Dietary Factors and Risks of Cardiovascular Diseases: An Umbrella Review.

Unhealthy diet is a significant risk factor for cardiovascular diseases (CVD). Therefore, this umbrella review aims to comprehensively review the effects of dietary factors, including dietary patterns, food groups, and nutrients on CVD risks. Medline and Scopus databases were searched through March 2020. Systematic reviews with meta-analyses (SRMA) of randomized controlled trials (RCTs) or observational studies measuring the effects of dietary factors on CVD risks were eligible. Fifty-four SRMAs, including 35 SRMAs of observational studies, 10 SRMAs of RCTs, and 9 SRMAs of combined RCT and observational studies, were included for review. Findings from the SRMAs of RCTs suggest the significant benefit of Mediterranean and high-quality diets for lowering CVD risk, with pooled risk ratios (RRs) ranging from 0.55 (95%CI: 0.39–0.76) to 0.64 (95%CI: 0.53–0.79) and 0.70 (95%CI: 0.57–0.87), respectively. For food nutrients, two SRMAs of RCTs found that high intake of n-3 polyunsaturated fatty acid (PUFA) significantly reduced CVD risks, with pooled RRs ranging from 0.89 (95%CI: 0.82, 0.98) to 0.90 (95%CI: 0.85–0.96), while evidence of efficacy of n-6 PUFA and combined n-3 and n-6 PUFA were inconsistent. Moreover, results from the SRMAs of RCTs did not find a significant benefit of a low-salt diet and low total fat intake for CVD prevention. For food groups, results from the SRMAs of cohort studies suggest that high intakes of legumes, nuts, and chocolate, as well as a vegetarian diet significantly reduced the risk of coronary heart disease, with pooled RRs of 0.90 (95%CI: 0.84–0.97), 0.68 (95%CI: 0.59–0.78), 0.90 (95%CI: 0.82–0.97), and 0.71 (95%CI: 0.57–0.87), respectively. Healthy dietary patterns had a significant benefit for CVD prevention. With the substitutional and synergistic interactions between different food groups and nutrients, dietary recommendations for CVD prevention should be focused more on healthy dietary patterns than single food groups or nutrients.


Introduction
Cardiovascular diseases (CVD), today's leading causes of death, accounts for one third of all mortality worldwide [1]. In the past decade, CVD mortalities have increased globally by 12.5% [2]. One significant risk factor of CVD is an unhealthy diet, which is also related to other CVD risk factors, such as hypertension, diabetes mellitus (DM), and obesity [3,4]. Therefore, encouraging healthy diet adherence is important in decreasing CVD morbidity and mortality.
CVD dietary factors is usually classified into three main types: dietary patterns (e.g., the Mediterranean diet and the Dietary Approaches to Stop Hypertension (DASH) diet), food groups

Dietary Patterns
A total of 18 SMRAs, (13 SRMAs of observational studies, two SRMAs of RCTs and observational studies, and three SRMAs of RCTs) assessed the effects of dietary patterns and CVD risks. For SRMAs of observational studies, dietary patterns were Mediterranean diet (eight SRMAs), DASH diet (three SRMAs), diets with high HEI and AHEI scores (two SRMAs), HEI/AHEI and cardiovascular health (CVH) scores (one SRMA), and Diet Inflammatory Index (DII) scores (one SRMA). For SRMAs of RCTs, the interventions were prescribing a Mediterranean diet (one SRMA) [18] and modifying diet

Dietary Patterns
A total of 18 SMRAs, (13 SRMAs of observational studies, two SRMAs of RCTs and observational studies, and three SRMAs of RCTs) assessed the effects of dietary patterns and CVD risks. For SRMAs of observational studies, dietary patterns were Mediterranean diet (eight SRMAs), DASH diet (three SRMAs), diets with high HEI and AHEI scores (two SRMAs), HEI/AHEI and cardiovascular health (CVH) scores (one SRMA), and Diet Inflammatory Index (DII) scores (one SRMA). For SRMAs of RCTs, the interventions were prescribing a Mediterranean diet (one SRMA) [18] and modifying diet quality by lowering the consumption of carbohydrates, fat, and calories, and increasing the consumption of fish, vegetables, complex carbohydrates, and fiber (two SRMA) [17,19]. Most of the SRMAs (16/18) considered the general population as the study's participants, while two included only high-risk populations (e.g., patients with obesity, hypertension, and DM). The mean age and percentage of male participants ranged from 18 to 104 years, and 0% to 100%, respectively (see Table S1). The effects of each dietary pattern are described in Figure 2A-D and Table S3. consumption of fish, vegetables, complex carbohydrates, and fiber (two SRMA) [17,19]. Most of the SRMAs (16/18) considered the general population as the study's participants, while two included only high-risk populations (e.g., patients with obesity, hypertension, and DM). The mean age and percentage of male participants ranged from 18 to 104 years, and 0% to 100%, respectively (see Table  S1). The effects of each dietary pattern are described in Figure 2A-D and Table S3.

Mediterranean Diet
For all-cause mortality, two SRMAs of observational studies [20,21] found that adherence to Mediterranean diet significantly decreased risk of all-cause mortality in the general population, with a pooled RR ranging from 0.91 (95%CI: 0.89-0.94) to 0.92 (95%CI: 0.90-0.94). However, a finding from an SRMA of RCTs [18] suggested a non-significant benefit of the Mediterranean diet in reducing allcause mortality in high-risk populations (pooled RR = 1.00; 95%CI: 0.86-1.15; see Figure 2A).
For CVD mortality, both SRMAs of observational studies and RCTs [20,22] found a significant benefit of the Mediterranean diet in decreasing CVD mortality in the general population, with pooled RRs ranging from 0.59 to 0.91 (see Figure 2B), while the SRMA of RCTs [18] found no significant effect in high-risk populations (pooled RR = 0.90; 95%CI: 0.72-1.11; see Table S3).  [20,21] found that adherence to Mediterranean diet significantly decreased risk of all-cause mortality in the general population, with a pooled RR ranging from 0.91 (95%CI: 0.89-0.94) to 0.92 (95%CI: 0.90-0.94). However, a finding from an SRMA of RCTs [18] suggested a non-significant benefit of the Mediterranean diet in reducing all-cause mortality in high-risk populations (pooled RR = 1.00; 95%CI: 0.86-1.15; see Figure 2A).
For CVD mortality, both SRMAs of observational studies and RCTs [20,22] found a significant benefit of the Mediterranean diet in decreasing CVD mortality in the general population, with pooled RRs ranging from 0.59 to 0.91 (see Figure 2B), while the SRMA of RCTs [18] found no significant effect in high-risk populations (pooled RR = 0.90; 95%CI: 0.72-1.11; see Table S3).

Diet Quality
Diet quality was measured by HEI/AHEI (two SRMAs) [31,32], HEI/AHEI and CVH (one SRMA) [33], and DII (one SRMA) [34] scores. The effects of diet quality on CVD risk are presented in Table S3. High HEI/AHEI and CVH scores reflect the high quality of diet, whereas high DII scores reveal a poor-quality diet. Participants of all SRMAs were the general population. All three SRMAs of observational studies have consistent findings that consuming diets with high HEI/AHEI and CVH scores significantly decreased the risk of all-cause mortality (pooled RRs ranging from 0.54 to 0.78), CVD mortality (pooled RRs ranging from 0.30 to 0.77), and CVD (pooled RR = 0.78). However, diets with a high DII score showed a significantly increased risk of CVD (pooled RR = 1.35; 95%CI: 1.11-1.63) and CVD mortality (pooled RR = 1.37; 95%CI: 1.11-1.70; see Table S3).

Food Groups
A total of 14 SRMAs (10 SRMAs of observational studies and four SRMAs of both observational studies and RCTs) assessed the effects of food groups on CVD risks. The food groups considered were (1) fruits and vegetables; (2) nuts, whole grains, and legumes; (3) fish; (4) a vegetarian diet; (5) olive oil; (6) chocolate; (7) coffee; and (8) green tea. Participants of all 14 SRMAs were from the general population. Mean age, male percentage, and total number of participants ranged from 20 to 100 years, 0% to 100%, and 51 to 454,775, respectively. The effects of each food group are presented in Figure 3 and Table S3. studies and RCTs) assessed the effects of food groups on CVD risks. The food groups considered were (1) fruits and vegetables; (2) nuts, whole grains, and legumes; (3) fish; (4) a vegetarian diet; (5) olive oil; (6) chocolate; (7) coffee; and (8) green tea. Participants of all 14 SRMAs were from the general population. Mean age, male percentage, and total number of participants ranged from 20 to 100 years, 0% to 100%, and 51 to 454,775, respectively. The effects of each food group are presented in Figure 3 and

Fruits and Vegetables
Three SRMAs of observational studies assessed the association between fruit and vegetable intake and CVD risks [10][11][12]. High fruit and vegetables intake significantly decreased the risk of stroke (pooled RR ranging from 0.77 (

Chocolate
Two SRMAs of observational studies investigated the association between chocolate consumption and CVD risks [43,44]. The results demonstrate that high chocolate consumption significantly decreased the risk of CHD (pooled RRs ranging from 0.71 (

Coffee and Green Tea
One SRMA of observational studies found that drinking one cup of coffee per day significantly reduced all-cause and CVD mortality, with pooled RRs of 0.92 (95%CI: 0.91-0.94) and 0.89 (95%CI: 0.86-0.91), respectively [45] (see Figure 3C,D). For green tea, consuming 1-3 cups of green tea per day significantly lowers the risk of stroke (pooled RR = 0.64 (95%CI: 0.47-0.86), while the risk of CVD, all-cause, and CVD mortality were not significantly different between green tea intake and non-intake [46].

Food Nutrients
Sixteen SRMAs of observational studies, seven SRMAs of RCTs, and one SRMA of combined observational studies and RCTs investigated the effect of food nutrients on CVD risk in the general population. Mean age, percentage male, and total number of study's participants ranged from 20-89 years, 0-100%, and 16-388,229, respectively. Most SRMAs studied fat intake (11/24) Table S3.
Sixteen SRMAs of observational studies, seven SRMAs of RCTs, and one SRMA of combined observational studies and RCTs investigated the effect of food nutrients on CVD risk in the general population. Mean age, percentage male, and total number of study's participants ranged from 20-89 years, 0-100%, and 16-388,229, respectively. Most SRMAs studied fat intake (11/24), followed by fiber      Figure 5. Pooled risk ratios for food nutrients and the risk of stroke and coronary heart disease. Results are from systematic reviews and meta-analyses of observational studies for (A) and (B), and systematic reviews and meta-analyses of RCTs for (C).
One meta-analysis of RCTs indicated a non-significant benefit of n-6 PUFA for the prevention of all-cause mortality ( Figure 4A), while there were inconsistent findings for the outcomes on CVD mortality and CHD risk. One meta-analysis of RCTs found that high n-6 PUFA intake significantly decreased the risk of CVD mortality (pooled RR = 0.81; 95%CI: 0.70-0.95) [54], while another meta-analysis of RCT found a non-significant effect (pooled RR = 1.17; 95%CI: 0.82-1.68) [7]. Two meta-analyses of cohort studies assessing the effect of n-6 PUFA on the risk of CHD also show conflicting results, with Farvid et al. suggesting the significant benefit of high n-6 intake for prevention of CHD (pooled RR = 0.81; 95%CI: 0.70-0.95) [54], and Skeaff et al. demonstrating a non-significant benefit (pooled RR = 1.05; 95%CI: 0.92-1.20) [6]. For combined n-3 and n-6 PUFA, the results from two SRMAs of RCTs [6,7] indicate that high PUFA intake was not significantly associated with CHD risk, all-cause, and CVD mortality.

Sodium
One SRMA of RCTs, one SRMA of cohort studies, and one SRMA of observational studies and RCTs found that low sodium intake did not significantly reduce risk of CVD, all-cause mortality, and CVD mortality [56,57]; however, meta-analysis of cohort studies suggested that low sodium intake could significantly decrease the risk of stroke, with pooled RRs ranging from of 0.81 (95%CI: 0.70-0.93) to 0.94; 95%CI: 0.90-0.98 [57,58] (see Figure 5A).

Flavonoid
Two SRMAs of observational studies found that the risk of CVD mortality and stroke in people having a high flavonoid intake was significantly lower than those with a low intake, with pooled RRs of 0.86 (95%CI: 0.75-0.98; Figure 4C) and 0.80 (95%CI: 0.65-0.98; Figure 5A) [59,60]. However, the risk of all-cause mortality was not significantly different between the two groups [59] (see Figure 4A).
One SRMA of observational studies assessed the effect of anthocyanins, which are a subtype of flavonoid, on CVD risk [61]. This SRMA found that a high intake of anthocyanins significantly decreased the risk of CVD mortality and CHD, with pooled RRs of 0.92 (95%CI: 0.87-0.97) and 0.91 (95%CI: 0.83-0.99), respectively (see Table S3). However, the risk of stroke did not significantly decrease in people with a high intake of anthocyanins.

Calcium
One meta-analysis of cohort studies assessed the effect of high calcium intake and CVD mortality [63], and found that a high calcium intake from diet and supplements did not significantly reduce the risk of all-cause mortality (pooled RR = 0.83; 95%CI: 0.70-1.00) or CVD mortality (pooled RR = 0.97; 95%CI: 0.89-1.07) (see Figure 4A,C).

Methodological Quality Assessment
Results of quality assessment are presented in Table S4. Forty-eight out of 54 studies were classified as critically low confidence, five studies were classified as low confidence, and one study was classified as high confidence, according to AMSTAR-2 criteria. All of the studies having critically low confidence did not registered the review protocols, and most of those studies (90%) did not provide a list of excluded studies and reasons for exclusion. Around 50% of included studies did not consider the results of the risk of bias assessment in individual studies when interpreted, nor discussed the results. Most of the studies (51/54) used more than two databases for searching relevant studies, but only nine studies searched in databases for grey literatures. Most of the studies used appropriated methods of meta-analysis (52/54) and reported publication bias (45/54).

Discussion
This umbrella review provides a comprehensive summary of evidence about the effect of dietary factors on the risk of CVD. Evidence from RCTs and observational studies confirms the benefit of healthy dietary patterns, especially the Mediterranean diet, for the prevention of CVD, but the benefit for lowering all-cause and CVD mortality were inconsistent. The DASH diet also exhibited the ability to prevent CVD, but was only supported by observational studies.
A high intake of whole grains, legumes, fish, chocolate, and drinking one cup of coffee per day significantly decreased the risk of CVD mortality; a high intake of legumes, nuts, and chocolate, as well as a vegetarian diet could reduce risk of CHD. In addition, high intakes of vegetables and fruits, olive oil, and nuts, as well as dinking 1-3 cups of green tea per day significantly lowered risk of stroke. Evidence from RCTs and observational studies found that total fat intake was not associated with all-cause and CVD mortality, or the risk of CVD. However, high TFA intake significantly increased the risk of CVD and CVD mortality. Evidence from RCTs indicated consistent and inconsistent benefits of n-3 PUFA and n-6 PUFA for CVD prevention, respectively. Observational studies suggest the benefit of SFA for lowering stroke risk, and MUFA for lowering all-cause mortality and risk of stroke. In addition, evidence from observational studies also linked high fiber to CVD prevention and flavonoids to lowering CVD mortality. High potassium and low sodium intake also reduced the risk of stroke.
Our results, found from both RCTs and observational studies, suggest the significant benefit of healthy dietary patterns, especially the Mediterranean diet, for prevention of CVD. These findings correspond with the 2019 American College of Cardiology/American Heart Association (ACC/AHA) guidelines, which recommends healthy dietary patterns for lowering CVD [64]. Common characteristics of healthy diet patterns include a lower intake of red and processed meats, refined carbohydrates, sugar-sweetened beverages, and whole-fat dairy products; a higher consumption of fruits, vegetables, whole grains, nuts, and legumes; and a moderate consumption of alcohol. These dietary patterns are not nutrient-oriented, but rather consider a combination of multiple food groups. This approach provides several advantages. Firstly, people eat foods, not nutrients. Therefore, providing recommendation by dietary patterns is more practical in both public health and routine clinical practices than by nutrient alone [65]. Secondly, single-nutrient recommendations usually fail to consider substitutional effects and food replacement. For instance, our study found no benefit of a low-fat diet for CVD prevention, which may have resulted from the excess intake of other high-risk foods, such as refined carbohydrates and sugar to substitute the energy from fat [66,67]. Since the effect of foods on health depends on both synergistic and antagonistic interactions between multiple nutrients [68], nutrient orientation in generating dietary recommendations is not appropriate for CVD prevention.
The Mediterranean diet, among all healthy dietary patterns, has been the primary focus of previous studies. Several SRMAs of RCTs consistently found that the Mediterranean diet significantly decreased CVD mortality, CHD, and stroke risk. The key features of Mediterranean diet are a low intake of meat, with very low consumption of red and processed meat; a high intake of vegetables, fruits, nuts, legumes, cereals; and moderate intake of alcohol [69], all of which food groups had differing effects in CVD prevention. For instance, our review found that high intake of nuts, whole grains, and legumes, with a moderate intake of fish (2-4 servings/week) significantly decreased CVD mortality and CVD risk. Fish contains long-chain n-3 PUFAs that have a beneficial effect on CVD outcomes. Findings from SRs of RCTs support the hypothesis that prescribing supplement of n-3 PUFAs significantly reduces CVD risk. In addition, a recent RCT found that icosapent ethyl significantly decreases CVD risk beyond cholesterol-lowering therapy, and now has been approved by the U.S. Food and Drug Administration (FDA) for CVD prevention in high-risk patients [70].
Olive oil and extra-virgin olive oil are major sources of fat in the Mediterranean diet. Bioactive polyphenols-agents postulated to prevent CVD [71]-are only found in extra-virgin olive oil but not in common olive oil, which could explain the lack of the link between olive oil and CHD prevention in our review. In contrast, previous RCTs have suggested the significant benefit of extra-virgin olive oil in Mediterranean diet [72]. Hence, recommendation of olive oil for CVD prevention should focus on extra-virgin olive oil rather than common olive oil. Apart from bioactive polyphenols, the cardioprotective property of olive oil may come from its high content of MUFA [73]. However, results from our review show inconsistent evidence of the benefits of MUFA for CVD prevention.
Evidence of dietary patterns have usually focused on western diets, such as the Mediterranean and DASH diets. However, diets in other regions are different from those in Western countries, and it is not practical to recommend Western dietary patterns to other regions, e.g., Asian countries. Recently, the Japan Collaborative Cohort study that evaluated the effect of high Japanese food scores (i.e., high of consumption of rice, miso soup, seaweeds, pickles, green and yellow vegetables, fish, and green tea, and low consumption of beef, pork, and coffee) found that adherence to a Japanese food score may have benefits for CVD prevention [74]. However, more evidence is needed to confirm the benefits of Japanese food and other Asian dietary patterns on CVD prevention. Recommendations of dietary patterns should take into account the food culture of each region.
SRMAs of cohort studies found no benefit of high vegetable and fruit intake, which is recommended in Mediterranean and other healthy diet patterns, on CVD outcomes. However, SRMAs of observational studies found significant benefits in high dietary fiber intake for lowering CVD mortality and CVD risk. These conflicting results might have resulted from dietary fiber consisting of vegetables, fruits, nuts, whole grains, and legumes. However, nuts, whole grains, and legumes consist of not only fiber, but also of plant protein, unsaturated fats, minerals, and phytochemicals that might be attributable for their CVD-prevention properties [75]. Evidence of CVD dietary risk factors have mainly focused on fat, which is divided into total fat intake and subtypes of fatty acids, such as saturated fat, MUFA, and n-3 and n-6 PUFA. Inconsistent results according to different outcomes were found for each subtype of fat. For instance, high saturated fat intake did not significantly increase risk of CHD and CVD mortality, but was significantly associated with lower stroke risk. Moreover, n-3 PUFA substantially decreased CVD risk, but did not decrease all-cause and CVD mortality risk. Evidence of n-6 PUFAs were conflicting for CVD mortality and CHD. Guidelines from ACC/AHD year 2019 recommends replacing saturated fat with dietary MUFA and PUFA to reduce CVD risk [64]. However, this guideline does not state clearly the type of PUFA that is beneficial for CVD prevention.
Studies regarding some food groups and nutrients, such as vegetables, fruits, a vegetarian diet, olive oil, flavonoids, green tea, and potassium also had inconsistent findings, according to CVD outcomes. These might have resulted from the complex relationship between dietary and other lifestyle factors (e.g., physical activity), which might have confounded the findings of the previous studies.

Strengths and Limitations
This is the first umbrella review that comprehensively reviews the evidence of diets and CVD outcomes. We considered the effect of all dietary patterns, food groups, and nutrients, according to all types of CVD outcomes, including all-cause mortality, CVD mortality, CVD, CHD, and stroke. Only systematic reviews and meta-analyses, which are in the top hierarchy of evidence, were included in the review. The quality of systematic reviews and meta-analyses were assessed using AMSTAR 2. However, our study has some limitations. Firstly, the quality of most included SRMAs were critically low. Moreover, evidence of food groups and some nutrients were from SRMAs of observational studies having more bias and confounding effects than SRMAs of RCTs. Therefore, the results from those SRMAs should be interpreted with caution. In addition, the results from the included SRMAs were qualitatively analyzed, and dietary effects on CVD outcomes cannot be exactly quantified.

Conclusions
Healthy dietary patterns, such as the Mediterranean diet, have significant beneficial effects on CVD risk. A high intake of food groups and nutrients like nuts, whole grains, legumes, and dietary fiber, and moderate intake of fish also acts to prevent CVD, while a high intake of trans fatty acids significantly increased CVD risk. With the substitutional and synergistic interactions between different food groups and nutrients, dietary recommendations for CVD prevention should be focused more on healthy dietary patterns than single food groups or nutrients.
Supplementary Materials: The following are available online at http://www.mdpi.com/2072-6643/12/4/1088/s1, Appendix S1: Search terms and search strategies from Medline and Scopus databases. Table S1: Characteristics of included systematic reviews and meta-analyses. Table S2: Details of diet intervention and dietary factors. Table S3: Pooled risk ratios of dietary factors, according to outcomes. Table S4: Results of methodological quality assessment.