Dietary intake of selected flavonols, flavones, and flavonoid-rich foods and risk of cancer in middle-aged and older women

Abstract

Background: Flavonoids may protect against cancer development through several biological mechanisms. However, epidemiologic studies on dietary flavonoids and cancer risk have yielded inconsistent results.

Objective: We prospectively investigated the association between the intake of selected flavonoids and flavonoid-rich foods and risk of cancers in the Women's Health Study.

Design: A total of 3234 incident cancer cases were identified during 11.5 y of follow-up among 38,408 women aged ≥45 y. Intake of individual flavonols (quercetin, kaempferol, and myricetin) and flavones (apigenin and luteolin) was assessed from food-frequency questionnaires. Cox regression models were used to estimate the relative risk (RR) of total and site-specific cancer across increasing intakes of total and individual selected flavonoids and flavonoid-rich foods (tea, apple, broccoli, onion, and tofu).

Results: The multivariate RRs of total cancer across increasing quintiles of total quantified flavonoid intake were 1.00, 1.00, 0.93, 0.94, and 0.97 (P for trend = 0.72). For site-specific cancers, the multivariate RRs in the highest quintile of total quantified flavonoid intake compared with the lowest quintile were 1.03 for breast cancer, 1.01 for colorectal cancer, 1.03 for lung cancer, 1.15 for endometrial cancer, and 1.09 for ovarian cancer (all P > 0.05). The associations for the individual flavonoid intakes were similar to those for the total intake. There was also no significant association between intake of flavonoid-rich foods and the incidence of total and site-specific cancers.

Conclusion: Our results do not support a major role of 5 common flavonols and flavones or selected flavonoid-rich foods in cancer prevention.

INTRODUCTION

Consumption of fruit and vegetables has been consistently associated with a reduced risk of human cancers at many sites (1, 2). Flavonoids are a group of potentially chemoprotective compounds widely distributed in fruit, vegetables, and beverages of plant origin and have similar structures that consist of 2 phenolic benzene rings linked to a heterocyclic pyre or pyrone (3). More than 5000 individual flavonoids have been identified, which are classified into ≥10 subgroups according to their chemical structure (4, 5). Flavonoids of 6 principal subgroups—flavonols, flavones, anthocyanidins, catechins, flavanones, and isoflavones—are relatively common in human diets (3). Main dietary sources of flavonoids vary substantially between the subgroups (6, 7). Flavonols (eg, quercetin, kaempferol, and myricetin) are the most abundant flavonoids in plant foods and are mainly present in leafy vegetables, apples, onions, broccoli, and berries. Flavones (eg, apigenin and luteolin) and anthocyanidins are present in relatively small quantities in grains, leafy vegetables, and herbs. Catechins (eg, catechin and epicatechin) are abundant in tea, apples, grapes, chocolate, and red wine. Flavanones (eg, naringenin and hesperetin) are predominantly contained in citrus fruit and their juices. Isoflavones (eg, daidzein and genistein) are mainly found in soybeans and soy-based products. Flavonoids have many biological effects that may play a role in cancer prevention, including free radical scavenging, antimutagenic and antiproliferative properties, regulation of cell signaling and cell cycle, and inhibition of angiogenesis (6, 8). These properties may underlie, in part, the well-established association between high consumption of fruit and vegetables and reduced cancer risk.

Previous case-control studies have shown that the intake of total flavonoids, subgroups of flavonoids, or individual flavonoids is associated with a reduced risk of lung cancer (9–12), gastric cancer (13, 14), colorectal cancer (15, 16), breast cancer (17–19), ovarian cancer (20, 21), endometrial cancer (22), and non-Hodgkin's lymphoma (23). However, evidence from prospective cohort studies on the association between baseline dietary flavonoid intake and subsequent cancer risk remains controversial, with inverse associations observed for incidence of total cancer (24), lung cancer (24–26), colorectal cancer (27), and prostate cancer (26) in some but not all studies (28–30). To further investigate the potential role of dietary flavonoids in cancer prevention, we examined the association between baseline flavonoid intake and the risk of total and site-specific cancers in a large prospective cohort of middle-aged and older US women.

SUBJECTS AND METHODS

Study cohort

The Women's Health Study (WHS) was a randomized, double-blind, placebo-controlled, 2 × 2 factorial trial that evaluated the risks and benefits of low-dose aspirin and vitamin E in the primary prevention of cardiovascular disease and cancer (31). A third component, β-carotene, was initially included in the trial but was no longer included after a median treatment time of 2.1 y (32). From September 1992 to May 1995, a total of 39,876 female US health professionals aged ≥45 y, who were free of cardiovascular disease and cancer (except nonmelanoma skin cancer), were randomly assigned into the WHS; 39,310 of these women completed a 131-item validated semiquantitative food-frequency questionnaire (SFFQ) (33–35). For this study, we excluded 21 women because of insufficient completion of the SFFQ (>70 items left blank), 829 women because of implausible daily energy intake (<600 or ≥3500 kcal/d), and 59 women with cardiovascular disease or cancer diagnosed before randomization but reported after randomization. A total of 38,408 women remained for analysis after these exclusions. Written informed consent was obtained from all WHS participants. The WHS was approved by the Institutional Review Board of Brigham and Women's Hospital, Boston, MA.

Assessment of flavonoid intake and other baseline variables

On the baseline SFFQ, a commonly used unit or portion size was specified for each food item. Participants were asked how often they had consumed that amount, on average, during the previous year. Nine possible responses ranging from “never or less than once per month” to “6+ per day” were recorded. The average daily intake for each food item was calculated by multiplying the intake frequency by the portion size of the specific items. Nutrient intake was computed by multiplying the intake frequency of each unit of food by the nutrient content of the specified portion size according to food-composition tables from the Harvard School of Public Health, Boston, MA (36). Each nutrient reported in the present analysis was adjusted for total energy intake by using the residual method (37).

The flavonoid database was created by first analyzing foods known to be important dietary sources of flavonoids in the United States (38) and then supplementing the database with values published in Europe (39, 40). Total flavonoids represented the sum of 5 quantified individual flavonoids—3 flavonols (quercetin, kaempferol, and myricetin) and 2 flavones (apigenin and luteolin)—in the database. Five foods (apple, broccoli, onion, tofu, and tea) were selected for analyses as major sources of flavonoids based on the flavonoid content database. The SFFQ used in the WHS has been shown to have reasonable validity and reproducibility as a measure of long-term average dietary intake (33–35). In study populations similar to the WHS, Pearson correlation coefficients for intake of the selected flavonoid-rich food estimated from the SFFQ and dietary records ranged from 0.46 for broccoli to 0.77 for tea (34).

On a baseline questionnaire, women reported age, race, weight, and height; lifestyle characteristics including smoking status, alcohol use, leisure-time physical activity, hormone replacement therapy use, multivitamin use, and oral contraceptive use; medical history including history of diabetes, hypertension, hypercholesterolemia, benign colorectal polyps, and benign breast disease; reproductive history including age when menstrual periods began, menopausal status, and number of pregnancies lasting ≥6 mo; and family history of colorectal cancer, ovary cancer, and breast cancer.

Ascertainment of cancer cases

Every 6 mo during the first year and then annually thereafter, follow-up questionnaires were mailed to WHS participants to collect information about the occurrence of newly diagnosed study endpoints, including cancers. Deaths of participants were identified by reports from family members, postal authorities, and a search of the National Death Index. When a cancer was reported by questionnaire or identified from death certificates, written consent for medical record review was requested from the participant or next of kin of the deceased, and medical records were obtained from hospitals or treating physicians. All relevant information was reviewed by the WHS Endpoints Committee, which consisted of physicians who were blinded to treatment assignment. Reports of cancer were confirmed on the basis of pathology or cytology reports (96.8%) or, rarely, strong clinical and radiologic or laboratory marker evidence when a pathology or cytology review was not conducted. Only confirmed cases of cancer, including any invasive cancer except nonmelanoma skin cancer, were included in the analyses.

Statistical analysis

Statistical analyses were performed by using SAS software (version 9; SAS Institute, Cary, NC). The intake of total and individual flavonoids was divided into quintiles. The intake of flavonoid-rich foods was categorized according to intake frequency reported on the SFFQ. Mean values or proportions of baseline cancer risk factors were compared across categories of flavonoid intake to identify potential confounding factors. The incidence of total invasive cancer as well as site-specific cancers was calculated for each intake category by dividing the number of cases by the person-years of follow-up. Person-years of observation for each woman were calculated from time of randomization to the time of report of a cancer, death, the last day in follow-up, or 16 March 2007, whichever came first. Cox regression models were used to estimate the relative risk (RR) and 95% CI of incident cancer across increasing intakes of total quantified flavonoids, individual quantified flavonoids, and flavonoid-rich foods with the lowest category as the reference. Initial models only adjusted for age, race, total energy intake, and randomized treatment assignment. Multivariate models additionally adjusted for known risk factors for cancer at baseline, including body mass index (BMI), smoking, alcohol use, physical activity, postmenopausal status, hormone replacement therapy use, multivitamin use, family history of cancer in a parent or sibling, and intake of fruit and vegetables, fiber, folate, and saturated fat. Linear trends across increasing categories of intake were tested by using the median value of each category as an ordinal variable. To control for potential bias in endpoint ascertainment due to cancer screening behaviors, we conducted supplemental analyses that included mammography (yes, no) and colonoscopy (yes, no) screening reported on the 12-mo follow-up questionnaire as covariates in the multivariate model after excluding the first year of follow-up; the study results did not change significantly. No significant interactions were found between intake of flavonoids or flavonoid-rich foods and randomized treatment (aspirin, vitamin E, and β-carotene) according to Wald chi-square tests; therefore, we present the results for all WHS participants regardless of the intervention.

RESULTS

The median intake of total quantified flavonoids ranged from 8.88 mg/d in the lowest quintile to 47.44 mg/d in the highest quintile among 38,408 middle-aged and older women. Of the 5 individual flavonoids, quercetin was the major contributor to intake of total quantified flavonoid (median intake across quintiles ranging from 6.49 to 32.79 mg/d), followed by kaempferol (0.86–13.06 mg/d), myricetin (0.15–2.83 mg/d), apigenin (0.13–1.35 mg/d), and luteolin (0.01–0.20 mg/d).

Baseline cancer risk factors across each quintile of total quantified flavonoid intake are shown in Table 1. Women who consumed more total flavonoids were older, were less likely to be current smokers, were more physically active, and consumed less alcohol. Women in the higher quintiles of total flavonoid intake were also more likely to be postmenopausal and to use hormone replacement therapy and were less likely to take oral contraceptives. Total flavonoid intake was positively associated with fruit and vegetable, fiber, and folate intakes and inversely associated with red meat, total fat, and saturated fat intakes.

TABLE 1.

Baseline characteristics of 38,408 women aged ≥45 y and free of cardiovascular disease and cancer according to quintiles of total quantified flavonoid intake

	Quintiles of total quantified flavonoid intake
	1	2	3	4	5	P1
No. of participants	7676	7686	7682	7681	7683
Range (mg/d)	0–11.55	11.56–16.30	16.31–22.53	22.54–34.54	≥34.55
Median intake (mg/d)	8.88	13.91	19.13	27.24	47.44
Age (y)	53.9 ± 6.72	54.3 ± 6.9	54.8 ± 7.1	55.1 ± 7.3	54.9 ± 7.2	<0.0001
BMI (kg/m2)	26.2 ± 5.3	25.9 ± 4.9	25.9 ± 4.9	25.9 ± 5.0	26.0 ± 5.1	0.24
Race (% white)	96.7	96.4	95.1	94.4	92.8	<0.0001
Total energy intake (kcal/d)	1725 ± 540	1745 ± 519	1754 ± 534	1741 ± 551	1667 ± 524	<0.0001
Current smoking (%)	18.1	12.8	11.3	10.3	12.5	<0.0001
Alcohol intake (g/d)	4.55 ± 9.6	4.46 ± 8.4	4.05 ± 7.6	3.85 ± 8.1	3.53 ± 7.6	<0.0001
Physical activity (kcal/wk)	758 ± 1005	948 ± 1166	1012 ± 1191	1073 ± 1326	1055 ± 1344	<0.0001
Postmenopausal (%)	51.5	53.5	55.3	56.0	56.1	<0.0001
Hormone replacement therapy use (%)	37.9	41.1	41.4	42.9	41.8	<0.0001
Multivitamin use (%)	28.2	29.7	29.9	29.8	28.8	0.10
Family history of breast cancer (%)	6.15	6.62	6.02	6.46	5.82	0.24
Family history of colorectal cancer (%)	10.5	10.2	10.4	10.5	10.1	0.90
Family history of ovarian cancer (%)	3.22	2.76	2.75	2.97	3.33	0.12
Colon/rectal polyps (%)	2.57	2.47	2.32	2.75	2.43	0.51
Oral contraceptive use (%)	71.3	71.0	70.2	67.8	68.0	< 0.0001
No. of pregnancies ≥6 mo	2.53 ± 1.6	2.52 ± 1.5	2.55 ± 1.6	2.57 ± 1.5	2.54 ± 1.5	0.56
Age ≥30 y at first pregnancy ≥6 mo (%)	9.86	10.3	10.1	9.66	9.94	0.70
Age at menarche (y)	12.5 ± 1.4	12.4 ± 1.4	12.4 ± 1.4	12.4 ± 1.5	12.4 ± 1.5	0.07
Food intake (servings/d)
Fruit and vegetables	3.98 ± 2.0	5.58 ± 2.4	6.43 ± 2.9	6.95 ± 3.3	7.22 ± 4.2	<0.0001
Whole grains	1.35 ± 1.3	1.43 ± 1.2	1.47 ± 1.2	1.44 ± 1.2	1.37 ± 1.2	0.42
Red meats	0.83 ± 0.6	0.74 ± 0.5	0.70 ± 0.5	0.68 ± 0.5	0.65 ± 0.5	<0.0001
Nutrient intakes3
Fiber (g/d)	15.3 ± 4.5	18.0 ± 4.6	19.7 ± 5.5	20.6 ± 5.8	21.4 ± 6.9	<0.0001
Folate (mg/d)	373 ± 218	418 ± 217	437 ± 221	443 ± 221	470 ± 232	<0.0001
Total fat (g/d)	62.2 ± 11.8	58.4 ± 10.9	56.4 ± 11.2	55.9 ± 11.6	55.6 ± 12.2	<0.0001
Saturated fat (g/d)	21.8 ± 5.0	20.1 ± 4.5	19.1 ± 4.5	18.8 ± 4.7	18.7 ± 5.0	<0.0001

¹The trend test was used for continuous variables, and the chi-square test was used for categorical variables.

²Mean ± SD (all such values).

³Energy-adjusted by using the residual method (37).

The incidence of total invasive cancer across quintiles of total and individual quantified flavonoid intakes is shown in Table 2. A total of 3234 cancer cases were identified during an average of 11.5 y of follow-up. After adjustment for age, race, energy intake, and randomized treatment, the RRs and 95% CIs of total cancer across increasing quintiles of total flavonoid intake were 1.00, 0.99 (0.89, 1.10), 0.91 (0.82, 1.02), 0.90 (0.81, 1.01), and 0.93 (0.83, 1.03) (P for trend = 0.14). Additional adjustment for other cancer risk factors further attenuated the nonsignificant trend of inverse association (RR: 1.00, 0.93, 0.94, and 0.97 for quintiles 2–5 compared with the lowest quintile; P for trend = 0.72). The associations of individual flavonoid intakes with risk of total invasive cancer were similar to those of total flavonoids. The multivariate RRs and 95% CIs of total invasive cancer in the highest compared with the lowest quintile were 1.00 (0.89, 1.13) for myricetin intake, 0.97 (0.86, 1.08) for kaempferol intake, 0.94 (0.83, 1.07) for quercetin intake 1.07 (0.95, 1.22) for luteolin intake, and 0.99 (0.87, 1.12) for apigenin intake.

TABLE 2.

Relative risks and 95% CIs of total cancer according to quintiles of flavonoid intake

	Quintiles of intake
	1	2	3	4	5	P for linear trend1
Total quantified flavonoid
Median (mg/d)	8.88	13.91	19.13	27.24	47.44
Range (mg/d)	0–11.55	11.56–16.30	16.31–22.53	22.54–34.54	34.55–236.38
Cases/person-years	665/88,485	670/88,645	631/88,692	631/88,157	637/88,322
Adjusted for age, race, energy, treatment2	1.00 (reference)	0.99 (0.89, 1.10)	0.91 (0.82, 1.02)	0.90 (0.81, 1.01)	0.93 (0.83, 1.03)	0.14
Multivariate adjusted3	1.00 (reference)	1.00 (0.89, 1.12)	0.93 (0.82, 1.05)	0.94 (0.83, 1.07)	0.97 (0.86, 1.10)	0.72
Myricetin
Median (mg/d)	0.15	0.47	0.77	1.33	2.83
Range (mg/d)	0–0.34	0.35–0.60	0.61–1.00	1.01–1.84	1.85–19.92
Cases/person-years	660/87,711	692/89,640	637/88,169	621/88,751	624/88,031
Adjusted for age, race, energy, treatment2	1.00 (reference)	1.00 (0.90, 1.11)	0.95 (0.85, 1.06)	0.90 (0.80, 1.00)	0.93 (0.84, 1.04)	0.14
Multivariate adjusted3	1.00 (reference)	1.03 (0.92, 1.15)	0.98 (0.87, 1.10)	0.95 (0.84, 1.06)	1.00 (0.89, 1.13)	0.83
Kaempferol
Median (mg/d)	0.86	1.80	3.31	5.68	13.06
Range (mg/d)	0–1.25	1.26–2.55	2.56–4.33	4.34–7.88	7.89–99.72
Cases/person-years	702/88,612	645/88,205	600/88,781	658/88,414	629/88,289
Adjusted for age, race, energy, treatment2	1.00 (reference)	0.92 (0.83, 1.02)	0.84 (0.76, 0.94)	0.93 (0.84, 1.04)	0.91 (0.82, 1.02)	0.46
Multivariate adjusted3	1.00 (reference)	0.96 (0.85, 1.07)	0.88 (0.79, 0.99)	0.98 (0.88, 1.10)	0.97 (0.86, 1.08)	0.91
Quercetin
Median (mg/d)	6.49	10.19	13.62	19.21	32.79
Range (mg/d)	0–8.52	8.53–11.84	11.85–15.92	15.93–24.26	24.27–229.2
Cases/person-years	669/88,497	634/88,941	660/88,273	638/88,405	633/88,186
Adjusted for age, race, energy, treatment2	1.00 (reference)	0.93 (0.83, 1.04)	0.95 (0.85, 1.06)	0.91 (0.81, 1.01)	0.91 (0.81, 1.01)	0.12
Multivariate adjusted3	1.00 (reference)	0.92 (0.82, 1.04)	0.96 (0.85, 1.08)	0.94 (0.83, 1.06)	0.94 (0.83, 1.07)	0.64
Luteolin
Median (mg/d)	0.01	0.04	0.05	0.08	0.20
Range (mg/d)	0–0.02	0.03–0.04	0.05–0.06	0.07–0.13	0.14–2.02
Cases/person-years	706/98,527	584/82,146	728/98,794	565/77,080	651/85,755
Adjusted for age, race, energy, treatment2	1.00 (reference)	0.98 (0.88, 1.10)	1.00 (0.90, 1.11)	0.98 (0.88, 1.10)	1.01 (0.90, 1.12)	0.83
Multivariate adjusted3	1.00 (reference)	1.00 (0.89, 1.12)	1.03 (0.92, 1.15)	1.03 (0.91, 1.16)	1.07 (0.95, 1.22)	0.23
Apigenin
Median (mg/d)	0.13	0.29	0.43	0.66	1.35
Range (mg/d)	0–0.22	0.23–0.35	0.36–0.51	0.52–0.92	0.93–16.81
Cases/person-years	663/89,741	652/85,704	626/91,287	637/87,451	656/88,118
Adjusted for age, race, energy, treatment2	1.00 (reference)	1.01 (0.91, 1.13)	0.94 (0.84, 1.05)	0.99 (0.88, 1.10)	0.93 (0.84, 1.04)	0.18
Multivariate adjusted3	1.00 (reference)	1.07 (0.95, 1.20)	0.98 (0.88, 1.10)	1.04 (0.92, 1.16)	0.99 (0.87, 1.12)	0.61

¹Linear trends were tested by using the median value of each intake category as an ordinal variable.

²Model adjusted for age (continuous), race (white, nonwhite), total energy intake (continuous), and randomized treatment assignment (aspirin, vitamin E, β-carotene, or corresponding placebos).

³Model additionally adjusted for smoking (never; former; current, <15 cigarettes/d; current, ≥15 cigarettes/d); alcohol use (0, >0 to <5, 5 to <15, ≥15 g/d); physical activity (<200, 200 to <600, 600 to <1500, ≥1500 kcal/wk); postmenopausal status (yes, no, uncertain); hormone replacement therapy use (never, former, current); multivitamin use (never, former, current); BMI (continuous); family history of colorectal cancer, ovary cancer, and breast cancer (yes, no); and intake of fruit and vegetables, fiber, folate, and saturated fat (continuous).

The association between total quantified flavonoid intake and the incidence of common site-specific cancers is shown in Table 3.The RRs and 95% CIs of site-specific cancers in the highest compared with the lowest quintile of total flavonoid intake were 1.03 (0.85, 1.25) for breast cancer, 1.01 (0.68, 1.49) for colorectal cancer, 1.03 (0.67, 1.58) for lung cancer, 1.15 (0.73, 1.82) for endometrial cancer, and 1.09 (0.60, 2.01) for ovarian cancer (all P for trend >0.05). This overall lack of association with incidence of common site-specific cancers extended to the nonsignificant linear trend tests across quintiles for each individual flavonoid (data not shown). Total and individual flavonoid intake was also not associated with other rare site-specific cancers such as stomach, pancreatic, bladder, brain, thyroid, and cervical cancers and lymphoma/leukemia (data not shown).

TABLE 3.

Multivariate relative risks and 95% CIs of site-specific cancers according to quintiles of total quantified flavonoid intake¹

	Quintiles of total quantified flavonoid intake
	1	2	3	4	5	P for linear trend2
Breast cancer
No. of cases	270	299	266	258	258
Multivariate adjusted3	1.00 (reference)	1.10 (0.92, 1.32)	1.03 (0.86, 1.24)	0.97 (0.80, 1.18)	1.03 (0.85, 1.25)	0.79
Colorectal cancer
No. of cases	68	58	50	61	68
Multivariate adjusted4	1.00 (reference)	0.89 (0.61, 1.29)	0.68 (0.46, 1.02)	0.88 (0.60, 1.29)	1.01 (0.68, 1.49)	0.47
Lung cancer
No. of cases	63	44	43	44	47
Multivariate adjusted	1.00 (reference)	0.85 (0.57, 1.28)	0.80 (0.52, 1.23)	0.93 (0.60, 1.43)	1.03 (0.67, 1.58)	0.56
Endometrial cancer
No. of cases	47	58	49	55	50
Multivariate adjusted3	1.00 (reference)	1.17 (0.77, 1.76)	1.00 (0.64, 1.55)	1.18 (0.76, 1.83)	1.15 (0.73, 1.82)	0.61
Ovarian cancer
No. of cases	22	30	27	30	32
Multivariate adjusted3	1.00 (reference)	1.20 (0.68, 2.10)	0.86 (0.47, 1.58)	0.96 (0.52, 1.76)	1.09 (0.60, 2.01)	0.86

¹Model adjusted for age (continuous); race (white, nonwhite); total energy intake (continuous); randomized treatment assignment (aspirin, vitamin E, β-carotene, or corresponding placebos); smoking (never; former; current, <15 cigarettes/d; current, ≥15 cigarettes/d); alcohol use (0, >0 to <5, 5 to <15, ≥15 g/d), physical activity (<200, 200 to <600, 600 to <1500, ≥1500 kcal/wk); postmenopausal status (yes, no, uncertain); hormone replacement therapy use (never, former, current); multivitamin use (never, former, current); BMI (continuous); family history of colorectal cancer, ovary cancer, and breast cancer (yes, no); and intake of fruit and vegetables, fiber, folate, and saturated fat (continuous).

²Linear trends were tested by using the median value of each category as an ordinal variable.

³Models additionally adjusted for age at menarche (continuous), number of pregnancies lasting ≥6 mo (continuous), age at first pregnancy lasting ≥6 mo (<19, 20–24, 25–29, ≥30 y), and oral contraceptive use (yes, no).

⁴Models additionally adjusted for history of benign colorectal polyps (yes, no).

The association between intake of selected flavonoid-rich foods and the incidence of total cancer is shown in Table 4. The risk of total cancer did not significantly differ by intake of flavonoid-rich foods. The multivariate RRs and 95% CIs of total cancer among women who consumed ≥2 servings/wk compared with those who consumed <1 serving/mo of flavonoid-rich foods were 1.02 (0.94, 1.11) for tea, 1.13 (0.97, 1.30) for apples, 1.05 (0.88, 1.25) for broccoli, 0.92 (0.82, 1.03) for onions, and 0.91 (0.61, 1.35) for tofu. The risk of common and rare site-specific cancers also did not materially change with the intake of flavonoid-rich foods (data not shown). Sensitivity analyses that categorized food intake using other cutoffs yielded similar results (data not shown).

TABLE 4.

Multivariate relative risks and 95% CIs of total cancer according to intake of selected flavonoid-rich foods¹

	Frequency of intake
	<1 serving/mo	1–3 servings/mo	1 serving/wk	≥2 servings/wk	P for linear trend2
Tea
No. of cases/no. of women	1055/12,376	532/6157	313/3719	1287/15,625
Multivariate adjusted	1.00 (reference)	1.08 (0.97, 1.20)	1.00 (0.88, 1.15)	1.02 (0.94, 1.11)	0.98
Apples
No. of cases/no. of women	295/3556	758/9178	737/8510	1400/16,781
Multivariate adjusted	1.00 (reference)	1.07 (0.92, 1.23)	1.14 (0.98, 1.32)	1.13 (0.97, 1.30)	0.25
Broccoli
No. of cases/no. of women	189/2170	770/9087	1170/14,093	1090/12,851
Multivariate adjusted	1.00 (reference)	1.00 (0.85, 1.19)	1.00 (0.85, 1.18)	1.05 (0.88, 1.25)	0.28
Onion
No. of cases/no. of women	1349/16,297	852/9827	526/6191	469/5693
Multivariate adjusted	1.00 (reference)	1.00 (0.91, 1.09)	0.95 (0.85, 1.06)	0.92 (0.82, 1.03)	0.11
Tofu
No. of cases/no. of women	2949/34,741	172/2141	48/ 664	27/392
Multivariate adjusted	1.00 (reference)	0.98 (0.83, 1.16)	0.89 (0.65, 1.21)	0.91 (0.61, 1.35)	0.46

¹Model adjusted for age (continuous); race (white, nonwhite); total energy intake (continuous); randomized treatment assignment (aspirin, vitamin E, β-carotene, or corresponding placebos); smoking (never; former; current, <15 cigarettes/d; current, ≥15 cigarettes/d); alcohol use (0, >0 to <5, 5 to <15, ≥15 g/d); physical activity (<200, 200 to <600, 600 to <1500, ≥1500 kcal/wk); postmenopausal status (yes, no, uncertain); hormone replacement therapy use (never, former, current); multivitamin use (never, former, current); BMI (continuous); family history of colorectal cancer, ovary cancer, or breast cancer (yes, no); and intake of fruit and vegetables, fiber, folate, and saturated fat (continuous).

²Linear trends were tested by using the median value of each intake category as an ordinal variable.

DISCUSSION

In this prospective cohort study, we found no association between the intake of total quantified flavonoids—the sum of quercetin, kaempferol, myricetin, apigenin, and luteolin—or individual flavonoids and the risk of total invasive cancer and site-specific cancers. There was also no association between the intake of flavonoid-rich foods and the risk of total or site-specific cancers.

Flavonoids potentially have protective effects on cancer development through several important biological mechanisms. The free radical scavenging ability of flavonoids has been well characterized in experimental models (41). More recently, in vitro and in vivo experimental studies suggest that flavonoids also influence signal transduction pathways (42, 43), induce apoptosis (44), counteract inflammation (45), and inhibit proliferation in human cancer cell lines (46). Selected flavonoids may further promote a chemopreventive effect by increasing transcription of detoxification enzymes, thereby enhancing clearance of carcinogens (47). Despite these findings in laboratory settings, results from the present epidemiologic study suggest that the potentially protective effects of flavonoids, even if present in animal models and in vitro systems, may not be able to significantly reduce cancer risk at levels commonly consumed in a typical American diet and are unlikely to play a major role in mediating the observed benefits of fruit and vegetable intake against cancer.

Previous case-control studies have reported inverse associations of site-specific cancer risk with intake of kaempferol (10, 12, 13), quercetin (9, 11–13, 15), catechin (15), epicatechin (12, 15), and total flavonoid or flavonoid subgroups (11, 13–23, 48–52). These data from case-control studies need to be interpreted cautiously because of the retrospective collection of dietary information and the long latency of cancer development. The results from prospective cohort studies linking flavonoid intake and cancer risk are inconsistent. Of 6 studies from 4 European cohorts, no association was found between individual flavonoid intake and the incidence of cancer in a Dutch male cohort (28, 29), whereas an inverse association was found between intake of total flavonoid, some subgroup, or individual flavonoids and risk of lung cancer (25, 26, 53), pancreatic cancer (54), and prostate cancer (26) in 3 Finnish cohorts. In the United States, a high intake of flavonoids was associated with a reduced risk of ovarian cancer in the Nurses' Health Study I (55) and the California Teachers Study (56); total cancer, lung cancer (57), and rectal cancer (27) in the Iowa Women's Health Study; and pancreatic cancer among current smokers in the Multiethnic Cohort Study (58). However, there was no association between flavonoid intake and risk of cancers at many other sites in several US cohort studies (27, 30, 59). Apparent differences in study designs, participant characteristics, dietary assessment, and nutrient database used for analyses offer potential explanations to the inconsistent findings on dietary flavonoid and cancer risk. In these studies, diet was assessed via structured interviews (24, 28) or self-administered FFQs (27, 30). Flavonoid intake was typically computed by using published food-composition data in the Netherlands (39, 40, 60), but supplemented with analyses on additionally selected food items (26, 27, 30). Moreover, amounts of flavonoid intake have varied considerably between different populations (61).

Our study did not find a clear association between intake of flavonoid-rich fruit and vegetables and cancer incidence, possibly reflecting a modest effect of single food items or due to a relatively low range of intakes. Partly because of the complex mixture of nutrients in foods, previous study findings for intake of individual flavonoid-rich foods in association with cancer risk have varied greatly. For apple intake, a good source of quercetin, an inverse association with risk of lung cancer was found in one case-control study (9) and one cohort study (24), but no association with cancer risk was found in many other studies (15, 29, 30, 54, 55, 57). Intake of broccoli, which has a high content of quercetin and kaempferol, was inversely associated with colorectal cancer in a case-control study (62), but was not associated with colorectal (30) or ovarian (56) cancer in cohort studies. Similarly, intake of onions, a good source of quercetin, was associated with cancer risk in a case-control study (9) but not in several large cohort studies (24, 30, 54, 55); drinking a high quantity of tea, a rich source of catechins (a flavonoid subgroup not assessed in our study), was associated with a lower risk of lung cancer (63) and cancers of the digestive tract (15) in case-control studies, but not in many prospective cohort studies (29, 54, 56, 64–67). Finally, the inverse association of soy food intake with cancer risk has been reported extensively in case-control studies among Asian populations (21, 68–72), in whom soy intake is notably high. However, cohort studies that specifically examined tofu intake found no significant association with risk of breast cancer (73) or ovarian cancer (56). Considering the low intake of soy products among US adults in the early 1990s, our observation of no association between tofu intake and cancer risk was not surprising.

Several alternative explanations for the null associations in our study are possible. First, flavonoids exist ubiquitously in plant-based foods; however, the SFFQ used in the present study was not specifically designed to estimate flavonoid intake. Missing information on certain flavonoid-containing food items can lead to an incomplete assessment of flavonoid intake. Second, the nutrient database used in our study provided intake assessment for only 5 individual flavonoids; therefore, many other flavonoids were unavailable for analyses. A more comprehensive flavonoid database may provide stronger associations with cancer risk. Third, the quantities of flavonoids in foods may differ by species variety, growth condition, maturation, preparation, and food-processing methods. These factors, which are not accounted for in the nutrient database, can contribute to random errors in intake assessment and result in bias toward the null. Similarly, the assessment of dietary intake from a single SFFQ is subject to random error of self-report, cannot capture cumulative intake and possible dietary changes during long-term follow-up, and tends to underestimate any associations. Fourth, although our cohort showed a reasonable range of flavonoid intakes, we cannot rule out the possibility that an association with cancer risk exists at different flavonoid intake amounts. Fifth, although the total number of cases in the entire cohort was large, the number of cases in each category may not be large enough to detect a modest effect of dietary intake. Moreover, the possibility cannot be excluded that the lack of association for some rare site-specific cancers may have been due to a small number of cases. Finally, our study results are based on a population of female health professionals who were initially free of cardiovascular disease and cancer. Studies in other populations are needed to confirm our findings.

In conclusion, we found no significant association between baseline intake of 5 common flavonoids or some flavonoid-rich foods and the risk of cancer in middle-aged and older, predominantly white women. Our findings do not support a major protective role of the flavonoids we assessed against cancer. Studies with a broader and more precise assessment of flavonoid intake, including the use of relevant biomarkers, are needed to confirm or refute our findings. In addition, more studies are needed to better characterize the bioaccessibility, bioavailability, and metabolism of flavonoids; to improve our knowledge of the specific actions of each flavonoid subclass and individual flavonoid; and to further clarify whether some flavonoids or flavonoid-rich foods may have a particularly important effect on cancer prevention.