Polyphenol Health Effects on Cardiovascular and Neurodegenerative Disorders: A Review and Meta-Analysis

Several studies have demonstrated that polyphenol-enriched diets may have beneficial effects against the development of degenerative diseases, including atherosclerosis and disorders affecting the central nervous system. This activity has been associated not only with antioxidant and anti-inflammatory properties, but also with additional mechanisms, such as the modulation of lipid metabolism and gut microbiota function. However, long-term studies on humans provided controversial results, making the prediction of polyphenol impact on health uncertain. The aim of this review is to provide an overview and critical analysis of the literature related to the effects of the principal dietary polyphenols on cardiovascular and neurodegenerative disorders. We critically considered and meta-analyzed randomized controlled clinical trials involving subjects taking polyphenol-based supplements. Although some polyphenols might improve specific markers of cardiovascular risk and cognitive status, many inconsistent data are present in literature. Therefore, definitive recommendations for the use of these compounds in the prevention of cardiovascular disease and cognitive decline are currently not applicable. Once pivotal aspects for the definition of polyphenol bioactivity, such as the characterization of pharmacokinetics and safety, are addressed, it will be possible to have a clear picture of the realistic potential of polyphenols for disease prevention.


Introduction
Interest in the health effects of polyphenols has exponentially increased in recent years. Since the 80s, more than 80,000 studies have been published in PubMed, reporting multiple effects on a broad range of diseases [1]. However, most of this evidence was derived from in vitro or animal studies [2], while human trials that evaluate long-term effects of polyphenols are particularly scant. Moreover, no studies with hard clinical end points have been conducted so far [3]. The main limitations of current published studies include inaccuracy in the selection of tested concentrations. In fact, much higher Int. J. Mol. Sci. 2019, 20, 351 3 of 26 described in the meta-analysis section, while the following narrative part discusses multiple endpoints from recent (last five years) trials.
Amongst all the investigated CVD hallmarks, hypertension, dyslipidemia, endothelial dysfunction and systemic inflammation were the most reported in selected literature. Specific markers have been used for evaluating cardiovascular health. Increased total or low density lipoprotein cholesterol (LDL-C), or diminished high density lipoprotein cholesterol (HDL-C) levels, are consolidated markers of increased cardiovascular risk and directly contribute to the atherosclerotic plaque physiopathology [11]. In addition, other proteins like apolipoprotein A-I or paraoxonase-1, which are constitutive components of HDL particles, mainly displaying atheroprotective effects, have been inversely correlated with cardiovascular risk. Specific markers of endothelial activation and dysfunction, causal drivers of vascular damage, may be measured both in plasma and through minimally invasive ultrasound techniques. For example, endothelin-1, soluble endothelial-derived adhesion molecules (sICAM-1, sVCAM-1), asymmetric dimethylarginine, exerting pro-inflammatory actions on vascular endothelium, as well as high-sensitive C-reactive protein (hs-CRP), tumor necrosis factor alpha (TNF-α) and interleukin-1 beta, systemic inflammation markers, are commonly used [12,13]. Moreover, the brachial artery flow mediated dilation (FMD) has found wide application. It is defined as the percent change in brachial artery diameter, following to reactive hyperaemia, typically induced by a five-minutes circulatory arrest through a supra-systolic cuff occlusion, causing NO release and vasodilation [14].
However, high heterogeneity in experimental design, study population, compounds and endpoints evaluated, as well as the paucity of data about the biological activity of the different metabolites hampered the possibility of clear comparisons or univocal conclusions.
In another study, 70 subjects were randomized to receive 162 mg/day of quercetin from onion peel extract or placebo in a double-blinded, placebo-controlled cross-over trial with six-week treatment periods separated by a six-week washout period [16]. This study is of particular interest because subjects were also controlled for plasma concentrations of quercetin, its monomethylated derivatives tamarixetin (4 -O-methyl quercetin), isorhamnetin (3 -O-methyl quercetin) and the dehydroxylated quercetin metabolite kaempferol. In the subgroup of the hypertensive subjects, quercetin significantly decreased systolic blood pressure (SBP) by −3.6 mmHg (p = 0.022) when compared with placebo (mean treatment difference, −3.9 mmHg; p = 0.049). Notwithstanding, vasoactive biomarkers including endothelin-1, sICAM-1, sVCAM-1, asymmetric dimethylarginine, angiotensin-converting enzyme activity, vascular/endothelial function (evaluated by peripheral arterial tonometry, a technology to assess the reactive hyperaemia index), parameters of oxidation, were not affected by quercetin in the total group and in the subgroup of hypertensives. In the same cohort [17], authors did not find any significant changes in serum concentrations of leptin and adiponectin, homeostasis model assessment-adiponectin (HOMA-AD), glucose, insulin, homeostasis model assessment of insulin resistance (HOMA-IR), blood biomarkers of liver and renal function, hematology, serum electrolytes, hs-CRP and plasma TNF-α. On the contrary, a randomized double-blind, placebo-controlled study involving 72 healthy, overweight, and obese participants, randomly assigned to receive 100 mg/day of quercetin for 12 weeks did not find any significant difference in blood pressure between treatment and placebo arm, nor versus baseline [18]. However, treatment was effective in ameliorating endothelial function evaluated by percent FMD (from 12.5 ± 5.2% to 15.2 ± 6.1%; p = 0.002), and circulating 4 of 26 endothelial progenitor cells counts by flow cytometry (44.2 ± 25.6% vs. 52.3 ± 18.6%; p = 0.005), compared with the baseline values.
A randomized, double-blind, placebo-controlled crossover trial evaluated the effects of pure (−)-epicatechin (100 mg/day) and quercetin-3-glucoside (160 mg/day) on biomarkers of endothelial dysfunction and inflammation [19] or vascular function and cardiometabolic health [20], in a cohort of 37 apparently healthy pre-hypertensive (SBP = 125=−160 mmHg) men and women aged 40-80 years. The first analysis showed reduced plasma levels of markers of vascular inflammation, such as soluble endothelial-selectin or interleukin-1 beta. However, the differences were not significant in most cases.
In the second analysis, epicatechin supplementation improved fasting plasma insulin and HOMA-IR, without affecting blood pressure, arterial stiffness, nitric oxide (NO), endothelin-1, or blood lipid profile. Quercetin-3-glucoside supplementation had no effect on FMD, insulin resistance, or other CVD risk factors. Although the compliance and polyphenol absorption were monitored after four weeks of treatment by measuring plasma and urine epicatechin and quercetin concentrations, no further analysis of metabolites was carried out. However, studying pure flavonoids instead of the original complex matrixes may raise some important issues. In fact, if on the one hand, this approach reduces the burden of potential confounding factors, on the other, it may exclude potential favorable interactions with other flavonoids and compounds naturally present in original sources, like cocoa or tea.

Mechanisms of Cardioprotection
The role of oxidative stress as a promoter of endothelial dysfunction, that in turn is a driver of early atherosclerosis and consequent cardiovascular related disorders, is no longer in doubt and provides a support for the anti-inflammatory and anti-oxidant strategy in the field of cardioprotection [21]. The widely described antioxidant properties of polyphenols relies on the presence of hydroxyl groups that can be readily oxidized to produce the corresponding O-quinones [22]. This conversion results in an effective scavenger activity towards reactive oxygen species, which occurs through the entrapment of free radicals into stabilized chemical complexes, thus preventing further reactions [23]. This so-called "biochemical scavenger theory" is currently the most validated one to explain the beneficial effects towards a broad range of non-communicable diseases, including CVD. However, it has to be noted that, although the antioxidant capacity of polyphenols has been largely tested, the results of in vitro tests not always translate in an increased antioxidant status in humans. This lack of consistency may be due to high variability of the single in vitro assays, as well as to individual-related factors [24].
Beyond the inhibition of oxidative stress, polyphenols also display indirect antioxidant effects, occurring through the activation of the transcription nuclear factor erythroid 2-related factor 2 (Nrf2). This event induces endogenous antioxidant systems and is likely to be responsible also for polyphenol-mediated maintenance of the correct redox balance of cells, achieved through the equilibrium of phase I and II enzyme activity [25]. The anti-inflammatory properties of polyphenols are strictly connected with the modulation of oxidative stress and of the balance of redox cellular homeostasis [26]. Multiple mechanisms, most of which are mediated by the inhibition of the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB), account for polyphenol's anti-inflammatory activity. These compounds are able to decrease the cellular production of pro-inflammatory mediators [27][28][29] and to inhibit the expression of adhesion molecules [30], thus impairing the chemotaxis of monocytes within the inflamed tissues. Additional mechanisms accounting for the cardioprotective action of polyphenols target lipid metabolism, whose impairment represents a causative factor of atherosclerosis development [31]. An exhaustive appraisal of polyphenol-enriched food capacity to beneficially modulate lipid and lipoprotein metabolism has been recently reviewed by our group [32]. The widely reported decrease of total and LDL-C following the intake of polyphenols is possibly related to mechanisms occurring at hepatic and intestinal level. In the former, the reduction of cholesterol synthesis, the increase of LDL receptor expression and activity [33] and the increase of the cholesterol transporters ATP-Binding Cassette G5/ATP-Binding Cassette G8 expression [34] have been described in in vivo models. In the latter, polyphenol (i.e., epigallocatechin gallate) capacity to displace cholesterol from intestinal micelles have been associated with increased cholesterol fecal elimination in vivo [35]. The effect on triglyceride plasma level is possibly related to the reduction of apolipoprotein B48 and apolipoprotein B100 production in the liver and intestine, as demonstrated in obese subjects [36], or to the interference with lipoprotein lipase expression, as evidenced in pigs [37]. The mechanisms accounting for the increase of HDL-C are limited to in vitro evidence. Among them, the increase in apolipoprotein A-I synthesis has been reported in cultured hepatic or intestinal cells exposed to cocoa polyphenols [38].
Polyphenols could also be able to positively affect endothelial function, whose impairment is an established key factor for the development of atherosclerosis. This activity has been demonstrated as an amelioration of FMD in humans, although the effective dose was significantly higher than the typical dietary intake [39]. The mechanism accounting for improved FMD probably relies on the increase of NO synthase activity, as suggested by in vitro [40] and human studies [41]. This NO-mediated vasodilation, together with the influence on the renin-angiotensin system [42], is responsible for the reduction of blood pressure, an additional cardioprotective activity of polyphenols. Interestingly, this effect is also evident upon consumption of low, habitual amount of polyphenol-rich food [43].
Finally, polyphenols cardiovascular benefit may be ascribed to the peculiar pharmacokinetic properties mentioned above. While these compounds mainly reach the distal tract of gastrointestinal system unchanged, once modified by the gut microbiota, they may exert a prebiotic-like activity [44], causing the selective growth of beneficial bacteria [45][46][47], together with the inhibition of harmful strains [46][47][48]. This effect has been demonstrated in intervention studies, as well as in ex vivo models of fecal fermentation, possibly accounting for the amelioration of markers of CVD.

Methods
The literature search was performed evaluating both MEDLINE and Embase databases, considering all manuscript published in English language until November 10, 2018. The following search strategy was performed: ((polyphenol) OR flavonoids) OR polyphenolic compounds) OR isoflavone) OR flavanol) OR phytoestrogen) OR resveratrol) AND atherosclerosis) OR cardiovascular) OR cardiovascular disorder) OR cardiovascular health. The literature search was repeated adding the words "administration" and "therapy" using Mesh terms. Inclusion criteria were: (i) double-blind, randomized, controlled study design; (ii) adult patients (older than 18 years); (iii) chronic polyphenol administration; (iv) administration of polyphenols as food supplement or isolated compounds. Thus, both sexes were considered eligible, as well as all polyphenol compounds and route of administration. Moreover, only treatment periods over 2 weeks were included. During data extraction, the following endpoints were analyzed: FMD, LDL-C, HDL-C, hs-CRP and blood pressure.
The meta-analysis of the data extracted from the included studies was conducted using the Review Manager 5.2 software (Version 5.2.4 Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2012). Data were combined using random or fixed effect models, according to data heterogeneity. Heterogeneity between the results of different studies was examined by inspecting the scatter in the data points and the overlap in their confidence intervals and by performing chi-square tests and I 2 statistics. When I 2 was higher than 60%, a random effect model was used since it provides a more conservative estimate of the overall effect, especially relevant when studies were of different design and duration. Weighted mean differences and 95% confidence intervals were estimated for each endpoint considered. Values of p < 0.05 were considered statistically significant.

Results
The flow chart of literature search is depicted in Figure 1 We initially detected a total of 2,370,680 studies. Once only studies involving humans in a clinical trial were considered, an amount of 123,773 studies was selected. Among these, 43 studies were evaluated to extract cardiovascular data. Finally, 9 studies were excluded (Table S1), while 34 studies were included in the meta-analysis (Table 1). We initially detected a total of 2,370,680 studies. Once only studies involving humans in a clinical trial were considered, an amount of 123,773 studies was selected. Among these, 43 studies were evaluated to extract cardiovascular data. Finally, 9 studies were excluded (Table S1), while 34 studies were included in the meta-analysis (Table 1). SBP was reported in 16 studies, evaluating a total of 1190 subjects (606 treated with polyphenols and 584 with placebo). The overall analysis showed a significant systolic pressure reduction after polyphenol administration (mean difference −1.01, 95%CI: −2.04; 0.02, p = 0.005) ( Figure 2).  In the same studies, diastolic blood pressure was significantly reduced by polyphenol administration (mean difference −1.32 mmHg, 95% CI: −2.37, −0.27 mmHg; p = 0.001) (Figure 3).   In the same studies, diastolic blood pressure was significantly reduced by polyphenol administration (mean difference −1.32 mmHg, 95% CI: −2.37, −0.27 mmHg; p = 0.001) ( Figure 3).

Figure 2.
Pooled analysis of association between polyphenol intake and systolic blood pressure (SBP). The green square represents the mean difference detected in each study. The black diamond shows the final overall mean difference.

Discussion
In an attempt to provide a quantitative assessment of the possible effects of polyphenols on cardiovascular health, we performed a meta-analysis of the main parameters evaluated in the selected studies. All the tested dosages tended to be higher than those reachable with a common diet, as our analysis specifically focused on polyphenol supplementation. This is particularly true for flavanols like quercetin (>100 mg/day) [16][17][18][19] and isoflavones (80 to 132 mg/day, Mainini 2013 and Hs-CRP was reported in nine studies, for a total of 611 patients (303 treated with polyphenol and 308 with placebo). The polyphenol administration did not change hs-CRP plasma levels (p < 0.430) (Figure 7).

Discussion
In an attempt to provide a quantitative assessment of the possible effects of polyphenols on cardiovascular health, we performed a meta-analysis of the main parameters evaluated in the selected studies. All the tested dosages tended to be higher than those reachable with a common diet, as our analysis specifically focused on polyphenol supplementation. This is particularly true for flavanols like quercetin (>100 mg/day) [16][17][18][19] and isoflavones (80 to 132 mg/day, Mainini 2013 and Jayagopal 2002), while in some cases, the doses of oligomeric proanthocyanidin complexes (200 mg/day [15]) ranged around those achievable by diet [49].
According to the inclusion and exclusion criteria, a satisfactory number of studies contributed to the analysis (17 for blood pressure, 21 for blood lipids and 9 for endothelial function and systemic inflammation) and the number of subjects evaluated ranges from 500 to 2000, for each parameter

Discussion
In an attempt to provide a quantitative assessment of the possible effects of polyphenols on cardiovascular health, we performed a meta-analysis of the main parameters evaluated in the selected studies. All the tested dosages tended to be higher than those reachable with a common diet, as our analysis specifically focused on polyphenol supplementation. This is particularly true for flavanols like quercetin (>100 mg/day) [16][17][18][19] and isoflavones (80 to 132 mg/day, Mainini 2013 and Jayagopal 2002), while in some cases, the doses of oligomeric proanthocyanidin complexes (200 mg/day [15]) ranged around those achievable by diet [49].
According to the inclusion and exclusion criteria, a satisfactory number of studies contributed to the analysis (17 for blood pressure, 21 for blood lipids and 9 for endothelial function and systemic inflammation) and the number of subjects evaluated ranges from 500 to 2000, for each parameter considered. However, high heterogeneity has been found, even because of the differences in the treatment, in terms of formulation, dose, source and identity of the evaluated polyphenol. These discrepancies hampered the possibility to carry out subgroup analysis. Notwithstanding, the overall analysis revealed a significant effect of polyphenols in positively modulating the cardiovascular parameters considered. Treatments lowered systolic and diastolic blood pressure, as well as plasma levels of LDL-C, while increasing HDL-C levels and percent FMD. Interestingly, the effect on blood pressure seems to be stronger in the diastolic than systolic one. Finally, no significant effects were detected for hs-CRP by our meta-analysis. Although these effects are significant from a statistical point of view, the detected differences are undoubtedly of modest size and their actual clinical benefit remains to be established. For example, the absolute reduction in LDL-C (mean difference: −4.39 mg/dL) may be of interest, since it has been shown that a linear relationship exists between LDL-C lowering and the rate of major vascular events [50]. The effect on HDL-C is far more complex to evaluate. In fact, the effect is due for about 95% to a single study [51] out of 21 total studies considered. Furthermore, according to the most recent evidences coming from both Mendelian randomization studies and clinical trials [52][53][54], increasing HDL-C in subjects with normal baseline values is not an effective approach for cardiovascular risk reduction. The value of these results can be further corroborated by a recent analysis, aimed at defining the appropriateness of outcome variables for cardiovascular health claims compliant with European Community Regulation 1924/2006 [55]. Accordingly, the study provided evidence of risk factors to be used for the substantiation of health claims in the context of CVD, such as LDL-C and SBP, while other factors, like HDL-C or triglycerides, were not recommended.

Impact of Polyphenols on Cognitive Functions in Humans
The significant advances in medicine have brought about a marked increase in lifespan over the last years. Despite the obvious positive consequences, population aging led to a significant raise in the onset of age-associated diseases, including neurodegenerative diseases [56]. Neurodegeneration is a common condition shared by several disorders, such as amyotrophic lateral sclerosis, Huntington's, Alzheimer's disease (AD), frontotemporal dementia, and Parkinson's disease [57]. Their etiology is multifactorial and still far from being fully understood. Among the several factors involved, chronic and neuroinflammation, cellular senescence, genome instability and proteostasis dysregulation are listed [58,59]. Notably, even exogenous factors and toxins may increase the risk of neurodegenerative diseases onset by determining a persistent inflammatory status and oxidative stress [60,61]. Although different in clinical presentation, all these diseases share common features and mechanisms, including abnormal accumulation and aggregation of disease-specific proteins that lead to progressive loss of specific neurons and synapses in distinct brain regions [61,62]. From a clinical point of view, neurodegenerative diseases are characterized by a progressive impairment of the cognitive and/or motor function, the former being the typical pre-clinical condition of AD and frontotemporal dementia [63]. Regarding AD, which represents the main cause of dementia worldwide (60−80%) [64], early symptoms are represented by a progressive memory loss, a decline in cognitive functions, and behavioral signs, including mood changes and depression [65]. The prodromal form of AD is defined as mild cognitive impairment (MCI) and is characterized by alterations of specific cognitive domains among memory, executive function, attention, visuospatial skills and language [66,67]. MCI subjects display a 50% risk in five years for progression of neurocognitive disorders [68]. The current treatments to halt cognitive decline are limited to counteract symptoms and have a positive impact on cognition and behavior only in a transient manner, without affecting the underlying pathology. This current scenario makes the development of alternative strategies to prevent or slow the neurodegeneration an urgent need. Among these approaches, the role of diet as a preventive tool for cognitive decline has been an object of intensive research in the last years.
Many observational and intervention studies suggest a relationship between the administration of polyphenol-rich products, such as berries, grape, cocoa and green tea, and an improvement in cognitive performance [69][70][71][72]. However, not all studies have univocally brought positive results, leaving the open question on the potential applications of polyphenols for cognitive health. First, a recent meta-analysis reported no significant benefit of the Mediterranean diet, whose supplying of polyphenol is about 1g/day [49], on incident cognitive impairment [73]. In addition, an overview of the effect of soy isoflavones concluded that their effects on cognition may be limited and influenced by several factors related to individual features, type of treatment, and the plant itself [74]. Similarly, a meta-analysis of 21 clinical trials indicated that the treatment of MCI patients with Ginkgo biloba extract brought inconsistent results on the cognitive endpoint [75].
Possible explanations for these discrepancies are the previously mentioned bioavailability issues and the inter-individual variability of gut microbiota composition and function, so that the same polyphenols may translate in different effects on people, depending on individual characteristics [76]. In summary, the lack of convincing results from the clinical trials and the above cited criticism make it difficult to draw a solid conclusion on their preventive effects on cognitive status and neurodegeneration.
In this work, we performed a meta-analysis of human trials, in order to help clarify the efficacy of polyphenols with respect to several aspects of cognitive decline. In detail, we focused on healthy elderly or MCI subjects, and we considered the impact of polyphenols on global cognitive health, as well as on specific cognitive domains, including memory, and verbal and executive functions.

Mechanisms of Neuroprotection
Multiple mechanisms are likely to account for polyphenol-induced prevention of age-related neurodegeneration [77,78]. Seminal observations attributed these benefits to the well described radical scavenging activity, allowing the reduction of brain cell damage. Even in the case of neuroprotection, we have to acknowledge that in vitro tests do not always reflect effective antioxidant capacity in humans, as stated above [24]. An additional antioxidant mechanism is the metal-chelating activity, which may prevent the enzymatic production of reactive oxygen species, catalyzed by the metals themselves [22]. This effect is likely to be responsible, at least in part, for the neuroprotective effective of some flavonoids [79]. Recent evidence suggests that polyphenols, in particular flavonoids, may affect specific processes involved in the onset of neurodegenerative disease. Among these, the reduction of neuropathological proteins accumulation [80] and the improvement of synaptic plasticity [81] have been described in animal models of neurocognitive dysfunction treated with physiological doses of polyphenols. Strikingly, the role of polyphenols in improving brain vascular function, thus maintaining cerebral blood flow underpinning cognitive activity, has been well established in young and old adults [82,83].
Less consistent results are obtained in the research for polyphenol-mediated protection of neuroinflammation. The potential anti-inflammatory mechanisms have been described in glial and neuronal cells exposed to supraphysiological concentrations of flavonoids and some studies on humans revealed flavonoid ability to reduce plasma levels of several inflammatory markers [84,85]. However, the anti-inflammatory potential of polyphenols remains uncertain, since these effects were not confirmed in other small trials [86,87]. An additional mechanism explaining the neuroprotective action of polyphenols against AD is the inhibition of cholinesterases, resulting in increased cholinergic activity and improved cognitive performance [88]. This activity has been described for resveratrol, which was able to reduce acetylcholinesterase in specific brain regions of diabetic rats, with a consequent amelioration of memory impairment [89].
The previously described inter-relationships between polyphenols and microbiota also have relevant implication in the setting of neuroprotection. Indeed, intestinal-derived metabolites of polyphenols have been shown to prevent amyloid oligomerization with potential protection against neurodegenerative disorders in rats [90].

Methods
The literature search was performed evaluating both MEDLINE and Embase databases, considering all manuscript published in English-language until November 10, 2018. The following search strategy was performed: ((polyphenol) OR flavonoids) OR polyphenolic compounds) OR Isoflavone) OR flavanol) OR phytoestrogen) OR resveratrol) AND neurocognitive function) OR cognitive function) OR cognition. The literature search was repeated adding the words "administration" and "therapy" using Mesh terms. Inclusion criteria were the same described above. During data extraction, the following endpoints were considered: Wisconsin Card Classification Test (WCST) [91], Reys Auditory Verbal Learning Task (RAVLT) [92], Trail Making test (TMT A and B) [93], Alzheimer's Disease Assessment Scale-cognitive subscale (ADAS-Cog) [94], Mini-Mental State Examination (MMSE) [95] and Wechsler Adult Intelligence Scale (WAIS) [96]. The meta-analysis of the data extracted from the included studies was conducted as described in the Section 2.3.1.

Results
The literature search detected 293,093 documents. Studies evaluating animal or cellular models were excluded and 19,907 remained for the abstract evaluation. Among these, 61 studies fulfilled inclusion criteria and were further evaluated for data extraction. Neurocognitive endpoints considered for the analysis were available in 21 studies, which were entered in the meta-analysis (Figure 8 and Examination (MMSE) [95] and Wechsler Adult Intelligence Scale (WAIS) [96]. The meta-analysis of the data extracted from the included studies was conducted as described in the Section 2.3.1.

Results
The literature search detected 293,093 documents. Studies evaluating animal or cellular models were excluded and 19,907 remained for the abstract evaluation. Among these, 61 studies fulfilled inclusion criteria and were further evaluated for data extraction. Neurocognitive endpoints considered for the analysis were available in 21 studies, which were entered in the meta-analysis ( Figure 8 and Table 2).    The excluded studies and reason for exclusion are reported in supplementary material (Table S2). The ADAS-Cog score was reported in five studies, for a total of 508 patients treated and 501 controls. The overall effects did not show a significant impact of polyphenols on ADAS-Cog (p = 0.310). However, the analysis showed a significant effect of Ginkgo biloba on ADAS-Cog, compared to the placebo (mean difference −2.02, 95% CI −3.79; −0.26, p = 0.020). (Figure 9). The excluded studies and reason for exclusion are reported in supplementary material (Table  S2).
The ADAS-Cog score was reported in five studies, for a total of 508 patients treated and 501 controls. The overall effects did not show a significant impact of polyphenols on ADAS-Cog (p = 0.310). However, the analysis showed a significant effect of Ginkgo biloba on ADAS-Cog, compared to the placebo (mean difference −2.02, 95% CI −3.79; −0.26, p = 0.020). (Figure 9). MMSE was reported in eight studies, for a total of 1868 patients treated and 1856 controls. The overall effect did not show any significant effect of polyphenols on MMSE outcome (p = 0.080) ( Figure  S1).
Considering WAIS, three outcomes were reported in more than three studies: Digital symbol, Block design and Digit Span. Thus, only these three WAIS field were meta-analysed. WAIS digital symbol field was reported in eight studies, for a total of 4032 patients (2029 patients and 2003 controls), without a significant effect of polyphenols (p = 0.380) ( Figure 10). Block design field, on the contrary, was evaluated in four studies, for a total of 3389 patients, suggesting a significant reduction (p = 0.005), in particular when Ginkgo biloba was used (mean difference −0.43, 95%CI: −0.73; −0.13, p = 0.005) ( Figure S2). MMSE was reported in eight studies, for a total of 1868 patients treated and 1856 controls. The overall effect did not show any significant effect of polyphenols on MMSE outcome (p = 0.080) ( Figure S1).
Considering WAIS, three outcomes were reported in more than three studies: Digital symbol, Block design and Digit Span. Thus, only these three WAIS field were meta-analysed. WAIS digital symbol field was reported in eight studies, for a total of 4032 patients (2029 patients and 2003 controls), without a significant effect of polyphenols (p = 0.380) (Figure 10). The excluded studies and reason for exclusion are reported in supplementary material (Table  S2).
The ADAS-Cog score was reported in five studies, for a total of 508 patients treated and 501 controls. The overall effects did not show a significant impact of polyphenols on ADAS-Cog (p = 0.310). However, the analysis showed a significant effect of Ginkgo biloba on ADAS-Cog, compared to the placebo (mean difference −2.02, 95% CI −3.79; −0.26, p = 0.020). (Figure 9). MMSE was reported in eight studies, for a total of 1868 patients treated and 1856 controls. The overall effect did not show any significant effect of polyphenols on MMSE outcome (p = 0.080) ( Figure  S1).
Considering WAIS, three outcomes were reported in more than three studies: Digital symbol, Block design and Digit Span. Thus, only these three WAIS field were meta-analysed. WAIS digital symbol field was reported in eight studies, for a total of 4032 patients (2029 patients and 2003 controls), without a significant effect of polyphenols (p = 0.380) ( Figure 10). Block design field, on the contrary, was evaluated in four studies, for a total of 3389 patients, suggesting a significant reduction (p = 0.005), in particular when Ginkgo biloba was used (mean difference −0.43, 95%CI: −0.73; −0.13, p = 0.005) ( Figure S2). Block design field, on the contrary, was evaluated in four studies, for a total of 3389 patients, suggesting a significant reduction (p = 0.005), in particular when Ginkgo biloba was used (mean difference −0.43, 95%CI: −0.73; −0.13, p = 0.005) ( Figure S2).
WAIS Digit Span field was reported in five studies for a total of 308 patients and 309 controls. The overall effect was not significantly affected by polyphenol administration (p = 0.600) (Figure 11). WAIS Digit Span field was reported in five studies for a total of 308 patients and 309 controls. The overall effect was not significantly affected by polyphenol administration (p = 0.600) (Figure 11). RAVLT was reported in three studies, for a total of 301 subjects (150 treated and 151 controls). Polyphenol administration significantly increased RAVLT immediate score (mean difference 1.63, 95% CI: 1.25; 2.00, p < 0.001) ( Figure 12). WCST was reported in four studies for a total of 456 patients. The overall effect was not significantly influenced by polyphenol administration (p = 0.070) ( Figure S3). Trail making test A was used in seven studies, for a total of 3473 patients. The polyphenol administration did not significantly improve this score (p=0.620) ( Figure S4).
Similarly, trail making test B was used in eight studies, for a total number of 3503 subjects (1765 patients and 1738 controls). The overall polyphenol effect was statistically significant (mean difference −1.59, 95% CI: −2.75; −0.43, p = 0.007), probably for the contribution of resveratrol, evaluated in two single studies (mean difference −1.85, 95% CI: −3.14, −0.55, p = 0.005) ( Figure 13). RAVLT was reported in three studies, for a total of 301 subjects (150 treated and 151 controls). Polyphenol administration significantly increased RAVLT immediate score (mean difference 1.63, 95% CI: 1.25; 2.00, p < 0.001) (Figure 12). WAIS Digit Span field was reported in five studies for a total of 308 patients and 309 controls. The overall effect was not significantly affected by polyphenol administration (p = 0.600) (Figure 11). RAVLT was reported in three studies, for a total of 301 subjects (150 treated and 151 controls). Polyphenol administration significantly increased RAVLT immediate score (mean difference 1.63, 95% CI: 1.25; 2.00, p < 0.001) ( Figure 12). WCST was reported in four studies for a total of 456 patients. The overall effect was not significantly influenced by polyphenol administration (p = 0.070) ( Figure S3). Trail making test A was used in seven studies, for a total of 3473 patients. The polyphenol administration did not significantly improve this score (p=0.620) ( Figure S4).
Similarly, trail making test B was used in eight studies, for a total number of 3503 subjects (1765 patients and 1738 controls). The overall polyphenol effect was statistically significant (mean difference −1.59, 95% CI: −2.75; −0.43, p = 0.007), probably for the contribution of resveratrol, evaluated in two single studies (mean difference −1.85, 95% CI: −3.14, −0.55, p = 0.005) ( Figure 13). WCST was reported in four studies for a total of 456 patients. The overall effect was not significantly influenced by polyphenol administration (p = 0.070) ( Figure S3). Trail making test A was used in seven studies, for a total of 3473 patients. The polyphenol administration did not significantly improve this score (p = 0.620) ( Figure S4).

Discussion
In the current meta-analysis, we investigated polyphenol effects on different features of cognitive functions, in order to provide an overall picture of the reported activities. As already specified, our meta-analysis focused on intervention studies using polyphenol-containing food supplements. In particular, the selected studies considered amounts of polyphenols-rich extract up to 480 mg, 300 mg and 100 mg for Gingko biloba, resveratrol and soy isoflavones, respectively. These dosages provide an amount of polyphenols that is very likely to be higher, compared to that reachable by diet [49].
The global cognitive functions, evaluated through MMSE and ADAS-Cog scores, were not significantly improved by polyphenols. The only exception was Gingko biloba, whose intake translated in an improvement on the ADAs-Cog score, one of the most frequently used tests measuring the extent of cognitive decline through the evaluation of language and memory. Our results strengthen the observation of a previous meta-analysis of two studies in which a not statistically significant trend towards improvement in ADAS-Cog score was observed in the Ginkgo biloba group [97]. Similarly, among the different fields analyzed within the battery of the WAIS test, only the block design, often thought as a prototype test for visuospatial ability, was improved by treatment with polyphenols. As for ADAS-Cog, the most significant results were obtained with Gingko biloba extract. It has to be noted that such positive result is driven by the results of one over the three studies analyzed [98]. Thus, this observation requires confirmation with further studies. Nowadays, a possible explanation for the efficacy of Gingko may be attributable to the presence of several active compounds in the extract, such as diterpenes, ginkgolides and a number of flavonoids [99]. All these molecules may act on multiple sites in the central nervous system, all together contributing to attenuating cognitive decline.
By specifically analyzing the impact of polyphenols on distinct cognitive domains, we found a significant improvement in the RAVLT score, which evaluates verbal learning and memory, although the total number of subjects was limited. In particular, the three studies included in the meta-analysis focused on the effect of soy isoflavones and resveratrol and all led to positive results. The observation is consistent with two previous meta-analyses that were specifically aimed at investigating the effect of resveratrol and soy [100,101]. However, the latter study utilized a unique score that took into account the results of all neuropsychological tests together, while in our study we discriminated the effects on specific cognitive domains. On the other hand, our results on resveratrol are inconsistent with those of another recent meta-analysis, in which a beneficial impact has been demonstrated on the mood profile, rather than on memory and cognition performance [102]. A possible explanation

Discussion
In the current meta-analysis, we investigated polyphenol effects on different features of cognitive functions, in order to provide an overall picture of the reported activities. As already specified, our meta-analysis focused on intervention studies using polyphenol-containing food supplements. In particular, the selected studies considered amounts of polyphenols-rich extract up to 480 mg, 300 mg and 100 mg for Gingko biloba, resveratrol and soy isoflavones, respectively. These dosages provide an amount of polyphenols that is very likely to be higher, compared to that reachable by diet [49].
The global cognitive functions, evaluated through MMSE and ADAS-Cog scores, were not significantly improved by polyphenols. The only exception was Gingko biloba, whose intake translated in an improvement on the ADAs-Cog score, one of the most frequently used tests measuring the extent of cognitive decline through the evaluation of language and memory. Our results strengthen the observation of a previous meta-analysis of two studies in which a not statistically significant trend towards improvement in ADAS-Cog score was observed in the Ginkgo biloba group [97]. Similarly, among the different fields analyzed within the battery of the WAIS test, only the block design, often thought as a prototype test for visuospatial ability, was improved by treatment with polyphenols. As for ADAS-Cog, the most significant results were obtained with Gingko biloba extract. It has to be noted that such positive result is driven by the results of one over the three studies analyzed [98]. Thus, this observation requires confirmation with further studies. Nowadays, a possible explanation for the efficacy of Gingko may be attributable to the presence of several active compounds in the extract, such as diterpenes, ginkgolides and a number of flavonoids [99]. All these molecules may act on multiple sites in the central nervous system, all together contributing to attenuating cognitive decline.
By specifically analyzing the impact of polyphenols on distinct cognitive domains, we found a significant improvement in the RAVLT score, which evaluates verbal learning and memory, although the total number of subjects was limited. In particular, the three studies included in the meta-analysis focused on the effect of soy isoflavones and resveratrol and all led to positive results. The observation is consistent with two previous meta-analyses that were specifically aimed at investigating the effect of resveratrol and soy [100,101]. However, the latter study utilized a unique score that took into account the results of all neuropsychological tests together, while in our study we discriminated the effects on specific cognitive domains. On the other hand, our results on resveratrol are inconsistent with those of another recent meta-analysis, in which a beneficial impact has been demonstrated on the mood profile, rather than on memory and cognition performance [102]. A possible explanation for these discrepancies is the selection criteria of the studies (in our case only aged or MCI patients were included) and very high inter-individual differences on resveratrol bioavailability that may result from poor water solubility, chemical instability and difference in hepatic and intestinal metabolism [103].
Another score that was ameliorated by polyphenols was TMT-B, which evaluates executive functions, while no effect was observed on attention, measured by TMT-A. Notably, the analysis was conducted on more than 3000 subjects, making this result clinically relevant. In addition, in this case, the positive effect on the outcome is driven by resveratrol. The superior efficacy of resveratrol in our meta-analysis is in agreement with the beneficial effect on cerebrovascular functions and cognition, demonstrated in a recent trial on post-menopausal women [104].
The main limitation of this meta-analysis is related to the wide number of scores that have been used across the trials to evaluate the impact of polyphenols on cognitive performances. In the effort of analyzing the effect on individual scores, reflecting a particular feature of the cognitive domain, we restricted the analysis to a small number of studies compared to those available in literature. To overcome this criticism, as for cardiovascular markers, there is the necessity to identify the most consolidated markers of neurocognitive improvement, also in order to accomplish to the requests of the European Food Safety Authority (EFSA) for health claims on foods. In this regard, a precise indication of which psychological test is more appropriate for evaluation of central nervous system functions is lacking [105]. However, an ongoing project aims to answer this question in the neurological context [55].
On the other hand, the strength of our work is the strict selection of subjects, which excluded patients already affected by dementia, in order to specifically highlight some potential positive effects of polyphenols in the prevention of cognitive decline. In conclusion, the results of this meta-analysis suggest that polyphenol supplementation might improve specific measures of cognitive status. In addition, among all compounds, stilbene resveratrol appears to be the most active. In the case of Gingko biloba, the reported benefits could be attributed to the whole phytocomplex, where polyphenolic and non-polyphenolic compounds are likely to play an active role.
Given the differences among the studies in terms of patient features and the low number of analyzed studies, as well as the literature inconsistency across several trials, further research is needed to verify whether long-term intake of either single or multiple polyphenols may be a promising strategy for the prevention of cognitive decline in age-related diseases.

Discussion
The global, objective evaluation of polyphenol health effects presents several points of criticism. One is the existence of almost 10,000 polyphenols with peculiar structures and physiological roles, making the elucidation of their short-and long-term health effects challenging. Second, the recently well-described role of microbiota in the metabolism of polyphenols accounts for different metabotypes, which can differently affect physiological functions, producing high inter-individual differences in the biological response. Third, it is hard to compare data coming from observational/epidemiological studies with intervention trials. While in the former, the global effect is the result of the synergistic interaction of multiple components of the diet, the latter, although limiting the burden of potential confounding factors, may suffer from low adherence to real life conditions, where polyphenols are consumed in small amounts throughout life.
As suggested by Mena in his recent commentary, research on polyphenols requires an urgent, rigorous experimental approach in order to unravel their preventive potential on the development of diseases [10]. According to this, the evaluation of pattern of consumption, metabolites in circulation, inter-individual variations and new biological targets, needs to become pivotal aspects to be considered in every study. As soon as these guidelines are accepted by all reasearchers, data collected in the field of polyphenols and human health will be more realistic.
In the present review, we critically evaluated and meta-analyzed only intervention studies using polyphenol-containing food supplements, finding promising, but not definitive indications of the preventive effects of some polyphenols on cardiovascular disease and cognitive decline.
We are aware of the limitations of this study: the selection of trials with food supplements prevents considerations on the health benefits derived from a naturally polyphenol-enriched diet, where the synergistic effects of different components may appear. On the same trail, the consumption of individual polyphenolic compounds may not produce the same benefits observed in epidemiological studies. In addition, data coming from studies on food supplements, where polyphenol intake is higher than normal, could not be realistic in terms of interaction with drugs or other dietary components. When polyphenols are consumed in drug formulations, they should be considered as nutraceuticals, whose safety should also be addressed, especially if their bioavailability and biological activity are increased.

Conclusions
Although most polyphenol-based supplements in the market claim beneficial effects based only on in vitro data, the amount and quality of rigorous randomized clinical trials has been increasing for years. However, the long-term effects are yet to be elucidated and studies with hard clinical end points are still lacking. Studies aiming to fully characterize polyphenols pharmacokinetics and their safety are necessary to unravel their potential preventive role in real life.