HCC whereas those with ≥ 90 drink-years had an OR of 2.5 for HCC. Results were similar when only subjects negative for both HBV and HCV serology were included in the analysis (Table 2).
A history of diabetes was more prevalent in patients with HCC (21%) than control subjects (9%). The mean age (± SD) at diagnosis of diabetes was 51 (± 13) years for cases and 49 (± 12) for controls. Only 1 case had diabetes before age 18 years (at age 5 years). Subjects with a history of diabetes were at a 2.7-fold (95% CI, 1.6–4.3) increased risk of HCC after adjustment for age, gender, race, level of education, smoking status, and number of alcoholic drinks per day. This positive association did not vary significantly by treatment status or time interval between the diagnoses of diabetes and HCC. Exclusion of 19 HCC cases whose diabetes was diagnosed within 5 years of the cancer diagnosis and 13 controls who had diabetes within 5 years of the cancer diagnosis of the index cases did not materially change the results. The multivariate adjusted OR of HCC associated with diabetes was 2.7 (95% CI, 1.5–4.7). Results were similar and remained statistically significant when only subjects negative for both HBV and HCV serology were included in the analysis (Table 2). After further excluding subjects who currently or recently smoked cigarettes or consumed ≥ 4 alcohol drinks per day, 32 cases and 49 controls remained. Of these, 8 (25%) cases and 2 (4%) controls had diabetes, yielding an OR of 9.3 (95% CI, 1.7–51.2).
Table 3 presents the pairwise interactions of diabetes, heavy alcohol consumption, and cigarette smoking on risk of HCC. Heavy (> 4 drinks per day) drinkers with a history of diabetes were at very high risk for HCC (OR =17.3; 95% CI, 3.9–77.6) relative to nondiabetics who consumed ≤ 4 drinks per day. The combined effect was more than the sum of the two individual effects. Table 4 shows that the combined effects of viral hepatitis and diabetes, and viral hepatitis and heavy alcohol intake, were both more than the sum of the individual effects.
|Factor 1||Factor 2||Cases (n = 295)||Controls (n = 435)||OR (95% CI)a||Interaction effect|
|Additive S (95% CI)b||Multiplicative OR (95% CI)a|
|No||≤ 4 drinks per day||157||352||1.0|
|No||≥ 4 drinks per day||76||45||3.4 (2.2–5.4)|
|Yes||≤ 4 drinks per day||43||36||2.5 (1.5–4.0)|
|Yes||> 4 drinks per day||19||2||17.3 (3.9–77.6)||4.2 (2.6–5.8)||2.0 (0.4–10.1)|
|No||Current/recent ex-smokers||106||125||1.5 (1.0–2.2)|
|Yes||Non/long-term ex-smokers||33||28||2.5 (1.4–4.4)|
|Yes||Current/recent ex-smokers||29||10||4.9 (2.2–10.9)||2.0 (0.9–3.1)||1.3 (0.5–3.5)|
|Alcohol drinking||Cigarette smokingc|
|≤ 4 drinks per day||Non/long-term ex-smokers||125||276||1.0|
|≤ 4 drinks per day||Current/recent ex-smokers||35||24||3.3 (1.8–6.1)|
|> 4 drinks per day||Non/long-term ex-smokers||75||112||1.5 (1.0–2.2)|
|> 4 drinks per day||Current/recent ex-smokers||60||23||5.9 (3.3–10.4)||1.7 (0.9–2.6)||1.2 (0.5–2.6)|
|Factor 1||Factor 2||Cases (n = 245)||Controls (n = 220)||OR (95% CI)a||Interaction effect|
|Additive S (95% CI)b||Multiplicative OR (95% CI)a|
|Positive||Yes||20||1||47.8 (6.0–377.5)||4.8 (2.7–6.9)||1.8 (0.2–16.3)|
|HBV/HCV markers||Alcohol drinking|
|Negative||≤ 4 drinks per day||80||173||1.0|
|Negative||> 4 drinks per day||29||22||2.6 (1.3–5.1)|
|Positive||≤ 4 drinks per day||85||23||8.1 (4.6–14.4)|
|Positive||> 4 drinks per day||51||2||48.3 (11.0–212.1)||5.5 (3.9–7.0)||2.3 (0.5–12.1)|
|HBV/HCV markers||Cigarette smokingd|
|Negative||Current/recent ex-smokers||47||56||1.7 (1.0–3.0)|
|Positive||Non/long-term ex-smokers||68||17||9.1 (4.7–17.6)|
|Positive||Current/recent ex-smokers||68||8||15.0 (6.4–34.9)||1.6 (0.6–2.6)||1.0 (0.3–2.9)|
Thirteen years ago, we reported that infections with HBV and HCV, heavy alcohol consumption, cigarette smoking, and a history of diabetes were independent risk factors for HCC among blacks and Hispanic and non-Hispanic whites in Los Angeles.6, 16 The current study, an extension of the previous investigation with an additional 221 HCC cases, confirmed our earlier findings based on a relatively small sample size. A new finding of the current study is the demonstration of synergistic (i.e., more than additive) effects on HCC between HBV/HCV infection, heavy alcohol consumption, and diabetes. More importantly, the observed positive association between diabetes and HCC risk in subjects who had no infection with HBV and HCV ruled out the possibility that diabetes is the consequence of chronic liver disease.
In the U.S., incidence rates of HCC in both men and women have increased steadily during the past three decades.11, 13 The reasons for this steady increase remain unknown. In the current study, the proportions of HCV-positive HCC cases for the periods 1984–1990, 1991–1995, and 1996–2001 were 33% (19 of 58), 36% (31 of 85), and 48% (49 of 102), respectively, among blacks, Hispanics, and non-Hispanic whites in Los Angeles. Conversely, the corresponding figures of HBV-related (HBsAg and/or anti-HBc positive) HCC cases were 40% (23 of 58), 41% (35 of 85), and 27% (28 of 102). The current study findings of the increasing proportion of HCV-related HCC are consistent with recent reports by others.17 Alcohol consumption in the U.S. had been increasing steadily since the end of World War II, peaked in the early 1980s, and have leveled since then. The per capita consumption of alcohol increased from 7.8 L in 1960 to 10.4 L in 1980, an increase of 33%. In 1960, beer accounted for 48% of the alcohol consumption in the U.S., spirits for 41%, and wine for 11%. The corresponding figures in 1980 were 50%, 38%, and 12%.18 It is conceivable that heavy alcohol consumption, HCV infection, and the current epidemic of obesity/diabetes have contributed to the increasing incidence of HCC in the U.S.
The current study examined, for the first time, the possible difference in risk of HCC across the three types of alcoholic beverages (wine, beer, and spirits). We found that the adverse effect of ethanol on HCC risk was mainly derived from beer and spirits. Relative to beer and spirits, ethanol from wine exhibited a considerably weaker, albeit still a positive relation with HCC risk. Oxidative stress due to free radical formation after alcohol intake has been implicated in the pathogenesis of alcohol-induced liver disease.19, 20 Polyphenolic compounds in wine, particularly red wine, have shown antioxidative properties.21 Intake of red wine increases antioxidant activity, leading to reduced oxidative by-products in the blood.22 A recent report showed that wine drinkers experienced a significantly lower risk of developing cirrhosis relative to beer/spirits-only drinkers with comparable amounts of ethanol intake.23 It is conceivable that potent antioxidants in wine offset the risk of alcohol-induced liver disease, including HCC, in wine drinkers.
Our observation of a positive association between diabetes and HCC risk is consistent with that from a recent hospital-based, case–control study of HCC in Texas.8, 24 Furthermore, the current study demonstrates that this positive association is present in subjects free of viral hepatitis, ruling out the possibility that diabetes is a consequence of chronic liver disease. Previous studies have found that approximately 10–20% of patients with viral cirrhosis have overt diabetes or impaired glucose tolerance.25, 26 The incidence of diabetes in the U.S. has been steadily increasing during the past 40 years. The prevalence of diabetes increased from 1% in 1960 to 3–5% around 1990, and to 8% in 2000.27–30 The increases were observed in both genders, all ages, all ethnic groups, and all levels of education. Obesity is the most important risk factor for diabetes, and the two conditions are highly correlated (r = 0.64, P < 0.001).28 In the U.S., the prevalence of obesity (body mass index > 30 kg/m2) has doubled during the past 30 years, from 15% in 1971–1974 to 31% in 1999–2000.31 Obesity has been found to be associated with a statistically, significant increased risk of death from liver carcinoma in U.S. men and women.32 One would predict an increasingly important role of diabetes/obesity in HCC in the U.S. as an increasing proportion of older Americans are afflicted with these two conditions.
Clinical studies have shown that 40–100% of patients with nonalcoholic steatohepatitis (NASH) are obese and that 20–75% of them have a history of adult-onset diabetes.33 Furthermore, the severity of fibrosis among patients with NASH is positively associated with obesity and diabetes.34 It also is known that a high percentage (≤ 70%) of patients with cryptogenic cirrhosis were obese and/or diabetic.35 These clinical data suggest that obesity/diabetes are risk factors for hepatic fibrosis and progression to cirrhosis, a recognized primary risk factor for HCC, regardless of the cause of cirrhosis.36
The biologic mechanism of obesity and/or diabetes causing HCC is unknown. Obesity can lead to insulin resistance and steatosis, which is associated with the release of inflammatory mediators such as tumor necrosis factor (TNF)-α in Kuppfer cells, which in turn enhances the production of cytokines including interleukin (IL)-6 and IL-8, leading to steatohepatitis. Recent data showed that plasma levels of IL-8 and IL-6 were 30–40% higher in obese than lean subjects. Obese subjects, after a 15% weight loss through an energy-restricted diet, had a 25–30% reduction in plasma levels of IL-6 and TNF-α, accompanied by improvement in insulin sensitivity.37 Parenchymal inflammation, hepatocyte necrosis, and ballooning hepatocyte degeneration are typical histologic characteristics of NASH.33 Therefore, obesity and diabetes can cause hepatic inflammation, leading to oxidative stress and lipid peroxidation of the phospholipid constituents of hepatocyte and intracellular membranes, resulting in hepatocyte injury and necrosis. The injury to hepatocytes leads to the activation of hepatic stellate cells, excessive secretion of extracellular matix components, the formation of hepatic fibrosis, and ultimately the development of cirrhosis.38
One important finding of the current study is the synergism (excess over additivity) on HCC risk among diabetes, chronic HBV/HCV infection, and heavy alcohol consumption. Our findings were consistent with those of previous studies.7, 8, 39, 40 Conversely, there was an absence of synergism between cigarette smoking and the other risk factors for HCC. Thus, relative to the other lifestyle-related risk factors for HCC, the impact on HCC risk from cigarette smoking was relatively modest. In addition, there is a possibility that the modest smoking–HCC association represents residual confounding by alcohol on HCC risk.
The current study, like any other case–control study, potentially could be limited by information bias. One of the limitations is the low recruitment rate of cases because of the rapidly fatal nature of HCC (the median time interval between diagnosis and death is approximately 3 months).13 Eighty-four percent of eligible patients with HCC died before our attempted contact. However, there is no evidence of a differential demographic profile between eligible patients excluded from the study and those recruited. Another limitation is that we did not collect blood samples from control subjects at the inception of the study. We observed a lower prevalence rate of anti-HCV positivity in our controls (0.5%) than the reported rate in another U.S. population (1.8%).41 This could be a true reflection of the actual HCV prevalence in Los Angeles County or a function of a selection bias because we did not collect blood samples from all eligible control subjects. However, there is no evidence that our control subjects with and without blood samples differed with regard to demographic and lifestyle factors such as age, gender, cigarette smoking, alcohol drinking, and history of diabetes. Another concern is the potential bias in the diagnosis of diabetes among HCC cases. Patients with viral hepatitis might be closely examined by their physicians and thus more likely to have their diabetes detected than those without viral hepatitis. This would result in a spurious association between diabetes and HCC risk. However, we noted a positive, and similarly strong diabetes–HCC association among subjects negative for HBV/HCV serology, thus ruling out the possibility of this potential detection bias.
The strengths of the current study include a population-based study design with a relatively large sample size. To our knowledge, this is the largest population-based, case–control study of HCC in U.S. blacks and whites. Our results should be generalizable to the entire non-Asian population in the U.S.
HBV and HCV infections, use of alcohol and tobacco, and a history of diabetes are independent predictors of HCC risk in Hispanic and non-Hispanic whites and blacks in Los Angeles County. Heavy alcohol consumption, diabetes, and viral hepatitis exert synergistic effects on risk of HCC. These factors are likely contributors to the rising incidence of HCC in the U.S.
- 1Cancer incidence in five continents. Volume VIII. IARC Scientific Pub. No. 155. Lyon: International Agency for Research on Cancer, 2002., , , , .
- 13Surveillance, Epidemiology, and End Results (SEER) program (available from URL: www.seer.cancer.gov). Public-use data (1973-2000). Bethesda: National Cancer Institute, DCCPS, Surveillance Research Program, Cancer Statistics Branch, 2003 [released April 2003, based on the November 2002 submission].
- 14Statistical methods in cancer research. Volume 1. The analysis of case-control studies. IARC Sci. Pub. No. 32. Lyon: International Association for Research on Cancer, 1980., .
- 24The role of diabetes in hepatocellular carcinoma: a case-control study among United States Veterans. Am J Gastroenterol. 2001; 96: 2462–2467., , .Direct Link:
- 27Prevalence and incidence of non-insulin-dependent diabetes in America. 2nd edition. Publication No. 95-1468. Bethesda: National Diabetes Data Group of the National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, 1995: 47–88., , .