1. Introduction
Cancer is a disease resulting from the uncontrolled growth and division of abnormal cells in the body and is the leading cause of death and an important obstacle to extending life expectancy worldwide [
1]. It was estimated that cancer ranked as the first or second leading cause of death in the population aged below 70 years in more than 60% of countries and ranked third or fourth in the remaining countries in 2019 [
1]. There were an estimated 19.3 million new cases and 10 million cancer deaths in 2020 [
1]. Primary prevention is a pivotal strategy with which to decrease the growing burden of cancer. However, establishing high-quality cancer prevention remains a challenge [
1]. Nearly one-third of cancer-related deaths could be prevented using balanced dietary improvements in the US [
2]. The adequate consumption of fruit and vegetables presented the considerable potential to reduce the occurrence and progression of cancer [
3,
4]. Dietary flavonoids are a group of natural polyphenols, present in fruit, cereal, vegetables, tea, and red wine [
2]. Two benzene rings connected by a heterocyclic pyran ring consist of the basic—structure of flavonoids [
5]. Flavonoids are subcategorized into six subclasses based on the linking arrangement and the saturation of the double bond on the pyran structure, including anthocyanins, flavan-3-ols, flavanones, flavones, flavonols, and isoflavones [
5]. Studies in vitro demonstrated that flavonoids could induce apoptosis and inhibit cell proliferation and metastasis by targeting the key molecules and signaling pathways in various tumor cells [
6,
7,
8,
9]. However, the results of epidemiological studies on the association between dietary flavonoid intake and cancer mortality remain inconsistent. Some previous studies on total flavonoid intake and all-cause cancer mortality did not reveal the inverse association [
10,
11]. More recent research demonstrated a beneficial association between total dietary flavonoid intake and cancer mortality [
12,
13]. Moreover, whether people with an unhealthy lifestyle could benefit from increased dietary flavonoid intake needs to be elucidated.
To explore the association between dietary flavonoid intake and cancer-related mortality, we downloaded all the publicly available data in the Database of Flavonoid Values for USDA Food Codes and Flavonoid Intake Data Files from What We Eat in America (WWEIA), National Health, and Nutrition Examination Survey (NHANES) [
14]. Until 16 December 2022, only the data in 2007-2010 and 2017-2018 were released by the USDA [
14]. Therefore, our study explored the association between dietary flavonoid intake and cancer-related mortality in 2007-2010 and 2017-2018, which included all the publicly accessible data [
14]. We revealed that in comparison to being in the first quartile of dietary flavonol intake, being in the second, third, and fourth quartile was inversely associated with cancer-related mortality. Furthermore, potential protective effects of dietary flavonol intake against cancer death was observed, especially in former smokers, mild drinkers, and people who used to drink, which suggests that changing an unhealthy lifestyle is also important. A nomogram based on dietary flavonol intake is clinically adaptable to assessing cancer mortality among individuals.
2. Materials and Methods
NHANES is a program of studies designed to evaluate the health and nutritional status of adults and children in the US. NHANES is a program of the National Center for Health Statistics, which is part of the Centers for Disease Control and Prevention [
15]. The NHANES directors can be accessed at [
16]. The protocols for NHANES surveys were approved by the NCHS Ethics Review Board. The approval number for the survey year cycles 2007-2008 and 2009-2010 was the Continuation of Protocol #2005-06. The approval numbers for the survey year cycle 2017–2018 were the Continuation of Protocol #2011-17 (Effective through 26 October 2017) and Protocol #2018-01 (Effective beginning 26 October 2017) [
17]. All participants signed a written informed consent form.
All data in our study are publicly available and without personally identifiable information. All analyses were conducted following the relevant guidelines and regulations [
18]. The survey process consisted of two steps. Demographic and health-related information was collected in the homes of participants during the first interview. After that, a standardized physical examination was carried out in a mobile examination center two weeks later, as well as a blood draw,24-h dietary recall, and other investigations, such as laboratory analysis of urine, microbiome sampling of the oral cavity, and so on. We obtained 18,538 participants aged 18 years or above in the continuous NHANES cycles of 2007–2010 and 2017–2018 with complete mortality information: among them were 14,490 participants with complete data from the dietary flavonoid intake assessment.
2.1. Dietary Flavonoid Intake Assessment
The dietary flavonoid values in our study were obtained from the database of flavonoid values for USDA Survey Foods and Beverages (flavonoid database for short), which was established in 2003–2004 [
19]. The flavonoid database provides the flavonoid values in foods and beverages in the USDA Food and Nutrient Database for Dietary Studies (FNDDS) [
20] and corresponding dietary data from WWEIA [
21] and NHANES. The amounts of 29 flavonoids (mg/100 g) in each food/beverage were determined by the USDA Nutrient Data Laboratory [
22]. The dietary intake of flavonoids was calculated on days 1 and 2, including the six main flavonoid subclasses commonly consumed in the US diet, namely total anthocyanins (cyanidin, delphinidin, malvidin, pelargonidin, peonidin, and petunidin), total flavan-3-ols ((-)-epicatechin, (-)-epicatechin 3-gallate, (-)-epigallocatechin, (-)-epigallocatechin 3-gallate, (+)-catechin, (+)-gallocatechin, theaflavin, theaflavin-3,3′-digallate, theaflavin-3′-gallate, theaflavin-3-gallate, and thearubigins), total flavanones (eriodictyol, hesperetin, and naringenin), total flavones (apigenin and luteolin), total flavonols (isohamnetin, kaempferol, myricetin, and quercetin), total isoflavones (diadzein, genistein, and glycitein), and subtotal catechins ((-)-epicatechin, (-)-epicatechin 3-gallate, (-)-epigallocatechin, (-)-epigallocatechin 3-gallate, (+)-catechin, and (+)-gallocatechin). The retention factors for cooked foods were introduced in the estimation of the flavonoid amounts. For moist-heat cooking, a loss of 50% was applied to anthocyanidins, as well as one reduction of 15% to flavonols, flavan-3-ols, flavanones, and flavones. No retention factors were implemented for isoflavones and dry heat cooking [
20]. The association between individual flavonoid intake grouped by flavonoid subclasses was analyzed using the Pearson correlation method.
Based on the stratified and multistage probability sampling designed in the NHANES, we used the mean of the two-day intake of each flavonoid, as well as the weights “wtdr2d” constructed for participants who completed two days of dietary recall in making estimates representative of the US non-hospitalized population [
18].
2.2. Mortality Ascertainment
The updated follow-up date was 31st December 2019. A total of 14,092 participants with an available survival status were included in our study. The time in months from the household interview to mortality was assessed as the follow-up time. A death caused by cancer was ascertained by the National Death Index and International Classification of Diseases-10 C00-C97 as a malignant neoplasm. A potential nonlinear association between total flavonol intake and cancer mortality was evaluated using restricted cubic splines. The association between flavonoid intake and cancer mortality was assessed using Cox proportional hazards analysis.
2.3. Covariate Assessment
Covariates, i.e., age, ethnicity, education, marital status, poverty income ratio (PIR), smoking status, alcohol use, and physical activity (PA), were collected via questionnaires. Educational status was classified according to the number of years of education as <9 years, 9–12 years, and >12 years. Marital status was divided into partnered and unpartnered. The body mass index (BMI) was calculated as weight (kg)/height squared (m
2). The term “never” regarding the smoking status was defined as fewer than 100 cigarettes during life; “former” as more than 100 cigarettes during life and not at all currently smoking; and “now” as more than 100 cigarettes during life and smoking some days or every day. The classification of alcohol usage was as described in [
23]. The healthy eating index (HEI) was calculated based on the 2015 version using the sum of food intake on days 1 and 2 for each participant [
24]. The dietary inflammatory index (DII) was calculated as described in [
25]. We used the total time and the total metabolic equivalent (MET) of PA for one week.
Regarding the perspectives of disease diagnosis, hyperlipidemia was defined as meeting one of the following conditions: triglyceride ≧ 150 mg/dL, low-density lipoprotein ≧ 130 mg/dL, high-density lipoprotein < 140 ng/dL, or usage of lipid-lowering drugs. Cardiovascular disease was defined as having ever had a heart attack or stroke. Participants were diagnosed as having the chronic obstructive pulmonary disease (COPD) when meeting one of the following conditions: the value of forced expiratory volume at first second/forced vital capacity (FEV1/FVC) < 0.7 after the application of a bronchodilator; having been told you had emphysema; using COPD drugs selective phos-phodiesterase-4 inhibitors, mast cell stabilizers leukotriene modifiers, and inhaled corticosteroids. Participants were diagnosed as having asthma when meeting one of the following conditions: having been told you had asthma; had an asthma attack; the application of selective phos-phodiesterase-4 inhibitors, or mast cell stabilizers leukotriene modifiers and inhaled corticosteroids. Participants with both COPD and asthma were coded as ACO. Stroke history was defined as having ever had a stroke. Cancer history was defined as having ever had cancer. Average blood pressure was calculated as [
26], and participants were diagnosed as having hypertension when meeting one of the following conditions: systolic pressure≧ 140 mmHg or diastolic pressure ≧ 90 mmHg on three occasions. Participants were diagnosed as having type 2 diabetes mellitus (DM) when meeting one of the following conditions: having been told you had diabetes; HbA1c ≧ 6.5%, fasting glucose ≧ 7.0 mmol/L, glucose ≧ 11.1 mmol/L, oral glucose tolerance test ≧ 11.1 mmol/L; or usage of antidiabetic drugs. Missing covariates were imputed by the R package “mice”.
2.4. Flavonoid Supplement Identification
All dietary flavonoid supplement names and ingredients from the NHANES database were text-mined for key phrases to identify the products. Search terms used to identify flavonoid supplements and relevant results are shown in
Supplementary Table S1.
2.5. Urinary Isoflavone Metabolite Assessment
The levels of isoflavone metabolites, including daidzein (ng/mL), equol (ng/mL), genistein (ng/mL), and O-desmethylangolensin (O-DMA, ng/mL) in the urine, were obtained from the NHANES, which were measured by high-performance liquid chromatography-atmospheric pressure photoionization-tandem mass spectrometry. Only the data in the year 2007–2010 were publicly accessible in the survey years of our study. The association between urinary isoflavone metabolite levels and cancer-related mortality was analyzed using Cox proportional hazards analysis. Meanwhile, the special weights “wtsb2yr” or “wtsa2yr” for special urine examination were employed in the Cox analysis.
2.6. Statistical Analysis
The R packages “NHANESR” and “survey” were employed for data preparation and statistical analysis. Continuous variables are demonstrated as the mean ± standard deviation, median, and percentile range, and categorical variables are presented as percentages. The potential nonlinear association was evaluated using restricted cubic splines with the R package “rms”. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (95% CIs). A
p value < 0.05 was used as a cut-off for statistical significance. A weighted survey nomogram with the intake of flavonols was established and validated via a calibration curve for predicting the 10-year survival probability amongst participants using the R packages “rms” and “SvyNom” [
27]. All analyses were conducted using R software (version 4.1.3, the R Foundation for Statistical Computing).
4. Discussion
The global cancer burden is expected to increase by 47% to 28.4 million cases by 2040 in comparison with 2020 [
1]. Primary prevention prioritizes reducing the personal, clinical, and socioeconomic burden of cancer. It is estimated that approximately one-third of cancers can be attributed to diet, nutrition, and physical activity in developed countries [
29]. The adequate intake of fruit and vegetables, enriched in flavonoids, has been recommended to prevent cancers [
29]. Recent epidemiological studies have indicated that the intake of flavonoids is inversely associated with cancer mortality [
12,
13]. Flavonoids can be categorized into six main subclasses, namely anthocyanidins, flavan-3-ols, flavanones, flavones, flavonols, and isoflavones. Among them, flavonols, mainly including quercetin, kaempferol, myricetin, and isorhamnetin, are present in tea, onions, and berries.
Our results demonstrated that the inverse association between flavonol intake and cancer mortality is coincidental with Danish results [
13]. Interestingly, differing from the result that the intake of flavonols was inversely associated with the risk of lung cancer amongst male smokers [
30], the potential protective effects of flavonol were observed amongst former smokers, former drinkers, and mild-drinkers in our study, which may suggest that quitting smoking and alcohol consumption might have priority in reducing cancer death. It is well-documented that flavonols play an important role in the carcinogenesis and progression of cancer in vitro. Kaempferol can inhibit the epithelial-mesenchymal transition and induce apoptosis and cell cycle arrest in the G2/M phase by targeting phosphoinositide 3-kinase/protein kinase B signaling pathways [
31]. Quercetin is accountable for suppressing proliferation by causing cell cycle arrest in the G1 phase through targeting cyclins [
32,
33]. Moreover, quercetin at a high concentration can inhibit cell cycle progression from G0/G1 to G2/M [
25,
26]. Besides influencing the cell cycle, quercetin can induce apoptosis through pro-apoptotic PI3K/Akt and mitogen-activated protein kinase signaling pathways [
34,
35]. Similar to other flavonols, myricetin can induce apoptosis, interfere with the cell cycle, and inhibit cell proliferation in cancer cells [
36]. As an antioxidant, myricetin can scavenge the elevated free radicals and reactive oxygen species in cancer cells [
36,
37,
38]. Myricetin can restrain cancer progression by downregulating the levels of various inflammatory markers [
39]. However, what should be paid attention to is that flavonoids could exert pro-oxidant properties at high doses [
5,
40,
41]. Flavonoids containing phenolic rings, when oxidized by oxidases, produce cytotoxic phenoxy radicals; co-oxidize unsaturated lipids, nucleic acids, ascorbic acid, nicotinamide adenine dinucleotide, and glutathione; and lead to the formation of reactive oxygen species and mitochondrial toxicity. Quercetin and other flavonoids were reported to induce significant frequencies of sister chromatid exchange and micronuclei, as well as inhibit cell proliferation due to the pro-oxidant effects in special conditions [
42,
43]. In our study, we found numerous supplements were taken by people, which may not be regulated by the Food and Drug Administration and therefore the potential toxicity has not been fully assessed (
Supplementary Table S1). A study on the effect of flavonoid dietary supplements on cancer survival is warranted.
A comprehensive understanding of the mechanisms of flavonoids in cancer could help define new strategies for cancer management and prevention. The bioavailability of flavonoids is affected by dietary patterns, interaction with the food matrix, host genetics, intestinal microbiota, and the phase I and II metabolism in the liver [
44,
45,
46], which may explain the inconsistent epidemiological findings. Our observation that the intake of flavonols tended to influence cancer mortality differently amongst different ethnicities may be caused by host genetics and dietary patterns. In our study, we found a distinct flavonoid intake pattern among ethnicities. Except for the increased intake of flavanones, the intake of the remaining flavonoids was lower amongst black people than white people. The low intake of flavonoids in our study may reflect a high consumption of the Southern dietary pattern, constituted by a high consumption of added fats, fried food, organ meats, processed meats, and sugar-sweetened beverages in the US black population [
47]. The focus of primary prevention should be on those groups.
Moreover, although the flavonoids have low solubility and low bioavailability [
48,
49], certain intestinal bacteria can degrade flavonoids or their glycosides into smaller and more bioavailable metabolites which, when absorbed, can exert their physiological effects
in vivo [
45]. For example, physiological effects of isoflavones are mainly due to their metabolites, such as equol and O-DMA, produced by intestinal microbiota [
50]. In comparison to daidzein, equol has a stronger estrogenic effect, antioxidant property, and anti-androgenic effect [
51,
52]. However, neither the intake of isoflavone nor the urinary levels of equol and O-DMA were associated with cancer mortality in our study. It would be interesting to study their association in estrogen-related diseases.
Notably, we found that the intakes of peonidin, naringenin, and catechin are inversely associated with cancer mortality. Peonidin, present in red wine and berries, can inhibit lung cancer metastasis [
53]. Naringenin, present in citrus, decreases proliferation and induces apoptosis in various cancer types [
54]. Catechin, abundant in green tea, can inhibit cancer growth and progression through its anti-oxidant property, cell cycle modulation, receptor tyrosine kinase pathway downregulation, immune response regulation, and epigenetic modification control [
55].
The estimated amounts of individual flavonoids in plants and foods were reported to be influenced by nonrepresentative sampling, different cultivars, different growing, and processing conditions, and analytical bias [
22]. To generate the qualified data for the flavonoid database, five criteria were implemented in the flavonoid database of USDA, including the sampling plan, the number of samples, sample handling, analytical method, and analytical quality control [
22]. For instance, multiple samples and multiple detections were employed in the flavonoid database. Moreover, the flavonoid amounts by cultivars of each food/beverage were well documented in the flavonoid database [
22]. Furthermore, the retention factors were introduced to assess the loss of flavonoids during cooking, as described in the materials and methods section.
In addition, the effects of seasonal variation on amounts of flavonoids in plants or on total flavonoid intake may be not as significant as other factors [
56,
57]. It is reported that only blueberries demonstrated a seasonality [
22]. Since most foods/beverages are available the whole year for consumers, the seasonal variation might be due to the different cultivars or storage methods [
22].
Considering that an individual’s dietary structure may change over a lifetime, we compared changes in flavonoid intake across a study population spanning ten years. We found that the flavonoid consumption remained relatively stable, although there were significant differences in the intake of some flavonoids. We employed the mean of the two-day intake of each flavonoid as well as the weights “wtdr2d” constructed for participants who completed two days of dietary recall in making estimates representative of the US non-hospitalized population. Our results could reflect the change in flavonoid consumptions in population. Our study had several limitations. The observational analysis only revealed an association (rather than causality). Notably, dietary flavonoid intake did not include the intake of flavonoid supplements, contributing to the limitations of our results. As discussed above, there are possible health risks associated with excess flavonoid intake. Future studies should include dietary and supplemental flavonoid intakes.