The Effect of Polyphenol-Rich Interventions on Cardiovascular Risk Factors in Haemodialysis: A Systematic Review and Meta-Analysis

End-stage kidney disease is a strong risk factor for cardiovascular-specific mortality. Polyphenol-rich interventions may attenuate cardiovascular disease risk factors; however, this has not been systematically evaluated in the hemodialysis population. Using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, the following databases were searched: Cochrane Library (http://www.cochranelibrary.com/), MEDLINE (https://health.ebsco.com/products/medline-with-full-text), Embase (https://www.elsevier.com/solutions/embase-biomedical-research), and CINAHL (https://www.ebscohost.com/nursing/products/cinahl-databases/cinahl-complete). Meta-analyses were conducted for measures of lipid profile, inflammation, oxidative stress, and blood pressure. Risk of bias was assessed using the Cochrane Collaboration Risk of Bias tool and quality of the body of evidence was assessed by the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) methodology. Twelve studies were included for review. Polyphenol-rich interventions included soy, cocoa, pomegranate, grape, and turmeric. Polyphenol-rich interventions significantly improved diastolic blood pressure (Mean Difference (MD) −5.62 mmHg (95% Confidence Interval (CI) −8.47, −2.78); I2 = 2%; p = 0.0001), triglyceride levels (MD −26.52 mg/dL (95% CI −47.22, −5.83); I2 = 57%; p = 0.01), and myeloperoxidase (MD −90.10 (95% CI −135.84, −44.36); I2 = 0%; p = 0.0001). Included studies generally had low or unclear risks of bias. The results of this review provide preliminary support for the use of polyphenol-rich interventions for improving cardiovascular risk markers in haemodialysis patients. Due to the limited number of studies for individual polyphenol interventions, further studies are required to provide recommendations regarding individual polyphenol intervention and dose.

Studies were eligible for inclusion if they (1) used a double blind, randomized, placebo-controlled trial study design; (2) had no concurrent intervention; (3) examined the effect of a polyphenol-rich intervention on CVD outcomes (e.g., lipid profile, blood pressure, oxidative stress); and (4) recruited haemodialysis patients only. Other ESKD populations were excluded, in an attempt to keep the study population homogenous. We used the Phenol-Explorer 3.6 database to characterize and inform our decision on known polyphenol-rich interventions [30,31].

Data Extraction
The screening of articles was independently conducted by two review authors (J.K. and W.M.), with disagreements in judgement resolved by consensus or third reviewer (S.M.). Relevant articles titles and abstracts were initially screened. If deemed potentially eligible, studies were selected for full text review. Data was extracted from relevant studies using the following parameters: author/date, study design, sample size, total study period, population characteristics (including age, gender and co-morbidities), intervention characteristics (including type of polyphenol, dose and duration of exposure), length of follow up, and country of origin. For all included studies, mean, standard deviation (SD), standard error or 95% confident intervals (CI) for all pre-specified outcome data that were reported at baseline and follow-up were extracted for analysis if a significant difference was reported. Data was extracted by one reviewer (S.M.) and checked for accuracy by a second reviewer (S.N.).

Assessment of Study and Evidence Quality
Bias assessment was preformed based on the Cochrane Risk of Bias tool [32]. This tool provides criteria for assessing the quality of the included studies. All studies were included in the review regardless of bias rating. A score of 'high' indicated a high risk of all bias categories. A score of 'unclear' was given when information available was inadequate to correctly comment. A score of 'low' indicated low risk of all forms of bias and was the most desirable outcome.
The certainty in the body of evidence for each CVD outcome category was assessed using the Grading of Recommendations, Assessment, Development and Evaluation (GRADE) assessment tool [33] Certainty in the body of evidence was informed by considering risk of bias, inconsistency, indirectness, imprecision, publication bias, effect size, dose-response and plausible confounding. Based on the pooled or combined data across studies informing these considerations, the certainty in the body of evidence was conserved as very low, low, moderate or high [34]. Determination of the GRADE level of evidence was determined independently by two reviewers (S.M. and J.K.), with disagreements managed by consensus and discussion with a third reviewer (W.M.).

Data Analysis
The overall treatment effect on primary and secondary outcomes was calculated as the difference between the intervention and comparison groups' from change scores from baseline to the end of follow-up, or end of intervention values, permitting no significant differences observed at baseline between groups.
Quantitative analysis was conducted for sufficiently homogeneous and adequately reported outcome measures by pooling data into Review Manager (Version 5.3, The Cochrane Collaboration 2014) for meta-analysis using raw data. The appropriate variance from each individual study was used, either as the SD or calculated from the standard error of the mean (SEM) or 95% CI. Studies that reported on Median and Inter-Quartile-Ranges were assumed to not be normally distributed data and therefore not included in the meta-analysis. Meta-analysis was performed using the DerSimonian and Laird random-effects model [35] The I 2 statistic was used to assess the inconsistencies between studies and describe the percentage of variability in effect and data was checked using the fixed-effect model to ensure robustness and susceptibility to potential outliers. Heterogeneity was considered substantial if the I 2 statistic was ≥50%. A statistically significant (p < 0.05) result was considered evidence of an effect.

Study Selection
As shown in Figure 1, the literature search identified 3521 citations after the removal of duplicates. Initial screening identified 50 papers as potentially relevant for full text review. From this, 39 studies were excluded as they did not meet the inclusion criteria. Hand searching identified 1 additional study for inclusion, leaving 12 total studies included in the review. robustness and susceptibility to potential outliers. Heterogeneity was considered substantial if the I 2 statistic was ≥50%. A statistically significant (p < 0.05) result was considered evidence of an effect.

Study Selection
As shown in Figure 1, the literature search identified 3521 citations after the removal of duplicates. Initial screening identified 50 papers as potentially relevant for full text review. From this, 39 studies were excluded as they did not meet the inclusion criteria. Hand searching identified 1 additional study for inclusion, leaving 12 total studies included in the review.   Table 1 provides a summary of the study designs of included studies. The total sample size of the included studies was 520 participants, with individual study sample sizes ranging from 27 to 101 participants. All studies used a double-blind, randomized placebo-controlled parallel study design.
The samples appear to be the same; where Pakfetrat et al. [27] excluded some participants after recruitment, which could explain slight differences in sample characteristics. However, Pakfetrat et al. 2015 [27] does not refer to the earlier study at all, and therefore it is not clear.

between groups
Fasting serum total cholesterol:

between groups
Fasting serum HDL-C:

between groups
Fasting serum LDL-C:

between groups
Fasting serum total cholesterol:

between groups
Fasting serum total cholesterol:

between groups
Fasting serum HDL-C:

between groups
Fasting serum LDL-C:

between groups
Plasma CRP:

between groups
Pulse wave velocity:

between groups
Carotid intima-media thickness:

between groups
Blood pressure Systolic blood pressure: Acute

between groups
Acute on chronic Systolic blood pressure:

between groups
Diastolic blood pressure:
Pomegranate juice and extract improved markers of oxidative stress in three studies [26,37,44]. Two studies reported significant reductions in advanced oxidation protein products, polymorphonuclear leukocyte priming, myeloperoxidase [26,37]. One study reported significant reductions in oxidized fibrinogen (p = 0.03) and MDA (p < 0.001) [37], and another reported a significant interaction effect (group × time) for a measure of HDL-C (High Density Lipoprotein Cholesterol; p value not reported) associated paranoxose-1 activity [44].
Soy supplementation reduced oxidized LDL-C in one study (p < 0.05) [43], and one study reported turmeric supplementation to improve measures of catalase (p = 0.039) and MDA (p = 0.040) [27]. No other significant between-group differences in measures of oxidative stress were reported.
Meta-analyses reported significant improvements in myeloperoxidase (MD −90.10 (95% CI −135.84, −44.36); I 2 = 0%; p = 0.0001; n = 2 studies; 1 polyphenol-rich intervention; n = 126 participants; Figure 2). There was insufficient data to conduct a meta-analysis on any other measure of oxidative stress due to insufficient numbers of common outcomes reported in the included studies. Pomegranate juice and extract improved markers of oxidative stress in three studies [26,37,44]. Two studies reported significant reductions in advanced oxidation protein products, polymorphonuclear leukocyte priming, myeloperoxidase [26,37]. One study reported significant reductions in oxidized fibrinogen (p = 0.03) and MDA (p < 0.001) [37], and another reported a significant interaction effect (group × time) for a measure of HDL-C (High Density Lipoprotein Cholesterol; p value not reported) associated paranoxose-1 activity [44].
Soy supplementation reduced oxidized LDL-C in one study (p < 0.05) [43], and one study reported turmeric supplementation to improve measures of catalase (p = 0.039) and MDA (p = 0.040) [27]. No other significant between-group differences in measures of oxidative stress were reported.
Meta-analyses reported significant improvements in myeloperoxidase (MD −90.10 (95% CI −135.84, −44.36); I 2 = 0%; p = 0.0001; n = 2 studies; 1 polyphenol-rich intervention; n = 126 participants; Figure 2). There was insufficient data to conduct a meta-analysis on any other measure of oxidative stress due to insufficient numbers of common outcomes reported in the included studies.
One study reported turmeric to reduce CRP (p = 0.012) [38] and one study reported pomegranate juice to reduce IL-6 (p < 0.001) and TNF-α (p = 0.03) [37] No other significant between-group differences in inflammatory measures were reported.
Pomegranate extract significantly improved systolic and diastolic blood pressure, and mean arterial pressure (p < 0.05 reported for all measures) [44]. Cocoa flavanols significantly improved flowmediated dilatation (p < 0.001) [28]. No other significant between-group differences were reported.
One study reported turmeric to reduce CRP (p = 0.012) [38] and one study reported pomegranate juice to reduce IL-6 (p < 0.001) and TNF-α (p = 0.03) [37] No other significant between-group differences in inflammatory measures were reported.
Pomegranate extract significantly improved systolic and diastolic blood pressure, and mean arterial pressure (p < 0.05 reported for all measures) [44]. Cocoa flavanols significantly improved flow-mediated dilatation (p < 0.001) [28]. No other significant between-group differences were reported.

Lipid Profiles
Four studies reported on changes to cholesterol profile (i.e., total-C, HDL-C, LDL-C and triglycerides) following pomegranate [36,44], and soy supplementation [41,42]. Pomegranate had no significant between-group differences on participant lipid profiles except for one study that reported significant improvements in HDL-C (p = 0.03) and triglycerides (p = 0.008) for a subset of participants with low HDL-C or high triglycerides, respectively [36]. Soy supplementation improved fasting total cholesterol (p < 0.05) in one study and another study reported significant improvements on fasting triglycerides and total cholesterol in a subset of hyperlipidemic participants only (p < 0.05) [41,42].

Lipid Profiles
Four studies reported on changes to cholesterol profile (i.e., total-C, HDL-C, LDL-C and triglycerides) following pomegranate [36,44], and soy supplementation [41,42]. Pomegranate had no significant between-group differences on participant lipid profiles except for one study that reported significant improvements in HDL-C (p = 0.03) and triglycerides (p = 0.008) for a subset of participants with low HDL-C or high triglycerides, respectively [36]. Soy supplementation improved fasting total cholesterol (p < 0.05) in one study and another study reported significant improvements on fasting triglycerides and total cholesterol in a subset of hyperlipidemic participants only (p < 0.05) [41,42].

Lipid Profiles
Four studies reported on changes to cholesterol profile (i.e., total-C, HDL-C, LDL-C and triglycerides) following pomegranate [36,44], and soy supplementation [41,42]. Pomegranate had no significant between-group differences on participant lipid profiles except for one study that reported significant improvements in HDL-C (p = 0.03) and triglycerides (p = 0.008) for a subset of participants with low HDL-C or high triglycerides, respectively [36]. Soy supplementation improved fasting total cholesterol (p < 0.05) in one study and another study reported significant improvements on fasting triglycerides and total cholesterol in a subset of hyperlipidemic participants only (p < 0.05) [41,42].

Lipid Profiles
Four studies reported on changes to cholesterol profile (i.e., total-C, HDL-C, LDL-C and triglycerides) following pomegranate [36,44], and soy supplementation [41,42]. Pomegranate had no significant between-group differences on participant lipid profiles except for one study that reported significant improvements in HDL-C (p = 0.03) and triglycerides (p = 0.008) for a subset of participants with low HDL-C or high triglycerides, respectively [36]. Soy supplementation improved fasting total cholesterol (p < 0.05) in one study and another study reported significant improvements on fasting triglycerides and total cholesterol in a subset of hyperlipidemic participants only (p < 0.05) [41,42].

Adverse Events
Eight studies provided data on measured adverse events or irregular biochemistry [26,28,[37][38][39][40]45,46]. Most studies did not report any adverse events during the intervention period. Minor gastrointestinal symptoms (e.g., constipation or diarrhea) were reported in one study [25] Severity of adverse events was only reported in one study, which reported one serious adverse event (bleeding) within the intervention group [28]. No study reported a statistical analysis to determine significant differences in adverse event rates between control and interventions arms. Three studies

Adverse Events
Eight studies provided data on measured adverse events or irregular biochemistry [26,28,[37][38][39][40]45,46]. Most studies did not report any adverse events during the intervention period. Minor gastrointestinal symptoms (e.g., constipation or diarrhea) were reported in one study [25] Severity of adverse events was only reported in one study, which reported one serious adverse event (bleeding) within the intervention group [28]. No study reported a statistical analysis to determine

Adverse Events
Eight studies provided data on measured adverse events or irregular biochemistry [26,28,[37][38][39][40]45,46]. Most studies did not report any adverse events during the intervention period. Minor gastrointestinal symptoms (e.g., constipation or diarrhea) were reported in one study [25] Severity of adverse events was only reported in one study, which reported one serious adverse event (bleeding) within the intervention group [28]. No study reported a statistical analysis to determine

Adverse Events
Eight studies provided data on measured adverse events or irregular biochemistry [26,28,[37][38][39][40]45,46]. Most studies did not report any adverse events during the intervention period. Minor gastrointestinal symptoms (e.g., constipation or diarrhea) were reported in one study [25] Severity of adverse events was only reported in one study, which reported one serious adverse event (bleeding) within the intervention group [28]. No study reported a statistical analysis to determine significant differences in adverse event rates between control and interventions arms. Three studies reported changes in potassium levels with all three studies reporting no change after grape powder [25], soy [43], and cacao [28].

Risk of Bias
Risk of bias was low or unclear for most studies in the following domains: detection (11/12 rated as low) and reporting bias (12/12 rated as low), selection bias (10/13 and 11/12 rated as unclear). Three studies reported high risks of attrition bias and three studies were determined to have other risks of bias such as possible or undeclared conflicts of interest ( Figure 9). reported changes in potassium levels with all three studies reporting no change after grape powder [25], soy [43], and cacao [28].

Risk of Bias
Risk of bias was low or unclear for most studies in the following domains: detection (11/12 rated as low) and reporting bias (12/12 rated as low), selection bias (10/13 and 11/12 rated as unclear). Three studies reported high risks of attrition bias and three studies were determined to have other risks of bias such as possible or undeclared conflicts of interest ( Figure 9).

Quality of Evidence
Using the GRADE tool, most outcomes were rated at moderate quality (4/12) or very low (5/12) quality with inconsistency and imprecision being the most common reasons for downgrading (Table  3). Of the pooled data with significant findings, there was moderate quality of evidence for the effect on myeloperoxidase (oxidative stress marker); high quality for the effect on diastolic blood pressure, and very low quality for the effect on triglycerides.

Quality of Evidence
Using the GRADE tool, most outcomes were rated at moderate quality (4/12) or very low (5/12) quality with inconsistency and imprecision being the most common reasons for downgrading (Table 3). Of the pooled data with significant findings, there was moderate quality of evidence for the effect on myeloperoxidase (oxidative stress marker); high quality for the effect on diastolic blood pressure, and very low quality for the effect on triglycerides.

Discussion
The aim of this systematic literature review was to synthesise results from existing randomized controlled trials to evaluate the effect of polyphenol-rich interventions on cardiovascular markers in haemodialysis patients. The results of individual studies included in this review indicate that polyphenol-rich interventions may improve cardiovascular risk in patients on haemodialysis by improving various markers of inflammation (i.e., CRP, IL-6, TNF-α), lipid profile (i.e., HDL-C and triglycerides), blood pressure, and oxidative stress (i.e., advanced oxidation protein products, polymorphonuclear leukocyte priming, myeloperoxidase, oxidized fibrinogen, catalase, glutathione peroxidase, and MDA); with varying effect sizes and precision across studies.
Despite individual studies reporting significant improvements, pooled results report no effect for most outcomes excepting myeloperoxidase, diastolic blood pressure and triglycerides. Only myeloperoxidase, a measure of oxidative stress, had a large pooled effect size. In addition, using the GRADE assessment, most outcomes were rated as moderate or very low quality which provides limited confidence that the effect sizes reported in the existing evidence is representative of the true effect. The exception is for diastolic blood pressure, which was rated as high quality.
Individual studies that investigated cacao [28], pomegranate [26,36,37,44], turmeric [27,38], and soy [41,43], reported significant improvements in cardiovascular measures. Sensitivity analyses indicate that some polyphenol-rich interventions may provide greater improvements in cardiovascular markers. However, due to the small number of available studies investigating individual interventions in the haemodialysis population, it is premature to conclude superiority of one polyphenol-rich intervention over another at this time. In addition, while polyphenol-rich interventions reported significant improvements in numerous cardiovascular markers, there was little consistency in reported outcomes between studies that measured the same outcome and/or used the same intervention (e.g., blood pressure in [36,37]). Hence, future studies are required to expand the currently limited evidence base and to address such limitations.
The low baseline levels of some cardiovascular markers may be a possible explanation for the null findings and/or small effect sizes reported in some included studies and pooled data as it may be unlikely that further reductions are possible. For example, Janiques et al. [25] reported no significant difference in CRP; however, reported baseline levels (range: 2.6-2.6 mg/dL) were in the normal range (<3 mg/dL). In contrast, Paketrat et al. [38] reported significant reductions in CRP in participants that had CRP levels above the normal range (range: 7.0-10.8 mg/dL). This is also supported by the results of Wu et al. [44], Shema-did et al. [36], and Chen et al. [42] that reported greater decreases in blood pressure or cholesterol measures in hypertensive or hyperlipidemic participants, respectively.
Due to the large number of foods that contain appreciable levels of polyphenols [30], the habitual diet of participants may be a significant influence on study results, if not appropriately controlled for. Few studies included in this review implemented measures to control for this; however, future studies may benefit from implementing methods such as recording habitual diet throughout the study through the use of food diaries and research dietitians as well as educating participants on high polyphenol foods to avoid during the trial duration.
Few adverse events (predominantly gastrointestinal complains, one significant bleeding event reported [28]) were reported during the included trials which provide preliminary evidence for polyphenol-rich interventions being relatively safe within the haemodialysis population. However, due to the additional dietary restrictions present in this population, close monitoring for adverse events are required with clinical use and future trials are required to further evaluate their safety. In particular, although not reported to significantly affect patients in the included studies, consumption of certain polyphenol-rich food items, such as pomegranate juice, can significantly increase potassium intake beyond what would be typically advised for dialysis patients and therefore, care should be taken with people with history of or at higher risks of hyperkalaemia.
A further consideration for future research is to address the poor bioavailability of specific polyphenols. Resveratrol and curcumin (found in turmeric) [45,46], for example, have been demonstrated in pharmacokinetic studies to have poor bioavailability and a short half-life which has been addressed in several studies by using various methods such as nanoencapsulation, lipid emulsions, and co-administering active compounds that interact with liver enzymes involved in drug metabolism [46,47]. Addressing limitations with bioavailability may provide greater treatment efficacy.
A related research area is to elucidate potential inter-individual differences in polyphenol metabolism as this will inform which patients are likely to benefit from polyphenol-rich interventions. Individual differences in gastrointestinal microbiota appear to significantly influence the metabolism of certain polyphenols [48]. For example, the soy isoflavone, daidzein, is metabolised to (S)-equol in only 25-60% of the population [49]. Metabolism of ellagic acid, found in foods such as pomegranate and berries, can also be affected by microbiota composition, affecting timing, quantity, and types of metabolites excreted [50]. The role of microbiota on polyphenol metabolism in patients with kidney disease may be further complicated due to the possible influence of chronic kidney disease on intestinal microbiota [51,52].
This review includes studies that have used polyphenol-rich interventions. However, food interventions are comprised of several bioactive nutritive (e.g., vitamins and mineral) and non-nutritive compounds (e.g., polyphenols) and therefore, the results of the included studies may have been influenced by these additional compounds. Future trials that use standardized polyphenol extracts are recommended to control for the influence of non-polyphenol compounds.
The findings of this study provide preliminary evidence regarding polyphenol-rich interventions; however, results and conclusions are limited by the heterogeneity of interventions, dosages, and durations as well as variability in the cardiovascular risk of included participants. Although polyphenol-rich interventions have reported benefits in non-ESKD patients, considering the inclusion criteria of this review, generalising results to patients with ESKD or chronic kidney disease who are not receiving dialysis should be avoided until further studies are conducted. Furthermore, studies with large sample sizes are required to sufficiently evaluate the adverse events of polyphenol-rich interventions in this population group.

Conclusions
This review evaluated the clinical evidence of various polyphenol-rich interventions for patients with ESKD receiving haemodialysis from double-blind placebo-controlled randomized trials. The results of this review provide preliminary support for the use of polyphenol-rich interventions as part of cardiovascular disease prevention and/or management in haemodialysis patients. At this stage, no specific polyphenol-rich intervention appears superior, which is likely due to the small number of available studies, small sample sizes, and lack of control of habitual diet. With this in mind, clinical recommendations are premature until further evidence addresses these limitations.
Author Contributions: W.M. was involved in all processes of the review, J.K. was provided the meta-analyses and GRADE assessment, S.M. contributed to the GRADE assessment and data extraction, S.N. contributed to the initial literature search, K.C. provided content support into the manuscript, C.I. was involved in the design of the study. All authors contributed to the development of the study manuscript.

Conflicts of Interest:
The authors have no conflicts of interest to declare.