Epidemiologic Evidence on Coffee and Cancer

Abstract

Coffee consumption is a major and frequent dietary exposure in diverse cultures around the globe whose safety has been questioned. A substantial body of epidemiologic evidence, consisting of over 500 papers relating the consumption of coffee to cancer of various sites, has accumulated to date. Numerous individual, site-specific meta analyses have been undertaken at various times. However, there is no comprehensive, up-to-date overview of the entirety of the knowledge base. To address this need, this review summarized the findings of the meta analyses and recent papers on site-specific human cancers among coffee consumers. For hepatocellular and endometrial cancers, there appears to be a strong and consistent protective association; for colorectal cancer, the direction of association is borderline protective. There appears to be no association with breast, pancreatic, kidney, ovarian, prostate, or gastric cancer. Risk of bladder cancer appears to be associated with heavy coffee consumption in some populations and among men. The associations with childhood leukemia and mother's consumption of coffee were ambiguous—with some suggestion of risk at high levels of daily consumption.

INTRODUCTION

Coffee, as one of the most widely consumed beverages in the world, can potentially impact the etiology of cancer of various sites along multiple pathways, ranging from carcinogenesis to cellular apoptosis. The significant volume of epidemiologic literature that has explored associations between coffee consumption and cancer occurrence at various sites has not been recently compiled and reviewed. A recent comprehensive analysis of diet and cancer, the World Cancer Research Report (WCRF), addressed only the relationships between coffee and risk of pancreatic and kidney cancer (1). In fact, epidemiologic studies have been published relating coffee consumption to cancers of 11 different organ sites (see Table 1) in addition to reviewing the total impact on all causes of mortality as described herein.

TABLE 1 Overview of relative risks of coffee drinking in 11 cancer sites ^a

Author	Countries	Study Design	N Controls	n Cases	Follow-up	Point Estimate	CIs	Dose
Pancreatic Cancer
Jick, Dinan (9)	United States	Hospital-based CC	156	78		0.7, 0.5	Not reported	1–5 cups/day; 6+ cups/day
Lin and Kessler(10)	Japan	Cohort	110,792	225 deaths	8 yr	Men 0.59, Women 0.94	0.26–1.33, 0.44–2.01	1 cups/day
MacMahon, Yen, Trichopoulos, Warren, and Nardi (11)	United States	CC	644	369		1.8, 2.7	1.0–3.0, 1.6–4.7	2 cups/day; 3+ cups/day
WCRF (1)	Meta-analysis	18 cohort, 37 CC, 11 ecological				Cohort 1.00, CC 1.04	0.94–1.07, 1.01–1.07
Luo, Inoue, Iwasaki, Sasazuki, Otane, et al. (14)	Japan	Population-based cohort	102,137	233	11 yr	0.8	0.6–1.2	1–2 cups/day
Renal Cell Cancer
WCRF (1)		5 cohort studies, 18 CC				Cohort women: 0.87; men: 0.15; CC: 0.99	0.66–1.16, 0.02–1.16, 0.96–1.01	Per cup
Lee, Giovannucci, Smith-Warner, Spiegelman, Willett, et al. (16)	United States	2 cohorts	88,759 women, 47,828 men	248	20 yr 14 yr	0.87	0.62–1.21	Highest vs. lowest quartile
Bravi, Bosetti, Scotti, Talamini, Montella, et al. (17)	Italy	CC	1,534	767		0.98	0.72–1.32	QV vs. QI
Lee, Hunter, Spiegelman, Adami, Bernstein, et al. (18)	Mixed	13 prospective	530,469 women, 244,483 men	1,478	7–20 yr	0.84	0.67–1.05	3+ cups/day
Colorectal Cancer
Tavani and La Vecchia (19)	Japan, France, Italy, Denmark, Netherlands, Norway, Sweden, Finland, Switzerland, Argentina and the United States	Meta-analysis: 5 cohorts, 12 CC	147,227	1,756 colon, 759 rectal, 8,713 colorectal, 8,058 colon, 4,865 rectal, 604 colorectal		0.76	0.66–0.89	High vs. low
Oba,	Japan	Cohort	138,940 men	29	8 yr	0.81	0.46–1.42	1+
Shimizu,			16,327 women	45		0.42	0.22–0.85	cups/day
Nagata, Shimizu, Kametani, et al. (20)
Lee, Inoue,	Japan	Population-	96,162	1,163	10 yr	1.10	0.82–1.47	3+
Otani,		based cohort	46,023 men
Iwasaki,			50,139 women			0.44	0.19–1.04	cups/day
Sasazuki. et al. (21)
Breast Cancer
Hirvonen, Mennen, de Bree, Castetbon, Galan, et al. (22)	French	Cohort	4,396 female	95	6.6 yr	1.07, 1.10	0.64–1.79, 0.66–1.84	112–252 ml/day; >252 ml/day
Michels, Holmberg, Bergkvist, and Wolk (23)	Swedish	Cohort	508,036 women	1,271	9.5 yr	0.94	0.75–1.28	4+ cups/day
Tavani, Pregnolato, La Veccia, Favero, and Franceschi (24)	Italy	Hospital-based CC	5,984	5,504		0.96	0.83–1.11	>4 cups/day
Baker, Beehler, Sawant, Jayaprakash, McCann, et al. (25)	US	Hospital based CC, premenopausal	1,895	1,932		0.62	0.39–0.98	4+ cups/day
Nkondjock, Ghadirian, Kotsopoulos, Lubinski, Lynch, et al. (26)	4 countries: United States, Canada, Poland, Israel	Matched CC analysis	1,690 BRCa+ women	1,690		0.90, 0.75, 0.31	0.72–1.12, 0.47–1.19, 0.13–0.71	1–3 cups/day; 4–5 cups/day; 6+ cups/day
Endometrial Cancer
Hirose, Niwa, Wakai, Matsuo, Nakanishi, et al. (28)	Japanese	Hospital-based CC	12,425	229		0.64, 0.41	0.43–0.94, 0.19–0.87	1–2 cups; 3+ cups/day
Terry, Vainio, Wolk, and Weiderpass (29)	Swedish	CC	2,887 postmenopausal women	709		0.7	0.5–1.0	4th quartile
Bravi, Scotti, Bosetti, Zucchetto, Talamini, et al. (30)	Italy	Hospital-based CC	908	454		0.50	0.29–0.89	QV vs. QI
Shimazu, Inoue, Sasazuki, Iwasaki, Kurahashi, et al. (31)	Japan	Population-based cohort	53,724	117	15 yr	0.38	0.16–0.91	3 cups/day
Koizumi, Nakaya, Okamura, Sato, Shimazu, et al. (32)	Japan	CC	214	107		0.4	0.2–0.9	Upper tertile
Ovarian Cancer
Steevens, Schouten, Verhage, Goldbohm, and van den Brandt (34)	Netherlands	Case-cohort and meta-analysis	2,083	280	13.3 yr	1.08, 1.32, 1.15	0.75–1.57, 0.99–1.77, 0.89–1.47	5+ cups/day
Hirose, Niwa, Wakai, Matsuo, Nakanishi, et al. (28)	Japan	CC	12,425	166		1.33	0.68–2.60	3+ cups/day
Silvera, Jain, Howe, Miller, and Rohan (33)	Canada	Cohort	49,613	264	16.4 yr	1.62	0.95–1.77	>4 cups/day
Baker, Boakye, McCann, Beehler, Rodabaugh, et al. (35)	United States	Hospital-based CC	868	414		1.05	0.73–1.52	4+ cups/day
Hepatocellular Cancer
Larsson, and Wolk (40)	13 cohorts	Meta-analysis	530,469 women, 244,483 men	1,478	7–20 yr	0.84	0.67–1.05	3+ cups/day
Inoue Yoshimi, Sobue, and Tsugane (41)	Japan	Population-based cohort	43,109 men and 47,343 women	250 men, 84 women	10 yr	HR = 0.49, 0.52, 0.48, 0.24	0.36–0.66, 0.38–0.73, 0.28–0.83, 0.08–0.77	Nondrinkers; 1–2 cups/day; 3–4 cups/day; 5
Kurozawa, Ogimoto, Shibata, Nose, Yoshimura, et al. (42)	Japan	Cohort	46,399 male and 64,289 female	258	9 yr	HR = 0.50	0.31–0.79	1+ cups/day
Tanaka, Hara, Sakamoto, Higaki, Mizuta, et al. (43)	Japan	CC	1,308 community controls, 275 hospital controls, 381 patients	209		0.33, 0.22	0.22–0.48, 0.11–0.43	Occasional; 3 cups/day
Ohishi, Fujiwara, Cologne, Suzuki, Akahoshi, et al. (44)	Japan	Nested CC	644	224		0.40	0.16–1.02	Daily
Wakai, Kurozawa, Shibata, Fujita, Kotani, et al. (45)	Japan	Nested CC	3,444	96		0.72, 0.49	0.45–1.15 0.25–0.96	<1 cups/day; 1 cups/day
Gelatti, Covolo, Franceschini, Pirali, Tagger, et al. (46)	Northern Italy	Hospital-based CC	500	250		0.8, 0.4, 0.3	0.4–1.3, 0.2–0.8, 0.1–0.7	1–2 cups/day; 3–4 cups/day; 5+ cups/day
Montella, Polesel, La Vecchia, Dal Maso, Crispo, et al. (47)	Italy	Hospital-based CC	412 controls	185		0.4	0.2–1.1	28cups/wk vs. <14 cups/wk
Larsson and Wolk (40)	Meta-analysis	Meta-analysis: 4 cohort, 5 CC	239,146	2,260		0.57	0.49–0.67	For increase +2 cups/day
Bravi, Bosetti, Tavani, Bagnardi, Gallus, et al. (39)	6 CC, 4 cohort: Japan, & Europe	Meta-analysis	CC: 1,551; cohort 709			0.54, 0.64, 0.70, 0.45	0.38–0.76, 0.56–0.74, 0.57–0.85, 0.38–0.53	Drinkers vs. non, low, moderate, high
Bladder Cancer
Zeegers, Tan, Goldbohm, and van den Brandt (48)	Meta-analysis	16 studies on men, 12 on women	58,279	491	6.3 yr	0.98 men, 0.44 women	0.68–1.40, 0.22–0.86	5–6 cups/day; referent is 4–5 cups/day
Villanueva, Cantor, King, Jaakkola, Cordier, et al. (49)	6 CCs	Pooled analysis, 6 CC	5,150	2,729		1.11 men, 0.93 women	0.98–1.25, 0.75–1.15	Per cup
De Stefani, Boffetta, Deneo-Pellegrini, Correa, Ronco, et al. (50)	Uruguay	CC	501	255		1.6	0.8–3.1	1–6 cups/day
Ugnat, Luo, Semenciw, and Mao (51)	Canada	Population-based CC	1,099	549		1.77, 4+ cups/day	1.11–2.82	4+ cups/day
Gastric Cancer
Botelho, Lunet, and Barros (53)	23:5 United States, 5; Japan, 6; Europe, 2; South America	Meta-analysis				1.26; 0.97 CC; 1.02 cohort; 0.90 Pop CC; 0.97, hospital-based CC; 0.97 United States; 0.98 Japan; 0.64 Europe	1.02–1.57, 0.86–1.09, 0.76–1.37, 0.70–1.15, 0.83–1.13, 0.82–1.14, 0.81–1.17, 0.47–0.86	Highest vs. lowest
Larsson,	Swedish	Population-	61,433 women	160	15.7 yr	1.49, 1.86	0.97–2.27,	2–3 cups/day
Giovannucci,		based cohort					1.07–3.25	4+ cups/day
and Wolk (54)
Childhood Leukemia
Peters, Preston-Martin, London, Bowman, Buckley, et al. (55)	United States	Population-based CC	232	232		Not reported—no association with maternal coffee
Petridou, Trichopoulos, Kalapothaki, Pourtsidis, Kogevinas, et al. (56)	Greece	CC	306	153		Not reported—“no substantial or significant elevation”
Menegaux, Steffen, Bellec, Baruchel, Lescoeur, et al (57)	France	Multicenter CC	288	280		No effect; 2.4, 3.1	1.3–4.7, 1.0–9.5	<3 cups/day; 4–8 cups/day; >8 cups/day
Menegaux, Ripert, Hemon, and Clavel (58)	France	Population-based CC	567	472 all		1.4, 1.9	0.9–2.3, 1.0–3.5	<3 cups/day vs. >3 cups/day
Prostate
Hsing, McLaughlin, Schuman, Bjelke, Gridley, et al. (61)	United States	Cohort	17,633	149	20 yr	0.8	0.6–1.2	3–4 cups/day vs. <3
Jain, Hislop, Howe, Burch, and Ghadirian (63)	Canada	CC	637	617		0.97	0.65–1.44
Hsieh, Thanos, Mitropoulos, Deliveliotis, Mantzoros, et al. (64)	Greece	CC	246	320			Probability test, 0.27	3cups/day vs. 0
Slattery and West (65)	United States	CC	685	362		1.09	0.75–1.60	>20 c/wk
Sharpe and Siemiatycki (66)	Canada	CC	621	399		1.1	0.6–2.0	Daily
Ellison (69)	Canada	Retrospective cohort	3,400	145		1.4	0.84–2.32	>250 ml/day
^aAbbreviations are as follows: CC, case-control study, CI, confidence interval; WCRF, World Cancer Research Report; HR, hazard ratio; QV, quintile five; QI, quintile one.

Total Mortality

To insure that protective effects for one disease are not overridden by detrimental effects for another, a first look at total mortality is prudent. Studies in Norway (2), Sweden (3), Croatia (4), and the United States (5, 6) have reported on this. At levels of consumption up to 6 cups/day, no unfavorable associations were seen, with a significant protective association reported among women in Iowa. In Evans County, Georgia, across 4 race and gender groups, no association between high and low coffee consumption and mortality was seen. In Croatia, a significant 37% reduction in all cause death was seen among women who drank 1 to 2 cups/day but not among men. In Norway, no significant positive associations were found with mortality; whereas in Sweden, there was an inverse relationship between mortality from all causes and coffee consumption. When the population was restricted to patients with Type 2 diabetes, a negative dose-response relationship between cups of coffee and total mortality was seen up to 6 cups/day, when it leveled out at a hazards ratio of 0.70 (7). A significant amount of literature exists on relationships between coffee consumption and human cancer occurrence at 11 organ sites.

Pancreatic Cancer

Coffee, first associated with cancer in the English medical literature in 1970 (8), became a major topic in 1981 with publications in the New England Journal of Medicine (9) and the Journal of the American Medical Association (10) suggesting that coffee causes pancreatic cancer. MacMahon et al. (11) reported odd ratios with various levels of coffee consumption of 2.6 to 2.8 for men and 2.0 to 3.3 for women. All drinkers’ risks were elevated equally for all levels of consumption greater than zero. The controversy triggered a review of coffee by the International Agency on Research on Cancer (IARC) that was published in 1990 (12). This report found 1) limited evidence in humans that coffee drinking is carcinogenic in the urinary bladder; 2) evidence suggesting lack of carcinogenicity of coffee drinking in the human female breast and in the large bowel; and 3) inadequate evidence in humans that coffee drinking is carcinogenic in the pancreas, ovary, and other body sites. IARC's overall evaluation determined that coffee is possibly carcinogenic to the human urinary bladder (IARC Category Group 2B) and that there is evidence of an inverse relationship between coffee drinking and cancer of the large bowel. Coffee drinking was unclassifiable as to its carcinogenicity to other organs.

A number of theories regarding the validity of the initial reports leading to the belief in a causal relationship with pancreatic cancer have been proposed. These included the possibility that a spurious association with pancreatic cancer may have been attributable to polystyrene leaching from disposable coffee cups, pesticides in the coffee bean, or antimony (stibium) contamination during shipping as the underlying toxins. Alternatively, there have been methodology concerns about inappropriate use of hospital-based patients with gastrointestinal disease as controls, as they are likely to reduce their coffee intake (13). It was also suggested that disturbed glucose tolerance due to preclinical pancreatitis resulted in increased thirst and increase coffee consumption. Both of these would lead to false conclusions because of reverse causation.

Since the IARC report, the WCRF report (1) examined 18 cohort studies, 37 case-control studies, and 11 ecological studies on human pancreatic cancer and coffee consumption. The authors estimated a summary relative risk of 1.00 for cohort studies and 1.04 for case-control studies. They concluded that a consistent homogenous lack of association is present and therefore “it is unlikely that coffee has any substantial effect on the risk of pancreatic cancer” (1). Since the WCRF report, one additional cohort study has been published on this topic. Luo et al. (14) studied the association between the drinking coffee and the risk of pancreatic cancer in a large population-based cohort study in Japan. Among 102,137 participants followed for an average of 11 yr in which 233 incident cases of pancreatic cancer were identified, there was no increased risk of pancreatic cancer with coffee intake. A reduced risk was apparent among men who drank at least 3 cups of coffee per day compared with those who did not drink any or only rarely drank coffee (14). Thus, the chapter on coffee and pancreatic cancer, based on a wealth of evidence, seems now to be closed.

Renal Cell Cancer

With due concern for the rising incidence in renal cell cancer in North America and northern Europe, particularly among African Americans, and its high mortality rate, the search for causes of renal cell cancer has addressed a possible link to diet (15). The etiology of renal cell cancer links risk to both cigarette smoking and obesity or high relative weight. An inverse association with intake of fruits and vegetables has been one of the few consistent dietary findings. No relation with coffee has been reported in the 26 studies on the topic. The WCRF report found 5 cohort studies, 18 case-control studies, and one ecologic study that have provided substantial and consistent evidence of a lack of association between coffee and kidney cancer (1). Three publications not included in their report support this lack of causation. Among the participants in the Nurses’ Health Study and the Health Professionals Follow-up Study, no association was seen between total fluid, beverage, or coffee intakes and risk of renal cell cancer (16). An Italian case-control study in a population of frequent coffee drinkers compared 767 patients less than 79 yr of age with incident, histologically confirmed, renal cell cancers and 1,534 controls and found no relationship to coffee consumption (17). A larger pooled analysis (18) of 13 prospective studies (530,469 women and 244,483 men) involving a total of 1,478 incident renal cell cancer cases arising during a follow-up of 7 to 20 yr found coffee consumption not to be associated with a risk of renal cell cancer (RR for 3+ cups/day vs. less than 1 cup/day = 0.84; 95% CI = 0.67–1.05). Thus, there is no indication that coffee consumption increases renal cell cancer risk (18).

Colorectal Cancer

A meta-analysis of colorectal cancer and coffee consumption was completed and published in 2004 (19). Based on a literature search of MEDLINE from 1990 to 2003, 3 cohorts and 15 case-control studies were identified from Japan, France, Italy, Denmark, The Netherlands, Norway, Sweden, Finland, Switzerland, Argentina, and the United States. Together these include the experiences from over 8,713 cases among 147,227 subjects. The authors found significant heterogeneity of effect by study design. The 3 cohort studies showed no association. Based on the case-control studies, there was a potentially protective effect of coffee on colon cancer, relating to a summary point estimate of a 24% lower risk of colon cancer per cup consumed across coffee drinkers in all of these countries. However, there was no consistent dose response seen within studies, suggesting either other differences between coffee consumers and nonconsumers or possibly a threshold effect.

Two studies conducted since then, both from Japan, found 60% reductions in risk of colorectal cancer with coffee consumption among women only (20, 21). The studies were both cohorts, one of 31,550 people and the other of 96,000 people. In total, there has been strong evidence of no harmful effect. Due to the heterogeneity of effect and the lack of dose-response effect, there is very weak evidence of a preventive effect.

Breast Cancer

No effect has been noted between breast cancer incidence and coffee consumption in the few publications on this topic among postmenopausal women. One example of these is the recent prospective study of 4,396 French female participants in the French Supplémentation en Vitamines et Minéraux Antioxydants (SUVIMAX) study in which 95 breast cancers were identified through clinical examinations over the 6.6 yr of follow-up (22). Another is the Swedish Mammography Screening Cohort, a large population-based, prospective, cohort study comprising 59,036 women aged 40 to 76 yr whose coffee consumption is among the highest per capita in the world. Among 508,267 person years of follow-up, 1,271 cases of invasive breast cancer arose (23). A third study conducted in Italy also observed no relationship between coffee intake and the risk of breast cancer (24). These studies had recruited largely postmenopausal women. Among premenopausal women, consumption of regular coffee was associated with statistically significant linear declines in breast cancer risk. Consumers of 4 or more cups per day experienced a 40% risk reduction [odds ratio (OR) = 0.62, 95% CI = 0.39–0.98] (25).

More than one group has assessed the association between coffee consumption and breast cancer risk among high-risk women carrying BRCA mutations. One of these is a matched case-control analysis of 1,690 women with a BRCA1 or BRCA2 mutation from 40 centers in 4 countries; lifetime coffee consumption was protective in these high-risk women. Odds ratios were 0.90 (95% CI = 0.72–1.12), 0.75 (95% CI = 0.47–1.19), and 0.31 (95% CI = 0.13–0.71; P trend = 0.02) for those who drank 1 to 3, 4 to 5, and 6 or more cups of coffee, respectively, as compared to nondrinkers. The effect was limited to the consumption of caffeinated coffee (26).

These findings motivated a closer look at who might be benefiting most from a preventive effect of coffee. Under the assumption that breast cancer risk is modulated by CYP1A2, the interaction with coffee and a common A to C polymorphism in the CYP1A2 gene associated with impaired caffeine metabolism was studied (27). Women with at least one variant C allele (AC or CC) who consumed coffee had a 64% reduction in breast cancer risk compared with women who never consumed coffee (OR = 0.36, 95% CI = 0.18–0.73). A protective effect of coffee consumption was not observed among women with the CYP1A2 AA genotype (OR = 0.93, 95% CI = 0.49–1.77), suggesting that caffeine may protect against breast cancer in this subset of women carrying the variant C allele of CYP1A2.

Endometrial Cancer

The 5 observational epidemiologic studies published on the relationship between endometrial cancer occurrence and prior coffee consumption have yielded homogenous results. One of these (28) is a hospital-based, case-control study of 229 endometrial cancer cases as compared with 12,425 Japanese control women. In this population in which coffee consumption is rare and rather recent, a statistically significant inverse association between risk of endometrial cancer and coffee consumption was noted. Compared to nondrinkers, the OR of daily drinking of 1 to 2 cups and 3 or more cups per day for endometrial cancer were 0.64 (95% CI = 0.43–0.94) and 0.41 (95% CI = 0.19–0.87), respectively, and the linear trend was statistically significant (P< 0.01) (28). A large, case-control study of Swedish-born postmenopausal women aged 50 to 74 (709 cases and 2,887 controls) with much higher coffee consumption patterns also found coffee to be associated with a reduction in risk of endometrial cancer (29). A hospital-based case control study from Italy, with 454 cases and a 2:1 control to case ratio, found the risk to be half at the upper quintiles of intake as compared to the lowest quintile (30). Two additional studies from Japan were published in 2008; one a population-based cohort of 53,724 women, 117 of which developed endometrial cancer during 15 years of follow-up (31), and the other a case control study with a similar number of cases (32)—both showed relative risks of 0.4 for those consuming 3 or more cups per day of coffee (cohort) or those in the upper tertile of consumption.

Ovarian Cancer

Four studies, one each in the United States, Canada, The Netherlands, and Japan, have examined relationships between coffee and ovarian cancer, including a meta-analysis of recent data. None of these has shown significant associations between coffee consumption and ovarian cancer incidence (25, 28, 33, 34). One of these is the previously mentioned Aichi Cancer Center study that reported no clear association between intake of coffee and risk among Japanese women among the 166 ovarian cancer cases as compared with 12,425 women free of cancer (28). A hospital-based, case-control design employed among 414 women in the United States with primary epithelial ovarian, fallopian, or peritoneal cancer and 868 age- and region-matched control women with nonneoplastic conditions also showed no association between any level of regular coffee consumption and risk of ovarian cancer (35). After a mean 16.4 yr of follow-up among 49,613 Canadian women enrolled in the National Breast Screening Study, 264 incidences of ovarian cancer occurred. Ovarian cancer was also not significantly associated with coffee intake (33).

A cohort study conducted (34) in the Netherlands also observed no significant association with coffee and ovarian cancer, the multivariable rate ratios being 1.04 (95% CI = 0.97–1.12) per cup per day among the 280 ovarian cancer cases and cohort comparisons. These authors also conducted a meta-analysis and found no significant associations between ovarian cancer occurrence and coffee consumption (34) across the various studies from various populations and differing consumption patterns. Thus, the findings of a null effect for this outcome are quite compelling.

Hepatocellular Cancer

Coffee consumption has been associated with lower serum liver enzyme levels and a reduced risk of liver cirrhosis in mice and man (36, 37, 38). Indication of a protective effect has also been seen in the cohort and case-control studies of coffee and hepatocellular carcinoma. These findings have been confirmed in the two independent meta-analyses published to this effect (39, 40).

Two of the studies contributing to this topic are cohorts and three are case-control studies from Japan. All five suggest a protective effect. One is a population-based cohort study of 43,109 men and 47,343 women within which 250 men and 84 women with hepatocellular carcinoma were identified during 10 yr of follow-up (41). Subjects in this study who consumed coffee on a daily or almost daily basis had a 50% lower hepatocellular carcinoma risk than those who almost never drank coffee [hazards ratio (HR) = 0.49, 95% CI = 0.36–0.66], and risk decreased with the amount of coffee consumed. Compared with nondrinkers, the HR for 1 to 2 cups/day was 0.52 (95% CI = 0.38–0.73), for 3 to 4 cups per day was 0.48 (95% CI = 0.28–0.83), and for 5 cups/day was 0.24 (95% CI = 0.08–0.77). The risk of liver cancer among almost-never coffee drinkers in this population was 547.2 cases per 100,000 people over 10 yr compared with 214.6 cases per 100,000 people drinking coffee on a daily basis. Similar associations were observed when the analysis was restricted to hepatitis C virus-positive patients. Daily drinkers had a 43% lower risk as compared with nondrinkers (HR = 0.57, 95% CI = 0.37–0.86) and 54% lower risk compared to subjects with no history of chronic liver disease (HR = 0.45, 95% CI = 0.30–0.67).

The second Japanese study, the Japanese Collaborative Cohort Study for Evaluation of Cancer Risk, enrolled 46,399 male and 64,289 female cohort members aged 40 to 79. The HR of death due to hepatocellular carcinoma for drinkers of 1 or more cups of coffee per day, compared with noncoffee drinkers, was also 0.50 (95% CI 0.31–0.79) (42).

A third Japanese study, case control in design, addressed individuals with hepatitis: 209 incident hepatocellular carcinoma cases and 3 different controls (1,308 community controls, 275 hospital controls, and 381 patients with chronic liver disease without hepatocellular carcinoma), aged 40 to 79 yr were tested for plasma hepatitis B surface antigen and antibodies to hepatitis C virus (43). After adjustment for sex, age, heavy alcohol use, smoking status, and hepatitis virus markers, coffee use during the last 1 to 2 yr was associated with a decreased risk of hepatocellular carcinoma compared with any of the 3 control groups. Coffee consumption 10 yr prior to diagnosis or interview was associated with a significant decrease in risk. The adjusted ORs for occasional use, 1 to 2 cups/day, and 3 cups/day as compared with no use and community controls were 0.33, 0.27, and 0.22 (P trend < 0.001), respectively. Two nested case-controls studies from Japan also published relative risks of 0.40 and 0.49 with greater coffee consumption (44, 45).

At the other (higher) end of the coffee consumption spectrum, two case control studies conducted in Italy have reported consistent findings. A hospital-based case-control study conducted in an area of northern Italy found coffee consumption associated with a decreasing risk of hepatocellular carcinoma. As compared to nondrinking subjects, the ORs were 0.8 (95% CI = 0.4–1.3) for 1 to 2 cups/day, 0.4 (95% CI = 0.2–0.8) for 3 to 4 cups/day, and 0.3 (95% CI = 0.1–0.7) for 5 or more cups/day (46).

A hospital-based case-control study conducted in Italy in 1999–2002 included 185 incidents of histologically confirmed cases of hepatocellular carcinoma in subjects aged 43 to 84 yr and 412 controls admitted to the same hospitals’ networks for acute, nonneoplastic diseases. It found the risk of hepatocellular carcinoma decreased with increasing levels of consumption (OR = 0.4, 95% CI = 0.2–1.1 for  28 cups/wk) as compared to people who drank <14 cups/wk of coffee (47).

Most of these data were summarized in a meta-analysis of epidemiologic studies published between 1966 and February 2007 (40). It included 4 cohort and 5 case-control studies involving 2,260 cases and 239,146 noncases and reported an inverse relation between coffee consumption and risk of liver cancer in all studies, which and in 6 studies was statistically significant. Their analysis found an increase in consumption of 2 cups of coffee/day to be associated with a 43% reduced risk of liver cancer (RR = 0.57, 95% CI = 0.49–0.67). To examine whether this might be related to biases related to disease status, they stratified the analysis by history of liver disease and still found the summary RRs of liver cancer for an increase in consumption of 2 cups of coffee/day to be 0.69 (95% CI = 0.55–0.87) for persons without a history of liver disease.

A separate meta analysis conducted on many of the same studies, and published almost in parallel, reported summary relative risks of 0.54 (95% CI = 0.38–0.76) for case-control studies and 0.64 (95% CI = 0.56–0.74) for cohort studies for coffee drinkers versus nondrinkers (39). The overall summary RR for low or moderate coffee drinkers was 0.70 (95% CI = 0.57–0.85), and that for high drinkers was 0.45 (95% CI = 0.38–0.53). The risk per cup of coffee was decreased by 23% in case-control studies and 25% in cohort studies. The authors noted the concern that subjects with liver conditions may selectively reduce their coffee consumption but suggested that the consistency of an inverse relation between coffee drinking and hepatocellular carcinoma across study design and geographic areas weighs against this observed protective effect being due to bias or confounding.

Bladder Cancer

Coffee consumption and cancer of the urinary track was systematically reviewed in 2001 (48). The authors incorporated data on adjusted summary ORs from 16 studies on men and 12 on women. They found coffee unlikely to be associated with bladder cancer in women, whereas a 26% elevation of risk was seen in men (48). In 2006, a much smaller pooled analysis with a focus on fluids combined the data from 6 case-control studies of bladder cancer including 2,729 cases and 5,150 controls and reported increased bladder cancer risks with an intake of >5 cups of coffee daily vs. <5 cups daily and also an increase in risk from tap water alone (49). A case-control study (50) conducted in Uruguay and published in 2007, with 255 incident cases with transitional cell carcinoma of the bladder and 501 patients treated in the same hospitals, also showed both coffee and tea to be associated with bladder cancer risk (OR coffee drinking = 1.6, 95% CI = 1.2–2.3; OR tea drinking = 2.3, 95% CI = 1.5–3.4). These results were confirmed in a separate analysis of never smokers (51). One earlier study calculated the population attributable fraction (PAF) estimate in Canada to be 17% for heavy coffee consumption (51). A recent review without meta-analysis presented the specifics from 4 cohorts and 17 case control studies on the topic and concluded that some studies have shown an increase in risk with high consumption, but no dose-response relationship has been demonstrated (52).

Gastric Cancer

A systematic review and meta-analysis of the literature on the association between coffee consumption and gastric cancer pooled effects from 23 studies (53) showed no adverse association. The OR for the overall association between coffee and gastric cancer (highest vs. lowest category of exposure) was 0.97 (95% CI = 0.86–1.09) and was similar for cohort (OR = 1.02, 95% CI = 0.76–1.37) and case-control studies regardless of the comparison population (population-based OR = 0.90; 95% CI = 0.70–1.15; hospital-based OR = 0.97, 95% CI = 0.83–1.13). The OR was 1.26 (95% CI = 1.02–1.57) when considering 5 studies conducted in the United States, 0.97 (95% CI = 0.82–1.14) for the 5 Japanese studies, 0.98 (95% CI = 0.81–1.17) for the 6 studies from Europe, and 0.64 (95% CI = 0.47–0.86) for the 2 studies from South America. Since then, a population-based cohort study of 61,433 Swedish women with a mean follow-up of 15.7 yr was published (54). Based on 160 incident cases of stomach cancer, the multivariate HRs were 1.49 (95% CI = 0.97–2.27) for women who drank 2 to 3 cups/day and 1.86 (95% CI = 1.07–3.25) for those who drank 4 or more cups/day (P for trend = 0.01) as compared with nondrinkers. However, these findings appear to be outliers in the context of the other 23 studies.

Childhood Leukemia

In the search for contributing causes of childhood leukemia, studies have examined possible maternal dietary consumption during pregnancy. One of these (55), a population-based case control involving 232 cases and 232 control children from birth to age 10 yr in Los Angeles County, California, drew cases from a tumor registry and controls from friends and by random-digit dialing. No association with maternal coffee consumption was seen (55). Three other articles have been published on coffee and childhood leukemia, 2 of which are by the same author. One is a case-control study of childhood leukemia in Greece with 153 incident cases and 2 hospital controls per case. Significant elevations were not found with respect to coffee drinking during pregnancy (56). The second is a French, multicenter, case-control study (57) of childhood acute leukemia involving 280 incident cases and 288 hospitalized controls. Maternal coffee consumption during pregnancy was not found to be associated with childhood acute lymphoid leukemia for consumption levels greater than 3 cups/day. Higher consumption levels were associated with higher ORs (OR = 2.4, 95% CI = 1.3–4.7 for 4–8 cups/day and OR = 3.1 95% CI = 1.0–9.5 for >8 cups/day) (57). No association with acute nonlymphoid leukemia was noted in this population. However, the number of cases was quite low (n= 40). The third study is a French, population-based, case-control study that compared 472 cases of childhood acute leukemia and 567 population controls and found maternal coffee consumption not to be significantly related to acute leukemia overall (OR = 1.4, 95% CI = 0.9–2.3) (58). These few findings are heterogeneous by country but suggestive of a positive association at very high levels of coffee consumption.

Prostate

Initial interest in relationships between coffee consumption and prostate cancer was triggered by ecologic analyses that had suggested a positive correlation between country mean coffee drinking and prostate cancer rates (59, 60). These associations have not been confirmed in numerous subsequent studies, both cohort and case-control in design, in which age, socioeconomic status, and race were controlled. An example is from the Lutheran Brotherhood Cohort study of 17,633 White males age 35 and older followed for 20 yr that identified 149 fatal prostate cancer cases, which found coffee consumption unrelated to risk (61). This conclusion was also reached by study of the even larger group of 50,000 former students from Harvard University and the University of Pennsylvania in which student coffee consumption was unrelated to prostate cancer (62).

Remarkably consistent findings have been seen in the case control studies publishing results on the relationship between prostate cancer and coffee. One, conducted in 3 geographical areas of Canada, included 617 incident cases and 637 population controls and found coffee was not associated with prostate cancer (63). Another undertaken in Athens, Greece, involving 320 patients with histologically confirmed incident disease and 246 patients without history or symptoms of benign prostatic hyperplasia, treated in the same hospitals as the cases for minor diseases or conditions, found no evidence for a positive association between drinking of coffee and prostatic cancer (64). A third, conducted in Utah, included 362 newly diagnosed cases of prostate cancer and 685 age-matched controls and also found coffee consumption not to be associated with prostate cancer risk (65). Crossing the border into Canada, a case-control study set in Montreal investigated the relations between the consumption of coffee and the development of prostate cancer among 3,999 men with incident prostate cancer and 621 controls, aged 45 to 70 yr, and also found no association between the consumption of coffee and the development of prostate cancer (66). The only study (67) that reported an increase in risk involved participants in the 1970–1972 Nutrition Canada Survey (NCS), and was a dichotomous analysis of coffee drinking vs. nondrinking among 3,400 survey participants, 145 of which developed prostate cancer. Compared to those reporting no coffee consumption, men who averaged more than 250 ml/day experienced a 40% greater risk (95% CI = 0.84–2.32) (67). In summary, there appears to be a wealth of evidence pointing to no effect and no evidence of a dose-response effect.

DISCUSSION

Coffee consumption has aroused passions for centuries. Over the past 30 yr, it has increasingly become an exposure of interest in regard to cancer etiology. There is a large body of epidemiologic research (close to 400 published studies) on coffee and cancer of various sites, much of which has not been systematically reviewed. The WCRF have recently reviewed coffee consumption and cancer of two sites and have substantiated the IARC conclusions of 1990. Results have been largely consistent across study designs and between countries despite the inevitable measurement error in the amount of active product consumed. However, an additional 8 sites were not included in their analyses.

For most cancer sites, there is a significant amount of evidence showing no detrimental effect of consumption of up to 6 cups of coffee/day in relation to cancer occurrence. In fact, some of the evidence presented above suggests that coffee might prevent some cancers. These findings are substantiated by evidence of anticarcinogenic mechanisms through which coffee consumption might exert an anticarcinogenic effect in some organs. For example, reduction in cholesterol, bile acid, and neutral sterol secretion in the colon is a direct effect of coffee consumption (68, 69) as is increased colonic motility, which can reduce exposure of epithelium to carcinogens. Some of the components of coffee that have received attention are its caffeine (a purine alkaloid), cafestol (a diterpene), kahweol (another diterpene), and chlorogenic acid (a dietary phenol). Cafestol and kahweol, active ingredients in the coffee oil, decrease mutagenesis and tumorigenesis in animal models. Diterpenes, found in coffee, reduce genotoxicity of carcinogens and lower DNA adduct formation. Caffeic acid and chlorogenic acid are antioxidants and have been reported to decrease DNA methylation (70). More subtly, caffeine stimulates gSt and caffeine itself appears to be protective—affecting cell cycle, proliferation, and apoptosis (71, 72, 73).

The epidemiology of coffee and cancer is based on observational studies of self reported coffee consumption either at the time of diagnosis or prior to diagnosis. Since coffee was not administered randomly, examination of possible confounding variables and alternative hypotheses is necessary. One alternative hypothesis might be that coffee drinkers are healthier than nondrinkers of the same age and gender. This could arise if with the onset of gastric distress, coffee drinkers revert to nondrinkers. Comparisons with individuals who never drank coffee could theoretically address this. Another alternative hypothesis might be that, since people generally drink either coffee or tea, effects attributed to coffee are simply due to the absence of effects related to tea consumption. This can best be addressed in populations in which there are a significant number of individuals who drink both.

Because caffeine has been related to cancer risk in animal models, one needs to consider whether the effect is due to coffee or due to caffeine in cases in which coffee appears to be protective (63). The ideal approach would be to control for caffeine intake by comparing caffeinated with decaffeinated coffee consumption; but in fact, this has rarely been studied, and the factors contributing to selection of decaffeinated coffee might cause biases. Another approach is to compare coffee consumption risks with another caffeine rich drink, such as tea, to see if there is a consistency of effect. However, since tea has many other active ingredients, this could be misleading.

Many studies have found no effect of coffee despite a priori hypotheses that coffee would be detrimental. In these cases, the power of the study to detect an effect deserves attention. Power is weakened by error in the measurement of the true dose of active ingredient consumed. Doses can differ greatly depending upon the amount and type of coffee bean used and the roasting process involved. Roasting, which is seldom accounted for in epidemiologic research, ranges from light to dark and determines both the taste and caffeine level. The type of bean also determines the levels of active ingredients as coffee Arabica, which must be grown at altitudes over 3,000 feet, has half the caffeine of Robusta, and twice the cafestol of Robusta beans. Robusta beans are more disease resistant, grow at lower altitudes with higher yields, deliver a more astringent flavor, and contain negligible amounts of kahweol. Despite the expectation of high measurement error in coffee consumption, it is likely to be much lower than error or biases in the reports of other dietary components, such as snack foods or fast foods, considered socially undesirable.

Another driver of the power of a study to detect an effect is the distribution of drinking within a population. In some cases, such as Norwegian countries where coffee consumption is very high or Asian countries where regular, long-term coffee consumption is rare and in lower doses, a true effect might be undetectable.

CONCLUSIONS

All in all, at this time, the human epidemiologic evidence relating coffee consumption to risk of cancer involves a wide range of doses and stems from over 500 publications from multiple countries in North America, Europe, and Japan. Across sites, the evidence largely points to an overall lack of effect. There is evidence of a preventive effect in hepatocellular and endometrial cancer and possibly breast cancer. Epidemiologic studies of colorectal cancer have demonstrated some evidence of protection; but overall, at this time, there is a disconcerting heterogeneity by study design and gender. There is also evidence suggestive of a positive association with childhood leukemia among offspring of women with very high daily coffee consumption during pregnancy.