Abstract

Background: Retinol and its derivatives (retinoids), which have antioxidant activity and promote cell differentiation, may protect against the development of hepatocellular carcinoma (HCC) by controlling hepatocellular differentiation and reducing inflammatory responses. Methods: We examined prospectively the relationship between prediagnostic serum concentrations of retinol, α-carotene; β-carotene; β-cryptoxanthin; lutein; lycopene; zeaxanthin; α-, γ-, and δ-tocopherols; and selenium and the risk of developing HCC among 213 patients with HCC and 1087 matched control subjects from a cohort of 18 244 men in Shanghai, China, who were monitored from 1986 through 2001. Odds ratios (ORs) and 95% confidence intervals (CIs) for men by quartile of serum concentrations of micronutrients were estimated by using logistic regression with adjustment for cigarette smoking status, alcohol intake, self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment, and seropositivity for hepatitis B surface antigen (HBsAg). All statistical tests were two-sided. Results: Men with high prediagnostic serum retinol levels had a lower risk of HCC than men in the lowest quartile (Q2 versus Q1, OR = 0.37, 95% CI = 0.22 to 0.61; Q3 versus Q1, OR = 0.30, 95% CI = 0.17 to 0.50; and Q4 versus Q1, OR = 0.13, 95% CI = 0.06 to 0.26; Ptrend <.001). A statistically significant interaction was observed between retinol and HBsAg seropositivity on HCC risk; HBsAg-positive men in the lowest tertile of retinol had a greater than 70-fold higher risk (OR = 72.7, 95% CI = 31.6 to 167.4) of HCC than HBsAg-negative men in the highest tertile of retinol ( Pinteraction = .018). No independent effect of serum levels of α-carotene; β-carotene; β-cryptoxanthin; lutein; lycopene; zeaxanthin; α-, γ-, and δ-tocopherols; or selenium on HCC risk were observed. Conclusion: High prediagnostic serum level of retinol is associated with a decreased risk of HCC in this population.

Chronic infection with hepatitis B virus (HBV) is by far the most important risk factor for hepatocellular carcinoma (HCC) in humans and is the primary cause of this cancer in high-risk areas, including China and Africa ( 1 , 2 ) . Chronic infection with hepatitis C virus (HCV) is another risk factor for HCC and has an increasingly prominent role in the development of HCC in countries where rates of infection with HBV are relatively low ( 3 ) . Other environmental risk factors for HCC include dietary aflatoxin, which plays a prominent role in high-risk areas such as China and Africa; excessive alcohol intake; cigarette smoking; diabetes; and obesity ( 46 ) .

Carcinogenesis is characterized by aberrant cell differentiation ( 7 ) . Retinol and its derivatives (retinoids) have vital roles in controlling cellular growth and differentiation ( 8 ) . Retinoids have been shown to inhibit the transformation (activation) of hepatic stellate cells ( 9 ) and to suppress the proliferation of human hepatoma cells ( 10 ) . Retinol has also been shown to inhibit the formation of aflatoxin B 1 (AFB 1 )–DNA adducts in hepatocytes ( 11 ) , a crucial step in aflatoxin-induced hepatocarcinogenesis, and to inhibit the development of preneoplastic lesions in the liver of rats treated with chemical carcinogens ( 12 ) .

Persistent inflammation caused by chronic infection with HBV and/or HCV is believed to be one mechanism of hepatocarcinogenesis and is probably mediated by inflammatory cytokines ( 13 ) . Retinoids inhibit the production of proinflammatory cytokines in cultured macrophages and reduce inflammatory reactions ( 14 ) . Reactive oxygen species induced by chronic inflammation can cause damage to DNA, proteins, and lipids. Reactive oxygen species–induced DNA damage is believed to be directly involved in hepatocarcinogenesis ( 15 , 16 ) . Antioxidants, such as retinol and carotenoids, can neutralize reactive oxygen species and can possibly prevent the development of HCC. Several antioxidants, including β-carotene, vitamin E, and selenium, have been shown to inhibit chemically induced liver cancer in rodents ( 1719 ) .

Prospective epidemiologic studies of serum retinol and other antioxidants in relation to HCC risk are scarce. In this study, we examined the relationship between concentrations of antioxidant micronutrients in prediagnostic serum samples and the risk of developing HCC in a Chinese cohort in Shanghai, China, after 15 years of follow-up. Micronutrients measured included retinol, specific carotenoids (α-carotene, β-carotene, β-cryptoxanthin, lutein, lycopene, and zeaxanthin); α-, γ-, and δ-tocopherols; and selenium.

S UBJECTS AND M ETHODS

Study Population

The design of the Shanghai Cohort Study has been described in detail elsewhere ( 20 , 21 ) . In brief, 18 244 men (approximately 80% of the eligible subjects) aged 45–64 years and had no history of cancer at recruitment were enrolled in the study between January 1, 1986, and September 30, 1989. Participants were interviewed in person by using a structured questionnaire to obtain information on demographic characteristics, use of tobacco and alcohol, usual adult diet, and medical history. At completion of the interview, a 10-mL nonfasting blood sample was collected from each participant. Multiple aliquots of serum from each study subject were then stored at −70 °C. Written informed consent was obtained from surviving participants at follow-up in 1997. The institutional review boards at the University of Southern California, the University of Minnesota, and the Shanghai Cancer Institute approved this study on an ongoing basis.

At recruitment, we asked each participant whether he had ever been told by a physician that he had any of the following diseases: hepatitis (infectious, serum, or unknown type), liver cirrhosis, other liver disease, diabetes, gastric ulcer, duodenal ulcer, gallstones, other gallbladder conditions, tuberculosis, high blood pressure, asthma, emphysema, hemorrhoid, diverticulosis, polyposis coli, ulcerative colitis, adenomatous polyps, schistosomiasis, other parasitic diseases, or cancer. If the answer was yes, he was asked to provide the year of the first diagnosis and the treatment for each disease. We also asked each participant whether he had ever drunk alcoholic beverages at least once a week for 6 months or more. If the answer was yes, he was asked to provide the typical amount of beer, wine, and/or spirits (each) consumed. One drink was defined as 360 g of beer (12.6 g of ethanol), 103 g of wine (12.3 g of ethanol), or 30 g of spirits (12.9 g of ethanol) ( 22 ) . Smokers were identified as men who smoked at least one cigarette per day for 6 months or more.

Case Patients

Incident cancer cases and deaths among cohort participants through December 31, 2001, were identified through routine reviews of reports from the population-based Shanghai Cancer Registry and the Shanghai Municipal Vital Statistics Office and by annual in-person re-interviews of surviving cohort members. A total of 422 (2.3%) cohort participants were lost to follow-up as of December 31, 2001.

As of December 31, 2001 (this study's cutoff date), the study had accumulated 229 966 person-years of observation. A total of 214 cohort participants who were free of cancer at recruitment had developed liver cancer. These 214 HCC case patients were diagnosed by histopathologic analysis (n = 39), elevated serum α-fetoprotein with consistent clinical and radiologic history (n = 53), positive computerized axial tomography scan and/or ultrasonography with consistent clinical history (n = 115), or by death certificate only (n = 7).

Control Subjects

For each case patient, multiple cancer-free control subjects, who were individually matched to the index case patients by date of birth (within 2 years), date of blood collection (within 1 month), and neighborhood of residence at recruitment, were randomly chosen from the cohort. Ten control subjects were matched to each of the first six identified case patients with HCC, and five control subjects were matched to each of the remaining 208 case patients.

Laboratory Tests

Serum samples in each matched set (the case patient and his matched control subjects) were arranged in random order, identified only by unique codes, and assayed in the same batch for all laboratory measurements. Test samples were processed in a dim room to avoid potential change in chemical structures of retinol and certain carotenoids by strong light. Serum concentrations of α-carotene; β-carotene; β-cryptoxanthin; lutein; lycopene; zeaxanthin; retinol; and α-, γ-, and δ-tocopherols were determined by high-performance liquid chromatography using methods described previously ( 23 ) . Serum selenium concentrations were measured by Zeeman graphite atomic absorption spectrophotometry ( 24 ) . The within-batch coefficients of variation were 0.8% for retinol, 1.3% for selenium, and 2.6%–5.7% for the other micronutrients measured. The between-batch coefficients of variation were 3.9% for retinol, 2.3% for selenium, and 4.4%–13.3% for other micronutrients.

We also tested all study samples for the presence of hepatitis B surface antigen (HBsAg), using a standard radioimmunoassay assay (AUSRIA; Abbott Laboratories, Abbott Park, IL). Assays for the antibodies to HCV and urinary biomarkers of dietary exposure to aflatoxin were described previously ( 20 , 25 , 26 ) .

One case patient and eight control subjects had missing serum retinol and carotenoid measurements. Thus, this study included 213 case patients and 1087 control subjects.

Statistical Analysis

The distributions of all serum micronutrients investigated were markedly skewed toward high values. Therefore, formal statistical testing was performed on logarithmically transformed values, and geometric (as opposed to arithmetic) means are presented. We used the analysis-of-covariance method ( 27 ) to compare differences in serum concentrations of micronutrients between control subjects who were 1) smokers and nonsmokers, 2) alcohol drinkers and nondrinkers, and 3) positive and negative for HBsAg serology. We also used analysis-of-covariance method to compare the differences in serum concentrations of micronutrients between HCC case patients and their matched control subjects.

We used standard statistical methods to analyze data from matched case–control sets ( 28 ) . Conditional logistic regression models were used to calculate odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) and P values. Study subjects were grouped into tertiles or quartiles based on the distributions of serum micronutrients among all control subjects (Supplementary Table 1; available at http://jncicancerspectrum.oxfordjournals.org/jnci/content/vol98/issue7 ). The linear trend test for the association between serum concentrations of micronutrient and HCC risk was based on ordinal values.

To evaluate the effect of a micronutrient independent of other nutrients on HCC risk, we used two approaches: 1) a multivariable logistic regression model with other nutrients as covariates and 2) subgroup analyses in subjects stratified by tertile levels of other nutrients. To examine whether the combined effect of retinol and HBsAg with HCC risk was greater than the multiplicative product of their individual effects, we used a conditional logistic regression model that simultaneously included the product term and the main effect terms of the two variables. To adjust for potential confounding effects of other identified risk factors on the association of specific nutrients with HCC risk, we included the following variables in multivariable logistic regression models: cigarette smoking (nonsmokers or ever smokers), heavy alcohol consumption (<4 drinks/day including nondrinkers or ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no or yes), and HBsAg seropositivity (negative or positive). For all stratified analyses, matched sets were broken and unconditional logistic regression models were used. The original matching factors (age, year of blood collection, and neighborhood of residence at recruitment) were adjusted for in all analyses on unmatched data sets.

Statistical analyses were performed using the SAS version 9.1 (SAS Institute Inc., Cary, NC) and Epilog windows version 1.0 (Epicenter Software, Pasadena, CA) statistical software packages. All P values are two-sided. P <.05 was considered statistically significant.

R ESULTS

We first compared the distributions of selected demographic, lifestyle, and risk factors of case patients with control subjects. The mean age (± standard deviation) of case patients (n = 213) at diagnosis of HCC was 63.5 (±5.8) years, and the mean age of control subjects (n = 1087) at the time of HCC diagnosis of the matched case patients was 63.4 (±5.7) years. The average time interval between blood draw and cancer diagnosis among case patients was 6.9 (±3.8) years (range = 1 month to 15.8 years). The case patients and control subjects had similar mean body mass indices (22.0 versus 22.2 kg/m 2 ; P = .31). Cigarette smoking, heavy alcohol consumption (≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis, HBsAg seropositivity, and exposure to dietary aflatoxin were risk factors for HCC in the study population ( Table 1 ). As previously reported ( 25 ) , the role of HCV infection in HCC development in this study population was negligible; only one of 76 case patients and one of 405 control subjects tested positive for the antibodies to hepatitis C virus. One (0.5%) case patient and nine (0.8%) control subjects reported a history of physician-diagnosed diabetes.

Table 1.

Distributions of risk factors for hepatocellular carcinoma (HCC) in HCC patients and control subjects, Shanghai Cohort Study, 1986–2001

Risk factor HCC patients, % (n = 213) Control subjects, % (n = 1087)  OR (95% CI) * 
Cigarette smoking    
    Nonsmokers 33.8 43.5 1.00 (Referent) 
    <20 cigarettes/day 31.5 27.0 1.50 (1.04 to 2.16) 
    ≥20 cigarettes/day 34.7 29.5 1.49 (1.05 to 2.11) 
Alcohol drinking     
    Nondrinkers 59.1 56.6 1.00 (Referent) 
    <4 drinks/day 31.5 36.4 1.09 (0.76 to 1.57) 
    ≥4 drinks/day 9.4 7.0 2.77 (1.49 to 5.15) 
Self-reported history of physician-diagnosed hepatitis or liver cirrhosis    
    No 57.7 88.4 1.00 (Referent) 
    Yes 42.3 11.6 5.39 (3.85 to 7.53) 
HBsAg serology    
    Negative 38.5 90.4 1.00 (Referent) 
    Positive 61.5 9.6 15.4 (10.4 to 22.9) 
Anti-HCV serology     
    Negative 98.7 99.8 1.00 (Referent) 
    Positive 1.3 0.2 4.47 (0.28 to 71.8) 
Urinary aflatoxin biomarkers §    
    Negative 24.0 50.2 1.00 (Referent) 
    Positive 76.0 49.8 3.25 (1.63 to 6.48) 
        AFB 1 -N 7 -Gua negative  40.0 38.1 2.22 (1.03 to 4.77) 
        AFB 1 -N 7 -Gua positive  36.0 11.7 7.24 (2.98 to 17.6) 
Risk factor HCC patients, % (n = 213) Control subjects, % (n = 1087)  OR (95% CI) * 
Cigarette smoking    
    Nonsmokers 33.8 43.5 1.00 (Referent) 
    <20 cigarettes/day 31.5 27.0 1.50 (1.04 to 2.16) 
    ≥20 cigarettes/day 34.7 29.5 1.49 (1.05 to 2.11) 
Alcohol drinking     
    Nondrinkers 59.1 56.6 1.00 (Referent) 
    <4 drinks/day 31.5 36.4 1.09 (0.76 to 1.57) 
    ≥4 drinks/day 9.4 7.0 2.77 (1.49 to 5.15) 
Self-reported history of physician-diagnosed hepatitis or liver cirrhosis    
    No 57.7 88.4 1.00 (Referent) 
    Yes 42.3 11.6 5.39 (3.85 to 7.53) 
HBsAg serology    
    Negative 38.5 90.4 1.00 (Referent) 
    Positive 61.5 9.6 15.4 (10.4 to 22.9) 
Anti-HCV serology     
    Negative 98.7 99.8 1.00 (Referent) 
    Positive 1.3 0.2 4.47 (0.28 to 71.8) 
Urinary aflatoxin biomarkers §    
    Negative 24.0 50.2 1.00 (Referent) 
    Positive 76.0 49.8 3.25 (1.63 to 6.48) 
        AFB 1 -N 7 -Gua negative  40.0 38.1 2.22 (1.03 to 4.77) 
        AFB 1 -N 7 -Gua positive  36.0 11.7 7.24 (2.98 to 17.6) 
*

Odds ratios (ORs) were calculated by using conditional logistic regression models that retained a matched set consisting of five to 10 control subjects who were individually matched to the index case patient by date of birth (within 2 years), date of blood draw (within 1 month), and neighborhood of residence at recruitment. CI = confidence interval.

Adjusted for serum level of retinol, Ptrend = .016.

Serology of the antibodies to hepatitis C virus (anti-HCV) was determined for 76 patients and for 405 control subjects.

§

Urinary aflatoxin biomarkers including aflatoxin B 1 (AFB 1 ), its metabolites AFM 1 and AFP 1 , and AFB 1 -N 7 -guanine adduct (AFB 1 -N 7 -Gua) were determined for 50 patients and 265 control subjects.

In this study, alcohol drinkers had a statistically significantly higher level of circulating retinol than nondrinkers ( Table 2 ). Therefore, adjustment for serum retinol concentration was made in the analysis of the association between alcohol intake and HCC risk. Before adjustment for retinol, the odds ratios for HCC in men who consumed less than four and four or more drinks of alcoholic beverages per day were similar to that of nondrinkers. However, after adjustment for retinol, the odds ratio for heavy drinkers (≥4 drinks per day) were statistically significantly different (OR = 2.77, 95% CI = 1.49 to 5.15; Ptrend = .016) from those of nondrinkers ( Table 1 ).

Table 2.

Geometric means of concentrations of serum micronutrients by selected risk factors for hepatocellular carcinoma among control subjects only, Shanghai Cohort Study, 1986–2001

  HBsAg serology *
 
   Alcohol drinking
 
   Cigarette smoking
 
  
Micronutrient Negative (n = 983) Positive (n = 104) P§ No (n = 615) Yes (n = 472) P No (n = 473) Yes (n = 614) P 
Retinol, μg/dL 45.4 39.7 <.001 40.1 45.0 <.001 43.1 41.9 .11 
α-Carotene, μg/dL 1.31 1.32 .77 1.32 1.31 .59 1.36 1.27 .02 
β-Carotene, μg/dL 11.1 11.4 .75 11.3 11.2 .86 12.1 10.4 <.001 
β-Cryptoxanthin, μg/dL 3.32 3.86 .05 3.68 3.47 .21 3.98 3.22 <.001 
Lutein, μg/dL 112.4 110.6 .72 106.2 117.0 <.001 116.8 106.4 <.001 
Lycopene, μg/dL 2.13 2.62 .02 2.34 2.39 .70 2.54 2.20 .008 
Zeaxanthin, μg/dL 1.50 1.82 .03 1.69 1.61 .35 1.64 1.66 .88 
α-Tocopherol, μg/mL 8.91 8.41 .18 8.74 8.58 .47 8.93 8.40 .02 
γ-Tocopherol, μg/mL 1.21 1.17 .52 1.20 1.17 .39 1.20 1.17 .51 
δ-Tocopherol, μg/mL 0.12 0.11 .39 0.12 0.11 .23 0.12 0.11 .23 
Selenium, μg/dL 11.0 10.8 .40 10.8 11.0 .12 11.1 10.8 .03 
  HBsAg serology *
 
   Alcohol drinking
 
   Cigarette smoking
 
  
Micronutrient Negative (n = 983) Positive (n = 104) P§ No (n = 615) Yes (n = 472) P No (n = 473) Yes (n = 614) P 
Retinol, μg/dL 45.4 39.7 <.001 40.1 45.0 <.001 43.1 41.9 .11 
α-Carotene, μg/dL 1.31 1.32 .77 1.32 1.31 .59 1.36 1.27 .02 
β-Carotene, μg/dL 11.1 11.4 .75 11.3 11.2 .86 12.1 10.4 <.001 
β-Cryptoxanthin, μg/dL 3.32 3.86 .05 3.68 3.47 .21 3.98 3.22 <.001 
Lutein, μg/dL 112.4 110.6 .72 106.2 117.0 <.001 116.8 106.4 <.001 
Lycopene, μg/dL 2.13 2.62 .02 2.34 2.39 .70 2.54 2.20 .008 
Zeaxanthin, μg/dL 1.50 1.82 .03 1.69 1.61 .35 1.64 1.66 .88 
α-Tocopherol, μg/mL 8.91 8.41 .18 8.74 8.58 .47 8.93 8.40 .02 
γ-Tocopherol, μg/mL 1.21 1.17 .52 1.20 1.17 .39 1.20 1.17 .51 
δ-Tocopherol, μg/mL 0.12 0.11 .39 0.12 0.11 .23 0.12 0.11 .23 
Selenium, μg/dL 11.0 10.8 .40 10.8 11.0 .12 11.1 10.8 .03 
*

Adjusted for age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, cigarette smoking (nonsmokers, ever smokers), alcohol intake (nondrinkers, ever drinkers), and self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes) as covariates.

Adjusted for age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, cigarette smoking (nonsmokers, ever smokers), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and seropositivity for hepatitis B surface antigen (HBsAg) (negative, positive).

Adjusted for age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, alcohol intake (nondrinkers, ever drinkers), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and seropositivity for HBsAg (no, yes).

§

P values (two-sided) were calculated using analysis of covariance regression models that also included all covariates listed above.

Self-reported history of physician-diagnosed hepatitis or liver cirrhosis at baseline was associated with the risk of developing HCC after adjustment for HBsAg seropositivity. Most HCC patients who self reported a history of hepatitis or liver cirrhosis (84%) tested positive for HBsAg, whereas only 45% patients who did not report such a history tested positive. Among control subjects, 20% of those with a history of hepatitis or liver cirrhosis versus 8% of those without a history tested positive for HBsAg. After adjustment for HBsAg seropositivity, self-reported history of hepatitis or liver cirrhosis remained statistically significantly associated with risk of HCC (adjusted OR = 2.95; 95% CI = 1.94 to 4.48). Therefore, both HBsAg seropositivity and self-reported history of hepatitis or liver cirrhosis were included as covariates in all regression models in evaluating the independent effect of micronutrients on HCC risk.

We examined the associations of cigarette smoking, alcohol intake, and chronic HBV infection with circulating levels of micronutrients studied among control subjects only. After adjustment for self-reported history of physician-diagnosed hepatitis or liver cirrhosis, HBsAg-positive subjects had a statistically significantly lower (by 13%) serum retinol level than that of HBsAg-negative subjects ( Table 2 ). By contrast, regular alcohol drinkers had statistically significantly higher serum level of retinol than that of nondrinkers, whereas smokers and nonsmokers had similar levels of retinol ( Table 2 ). For other micronutrients studied, smoking was associated with statistically significantly lower levels of α-tocopherol, selenium, and all specific carotenoids, except zeaxanthin, whereas alcohol intake was associated with increased levels of lutein. HBsAg-positive subjects had slightly higher levels of β-cryptoxanthin, lycopene, and zeaxanthin than those of HBsAg-negative subjects ( Table 2 ).

We compared serum levels of micronutrients for HCC case patients with those of control subjects. Overall, case patients had statistically significantly lower geometric means of retinol, β-carotene, β-cryptoxanthin, and α- and γ-tocopherol than did control subjects after adjustment for HBsAg seropositivity, history of hepatitis or liver cirrhosis at recruitment, heavy alcohol consumption, and cigarette smoking ( Table 3 ). After exclusion of HBsAg-positive subjects, the differences in levels of retinol and β-carotene between the case and control subgroups remained statistically significant ( Table 3 ). Further exclusion of subjects with a positive history of hepatitis or liver cirrhosis did not materially change the results (data not shown). The geometric means of all other micronutrients measured were not statistically significantly different between case patients and control subjects ( Table 3 ).

Table 3.

Geometric mean of prediagnostic serum concentrations of micronutrients in hepatocellular carcinoma (HCC) patients and control subjects, Shanghai Cohort Study, 1986–2001

  All subjects *
 
   Subjects negative for HBsAg
 
  
Micronutrients HCC patients (n = 213) Control subjects (n = 1087) P HCC patients (n = 82) Control subjects (n = 983) P 
Retinol, μg/dL 32.5 41.8 <.001 38.1 45.6 <.001 
α-Carotene, μg/dL 1.19 1.25 .12 1.29 1.37 .24 
β-Carotene, μg/dL 9.44 11.57 .001 9.23 11.06 .04 
β-Cryptoxanthin, μg/dL 2.99 3.42 .03 2.92 3.31 .14 
Lutein, μg/dL 111.3 114.8 .41 101.3 110.9 .07 
Lycopene, μg/dL 2.09 2.18 .44 2.11 2.32 .32 
Zeaxanthin, μg/dL 1.49 1.58 .40 1.49 1.49 .99 
α-Tocopherol, μg/mL 8.13 8.79 .02 8.10 8.59 .22 
γ-Tocopherol, μg/mL 1.08 1.18 .03 1.09 1.16 .26 
δ-Tocopherol, μg/mL 0.10 0.12 .12 0.11 0.12 .32 
Selenium, μg/dL 10.9 10.9 .58 10.8 10.9 .72 
  All subjects *
 
   Subjects negative for HBsAg
 
  
Micronutrients HCC patients (n = 213) Control subjects (n = 1087) P HCC patients (n = 82) Control subjects (n = 983) P 
Retinol, μg/dL 32.5 41.8 <.001 38.1 45.6 <.001 
α-Carotene, μg/dL 1.19 1.25 .12 1.29 1.37 .24 
β-Carotene, μg/dL 9.44 11.57 .001 9.23 11.06 .04 
β-Cryptoxanthin, μg/dL 2.99 3.42 .03 2.92 3.31 .14 
Lutein, μg/dL 111.3 114.8 .41 101.3 110.9 .07 
Lycopene, μg/dL 2.09 2.18 .44 2.11 2.32 .32 
Zeaxanthin, μg/dL 1.49 1.58 .40 1.49 1.49 .99 
α-Tocopherol, μg/mL 8.13 8.79 .02 8.10 8.59 .22 
γ-Tocopherol, μg/mL 1.08 1.18 .03 1.09 1.16 .26 
δ-Tocopherol, μg/mL 0.10 0.12 .12 0.11 0.12 .32 
Selenium, μg/dL 10.9 10.9 .58 10.8 10.9 .72 
*

Geometric means were calculated using analysis-of-covariance regression models that retained a matched set consisting of five to 10 control subjects who were individually matched to the index case by date of birth (within 2 years), date of blood draw (within 1 month), and neighborhood of residence at recruitment. Covariates were cigarette smoking (nonsmokers, ever smokers), heavy alcohol consumption (nondrinkers or <4 drinks/day, ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and seropositivity for hepatitis B surface antigen (HBsAg) (negative, positive).

A total of 131 case patients and 104 control subjects who tested positive for HBsAg were excluded from these analyses. The matched case–control sets were not retained in analysis-of-covariance regression models that included age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, neighborhood of residence at recruitment, and all other covariates listed above.

P values (two-sided) were calculated by using analysis-of-covariance regression models that also included all covariates listed above.

We examined the association between quartile levels of prediagnostic serum micronutrients and HCC risk. Before adjustment for potential confounders, increased serum levels of retinol; α- and β-carotene; and α-, γ-, and δ-tocopherol were associated with statistically significantly reduced risk of HCC (all Ptrend <.05, Table 4 ). After adjustment for HBsAg seropositivity, self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment, heavy alcohol consumption (≥4 drinks/day), and cigarette smoking, the inverse associations with retinol, β-carotene, and γ-tocopherol remained statistically significant; the adjusted inverse relationship between α-tocopherol and HCC risk was not. Further adjustment for serum levels of β-carotene, α-tocopherol, and γ-tocopherol did not materially change the inverse association between increasing retinol and reduced HCC risk ( Ptrend <.001, Table 4 ). However, further adjustment for retinol considerably diminished the inverse associations between α- and γ-tocopherols and HCC risk ( Ptrend = .92 and Ptrend = .16, respectively). The retinol-adjusted inverse association between increasing β-carotene level and reduced HCC risk remained statistically significant ( Ptrend = .03).

Table 4.

Odds ratios (ORs) and 95% confidence intervals (CIs) of hepatocellular carcinoma (HCC) in relation to prediagnostic serum levels of micronutrients, Shanghai Cohort Study, 1986–2001

  OR (95% CI) by quartile level of micronutrient *
 
    
Micronutrient 1 (low) 4 (high) Ptrend 
Retinol, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.22 (0.14 to 0.32) 0.17 (0.11 to 0.27) 0.08 (0.05 to 0.15) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.37 (0.22 to 0.61) 0.30 (0.17 to 0.50) 0.13 (0.06 to 0.26)  <.001  
α-Carotene, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.82 (0.51 to 1.32) 0.75 (0.46 to 1.22) 0.50 (0.28 to 0.90) .02 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.99 (0.54 to 1.82) 0.80 (0.42 to 1.49) 0.69 (0.34 to 1.40) .27 
β-Carotene, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.86 (0.58 to 1.27) 0.58 (0.37 to 0.89) 0.57 (0.37 to 0.89) .004 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.71 (0.43 to 1.17) 0.38 (0.22 to 0.67) 0.59 (0.34 to 1.02)  .01  
β-Cryptoxanthin, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.83 (0.55 to 1.24) 0.77 (0.50 to 1.18) 0.69 (0.45 to 1.08) .10 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.92 (0.55 to 1.53) 0.87 (0.50 to 1.51) 0.85 (0.48 to 1.51) .56 
Lutein, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.77 (0.51 to 1.16) 0.66 (0.43 to 1.02) 0.92 (0.61 to 1.40) .55 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.71 (0.42 to 1.18) 0.56 (0.33 to 0.97) 0.82 (0.49 to 1.37) .33 
Lycopene (μg/dL)      
    OR (95% CI)  1.00 (Referent) 0.77 (0.49 to 1.23) 0.83 (0.52 to 1.33) 0.92 (0.53 to 1.60) .58 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.82 (0.46 to 1.46) 0.78 (0.44 to 1.41) 0.84 (0.42 to 1.68) .47 
Zeaxanthin, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.91 (0.51 to 1.62) 0.85 (0.47 to 1.54) 1.14 (0.60 to 2.14) .96 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.85 (0.40 to 1.81) 0.71 (0.34 to 1.47) 1.20 (0.54 to 2.69) .79 
α-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.91 (0.61 to 1.34) 0.65 (0.42 to 0.99) 0.37 (0.23 to 0.61) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.91 (0.55 to 1.52) 0.83 (0.50 to 1.39) 0.56 (0.31 to 1.01)  .06  
γ-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.71 (0.48 to 1.06) 0.56 (0.37 to 0.84) 0.45 (0.29 to 0.70) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.88 (0.54 to 1.46) 0.51 (0.30 to 0.86) 0.60 (0.35 to 1.02)  .02  
δ-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.73 (0.49 to 1.10) 0.81 (0.54 to 1.23) 0.51 (0.33 to 0.81) .01 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.77 (0.46 to 1.29) 0.87 (0.52 to 1.45) 0.60 (0.34 to 1.05) .13 
Selenium (μg/dL)      
    OR (95% CI)  1.00 (Referent) 0.97 (0.65 to 1.46) 0.74 (0.48 to 1.16) 0.82 (0.52 to 1.29) .24 
    Adjusted OR (95% CI)  1.00 (Referent) 0.96 (0.58 to 1.59) 0.61 (0.34 to 1.09) 0.80 (0.46 to 1.41) .27 
  OR (95% CI) by quartile level of micronutrient *
 
    
Micronutrient 1 (low) 4 (high) Ptrend 
Retinol, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.22 (0.14 to 0.32) 0.17 (0.11 to 0.27) 0.08 (0.05 to 0.15) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.37 (0.22 to 0.61) 0.30 (0.17 to 0.50) 0.13 (0.06 to 0.26)  <.001  
α-Carotene, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.82 (0.51 to 1.32) 0.75 (0.46 to 1.22) 0.50 (0.28 to 0.90) .02 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.99 (0.54 to 1.82) 0.80 (0.42 to 1.49) 0.69 (0.34 to 1.40) .27 
β-Carotene, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.86 (0.58 to 1.27) 0.58 (0.37 to 0.89) 0.57 (0.37 to 0.89) .004 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.71 (0.43 to 1.17) 0.38 (0.22 to 0.67) 0.59 (0.34 to 1.02)  .01  
β-Cryptoxanthin, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.83 (0.55 to 1.24) 0.77 (0.50 to 1.18) 0.69 (0.45 to 1.08) .10 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.92 (0.55 to 1.53) 0.87 (0.50 to 1.51) 0.85 (0.48 to 1.51) .56 
Lutein, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.77 (0.51 to 1.16) 0.66 (0.43 to 1.02) 0.92 (0.61 to 1.40) .55 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.71 (0.42 to 1.18) 0.56 (0.33 to 0.97) 0.82 (0.49 to 1.37) .33 
Lycopene (μg/dL)      
    OR (95% CI)  1.00 (Referent) 0.77 (0.49 to 1.23) 0.83 (0.52 to 1.33) 0.92 (0.53 to 1.60) .58 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.82 (0.46 to 1.46) 0.78 (0.44 to 1.41) 0.84 (0.42 to 1.68) .47 
Zeaxanthin, μg/dL      
    OR (95% CI)  1.00 (Referent) 0.91 (0.51 to 1.62) 0.85 (0.47 to 1.54) 1.14 (0.60 to 2.14) .96 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.85 (0.40 to 1.81) 0.71 (0.34 to 1.47) 1.20 (0.54 to 2.69) .79 
α-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.91 (0.61 to 1.34) 0.65 (0.42 to 0.99) 0.37 (0.23 to 0.61) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.91 (0.55 to 1.52) 0.83 (0.50 to 1.39) 0.56 (0.31 to 1.01)  .06  
γ-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.71 (0.48 to 1.06) 0.56 (0.37 to 0.84) 0.45 (0.29 to 0.70) <.001 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.88 (0.54 to 1.46) 0.51 (0.30 to 0.86) 0.60 (0.35 to 1.02)  .02  
δ-Tocopherol, μg/mL      
    OR (95% CI)  1.00 (Referent) 0.73 (0.49 to 1.10) 0.81 (0.54 to 1.23) 0.51 (0.33 to 0.81) .01 
    Adjusted OR (95% CI) § 1.00 (Referent) 0.77 (0.46 to 1.29) 0.87 (0.52 to 1.45) 0.60 (0.34 to 1.05) .13 
Selenium (μg/dL)      
    OR (95% CI)  1.00 (Referent) 0.97 (0.65 to 1.46) 0.74 (0.48 to 1.16) 0.82 (0.52 to 1.29) .24 
    Adjusted OR (95% CI)  1.00 (Referent) 0.96 (0.58 to 1.59) 0.61 (0.34 to 1.09) 0.80 (0.46 to 1.41) .27 
*

See Supplementary Table 1 (available at http://jncicancerspectrum.oxfordjournals.org/jnci/content/vol98/issue7 ) for the quartile cutpoints of serum concentrations of various micronutrients and number of cancer patients and control subjects in each category of a given micronutrient.

Ptrend values (two-sided) for micronutrient levels in quartiles were calculated by using conditional logistic regression models that also included all covariates listed above.

Odds ratios were calculated by using conditional logistic regression models that retained a matched set consisting of five to 10 control subjects who were individually matched to the index case patient by date of birth (within 2 years), date of blood draw (within 1 month), and neighborhood of residence at recruitment.

§

Adjusted for cigarette smoking (nonsmokers, ever smokers), heavy alcohol consumption (nondrinkers or <4 drinks/day, ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and seropositivity for hepatitis B surface antigen (negative, positive).

Further adjustment for serum levels of β-carotene, α-tocopherol, and γ-tocopherol did not materially change the inverse association between retinol levels and HCC risk ( Ptrend <.001).

Further adjustment for serum retinol level considerably diminished the inverse relation for HCC risk to α-tocopherol ( Ptrend = .92) and γ-tocopherol ( Ptrend = 0.16). However, the retinol-adjusted inverse association between β-carotene level and HCC risk remained statistically significant ( Ptrend = .03).

We also examined interactions of serum β-carotene and retinol levels on HCC risk ( Table 5 ). Within each tertile level of serum retinol, there was no evidence of an association between β-carotene and HCC risk. By contrast, a strong, dose-dependent inverse retinol–HCC risk association (all Ptrend <.001) was seen within each tertile level of serum β-carotene.

Table 5.

Odds ratios (ORs) and 95% confidence intervals (CIs) of hepatocellular carcinoma in relation to the joint levels of retinol and β-carotene in prediagnostic serum, Shanghai Cohort Study, 1986–2001

  β-Carotene levels by tertile, μg/dL
 
      
Retinol levels by tertile, μg/dL  Low (<8.0)
 
  Intermediate (8.0–15.6)
 
  High (>15.6)
 
  
  Ca/Co *  OR (95% CI)   Ca/Co *  OR (95% CI)   Ca/Co *  OR (95% CI)  Ptrend 
Low (<41.1) 62/131 1.00 (Referent) 57/123 0.58 (0.32 to 1.07) 40/109 0.57 (0.30 to 1.11) .20 
Intermediate (41.1–51.4) 14/113 0.40 (0.19 to 0.84) 8/133 0.11 (0.04 to 0.27) 13/120 0.38 (0.18 to 0.82) .96 
High (>51.4) 8/119 0.16 (0.06 to 0.42) 6/106 0.12 (0.04 to 0.35) 5/133 0.09 (0.03 to 0.26) .57 
Ptrend  <.001  <.001  <.001  
  β-Carotene levels by tertile, μg/dL
 
      
Retinol levels by tertile, μg/dL  Low (<8.0)
 
  Intermediate (8.0–15.6)
 
  High (>15.6)
 
  
  Ca/Co *  OR (95% CI)   Ca/Co *  OR (95% CI)   Ca/Co *  OR (95% CI)  Ptrend 
Low (<41.1) 62/131 1.00 (Referent) 57/123 0.58 (0.32 to 1.07) 40/109 0.57 (0.30 to 1.11) .20 
Intermediate (41.1–51.4) 14/113 0.40 (0.19 to 0.84) 8/133 0.11 (0.04 to 0.27) 13/120 0.38 (0.18 to 0.82) .96 
High (>51.4) 8/119 0.16 (0.06 to 0.42) 6/106 0.12 (0.04 to 0.35) 5/133 0.09 (0.03 to 0.26) .57 
Ptrend  <.001  <.001  <.001  
*

Ca/Co = number of cases patients/number of control subjects.

Odds ratios were calculated by using a conditional logistic regression model that retained a matched set consisting of five to 10 control subjects who were individually matched to the index case patient by date of birth (within 2 years), date of blood collection (within 1 month), and neighborhood of residence at recruitment. The conditional logistic regression model also included cigarette smoking (nonsmokers, ever smokers), heavy alcohol consumption (nondrinkers or <4 drinks/day, ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and seropositivity for hepatitis B surface antigen (negative, positive).

Ptrend values (two-sided) for retinol (or β-carotene) level within each tertile level of β-carotene (or retinol) were calculated by using unconditional logistic regression models that also included age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, neighborhood of residence at recruitment, and all covariates listed above.

Tertile levels of retinol and HCC risk were then stratified by HBsAg serology and duration of follow-up ( Table 6 ). A strong, inverse association between prediagnostic serum level of retinol and risk of developing HCC was present independent of HBsAg serology status and duration of follow-up.

Table 6.

Odds ratios (ORs) and 95% confidence intervals (CIs) of hepatocellular carcinoma (HCC) in relation to prediagnostic serum levels of retinol by hepatitis B surface antigen (HBsAg) serology and time interval between blood draw and diagnosis of cancer, Shanghai Cohort Study, 1986–2001

Category, tertile of retinol * No. of control subjects  All cancer patients
 
  Patients who developed HCC within 7 years of enrollment
 
  Patients who developed HCC after 7 years of enrollment
 
 
   OR (95% CI)   OR (95% CI)   OR (95% CI)  
Total subjects        
    Low 363 159 1.00 (Referent) 83 1.00 (Referent) 76 1.00 (Referent) 
    Intermediate 366 35 0.36 (0.23 to 0.56) 18 0.31 (0.17 to 0.56) 17 0.36 (0.20 to 0.65) 
    High 358 19 0.17 (0.10 to 0.30) 0.12 (0.05 to 0.29) 11 0.20 (0.10 to 0.41) 
Ptrend§   <.001  <.001  <.001 
HBsAg-negative        
    Low 312 46 1.00 (Referent) 25 1.00 (Referent) 21 1.00 (Referent) 
    Intermediate 338 26 0.53 (0.32 to 0.89) 11 0.41 (0.19 to 0.85) 15 0.68 (0.34 to 1.34) 
    High 333 10 0.21 (0.10 to 0.42) 0.15 (0.05 to 0.46) 0.26 (0.10 to 0.65) 
Ptrend§   <.001  <.001  .004 
HBsAg-positive        
    Low 51 113 1.00 (Referent) 58 1.00 (Referent) 55 1.00 (Referent) 
    Intermediate 28 0.11 (0.04 to 0.29) 0.18 (0.06 to 0.54) 0.04 (0.01 to 0.24) 
    High 25 0.12 (0.05 to 0.33) 0.08 (0.02 to 0.37) 0.13 (0.04 to 0.46) 
Ptrend§   <.001  <.001  <.001 
Category, tertile of retinol * No. of control subjects  All cancer patients
 
  Patients who developed HCC within 7 years of enrollment
 
  Patients who developed HCC after 7 years of enrollment
 
 
   OR (95% CI)   OR (95% CI)   OR (95% CI)  
Total subjects        
    Low 363 159 1.00 (Referent) 83 1.00 (Referent) 76 1.00 (Referent) 
    Intermediate 366 35 0.36 (0.23 to 0.56) 18 0.31 (0.17 to 0.56) 17 0.36 (0.20 to 0.65) 
    High 358 19 0.17 (0.10 to 0.30) 0.12 (0.05 to 0.29) 11 0.20 (0.10 to 0.41) 
Ptrend§   <.001  <.001  <.001 
HBsAg-negative        
    Low 312 46 1.00 (Referent) 25 1.00 (Referent) 21 1.00 (Referent) 
    Intermediate 338 26 0.53 (0.32 to 0.89) 11 0.41 (0.19 to 0.85) 15 0.68 (0.34 to 1.34) 
    High 333 10 0.21 (0.10 to 0.42) 0.15 (0.05 to 0.46) 0.26 (0.10 to 0.65) 
Ptrend§   <.001  <.001  .004 
HBsAg-positive        
    Low 51 113 1.00 (Referent) 58 1.00 (Referent) 55 1.00 (Referent) 
    Intermediate 28 0.11 (0.04 to 0.29) 0.18 (0.06 to 0.54) 0.04 (0.01 to 0.24) 
    High 25 0.12 (0.05 to 0.33) 0.08 (0.02 to 0.37) 0.13 (0.04 to 0.46) 
Ptrend§   <.001  <.001  <.001 
*

Tertile cut points among all control subjects in Table 5 .

The mean time interval between blood draw at recruitment and cancer diagnosis was 6.9 years for all patients, 3.8 years for patients who developed hepatocellular carcinoma (HCC) within 7 years of enrollment, and 10.2 years for patients who developed HCC after 7 years of enrollment.

Odds ratios were calculated by using unconditional logistic regression models that also included age at recruitment (years), years between blood draw and laboratory measurements of serum micronutrients, neighborhood of residence at recruitment, cigarette smoking (nonsmokers, ever smokers), heavy alcohol consumption (nondrinkers or <4 drinks/day, ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis (no, yes), and, for total subjects, HBsAg seropositivity (negative, positive).

§

Ptrend values (two-sided) for retinol level within each category by HBsAg status and duration of follow-up were calculated by using unconditional logistic regression models that also included all covariates listed above.

We observed an interaction between retinol and HBsAg on HCC risk ( Table 7 ). Among HBsAg-negative subjects, the lowest tertile level of serum retinol was associated with a fivefold increased risk of HCC (OR = 4.98, 95% CI = 2.38 to 10.4). Among subjects with the highest tertile of serum retinol, HBsAg seropositivity was associated with a 10-fold increased risk of HCC (OR = 10.6, 95% CI = 3.50 to 32.3). There was evidence of a more than multiplicative interaction between HBsAg seropositivity and serum retinol level on HCC risk. That is, HBsAg-positive subjects in the lowest tertile level of retinol had a greater than 70-fold higher risk of HCC (OR = 72.7, 95% CI = 31.6 to 167.4) than did HBsAg-negative subjects in the highest tertile of retinol ( Pinteraction = .018).

Table 7.

Odds ratios (ORs) and 95% confidence intervals (CIs) of hepatocellular carcinoma in relation to combined exposures of prediagnostic serum levels of retinol and hepatitis B surface antigen (HBsAg) serology, Shanghai Cohort Study, 1986–2001

  HBsAg-negative
 
   HBsAg-positive
 
  
Tertile of retinol * No. of patients No. of control subjects  OR (95% CI)  No. of patients No. of control subjects  OR (95% CI)  
High 10 333 1.00 (Referent) 25 10.6 (3.50 to 32.3) 
Intermediate 26 338 2.91 (1.33 to 6.35) 28 8.92 (3.13 to 25.4) 
Low 46 312 4.98 (2.38 to 10.4) 113 51  72.7 (31.6 to 167.4)  
  HBsAg-negative
 
   HBsAg-positive
 
  
Tertile of retinol * No. of patients No. of control subjects  OR (95% CI)  No. of patients No. of control subjects  OR (95% CI)  
High 10 333 1.00 (Referent) 25 10.6 (3.50 to 32.3) 
Intermediate 26 338 2.91 (1.33 to 6.35) 28 8.92 (3.13 to 25.4) 
Low 46 312 4.98 (2.38 to 10.4) 113 51  72.7 (31.6 to 167.4)  
*

Tertile cutpoints among all control subjects in Table 5 .

Odds ratios were calculated by using a conditional logistic regression model that retained a matched set consisting of five to 10 control subjects that were individually matched to the index case patient by date of birth (within 2 years), date of blood draw (within 1 month), and neighborhood of residence at recruitment. The conditional logistic regression model also included cigarette smoking (nonsmokers, ever smokers), heavy alcohol consumption (nondrinkers or <4 drinks/day, ≥4 drinks/day), self-reported history of physician-diagnosed hepatitis or liver cirrhosis at recruitment (no, yes), and HBsAg seropositivity (negative, positive).

P (two-sided) = .018 for the interaction between serum retinol level and HBsAg seropositivity on risk of developing hepatocellular carcinoma on a multiplicative scale. The P value was calculated by using the conditional logistic regression model that simultaneously included the product term and the main effect terms of serum retinol and HBsAg seropositivity, as well as all covariates listed above.

D ISCUSSION

This study demonstrated a strong inverse association between the concentration of retinol in prediagnostic serum and the risk of developing HCC in middle-aged and older Chinese men in Shanghai, China. The inverse association was present in both HBsAg-positive and -negative men. Furthermore, there was evidence of a more than multiplicative interaction between chronic HBV infection and serum retinol level on HCC risk.

Although most HCC was associated with chronic infection with HBV in our population (61% of case patients tested positive for HBsAg), HCC is not an inevitable consequence of chronic infection with HBV. Only a fraction of chronic HBV carriers eventually develop HCC ( 29 ) , suggesting that other factors contribute to HCC development among chronic carriers of HBV. Previously, we demonstrated a substantial modifying effect of dietary aflatoxin exposure on HBV-related HCC risk in this study population ( 20 , 26 ) . In a previous study ( 13 ) conducted in Guangxi, China, we showed that genetic mutations of cytokine genes played a role in determining individual susceptibility to HBV-related HCC ( 13 ) .

Vitamin A may protect against the development of HCC in humans by controlling hepatocellular differentiation and immunoresponse to viral infection, i.e., HBV. Vitamin A is derived primarily from animal-based food, whereas vegetables and fresh fruit are major sources of provitamin A carotenoids. In humans, the carotenoids possessing provitamin A activity that are present in substantial quantities in the blood include α-carotene, β-carotene, and β-cryptoxanthin. Provitamin A carotenoids can be converted to retinol in vivo. The rate of conversion depends largely on the type and amount of provitamin A carotenoids consumed and the subject's vitamin A status ( 30 ) . A recent study found a mean conversion rate of approximately 9 μg of β-carotene to 1 μg of retinol in healthy, well-nourished adults in the United States ( 31 ) . The conversion for α-carotene and β-cryptoxanthin to retinol is less efficient (approximately half that for β-carotene) ( 32 ) . Whether the measurement of serum retinol levels provides an accurate representation of hepatic levels in well-nourished populations has been debated ( 33 , 34 ) . However, it is widely accepted that there is a direct relationship between serum and hepatic vitamin A levels in vitamin A–deficient populations ( 33 ) .

Clinical studies have demonstrated that patients with chronic liver diseases have reduced levels of retinol in the blood and in the liver ( 35 , 36 ) . Patients with severe liver cirrhosis have lower serum retinol levels than those with less severe cirrhosis ( 36 ) . Our findings are consistent with these reported data; a statistically significantly lower serum level of retinol was observed among control subjects who were positive for serum HBsAg or had a self-reported history of liver disease (i.e., physician-diagnosed hepatitis or liver cirrhosis).

Because reduced serum levels of retinol could be the consequence of HCC or its precursor lesions, it is problematic to assess the etiologic role of retinol in the development of HCC in a retrospective case–control study or in a prospective cohort study with relatively short follow-up. An earlier prospective study of Chinese men showed an inverse relation between baseline serum retinol level and HCC risk. However, that study was based on only 35 patients and a relatively short follow-up (5 years). Moreover, the inverse association was observed in chronic carriers of HBV only ( 37 ) . Nonetheless, given that the liver is the main storage site of vitamin A in vivo and that chronic liver disease is associated with a reduced level of circulating retinol, a noncausal interpretation of the observed retinol–HCC association cannot be completely ruled out. Results from intervention trials would strengthen the evidence for a role of retinol in the etiology of and prevention against HCC in humans. A randomized controlled trial involving 89 HCC patients who were disease free after treatment showed that those who received oral administration of a synthetic retinoid on a daily basis for 12 months had statistically significantly lower incidence of second primary or recurrent HCCs and experienced statistically significantly longer survival than those who received placebo after a median follow-up of 62 months ( 38 , 39 ) . However, there are no intervention trials evaluating the efficacy of retinol or its derivatives on primary prevention of HCC in humans.

We also examined the associations between serum levels of α-carotene; β-carotene; β-cryptoxanthin; lutein; lycopene; zeaxanthin; and α-, γ-, and δ-tocopherols and HCC risk in this study population. The observed inverse relationships for α-carotene; β-carotene; and α-, γ-, and δ-tocopherols in univariate analyses were explained by cigarette smoking, alcohol drinking, and serum levels of retinol. No independent effect of specific tocopherols and carotenoids on HCC risk was observed, which was consistent with findings from a previous study ( 40 ) .

Dietary supplements with selenium show an inhibitory effect on aflatoxin-induced DNA damage and hepatocarcinogenesis in rats ( 19 , 41 ) but have no effect on incidence of diethylnitrosamine-induced liver cancers ( 42 ) . A study of Chinese men in Taiwan that was based on a small sample size and a relatively short follow-up period reported that patients with HCC had lower levels of serum selenium than those of control subjects ( 43 ) . This study, with a larger sample size and longer follow-up period, did not confirm these previous results.

Subjects with alcoholic hepatitis had statistically significantly lower levels of hepatic retinol than those with chronic viral hepatitis, although the serum levels of retinol were comparable between the two hepatitis groups ( 44 ) . The total liver vitamin A contents were statistically significantly lower in cirrhotic patients with alcohol abuse than those with hepatitis C infection ( 45 ) . Experimental studies have demonstrated that alcohol intake can enhance mobilization of retinol from the liver to extrahepatic tissues ( 46 ) . A statistically significant increase in circulating retinol level was observed in alcohol drinkers compared with nondrinkers in our study population. Therefore, it is important to adjust for serum retinol levels when assessing the role of alcohol on HCC risk. Here, adjustment for serum retinol levels strengthened the alcohol–HCC association. The depletion of retinol in the liver by alcohol might contribute to alcohol-induced hepatocarcinogenesis. Low retinol intake could exacerbate the effect of alcohol on HCC risk in heavy alcohol drinkers.

Our study had several strengths. The prospective study design and the availability of prediagnostic serum specimens minimized the possible influence of disease symptoms on dietary intake of various micronutrients and other lifestyle factors. The measurements of multiple serum micronutrients allowed for simultaneous evaluations of the relationships between measured micronutrients and HCC risk.

Our study also had several limitations. Although the study was based on a cohort of more than 18 000 Chinese men with 15 years of follow-up, the number of incident HCC case patients was relatively small, and hence confidence intervals for the interaction between retinol level and HBsAg on HCC risk were wide. Another limitation was that the assessment of serum micronutrients was conducted at baseline serum samples only. Changes in diet during the course of follow-up were not captured. However, dietary changes leading to either increasing or decreasing intakes of micronutrients, when they occurred equally in both case patients and control subjects, would shift the estimated odds ratios toward 1.0 (i.e., null effect). Thus the failure of this study to detect a statistically significant association between carotenoids, tocopherols, or selenium and HCC risk could be due to this limitation, especially when the true effect of these micronutrients on disease risk is moderate.

In summary, this study demonstrates that higher prediagnostic serum levels of retinol was associated with statistically significantly reduced risk of developing HCC in middle-aged or older Chinese men in Shanghai, China. The association between serum retinol levels and HCC risk was present in both chronic carriers and noncarriers of HBV. There is a statistically significant interaction between low retinol levels and HBsAg positivity on HCC risk. Given that HCC is a highly fatal malignancy, our findings may have clinical and prevention implications.

Supported by the United States National Institutes of Health (grants R01 CA43092 [R. K. Ross] and R01 CA98497 [J.-M. Yuan]).
We thank Xue-Li Wang, Yue-Lan Zhang, and Jia-Rong Cheng of the Shanghai Cancer Institute for their assistance in data collection and management, and the staff of the Shanghai Cancer Registry for their assistance in verifying cancer diagnoses in study participants. We also thank Bee-Lan Lee of National University of Singapore for the technical support in micronutrient assays. The funding agency had no role in the study design, data collection, analysis, interpretation of the results, or the preparation of the manuscript.

References

(1)
Yeh FS, Yu MC, Mo CC, Luo S, Tong MJ, Henderson BE. Hepatitis B virus, aflatoxins, and hepatocellular carcinoma in southern Guangxi, China.
Cancer Res
 
1989
;
49
:
2506
–9.
(2)
Kew MC, Yu MC, Kedda MA, Coppin A, Sarkin A, Hodkinson J. The relative roles of hepatitis B and C viruses in the etiology of hepatocellular carcinoma in southern African blacks.
Gastroenterology
 
1997
;
112
:
184
–7.
(3)
Donato F, Boffetta P, Puoti M. A meta-analysis of epidemiological studies on the combined effect of hepatitis B and C virus infections in causing hepatocellular carcinoma.
Int J Cancer
 
1998
;
75
:
347
–54.
(4)
Yu MC, Yuan JM. Environmental factors and risk for hepatocellular carcinoma.
Gastroenterology
 
2004
;
127
:
S72
–8.
(5)
Caldwell SH, Crespo DM, Kang HS, Al-Osaimi AM. Obesity and hepatocellular carcinoma.
Gastroenterology
 
2004
;
127
:
S97
–103.
(6)
Yuan JM, Govindarajan S, Arakawa K, Yu MC. Synergism of alcohol, diabetes, and viral hepatitis on the risk of hepatocellular carcinoma in blacks and whites in the U.S.
Cancer
 
2004
;
101
:
1009
–17.
(7)
Prasad KN, Hovland AR, Nahreini P, Cole WC, Hovland P, Kumar B, et al. Differentiation genes: are they primary targets for human carcinogenesis?
Exp Biol Med
 
2001
;
226
:
805
–13.
(8)
Ross AC, Gardner EM. The function of vitamin A in cellular growth and differentiation, and its roles during pregnancy and lactation.
Adv Exp Med Biol
 
1994
;
352
:
187
–200.
(9)
Chi X, Anselmi K, Watkins S, Gandhi CR. Prevention of cultured rat stellate cell transformation and endothelin-B receptor upregulation by retinoic acid.
Br J Pharmacol
 
2003
;
139
:
765
–74.
(10)
Piao YF, Shi Y, Gao PJ. Inhibitory effect of all-trans retinoic acid on human hepatocellular carcinoma cell proliferation.
World Journal of Gastroenterology
 
2003
;
9
:
2117
–20.
(11)
Yu MW, Zhang YJ, Blaner WS, Santella RM. Influence of vitamins A, C, and E and beta-carotene on aflatoxin B1 binding to DNA in woodchuck hepatocytes.
Cancer
 
1994
;
73
:
596
–604.
(12)
Berberian I, Chen LC, Robinson FR, Glauert HP, Chow CK, Robertson LW. Effect of dietary retinyl palmitate on the promotion of altered hepatic foci by 3,3′,4,4′-tetrachlorobiphenyl and 2,2′,4,4′,5,5′-hexachlorobiphenyl in rats initiated with diethylnitrosamine.
Carcinogenesis
 
1995
;
16
:
393
–8.
(13)
Nieters A, Yuan JM, Sun CL, Zhang ZQ, Stoehlmacher J, Govindarajan S, et al. Effect of cytokine genotypes on the hepatitis B virus-hepatocellular carcinoma association.
Cancer
 
2005
;
103
:
740
–8.
(14)
Brinckerhoff CE, Coffey JW, Sullivan AC. Inflammation and collagenase production in rats with adjuvant arthritis reduced with 13-cis-retinoic acid.
Science
 
1983
;
221
:
756
–8.
(15)
Moriya K, Nakagawa K, Santa T, Shintani Y, Fujie H, Miyoshi H, et al. Oxidative stress in the absence of inflammation in a mouse model for hepatitis C virus-associated hepatocarcinogenesis.
Cancer Res
 
2001
;
61
:
4365
–70.
(16)
Loguercio C, Federico A. Oxidative stress in viral and alcoholic hepatitis.
Free Radic Biol Med
 
2003
;
34
:
1
–10.
(17)
Moreno FS, T SW, Naves MM, Silveira ER, Oloris SC, da Costa MA, et al. Inhibitory effects of beta-carotene and vitamin A during the progression phase of hepatocarcinogenesis involve inhibition of cell proliferation but not alterations in DNA methylation.
Nutr Cancer
 
2002
;
44
:
80
–8.
(18)
Factor VM, Laskowska D, Jensen MR, Woitach JT, Popescu NC, Thorgeirsson SS. Vitamin E reduces chromosomal damage and inhibits hepatic tumor formation in a transgenic mouse model.
Proc Natl Acad Sci U S A
 
2000
;
97
:
2196
–201.
(19)
Lei DN, Wang LQ, Ruebner BH, Hsieh DP, Wu BF, Zhu CR, et al. Effect of selenium on aflatoxin hepatocarcinogenesis in the rat.
Biomed Environ Sci
 
1990
;
3
:
65
–80.
(20)
Ross RK, Yuan JM, Yu MC, Wogan GN, Qian GS, Tu JT, et al. Urinary aflatoxin biomarkers and risk of hepatocellular carcinoma.
Lancet
 
1992
;
339
:
943
–6.
(21)
Yuan JM, Ross RK, Wang XL, Gao YT, Henderson BE, Yu MC. Morbidity and mortality in relation to cigarette smoking in Shanghai, China. A prospective male cohort study.
JAMA
 
1996
;
275
:
1646
–50.
(22)
Adames CF. Nutritive value of American foods in common unit. Agriculture Hand Book No 456. Washington (D.C.): U.S. Department of Agriculture, U.S. Government Printing Office;
1975
: p. 31.
(23)
Lee BL, New AL, Ong CN. Simultaneous determination of tocotrienols, tocopherols, retinol, and major carotenoids in human plasma.
Clin Chem
 
2003
;
49
:
2056
–66.
(24)
Hughes K, Ong CN. Vitamins, selenium, iron, and coronary heart disease risk in Indians, Malays, and Chinese in Singapore.
J Epidemiol Community Health
 
1998
;
52
:
181
–5.
(25)
Yuan JM, Ross RK, Stanczyk FZ, Govindarajan S, Gao YT, Henderson BE, et al. A cohort study of serum testosterone and hepatocellular carcinoma in Shanghai, China.
Int J Cancer
 
1995
;
63
:
491
–3.
(26)
Qian GS, Ross RK, Yu MC, Yuan JM, Gao YT, Henderson BE, et al. A follow-up study of urinary markers of aflatoxin exposure and liver cancer risk in Shanghai, People's Republic of China.
Cancer Epidemiol Biomarkers Prev
 
1994
;
3
:
3
–10.
(27)
Winer BJ. Statistical principles in experimental design. 2nd ed. New York: McGraw-Hill;
1971
.
(28)
Breslow NE, Day NE. Statistical methods in cancer research. Volume 1. The analysis of case-control studies. Lyon (France): IARC Scientific Publication;
1980
.
(29)
Yang HI, Lu SN, Liaw YF, You SL, Sun CA, Wang LY, et al. Hepatitis B e antigen and the risk of hepatocellular carcinoma.
N Engl J Med
 
2002
;
347
:
168
–74.
(30)
Wang Z, Yin S, Zhao X, Russell RM, Tang G. Beta-carotene-vitamin A equivalence in Chinese adults assessed by an isotope dilution technique.
Br J Nutr
 
2004
;
91
:
121
–31.
(31)
Tang G, Qin J, Dolnikowski GG, Russell RM. Short-term (intestinal) and long-term (postintestinal) conversion of beta-carotene to retinol in adults as assessed by a stable-isotope reference method.
Am J Clin Nutr
 
2003
;
78
:
259
–66.
(32)
Institute of Medicine. Dietary reference intakes for vitamin A, vitamin K, arsenic, boron, chromiun, copper, iodine, iron, manganese, molybdenum, nickel, silicon, vanadium and zinc. Washington (DC): National Academy Press;
2001
.
(33)
Olson JA. Serum levels of vitamin A and carotenoids as reflectors of nutritional status.
J Natl Cancer Inst
 
1984
;
73
:
1439
–44.
(34)
Ross AC, Zolfaghari R. Regulation of hepatic retinol metabolism: perspectives from studies on vitamin A status.
J Nutr
 
2004
;
134
:
269S
–75S.
(35)
Clemente C, Elba S, Buongiorno G, Berloco P, Guerra V, Di Leo A. Serum retinol and risk of hepatocellular carcinoma in patients with child-Pugh class A cirrhosis.
Cancer Lett
 
2002
;
178
:
123
–9.
(36)
Newsome PN, Beldon I, Moussa Y, Delahooke TE, Poulopoulos G, Hayes PC, et al. Low serum retinol levels are associated with hepatocellular carcinoma in patients with chronic liver disease.
Aliment Pharmacol Ther
 
2000
;
14
:
1295
–301.
(37)
Yu MW, Hsieh HH, Pan WH, Yang CS, CJ CH. Vegetable consumption, serum retinol level, and risk of hepatocellular carcinoma.
Cancer Res
 
1995
;
55
:
1301
–5.
(38)
Muto Y, Moriwaki H, Ninomiya M, Adachi S, Saito A, Takasaki KT, et al. Prevention of second primary tumors by an acyclic retinoid, polyprenoic acid, in patients with hepatocellular carcinoma. Hepatoma Prevention Study Group.
N Engl J Med
 
1996
;
334
:
1561
–7.
(39)
Muto Y, Moriwaki H, Saito A. Prevention of second primary tumors by an acyclic retinoid in patients with hepatocellular carcinoma.
N Engl J Med
 
1999
;
340
:
1046
–7.
(40)
Yu MW, Chiu YH, Chiang YC, Chen CH, Lee TH, Santella RM, et al. Plasma carotenoids, glutathione S-transferase M1 and T1 genetic polymorphisms, and risk of hepatocellular carcinoma: independent and interactive effects.
Am J Epidemiol
 
1999
;
149
:
621
–9.
(41)
Shi CY, Chua SC, Lee HP, Ong CN. Inhibition of aflatoxin B1-DNA binding and adduct formation by selenium in rats.
Cancer Lett
 
1994
;
82
:
203
–8.
(42)
Aquino TM, Porta EA, Sablan HM, Dorado RD. Effects of selenium supplementation on hepatocarcinogenesis in rats.
Nutr Cancer
 
1985
;
7
:
25
–36.
(43)
Yu MW, Horng IS, Hsu KH, Chiang YC, Liaw YF, Chen CJ. Plasma selenium levels and risk of hepatocellular carcinoma among men with chronic hepatitis virus infection.
Am J Epidemiol
 
1999
;
150
:
367
–74.
(44)
Leo MA, Lieber CS. Hepatic vitamin A depletion in alcoholic liver injury.
N Engl J Med
 
1982
;
307
:
597
–601.
(45)
Ukleja A, Scolapio JS, McConnell JP, Spivey JR, Dickson RC, Nguyen JH, et al. Nutritional assessment of serum and hepatic vitamin A levels in patients with cirrhosis.
JPEN J Parenter Enteral Nutr
 
2002
;
26
:
184
–8.
(46)
Leo MA, Kim C, Lieber CS. Increased vitamin A in esophagus and other extrahepatic tissues after chronic ethanol consumption in the rat.
Alcohol Clin Exp Res
 
1986
;
10
:
487
–92.