Coffee consumption and the risk of primary liver cancer: Pooled analysis of two prospective studies in Japan
Article first published online: 8 MAR 2005
Copyright © 2005 Wiley-Liss, Inc.
International Journal of Cancer
Volume 116, Issue 1, pages 150–154, 10 August 2005
How to Cite
Shimazu, T., Tsubono, Y., Kuriyama, S., Ohmori, K., Koizumi, Y., Nishino, Y., Shibuya, D. and Tsuji, I. (2005), Coffee consumption and the risk of primary liver cancer: Pooled analysis of two prospective studies in Japan. Int. J. Cancer, 116: 150–154. doi: 10.1002/ijc.20989
- Issue published online: 12 MAY 2005
- Article first published online: 8 MAR 2005
- Manuscript Accepted: 3 JAN 2005
- Manuscript Received: 24 SEP 2004
- Japanese Ministry of Health and Welfare
- Japanese Ministry of Education, Science, and Culture
- liver neoplasms;
- prospective studies;
Although case-control studies suggested that coffee consumption is associated with a decreased risk of liver cancer, no prospective cohort study has been carried out. To examine the association between coffee consumption and the risk of liver cancer, we conducted a pooled analysis of data available from 2 cohort studies in Japan. A self-administered questionnaire about the frequency of coffee consumption and other health habits was distributed to 22,404 subjects (10,588 men and 11,816 women) in Cohort 1 and 38,703 subjects (18,869 men and 19,834 women) in Cohort 2, aged 40 years or more, with no previous history of cancer. We identified 70 and 47 cases of liver cancer among the subjects in Cohort 1 (9 years of follow-up with 170,640 person-years) and Cohort 2 (7 years of follow-up with 284,948 person-years), respectively. We used Cox proportional hazards regression analysis to estimate the relative risk (RR) and 95% confidence interval (CI) of liver cancer incidence. After adjustment for potential confounders, the pooled RR (95% CI) of drinking coffee never, occasionally and 1 or more cups/day were 1.00 (Reference), 0.71 (0.46–1.09) and 0.58 (0.36–0.96), respectively (p for trend = 0.024). In the subgroup of subjects with a history of liver disease, we found a significant inverse association between coffee consumption and the risk of liver cancer. Our findings support the hypothesis that coffee consumption decreases the risk of liver cancer. Further studies to investigate the role of coffee in prevention of liver cancer among the high-risk population are needed. © 2005 Wiley-Liss, Inc.
Primary liver cancer is the third most common cause of death from cancer worldwide.1 The incidence of liver cancer is highest in Eastern Asia, including Japan.2 Although its incidence is lower in Europe3, 4 and the United States,5 it has been increasing over the last few decades.
Hepatitis B virus (HBV) and hepatitis C virus (HCV) infections are established causes of liver cancer,6 and 59.6% and 23.6% of liver cancers worldwide are considered attributable to HBV and HCV, respectively.7 Epidemiologic studies have indicated that alcohol drinking8, 9 and tobacco smoking10 are also associated with an increased risk of liver cancer.
There are several sets of data supporting the possibility that coffee consumption has a preventive effect against liver cancer. Animal experiments have indicated that coffee has inhibitory effects against chemical carcinogenesis in liver tissue.11 Furthermore, epidemiologic studies have demonstrated that coffee consumption is inversely related to serum liver enzyme activity,12, 13, 14, 15 and has an inverse association with the incidence of liver cirrhosis.16, 17 Recent case-control studies in Italy and Greece have suggested that coffee consumption is associated with a decreased risk of liver cancer.18, 19, 20
All the existing epidemiologic evidence related to coffee consumption and liver cancer has been derived only from case-control studies.18, 19, 20 To further clarify the association between coffee consumption and the risk of liver cancer, a prospective cohort study is essential. Our present study was conducted to examine the association between coffee consumption and the risk of primary liver cancer based on population-based prospective cohort studies in Japan.
Material and methods
Our present study was based on a pooled analysis of 2 prospective cohort studies in Japan. The study designs for the 2 studies have been described in detail elsewhere.21, 22, 23 Briefly, for Cohort 1, we delivered a self-administered questionnaire in January 1984 to 33,453 residents, 40 years of age or older, in 3 municipalities of Miyagi Prefecture. Usable questionnaires were returned from 31,345 (93.7%) of the subjects. For Cohort 2, we delivered a self-administered questionnaire between June–August 1990 to 51,921 residents, 40–64 years of age, in 14 municipalities of Miyagi Prefecture. Usable questionnaires were returned from 47,605 (91.7%) of the subjects. Study protocols for the 2 cohorts were approved by the institutional review board of Tohoku University Graduate School of Medicine. We considered that the return of the self-administered questionnaires signed by the subjects implied their consent to participate in the study.
In both cohorts, the questionnaire included items inquiring about the frequency of recent consumption of 3 kinds of beverages (coffee, green tea, black tea) and food items, as well as questions on smoking status and history of disease. In the question about history of liver disease, the subjects were simply asked, “Have you had any liver disease?” Thus, we did not ascertain the name of the liver disease. Alcohol consumption was assessed by asking if the subject had never drunk, or was a former, or current, drinker. Current drinkers were also asked about their frequency of drinking and the amount of alcohol consumed on one occasion.
We asked the subjects about their frequency of coffee consumption according to 5 categories: never, occasionally, 1–2 cups per day, 3–4 cups per day and 5 or more cups per day. No question about the method used to brew the coffee was asked. The volume of a typical cup of coffee was 150 ml in the study region. The validation study of beverage consumption indicated that the self-reported frequency of coffee consumption among the subjects was satisfactorily valid and reliable. One hundred thirteen subjects in the study population responded to the questionnaire twice, 1 year apart, and provided four 3-day diet records within the year. Spearman's coefficient for the correlation between the amounts of coffee consumed according to the questionnaire and the amounts consumed according to the diet records was 0.70, and the correlation between consumption measured by the 2 questionnaires over 1 year was 0.72.
The end point in our analysis was incidence of primary liver cancer defined as the topography code C22.0 and fifth digit behavior code for neoplasms/3 according to the International Classification of Diseases for Oncology (2nd Ed.; ICD-O-2).24
For both cohorts, we followed the vital and residential status of each subject using a population registry for each municipality. We ascertained the incidence of cancer using the Miyagi Prefectural Cancer Registry, one of the earliest and most accurate population-based cancer registries in Japan.25 In this registry, the relevant cases were abstracted from medical records of hospitals by a medical doctor or trained medical record reviewer, except for the cases reported from an institution to the registry. A follow-up was conducted from 1 January 1984–31 December 1992 for Cohort 1, and from 1 August 1990–31 March 1997 for Cohort 2.
We excluded cancer cases prevalent at the baseline (541 cases in Cohort 1 and 1,110 cases in Cohort 2). Then, we excluded subjects who did not answer the question about coffee consumption (8,400 subjects in Cohort 1 and 7,792 subjects in Cohort 2). Consequently, our analysis included 22,404 subjects (10,588 men and 11,816 women) including a total of 70 cases of liver cancer (50 men and 20 women) in Cohort 1, and 38,703 subjects (18,869 men and 19,834 women) including a total of 47 cases of liver cancer (41 men and 6 women) in Cohort 2.
Diagnosis of the 117 primary liver cancer cases was confirmed by medical records (n = 90, 76.9%) or death certificates alone (n = 27, 23.1%). In the 90 cases of primary liver cancer reviewed from medical records, the diagnosis was confirmed by histologic or cytologic examination in 43 cases, and by imaging (ultrasonography, computed tomography, magnetic resonance imaging, or angiography) alone in 35 cases. Although medical records had been reviewed, no further information on the basis of diagnosis was obtainable from the registered data in 12 cases. Among the 43 cases of primary liver cancer established by histologic or cytologic examination, the histological types were hepatocellular carcinoma (ICD-O-2 morphology code M-8170/3, n = 35), unspecified cancer (M-8000/3, n = 6), adenocarcinoma (M-8140/3, n = 1) and hemangiosarcoma (M-9120/3, n = 1).
We counted the number of person-years of follow-up for each subject from the beginning of follow-up until the date of diagnosis of liver cancer, the date of emigration from the study districts, the date of death or the end of follow-up, whichever occurred first. Total person-years accrued were 170,640 for Cohort 1 and 284,948 for Cohort 2. We combined the upper 3 categories of coffee consumption into the single category “1 or more cups/day” because of the small number of subjects in each category. In fact, the numbers of patients with liver cancer who reported drinking coffee 1–2, 3–4 or 5 or more cups/day were 19, 11 and 0, respectively. Relative risk was computed as the incidence rate among subjects in each category of coffee consumption divided by the rate among those who had never drunk coffee. We considered subjects who had never drunk coffee as the reference group.
We used Cox proportional hazards regression analysis to estimate the relative risk (RR) and 95% confidence interval (CI) of liver cancer incidence according to categories of coffee consumption and to adjust for potentially confounding variables, using SAS version 8.2 statistical software (SAS Inc., Cary, NC).
As the primary outcome, we examined the association between coffee consumption and the risk of incidence of primary liver cancer. We considered the following variables to be potential confounders: age (in years), gender, history of liver disease (yes or no), alcohol consumption (never drinker, former drinker, occasional drinker [current drinker less often than daily], daily drinker who consumed <45.6 g alcohol/day, or 45.6 g or more alcohol/day), and smoking status (never smoker, former smoker, currently smoking 1–19 cigarettes/day, currently smoking at least 20 cigarettes/day).
To obtain a summary measure of the results from Cohort 1 and Cohort 2, we used the general variance-based method.26 The p-values for the analysis of linear trends were calculated by treating the coffee consumption category as an ordinal variable. All reported p-values are 2-tailed, and differences at p < 0.05 were considered statistically significant.
Table I compares the characteristics of subjects according to coffee consumption. The subjects with higher coffee consumption tended to be younger and male, and were more likely to be drinkers and heavy smokers (20 cigarettes or more/day) and were less likely to have a history of liver disease. The consumption of green tea did not vary according to the consumption of coffee. We observed a similar tendency in Cohort 2.
|Coffee consumption (cups/day)|
|Cohort 1||Cohort 2|
|No. of subjects||4938||9507||7959||6954||14130||17619|
|Age (years), mean ± SD2||60.4 ± 11.8||56.0 ± 10.5||52.4 ± 9.7||54.2 ± 7.0||52.8 ± 7.2||48.9 ± 7.1|
|History of liver disease (%)||5.4||4.7||4.5||6.6||4.8||4.4|
|Alcohol drinking (%)|
|Daily, <45.6 g/d||7.0||8.5||9.4||8.0||8.3||9.2|
|Daily, ≥45.6 g/d||15.4||15.7||17.4||21.1||18.2||17.6|
|Daily, <19 cigarettes||10.6||10.7||12.7||10.5||10.7||11.0|
|Daily, ≥20 cigarettes||13.2||16.3||27.7||18.6||20.8||31.7|
|Daily consumption (%)|
|Green tea (≥3 cups/day)||57.0||68.6||60.3||47.9||52.2||39.6|
|Black tea (≥3 cups/day)||0.7||0.9||2.7||0.3||0.4||0.8|
Table II shows the association between coffee consumption and the risk of primary liver cancer. We found that higher coffee consumption was significantly associated with a lower risk of incidence of liver cancer. The pooled multivariate RR (95% CI) of liver cancer in subjects who drank coffee never, occasionally, and 1 or more cups/day were 1.00, 0.71 (0.46–1.09) and 0.58 (0.36–0.96), respectively (p for trend = 0.024). In the analysis of each cohort, a similar trend was observed. Results remained essentially the same when we excluded the 41 cases (22 cases [13 men and 9 women] in Cohort 1, 19 cases [17 men and 2 women] in Cohort 2) with liver cancer diagnosed in the first 3 years of follow-up (data not shown).
|Variable||Coffee consumption (cups/day)||p for Trend|
|No. of cases of liver cancer/person-years|
|Age, gender-adjusted RR (95% CI)|
|Cohort 1||1.00||0.48 (0.28–0.82)||0.44 (0.24–0.83)||0.0076|
|Cohort 2||1.00||0.92 (0.45–1.87)||0.61 (0.28–1.33)||0.19|
|Pooled||1.00||0.61 (0.40–0.94)||0.50 (0.31–0.82)||0.0041|
|Multivariate RR1 (95% CI)|
|Cohort 1||1.00||0.56 (0.33–0.97)||0.53 (0.28–1.00)||0.038|
|Cohort 2||1.00||1.05 (0.52–2.16)||0.68 (0.31–1.51)||0.30|
|Pooled||1.00||0.71 (0.46–1.09)||0.58 (0.36–0.96)||0.024|
When we did not count 27 cases confirmed by death certificate only (DCO) as primary liver cancer, the point estimate of the RR of liver cancer had a similar trend. The pooled multivariate RR of liver cancer in subjects who drank coffee never, occasionally, and 1 or more cups/day were 1.00, 0.87 (0.52–1.45) and 0.73 (0.41–1.29), respectively (p for trend = 0.25).
Table III shows the association between coffee consumption and the risk of liver cancer in subgroup analyses. The RR of liver cancer were below unity irrespective of whether the subjects were younger or older, male or female, current drinkers or not, current smokers or not and had had liver disease or not. A significant inverse association between coffee consumption and the risk of liver cancer was observed in the subjects with a history of liver disease (p for trend = 0.047), whereas the association was not significant in the subjects without a history of liver disease.
|Coffee consumption (cups/day)||p for Trend1|
|40-59 (n = 46,718)|
|No. of cases||14||23||20|
|Multivariate RR1 (95% CI)||1.00||0.88 (0.44–1.74)||0.81 (0.40–1.64)||0.59|
|60- (n = 14,389)|
|No. of cases||27||23||10|
|Multivariate RR1 (95% CI)||1.00||0.58 (0.32–1.09)||0.44 (0.21–0.93)||0.015|
|Male (n = 29,457)|
|No. of cases||28||36||27|
|Multivariate RR1 (95% CI)||1.00||0.73 (0.44–1.21)||0.64 (0.37–1.12)||0.11|
|Female (n = 31,650)|
|No. of cases||13||10||3|
|Multivariate RR1 (95% CI)||1.00||0.66 (0.28–1.57)||0.54 (0.14–2.07)||0.12|
|Never (n = 21,914)|
|No. of cases||14||10||4|
|Multivariate RR1 (95% CI)||1.00||0.69 (0.29–1.65)||0.46 (0.14–1.52)||0.095|
|Former (n = 2,974)|
|No. of cases||8||6||6|
|Multivariate RR1 (95% CI)||1.00||0.60 (0.20–1.78)||0.74 (0.23–2.39)||0.58|
|Current (n = 28,750)|
|No. of cases||15||28||14|
|Multivariate RR1 (95% CI)||1.00||0.90 (0.47–1.71)||0.56 (0.24–1.29)||0.097|
|Never (n = 27,233)|
|No. of cases||12||14||2|
|Multivariate RR1 (95% CI)||1.00||0.90 (0.38–2.11)||0.27 (0.06–1.32)||0.10|
|Former (n = 6,164)|
|No. of cases||14||9||2|
|Multivariate RR1 (95% CI)||1.00||0.53 (0.21–1.34)||0.18 (0.04–0.84)||0.012|
|Current (n = 18,334)|
|No. of cases||11||16||19|
|Multivariate RR1 (95% CI)||1.00||0.80 (0.36–1.75)||0.90 (0.41–1.97)||0.84|
|History of liver disease|
|yes (n = 2,985)|
|No. of cases||23||17||13|
|Multivariate RR1 (95% CI)||1.00||0.51 (0.27–0.97)||0.52 (0.25–1.07)||0.047|
|no (n = 58,122)|
|No. of cases||18||29||17|
|Multivariate RR1 (95% CI)||1.00||0.98 (0.53–1.80)||0.75 (0.37–1.50)||0.33|
We also examined the relationship between green tea consumption and the risk of primary liver cancer, but the relationship was null. After adjustment for the same covariates as those used for analysis of coffee consumption, the pooled RR (95% CI) of primary liver cancer in subjects who drank 2 or less, 3–4, and 5 or more cups of green tea/day were 1.00, 1.20 (0.75–1.94) and 0.90 (0.56–1.44), respectively (p for trend = 0.70). We were unable to estimate the relationship between consumption of black tea and liver cancer incidence because the proportion of subjects who drank 1 or more cups of black tea/day was only 7.8% in Cohort 1 and 2.8% in Cohort 2.
In this pooled analysis of 2 prospective cohorts, we found a statistically significant inverse association between coffee consumption and the incidence risk of primary liver cancer. This result is consistent with recent case-control studies in Italy and Greece.20 Consumption of coffee in our subjects was not particularly low in comparison with the Western population. The proportion of subjects who reported drinking 1 or more cups of coffee/day was 41.9% in our study and 58.5% in the United States.16 Almost half of the liver cancer cases occurred in the 2,985 subjects who reported a history of liver disease at the baseline. Although we had no specific information on liver diseases, this result was consistent with the strong association between chronic liver diseases such as chronic hepatitis or liver cirrhosis and the risk of liver cancer.27
Our study had several strengths. We recruited our subjects from the general population and identified a large number of cases of liver cancer among them. The information on coffee consumption and other variables was obtained before the cases of liver cancer were diagnosed, thus avoiding any effect of recall bias. The questionnaire used for measuring coffee consumption had a reasonably high level of validity and reproducibility. In addition, the inverse association between coffee consumption and the risk of liver cancer was unchanged after adjustment for, and stratification by, potential confounders. Moreover, to avoid any potential bias from subclinical conditions, we excluded subjects in whom liver cancer was diagnosed in the first 3 years of follow-up. The inverse association was unchanged after this exclusion.
Our study also had some limitations. First, we had no information about history of HBV or HCV infection. The prevalence of hepatitis B surface antigen (HBsAg) and antibodies against HCV (anti-HCV) among subjects 40 years of age or older in this area was 1.87% and 2.19%, respectively.28 In Japan, 28% and 43% of liver cancers are estimated to be attributable to HBV and HCV, respectively.7 If these viral infections were related to change in coffee consumption, the association between coffee consumption and the risk of liver cancer would be confounded. In our study, the RR of liver cancer were below unity, irrespective of whether the subjects had liver disease or not. In a previous study,17 the inverse relationship between coffee consumption and the odds ratio of liver cirrhosis was independent of HCV and HBV infection. Because of the strong association between these virus infections and the risk of liver cancer, however, even a weak inverse association between these viral infections and coffee consumption could introduce negative confounding, which would lead to overestimation of the effect of coffee consumption on the decreased liver cancer risk. Measurement of HBV and HCV infections would be needed in further prospective studies.
Second, primary liver cancer cases identified on the basis of death certificates alone without confirmation by medical records might have a possibility of misclassifying secondary metastasis to the liver as primary liver cancer. We carried out an additional analysis not considering the DCO cases as primary liver cancer. The inverse association between coffee consumption and the risk of primary liver cancer was not materially changed. We believe it is unlikely that the DCO cases distorted the inverse association substantially.
Third, we excluded 16,192 subjects because they did not answer the question on coffee consumption. Fifty-one cases of liver cancer were diagnosed in this group. We considered that the characteristics of subjects who did not report their coffee consumption were essentially similar to those of subjects who did. The 2 groups were similar with respect to the prevalence of current smokers (35.5% and 35.4% of the groups, respectively), current alcohol drinkers (51.4% and 54.2%, respectively), and a history of liver disease (4.2% and 4.9%, respectively), apart from the distribution of age classes (subjects 40–59 years of age made up 53.6% and 76.5% of the groups, respectively) and gender (men made up 41.3% and 48.2%, respectively). The pooled multivariate RR (95% CI) of liver cancer in the subjects who did not answer the question about their coffee consumption, as compared to those who did, was 1.23 (0.87–1.74). Thus, our result might not have been substantially biased by exclusion of the subjects who did not answer the question on coffee consumption.
Fourth, we were unable to distinguish between never and former coffee drinkers, as this information was not collected at the baseline. Such information would allow more precise estimation of the effects of coffee on liver cancer in further studies. Finally, we did not investigate the method used for brewing coffee. For practical purposes, however, we can consider that most of the subjects would have consumed instant or filtered coffee because unfiltered coffee is rarely consumed in Japan.29
Among the subjects with a history of liver disease, we observed a significant inverse relationship between coffee consumption and liver cancer. We speculate that coffee may prevent liver cancer more effectively among subjects with liver disease than among those without liver disease. If the subjects with a history of liver disease had reduced their coffee consumption at the time of baseline data collection because of ill health, an inverse association between coffee consumption and liver cancer would have been observed. Among the subjects with a history of liver disease, we did not observe a decreasing trend in the proportion of former alcohol drinkers and former smokers, according to the frequency of coffee consumption, who might have quit drinking and smoking due to ill health. In our data, among the subjects with a history of liver disease, the proportions of former alcohol drinkers who drank coffee never, occasionally or 1 or more cups/day were 13.5%, 11.5% and 13.1% respectively, and the corresponding proportions of former smokers were 17.7%, 19.3% and 17.6%, respectively. Kuper et al.19 failed to estimate the odds ratio of liver cancer for coffee drinkers among a subgroup of subjects with HBsAg or anti-HCV because there were no controls who did not drink coffee among these subjects. Further studies to elucidate the preventive effects of coffee consumption against liver cancer among subjects with chronic hepatitis or liver cirrhosis are needed.
Meanwhile, among the subjects without a history of liver disease, the inverse association between coffee consumption and the risk of liver cancer was not significant, but the RR of liver cancer was below unity. Among the subjects without a history of liver disease, we could not conclude from our data whether we might fail to detect a significant inverse association between coffee consumption and the risk of liver cancer due to insufficient statistical power or whether there might be no association.
It remains unclear which ingredient(s) of coffee is protective against liver cancer. Mutagenic and antimutagenic effects of coffee and caffeine on cultured cells of bacterial and mammalian origin have been demonstrated, but mutagenic effects would be almost non-existent at the usual levels of coffee consumption in humans.30 The caffeine concentration in coffee and green tea is 0.06% and 0.02%, respectively.31 Caffeine might not have a protective effect against liver cancer because our study indicated that consumption of green tea was not associated with the risk of liver cancer. Coffee also contains chlorogenic acid, a phenolic compound, whose inhibitory effects on chemical carcinogenesis in the liver have been demonstrated in an animal model.11 The diterpenes cafestol and kahweol, both present in coffee, have been implicated in anticarcinogenic activity,32 but it seems unlikely that they would have had a protective effect against liver cancer in this study group because their quantity is almost negligible in instant and filtered coffee.33
In conclusion, we have found that coffee consumption is significantly associated with a decreased incidence of liver cancer. In addition, subgroup analysis among our subjects with a history of liver disease showed an inverse association between coffee consumption and the risk of liver cancer. Further studies to clarify the role of coffee in prevention of liver cancer among the population at high risk are needed.
- 2GLOBOCAN 2000: cancer incidence, mortality and prevalence worldwide. version 1.0. IARC Cancer Base No. 5. Lyon: IARC, 2001., , , .
- 6IARC. Monographs on evaluation of carcinogenic risks to humans. Hepatitis viruses. vol. 59. Lyon: IARC, 1994.
- 8IARC. Monographs on evaluation of carcinogenic risks to humans. Alcohol drinking. vol. 44. Lyon: IARC, 1998.
- 9World Cancer Research Fund. Food, nutrition and the prevention of cancer: a global perspective. Liver. Washington, DC: American Institute for Cancer Research, 1997. p 202– 14.
- 10IARC. Monographs on evaluation of carcinogenic risks to humans. Tobacco smoking and involuntary smoking. vol. 83. Lyon: IARC, 2002.
- 21A cohort study on the relation of lifestyle, personality and biologic markers to cancer in Miyagi, Japan: study design, response rate and profiles of the cohort subjects. J Epidemiol 1995; 5: 153–57., , , , , , , , , , , .
- 24PercyC, Van HoltenV, MuirC International classification of diseases for oncology, 2nd ed. Geneva: WHO, 1990.
- 25Japan, Miyagi Prefecture. In: ParkinDM, WhelanSL, FerlayJ, RaymondL, YoungJ, eds. Cancer incidence in five continents. International Agency for Research on Cancer, vol. 7. Lyon: IARC, 1997. p 386–89., .
- 26Statistical methods in meta-analysis. In: meta-analysis, decision analysis, and cost-effectiveness analysis: methods for quantitative synthesis in medicine. New York: Oxford University Press, 1994.p 90–114..
- 28Study on the result of hepatitis virus test for essential health check. Yobou Igaku J 2003; 388: 32–36., , .
- 29All Japan Coffee Association. A basic survey for monitoring trends in the demand for coffee. Tokyo: All Japan Coffee Association, 2003.
- 31Resources Council, Science and Technology Agency, Japan. Standard tables of food composition in Japan, 5th ed. Tokyo: Printing Bureau, Ministry of Finance, 2000.