Flavonoids and Cancer Prevention: What Is the Evidence in Humans?

Abstract

Diets high in fruits and vegetables are associated primarily with a reduced risk of human cancers of the gastrointestinal tract, as well as some other common cancers including those of the lung, breast, and prostate. Plants contain hundreds of bioactive substances, including flavonoids, which are a large group of compounds consisting of two phenolic benzene rings linked to a heterocyclic pyran or pyrone. This report reviews published epidemiologic studies that have examined associations of flavonoid intake with cancer risk in humans. Findings to date are extremely limited, but there is relatively consistent evidence from these studies that flavonoids, especially quercetin, reduce risk of lung cancer. Evidence is modest for other epithelial cancers. Further research is needed to confirm these findings before public health recommendations about flavonoids can be offered to consumers.

Keywords:

• Cancer prevention
• clinical studies
• epidemiology
• flavonoids

Introduction

Fruit and vegetable intake is consistently associated with a reduced risk of cancer (Steinmetz & Potter, 1991a, 1991b, 1996; Ziegler et al., 1996). In 1997, the World Cancer Research Fund summarized data from the published epidemiologic literature at that time and concluded that there was convincing evidence to suggest that fruit and vegetable intake was inversely associated with cancers of the lung, stomach, mouth, pharynx, esophagus, colon, and rectum (World Cancer Research Fund AICR, 1997). These general observations are important, but data are still needed on identifying whether specific fruits and vegetables, or bioactive compounds found in fruits and vegetables, may have unique protective properties for specific cancers. These kinds of data are important for formulating focused published health recommendations both for the general population and for those who are at increased risk of specific cancers due to family history or lifestyle behaviors (i.e., smoking).

Plant foods are the source of myriad bioactive compounds, which may influence the risk of human carcinogenesis. For many years, β-carotene was presumed to be the key compound in relation to cancer prevention (Ziegler et al., 1991, 1996; Mayne et al., 1994). However, the three National Cancer Institute sponsored β-carotene chemoprevention trials, whose objectives were to reduce lung cancer risk, found that β-carotene supplements unexpectedly increased lung cancer risk (The Alpha-Tocopherol Beta Carotene Cancer Prevention Study Group, 1994; Omenn et al., 1996; Hennekens et al., 1996). Partly as a result of the β-carotene chemoprevention studies, there has been an increasing amount of research on other phytochemicals (Craig, 1997), such as isoflavones (Adlercreutz et al., 2000; Lampe, 2003), glucosinolates (Brooks et al., 2001; Shapiro et al., 1998), and allium compounds (Hageman et al., 1997). One of the most promising plant compounds under investigation are flavonoids, a large group of compounds with a similar chemical structure; namely, two phenolic benzene rings linked to a heterocyclic pyran or pyrone (Aherne & O'Brien, 2002). Because of the considerable scientific interest in these unique plant compounds, this report is designed to review the published epidemiologic literature on associations of dietary flavonoid intake with human cancer risk. The review included all manuscripts published in the English language that are cited in the MEDLINE database of the National Library of Medicine. The sole restriction is that the review is limited to reports of flavonoids found in fruits and vegetables, as two reviews of tea polyphenols in relation to cancer risk have previously been published (Yang et al., 2001; McKay & Blumberg, 2002).

Food sources of flavonoids

Flavonoids are widespread in the food supply. Until recently, though, investigations of dietary flavonoid intake with cancer risk have been limited by the lack of reliable data on the flavonoid content of foods. The USDA database for flavonoids for selected foods was published in early 2003. This database lists the flavonoid values for 19 compounds from 255 foods. All values in the database were generated by high-performance liquid chromatography (HPLC), which facilitates adequate separation of the myriad flavonoid compounds. Table 1 lists the subclasses of compounds included in the database, and details can be found at the Nutrient Data Laboratory Web site (http://www.nal.usda.gov/fnic/foodcomp). Importantly, the investigations reviewed in this report were published prior to the release of the 2003 flavonoid database, and therefore the quantities reported here would likely be different than those that will be reported in future studies using the 2003 database.

Table 1 Bioactive compounds and their food sources in the 2003 USDA database for the flavonoid content of selected foods^a.

Subclass of flavonoid	Specific compounds	Principal food sources
Flavonols	Quercetin, kaempferol, myricetin, isohamnetin	Apples, black tea, blueberries, broccoli, buckwheat, cocoa, cranberries, green beans, green tea, kale, onions, red wine
Flavones	Luteolin, apigenin	Celery, parsley, peppermint
Flavonones	Hesperetin, naringenin, eriodictyol	Chili peppers, citrus fruits
Flavon-3-ols	(+)-Catechin, (+)-gallocatechin, (−)-epicatechin, (−)-epigallocatechin, (−)-epicatechin 3-gallate, (−)-epigallocatechin 3-gallate, theaflavin, theaflavin 3-gallate, theaflavin 3′-gallate, theaflavin 3,3′-digallate, thearubigins	Apricots, blackberries, black tea, broadbeans, chocolate, grapes, green tea, red wine
Anthocyanidins	Cyanidin, delphinidin, malvidin, pelargonidin, peonidin, petunidin	Blueberries, cherries, elderberries, raspberries, red wine

^aAvailable at http://www.nal.usda.gov/fnic/foodcomp.

Cohort studies of dietary flavonoids and cancer risk

Cohort studies are also known as “follow-up studies.” In this design, individuals enter the cohort free of disease and are followed over a period of time to assess disease occurrence. Statistical tests are conducted to investigate whether those who had similar exposures (e.g., to dietary factors or smoking) had similar incidence rates for the disease during the follow-up period. Cohort studies are particularly useful because of the extended follow-up time and the capability to examine many exposures, including diet, in relation to cancer risk. Further, because of the follow-up time inherent with the cohort study design, the temporal relationship between exposure and outcome can be more easily confirmed compared to other observational designs. This point is particularly important in relation to diet and cancer, because we do not yet know the point in the carcinogenesis process at which flavonoid intake may exert its protective effect (i.e., initiation or promotion). Cohort studies may offer important information in that regard.

In Table 2, results are presented from five publications, conducted in four cohorts, which examined whether flavonoids were associated with cancer risk in humans (Hertog et al., 1994; Hirvonen et al., 2001; Arts et al., 2001, 2002; Knekt et al., 2002). None of these studies was specifically designed to examine whether flavonoids are associated with cancer risk. Each of these studies had other primary objectives, but as cohort studies often collect a multitude of data over a period of many years concurrent with the collection of many distinct disease end-points, they become excellent opportunities to test new hypotheses. For example, the Zutphen Elderly Study reported a 44% non-statistically significant reduced risk of cancer associated with high compared to low consumption of flavonoids (Hertog et al., 1994). A subsequent analysis from this cohort focusing only on catechin reported no association with cancer risk (Arts et al., 2001). Hirvonen et al. examined flavonoid intake and cancers at several sites among men enrolled in the Alpha-Tocopherol Beta-Carotene Cancer Prevention Study. These authors reported that for men whose flavonoid intake was in the fourth versus the first quartile (reference group) at the time of study entry, there was a 44% reduced risk of lung cancer after a median follow-up time of 6 years. There was a near statistically significant decreased risk of renal cell carcinoma and increased risk of colorectal cancer for the highest compared to the lowest quartiles of total flavonoid consumption. Risk for other cancers did not appear to be related to flavonoid intake (Hirvonen et al., 2001). Among postmenopausal women in The Iowa Women's Health Study, the highest versus lowest quintile of baseline catechin intake was associated with an approximate halving in risk of rectal cancer after 13 years of follow-up (Arts et al., 2002).

Table 2 Cohort studies of dietary flavonoids and cancer risk.

Investigators, year, location	Subjects	Cancer	Relative Risk^a	95% confidence interval	Comments
Hertog et al. (1994) The Netherlands	738 elderly men 75 cases	Any cancer	Total flavonoids^b	0.57 (0.31, 1.08)^c p, trend = 0.08
Hirvonen et al. (2001) Finland	27,100 male smokers 791 cases	Lung cancer	Total flavonoids^b	0.56 (0.45, 0.69)^d p, trend < 0.001	Participants were enrolled in the Alpha-Tocopherol, β-Carotene Cancer Prevention Study
	156 cases	Urothelial cancer		1.2 (0.73, 1.8)^d p, trend = 0.77
	59 cases	Renal cancer		0.63 (0.36, 1.1)^d p, trend = 0.10
	226 cases	Prostate cancer		1.3 (0.87. 1.8)^dp, trend = 0.24
	133 cases	Colorectal cancer		1.7 (1.0, 2.7)^d p, trend = 0.10
	111 cases	Stomach cancer		1.2 (0.71, 1.9)^d p, trend = 0.51
Arts et al. (2001)	728 elderly men
The Netherlands	96 cases	Epithelial cancer^e	Catechin	0.94 (0.56, 1.59)^f p, trend = 0.82	The cohorts for the Netherlands cohorts are identical
Arts et al. (2002) Iowa, USA	34,651 postmenopausal women
	276 cases	Uterus	Catechin	1.00 (0.73, 1.36)^g,h p, trend = 0.54
	151 cases	Ovary		0.73(0.44, 1.24)^g,h p, trend = 0.21
	130 cases	Pancreas		0.74 (0.46, 1.20)^g p, trend = 0.77
	114 cases	Kidney		0.73 (0.40, 1.32)^g p, trend = 0.12
	103 cases	Bladder		1.12 (0.65, 1.93)^g p, trend = 0.93
Knekt et al. (2002) Finland	9865 adult males and females
	74 cases	Stomach	Quercetin	1.03 (0.52, 2.07)^i p, trend = 0.82
			Kaempferol	1.14 (0.59, 2.22)^i p, trend = 0.98
			Myricetin	1.16 (0.59, 2.26)^i p, trend = 0.29
			Hesperetin	0.88 (0.43, 1.80)^i p, trend = 0.67
			Naringenin	0.94 (0.47, 1.88)^i p, trend = 0.67
	90 cases	Colorectum	Quercetin	0.62 (0.33, 1.17)^i p, trend = 0.22
			Kaempferol	1.13 (0.60, 2.12)^i p, trend = 0.96
			Myricetin	1.31 (0.71, 2.43)^i p, trend = 0.39
			Hesperetin	0.97 (0.50, 1.90)^i p, trend = 0.84
			Naringenin	0.93 (0.48, 1.82)^i p, trend = 0.95
	81 cases	Urinary organs	Quercetin	0.87 (0.44, 1.72)^i p, trend = 0.49
			Kaempferol	0.67 (0.34, 1.31)^i p, trend = 0.11
			Myricetin	0.78 (0.41, 1.49)^i p, trend = 0.23
			Hesperetin	0.83 (0.40, 1.70)^i p, trend = 0.94
			Naringenin	0.81 (0.39, 1.66)^i p, trend = 0.90
	169 cases	Lung	Quercetin	0.42 (0.25, 0.72)^i p, trend = 0.001	Men only (n = 9865 males in cohort)
			Kaempferol	0.81 (0.51, 1.28)^i p, trend = 0.26
	169 cases	Lung	Myricetin	1.20 (0.78, 1.83)^i p, trend = 0.98
			Hesperetin	0.74 (0.46, 1.18)^i p, trend = 0.07
			Naringenin	0.64 (0.39, 1.04)^i p, trend = 0.02
	95 cases	Prostate	Quercetin	0.76 (0.40, 1.42)^i p, trend = 0.35	5128 males in cohort
			Kaempferol	1.03 (0.53, 2.02)^i p, trend = 0.54
			Myricetin	0.43 (0.22, 0.86)^i p, trend = 0.002
			Hesperetin	1.47 (0.80, 2.73)^i p, trend = 0.26
			Naringenin	1.48 (0.80, 2.71)^i p, trend = 0.27
	125 cases	Breast	Quercetin	0.62 (0.37, 1.03)^i p, trend = 0.25	4647 females in cohort
			Kaempferol	0.87 (0.53, 1.41)^i p, trend = 0.70
			Myricetin	0.95 (0.57, 1.60)^i p, trend = 0.63
			Hesperetin	1.08 (0.63, 1.86)^i p, trend = 0.93
			Naringenin	1.14 (0.67, 1.94)^i p, trend = 0.82

^aRelative risks are for highest relative to lowest quantile of intake.

^bSum of quercetin, kaempherol, myricetin, luteolin, and apignin.

^cAdjusted for age, smoking, body mass index, physical activity, and intake of energy, alcohol, vitamin C, β-carotene, vitamin E, and fiber.

^dRelative risks for lung cancer are adjusted for age, trial arm (study vitamins vs. placebo), and smoking. Relative risks for prostate, colorectal, and stomach cancer are adjusted for age and trial arm only.

^eIncludes cancers of the oropharynx, esophagus, stomach, colon, rectum, liver, gallbladder, pancreas, kidney, bladder, and lung.

^fAdjusted for age, physical activity, total energy, alcohol, smoking, body mass index, and intake of coffee, fiber, vitamin C, vitamin E, and β-carotene.

^gAdjusted for age, body mass index, waist-hip ratio, physical activity, smoking, and intake of energy, alcohol, and fruits and vegetables.

^hRelative risks for breast, ovary, and uterus cancer exclude women with baseline mastectomy, hysterectomy, or oophorectomy, leaving 21,502 for analysis. Additional adjustments for use of estrogen replacement therapy, age, and menarche, age at menopause, and age at first pregnancy.

ⁱAdjusted for age, sex, geographic area, occupation, smoking, and body mass index.

The Finnish Mobile Health Examination Survey investigated five flavonoids and cancer risk at six specific sites after a maximum of 30 years of follow-up (Knekt et al., 2002). There was a statistically significant 58% and 36% reduction in lung cancer risk for men in the highest quartile of baseline quercetin and naringenin consumption, respectively, compared to men in the lowest quartile (Knekt et al., 2002). Hesperetin (a flavonoid found in high levels in citrus fruits) intake was associated with a 42% reduction in lung cancer risk for the third compared to the first quartile of consumption, a finding that was statistically significant. However, there was no further protection at higher levels of consumption, suggesting that there may be a threshold effect beyond which no further protection is realized. Prostate cancer risk also appeared to be modestly associated with flavonoid intake in the Finnish Mobile Health Survey. For example, compared to men in the lowest quartile of myricetin intake, the risk of prostate cancer for those in the top quartile of consumption was reduced by 57%. However, none of the other flavonoids was significantly associated with cancer risk at any organ site in this large and well-followed cohort (Knekt et al., 2002).

Case-control studies of flavonoids and cancer risk

Case-control studies select study participants based on whether or not they have been diagnosed with the disease end-point of interest. The cases and controls are compared with regard to exposures (usually in the distant past), and statistical tests are conducted to determine if the proportion of cases with a particular exposure (in this case, exposure to dietary flavonoids) is different than that of the controls. Conclusions may then be drawn about whether or not the exposure of interest may be a risk factor for the studied outcome. It is important to remember that cause and effect cannot be concluded from case-control studies, only measures of association.

Only five case-control studies have examined associations of dietary flavonoid intake and cancer risk (Table 3) (Garcia-Closas et al., 1998, 1999; De Stefani et al., 1999; Garcia et al., 1999; Le Marchand et al., 2000a). Total flavonoids and quercetin were associated with a 30–42% reduced risk of lung cancer in two published studies but a non-significant increased risk for lung cancer in a third study (Garcia-Closas et al., 1998; De Stefani et al., 1999; LeMarchand et al., 2000b). With regard to other flavonoids, high versus low quercetin and kaempferol intakes were associated with 40% and 50% reduction in risk, respectively, for stomach cancer. Flavonoid intake did not appear to be statistically significantly associated with bladder cancer risk (Garcia et al., 1999).

Table 3 Case-control studies of dietary flavonoids and cancer risk.

Investigators, year, location	Subjects	Cancer	Odds Ratio^a	95% Confidence Interval	Comments
Garcia-Closas et al. (1998) Spain	103 cases and 206 hospital controls	Lung cancer	Quercetin	1.89 (0.72, 4.92)^b p, trend = 0.19	Females only. Cases and controls matched on age, area of residence, and hospital.
			Kaempferol	0.51 (0.22, 1.17)^b p, trend = 0.10
			Luteolin	0.59 (0.24, 1.43)^b p, trend = 0.40
De Stefani et al. (1999) Uruguay	541 cases and 540 hospital controls	Lung cancer	Total flavonoids	0.59 (0.40, 0.87)^c p, trend = 0.01	Males only. Cases and controls matched on age
			Quercetin	0.58 (0.39, 0.85)^c p, trend = 0.007	(10 year increments), residence, and urban/rural status.
			Kaempferol	0.79 (0.55, 1.17)^c p, trend = 0.16
Garcia et al. (1999) Spain	497 cases and 547 neighborhood controls + 566 hospital controls	Bladder cancer	Quercetin	1.21 (0.8, 1.9) p, trend = 0.94	Cases and controls matched by sex, age, area of residence, and hospital
			Kaempferol	1.35 (0.9, 2.1) p, trend = 0.11
			Luteolin	0.95 (0.6, 1.4) p, trend = 0.40
			Myricetin	0.82 (0.6, 1.2) p, trend = 0.20
Garcia-Closas et al. (1999) Spain	354 cases and 354 hospital controls	Stomach cancer	Quercetin	0.62 (0.35, 1.10)^d p, trend = 0.02	Cases and controls matched by age, sex, area of residence, and hospital
			Kamepferol	0.48 (0.26, 0.88)^d p, trend = 0.04
			Myricetin	1.12 (0.67, 1.85)^d p, trend = 0.45
LeMarchand et al. (2000) Hawaii	582 cases selected from the Hawaii SEER^e cancer  registry and 582 population controls	Lung cancer	Quercetin	0.7 (0.4, 1.1)^f p, trend = 0.07	Cases and controls matched on sex, ethnicity, and age
			Kaempferol	0.9 (0.5, 1.4)^f p, trend = 0.41
			Myricetin	1.0 (0.6, 1.6)^f p, trend = 0.42
			Herperidin	1.2 (0.7, 2.0)^f p, trend = 0.54
			Naringenin	0.7 (0.5, 1.1)^f p, trend = 0.17

^aOdds Ratios are for the highest compared to the lowest quantile of intake.

^bAdjusted for intake of vitamin E, vitamin C, total carotenoids, and each of the other flavonoids.

^cAdjusted for age, education, family history of lung cancer, body mass index, smoking, total energy, and total fat intake.

^dAdjusted for intake of total energy, nitrites, nitrosamines, vitamin C, total carotenoids, and each of the other flavonoids.

^eSEER (Surveillance, Epidemiology and End Results) is a cancer registry program of the National Cancer Institute.

^fAdjusted for smoking and intake of β-carotene and saturated fat.

Other study designs

No cross-sectional studies and no randomized controlled trials have been published on flavonoids and cancer risk. The available data are limited to cohort and case-control studies.

Discussion

This report provides modest evidence that flavonoid intake may protect humans from a variety of epithelial cancers. The evidence is strongest for an inverse association of total flavonoids, foods rich in flavonoids and quercetin, with lung cancer risk (Hirvonen et al., 2001; Knekt et al., 2002; De Stefani et al., 1999; Le Marchand et al., 2000a). A single study examined rectal cancer, which reported a halving of risk associated with high versus low catechin intake (Arts et al., 2002). Clearly, because there are only a limited number of studies to date, any conclusions should be regarded as preliminary but worthy of further investigation.

Flavonoids influence several important biological functions, which may explain the observed inverse associations of flavonoids with cancer risk. The free-radical scavenging ability of flavonoids has been fairly well characterized in experimental systems. More recently, in vitro and animal model systems suggest that flavonoids influence signal transduction pathway (Morrow et al., 2001; Frigo et al., 2002), stimulate apoptosis (Choi et al., 2001), inhibit inflammation (Cho et al., 2001), and inhibit proliferation in human cancer cell lines (Manthey & Guthrie, 2002). Selected flavonoids may also increase transcription of phase II detoxifying enzymes, which supports a cancer protective effect via the clearance of procarcinogenic substances that are detoxified and eliminated by phase II enzyme products (Valerio et al., 2001). A study conducted with azoxymethane (AOM)-treated mice who were fed with either a standard diet, a standard diet plus rutin, or a standard diet plus quercetin showed that the flavonoids substantially decreased the number of focal areas of dysplasia that were induced by the AOM exposure (Yang et al., 2000). This type of evidence is intriguing because it suggests that flavonoids may be related to events early in the carcinogenesis pathway. With this in mind, future studies in human populations may benefit most from cohort designs. As noted above, this design facilitates dietary assessment over a prolonged period of time and captures dietary exposures that may influence early events in carcinogenesis. It is important to remember, however, much of the mechanistic evidence is derived from in vitro and animal model studies, which may not accurately represent the action of flavonoids in the context of the food matrix and in a mixed diet among free-living humans. Although the randomized controlled trial is considered the gold standard to the establishment of cause and effect, there is not yet enough evidence to warrant such a design with regard to flavonoids and cancer risk. However, human feeding studies may be able to identify some of the mechanisms by which flavonoids influence gene transcription and enzyme activity.

All studies have limitations that must be noted. Nutritional epidemiology is limited by the fact that bioactive compounds in foods are highly correlated. The influence of any one nutrient or compound is not completely independent of other nutrients (Willett, 1998). The possibility cannot be ruled out that the protective associations observed for flavonoids are simply either markers of unmeasured constituents of plants or a marker of a generally healthy lifestyle. Moreover, there may be residual confounding and/or measurement error (a characteristic of all self-report) that could have introduced bias in the relative risk or odds ratio estimates. Finally, a variety of dietary assessment instruments were used in these studies (i.e., dietary history, food frequency questionnaires), and the methods are not necessarily comparable. Therefore, readers should exercise caution when interpreting the results presented in this manuscript.

Conclusions

There is modest evidence that flavonoids, particularly quercetin, are inversely associated with cancer risk. Because a very limited number of epidemiologic studies have been conducted to examine the associations of dietary intake of flavonoids with cancer risk, it is premature to make public health recommendations at this time. However, the data to date are interesting and suggest the need for further investigations of these important bioactive plant compounds. Additional studies using updated dietary databases with HPLC values for flavonoid estimates from food will provide evidence regarding the strength of association of these plant compounds with cancer risk. Moreover, human experimental nutrition studies that are designed to characterize the absorption, metabolism, and disposition of plant flavonoids and their interaction with key enzyme systems could provide crocial information about the biological basis for an inverse association of flavonoid intake with cancer risk. Until more specific recommendations can be offered, consumers should follow current dietary guidance for cancer prevention, which includes consumption of at least five or more servings of fruits and vegetables per day.