Prospective study of vegetable consumption and liver cancer in Japan
We examined the relationship between vegetable consumption and the risk of death from liver cancer in a cohort study in Japan. This analysis is based on data from 6,049 subjects aged 40 to 79 years enrolled in a cohort study conducted in Fukuoka Prefecture, Japan. The follow-up period was from 1986 to 1999. All liver cancer deaths were recorded. The vegetable consumption was classified into 3 groups: “once per week or less,” “2–4 times per week” and “daily intake.” The Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence interval (95% CI). A total of 51 male and 22 female liver cancer deaths were recorded during 62,343 person-years of follow-up. The “once per week or less” group was considered the referent group. In males, the multivariate HRs of liver cancer deaths were 0.61 (95% CI: 0.33–1.14) and 0.25 (95% CI: 0.11–0.59) in the “2–4 times per week” and “daily intake” groups, respectively. In females, the multivariate HRs were 0.44 (95% CI: 0.13–1.51) and 0.51 (95% CI: 0.16–1.69), respectively. The multivariate HRs were also reported by history of hepatitis and cirrhosis. In those without a history of these conditions, the multivariate HRs were 0.54 (95% CI: 0.27–1.09) and 0.36 (95% CI: 0.16–0.83). In those with a history of these conditions, the multivariate HRs were 0.58 (95% CI: 0.22–1.56) and 0.37 (95% CI: 0.13–1.06), respectively. Our study reveals an inverse association between vegetable consumption and the risk of death from liver cancer. These results provide further evidence of the protective effect of vegetables against liver cancer. © 2006 Wiley-Liss, Inc.
Liver cancer is one of the malignancies for which the major etiologic factors have been well established. Among these, chronic infection with either hepatitis B virus (HBV) or hepatitis C virus (HCV) has been well documented as a major cause.1–3 Infection either with HBV or HCV increases the risk of this cancer more than 10-fold.4 Many other possible etiologic factors, including exposure to excessive alcohol consumption, dietary aflatoxin, diabetes mellitus, cirrhosis of any etiology and others, have been shown to play a role in the development of this cancer.1, 5
A large number of epidemiological studies have been conducted to examine the relationship of high intake of vegetables with a wide range of cancers.6, 7 In 1997, an international review panel (World Cancer Research Fund)8 reviewed the results of published studies and reported that diets high in vegetables were considered to decrease the risk of most cancer sites. With regard to liver cancer, however, we have identified only a few studies, which examined the association between vegetable consumption and liver cancer, and the results of these were inconsistent.9–12 To contribute to clarifying this relationship, we conducted this study to investigate the association between vegetable consumption and liver cancer mortality, based on a population-based prospective cohort study in Japan.
Material and methods
The present survey, named the Miyako Study, was a prospective cohort study established to evaluate the relationship between lifestyle and health. This study was conducted in a general population derived from 4 areas of Fukuoka Prefecture, a prefecture in western Japan with a high liver cancer mortality rate.13 The follow-up has been described in detail elsewhere.14, 15 Briefly, the baseline survey was conducted from 1986 to 1989. We invited all inhabitants aged from 30 to 79 years living in A town, B village, and selected districts of C city and D town (15,417 subjects in total) to participate in a self-administrated questionnaire survey.
Baseline characteristics of the study were obtained through self-questionnaires. All subjects were asked to provide information about health-related factors, including smoking, alcohol, diet, disease history and others. The response rate was 86.1%, equivalent to 13,270 subjects who constituted the cohort. Of those, 270 inhabitants aged 80 or over, living together with the study participants, completed the questionnaires and were also included in the cohort. The vital status of all study participants was checked with the collaboration of each municipal office annually until March 31, 1999 for A town and C city; until August 31, 1999 for D town; and until November 30, 1999 for B village. Subjects who moved out from the study areas were treated as censored subjects. For deceased subjects, the cause of death was determined from the death certificate, held at the regional heath center, with the permission of the Management and Co-ordination Agency of the Japanese Government. Causes of death were coded based on the International Classification of Diseases and Injuries (ICD), 9th Revision (ICD-9). In the present analysis, liver cancer was defined as primary liver cancer (155.0 in the ICD-9); malignant neoplasm of intrahepatic bile ducts (155.1 in the ICD-9); and malignant neoplasm of liver, not specified as primary or secondary (155.2 in the ICD-9).
The research protocol of this study was approved by the Ethics Committee of the Medical Care and Research at the University of Occupational and Environmental Health, Kitakyushu, Japan.
Among the 254 items of the self-administered questionnaire, subjects were asked to provide information on their dietary habits regarding food consumption, including vegetables. Vegetables included in the present analysis were limited to green–yellow vegetables such as squash, pumpkin, carrots and all kinds of green leafy vegetables. On the questionnaire, the vegetable consumption was assessed by 5 levels, including “twice or more per day,” “once a day,” “2–4 times per week,” “2–4 times per month” and “seldom or never.” For the present analysis, vegetable consumption levels were converted into 3 groups. The group “twice or more per day” and group “once a day” were combined and named “daily intake” group. The group “2–4 times per week” was kept as it was. The group “2–4 times per month” and “seldom or never” were also combined and named “once per week or less.”
We excluded subjects aged less than 40 and those aged more than 79 years (2,963 subjects). We excluded cancer cases before the study baseline (172 cases) and excluded subjects with missing data on vegetable consumption (363 subjects). Then, we excluded one study area (A town) from the analysis, because history of hepatitis and cirrhosis had not been investigated (3,723 subjects) in this town. Finally, the analysis was restricted to 6,049 people (2,694 males and 3,355 females).
The baseline characteristics of all study subjects were compared according to the intake groups of vegetables using chi-square test for categorical variables and analysis of variance for continuous variables. We counted the number of person-years of follow-up for each subject from the baseline of the study until the date of death, or the date of migration from the study areas or the end of follow-up, whichever came first. We used the Cox proportional hazards regression to estimate the hazard ratios (HRs) and their 95% confidence intervals (95% CI) of liver cancer mortality according to categories of vegetable consumption. The “once per week or less” group was considered the referent group, and analyses were performed separately by sex. A test for trend was performed to assess statistic significance across exposure groups by including ordinal terms for each intake group of vegetable and entering the variable as a continuous term in the model. Patterns Schoenfeld residuals with time were considered to identify possible violation of the proportional hazards model16 and found valid in our analysis.
In addition to sex and age, the following variables were considered as potential confounders and included in the proportional hazards model (all missing data was considered as an additional category, termed the unknown category, and included in the analyses): body mass index (computed as weight in kilograms divided by square of height in meters); smoking (never smoker, ex-smoker and current smoker, and unknown); alcohol (never drinker, ex-drinker and current drinker, and unknown); coffee (never, a cup/day or less and more than a cup/day, and unknown); history of transfusion (yes or no); history of hepatitis (yes or no); history of cirrhosis (yes or no); history of diabetes (yes or no); and study area (B village, C City, or D town). All statistical analyses were performed using the Stata version 8.2 software package.
During an average follow-up period of 10.3 years (follow-up period ranged from 0.1 to 11.8 years), there were 73 liver cancer deaths recorded, of which 57 cases were primary liver cancer (155.0 in the ICD-9); 2 cases were malignant neoplasm of intrahepatic bile ducts (155.1 in the ICD-9); and 14 cases were malignant neoplasm of liver, not specified as primary or secondary (155.2 in the ICD-9).
Characteristics of subjects at baseline by vegetable consumption are shown in Table I. The subjects with high consumption of vegetables tended to be older for both sexes, with p value less than 0.05. The proportion of subjects with a history of diabetes in the daily intake group of vegetables was higher in both sexes (p < 0.05). No difference in history of hepatitis and cirrhosis by vegetable consumption was seen in males (p > 0.05), although males with higher consumption had a more frequent history of blood transfusion (p = 0.04). In females, the proportion of subjects with a history of hepatitis was higher in the daily intake group (p = 0.03), while the proportions of history of blood transfusion and history of cirrhosis were closely similar among the 3 groups (p > 0.05).
Table I. Baseline Characteristics of Subjects according to Vegetable Consumption and Sex
|Mean age (SD)2||54.1 (9.9)||55.5 (10.0)||58.5 (10.5)||<0.01||55.4 (10.5)||56.8 (10.6)||57.0 (10.6)||0.03|
|Mean body mass index3 (SD)2||23.1 (2.85)||22.9 (2.84)||22.8 (2.80)||0.11||22.9 (3.22)||23.1 (3.10)||22.8 (3.00)||0.07|
| Never smokers||16.3||16.7||19.7||<0.01||64.8||68.3||71.4||<0.01|
| Current smokers||55.3||50.4||39.5||12.7||8.1||5.5|
| Never drinkers||11.6||10.3||11.1||<0.01||35.4||35.0||37.6||0.36|
| Current drinkers||81.5||83.5||77.8||44.1||45.8||41.4|
| ≤1 cup/day||39.2||39.2||37.8||37.7||41.0||41.8|
| >1 cup/day||37.8||36.8||33.3||35.7||33.4||30.1|
|History of transfusion (%)3||10.3||11.2||14.3||0.04||11.7||9.5||10.5||0.32|
|History of hepatitis (%)3||8.8||8.3||9.3||0.74||3.5||4.1||5.9||0.03|
|History of cirrhosis (%)3||0.6||0.9||1.0||0.41||0.3||0.3||0.6||0.38|
|History of diabetes (%)3||4.8||4.4||6.2||0.03||2.0||2.8||4.3||0.02|
Adjusted HRs and 95% CI for liver cancer mortality by vegetable consumption according to sex are shown in Table II. Results showed a consistent inverse association in males, with multivariate HR of 0.61 (95% CI: 0.33–1.14) and 0.25 (95% CI: 0.11–0.59) for the “2–4 times per week” and “daily intake” groups, respectively. p for trend was less than 0.01. In females, multivariate HRs were 0.44 (95% CI: 0.13–1.51) and 0.51 (95% CI: 0.16–1.69) for the “2–4 times per week” and “daily intake” groups, respectively. p for trend was 0.36. The results calculated for both sexes also showed an inverse association, with the multivariate HR of 0.56 (95% CI: 0.32–0.98) and 0.35 (95% CI: 0.19–0.68) for the “2–4 times per week” and “daily intake” groups, respectively. p for trend was less than 0.01. We included 14 liver cancer deaths, not specified as primary or secondary (155.2 in the ICD-9) in our analysis. However, this fact did not change the results substantially. We additionally estimated the HRs of liver cancer deaths, after excluding these 14 cases. Multivariate HRs for males were 0.60 (95% CI: 0.31–1.17) and 0.18 (95% CI: 0.07–0.49) for the “2–4 times per week” and “daily intake” groups, and multivariate HRs for females were 0.32 (95% CI: 0.09–1.15) and 0.21 (95% CI: 0.05–0.84) (data not shown).
Table II. Hazard Ratios (HR) of Liver Cancer Deaths by Vegetable Consumption, according to Sex
| No. of cases||18||24||9|| |
| No. of subjects||640||1,278||776|
| Person-years of follow-up||6,493||13,013||7,638|
| Age-adjusted HR (95%CI)||1.0||0.59 (0.32–1.09)||0.31 (0.14–0.71)||<0.01|
| Multivariate HR1 (95%CI)||1.0||0.61 (0.33–1.14)||0.25 (0.11–0.59)||<0.01|
| No. of cases||4||8||10|| |
| No. of subjects||401||1,639||1,315|
| Person-years of follow-up||4,264||17,315||13,620|
| Age-adjusted HR (95%CI)||1.0||0.42 (0.13–1.40)||0.63 (0.20–2.01)||0.83|
| Multivariate HR1 (95%CI)||1.0||0.44 (0.13–1.51)||0.51 (0.16–1.69)||0.36|
|Males and females|
| No. of cases||22||32||19|| |
| No. of subjects||1,041||2,917||2,091|
| Person-years of follow-up||10,757||30,328||21,258|
| Sex, age-adjusted HR (95%CI)||1.0||0.53 (0.31–0.92)||0.42 (0.22–0.78)||<0.01|
| Multivariate HR2 (95%CI)||1.0||0.56 (0.32–0.98)||0.35 (0.19–0.68)||<0.01|
Given the importance of a history of hepatitis and cirrhosis at baseline as predominant risk factors for liver cancer death, we also estimated HR and 95% CI for a history of either or both of these conditions. Multivariate HR of liver cancer with a history of hepatitis or cirrhosis was 6.51 (95% CI: 3.55–12.00) for males and 16.62 (95% CI: 6.55–42.14) for females (data not shown). We then stratified the data according to hepatitis or cirrhosis history (Table III). A significant and inverse association between vegetable consumption and risk was observed in subjects without a history of these conditions, with a multivariate HR of 0.54 (95% CI: 0.27–1.09) for the “2–4 times per week” group and 0.36 (95% CI: 0.16–0.83) for the “daily intake” group. In those with a history of these conditions, HRs were 0.58 (95% CI: 0.22–1.56) and 0.37 (95% CI: 0.13–1.06), respectively.
Table III. Hazard Ratios (HR) of Liver Cancer Deaths by Vegetable Consumption and History of Hepatitis, Cirrhosis or Both
|Subjects without a history of hepatitis and/or cirrhosis|
| No. of cases||14||20||10|| |
| No. of subjects||953||2,712||1,919|
| Person-years of follow-up||10,056||28,548||19,678|
| Sex- and age-adjusted HR (95%CI)||1.0||0.51 (0.26–1.03)||0.36 (0.16–0.81)||0.02|
| Multivariate HR1 (95%CI)||1.0||0.54 (0.27–1.09)||0.36 (0.16–0.83)||0.02|
|Subjects with a history of hepatitis and/or cirrhosis|
| No. of cases||8||12||9|| |
| No. of subjects||66||173||153|
| Person-years of follow-up||702||1,779||1,580|
| Sex- and age-adjusted HR (95%CI)||1.0||0.60 (0.24–1.50)||0.39 (0.15–1.03)||0.06|
| Multivariate HR1 (95%CI)||1.0||0.58 (0.22–1.56)||0.37 (0.13–1.06)||0.07|
In the present study, a significant association between higher vegetable consumption and decreased liver cancer mortality was revealed. The results of our study are consistent with those of some previous studies.9, 10, 12 A prospective cohort study conducted in Japan reported that daily green–yellow vegetable intake provided an approximately 8% reduction in risk for total cancer mortality and a 25% reduction for liver cancer mortality;9 a case–control study in Italy also showed a significant inverse association for subjects with a high frequency of vegetable consumption;10 a cohort study of Taiwanese men showed a 4.7-fold increased risk of hepatocellular carcinoma in those who had a weekly vegetable consumption frequency of less than 6 meals a week, compared with those who had a frequency of 6 or more meals. This association was observed in HBsAg carriers but not in noncarriers;12 while another case–control study in Greece found no evidence that vegetable intake reduces the risk of liver cancer.11 The discrepancies and the strength of association in the results of previous studies may be due to the difference in population and in study design (cohort study or case–control study).
The strength of this study is its prospective design conducted among a general population. Information on vegetable consumption was collected before the subsequent diagnosis of any cancer, avoiding the exposure recall bias inherent in case–control studies. In addition, the high response rate (86.1%) to the baseline study and low proportion of migration (7.4%) over more than 10 years of follow-up are further strengths.
The possible mechanism to explain this association is discussed below. Vegetables contain nutrients, including carotenoids (pro-vitamin A), retinoids (pre-form vitamin A), ascorbic acid (vitamin C) and α-tocopherol (vitamin E). The potential protective effect of vegetables is thought to be mediated through the antioxidant effects of their wide range of these nutrients, which may act together to produce this effect via joint or additive activity. In experimental studies, these micronutrients have also been shown to suppress carcinogenesis.17–19 β-Carotene showed lower hepatic bromodeoxyuridine labeling indexes in neoplastic lesions, suggesting an inhibitory action of these substances on cell proliferation.19 Acyclic retinoid has the potential to inhibit cell proliferation activity in lesions and in nontumorous areas, as observed in male rats.20 In a cohort study in Taiwan, an inverse association of hepatocellular carcinoma (HCC), according to serum retinol levels, was also reported, based on a nested case–control study inside that cohort study.12
The association between vegetable consumption and the risk of liver cancer mortality was further analyzed according to history of hepatitis and cirrhosis. Results showed that daily intake of vegetables tended to decrease liver cancer mortality in subjects both with and without a history of these conditions, suggesting that vegetables exert a protective effect against the process of hepatocarcinogenesis regardless of a history of hepatitis or cirrhosis. Low plasma levels of antioxidants, such as retinal and α-tocopherol, have been reported in people with cirrhosis, even in early stages of the disease,21 which is strongly associated with liver cancer.
Several limitations of this study should be cited. First, we did not examine viral biomarkers. Although we accounted for a history of hepatitis, some people are not aware of having hepatitis, particularly in asymptomatic cases; under-reporting of hepatitis may therefore have occurred. In addition, we could not determine the hepatitis due to HBV or HCV, although approximately over 80% of liver cancer patients in Japan have antibodies to HCV.22 Second, history of hepatitis and history of cirrhosis were not investigated in the A town, which led to the exclusion of one third of the study population. However, the HRs adjusted for several factors, including history of these conditions, are not substantially different from those adjusted for age only (Table II). Additional calculation of multivariate HRs, including 3,723 subjects living in the A town also show an inverse association. Multivariate HRs were 0.61 (95% CI: 0.36–1.03) and 0.43 (95% CI: 0.23–0.80) for males; and multivariate HRs were 0.41 (95% CI: 0.15–1.13) and 0.56 (95% CI: 0.21–1.45) for females (data not shown).
In summary, we found an inverse association between vegetable consumption and the risk of death from liver cancer in this prospective cohort study. These results provide further evidence to support the hypothesized protective effect of vegetables against liver cancer.
The authors gratefully acknowledge the public health authorities in the municipal offices and public health staff of Miyako Health Center and Munakata Health Center for their valuable co-operations.