Green tea consumption and risk of stomach cancer: A meta-analysis of epidemiologic studies
Article first published online: 30 OCT 2008
Copyright © 2008 Wiley-Liss, Inc.
International Journal of Cancer
Volume 124, Issue 3, pages 670–677, 1 February 2009
How to Cite
Myung, S. K., Bae, W. K., Oh, S. M., Kim, Y., Ju, W., Sung, J., Lee, Y. J., Ko, J. A., Song, J. I. and Choi, H. J. (2009), Green tea consumption and risk of stomach cancer: A meta-analysis of epidemiologic studies. Int. J. Cancer, 124: 670–677. doi: 10.1002/ijc.23880
- Issue published online: 18 NOV 2008
- Article first published online: 30 OCT 2008
- Manuscript Accepted: 14 JUL 2008
- Manuscript Received: 9 MAY 2008
Vol. 124, Issue 6, 1496, Article first published online: 9 DEC 2008
- green tea;
- stomach cancer;
- systematic review
This meta-analysis investigated the quantitative association between the consumption of green tea and the risk of stomach cancer in epidemiologic studies using crude data and adjusted data. We searched MEDLINE, EMBASE and the Cochrane Review in August 2007. All the articles searched were independently reviewed and selected by 3 evaluators according to predetermined criteria. A total of 13 epidemiologic studies were included. When all the case–control and cohort studies were pooled, the relative risks (RR) of stomach cancer for the highest level of green tea consumption when compared with the lowest level of consumption were shown to be 1.10 (95% confidence interval (CI), 0.92–1.32) using the crude data and 0.82 (95% CI, 0.70–0.96) using the adjusted data. In the meta-analyses of case–control studies, no significant association was seen between green tea consumption and stomach cancer using the crude data (RR, 0.79; 95% CI, 0.58–1.07), but green tea was shown to have a preventive effect on stomach cancer using the adjusted data (RR, 0.73; 95% CI, 0.64–0.83). In the meta-analyses of the recent cohort studies, the highest green tea consumption was shown to significantly increase stomach cancer risk using the crude data (RR, 1.59; 95% CI, 1.16–2.18), but no significant association between them was seen when using the adjusted data (RR, 1.04; 95% CI, 0.93–1.17). Unlike the case–control studies, no preventive effect on stomach cancer was seen for the highest green tea consumption in the meta-analysis of the recent cohort studies. Further clinical trials are needed. © 2008 Wiley-Liss, Inc.
Green tea is consumed mainly in Asian countries, such as Korea, Japan, China and a few countries in the Middle East and North Africa.1 The main constituents of green tea extracts are polyphenols known as “catechins,” which have been reported to have anticarcinogenic, antimutagenic and antioxidant activity in experimental studies using in vivo animal models.1–3 It has been hypothesized that green tea consumption has a protective effect against the development of stomach cancer, a disease that is still prevalent in several Asian countries, including Korea, Japan and China. Since 1985, 6 case–control studies have reported that a high consumption of green tea significantly decreases the risk of stomach cancer,4–9 whereas 4 case–control studies have found a nonsignificant inverse association between green tea consumption and the risk of stomach cancer.10–13 Only 1 case–control study, using univariate analysis, has reported a significant association between green tea drinking and an increased risk of stomach cancer, although the association was found to be nonsignificant when a multiple logistic regression analysis was performed.14
In contrast to case–control studies, however, 7 prospective cohort studies, which were published from 1988 to the present, have reported that there is no significant association between green tea consumption and stomach cancer risk.15–21
Regarding those controversies, a recent qualitative review concluded that most of the case–control studies showed that green tea consumption caused a risk reduction for stomach cancer, whereas prospective studies showed no inverse association between them.22 Also, a recent quantitative meta-analysis of 14 epidemiologic studies (10 case–control studies and 4 cohort studies) indicated that there was no clear evidence of the preventive effect of green tea consumption on stomach cancer.23 When we compute an odds ratio (OR) or relative risk (RR), and its confidence interval, when conducting a meta-analysis, we are able to use 2 types of data: crude data (or unadjusted data) and adjusted data. Crude data refers to the number of cases and controls per category of exposure for case–control studies or the number of cases per category of exposure for cohort studies, without adjustment. Adjusted data refers to an OR or RR, and its corresponding 95% confidence interval (CI), already adjusted by various factors, such as age, sex, and other individual characteristics, for each of the studies. However, there has been no meta-analysis of green tea consumption and stomach cancer risk showing the differences in the findings using the crude data compared with the use of the adjusted data.
The purpose of this meta-analysis was to examine the quantitative association between the consumption of green tea and the risk of stomach cancer in epidemiologic studies using both the crude data and adjusted data.
Material and methods
Data sources and keywords
We searched and reviewed MEDLINE (PubMed: 1979 to August 2007), EMBASE (1984 to August 2007) and the Cochrane Library (2002 to August 2005), using selected common key words related to green tea consumption and stomach cancer risk in case–control and cohort studies. We also scanned bibliographies of relevant articles in order to identify additional studies. As the keywords for the literature search, we selected “green tea,” “polyphenol” and “catechin” for the exposure factors, and “stomach cancer,” “gastric cancer,” “stomach neoplasm” and “gastric neoplasm” for the outcome factors.
We included case–control studies, cohort studies and prospective studies reporting an association between green tea consumption and stomach cancer risk, and presenting the number of stomach cancer cases and controls for case–control studies and the number of stomach cancer cases and person-years or population size in the highest and lowest green tea consumption groups. We only selected articles written in English and excluded those studies with no available data for outcome measures.
Selection of relevant studies
All of the studies retrieved from the databases were independently evaluated by 3 evaluators (Dr. Myung, Dr. Bae, and Dr. Oh), each of whom is a coauthor of this study. When there were disagreements between the evaluators concerning the selected studies, these differences of opinion were resolved by discussion. In instances where the data were insufficient or missing, we attempted to contact the authors of the articles in order to request the relevant data. Of the articles found in the 3 databases, duplicate articles and those that did not meet the selection criteria were excluded. From those studies finally selected, we extracted the following data: study name (first author, year of publication), journal name, country and design, study period (years), population size, green tea consumption level, OR or RR with 95% CI, adjustments and datasets of the highest and lowest consumption levels.
Statistical analyses were performed for the case–control and cohort studies considered together and as subgroups. We used both crude data (unadjusted) and adjusted data (adjusted OR or RR with 95% CI) for meta-analysis. For the crude data analysis, we extracted the number of stomach cancer cases and controls for the case–control studies and the number of stomach cancer cases and the person-years for minus cases in the cohort studies for the highest and lowest green tea consumption levels: those numbers in the highest and lowest consumption groups from 7 studies,6, 9, 15, 16, 18, 20, 21 those numbers in 2 consumption groups from 5 studies,5, 7, 8, 13, 14 and those numbers in the highest and lower consumption groups from 1 study.4 For 1 prospective study,15 we calculated person-years using an average follow-up period of 14.8 years (number of noncases × 14.8 years), because the study reported the number of noncases, but not the number of person-years. The numbers were calculated from the percentages of the subjects in the reference and case groups reported in a case-reference study.6 Because the green tea consumption categories varied between studies, we chose the highest and lowest levels of green tea consumption categorized in each of the studies. Definitions of the levels used in each of the studies are shown in the “green tea consumption levels” column of Table I. For the adjusted data analysis, we used the individual OR or RR with 95% CI represented in each study.
|Study (reference)||Journal||Country; design||Study period (years)||Population||Green tea consumption levels||OR or RR (95% CI)||Adjustments||Dataset1|
|1||Kono et al, 1988 (4)||Jpn J Cancer Res||Japan; population and hospital-based case-control||1979–1982 (3)||139 cases and 2574 hospital controls||≥10 cups/day vs. Less||0.5 (0.3–1.1)||Age, sex, smoking, mandarin oranges and other fruits.||8/255, 131/2319|
|2||Lee et al, 1990 (14)||Anticancer Res||Taiwan; hospital-based case-control||Not stated||210 cases and 810 hospital controls||Yes vs. no||2.00 (p < 0.10)||Cigarette smoking, alcohol drinking, salted meat consumption, fried food consumption, fermented beans consumption and milk consumption.||10/14, 200/796|
|3||Yu et al, 1995 (5)||Cancer Causes Control||China; population-based case-control||1991–1993 (2)||711 cases and 711 matched controls||Drinkers vs. nondrinkers||0.71 (0.54–0.93)||Age, sex, place of residence, education, birthplace, alcohol consumption, and cigarette smoking.||222/247, 489/464|
|4||Ji et al, 1996 (13)||Cancer||China; population-based case-control||1988–1989 (1)||1124 cases and 1451 frequency-matched controls||Regular drinker vs. nondrinker||Men: 0.96 (0.77–1.21) Women: 0.77 (0.52–1.13)||Age, income and education among women; further adjusted for smoking and alcohol drinking among men.||409/506, 620/841|
|5||Inoue et al, 1998 (6)||Cancer Causes Control||Japan; hospital-based case–control||1990–1995 (5)||893 cases and 21128 noncancer outpatients controls||≥7 cups/day vs. rarely||0.69 (0.48–1.00)||Age, sex, year and season at first hospital-visit, habitual smoking, habitual alcohol drinking, regular physical exercise and intake of coffee, black tea, fruit, rice and beef.||87/1723, 53/1221|
|6||Ye et al, 1998 (7)||World J Gastroenterol||China; population-based case–control||1994–1995 (1)||272 cases and 544 matched controls||>0.75 kg/year vs. ≤0.75 kg/year||1.72 (1.26–2.36); reference = highest level||None||47/144, 225/400|
|7||Mu et al, 2005 (9)||Int J Cancer||China; population-based case–control||2000 (0.5)||206 cases and 415 healthy controls||>250 g/month vs. never||0.39 (0.17–0.91)||Age, sex, education, income, body mass index, pack-year of smoking, alcohol drinking, very hot food eating habit, H.pylori infection, stomach disease history and family history of stomach cancer.||12/70, 131/216|
|8||Setiawan et al, 2001 (8)||Int J Cancer||China; population-based case–control||1995 (0.5)||133 cases and 433 healthy controls||Drinking vs. no drinking||0.52 (0.29–0.94)||Age, sex, education, body mass index, pack-years of smoking and alcohol drinking.||37/173, 95/250|
|1||Galanis et al, 1998 (15)||Int J Epidemiol||U.S (Hawaii); prospective cohort study||14.8 years||108 cases among 11,907 randomly selected Japanese residents of Hawaii||2 or more cups/day vs. none||1.5 (0.9–2.3)||Age, sex, years of education and Japanese place of birth.||32/1471, 57/9012 (No. of cases/No. of population at baseline)|
|2||Tsubono et al, 2001 (16)||N Engl J Med||Japan; prospective cohort||1984–1992 (8)||419 cases among 26,311 residents in Miyagi prefecture (11,902 men and 14,409 women)||≥5 cups/day vs. <1 cups/day||1.2 (0.9–1.6)||Age, sex, type of health insurance, history of peptic ulcer, cigarette smoking, alcohol consumption, consumption of rice, black tea, coffee, meat, green or yellow vegetables, pickled vegetables, other vegetables, fruits and bean-paste soup.||206/85299, 66/36572|
|3||Hoshiyama et al, 2002 (18)||Br J Cancer||Japan; prospective cohort||1988–1997 (9)||359 deaths from stomach cancer among 72,851 inhabitants of 45 areas (the JACC study)||≥10 cups/day vs. <1 cups/day||Men: 1.0 (0.5–2.0) Women: 0.7 (0.3–2.0)||Age, smoking status, history of peptic ulcer, family history of stomach cancer, consumption of rice, miso soup, green-yellow vegetables, white vegetables, fruits and preference for salty foods.||47/58288, 44/105313|
|4||Sasazuki et al, 2004 (20)||Cancer Causes Control||Japan; prospective cohort||1990–2001 (11) for Cohort I 1993–1999 (6) for Cohort II||892 cases among 72,943 subjects (the JPHC study)||≥5 cups/day vs. <1 cups/day||Men: 0.98 (0.77–1.25) Women: 0.67 (0.43–1.04)||Age, area, cigarette smoking and consumption of fruit, green or yellow vegetables, fish gut, miso soup, rice, black tea and coffee.||260/157740, 163/135003|
|5||Sauvaget et al, 2005 (21)||Cancer Causes Control||Japan; prospective cohort||1980–1999 (19)||1270 cases among 38,576 atomic-bomb survivors (the Life Span Study)||≥5 cups/day vs. <2 cups/day||1.06 (0.89–1.25)||Sex-specific age, sex, city, radiation dose, sex-specific smoking habits and education level.||348/124704, 242/92393|
We estimated a pooled OR or RR with 95% CI based on both fixed-effects and random-effects models. Heterogeneity was assessed using Higgins I2, which measures the percentage of the total variation across studies that is due to heterogeneity, rather than chance.24I2 is calculated as follows:
where Q is Cochran's heterogeneity statistic and df is its degrees of freedom. Negative values of I2 are set at zero so that I2 lies between 0% (no observed heterogeneity) and 100% (maximal heterogeneity). An I2 value greater than 50% may be considered to represent substantial heterogeneity.
We used the Mantel-Haenszel method for a fixed effects analysis and the DerSimonian and Laird method for a random effects analysis. Begg's funnel plot and Egger's test were used to identify publication bias. Funnel plots are scatter plots of the log ORs of individual studies on the x-axis against the 1/standard error (or standard error for sample size; a measure of precision) of each study on the y-axis. If there is no publication bias, the log ORs of small studies scatter widely at the bottom of the graph, with the spread narrowing among large studies. Such a plot resembles a symmetrical inverted funnel. However, in the presence of publication bias, the plot becomes asymmetrical, because some small studies demonstrating no statistical significance tend to go unpublished. Egger's test is a test for a linear regression of normalized effect estimate (log OR/standard error) against its precision (1/standard error). If the p-value for Egger's test was less than 0.05, we estimated that there was a publication bias.
The Stata SE version 9.1 software package (StataCorp, College Station, Texas) was used for all of the statistical analyses.
Our study included a total of 13 epidemiologic studies (8 case–control studies4–9, 13–14 and 5 cohort studies15, 16, 18, 20, 21) published between 1988 and 2005. Figure 1 displays a flow diagram of the procedure used to identify the relevant studies. Searches of the 3 databases and the bibliographies of relevant articles yielded 222 articles. After the exclusion of duplicates (n = 67), we reviewed all of the remaining screened articles; of these, 13 studies were included in the final analysis. The main reasons for excluding the studies were as follows: written in a language other than English25–29 (n = 5); insufficient data10–12, 30–33 (n = 7); not a case–control or cohort study19, 22, 34–39 (n = 8), for example, animal experiments, cancer cell lines experiments, correlation analyses and review articles; identical populations17, 40, 41 (n = 3); or the use of an acute biomarker as an exposure source42 (n = 1). Regarding the identical population studies, we selected a study with a larger sample size21 and a cohort study18 rather than a nested case–control study.40 Table I shows the main characteristics of the studies included in the analysis. Of the 8 case–control studies, 5 studies were conducted in China, 2 in Japan and 1 in Taiwan. Of the 5 cohort studies, 4 studies were conducted in Japan and 1 among the Japanese residents of Hawaii, the United States. Four of the 5 cohort studies were published in the 2000s, whereas 6 of the 8 case–control studies were published between the late 1980s and the 1990s. The study periods of the cohort studies ranged from 6 years to 19 years.
All of the case–control and cohort studies using crude data and adjusted data
As shown in Figure 2, no significant association was seen between green tea consumption (highest vs. lowest consumption level) and stomach cancer risk when using the crude data: the pooled RR of the highest green tea consumption when compared with the lowest consumption for stomach cancer risk was 1.10 (95% confidence interval, 0.92–1.32) in a random effects analysis. There was a statistically significant heterogeneity across studies (I2 = 87.2%). In contrast, when using the adjusted data (adjusted OR or RR with 95% CI), stomach cancer risk was seen to be significantly decreased in the highest green tea consumption group (RR, 0.82; 95% CI, 0.70–0.96), with significant heterogeneity (I2 = 63.1%). Figure 3 depicts a symmetric Begg's funnel plot, indicating that there was no publication bias. Further, there was no evidence of bias using Egger's test (p for bias = 0.81).
Case–control studies using crude data and adjusted data
As shown in Figure 4, in case–control studies, although no significant association was seen between green tea consumption and stomach cancer risk using the crude data (RR, 0.79; 95% CI, 0.58–1.07; I2 = 82.2%), green tea consumption was seen to have a preventive effect on stomach cancer using the adjusted data (RR, 0.73; 95% CI, 0.64–0.83; I2 = 44.5%).
Cohort studies using crude data and adjusted data
Figure 5 shows that the highest green tea consumption was seen to significantly increase the stomach cancer risk in cohort studies when using the crude data (RR, 1.59; 95% CI, 1.16–2.18; I2 = 86.1%). In contrast, no significant association was seen between them when using the adjusted data (RR, 1.04; 95% CI, 0.93–1.17; I2 = 25.5%).
We found that there were discrepancies in the effects of green tea consumption on stomach cancer risk between case–control studies and cohort studies as well as between the crude data and adjusted data in our meta-analysis of published epidemiologic studies. When using the crude data, though no significant association between green tea consumption and stomach cancer was shown in either the meta-analysis of the case–control studies or the combined meta-analysis of the case–control and cohort studies, the highest green tea consumption was seen to significantly increase the stomach cancer risk in the cohort studies. However, when using the adjusted data, we found that green tea consumption at the highest level, when compared with that at the lowest level, was seen to significantly decrease the risk of stomach cancer in a meta-analysis of the case–control studies and in a combined meta-analysis of the case–control and cohort studies. In addition, no significant association was seen in a meta-analysis of the cohort studies.
These discrepancies concerning the effects of green tea consumption on stomach cancer between the case–control studies and cohort studies may be partly associated with the 2 main biases of case–control studies, recall and selection: cancer patients may recall their dietary habits differently from healthy controls, and healthy controls are rarely representative of the population as a whole and tend to report a healthy dietary habit.43, 44 In retrospective studies, such as case–control studies, the decreased consumption of the green tea after abdominal symptoms due to stomach cancer, which may have biased the patient's recall of past consumption, might result in the patient's underestimating their true intake.16 In addition, the accuracy of recalling past dietary habits is known to be influenced by present dietary habits.45
Our nonsignificant or negative findings regarding the effects of green tea consumption on stomach cancer in the meta-analyses of the cohort studies contradict the results of previous experimental studies on this topic using in vivo animal models and in vitro cancer cell lines.39, 46, 47 These experimental studies have suggested that green tea polyphenols might have a protective effect against stomach cancer due to apoptosis-inducing, antimutagenic and antioxidant activities. The reason for this discrepancy between the results of experimental studies and our meta-analyses is unclear. However, we suggest 2 possible explanations. The first is that, in the context of stomach cancer, there might be a difference in the biological activities of polyphenol as an individual constituent extracted from green tea and the green tea taken as a whole. Several constituents of green tea have been identified, including chlorophylls, pheophytins, carotenoids, theanine (amino acid), caffeine and other phytochemicals, which are known to have anticarcinogenic and antioxidant activities similar to polyphenols.48–52 Despite the protective role of an individual constituent against the development of cancer, it is possible that the adverse effect of green tea taken as a whole is due to the interactions and complex biological mechanisms of its multiple constituents. This assumption may offer an explanation for the unexpected results obtained from the ATBC (the Alpha-Tocopherol, Beta-Carotene Cancer Prevention Study)53 and CARET (the Beta-Carotene and Retinol Efficacy Trial)54 studies, which demonstrated that beta-carotene supplementation increases the incidence of lung cancer. It was suggested that the unexpected findings could be explained by the fact that beta-carotene supplements adversely affect the bioavailability of other dietary carotenoids.54 However, to date, there has been no definite evidence in support of this assumption.
The second explanation was discussed in a meta-analysis of antioxidant supplements and mortality in randomized trials.55 It suggests the following scenario: Oxidative stress is a probable consequence of certain pathological conditions.56 However, the elimination of free radicals from the human body following the excessive consumption of green tea may interfere with certain defensive activities, such as apoptosis, phagocytosis and detoxification.57, 58 This hypothesis should be explored through further studies in order to elucidate the biological mechanism and the effect of antioxidants in relation to the development of stomach cancer.
Our study does, however, have several limitations. First, because we only analyzed cohort studies of the Japanese population, we are not able to generalize our findings for all populations. Second, the exclusion of non-English language articles might bias our findings. However, to the best of our knowledge, there have been few cohort studies on this topic in other countries that have not been published in English; therefore, this exclusion criterion should not have substantially altered the results of our analysis of cohort studies. Lastly, because we used the extreme categories of highest and lowest green tea consumption levels as measures of exposure, we were not able to conclude that the consumption of green tea itself increases stomach cancer risk.
To summarize, we found that no preventive effect on stomach cancer was seen for the highest green tea consumption in the meta-analysis of recent cohort studies, which have fewer biases than case–control studies and are more persuasive, even though a preventive effect was seen for green tea consumption in case–control studies. Furthermore, from the findings of a meta-analysis of cohort studies using crude data, the highest green tea consumption group had an increased risk of stomach cancer. Therefore, the potential effects of green tea consumption (either beneficial or detrimental) on our health, particularly in relation to stomach cancer risk, should not be overemphasized. Our findings and explanations should be explored in future research, including clinical trials and further epidemiologic studies.
- 25Vegetables and fruits and risk of stomach cancer. Medicina (Kaunas) 2005; 41: 733–40. [Article in Lithuanian], , , , , .
- 26Study on the protective effect of green tea on gastric, liver, and esophageal cancer. Zhonghua Yu Fang Yi Xue Za Zhi 2003; 37: 171–3. [Article in Chinese], , , , , , .
- 27A case-control study on drinking green tea and decreasing risk of cancers in the alimentary canal among cigarette smokers and alcohol drinkers. Zhonghua Liu Xing Bing Xue Za Zhi 2003; 24: 192–5. [Article in Chinese], , , , , , , , , .
- 45Nutritional epidemiology, 2nd ed. New York: Oxford University Press, 1988..
- 47Induction of apoptosis in human stomach cancercells by green tea catechins. Oncol Rep 1998; 5: 27–9., , , , , , , .
- 51Potent suppressing activity of the non-polyphenolic fraction of green tea (Camellia sinensis) against genotoxin-induced umu C gene expression in Salmonella typhimurium (TA 1535/pSK1002)—association with pheophytins a and b. Cancer Lett 1997; 120: 117–23., .