Flavonoid intake and survival after diagnosis of colorectal cancer: a prospective study in 2 US cohorts

Abstract

Background

Although experimental evidence supports anticancer effects of flavonoids, the influence of flavonoid intake on colorectal cancer (CRC) survival remains unknown.

Objectives

This study aimed to assess the association of postdiagnostic flavonoid intake with mortality.

Methods

We prospectively assessed the association of postdiagnostic flavonoid intake with CRC-specific and all-cause mortality in 2552 patients diagnosed with stage I–III CRC in 2 cohort studies—the Nurses’ Health Study and the Health Professionals Follow-up Study. We assessed the intake of total flavonoids and their subclasses using validated food frequency questionnaires. We used the inverse probability–weighted multivariable Cox proportional hazards regression model to calculate the hazard ratio (HR) of mortality after adjusting for prediagnostic flavonoid intake and other potential confounders. We performed spline analysis to evaluate dose–response relationships.

Results

The mean [standard deviation (SD)] age of patients at diagnosis was 68.7 (9.4) y. During 31,026 person-y of follow-up, we documented 1689 deaths, of which 327 were due to CRC. The total flavonoid intake was not associated with mortality, but a higher intake of flavan-3-ols was suggestively associated with lower CRC-specific and all-cause mortality, with multivariable HR (95% CI) per 1-SD increases of 0.83 (0.69–0.99; P = 0.04) and 0.91 (0.84–0.99; P = 0.02), respectively. The spline analysis showed a linear relationship between postdiagnostic flavan-3-ol intake and CRC-specific mortality (P = 0.01 for linearity). As the major contributor to flavan-3-ol intake, tea showed an inverse association with CRC-specific and all-cause mortality, with multivariable HRs per 1 cup/d of tea of 0.86 (0.75–0.99; P = 0.03) and 0.90 (0.85–0.95; P < 0.001), respectively. No beneficial associations were found for other flavonoid subclasses.

Conclusions

Higher intake of flavan-3-ol after CRC diagnosis was associated with lower CRC-specific mortality. Small, readily achievable increases in the intake of flavan-3-ol–rich foods, such as tea, may help improve survival in patients with CRC.

Introduction

Despite advances in screening and treatment, colorectal cancer (CRC) remains the second leading cause of cancer death worldwide. Approximately 930,600 individuals died from the disease in 2020 [1]. Recommendations from the American Cancer Society highlight the importance of a healthy lifestyle, including weight management, physical activity, and a healthy diet, for cancer survivor care [2]. Among the dietary factors, a higher intake of flavonoids has been associated with a lower risk of CRC in some epidemiological studies [3]. However, the influence of flavonoids on prognosis in patients with CRC remains unclear.

Flavonoids are widely distributed in fruits, vegetables, and plants [4]. Based on their molecular structure, flavonoids are divided into several subclasses (expressed as aglycones), including proanthocyanins, flavan-3-ols, flavanones, flavonols, anthocyanins, and flavones. Flavonoids may exert their anticancer activity via the regulation of multiple signaling pathways, including apoptosis, cellular proliferation, inflammation, and modulation of the gut microbiome [5,6].

Therefore, we performed this prospective study to assess the postdiagnostic intake of total flavonoids and their major subclasses in relation to CRC-specific and all-cause mortality in patients with stage I–III CRC. We also performed a detailed dose–response analysis and examined the major food contributors to intake.

Methods

Study population

We used data from 2 large prospective cohort studies in the United States, the Nurses’ Health Study (NHS), which included 121,701 women aged 30–55 y beginning in 1976, and the Health Professionals Follow-up Study (HPFS), which included 51,529 men aged 40–75 y beginning in 1986 [7,8]. Every 2 y, follow-up questionnaires were administered to collect updated information from participants on their lifestyle, medication use, and newly diagnosed diseases. Every 4 y, food frequency questionnaires (FFQs) were self-administered to assess the consumption frequency of ∼150 food items during the past year [9]. Over a 90% response rate was reached in each follow-up cycle in the 2 cohorts. The present study defined the baseline as the questionnaire return date in 1980 in the NHS and in 1986 in the HPFS, when detailed dietary data were first collected. Informed consent was obtained from all study participants at study enrollment. Patients with stage IV disease were excluded because of their different clinical management from patients with stage I–III CRC and their poor prognosis, due to which only a small proportion of patients returned the postdiagnosis questionnaire. The study protocol was approved by the institutional review boards of Brigham and Women’s Hospital (1999P011117) and Harvard T.H. Chan School of Public Health and those of participating state cancer registries as needed.

Determination of CRC cases

Newly diagnosed CRCs in participants in the past 2 y were recorded in each follow-up questionnaire. After obtaining informed consent from the study participants, we acquired medical records, and our study physicians, who were blinded to the exposure data, reviewed the medical records to confirm CRC diagnosis and extract the histopathological information on CRC, including tumor subsite, differentiation, and stage according to the American Joint Committee on Cancer (AJCC) [10]. After excluding patients with missing data on pre- or postdiagnostic flavonoid intake, we documented a total of 2552 patients with stage I–III CRC from baseline up to 1 June, 2016, for the NHS (n = 1766) and 31 January, 2016, for the HPFS (n = 786; see Supplemental Figure 1 for the flowchart).

Ascertainment of deaths

Death status was confirmed via the review of the National Death Index and reports from family members and the postal system in response to follow-up questionnaires [11]. Study physicians confirmed the causes of death via a review of medical records and death certificates.

Assessment of flavonoid intake

We assessed the intake of total flavonoids and major flavonoid subclasses from the FFQs, including proanthocyanins, flavan-3-ols (catechin and epicatachin), flavanones (eriodictyol, hesperetin, and naringenin), flavonols (quercetin, kaempferol, myricetin, and isohamnetin), anthocyanins (cyanidin, delphinidin, malvidin, pelargonidin, petunidin, and peonidin), and flavones (luteolin and apigenin). According to the composition database from the U.S. Department of Agriculture, flavonoid intake was calculated by multiplying a weight proportional to the frequency of consumption of each food item by the amount of flavonoids contained and summing across food sources from the FFQs [12,13]. We adjusted for total energy intake using the nutrient residual method [14]. In our cohorts, the validity of FFQ assessment of flavonoid intake was high, with a correlation coefficient of 0.77 in men and 0.74 in women, when compared with the intake assessed by two 7-d dietary records, as previously described [13]. We defined the postdiagnostic intake based on the first FFQ collected at least 6 mo after CRC diagnosis to minimize the influence of active cancer treatment on diet. Prediagnostic intake was based on the latest available FFQ prior to CRC diagnosis. The median time between CRC diagnosis and the assessment of postdiagnostic diet was 2.4 y (interquartile range: 1.4–3.3 y). The alternate Healthy Eating Index (AHEI) was calculated on foods and nutrients to measure adherence to a healthy eating pattern that has been linked to a lower risk of chronic diseases [15].

Assessment of covariates

We assessed covariates based on the information from the follow-up questionnaires and medical records, including age at diagnosis, year of diagnosis, race (white and nonwhite), postmenopausal hormone use (no, yes, and unspecified), tumor differentiation (well differentiated, moderately differentiated, poorly differentiated, and unspecified), tumor subsite (proximal colon including the cecum, ascending colon, and transverse colon; distal colon including the descending colon and sigmoid colon; rectum; and unspecified), cancer stage (I, II, III, and unspecified), BMI (kg/m²), physical activity [metabolic equivalent (MET) h/wk], regular aspirin use, pack-years of smoking, and AHEI (in quartiles). Race was self-reported by the study participants. Details of the covariate assessment were documented in prior publications [16,17]. We also calculated from the FFQs the intake of other dietary factors that had been associated with CRC survival in prior studies, including alcohol, vitamin D, folate, calcium, marine omega-3 PUFAs, and dietary fiber [12,16,18].

Statistical analysis

Each participant contributed person-time of follow-up from the date of the return of the postdiagnostic FFQs to the date of death or the end of follow-up (31 December, 2019, for the NHS and the HPFS), whichever came first. We used time since diagnosis as the time scale, accounting for left truncation due to between-patient variation in the timing of postdiagnostic assessment [19]. The primary endpoint of the study was CRC-specific mortality, for which death from other causes was censored, and the secondary endpoint was all-cause mortality.

We used the Cox proportional hazards regression model to calculate the HRs and 95% CIs of mortality associated with quintiles of flavonoid intake. We calculated P for trend using the median intake of each quintile of flavonoids as a continuous variable in the model. As Seaman and White describe methods for a separate cohort study, we applied inverse probability weighting (IPW) in the Cox models to reduce potential selection bias due to the exclusion of patients who had missing postdiagnostic dietary data from the study [20]. IPW weighted complete cases by the inverse of the probability of having provided the postdiagnostic dietary data estimated by a multivariable logistic regression model (see details in the Supplementary material) [9]. To better control confounding, we stratified the model by age, sex, and cancer stage. The multivariable analysis further adjusted for prediagnostic flavonoid intake and other potential confounders, including the year of diagnosis; race; postmenopausal hormone use; tumor differentiation; tumor subsite; BMI; AHEI; regular aspirin use; physical activity; pack-years of smoking; alcohol consumption; and intake of total vitamin D, fiber, calcium, marine omega-3 PUFA, and folate after diagnosis. To assess the influence of individual lifestyle factors, we examined their correlations with flavonoid intake and performed a sensitivity analysis by excluding each of the covariates from our multivariable model. We tested the proportional hazards assumption by including the interaction term between flavonoid intake and follow-up time in the model and did not find evidence of a violation of this assumption.

We used restricted cubic splines to evaluate the dose–response relationship between flavonoid consumption and mortality [21]. We also examined the association of prediagnostic intake with mortality, performed cohort (sex)–specific analysis, and tested for interaction by cohort by including a cohort-exposure interaction term in the model. Finally, we examined the major foods that contributed to at least 5% of flavonoid intake in our study population and assessed the association between the major food contributor of flavonoids and mortality. All analyses were performed using SAS 9.4. All P values were 2-sided. To account for multiple testing, we considered a P value of <0.005 as statistically significant and that of <0.05 as suggestively significant.

Results

Basic characteristics of patients with CRC

Among 2552 patients with incident stage I–III CRC, the mean age at diagnosis (SD) was 68.7 (9.4) y, and 2464 participants (96%) self-reported as white. During 31,026 person-y of follow-up, we documented 1689 deaths, of which 327 were due to CRC. The mean (SD) intake of total flavonoids was 354 (291) mg per day. Among the major flavonoid subclasses, proanthocyanins accounted for 62% of the total flavonoid intake, flavan-3-ols and flavanones each accounted for 13%, flavonols and anthocyanins for 5.4%, and flavones for 0.7%. Pre- and postdiagnostic intakes of total and subclasses of flavonoids were modestly correlated (Spearman’s correlation coefficient r = 0.55 for total flavonoids; Supplemental Table 1).

As shown in Table 1, patients with higher flavonoid consumption tended to have a healthier diet and lifestyle profile, including higher physical activity; higher intake of dietary fiber, folate, and vitamin D; and less alcohol consumption. In contrast, no differences were found in cancer subsite, differentiation, or stage across quintiles of total flavonoid intake. Postdiagnostic total and subclasses of flavonoid intake weakly correlated with other dietary and lifestyle covariates (Spearman’s correlation coefficient r ≤ 0.40; Supplemental Table 2).

TABLE 1.

Age-standardized characteristics of patients with colorectal cancer at diagnosis according to quintiles of total flavonoid intake (n = 2552)¹

	Quintile 1 (n = 511)	Quintile 2 (n = 514)	Quintile 3 (n = 508)	Quintile 4 (n = 513)	Quintile 5 (n = 506)
Male, %	32	31	31	30	30
Age, yr	67.4	69.2	68.6	69.6	68.9
BMI, kg/m2	26.4	26.3	26.4	26.2	26.2
Physical activity, MET-h/wk	13.5	17.6	18.6	19.6	20.8
Pack-years of smoking	25.6	16.0	15.5	11.9	15.6
Current smokers, %	11	6	5	4	5
Regular use of aspirin, %2	52	53	52	54	48
Food and nutrient intake
Total fiber, g/d	17.7	20.3	21.7	22.7	22.3
Total folate, μg/d	606	646	667	709	703
Calcium, mg/d	1120	1170	1232	1266	1144
Vitamin D, IU/d	533	539	567	592	598
Marine ω-3 PUFAs, mg/d3	246	265	310	328	320
Alcohol, g/d	9.7	7.3	7.0	6.8	6.4
Tea, cup/wk	0.1	0.4	1.0	2.7	12
Total flavonoids, mg/d	118	199	273	386	797
Proanthocyanins, mg/d	63	109	155	237	546
Flavan-3-ols, mg/d	11	17	25	45	136
Flavanones, mg/d	26	43	55	53	52
Flavonols, mg/d	11	15	17	20	34
Anthocyanins, mg/d	6.3	12	17	27	31
Flavones, mg/d	1.6	2.2	2.7	2.8	2.7
Cancer subsite, %
Proximal colon	50	41	45	46	43
Distal colon	27	30	29	32	30
Rectum	19	25	23	19	22
Unspecified	4	5	3	3	5
Cancer differentiation, %
Well differentiated	12	14	14	17	14
Moderately differentiated	61	59	61	55	59
Poorly differentiated	14	13	10	12	15
Unspecified	13	14	15	16	12
Cancer stage, %
I	33	37	31	34	31
II	33	28	33	28	34
III	20	23	24	23	26
Unspecified	13	13	12	15	9

IU, international unit; MET, metabolic equivalent; ω-3 PUFA, omega-3 PUFA.

¹Quintiles are created in women and men separately. Means are calculated for continuous variables. All variables are age-standardized except age.

²Regular users are defined as those who use at least 2 standard (325 mg) tablets of aspirin per week.

³Marine ω-3 PUFAs include EPA, DHA, and DPA.

Association of total and flavonoid subclass consumption with mortality

Total flavonoid intake was associated with lower CRC-specific and all-cause mortality in the age-, sex-, and stage-adjusted model (P ≤ 0.01), but this association was attenuated after multivariable adjustment (Table 2). The multivariable HRs (95% CI) per 1-SD increment (300 mg) were 0.85 (0.72–1.01; P = 0.07) for CRC-specific mortality and 0.95 (0.88–1.03; P = 0.20) for all-cause mortality. Spline analysis detected a nonlinear relationship between postdiagnostic total flavonoid intake and all-cause mortality, and the beneficial association did not emerge until the intake reached ∼800 mg/d (P = 0.01 for nonlinearity; Supplemental Figure 2). The prediagnostic intake of total flavonoids was suggestively associated with a lower risk of CRC-specific and all-cause mortality (P = 0.04 and 0.02, respectively; Supplemental Table 3).

TABLE 2.

Association of postdiagnostic consumption of total and flavonoid subclasses with mortality among colorectal cancer patients (n = 2552)¹

A	Quintile 1	Quintile 2	Quintile 3	Quintile 4	Quintile 5	HR (95% CI) per 1 SD	P for linear trend	P for nonlinearity
Total flavonoid
Intake, median (IQR), mg/d	120 (91–145)	200 (181–217)	271 (251–294)	382 (348–421)	668 (545–909)
CRC-specific mortality
No. of events (n = 327)	72	67	66	66	56
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.98 (0.73–1.31)	0.96 (0.72–1.28)	0.90 (0.67–1.21)	0.71 (0.52–0.96)	0.82 (0.70–0.96)	0.01	NA
MV-adjusted HR (95% CI)3	Reference	1.17 (0.86–1.60)	1.15 (0.84–1.56)	1.18 (0.85–1.62)	0.80 (0.56–1.13)	0.85 (0.72–1.01)	0.07	NA
All-cause mortality
No. of events (n = 1689)	373	347	331	309	329
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.91 (0.79–1.03)	0.88 (0.77–1.01)	0.77 (0.67–0.88)	0.80 (0.70–0.91)	0.89 (0.83–0.95)	<0.001	NA
MV-adjusted HR (95% CI)3	Reference	1.07 (0.94–1.23)	1.06 (0.92–1.22)	0.97 (0.84–1.13)	0.96 (0.82–1.12)	0.95 (0.88–1.03)	0.20	0.01
Proanthocyanins
Intake, median (IQR), mg/d	57 (42–70)	107 (96–117)	158 (143–175)	236 (210–267)	456 (354–631)
CRC-specific mortality
No. of events (n = 327)	75	68	58	68	58
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.85 (0.64–1.14)	0.79 (0.59–1.06)	0.89 (0.67–1.18)	0.69 (0.51–0.93)	0.85 (0.73–0.98)	0.03	NA
MV-adjusted HR (95% CI)3	Reference	1.04 (0.77–1.41)	0.90 (0.66–1.23)	1.14 (0.84–1.55)	0.74 (0.53–1.03)	0.86 (0.72–1.01)	0.07	NA
All-cause mortality
No. of events (n = 1689)	377	350	324	306	332
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.89 (0.78–1.01)	0.81 (0.71–0.93)	0.77 (0.67–0.88)	0.82 (0.72–0.93)	0.92 (0.86–0.98)	0.01	NA
MV-adjusted HR (95% CI)3	Reference	1.02 (0.89–1.17)	0.98 (0.85–1.13)	0.98 (0.84–1.13)	0.98 (0.84–1.14)	0.98 (0.91–1.06)	0.62	0.002
Flavan-3-ols
Intake, median (IQR), mg/d	8.4 (6.4–10)	15 (13–17)	25 (22–28)	43 (37–54)	110 (78–177)
CRC-specific mortality
No. of events (n = 327)	73	71	53	69	61
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.86 (0.65–1.15)	0.70 (0.51–0.94)	0.90 (0.68–1.19)	0.67 (0.50–0.90)	0.84 (0.71–0.98)	0.03	NA
MV-adjusted HR (95% CI)3	Reference	1.00 (0.74–1.34)	0.81 (0.59–1.12)	1.02 (0.76–1.37)	0.71 (0.51–0.99)	0.83 (0.69–0.99)	0.04	NA
All-cause mortality
No. of events (n = 1689)	375	362	302	306	344
Age, sex, stage-adjusted HR (95% CI)2	Reference	1.04 (0.92–1.19)	0.75 (0.65–0.86)	0.85 (0.74–0.98)	0.80 (0.71–0.92)	0.89 (0.83–0.96)	0.002	NA
MV-adjusted HR (95% CI)3	Reference	1.18 (1.03–1.35)	0.91 (0.79–1.06)	1.00 (0.86–1.15)	0.91 (0.78–1.05)	0.91 (0.84–0.99)	0.02	0.009
Flavanones
Intake, median (IQR), mg/d	3.9 (1.1–6.8)	17 (14–22)	37 (32–43)	61 (52–66)	97 (83–120)
CRC-specific mortality
No. of events (n = 327)	67	50	69	78	63
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.69 (0.50–0.95)	0.90 (0.67–1.21)	1.05 (0.79–1.40)	0.83 (0.62–1.12)	1.00 (0.88–1.12)	0.93	NA
MV-adjusted HR (95% CI)3	Reference	0.72 (0.52–1.00)	0.98 (0.72–1.32)	1.13 (0.83–1.54)	0.96 (0.71–1.32)	1.05 (0.92–1.20)	0.45	NA
All-cause mortality
No. of events (n = 1689)	346	296	325	348	374
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.72 (0.63–0.83)	0.77 (0.67–0.89)	0.75 (0.66–0.86)	0.80 (0.71–0.92)	0.95 (0.90–1.00)	0.06	<0.001
MV-adjusted HR (95% CI)3	Reference	0.71 (0.62–0.82)	0.84 (0.73–0.97)	0.86 (0.74–0.98)	0.89 (0.78–1.02)	1.00 (0.95–1.06)	0.91	NA
Flavonols
Intake, median (IQR), mg/d	7.8 (6.6–8.9)	12 (11–13)	16 (15–17)	22 (20–23)	34 (29–42)
CRC-specific mortality
No. of events (n = 327)	74	82	59	51	61
Age, sex, stage-adjusted HR (95% CI)2	Reference	1.00 (0.77–1.32)	0.67 (0.50–0.91)	0.56 (0.41–0.77)	0.79 (0.60–1.06)	0.86 (0.75–0.98)	0.02	NA
MV-adjusted HR (95% CI)3	Reference	1.10 (0.83–1.47)	0.79 (0.58–1.07)	0.66 (0.47–0.93)	0.89 (0.64–1.24)	0.91 (0.78–1.05)	0.21	NA
All-cause mortality
No. of events (n = 1689)	395	360	328	294	312
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.93 (0.81–1.05)	0.79 (0.69–0.90)	0.69 (0.60–0.79)	0.82 (0.71–0.93)	0.90 (0.85–0.96)	<0.001	NA
MV-adjusted HR (95% CI)3	Reference	1.07 (0.93–1.22)	0.91 (0.79–1.04)	0.82 (0.70–0.95)	0.99 (0.85–1.15)	0.98 (0.91–1.04)	0.48	NA
Anthocyanins
Intake, median (IQR), mg/d	2.3 (1.1–3.2)	5.9 (4.8–7.2)	12 (10–14)	19 (17–22)	40 (31–66)
CRC-specific mortality
No. of events (n = 327)	68	81	59	68	51
Age, sex, stage-adjusted HR (95% CI)2	Reference	1.08 (0.82–1.43)	0.83 (0.62–1.11)	1.07 (0.80–1.43)	0.85 (0.62–1.17)	0.90 (0.76–1.08)	0.26	<0.05
MV-adjusted HR (95% CI)3	Reference	1.26 (0.94–1.69)	1.04 (0.76–1.42)	1.37 (0.99–1.87)	1.12 (0.79–1.59)	1.03 (0.85–1.25)	0.79	0.002
All-cause mortality
No. of events (n = 1689)	381	367	340	317	284
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.90 (0.79–1.02)	0.86 (0.76–0.98)	0.89 (0.78–1.02)	0.91 (0.80–1.05)	0.98 (0.90–1.06)	0.58	NA
MV-adjusted HR (95% CI)3	Reference	1.07 (0.93–1.22)	1.07 (0.94–1.23)	1.16 (1.00–1.33)	1.29 (1.10–1.51)	1.16 (1.07–1.27)	<0.001	<0.001
Flavones
Intake, median (IQR), mg/d	0.6 (0.4–0.8)	1.3 (1.1–1.4)	2.0 (1.8–2.2)	2.8 (2.6–3.1)	4.6 (3.9–5.9)
CRC-specific mortality
No. of events (n = 327)	64	69	57	73	64
Age, sex, stage-adjusted HR (95% CI)2	Reference	1.11 (0.83–1.49)	0.83 (0.61–1.13)	1.03 (0.76–1.39)	0.98 (0.72–1.32)	0.97 (0.85–1.12)	0.68	0.004
MV-adjusted HR (95% CI)3	Reference	1.20 (0.89–1.62)	0.95 (0.68–1.31)	1.36 (0.98–1.87)	1.23 (0.88–1.72)	1.09 (0.94–1.27)	0.27	<0.001
All-cause mortality
No. of events (n = 1689)	335	336	325	338	355
Age, sex, stage-adjusted HR (95% CI)2	Reference	0.98 (0.86–1.13)	0.90 (0.78–1.03)	0.88 (0.77–1.01)	0.91 (0.79–1.04)	0.95 (0.89–1.01)	0.07	NA
MV-adjusted HR (95% CI)3	Reference	1.08 (0.94–1.24)	1.04 (0.90–1.20)	1.11 (0.96–1.28)	1.10 (0.94–1.27)	1.03 (0.96–1.10)	0.41	NA

CRC, colorectal cancer; MV, multivariable; IQR, interquartile range; NA, data not available.

¹Postdiagnostic intake was assessed at least 6 mo after diagnosis to minimize the influence of active treatment.

²Cox proportional hazards regression model stratified by age groups at diagnosis (<60, 60–64, 65–69, 70–74, and 75 y), sex, and cancer stage (I, II, III, and unspecified), with additional adjustment for age at diagnosis (continuous).

³Further adjusted for intake of total flavonoids before diagnosis (continuous); year of diagnosis (continuous); race (white, nonwhite); postmenopausal hormone use (no, yes, and unspecified); BMI (<23.0, 23.0–24.9, 25.0–27.4, 27.5–29.9, and 30.0 kg/m²); regular aspirin use; physical activity (women: <5.0, 5.0–11.4, 11.5–21.9, and 22 metabolic equivalent h/wk; men: <7.0, 7.0–14.9, 15.0–24.9, and 25.0 metabolic equivalent h/wk); pack-years of smoking (0, 1–15, 16–25, 26–45, and >45 y); alcohol consumption (<0.15, 0.15–1.9, 2.0–7.4, and 7.5 g/d); tumor differentiation (well differentiated, moderately differentiated, poorly differentiated, and unspecified); tumor subsite (proximal colon, distal colon, rectum, and unspecified); intake of total vitamin D, fiber, calcium, marine ω-3 PUFA, and folate after diagnosis (continuous); and alternate healthy eating index (in quartiles).

Among the 6 major subclasses of flavonoids, a higher intake of flavan-3-ols was suggestively associated with lower mortality (Table 2), with a multivariable HR (95% CI) per 1-SD (62 mg) increment of 0.83 (0.69–0.99; P = 0.04) for CRC-specific mortality and that of 0.91 (0.84–0.99; P = 0.02) for all-cause mortality. Similar associations were found for the prediagnostic intake of flavan-3-ols (Supplemental Table 3). In the spline analysis, a linear relationship was found for CRC-specific mortality (P = 0.01 for linearity) and a nonlinear relationship for all-cause mortality (P = 0.009 for nonlinearity; Figure), although the nonlinearity occurred only to the very high intake end (>100 mg/d). In sensitivity analysis, exclusion of each of the dietary and lifestyle covariates from the multivariable model did not essentially change the results, indicating a collective confounding effect by these factors (Supplemental Table 4). We also performed a sensitivity analysis by excluding 320 cases with unspecified tumor stage, and the results remained essentially the same (Supplemental Table 5). The cohort (sex)–specific analysis found that flavan-3-ols were more strongly associated with lower CRC-specific and all-cause mortality in women than in men, although P for interaction by sex was >0.05 (Supplemental Table 6).

FIGURE.

Dose–response relationship between the intake of flavan-3-ols after diagnosis and mortality among patients with colorectal cancer. Colorectal cancer–specific mortality (A) and all-cause mortality (B). Dashed lines represent the 95% CIs of the HR. The multivariable model was adjusted for the same set of covariates as in Table 2. For flavan-3-ols intake and colorectal cancer–specific mortality (A), no spline variables were selected into the model and the relationship was linear with P = 0.01 for linearity; for all-cause mortality (B), the relationship was nonlinear with P = 0.009 for nonlinearity and P < 0.001 for the overall significance.

No statistically significant associations were found for other flavonoid subclasses after diagnosis, including proanthocyanin, flavanones, flavonols, or flavones (Table 2). Interestingly, anthocyanins were associated with higher all-cause but not CRC-specific mortality, with HRs (95% CI) per 1-SD increment of 1.16 (1.07–1.27; P < 0.001) and 1.03 (0.85–1.25; P = 0.79), respectively. When assessed by causes of death, a positive association was found for both cardiovascular death and other mortality, with the HRs (95% CI) of 1.22 (1.00–1.49; P < 0.05) and 1.28 (1.11–1.47; P < 0.001), respectively (Supplemental Table 7).

Association of flavonoid-rich foods with mortality

We also examined the major foods that contributed to at least 5% of flavonoid intake in our study population (Supplemental Figure 3). The main food contributors to total flavonoids were tea (36%), apples (14%), blueberries (6%), and orange juice (6%). For subclasses of flavonoids, tea was the predominant source for flavan-3-ols (60%) and proanthocyanins (43%). Orange juice was the predominant source of flavones (56%) and flavanones (32%). Blueberries were the main source of anthocyanins (48%), and onions were the main source of flavonols (24%).

As the major contributor to flavan-3-ols, we assessed tea consumption in relation to mortality (Table 3). Higher tea consumption was associated with a lower risk of CRC-specific and all-cause mortality, with multivariable HRs per 1 cup/d of tea of 0.86 (0.75–0.99; P = 0.03) and 0.90 (0.85–0.96; P < 0.001), respectively. These associations were attenuated to null after further adjusting for flavan-3-ol intake, suggesting that flavan-3-ols were the major driver for the beneficial association of tea intake.

TABLE 3.

Association of postdiagnostic consumption of tea with mortality among patients with colorectal cancer (n = 2500)

	Nondrinkers (n = 1178)	>0 to 1 cup/d (n = 1099)	>1 cups/d (n = 223)	HR (95% CI) per 1 cup/d	P for linear trend
CRC-specific mortality
No. of events (n = 321)	151	144	26
Age, sex, stage-adjusted HR (95%  CI)1	Reference	0.94 (0.77–1.15)	0.79 (0.55–1.14)	0.90 (0.80–1.01)	0.07
MV-adjusted HR (95% CI)2	Reference	0.92 (0.75–1.13)	0.74 (0.49–1.13)	0.86 (0.75–0.99)	0.03
MV- + postdiagnostic flavan-3-ols-adjusted HR (95% CI)	Reference	1.03 (0.82–1.29)	1.52 (0.77–2.98)	1.09 (0.79–1.48)	0.61
MV- + postdiagnostic proanthocyanin-adjusted HR (95% CI)	Reference	0.97 (0.78–1.21)	1.06 (0.58–1.93)	0.89 (0.70–1.15)	0.38
All-cause mortality
No. of events (n = 1651)	781	721	149
Age, sex, stage-adjusted HR (95% CI)1	Reference	0.90 (0.82–0.98)	0.80 (0.68–0.94)	0.92 (0.88–0.97)	0.001
MV-adjusted HR (95% CI)2	Reference	0.93 (0.85–1.03)	0.79 (0.66–0.94)	0.90 (0.85–0.95)	<0.001
MV- + postdiagnostic flavan-3-ols-adjusted HR (95% CI)	Reference	0.97 (0.88–1.07)	1.03 (0.78–1.37)	0.92 (0.81–1.04)	0.17
MV- + postdiagnostic proanthocyanin-adjusted HR (95% CI)	Reference	0.94 (0.85–1.04)	0.85 (0.65–1.10)	0.83 (0.75–0.93)	<0.001

CRC, colorectal cancer; MV, multivariable.

¹Cox proportional hazards regression model stratified by age groups at diagnosis, sex, and cancer stage, with additional adjustment for age at diagnosis.

²Further adjusted for the intake of total flavonoids before diagnosis; year of diagnosis; race; postmenopausal hormone use; BMI; regular aspirin use; physical activity; pack-years of smoking; alcohol consumption; tumor differentiation; tumor subsite; and intake of total vitamin D fiber, calcium, marine ω-3 PUFA, and folate after diagnosis; and alternate healthy eating index.

Discussion

To the best of our knowledge, this study is the largest prospective study to examine the association of postdiagnostic total flavonoids and major subclasses with mortality in patients with CRC. Among the 6 major subclasses of flavonoids, a higher intake of flavan-3-ols was suggestively associated with lower CRC-specific and all-cause mortality. A similar beneficial association was found for tea, which was the major source of flavan-3-ols and proanthocyanins. These findings suggest the survival-improving benefit of a higher intake of flavan-3-ols in survivors of CRC and provide novel evidence for dietary modifications in cancer management [22].

Flavonoids are polyphenolic compounds found in various fruits, vegetables, and plants. Owing to their capability of interfering with epigenetic signaling cascades responsible for tumorigenesis and metastasis, these polyphenols have been suggested to reduce cancer recurrence [23]. However, epidemiologic evidence remains limited. Only one prior study has assessed the association between flavonoid intake and the survival of patients with CRC. That study collected prediagnostic intake data only and found no association between the intake of total flavonoids or flavonoid subclasses and CRC recurrence among 409 patients [24]. This is in line with our null findings for the association of prediagnostic intake with CRC-specific mortality, and suggests that the postdiagnostic period may be the relevant window when flavonoids can exert beneficial effects on CRC survival.

Through a detailed dose–response analysis, we found that the beneficial association of flavan-3-ols with CRC-specific mortality did not emerge until ∼80 mg/d intake. According to the data from the USDA, each 100 mL of brewed black tea with 1% infusion (1 g of tea leaves/100 mL of boiling water) contains ∼115 mg of flavan-3-ols [25]. This result is consistent with our observation that patients who consumed 1–3 cups of tea (8 oz. per cup) already had a lower CRC-specific mortality than nondrinkers. Although the content of flavan-3-ols is known to vary by the type of tea [26], in USA, black tea was the predominant type during most of the follow-up time of our cohorts [27]. Moreover, the dose–response relationship may also explain the stronger association of flavan-3-ols and proanthocyanins with reduced mortality in women than in men, because in our cohorts, women had a wider range of intake than men (e.g., median intake flavan-3-ols ranged from 7.9 mg/d in quintile 1 to 116 mg/d in quintile 5 in women compared with 11–101 mg/d in men).

Mechanistic data indicate the anti-CRC properties of flavan-3-ols and proanthocyanins [5,28]. In vitro studies indicated that some flavan-3-ols, such as catechins, reduced CRC progression by inhibiting the nuclear factor-kappa B pathway and suppressing the overexpression of the inflammatory enzyme prostaglandin-endoperoxide synthase 2 [29,30]. Epigallocatechin gallate (EGCG) is the main flavan-3-ol found in green tea [31] and may protect against CRC [32]. Mukherjee et al. showed that EGCG reverted the inflammatory response by inhibiting cytokines, chemokines, and metalloproteinase activity [33]. EGCG administration to colon cancer cells reduced the methylation of CRC-related genes [34]. Moreover, increasing data indicate the role of gut microbes in metabolizing dietary flavonoids and moderating biological effects [35,36]. Green tea increases the abundance of short-chain fatty acid–producing bacteria, such as Lachnospiraceae, Bifidobacteriaceae, and Ruminococcaceae, and decreases the abundance of potentially proinflammatory bacteria, such as Prevotella, which is increased in patients with CRC [37]. Procyanidins, which are primarily composed of catechin and epicatechin [38], also have a wide range of beneficial properties, such as antioxidant, antitumor, and immunostimulating effects [39]. Oligomeric proanthocyanins decrease xenograft tumor growth and inhibit the formation of organoids in CRC [40]. Taken together, these mechanistic data support our epidemiological findings. In particular, given the increasing data indicating the importance of the gut microbiome in cancer treatment and prognosis [41] and the role of flavonoids in modulating the gut microbial composition and function [42], further studies are needed to investigate the interplay between flavan-3-ols and procyanidins and the gut microbiome in patients with CRC, and the translational potential for novel therapeutics.

Surprisingly, we found a positive association between anthocyanin intake and all-cause mortality, including both cardiovascular and other mortality. In contrast to flavan-3-ols and procyanidins whose predominant food source is tea, anthocyanins were mainly consumed from fruits, such as blueberries, strawberries, and apples (Supplemental Figure 3). Epidemiological studies showed that regular, moderate intake of blueberries and/or anthocyanins was associated with a reduced risk of cardiovascular disease and death [43]. Thus, given the numerous tests conducted in our study, our findings may be due to chance, indicating the need for further studies.

Our study has several strengths, including the prospective design, large sample size, relatively long follow-up time, established validity of the FFQs, and detailed data on confounding factors [13]. Several limitations are also notable. First, information on CRC treatment and recurrence was not systematically collected. However, the stage may be a good proxy for CRC treatment, especially in our cohorts, with all participants being health professionals. We excluded patients with stage IV disease and performed stage-stratified analysis to minimize the potential confounding effect of treatment. Second, as an observational study, we cannot exclude the possibility of residual confounding. However, we adjusted for a variety of clinical and lifestyle factors that may influence CRC prognosis, and our sensitivity analysis indicated the robustness of our findings to individual confounding factors. Third, the generalizability of our findings may be limited because the study participants were predominantly white and all health professionals. However, the homogeneity of our study population enhances internal validity because it reduces the variability in unmeasured confounding factors (e.g., cancer treatment and palliative care). Fourth, measurement error in dietary assessment by FFQs is inevitable due to the incompleteness of the food composition table, although the validity of FFQ assessment of flavonoid intake has been established. Finally, multiple testing may amplify the probability of false-positive findings. To address that, we considered a more stringent P value of <0.005 as statistically significant, and in addition to statistical significance, we considered biological plausibility and coherence (e.g., between the food and nutrient analysis and between the primary and sensitivity analyses) in the interpretation of our findings [44]. Furthermore, we acknowledge the exploratory nature of our study and emphasize the need for replication.

In summary, we found that higher consumption of flavan-3-ols and tea after CRC diagnosis was associated with better survival in patients with nonmetastatic CRC. A positive association was found between anthocyanidin intake and cardiovascular and all-cause mortality, which was likely due to chance and needs to be assessed in further studies. More prospective studies are warranted to replicate and confirm our findings, and further laboratory studies are necessary to better understand the biological mechanisms.

The authors acknowledge the contribution to this study from central cancer registries supported by the Centers for Disease Control and Prevention’s National Program of Cancer Registries and/or the National Cancer Institute’s Surveillance, Epidemiology, and End Results (SEER) Program. Central registries may also be supported by state agencies, universities, and cancer centers. Participating central cancer registries are as follows: Alabama, Alaska, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Hawaii, Idaho, Indiana, Iowa, Kentucky, Louisiana, Massachusetts, Maine, Maryland, Michigan, Mississippi, Montana, Nebraska, Nevada, New Hampshire, New Jersey, New Mexico, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Puerto Rico, Rhode Island, Seattle SEER Registry, South Carolina, Tennessee, Texas, Utah, Virginia, West Virginia, and Wyoming.

The authors’ responsibilities were as follows – SS, MS: have full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis; EBR, KW, KN: collated the data; KW, RZ: performed the statistical analysis; AC, ELG, SO, JAM, KN, ATC: critically revised the manuscript for important content; and all authors read and approved the final manuscript.

KW is currently a stakeholder and employee of Vertex Pharmaceuticals. The study was not funded by this entity. The remaining authors report no conflicts of interest.

Funding

This work was supported by U.S. NIH grants (P01 CA87969 and UM1 CA186107, to MJS; U01 CA167552, to WCW and LAM; R35 CA197735, to SO; R00CA215314, U01CA261961, and R01CA263776, to MS).

The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

Study data is not publicly available due proprietary restrictions but is available upon request.

Appendix A. Supplementary data

The following is the Supplementary data to this article: