- Split View
-
Views
-
Cite
Cite
Shamima Akter, Ikuko Kashino, Tetsuya Mizoue, Keitaro Matsuo, Hidemi Ito, Kenji Wakai, Chisato Nagata, Tomio Nakayama, Atsuko Sadakane, Keitaro Tanaka, Akiko Tamakoshi, Yumi Sugawara, Norie Sawada, Manami Inoue, Shoichiro Tsugane, Shizuka Sasazuki, ; for the Research Group for the Development and Evaluation of Cancer Prevention Strategies in Japan, Coffee drinking and colorectal cancer risk: an evaluation based on a systematic review and meta-analysis among the Japanese population, Japanese Journal of Clinical Oncology, Volume 46, Issue 8, August 2016, Pages 781–787, https://doi.org/10.1093/jjco/hyw059
- Share Icon Share
Abstract
It remains unclear whether coffee drinking is associated with colorectal cancer risk. We performed a systematic review and meta-analysis of epidemiologic studies on this issue among the Japanese population.
Original data were obtained from MEDLINE searches using PubMed or from searches of the ‘Ichushi’ database, complemented with manual searches. Meta-analysis was performed by using the random effects model to estimate the summary relative risk with 95% confidence interval according to the study design. The final judgment was made based on a consensus of the research group members with consideration for both epidemiological evidence and biological plausibility.
We identified five cohort studies and nine case–control studies. Of these, one cohort study reported a strong inverse association (in women only), whereas three case–control studies reported a strong inverse association with colon or rectal cancer. In meta-analysis, high consumption of coffee was not appreciably associated with colorectal cancer risk among cohort studies, whereas it was associated with significantly lower risk of colorectal or colon cancer among case–control studies. The summary relative risk/odds ratio (95% confidence interval) for the highest versus lowest categories of coffee consumption was 0.95 (0.77–1.17) and 0.78 (0.65–0.95) for cohort and case–control studies, respectively.
The evidence is insufficient to support that coffee drinking increases or decreases the risk of colorectal cancer among the Japanese population.
Introduction
Colorectal cancer is a major cause of morbidity and mortality throughout the world (1). It is the third most common cancer in the world, with nearly 1.4 million new cases diagnosed in 2012 (2). The burden of colorectal cancer is increasing rapidly in Asia (3). In Japan, there was a marked increase of colorectal cancer from 1975 to 2000 (4), with an current age-standardized incidence rate of 32.2 per 100 000 population (2). Such increase in Japan has been attributed to environmental factors including dietary factors (5,6). Coffee is one of the most commonly consumed beverages in the world. Increasing attention has been paid to the role of coffee as a protective dietary factor against cancer, including colorectal cancer (7). Coffee is a rich source of several bioactive compounds including polyphenols, diterpenes and caffeine, which may play a protective role in colorectal carcinogenesis (8–10). Coffee consumption has also been inversely associated with biomarkers of inflammation (11), insulin resistance (12) and insulin secretion (13), which have been linked to colorectal cancer risk (14,15).
Numerous studies have investigated the association between coffee consumption and the risk of colorectal cancer worldwide, and several meta-analyses on this issue have been published, resulting in inconsistent conclusions. Meta-analyses of case–control studies have consistently shown that coffee consumption was significantly associated with decreased risk of colorectal cancer (16–18), whereas most (16,18,19) but not all (20) meta-analyses of cohort studies reported no association. Data included in these meta-analyses, however, were derived mainly from Westerners, who had greater body fat (21) and insulin secretion capacity (22) than Asians. Thus, the effect of coffee drinking on glucose metabolism in Asians may differ from that of Westerners. Given that coffee drinking is increasingly popular in Japan (mean consumption among adults, ~11 cups/week) (23), it would be of great interest to know whether coffee consumption is associated with risk of colorectal cancer in this population. A recent meta-analysis among studies worldwide reported lower risk of colon cancer associated with coffee drinking for Asian women (16). That study, however, did not include two cohort (24,25) and several case–control studies (26–29) in Japan. Here, we conducted a systematic review and meta-analysis of the association between coffee consumption and colorectal cancer among the Japanese population.
Patients and methods
Relevant epidemiological studies were identified by searching MEDLINE for literature published up to August 2015. A search of the ‘Ichushi’ (Japana Centra Revuo Medicina) database was also conducted to identify the studies written in Japanese. These methods of literature identification were complemented by manual searches of references from pertinent articles where necessary. We employed the term ‘coffee’ combined with ‘colorectal cancer’, ‘colon cancer’, ‘rectal cancer’, ‘case–control studies’, ‘cohort studies’, ‘Japan’ and ‘Japanese’. Articles written in either English or Japanese were reviewed. Only studies on Japanese populations living in Japan were included. If results for incidence and mortality of colorectal cancer were reported from the same data source, we used the data reporting incidence. Individual results were summarized in tables separately according to study design as cohort or case–control studies.
The studies were evaluated on the basis of the magnitude of association and the strength of evidence. First, relative risk (RR) or odds ratio (OR) for the highest consumption group compared with the lowest was classified according to the magnitude of association, considering statistical significance (statistically significant (SS) or statistically not significant (NS)) into: strong (symbol ↑↑↑ or ↓↓↓), <0.5 or >2.0 (SS); moderate (symbol ↑↑ or ↓↓ ), (i) <0.5 or >2.0 (NS), (ii) >1.5 to 2.0 (SS) or (iii) 0.5 to <0.67 (SS); weak (symbol ↑ or ↓), (i) >1.5 to 2.0 (NS), (ii) 0.5 to <0.67 (NS) or (iii) 0.67 to 1.5 (SS) or no association (symbol −), 0.67 to 1.5 (NS). We chose 0.67 for the cutoff for decreased risk by dividing 1 by 1.5 (the cutoff for increased risk). A two-sided P value <0.05 or, if P value not available, a 95% confidence interval (CI) not including 1 was considered SS. The strength of evidence for epidemiological studies was evaluated by applying a method similar to that used in the WHO/FAO Expert Consultation Report (30), where evidence was classified as ‘convincing’, ‘probable’, ‘possible’ and ‘insufficient’. Despite the use of this quantitative assessment rule, an arbitrary evaluation is inevitable when considerable variation exists in the magnitude of association among studies. We assumed that biological plausibility is based on evidence in experimental models, human studies and other pertinent data from International Agency for Research on Cancer (31). With consideration for both epidemiological evidence and biological plausibility, the final judgment was made based on a consensus of the research group members, and it was therefore not necessarily objective.
Two papers presented results on this issue using the same data source (32,33), one reporting incidence and another reporting mortality, and we used incidence data (33). After excluding two case–control studies that did not show 95% CI (34,35), we conducted a meta-analysis using random effects model (36) to estimate the summary RR/OR and 95% CI of colorectal cancer for the highest versus lowest category of coffee consumption. We quantified heterogeneity among studies using the I2 index (37); I2 values range from 0 to 100%, with 0% indicating no heterogeneity and greater values expressing higher heterogeneity (38). We also assessed potential publication bias by using Funnel plot and Egger’s test. In addition, we performed several sensitivity analyses, in which studies with a small number of cases, using unique cutoff of coffee consumption, or based on mortality data only were excluded. We also conducted sex-specific analysis to explore the difference in association between men and women, which has been documented in a previous meta-analysis (16). A two-sided P value <0.05 was considered SS. All statistical analyses were performed using Stata (version 13.0; StataCorp, College Station, TX, USA).
Results
Five cohort studies (24,25,33,39,40) and nine case–control studies (26–29,34,35,41–43) met the inclusion criteria. As regards cohort studies (Table 1 and Supplementary Table 1), all reported results for men and women separately (24,25,33,39,40); of these, one additionally reported results for men and women combined (33). One study showed RRs for colon and rectum combined, but not separated (24); another reported result for colon only (39) and the remaining studies displayed results for both sites combined as well as for colon and rectum separately (25,33,40). In men, one study showed a moderate positive association with colon cancer (25), but the majority of cohort studies observed no association between coffee intake and colon cancer (33,39,40). In women, two studies showed a weak (40) and strong (39) inverse association with colon cancer, whereas another study showed a moderate positive association with colon cancer (25). All studies showed no association with rectal cancer in both men and women. One study reported a moderate inverse association with colorectal cancer in men (24), whereas another study showed a weak positive association with colorectal cancer in men (25).
References . | Study period . | Study population . | Magnitude of associationa . | ||||||
---|---|---|---|---|---|---|---|---|---|
Sex . | Number of subjects . | Age range . | Event . | Number of incident cases or deaths . | Colon . | Rectum . | Colorectum . | ||
Khan et al. (24) | 1984–2002 | Men | 1524 | 40–97 years | Death | 15 | NA | NA | ↓↓ |
Women | 1634 | 40–97 years | Death | 14 | NA | NA | |||
Oba et al. (39) | 1992–2000 | Men | 13 894 | 35–98 years | Incidence | 111 | — | NA | NA |
Women | 16 327 | 35–101 years | Incidence | 102 | ↓↓↓ | NA | NA | ||
Naganuma et al. (33) | 1990–2001 | Men | 18 867 | 40–64 years | Incidence | 284 | — | — | — |
Women | 19 834 | 40–64 years | Incidence | 173 | — | — | — | ||
Lee et al. (40) | 1990–2002 | Men | 46 023 | 40–69 years | Incidence | 726 | — | — | — |
Women | 50 139 | 40–69 years | Incidence | 437 | ↓ | — | — | ||
Yamada et al. (25) | 1988–2009 | Men | 23 607 | 40–79 years | Incidence | 557 | ↑↑ | — | ↑ |
Women | 34 614 | 40–79 years | Incidence | 444 | ↑↑ | —b | — |
References . | Study period . | Study population . | Magnitude of associationa . | ||||||
---|---|---|---|---|---|---|---|---|---|
Sex . | Number of subjects . | Age range . | Event . | Number of incident cases or deaths . | Colon . | Rectum . | Colorectum . | ||
Khan et al. (24) | 1984–2002 | Men | 1524 | 40–97 years | Death | 15 | NA | NA | ↓↓ |
Women | 1634 | 40–97 years | Death | 14 | NA | NA | |||
Oba et al. (39) | 1992–2000 | Men | 13 894 | 35–98 years | Incidence | 111 | — | NA | NA |
Women | 16 327 | 35–101 years | Incidence | 102 | ↓↓↓ | NA | NA | ||
Naganuma et al. (33) | 1990–2001 | Men | 18 867 | 40–64 years | Incidence | 284 | — | — | — |
Women | 19 834 | 40–64 years | Incidence | 173 | — | — | — | ||
Lee et al. (40) | 1990–2002 | Men | 46 023 | 40–69 years | Incidence | 726 | — | — | — |
Women | 50 139 | 40–69 years | Incidence | 437 | ↓ | — | — | ||
Yamada et al. (25) | 1988–2009 | Men | 23 607 | 40–79 years | Incidence | 557 | ↑↑ | — | ↑ |
Women | 34 614 | 40–79 years | Incidence | 444 | ↑↑ | —b | — |
NA, not available
a↑↑↑ or ↓↓↓, strong; ↑↑ or ↓↓, moderate; ↑ or ↓, weak; —, no association (see text for more detailed definition)
bNo case occurred in the highest category.
References . | Study period . | Study population . | Magnitude of associationa . | ||||||
---|---|---|---|---|---|---|---|---|---|
Sex . | Number of subjects . | Age range . | Event . | Number of incident cases or deaths . | Colon . | Rectum . | Colorectum . | ||
Khan et al. (24) | 1984–2002 | Men | 1524 | 40–97 years | Death | 15 | NA | NA | ↓↓ |
Women | 1634 | 40–97 years | Death | 14 | NA | NA | |||
Oba et al. (39) | 1992–2000 | Men | 13 894 | 35–98 years | Incidence | 111 | — | NA | NA |
Women | 16 327 | 35–101 years | Incidence | 102 | ↓↓↓ | NA | NA | ||
Naganuma et al. (33) | 1990–2001 | Men | 18 867 | 40–64 years | Incidence | 284 | — | — | — |
Women | 19 834 | 40–64 years | Incidence | 173 | — | — | — | ||
Lee et al. (40) | 1990–2002 | Men | 46 023 | 40–69 years | Incidence | 726 | — | — | — |
Women | 50 139 | 40–69 years | Incidence | 437 | ↓ | — | — | ||
Yamada et al. (25) | 1988–2009 | Men | 23 607 | 40–79 years | Incidence | 557 | ↑↑ | — | ↑ |
Women | 34 614 | 40–79 years | Incidence | 444 | ↑↑ | —b | — |
References . | Study period . | Study population . | Magnitude of associationa . | ||||||
---|---|---|---|---|---|---|---|---|---|
Sex . | Number of subjects . | Age range . | Event . | Number of incident cases or deaths . | Colon . | Rectum . | Colorectum . | ||
Khan et al. (24) | 1984–2002 | Men | 1524 | 40–97 years | Death | 15 | NA | NA | ↓↓ |
Women | 1634 | 40–97 years | Death | 14 | NA | NA | |||
Oba et al. (39) | 1992–2000 | Men | 13 894 | 35–98 years | Incidence | 111 | — | NA | NA |
Women | 16 327 | 35–101 years | Incidence | 102 | ↓↓↓ | NA | NA | ||
Naganuma et al. (33) | 1990–2001 | Men | 18 867 | 40–64 years | Incidence | 284 | — | — | — |
Women | 19 834 | 40–64 years | Incidence | 173 | — | — | — | ||
Lee et al. (40) | 1990–2002 | Men | 46 023 | 40–69 years | Incidence | 726 | — | — | — |
Women | 50 139 | 40–69 years | Incidence | 437 | ↓ | — | — | ||
Yamada et al. (25) | 1988–2009 | Men | 23 607 | 40–79 years | Incidence | 557 | ↑↑ | — | ↑ |
Women | 34 614 | 40–79 years | Incidence | 444 | ↑↑ | —b | — |
NA, not available
a↑↑↑ or ↓↓↓, strong; ↑↑ or ↓↓, moderate; ↑ or ↓, weak; —, no association (see text for more detailed definition)
bNo case occurred in the highest category.
Table 2 and Supplementary Table 2 show results of case–control studies. All the studies presented results for men and women combined only (26–29,34,35,41–43) and for the colon and rectal cancer separately (26–29,34,35,41–43); one additionally showed the result for colon and rectal cancers combined (29). Three studies found weak (34) and strong (41,42) inverse associations with colon cancer, whereas other studies showed no association with colon cancer (26–29,35,43). Three studies showed a weak to strong inverse association with rectal cancer (34,41,43), whereas one study reported a moderate positive association with rectal cancer (27).
References . | Study period . | Study subjects . | Magnitude of associationa . | |||||
---|---|---|---|---|---|---|---|---|
Sex . | Age range . | Number of casesb . | Number of controlsb . | Colon . | Rectum . | Colorectum . | ||
Kondo (34) | 1967–73 | Men and women | Not specified | 393 (M: 205, W: 188) | 582 (M: 408, W: 174) | ↓ | ↓↓ | NA |
Watanabe et al. (26) | 1977–83 | Men and women | Not specified | 203 (M: 110, W: 93) | 203 (M: 110, W: 93) | — | — | NA |
Tajima and Tominaga (35) | 1981–83 | Men and women | 40–70 years | 93 (M: 52, W: 41) | 186 (M: 111, W: 75) | — | — | NA |
Kato et al. (41) | 1986–90 | Men and women | Not specified | 223 (M: 139, W: 84) | 578 (M: 377, W: 201) | ↓↓↓ | ↓ | NA |
Hoshiyama et al. (42) | 1984–90 | Men and women | 40–69 years | 181 (M: 98, W: 83) | 653 (M: 343, W: 310) | ↓↓↓ | — | NA |
Kotake et al. (27) | 1992–94 | Men and women | Not specified | 363 (M: 214, W: 149) | 363 (M: 214, W: 149) | — | ↑↑ | NA |
Nishi et al. (28) | 1987–90 | Men and women | Not specified | 330 (M: 171, W: 159) | 660 (M: 342, W: 318) | — | — | NA |
Inoue et al. (43) | 1990–95 | Men and women | 40+ years | 628 (M: 386, W: 242) | 21128 (M: 6307, W: 14821) | — | ↓↓↓ | NA |
Wang et al. (29) | 2000–03 | Men and women | 20–74 years | 816 (%men: 60) | 815 (%men: 62) | — | — | — |
References . | Study period . | Study subjects . | Magnitude of associationa . | |||||
---|---|---|---|---|---|---|---|---|
Sex . | Age range . | Number of casesb . | Number of controlsb . | Colon . | Rectum . | Colorectum . | ||
Kondo (34) | 1967–73 | Men and women | Not specified | 393 (M: 205, W: 188) | 582 (M: 408, W: 174) | ↓ | ↓↓ | NA |
Watanabe et al. (26) | 1977–83 | Men and women | Not specified | 203 (M: 110, W: 93) | 203 (M: 110, W: 93) | — | — | NA |
Tajima and Tominaga (35) | 1981–83 | Men and women | 40–70 years | 93 (M: 52, W: 41) | 186 (M: 111, W: 75) | — | — | NA |
Kato et al. (41) | 1986–90 | Men and women | Not specified | 223 (M: 139, W: 84) | 578 (M: 377, W: 201) | ↓↓↓ | ↓ | NA |
Hoshiyama et al. (42) | 1984–90 | Men and women | 40–69 years | 181 (M: 98, W: 83) | 653 (M: 343, W: 310) | ↓↓↓ | — | NA |
Kotake et al. (27) | 1992–94 | Men and women | Not specified | 363 (M: 214, W: 149) | 363 (M: 214, W: 149) | — | ↑↑ | NA |
Nishi et al. (28) | 1987–90 | Men and women | Not specified | 330 (M: 171, W: 159) | 660 (M: 342, W: 318) | — | — | NA |
Inoue et al. (43) | 1990–95 | Men and women | 40+ years | 628 (M: 386, W: 242) | 21128 (M: 6307, W: 14821) | — | ↓↓↓ | NA |
Wang et al. (29) | 2000–03 | Men and women | 20–74 years | 816 (%men: 60) | 815 (%men: 62) | — | — | — |
NA, not available
a↑↑↑ or ↓↓↓, strong; ↑↑ or ↓↓, moderate; ↑ or ↓, weak; —, no association (see text for more detailed definition)
bM, men; W, women.
References . | Study period . | Study subjects . | Magnitude of associationa . | |||||
---|---|---|---|---|---|---|---|---|
Sex . | Age range . | Number of casesb . | Number of controlsb . | Colon . | Rectum . | Colorectum . | ||
Kondo (34) | 1967–73 | Men and women | Not specified | 393 (M: 205, W: 188) | 582 (M: 408, W: 174) | ↓ | ↓↓ | NA |
Watanabe et al. (26) | 1977–83 | Men and women | Not specified | 203 (M: 110, W: 93) | 203 (M: 110, W: 93) | — | — | NA |
Tajima and Tominaga (35) | 1981–83 | Men and women | 40–70 years | 93 (M: 52, W: 41) | 186 (M: 111, W: 75) | — | — | NA |
Kato et al. (41) | 1986–90 | Men and women | Not specified | 223 (M: 139, W: 84) | 578 (M: 377, W: 201) | ↓↓↓ | ↓ | NA |
Hoshiyama et al. (42) | 1984–90 | Men and women | 40–69 years | 181 (M: 98, W: 83) | 653 (M: 343, W: 310) | ↓↓↓ | — | NA |
Kotake et al. (27) | 1992–94 | Men and women | Not specified | 363 (M: 214, W: 149) | 363 (M: 214, W: 149) | — | ↑↑ | NA |
Nishi et al. (28) | 1987–90 | Men and women | Not specified | 330 (M: 171, W: 159) | 660 (M: 342, W: 318) | — | — | NA |
Inoue et al. (43) | 1990–95 | Men and women | 40+ years | 628 (M: 386, W: 242) | 21128 (M: 6307, W: 14821) | — | ↓↓↓ | NA |
Wang et al. (29) | 2000–03 | Men and women | 20–74 years | 816 (%men: 60) | 815 (%men: 62) | — | — | — |
References . | Study period . | Study subjects . | Magnitude of associationa . | |||||
---|---|---|---|---|---|---|---|---|
Sex . | Age range . | Number of casesb . | Number of controlsb . | Colon . | Rectum . | Colorectum . | ||
Kondo (34) | 1967–73 | Men and women | Not specified | 393 (M: 205, W: 188) | 582 (M: 408, W: 174) | ↓ | ↓↓ | NA |
Watanabe et al. (26) | 1977–83 | Men and women | Not specified | 203 (M: 110, W: 93) | 203 (M: 110, W: 93) | — | — | NA |
Tajima and Tominaga (35) | 1981–83 | Men and women | 40–70 years | 93 (M: 52, W: 41) | 186 (M: 111, W: 75) | — | — | NA |
Kato et al. (41) | 1986–90 | Men and women | Not specified | 223 (M: 139, W: 84) | 578 (M: 377, W: 201) | ↓↓↓ | ↓ | NA |
Hoshiyama et al. (42) | 1984–90 | Men and women | 40–69 years | 181 (M: 98, W: 83) | 653 (M: 343, W: 310) | ↓↓↓ | — | NA |
Kotake et al. (27) | 1992–94 | Men and women | Not specified | 363 (M: 214, W: 149) | 363 (M: 214, W: 149) | — | ↑↑ | NA |
Nishi et al. (28) | 1987–90 | Men and women | Not specified | 330 (M: 171, W: 159) | 660 (M: 342, W: 318) | — | — | NA |
Inoue et al. (43) | 1990–95 | Men and women | 40+ years | 628 (M: 386, W: 242) | 21128 (M: 6307, W: 14821) | — | ↓↓↓ | NA |
Wang et al. (29) | 2000–03 | Men and women | 20–74 years | 816 (%men: 60) | 815 (%men: 62) | — | — | — |
NA, not available
a↑↑↑ or ↓↓↓, strong; ↑↑ or ↓↓, moderate; ↑ or ↓, weak; —, no association (see text for more detailed definition)
bM, men; W, women.
We performed meta-analysis among five cohort studies (24,25,33,39,40) and seven case–control studies (26–29,41–43). In cohort studies, coffee drinking was not appreciably associated with risk of colorectal cancer (RR: 0.95, 95% CI: 0.77–1.17; I2 = 28.1%), colon cancer (RR: 0.98, 95% CI: 0.70–1.36; I2 = 55.5%) and rectal cancer (RR: 0.99, 95% Cl: 0.72–1.37; I2 = 0.0%) (Fig. 1). In a sensitivity analysis that excluded data for women in the study by Yamada et al. (25), which had only five cases in the highest category of coffee consumption, result was materially unchanged; RR (95% CI) for colorectal cancer was 0.92 (0.75–1.13). In another sensitivity analysis excluding the study by Khan et al. (24), which used mortality data, adopted unique cutoff of coffee consumption (weekly or daily versus less than weekly) and adjusted for fewer covariates than other cohort studies, similar result was obtained; the pooled RR (95% CI) for colorectal cancer was 0.98 (0.80–1.22). In sex-specific analysis among cohort studies, coffee drinking was not significantly associated with colorectal cancer in both sexes; the pooled RR (95% CI) of colorectal, colon and rectal cancer in men were 1.05 (0.83–1.31), 1.12 (0.82–1.52) and 1.01 (0.70–1.47), respectively; the corresponding values in women were 0.82 (0.55–1.23), 0.83 (0.43–1.60) and 0.94 (0.50–1.76). In case–control studies, high coffee consumption was associated with a 22% SS decrease in the risk of colorectal cancer (95% CI: 0.65–0.95; I2 = 51.6%). Pooled ORs were 0.75 (95% CI: 0.59–0.94; I2 = 43.4%) and 0.83 (95% CI: 0.59–1.15; I2 = 61.2%) for colon cancer and rectal cancer, respectively (Fig. 2). No evidence of publication bias was suggested for both cohort (P = 0.3) and case–control studies (P = 0.3).
Discussion
In the present systematic review of Japanese studies, we identified five cohort studies and nine case–control studies that reported the association between coffee consumption and colorectal cancer. Findings of cohort studies are inconsistent, whereas those of case–control studies suggest an inverse association. In meta-analysis, summarized RR of cohort studies was nearly equal to unity, whereas summarized OR of case–control studies was statistically significantly decreased for colorectal cancer and colon cancer.
The null finding among cohort studies in Japan is consistent with a recent systematic review and meta-analysis of cohort studies (RR = 0.94; 95% CI = 0.88–1.01) (16), which included eight studies from Europe, four from America and four from Asia including Japan (three studies). That meta-analysis did not detect any significant variation in association by study region (16). Some other meta-analyses (18,19) and pooled analysis of prospective cohort studies (44) also reported no association between coffee consumption and colorectal cancer. In contrast, another meta-analysis of cohort studies reported a significant inverse association (20). The difference may be partly ascribed to the fact that the latter meta-analysis with a significant finding (20) included two earlier studies (45,46), of which one large study with a greater weight (46) showed a significant risk reduction associated with coffee drinking.
We found a significant 22% risk reduction in the meta-analysis of case–control studies in Japan. This finding is generally consistent with two recent meta-analyses of case–control studies (16,17) including three Japanese studies (41–43), and with an earlier meta-analysis (18) including one Japanese study (41). Due to limitations inherent to case–control study, however, findings from this type of study should be interpreted with caution. For instance, patients with colorectal cancer may tend to underreport coffee consumption in the past as a result of decreased dietary intake after cancer diagnosis. Additionally, researchers of most case–control studies reviewed here selected controls from among outpatients or participants of health checkup or screening, raising a possibility of selection bias.
An apparent strength of the present meta-analysis is the inclusion of all relevant studies among Japanese populations, encompassing recent well-designed cohort studies. Several limitations of this study must be considered. First, risk factors for colorectal cancer including obesity, smoking, alcohol drinking, physical inactivity and high intake of red and processed meat may confound the association in each study. We confirmed that the result was materially unchanged when we repeated the analysis among cohort studies with adjustment for most of these potential confounders. Second, each study estimated coffee consumption using data on drinking frequency only (not on the size of the cup), leading to a misclassification of coffee consumption. We assumed, however, that the size of standard cup may not largely differ among the Japanese studies reviewed here. Finally, the definition of the lowest and highest categories for coffee consumption varied across studies, and this would a source of heterogeneity of results. However, we confirmed that the exclusion of one study (24) that used a unique cutoff of coffee consumption did not appreciably influence the pooled results.
Conclusion
From results of the present systematic review and meta-analysis, we conclude that the evidence is insufficient to support that coffee drinking increases or decreases colorectal cancer risk among the Japanese population.
Supplementary data
Supplementary data are available at http://www.jjco.oxfordjournals.org.
Funding
This study was supported by the National Cancer Center Research and Development Fund.
Conflict of interest statement
None declared.
References
Appendix
Research group members: Shizuka Sasazuki [principal investigator], Shoichiro Tsugane, Manami Inoue, Motoki Iwasaki, Tetsuya Otani [until 2006], Norie Sawada [since 2007], Taichi Shimazu [since 2007], Taiki Yamaji [since 2007] (National Cancer Center, Tokyo), Ichiro Tsuji [since 2004], Yoshitaka Tsubono [in 2003] (Tohoku University, Sendai); Yoshikazu Nishino [until 2006] (Miyagi Cancer Research Institute, Natori); Akiko Tamakoshi [since 2010] (Hokkaido University, Sapporo); Keitaro Matsuo [until 2010, since 2012] (Kyushu University, Fukuoka), Hidemi Ito [since 2010 until 2011] (Aichi Cancer Center, Nagoya); Kenji Wakai (Nagoya University, Nagoya); Chisato Nagata (Gifu University, Gifu); Tetsuya Mizoue (National Center for Global Health and Medicine, Tokyo); Keitaro Tanaka (Saga University, Saga)