Background

Flavonoids are a group of plant metabolites widely distributed in the plant kingdom with antioxidant properties, which can be classified into seven subgroups based on their chemical structure: flavanones, flavones, flavonols, flavan-3-ols, anthocyanidins, isoflavones and proanthocyanidins [1]. These compounds are present in small quantities in fruits, vegetables, tea, wine, nuts, seeds, herbs, spices, cocoa, soybean [2–4].

A wide spectrum of health benefits, such as antioxidant, anti-inflammatory, antibacterial, antithrombotic, anti-carcinogenetic properties have been reported for flavonoids [5].

Many epidemiological studies have reported inverse associations between the total flavonoid intake or the intake of specific classes of flavonoids and the incidence or mortality for cardiovascular (CV) diseases [6–20]. However, not all studies agreed about the cardio-protective effects of these compounds [21–27]. Many studies referred to specific foods, which are the main sources of flavonoids in different populations, such as tea [13, 28, 29], cocoa [28, 30], soy [17].

We have studied a population-based cohort of middle-aged adults, coming from a semi-rural area, where the consumption of some flavonoid-rich foods, such as cocoa and soybean is almost negligible, while the most important sources of flavonoids are fruits and red wine.

The primary objective of this study was establishing if the consumption of flavonoids was inversely associated with the CV risk evaluated after 12-year follow-up; the secondary aims were evaluating the associations between flavonoid intake and CV incidence, CV mortality, and all-cause mortality in our cohort.

Methods

All the Caucasian patients (n = 1,877), aged 45–64 years, of six family physicians were invited to participate in a metabolic screening in 2001–2003. These subjects were representative of the Local Health Units of the province of Asti (northwestern Italy) [31]. Exclusion criteria were: inability to go to the office of the family physician and to give the informed consent.

Of these, 1,658 (88.3%) subjects gave their written informed consent to participate and 219 patients declined. Both the participants and non-participants were similar to the resident population of a corresponding age range with respect to the percentage of males, level of education, prevalence of known diabetes, and residence in a rural area [31]. The study was approved by the local ethics committee. All procedures conformed to the principles of the Helsinki Declaration.

Methods

In the morning and after fasting, weight, height, waist circumference, and blood pressure were measured in the office of the family physicians. Glucose, insulin, total cholesterol, HDL-cholesterol, triglyceride, uric acid and high-sensitivity C-reactive protein (CRP) levels were determined. If the serum glucose value was ≥110 mg/dl, a second fasting glucose determination was performed. Two blood pressure measurements were performed with mercury sphygmomanometers and the appropriate cuff sizes after a 10-min rest in the sitting position, and the values reported are the means of the two measurements. The waist circumference was measured by a plastic tape meter at the level of the umbilicus. The measurements were performed by trained physicians holding a grant.

Patients completed the Minnesota Leisure Time Physical Activity questionnaire [32]. The physical activity level was calculated as the product of the duration and frequency of each activity (in hours/week), weighted by an estimate of the metabolic equivalent of the activity and summed for the activities performed.

From January to November 2014, the patients were submitted to a blood sample analysis and a follow-up visit by their family physicians. Information on the vital status of each patient and the causes of death of those who died was collected from the demographic files of the town of residence or death.

The laboratory methods have been described previously [31, 33]. All samples were run blindly.

Ascertainment of flavonoid intake

The semi-quantitative food-frequency questionnaire used in the EPIC (European Prospective Investigation into Cancer and Nutrition) studies was used for all subjects [34]. It assessed average frequency and portion intake of 148 foods consumed during the 12 months before the enrolment. For each food item, the participants had to mark if the food or dish was consumed or not during the previous year. For all food items consumed, the subjects should select their typical portion size with the help of photographs, the consumption frequency and the time period (day, week, month or year), which suited them best. Questions about the type of fat for cooking were also included. This tool has been previously validated [34]. A dietician, blinded to the study details, checked all questionnaires for completeness, internal coherence, and plausibility. In case of uncertain answers, the patients were interviewed by the dietician.

Each nutrient was adjusted for total energy, using the residual method [35]. The reliability of the reported energy intake was assessed by calculating the ratio of estimated energy intake to predicted basal metabolic rate, using age- and sex-specific formulas derived by Schofield [36]. Subjects with a ratio <0.88 were classified as under-reporters [37].

Dietary intake of total and subclasses of flavonoids were estimated by using the latest detailed food composition tables published by the US Department of Agriculture (USDA) on the seven major classes of flavonoids [2–4] and extended with information from an European database [38]. Merging of the databases gave a single data-file. Flavonoid intake was computed by multiplying the specific flavonoid content of the serving of each food item (mg aglicone equivalent/100 g food) by the daily consumption (g/day) of the selected food item. Estimated total intake of individual flavonoids was the sum of individual flavonoid intakes from all food sources reported in the questionnaire. Total flavonoid intake was calculated by summing up the seven subclasses (flavanones, flavones, flavonols, flavan-3-ols, anthocyanidins, isoflavones and proanthocyanidins), and were expressed as mg/day aglycones.

The contribution of each food to the total intake of subgroup and total flavonoids was calculated as a percentage; single foods were then grouped into large categories.

All flavonoid subgroups that were estimated, their respective compounds, and the main food sources are shown in Table 1.

Table 1 Flavonoid classes and compounds, and respective dietary intakes and main food sources in the whole cohort
Full size table

Definitions

Alcohol intake was assessed by multiplying the mean daily consumption of each beverage by its ethanol content, to give grams of alcohol/day. Moderate and heavy drinkers were considered in the case of consumption of ≤30 and >30 g/day alcohol, respectively, in line with Italian guidelines [39].

Diabetes mellitus was defined according to published recommendations [40]. Estrogen use included both contraceptive medications or estrogen replacement therapy. The use of nutritional supplements was infrequent in this cohort and was limited to multivitamin, iron, calcium or, less frequently, magnesium.

The CV risk score was estimated with the Framingham risk score [41]. The diagnosis of CV disease was based on documented events that were recorded by the family physician (i.e. angina, previous myocardial infarction, coronary artery by-pass graft or another invasive procedure to treat coronary artery disease, transient ischemic attack, stroke, gangrene, amputation, vascular surgery, intermittent claudication, absent foot pulses and abnormal brachial and posterior tibial blood pressure using Doppler techniques).

The underlying cause of death was available for all the deceased patients and was derived from the death certificate and coded according to the ICD-9 (International Classification of Diseases, Ninth Revision). Deaths due to CV diseases corresponded to ICD codes 410–414 (coronary artery diseases), 430–438 (strokes), 440 (peripheral artery diseases) and other ICD codes between 390–459 and 798.1 (other CV diseases).

Statistical analyses

Considering a type I error of 0.05 and a type II error of 0.90, a minimum of 83 subjects were needed for each tertile to detect a 10% difference in the CV scores between the tertiles of total flavonoid intakes.

Dietary total flavonoid intakes of the cohort were divided into tertiles, separately per sex. Cut-off points were 191.5, 401.2 and 138.3, 322.3 mg/day, respectively for men and women.

The distributions of flavonoid intake, fasting insulin, triglycerides, CRP values were skewed. The characteristics of the cohort according to the tertiles of flavonoid intakes were analyzed by ANOVA, Kruskal–Wallis tests (for not-normally distributed variables) or the χ2 test, as appropriate.

A multiple regression was performed to assess the association between the 2014 CV score and the variations from baseline to follow-up (values at 2014 minus values at baseline) in the CV risk score, and the tertiles of flavonoid intakes, after adjusting for BMI, education (primary/secondary/university), living in a rural area, METs (hour/week), alcohol intake (g/day), history of CV diseases, values of fasting glucose, log-CRP, fiber, and saturated fatty acid intakes. We did not include age, sex, total and HDL-cholesterol, smoking habits and blood pressure values, because these variables were included in the CV score calculation, to avoid over-controlling. However, when we controlled for these variables, results were not significantly different.

The relationships between tertiles of flavonoid intakes and all-cause mortality and CV mortality and incidence were assessed by estimating the hazard ratio (HR) and its 95% confidence intervals (CI) in Cox regression models, adjusted for age, sex, BMI, education, living in a rural area, METs (hour/week), alcohol, fiber, and saturated fatty acid intakes, smoking, values of systolic and diastolic blood pressure, total and HDL-cholesterol, fasting glucose, CRP, statin and aspirin use.

In all these analyses, individuals in the first (lower) tertile of flavonoid intakes were considered as the reference group, and the other groups were introduced as dummy variables (IBM SPSS Statistical Software Version 22).

Results

Out of 1,658 subjects, 138 (8.3%) resulted under-reporters. Among the tertiles of flavonoid intake, the percentage of under-reporters did not differ (8.2, 8.5 and 8.3% in the first, second and third tertiles, respectively).

Mean and median intake of total flavonoids were 320 and 251 mg/day, respectively (Table 1).Pearson correlations between flavonoids ranged from weak (r = 0.04 for flavan-3-ols with isoflavones) to high (r = 0.80 for flavan-3-ols with proanthocyanids).

Descriptive characteristics of the cohort by tertiles of flavonoid intakes are shown in Table 2.

Table 2 Baseline characteristics by tertiles of flavonoid intake (the first the lower; the third, the higher)
Full size table

In the lowest tertile, there was a higher percentage of smokers, alcohol abstainers, less educated individuals, hypertensive and diabetic patients (Table 2). On the other hand, subjects with the highest flavonoid intake were more frequently heavy drinkers living in a rural area, were more physically active, ate more calories, fiber and antioxidant vitamins, and less total fat and saturated fat. In individuals within the lowest tertile, the metabolic pattern was significantly worse, CRP values increased, and the CV risk score higher.

The 2014 CV risk score was significantly increased in the individuals with the lowest intake of total flavonoids and their subclasses, with the exception of isoflavones (Table 3). Similarly, after a mean 12-year follow-up, the difference in the scores (2014 score minus baseline score) was higher in those subjects. In a multiple regression models, being in the third (higher) tertile of flavonoid intake was inversely associated with the 2014 CV score and with change in score values from baseline to follow-up, after adjusting for BMI, education, living in a rural area, METs (hour/week), history of CV diseases, values of fasting glucose, log-CRP, alcohol, fiber, and saturated fatty acid intakes.

Table 3 CV risk score by tertiles of flavonoid intake (the first the lower; the third, the higher) in a multiple regression model
Full size table

During follow-up, 125 incident CV events were diagnosed and 220 deaths occurred, 84 of which due to CV causes (Table 4). The incidence of CV events was significantly lower in individuals with the higher intake of total flavonoids and with higher intake of all subclasses of flavonoids, but flavones and isoflavones, in Cox-regression models after adjustments for age, sex, BMI, education, living in a rural area, METs (hour/week), alcohol, fiber, and saturated fatty acid intakes, smoking, values of systolic and diastolic blood pressure, total and HDL-cholesterol, fasting glucose, CRP, statin and aspirin use. HRs ranged from 0.42 for the higher tertile of flavan-3-ols to 0.56 for the higher tertile of anthocyanidins in the Cox model after multiple adjustments.

Table 4 Cardiovascular events and all-cause and cardiovascular mortality by tertiles of flavonoid intake (the first the lower; the third, the higher)
Full size table

Total and subclasses of flavonoids were not significantly associated with the risk of CV mortality in the same Cox model (Table 4).

Being in the third tertile of flavan-3-ols (HR = 0.68; 95% CI 0.48–0.96), anthocyanidins (HR = 0.66; 95% CI 0.46–0.95) and flavanones (HR = 0.59; 95% CI 0.40–0.85) was inversely associated with all-cause mortality.

Data did not change after excluding the 138 under-reporters, the 79 women on estrogen therapy, the 55 individuals on nutritional supplements, and after adjusting for antioxidant vitamin intakes.

Discussion

The results of this population-based cohort study suggest that higher dietary intakes of flavonoids may be associated with a reduced CV risk score and a 40–50% lower risk of non-fatal CV events.

This is intriguing since in our cohort the consumption of some flavonoid-rich foods inversely associated with CV risk such as cocoa, soybean and tea [13, 17, 28, 29], is infrequent, being fruits and red wine the main sources of flavonoids. Epidemiological studies have suggested that a Mediterranean diet reduces the CV risk [42] and a high concentration of flavonoids has been found in fruits, vegetables, red wine and other elements of the Mediterranean diet. However, there is inconsistent evidence on the role of flavonoids derived from these foods and CV risk, since previous studies reported either a decreased CV incidence and mortality with increased intake of apples, pears, and red wine [8, 12, 16, 24, 25], or no significant effect [6, 10, 21, 26, 27].

We have found both a lower CV risk score at baseline and at follow-up in the higher tertile of flavonoid intake. Intriguingly, the increase in the score from enrolment to the end of follow-up was higher in those individuals. Accordingly, the consumption of flavonoid-rich food has been associated with lower systolic blood pressure [15, 43, 44], lower total cholesterol [44], higher HDL cholesterol values [44–46].

Benefits of flavonoids on blood pressure, lipid values, insulin resistance, and flow-mediated dilatation seem to derive above all from soy, cocoa and tea, as suggested by systematic reviews [47, 48]. However, more recently, flavonoids from fruits and vegetables have been reported to reduce the risk of diabetes mellitus and to improve microvascular reactivity and inflammatory status [49–51]. Accordingly, although a small number of incident CV events occurred in our cohort, the risk of non-fatal CV events was significantly lower in individuals with the higher intake of total and all subclasses of flavonoids, but flavones and isoflavones, which were consumed at negligible concentrations in our cohort. Therefore, the dietary intakes of flavonoids seems relevant for healthy CV outcomes at relatively low concentrations, since most inverse associations with CV risk score and non-fatal CV events appeared with intermediate or low intakes of specific subclasses, suggesting that even small amounts may be beneficial. However, a threshold of intake is probably needed, under which these compounds are unlike to be active.

Flavonoids can inhibit or induce a large variety of enzyme systems, involved in pathways regulating platelet aggregation, inflammatory and immune responses [1, 52, 53]. Furthermore, by their antioxidant properties, flavonoids may protect tissues against oxygen free radicals and lipid peroxidation, thus contributing to the prevention of atherosclerosis, chronic inflammation and cancer [1, 52, 53]. Because of their antioxidant and chelating properties, flavonoids may inactivate reactive oxygen species (ROS) and counteract the oxidation of circulating LDL particles [52–54]. Other anti-atherogenic actions proposed for these compounds are: reduction of the activity of enzymes increasing ROS production; inhibition of HMG-CoA reductase, cholesteryl ester transfer protein (CEPT), angiotensin-converting enzyme, signal transducers and activators of transcription (STAT), and glucose transporters; synthesis of nitric oxide; inhibition of platelet activation and function; anti-angiogenetic effects; improvement in endothelial function, vascular fragility, cellular permeability [54–56]. The anti-inflammatory properties of flavonoids may be due to the inhibition of NF-κB activation and adhesion molecule expression; suppression of the activity and secretion of inflammatory cells; reduction of the concentrations of CRP and cytokines [57, 58].

The associations with fatal events were controversial in our cohort. No significant association was found with CV mortality, probably because the number of fatal CV events was low. Otherwise, many flavonoid subclasses, such as flavan-3-ols, anthocyanidins and flavanones were inversely associated with all-cause mortality. Previous studies have reported a reduced total and/or CV mortality with proanthocyanids [19], flavan-3-ols, [11, 19, 25], anthocyanidins [16, 19], flavonols [13, 19], flavanones [16, 18], flavones [16, 19], and isoflavones [17, 44–47]. On the other hand, other authors reported no protective effects of total or specific subclasses of flavonoids on mortality [7, 12, 21, 22, 27].

These highly divergent results among studies might be due to differences in nutritional, socio-cultural and ethnic characteristics.

The median intakes of flavonoids are highly variable among studies, and values ranging from 50 to 450 mg have been reported in European studies [54]. In particular, the following median intakes have been described for Mediterranean countries: 92 mg/day in Greece [59] and 332.4 mg/day in Spain [60]. On the other hand, in non-Mediterranean countries, the median consumption of flavonoids was much lower, varying from 203 mg/day in US population [19] to 88 mg/day in Sweden, and 13 mg/day in Finland [61]. Our values were between these extreme intakes, in line with other Italian data [62, 63]. The high consumption of red wine and fruits, such as apples and citrus fruits, in our Italian cohort justify the higher intake of total flavonoids and proanthocyanidins with respect to other non-Mediterranean cohorts [19, 61]. On the other hand, the low consumption of tea, justified the lower intakes of flavon-3-ols (in particular epigallocatechin 3-gallate, epicatechin 3-gallate and epigallocatechin) with respect to UK and Ireland [61], and the negligible use of soy explain why the intake of isoflavones and flavones was much lower in our cohort when compared with Asian studies [17].

In most studies, the higher consumption of flavonoids was associated with an overall healthy dietary and metabolic pattern, in line with our results [8, 10–12, 16–21, 25, 26, 49]. Our cohort indeed included individuals with a low level of education, differently from previous studies performed in samples where most participants had at least a high school education [13, 16, 17, 19, 26].

Finally, many compounds tend to be present in the same foods: for example, in our cohort, individuals with lower intakes of flavonoids, ate less fiber and antioxidant vitamins and more saturated fats. It is therefore difficult to ascertain the independent effect of dietary components because of multicollinearity. However, our associations remained significant after adjusting for these dietary factors, thus suggesting that a higher flavonoid intake might not merely be an indicator of a healthier lifestyle.

Limitations

The EPIC questionnaire was not originally designed to measure flavonoid intake, but it has been extensively used and validated for this purpose [60, 64, 65].

The flavonoid intake might have been underestimated because of the limitations of the food composition databases. It should be noted that the presence of particular flavonoids in vegetables and fruits depends on the crop variety, location and type of cultivation. The adaptability of the USDA database to the Italian diet is questionable. The absorption and microbial transformation in the gut of specific subclasses of flavonoids vary considerably, therefore the different flavonoid bioavailability could have an impact on the associations between the assumption of these compounds and chronic diseases. In general, flavonoid subclasses are present simultaneously in foods and establishing which of the compound is responsible for the potential biological effect is difficult. We relied on dietary intake from the questionnaire administered at one point in time; thus misclassification of dietary exposure might have occurred if individuals have changed their diets during the follow-up. Furthermore, measurement error in collecting self-reported dietary intake is inevitable and our observational study was prone to the possibility of unmeasured confounding.

However, the recent versions of the USDA database includes more cooked foods [2], because in culinary preparations important losses in flavonoid content occur, and is the most complete and used database in the estimation of flavonoid intake. Moreover, we have referred also to a European database, and the USDA has been already used for the Italian population [62–64]. We have used a validated instrument and, both at baseline and at follow-up, the associations between flavonoid intakes and the CV risk score were consistent. Measurement errors and misclassification was likely to be random and would have attenuated the association found. We have took care to adjust for many potential confounders. Finally, we have studied a large population-based cohort from a localized region, with a high level of participation, which could have limited the number of potential confounders.

Conclusions

Individuals with higher intakes of flavonoids showed a lower CV risk after a mean 12-year follow-up, and a reduced risk of non-fatal CV events. If these results will be confirmed in larger prospective cohorts, it would be useful to obtain reliable markers of flavonoid intake in order to define the optimal doses of specific flavonoids for CV protection.