• soya;
  • isoflavone;
  • hepatitis b virus;
  • hepatitis c virus;
  • hepatocellular carcinoma


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

To determine if the risk of hepatocellular carcinoma (HCC) is reduced by consumption of soya foods, we conducted a case-control study within a cohort of Japanese A-bomb survivors. We compared the prediagnosis consumption of isoflavone-rich miso soup and tofu to HCC risk, adjusting for hepatitis B (HBV) and C (HCV) viral infections, the major HCC risk factors in this population. The study included 176 pathologist-confirmed cases of HCC diagnosed in 1964–1988 and 560 controls who died of diseases other than liver cancer. We examined dietary information collected at least 2 years before diagnosis or death and tissue-based measures of viral hepatitis. Using logistic regression, crude ORs were 0.5 (95% CI 0.29–0.95) and 0.5 (95% CI 0.20–0.99) for high vs. low miso soup and tofu intake, respectively. Adjusting for year of birth, sex, HBV, HCV and other factors, the OR for miso soup was unchanged at 0.5 (95% CI 0.14–1.55), and miso results were similar when ORs were recalculated separately for earlier and later birth cohorts to assess consistency of results. The adjusted OR for tofu was 0.9 (95% CI 0.20–3.51). We also found a statistically significant (p < 0.0001) interaction between sex and HCV, with risk of HCC being substantially higher for women. We conclude that consumption of miso soup and other soya foods may reduce HCC risk. © 2005 Wiley-Liss, Inc.

HCC is one of world's most common cancers, usually occurring in patients with chronic liver disease. Infection with HBV or HCV is the most important cause of HCC worldwide, each virus increasing risk of this cancer more than 10-fold.1 HCC has consistently been found to be more common in men than women, suggesting that sex hormones and/or X chromosome–linked genes may be involved in hepatocarcinogenesis.2 Liver cancer and HCC have also been linked to variations in hormone levels in men and women. Investigators have found significant associations between higher testosterone levels and increased HCC risk among men in Taiwan2 and Japan.3 Liver cancer risk in women is significantly increased by use of oral contraceptives4, 5 and tends to rise with earlier menarche and to be significantly increased by delayed menopause.6

Isoflavones, the most frequently studied phytoestrogen, have structural similarity to mammalian estrogens and are present in large amounts in soybeans and soy products such as miso and tofu.7 Japan has one of the world's highest consumption levels of soya foods; in comparison to the United Kingdom, mean serum levels of the isoflavone genistein are about 15 times higher in Japanese men (mean 492.7 vs. 33.3 nmol/l) and about 18 times higher in Japanese women (mean 501.9 vs. 27.7 nmol/l).8 There is evidence that soy isoflavones prevent liver cancer both in vitro and in vivo, 9 and isoflavones have been linked to reduced risk of several hormone-related cancers in humans, including prostate,10 endometrial11 and breast cancers.12, 13 There is limited evidence that isoflavones may reduce risk of liver cancer with its hormone-related etiology as well.14

Our goal was to use dietary data collected at several time points at least 2 years before HCC diagnosis to examine the relationship of soya food consumption to the etiology of this tumor, taking into account HBV and HCV infections as well as other factors.

Material and methods

  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Selection of cases and controls

The RERF LSS cohort is a group of 120,321 atomic bomb–exposed and unexposed persons who were official residents of Hiroshima or Nagasaki in 1945 and who were alive and residing there at the time of censuses conducted in 1950–1952.15 Mortality follow-up of this cohort has been ongoing since 1950 through death certificate analysis, and incident cases of liver and other cancers have been ascertained since 1957 by linkage to population-based cancer registries in Hiroshima and Nagasaki.16 A subgroup of about 20,000 LSS cohort members, the AHS, was selected in 1958 for closer clinical follow-up, and these subjects have received biennial health examinations since that time.

Our study included HCC cases who were diagnosed in 1965–1988 during an extensive pathology review of LSS cohort members, some of whom also received clinical follow-up through the AHS. Clinical material was reviewed by a panel of 3 pathologists, who reached consensus opinions on type of liver cancer (if present), cirrhosis status and HBV status, as described in more detail elsewhere.17, 18 Controls were selected from a group of cohort members autopsied in 1965–1988 who were determined not to have liver cancer. Prior to case review, 894 controls, for whom tissue samples were available, were matched to all potential cases on sex, city of residence at time of bombings, radiation exposure, attained age and, to the extent possible given the restricted period of the autopsy program, time period of death or liver cancer diagnosis. Our study was limited to cases and controls for whom both tissue samples and questionnaire data collected at least 2 years before HCC diagnosis or death were available, a total of 176 HCC cases and 560 controls. A total of 302 cohort members were confirmed during pathology review as HCC cases diagnosed in 1958–1988;18 126 were excluded because of lack of acceptable smoking or dietary data, the first questionnaire not being administered to this cohort until 1963. About 80% of cases and all controls were autopsied. Histologic classification was made in accordance with the standards proposed by the WHO.19

Determination of HBV and HCV status

To detect HBsAg, the pathology panel reviewed nonneoplastic, formalin-fixed and paraffin-embedded liver tissue slides stained with orcein20 and anti-HBV immunohistochemical material (Dako, Carpinteria, CA; LSAB kit, Universal K681). We also used PCR to test archival tissues from nontumor areas of the liver for the presence of genes encoding HBV antigens, as described elsewhere.21 If either the S, pre-C or X regions of the HBV genome could be amplified or if either the orcein or immunohistochemical stain was positive for HBV, the subject was considered positive for HBsAg, indicating active HBV infection. Agreement between the orcein and immunohistochemical staining methods for HBV was good (κ = 0.87). Assuming that persons positive on at least one of the staining tests were in fact HBV-positive, agreement between those results and the PCR results was slightly better (κ = 0.95).

To determine HCV status, we extracted RNA from one 5 μm-thick section of paraffin-embedded liver tissue. As described in detail elsewhere,22 we amplified the 5′-untranslated region of the HCV genome using specific primer sets and identified positive samples by hybridizing with a radiolabeled oligomer probe that recognizes a sequence between the 2 primers. We classified anyone testing HBV- or HCV-positive by any tissue test as positive, to maximize the sensitivity of our tests, and subjects without positive results as negative.

Measurement of dietary and other risk factors

Table I shows the numbers of participants, years of administration and response rates for the FFQs and other surveys used. FFQ-A included questions about the number of servings of tofu and miso soup consumed per week, with responses limited to ≤1 time or 2–4 times/week and “almost every day”. FFQ-B similarly assessed the number of times per week subjects ate these foods but used slightly different response categories. The ≤1 time/week category was divided into 2 subcategories, “never” and “≤1 time/week”; the other 2 response categories were the same as in FFQ-A. When information from both FFQ-A and FFQ-B was available, we assigned the following values to the FFQ response categories: 0, never; 0.5, ≤1 time/week; 3, 2–4 times/week; and 6, almost every day. We then calculated mean consumption levels.

Alcohol-related questions with similar formats were included in both FFQs and in the 1963–1967 AHS clinical survey but, unfortunately, excluded from the 1965 LSS mail survey (Table I). Alcohol consumption per volume of each beverage was quantified as follows: sake, 15%; beer, 4.2%; shochu, 15%; whiskey, 40%; and wine 13%. We calculated summary amounts as grams per day. Mean daily amounts were calculated for subjects who completed more than one survey. Subjects who said they had stopped drinking were classified as former drinkers.

Table I. Number of Cases and Controls Completing Surveys and Questionnaires About Diet, Alcohol Intake and Smoking (Hiroshima and Nagasaki, Japan)
Questionnaires and surveys completedTime frameResponse rateControls (n = 560)Cases (n = 176)
  • 1

    Calculated for FFQ-A form administrated in the AHS 1964–1967; response rate for FFQ-A mailed to female LSS respondents 1969–1970 is unknown.

FFQ-A and FFQ-B  183.22715.3
AHS clinical survey1963–1967>90%264.610.6
LSS mail survey (1965)196592%28050.06536.9
Both AHS and LSS surveys  132.310.6

Smoking-related questions were included on both FFQs and on both the 1963–1967 AHS survey and the 1965 LSS mail survey. The 4 different questionnaires recorded if subjects were current, past or never-smokers and the number of cigarettes per day that were usually smoked. We calculated the mean number of cigarettes per day for subjects with multiple sources of smoking data; subjects who said they had stopped smoking were classified as former smokers.

Measures of liver irradiation from the atomic bombs were derived from the Dosimetry System 1986, as described elsewhere.15 This system estimates the liver dose based on physical calculations of neutron particle and γ-ray bomb yields, interviews with cohort members about their locations and shielding by buildings and terrain during the bombings and radiation shielding by body tissue. Because of its increased biologic activity, the neutron dose was multiplied by 10 and added to the γ dose.

Statistical methods

Controls were matched to potential cases before pathology review, and although some potential or confirmed cases were subsequently rejected due to non-HCC diagnoses or inadequate tissue samples, their matched controls were retained in the study. Inadequate tissue samples resulting in failed hepatitis tests also created missing data for some controls. To retain statistical power in spite of matched sets lacking either a case or both controls, we performed an unmatched analysis using unconditional logistic regression without regard to case-control sets. We adjusted for the 5 matching factors as well as for HBV and HCV, also considering cirrhosis status in some analyses. In addition, to determine if the results of matched and unmatched analyses were consistent, we conducted a matched analysis using conditional regression. To examine the joint effects of sex, HBV and HCV with other risk factors, we multiplied continuous measures of exposure with indicator variables for infection status for each virus and for sex, considering the 5 matching factors, HBV and HCV. We calculated maximum likelihood 95% CIs. SAS version 8 was used to perform all analyses (SAS Institute, Cary, NC).


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Controls were born significantly earlier than cases (p < 0.0001) because controls were limited to those dying during the period of the autopsy program, and while all cases were also diagnosed during this time frame, some died afterward. There were no significant differences between cases and controls in attained age, sex, city of residence at time of bombing or radiation exposure.

Table II lists numbers of controls and cases according to HBV and HCV infection status, smoking status and alcohol intake, with adjusted ORs and 95% CIs derived from unconditional logistic regression models. Although, overall, we found no association between smoking and HCC risk, there was a statistically significant interaction between smoking and HCV infection (p = 0.03). Among subjects infected with HCV, the OR of HCC for smoking 20 or more cigarettes per day vs. never smoking was 2.9 (95% CI 0.25–42.78) (footnote 2, Table II). Further analysis showed that this elevated risk for smoking among HCV-infected persons was found only for HCC with cirrhosis (p = 0.06); the corresponding p value of HCC without cirrhosis was 0.36 (data not shown). Alcohol consumption and having stopped drinking were associated with elevated ORs of HCC, but these differences were not statistically significant (Table II). The OR of HCC with cirrhosis for having stopped drinking was 3.3 (95% CI 0.51–23.38); that of HCC without cirrhosis was 0.5 (95% CI 0.01–21.82) (data not shown).

Table II. Adjusted OR and 95% CI of HCC for Viral Hepatitis, Smoking and Alcohol Intake
Risk factorControlsCasesOR of HCC (95% CI)1p value for interaction terms
  • 1

    Adjusted for sex, city, liver irradiation level, attained age, year of death, HBV and HCV.

  • 2

    Among HCV-infected subjects, the adjusted OR (95% CI) of HCC for smoking ≥20 cigarettes/day was 2.9 (0.25–42.78).

HBV infection      
 Yes253413.5 (5.95–32.27) 0.030.38
HCV infection      
 Yes34579.1 (4.84–18.85)0.03 <0.0001
Smoking status      
 Stopped64320.9 (0.30–2.40)   
 1–19 cigarettes/day198601.0 (0.43–2.31)   
 ≥20 cigarettes/day123391.1 (0.40–2.93)   
 Each cigarette/day  1.00 (0.96–1.04)0.300.0320.36
Alcohol intake      
 Stopped20171.7 (0.25–11.61)   
 1–29 g/day48290.9 (0.20–4.02)   
 ≥30 g/day61242.6 (0.53–13.58)   
 Each g/day  1.00 (0.98–1.02)

As shown in Table II, there was a statistically significant interaction (p < 0.0001) between sex and HCV in the etiology of HCC, with women being at greater risk than men. HCV infections were found among 78% of female cases compared to 37% of male cases and 7% and 11% of female and male controls, respectively. HBV infections were found among about 20% of both male and female cases, and ORs of HCC for HBV were similar for both sexes. Inclusion of alcohol and smoking information in logistic regression models did not affect OR estimates, and results for viral hepatitis, alcohol intake and smoking were similar when a matched analysis was performed.

Because both dietary surveys were available for 12% of participants who answered questions about soya food consumption, we were able to examine the consistency of reported miso soup and tofu intake over time. Forty-four percent of subjects reported the same frequency of miso soup intake on both FFQs; 50% reported the same frequency of tofu intake. The proportions of subjects reporting extremely different consumptions in the 2 surveys (i.e., ≤1 time/week changed to ≥5 times/week or vice versa) were 24% for miso soup and 10% for tofu. In an earlier validation study for FFQ-B, investigators compared the responses of 3,005 AHS cohort members who completed both FFQ-B in 1978–1981 and a 24 hr dietary survey in 1984–1985.23 Their estimates of consumption levels of miso soup and tofu for each of the FFQ-A and -B response categories for these foods are included in Table III. Based on this previous study, it is estimated that subjects in our lowest consumption categories (≤1 time/week) for miso soup and tofu consumed about one-third to one-half as much of these foods as subjects reporting the highest consumption levels (≥5 times/week) (Table III).23

Table III. OR and 95% CI of HCC for Soya Food Consumption
Soya food intakeMean intake (g/day)1NumberCrude OR (95% CI)2Adjusted OR (95% CI)3p value for interaction terms
  • 1

    Based on data presented in published validation study23 for FFQ-B which provided separate estimates for never and ≤1/week consumers.

  • 2

    A total of 156 controls and 63 or 64 cases were included in the adjusted tofu and miso analyses, respectively. Restricting the crude analysis to these subjects, the ORs (95% CIs) were 0.6 (0.26–1.25) and 0.4 (0.20–0.81) for the 2 categories of miso intake and 0.8 (0.4–1.48) and 0.6 (0.21–1.34) for the 2 categories of tofu intake. Crude ORs per serving were 0.86 (0.76–0.98) for miso and 0.93 (0.79–1.09) for tofu.

  • 3

    Adjusted for sex, city, liver irradiation level, attained age, year of death, HBV and HCV. Dividing subjects into those born before and after March 1,1907, the crude ORs (95% CI) per miso soup serving were 0.89 (0.68–1.02) for the first birth cohort and 0.88 (0.69–1.13) for the second and adjusted ORs (95% CIs) were 0.80 (0.49–1.23) and 0.88 (0.69–1.13), respectively.

Miso soup        
 Never or ≤1/week8.05 or 10.9952291.01.0   
 2–4/week12.6559361.1 (0.59–2.03)0.6 (0.16–2.27)   
≥5/week17.14126370.5 (0.29–0.95)0.5 (0.14–1.55)   
 Per serving3   0.89 (0.80–0.98)0.86 (0.69–1.07)0.300.430.34
 Never or ≤1/week25.87 or 52.0875371.01.0   
 2–4/week60.14118550.9 (0.57-1.58)0.9 (0.31-2.66)   
≥5/week76.2044100.5 (0.20-0.99)0.9 (0.20-3.51)   
Per serving3   0.92 (0.81-1.05)1.00 (0.77-1.29)0.600.370.69

As shown in Table III, the crude analysis showed statistically significant (approx. 50%) lower HCC risk for the highest vs. lowest miso soup and tofu consumptions. Mean servings of miso soup per week for controls and cases were 4.1 (SD = 2.2) and 3.4 (SD = 2.2), respectively (p = 0.02) and those for tofu, 2.8 (SD = 1.9) and 2.5 (SD = 1.6), respectively (p = 0.26). Comparing subjects eating miso soup ≥5 times/week to those eating it ≤1 time/week, the crude and adjusted ORs were the same, though the latter 95% CI was wider and included the null value. ORs calculated per increased weekly serving of miso soup were consistent in the crude (OR = 0.89, 95% CI 0.80–0.98) and adjusted (OR = 0.86, 95% CI 0.69–1.07) analyses and when crude and adjusted ORs were calculated after dividing subjects into 2 equal-sized groups according to year of birth (footnote 2, Table III). We did not detect a statistically significant interaction of soya food consumption with either sex or viral hepatitis, and adjustment of tofu ORs for miso soup and vice versa and for alcohol intake and smoking did not substantially affect OR estimates.

Although all subjects with miso soup or tofu data were included in the crude analysis shown in Table III, the adjusted analysis shown here was limited to the approximately two-thirds of these subjects for whom information about viral hepatitis status and radiation exposure was also available. Recalculating crude ORs and 95% CIs for high vs. low soya food intake by limiting subjects to those included in the adjusted analysis disclosed crude estimates that were similar to those calculated using the full data set; crude ORs and CIs for high vs. low miso soup consumption dropped from 0.5 (95% CI 0.29–0.95) to 0.4 (95% CI 0.20–0.81), and the corresponding crude estimates for tofu were 0.5 (95% CI 0.20–0.99) and 0.6 (95% CI 0.21–1.34) (footnote 2, Table III).

Results of matched analysis limited to cases and controls included in matched sets disclosed unadjusted ORs and 95% CIs of 0.87 (95% CI 0.38–1.95) for weekly miso soup intake of 2 or more servings vs. less and 0.75 (95% CI 0.36–1.54) for weekly tofu intake of 2 or more servings vs. less (data not shown). Although inclusion of attained age and year of death in all adjusted analyses limited the effect of birth cohort differences on our results, we conducted an additional analysis of soya food risk estimates, recalculating estimates after dividing subjects into 2 groups by median date of birth. Dividing subjects by birth year and recalculating risk estimates resulted in similar ORs and CIs for soya food consumption for both the pre- and post-1907 birth cohorts (footnote 3, Table III).


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

Our results suggest that consumption of soya foods is associated with reduced risk of HCC. Crude analysis showed that eating miso soup or tofu ≥5 times/week, as opposed to ≤1 time/week, was associated with about 50% lower HCC risk, a relationship that was statistically significant. The reduction in crude HCC risks was slightly greater for miso soup than for tofu, the reduction per additional serving being 0.89 (95% CI 0.80–0.98) for miso soup and 0.92 (95% CI 0.81–1.05) for tofu. Similar crude and adjusted ORs were found for high vs. low miso soup intake and per additional serving of miso soup, though the 95% CIs for these adjusted ORs included the null value. Because ORs for miso soup did not increase following adjustment, the adjusting variables do not appear to be confounders and their inclusion in the logistic regression models may not be necessary to obtain unbiased estimates. Thus, comparison of adjusted and unadjusted ORs for high vs. low miso soup consumption support the idea that the unadjusted estimates, which exclude the null value, may be valid estimates of the relationship between this risk factor and HCC. Protective effects of both soya foods were found when the study group was divided according to year of birth, with risk estimates for both foods being similar in both groups, suggesting that differences in birth year for cases and controls did not affect results. The results of matched and unmatched analyses were also consistent.

The major limitation of our study is measurement of isoflavone exposure with FFQs. Our comparison of the miso soup and tofu answers of persons answering FFQs both in 1964–1970 and in 1978–1981 shows that about half of our subjects had similar answers to these questions in the 2 time periods, while a small proportion had completely different answers. This misclassification of soya food consumption should be random for cases and controls because FFQs were completed at least 2 years before HCC diagnosis or death; thus, while soya food risk estimates may be attenuated, they should not be overestimated. Misclassification was reduced for subjects completing both FFQs since we averaged their responses; unfortunately, subjects with such data constituted a small portion of the full study group. It is likely that subjects more accurately recalled their consumption of miso soup than tofu because miso soup is eaten as a discrete food item, while tofu is mixed into many foods prepared in Japan. Thus, FFQ respondents may have had more difficulty quantifying amounts of tofu eaten, thereby increasing misclassification of tofu consumption levels.

A second limitation is the wide popularity of soya foods in Japan, which meant that the difference between our lowest and highest soya food consumers was relatively small. The inclusion in our reference groups for miso soup and tofu of persons consuming from one-third to one-half as much of these foods as our highest-level consumers would be expected to attenuate ORs and widen 95% CIs, suggesting that our study may have underestimated the relationship between soya foods and HCC. The FFQs did not include measurement of natto, another form of fermented soybeans that is often eaten with miso soup and that is also a major source of phytoestrogen in the Japanese diet, though less frequently eaten in Hiroshima and Nagasaki. It is not known how exclusion of natto from the FFQs would affect the ORs for soya foods reported here, but it did limit our ability to estimate total phytoestrogen intake.

We found a statistically significant interaction between sex and HCV, with the HCV risk of HCC being significantly higher for women. Although HCV infection rates among controls in our study were similar for both sexes, 78% of female cases were HCV-infected compared to 37% of male cases, thereby increasing ORs of HCC for HCV among females. HBV prevalence rates were about 20% among both male and female cases, and ORs of HCC for HBV were similar for both sexes. Thus, HCC among women was primarily attributable to HCV. It is not clear if the increased risk for HCV among women is completely a function of their lower background risk for HCC because of reduced exposures to alcohol and other risk factors or if it is also partly a function of sex hormone–related differences that interact with HCV.

We did not find an increased risk of HCC associated with smoking regardless of whether HCC was accompanied by cirrhosis. We did find a statistically significant, greater than multiplicative interaction between HCV and smoking ≥20 cigarettes/day (p = 0.03), with the relationship being primarily confined to HCC with cirrhosis. HCV-infected subjects who smoked ≥20 cigarettes/day were at a 2.9-fold increased HCC risk (95% CI 0.25–42.78) compared to HCV-infected subjects who did not smoke. An earlier study of the entire A-bomb survivor cohort documented a 2.2-fold increase in risk of primary liver cancer for current compared to never-smokers,24 and another study in Japan showed that among HCV-infected patients HCC risk was about 2.5-fold higher for heavy smokers vs. never-smokers (p = 0.03).25

A large number of clinical and laboratory-based studies have pointed to a role for hormones in the development of liver cancer. Epidemiologic studies have consistently shown the incidence of HCC to be higher in males than females,26 a finding that is recapitulated in rodents, where males are more susceptible to both viral and chemical-induced HCC.27 The hypothesis that androgen may play a role in the development of HCC is supported by the increased expression and activation of androgen receptors in HCC tumor cells and by the ability of antiandrogenic compounds and castration to inhibit tumor development in laboratory animals.28, 29 However, use of antiandrogenic compounds in clinical trials for HCC treatment has been disappointing,30, 31 arguing either that androgens do not play a role in HCC or that they act only early in the tumorigenic process. Chronic liver diseases cause an alteration in the sex hormone balance, which is more pronounced in men.28 Studies have shown that HCC is predominantly found among male cirrhotics, who have a hormone imbalance that results in a hyperestrogenic state,32, 33 suggesting that feminization of males may play a role in the disease etiology. Estrogen is a well-known carcinogen in estrogen-responsive tissues, such as the breast and uterus; and it plays an important role in hepatocyte proliferation and as a tumor promoter in liver tissue.28 Although the original clinical trials using the antiestrogenic drug tamoxifen as a treatment for HCC were promising,28 these results did not hold up in larger studies.34

Evidence that sex hormones play an important role in the etiology of HCC also comes from epidemiologic studies. Yu and Chen35 showed that higher prediagnosis levels of serum testosterone were significantly associated with increased risk of HCC in Taiwanese men. A later study of HBV-positive men in Taiwan showed that higher testosterone levels combined with a reduced ability to block androgens was associated with a 4-fold increase in HCC risks.36 A study of male cirrhotics in Japan found that higher testosterone levels were associated with significantly higher HCC risk after adjustment for viral hepatitis, age, serum albumin and other clinical biologic factors.3 That study also found in univariate analysis that a higher ratio of testosterone to estradiol was a significant predictor of HCC, though those results were not statistically significant in multivariate analysis.3 The significantly higher occurrence of variant estrogen receptors in early-stage chronic liver disease, especially among HBV-infected men, suggests that alteration of estrogen receptors, favoring uncontrolled proliferation might play a role in neoplastic transformation of the liver.37

Epidemiologic studies have found that increased exposure to estrogen is significantly associated with elevated HCC risk among women and that isoflavones may reduce estrogen levels. Significantly increased risks of primary liver cancer and of HCC, specifically, were found in case-control studies of ever-users and long-term users of oral contraceptives in the United States38 and Italy.4 A case-control study in Greece found that menarche occurred a mean 1.4 years earlier in HCC cases than controls (p = 0.17) and that menopause occurred a mean 2.0 years later (p = 0.05).6 Results of a US clinical trial, in which 68 women were randomized to receive either soya supplements or a placebo for 12 weeks, showed that increased soya intake was associated with significantly longer menstrual cycles (p = 0.04);39 soya supplementation also increased follicular phase length and delayed menstruation in a UK study.40 Because estrogen is known to increase cell proliferation,41 such changes in menstruation patterns reducing a woman's number of lifetime menstrual periods would reduce her exposure to the peak estradiol levels associated with menstruation, possibly reducing cell proliferation.

Soya isoflavones, including genistein and daidzein, have high affinity for the active site of estrogen receptors but do not appear to have estrogenic effects, arguing that they could act as estrogen antagonists.42 However, due to the fact that estrogen antagonists have no effect in treating HCC, it is possible either that isoflavones have independent antitumor effects or that the antagonistic effects are greater during tumor development. Microarray studies comparing isoflavone-responsive genes to those induced by estrogen offer support to the first possibility. A number of epidemiologic studies, cell culture studies, animal studies and clinical trials have been conducted to assess the effects of these agents in preventing estrogen-sensitive cancers. Studies with cultured human hepatoma cell lines demonstrated that dietary isoflavones inhibited cell proliferation and induced apoptosis and that daily injection of isoflavones into hepatoma xenographs markedly reduced tumor size.9 Another group used the nitrosamine precursor DBA and nitrite to induce bladder and liver cancers in Swiss albino mice and found that addition of soybeans to their water or feed reduced dysplastic features and completely prevented the occurrence of both bladder and liver tumors.43 Laboratory-based studies suggest that isoflavones may inhibit tumorigenesis either by inhibiting cell cycle progression44, 45 or by inducing apoptosis.9, 45 In addition, in vitro studies have shown isoflavones to inhibit angiogenesis45 and proliferation of hepatic stellate cells, a key player in liver fibrosis.46 An Italian epidemiologic study that compared 92 cirrhotics to 32 HCC cases found that HCC cases at diagnosis had significantly lower (p = 0.001) intakes of genistein.14

Thus, our findings of lower HCC risk associated with higher soya intake are consistent with the results of epidemiologic, animal and laboratory-based studies as well as some clinical trials. This may reflect a counteracting effect of isoflavones on estrogen and testosterone levels that reduces HCC risk, perhaps by modifying the hormonal milieu and reducing the cell proliferation associated with increased cancer risk, and/or it may reflect an independent antitumor effect related to inhibition of angiogenesis or induction of apoptosis.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References

We thank the pathologists who helped conduct the liver cancer case review: Drs. H. Itakura and M. Yamamoto. We appreciate the help of Drs. K. Koyama, M. Soda and M. Tokunaga in identifying liver cancer cases and thank Drs. D. Pierce, J. Cologne, K. Nakachi, S. Izumi and L. Hendershot for helpful discussions. Mr. E. Grant, Ms. M. Hayashi and Ms. N. Sakata helped with data analysis. The Radiation Effects Research Foundation (Hiroshima and Nagasaki, Japan) is a private, nonprofit foundation funded by the Japanese Ministry of Health, Labor and Welfare and the U.S. Department of Energy, the latter through the National Academy of Sciences.


  1. Top of page
  2. Abstract
  3. Material and methods
  4. Results
  5. Discussion
  6. Acknowledgements
  7. References
  • 1
    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: 34754.
  • 2
    Yu MW, Cheng SW, Lin MW, Yang SY, Liaw YF, Chang HC, Hsiao TJ, Lin SM, Lee SD, Chen PJ, Liu CJ, Chen CJ. Androgen-receptor gene CAG repeats,plasma testosterone levels, and risk of hepatitis B-related hepatocellular carcinoma. J Natl Cancer Inst 2000; 92: 20238.
  • 3
    Tanaka K, Sakai H, Hashizume M, Hirohata T. Serum testosterone: estradiol ratio and the development of hepatocellular carcinoma among male cirrhotic patients. Cancer Res 2000; 60: 510610.
  • 4
    Tavani A, Negri E, Parazzini F, Franceschi S, La Vecchia C. Female hormone utilisation and risk of hepatocellular carcinoma. Br J Cancer 1993; 67: 6357.
  • 5
    Trichopoulos D. Etiology of primary liver cancer and the role of steroidal hormones. Cancer Causes Control 1992; 3: 35.
  • 6
    Mucci LA, Kuper HE, Tamimi R, Lagiou P, Spanos E, Trichopoulos D. Age at menarche and age at menopause in relation to hepatocellular carcinoma in women. BJOG 2001; 108: 2914.
  • 7
    PeetersPH, Keinan-BokerL, van der SchouwYT, GrobbeeDE, eds. Phytoestrogens and breast cancer risk: review of the epidemiological evidence. IARC Sci Publ 156. Lyon: IARC, 2002; 3316.
  • 8
    Morton MS, Arisaka O, Miyake N, Morgan LD, Evans BA. Phytoestrogen concentrations in serum from Japanese men and women over forty years of age. J Nutr 2002; 132: 316871.
  • 9
    Su SJ, Chow NH, Kung ML, Hung TC, Chang KL. Effects of soy isoflavones on apoptosis induction and G2–M arrest in human hepatoma cells involvement of caspase-3 activation, Bcl-2 and Bcl-XL downregulation, and Cdc2 kinase activity. Nutr Cancer 2003; 45: 11323.
  • 10
    Sonoda T, Nagata Y, Mori M, Miyanaga N, Takashima N, Okumura K, Goto K, Naito S, Fujimoto K, Hirao Y, Takahashi A, Tsukamoto T, et al. A case-control study of diet and prostate cancer in Japan: possible protective effect of traditional Japanese diet. Cancer Sci 2004; 95: 23842.
  • 11
    Xu WH, Zheng W, Xiang YB, Ruan ZX, Cheng JR, Dai Q, Gao YT, Shu XO. Soya food intake and risk of endometrial cancer among Chinese women in Shanghai: population based case-control study. BMJ 2004; 328: 12859.
  • 12
    Adlercreutz H. Phytoestrogens and breast cancer. J Steroid Biochem Mol Biol 2002; 83: 11318.
  • 13
    Lee HP, Gourley L, Duffy SW, Esteve J, Lee J, Day NE. Dietary effects on breast-cancer risk in Singapore. Lancet 1991; 337: 1197200.
  • 14
    Lei B, Roncaglia V, Vigano R, Cremonini C, De Maria N, Del Buono MG, Manenti F, Villa E. Phytoestrogens and liver disease. Mol Cell Endocrinol 2002; 193: 814.
  • 15
    Preston DL, Shimizu Y, Pierce DA, Suyama A, Mabuchi K. Studies of mortality of atomic bomb survivors. Report 13. Solid cancer and noncancer disease mortality: 1950–1997. Radiat Res 2003; 160: 381407.
  • 16
    Mabuchi K, Soda M, Ron E, Tokunaga M, Ochikubo S, Sugimoto S, Ikeda T, Terasaki M, Preston DL, Thompson DE. Cancer incidence in atomic bomb survivors. Part I: Use of the tumor registries in Hiroshima and Nagasaki for incidence studies. Radiat Res 1994; 137: S116.
  • 17
    Sharp GB, Cologne JB, Fukuhara T, Itakura H, Yamamoto M, Tokuoka S. Temporal changes in liver cancer incidence rates in Japan: accounting for death certificate inaccuracies and improving diagnostic techniques. Int J Cancer 2001; 93: 7518.
  • 18
    Fukuhara T, Sharp GB, Mizuno T, Itakura H, Yamamoto M, Tokunaga M, Tokuoka S, Cologne JB, Fujita Y, Soda M, Mabuchi K. Liver cancer in atomic-bomb survivors: histological characteristics and relationships to radiation and hepatitis B and C viruses. J Radiat Res (Tokyo) 2001; 42: 11730.
  • 19
    Gibson JB, Sobin LH. Histological typing of tumours of the liver, biliary tract and pancreas. International histological classification of tumours. Geneva: WHO, 1978.
  • 20
    Shikata T, Uzawa T, Yoshiwara N, Akatsuka T, Yamazaki S. Staining methods of Australia antigen in paraffin section—detection of cytoplasmic inclusion bodies. Jpn J Exp Med 1974; 44: 2536.
  • 21
    Iwamoto KS, Mizuno T, Ito T, Akiyama M, Takeichi N, Mabuchi K, Seyama T. Feasibility of using decades-old archival tissues in molecular oncology/epidemiology. Am J Pathol 1996; 149: 399406.
  • 22
    Mizuno T, Nagamura H, Iwamoto KS, Ito T, Fukuhara T, Tokunaga M, Tokuoka S, Mabuchi K, Seyama T. RNA from decades-old archival tissue blocks for retrospective studies. Diagn Mol Pathol 1998; 7: 2028.
  • 23
    Sauvaget C, Allen N, Hayashi M, Spencer E, Nagano J. Validation of a food frequency questionnaire in the Hiroshima/Nagasaki Life Span Study. J Epidemiol 2002; 12: 394401.
  • 24
    Goodman MT, Moriwaki H, Vaeth M, Akiba S, Hayabuchi H, Mabuchi K. Prospective cohort study of risk factors for primary liver cancer in Hiroshima and Nagasaki, Japan. Epidemiology 1995; 6: 3641.
  • 25
    Chiba T, Matsuzaki Y, Abei M, Shoda J, Tanaka N, Osuga T, Aikawa T. The role of previous hepatitis B virus infection and heavy smoking in hepatitis C virus–related hepatocellular carcinoma. Am J Gastroenterol 1996; 91: 1195203.
  • 26
    El Serag HB, Mason AC. Rising incidence of hepatocellular carcinoma in the United States. N Engl J Med 1999; 340: 74550.
  • 27
    Moriya K, Fujie H, Shintani Y, Yotsuyanagi H, Tsutsumi T, Ishibashi K, Matsuura Y, Kimura S, Miyamura T, Koike K. The core protein of hepatitis C virus induces hepatocellular carcinoma in transgenic mice. Nat Med 1998; 4: 10657.
  • 28
    De Maria N, Manno M, Villa E. Sex hormones and liver cancer. Mol Cell Endocrinol 2002; 193: 5963.
  • 29
    Vesselinovitch SD, Mihailovich N. The effect of gonadectomy on the development of hepatomas induced by urethane. Cancer Res 1967; 27: 178891.
  • 30
    Chao Y, Chan WK, Huang YS, Teng HC, Wang SS, Lui WY, Whang-Peng J, Lee SD. Phase II study of flutamide in the treatment of hepatocellular carcinoma. Cancer 1996; 77: 6359.
  • 31
    Grimaldi C, Bleiberg H, Gay F, Messner M, Rougier P, Kok TC, Cirera L, Cervantes A, De Greve J, Paillot B, Buset M, Nitti D, et al. Evaluation of antiandrogen therapy in unresectable hepatocellular carcinoma: results of a European Organization for Research and Treatment of Cancer multicentric double-blind trial. J Clin Oncol 1998; 16: 411417.
  • 32
    Farinati F, De Maria N, Marafin C, Fagiuoli S, Della LG, Naccarato R. Hepatocellular carcinoma in alcoholic cirrhosis: is sex hormone imbalance a pathogenetic factor? Eur J Gastroenterol Hepatol 1995; 7: 14550.
  • 33
    Nagasue N, Ogawa Y, Yukaya H, Ohta N, Ito A. Serum levels of estrogens and testosterone in cirrhotic men with and without hepatocellular carcinoma. Gastroenterology 1985; 88: 76872.
  • 34
    CLIP Group. Tamoxifen in treatment of hepatocellular carcinoma: a randomised controlled trial. CLIP Group (Cancer of the Liver Italian Programme). Lancet 1998; 352: 1720.
  • 35
    Yu MW, Chen CJ. Elevated serum testosterone levels and risk of hepatocellular carcinoma. Cancer Res 1993; 15: 7904.
  • 36
    Yu MW, Yang YC, Yang SY, Cheng SW, Liaw YF, Lin SM, Chen CJ. Hormonal markers and hepatitis B virus–related hepatocellular carcinoma risk: a nested case-control study among men. J Natl Cancer Inst 2001; 93: 164451.
  • 37
    Villa E, Dugani A, Moles A, Camellini L, Grottola A, Buttafoco P, Merighi A, Ferretti I, Esposito P, Miglioli L, Bagni A, Troisi R, et al. Variant liver estrogen receptor transcripts already occur at an early stage of chronic liver disease. Hepatology 1998; 27: 9838.
  • 38
    Hsing AW, Hoover RN, McLaughlin JK, Co-Chien HT, Wacholder S, Blot WJ, Fraumeni JFJr. Oral contraceptives and primary liver cancer among young women. Cancer Causes Control 1992; 3: 438.
  • 39
    Kumar NB, Cantor A, Allen K, Riccardi D, Cox CE. The specific role of isoflavones on estrogen metabolism in premenopausal women. Cancer 2002; 94: 116674.
  • 40
    Cassidy A, Bingham S, Setchell KD. Biological effects of a diet of soy protein rich in isoflavones on the menstrual cycle of premenopausal women. Am J Clin Nutr 1994; 60: 33340.
  • 41
    Yager JD. Endogenous estrogens as carcinogens through metabolic activation. J Natl Cancer Inst Monogr 2000; 27: 6773.
  • 42
    Pike AC, Brzozowski AM, Hubbard RE, Bonn T, Thorsell AG, Engstrom O, Ljunggren J, Gustafsson JA, Carlquist M. Structure of the ligand-binding domain of oestrogen receptor beta in the presence of a partial agonist and a full antagonist. EMBO J 1999; 18: 460818.
  • 43
    Mokhtar NM, el Aaser AA, el Bolkainy MN, Ibrahim HA, el Din NB, Moharram NZ. Effect of soybean feeding on experimental carcinogenesis—III. Carcinogenicity of nitrite and dibutylamine in mice: a histopathological study. Eur J Cancer Clin Oncol 1988; 24: 40311.
  • 44
    Park JH, Oh EJ, Choi YH, Kang CD, Kang HS, Kim DK, Kang KI, Yoo MA. Synergistic effects of dexamethasone and genistein on the expression of Cdk inhibitor p21WAF1/CIP1 in human hepatocellular and colorectal carcinoma cells. Int J Oncol 2001; 18: 9971002.
  • 45
    Fotsis T, Pepper M, Adlercreutz H, Hase T, Montesano R, Schweigerer L. Genistein, a dietary ingested isoflavonoid, inhibits cell proliferation and in vitro angiogenesis. J Nutr 1995; 125: 790S7S.
  • 46
    Kang LP, Qi LH, Zhang JP, Shi N, Zhang M, Wu TM, Chen J. Effect of genistein and quercetin on proliferation, collagen synthesis, and type I procollagen mRNA levels of rat hepatic stellate cells. Acta Pharmacol Sin 2001; 22: 7936.