1. Introduction
Cardiovascular diseases (CVDs) account for almost one third of deaths worldwide [
1]; diet plays a decisive role among the modifiable environmental factors involved in the aetiology of CVDs [
2]. Polyphenols are bioactive phytochemicals present in plant foods and beverages, classified into four major classes according to their chemical structure (flavonoids, phenolic acids, stilbenes, and lignans), each class being further divided into subclasses [
3]. A large number of preclinical and clinical studies suggest a protective role of dietary polyphenols against CVDs [
4], and several major risk factors of CVDs, such as hypertension, overweight, dyslipidaemia, or diabetes [
5]. A better understanding of the molecular mechanisms involved in CVD progression has led to the hypothesis that the combination of various protective agents—including different types of polyphenols—targeting multiple signaling pathways, might be particularly effective in the prevention of CVDs [
4]. Cardiovascular protection by polyphenol intakes may encompass various mechanisms [
6,
7,
8], including anti-inflammatory properties [
9], an increase in antioxidant capacity [
10], and an inhibition of platelet aggregation and antithrombotic properties [
11].
In vivo, studies have also shown that polyphenols may improve endothelial and overall vascular health through the formation of vasoprotective factors, such as nitric oxide [
4,
12,
13], and their effects on the regulation of signaling pathways involved in metabolic homeostasis [
14].
Translating these experimental findings to prevention measures for the general population is, however, still limited because of a too limited number of studies focused on primary prevention. These studies have often included a too limited number of individual polyphenols [
15], which is far from representing the large diversity of these compounds found in the diet. The Phenol-Explorer database includes food composition data on 502 different polyphenols from 452 different foods [
16]. This database has been used in a few longitudinal studies to assess the association between polyphenols intakes and CVDs or mortality risk [
17,
18,
19,
20,
21]. However, currently available results are somewhat inconsistent with respect to the role of specific classes and subclasses of polyphenols in CVDs.
Therefore, the aim of the present prospective study was to assess the association between total polyphenol intakes, as well as classes and subclasses of polyphenols, and the risk of incident CVDs, coronary heart diseases (CHD), and cerebrovascular diseases (CD) in the French NutriNet-Santé cohort.
4. Discussion
In this large prospective study, three categories of polyphenols showed strong linear associations with a decreased risk of CVDs: Anthocyanins, catechins, and flavonols. Dihydrochalcones, dihydroflavonols, hydroxybenzoic acids, stilbenes, proanthocyanidins, and total polyphenols were also related to a decreased risk of CVDs (19 to 27%).
To our knowledge, few prospective studies on the association between polyphenol classes and subclasses and CVD risk were based on such detailed polyphenol intake data considering coefficient cooking factors as well as a broad range of classes of polyphenols. Most previous studies only focused on flavonoids. Consistent with our findings, studies evaluating the associations between flavonoids and the risk of CVDs have shown significant inverse associations [
36,
37]. In particular, the results of a meta-analysis published by Wang et al. reported an inverse association between flavonoids and the risk of CVDs [
36]. This result was also reported in a recent meta-analysis between dietary flavonoids and all-cause mortality [
38]. In particular, one of the strong linear associations observed in our study was the inverse relationship between catechin intake and the risk of CVDs. We have also observed significant inverse associations for the flavonoids, ‘dihydrochalcones’ and ‘dihydroflavonols’, even though few studies have investigated these categories, since these categories are rarely quantified and only a few milligrams are consumed per day. Very few studies have shown an inverse linear relationship of anthocyanin intake with the risk of CVDs, except for investigations of data from the PREDIMED study and from the Nurses’ Health Study [
18,
39]. Flavanols have only been shown to be inversely associated with the risk of CVD in a single study [
40], since this category has often been grouped with other flavonoids [
21,
41,
42,
43,
44,
45]. To our knowledge, the only study that showed a link between the risk of CVD and proanthocyanidins has focused on myocardial infarction mortality, using a US polyphenol composition table [
44].
In the present study, we have also observed inverse associations for stilbenes, hydroxybenzoic acids, and total intake of polyphenols with the risk of CVD, although RCS procedure revealed a non-linear shape. Few other studies have been able to study these relationships in observational cohorts, since composition data for these classes of polyphenols have only been available recently in the Phenol-Explorer database [
46]. Stilbenes and hydroxybenzoic acids were inversely associated with CVD risk, mostly in one study, a cohort of participants over 55 years of age of the PREDIMED study [
17,
18].
Regarding the total intake of polyphenols, only the PREDIMED study evaluated the total intake of polyphenols and the prospective risk of CVD, and showed, concordantly with our study, an inverse association between total polyphenol intake, assessed as the sum of individual polyphenols via the Phenol-Explorer database, and the risk of CVD and all-cause mortality [
17,
18]. Intriguingly, in contrast to the PREDIMED study, which used the sum of individual polyphenols as an indicator, we only found a significant result when estimating total polyphenol intakes on the basis of the Folin-Ciocalteu method. This spectrophotometric method is the most widely used experimental method to quantify total polyphenols, but is less specific since it includes antioxidant compounds other than polyphenols and better reflects the total antioxidant activity in foods [
47]. Several studies showed that higher total antioxidant activity improves endothelial function [
48], and reduces systemic inflammation [
49]. In our study, it is possible that total antioxidant activity included other constituents, such as ascorbic acid [
50], that may have had beneficial or synergetic effects on CVDs, in contrast to analyses in which we specifically focused on polyphenols. This may explain the divergence between our findings concerning total antioxidant activity and those concerning total and specific polyphenols.
In turn, no significant result concerning flavones, flavanones, isoflavonoids, hydroxycinnamic acids, and lignans were observed in our study. This is consistent with the fact that very few studies have found significant results for these specific polyphenol groups [
17,
18,
19,
41,
51,
52,
53,
54], except for flavanones [
19,
41,
51,
53], mainly present in citrus and orange and derivative juices. In addition to a possible lower consumption of these products in our cohort, various methodological differences between our study and those investigations that have obtained significant findings for flavanones could explain these inconsistencies. Indeed, significant findings for these polyphenols have been observed in studies that used the USDA’s US composition table [
55], which could provide more comprehensive estimates of isoflavonoids intake [
56], that estimated dietary intakes on the basis of food frequency questionnaires [
56], and the investigated outcomes differed (stroke and CVD mortality) from those that the present study focused on [
41,
44]. In addition, one study focused on a specific sample composed by menopausal women [
51], and only one study using the Phenol-Explorer
® table and the USDA table showed an inverse association between flavanones and all-cause mortality in a Spanish cohort [
19]. In the EPIC cohort, lignans were not associated with CVD mortality risk [
20], while in the PREDIMED study, lignans were inversely associated with both CVD risk and CVD mortality risk [
19]. Thus, a possible reason for the lack of association observed in our cohort, and the contrasted results observed in the literature, could be the limited consumption of lignans main dietary sources in our study (oilseed-rich foods) [
57], as compared to the Spanish cohort from PREDIMED [
19,
58]. To our knowledge, only the PREDIMED study has evaluated phenolic acids in relation to CVDs, and similarly to our study, found no significant association for hydroxycinnamic acids (for which, the main contributor is coffee), but a significant inverse association between hydroxybenzoic acids (for which, the main contributor is tea) and the risk of CVD [
19].
Analyses stratified by the type of cardiovascular events in our study, revealed that the associations observed for stilbenes and flavonols were only found for CD, whereas the associations observed for anthocyanins and catechins were found for both CHD and CD, suggesting a particularly important role of anthocyanins and catechins for the prevention of CVD. In addition, our findings indicate that for stilbenes and flavonols, there may be a pathophysiological mechanism that is specifically related to the aetiology of CD. Flavonols have already been suggested to be inversely associated with CD risk by Hollman et al. [
59,
60]. Concerning stilbenes, resveratrol is probably the most well-known and studied polyphenol in the context of CVDs, and could explain the potential, but controversial, protective role of red wine consumption that has been observed concerning cerebrovascular events [
61,
62,
63,
64]. A diet rich in fruits and vegetables and, to a lesser extent, tea (the main contributor to flavonols intake), and probably/controversially a diet containing red wine (the main contributor of stilbenes) could prevent CD—which is in line with findings concerning a protective role of the Mediterranean diet [
65].
Strengths of our study include its prospective design and large sample size, detailed data on risk factors and confounders related to CVDs risk, the precise evaluation of polyphenol intakes using validated 24-h dietary records coupled with the most complete and detailed polyphenol database, Phenol-Explorer database, as well as the validation of cardiovascular events by a medical committee. Generalization of our findings to populations other than French adults may be limited due to the mode of consumption and the different diets of the countries. For example, in our study, cherries and strawberries were the main contributors to anthocyanin intake while in Nordic countries, anthocyanins intake is generally provided by other berries and red fruit juices. Another limitation of our study was the limited statistical power to stratify our results on the type of CVDs. Furthermore, the multiple testing issues and the number of polyphenols have limited our ability to test each potential interaction separately. Finally, the hypothesis of residual confounding resulting from unmeasured behavioural factors and/or imprecision in the measure of included covariates cannot be entirely excluded owing to the observational design of this study. However, given our detailed adjustment for a comprehensive set of confounders, it is unlikely that these unmeasured factors would account fully for the observed results. It is possible that our findings might have been caused by other constituents found in the foods that contribute most to the classes or subclasses of polyphenols. However, the addition of other constituents of fruits, like fibers and folates, to our multivariable model did not substantially attenuate the observed associations, suggesting that polyphenols may have a protective effect against CVDs.
In conclusion, this study showed that high intakes of anthocyanins, dihydrochalcones, dihydroflavonols, catechins, proanthocyanidins, flavonols, hydroxybenzoic acids, and stilbenes, and total polyphenols were associated with a decreased risk of CVD. These subgroups of polyphenols may hence be of particular interest in the primary prevention of CVDs. Overall, our findings may have important public health implications, but will need to be confirmed in long-term randomized controlled trials, including biomarkers of CVDs risk, to elucidate mechanisms.